keentelengineering

Test Energy Before COD in WECC: CAISO BESS Guide

Fri, 28 Aug 2026 21:46:44 GMT

Aug 28, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Challenge : Frequent false tripping using conventional electromechanical relays
Solution: SEL-487E integration with multi-terminal differential protection and dynamic inrush restraint
Result: 90% reduction in false trips, saving over $250,000 in downtime

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

1. Executive Summary

An electric grid must remain in continuous balance — generation onto the grid must equal consumption from it at every instant. PJM achieves this balance, and prices it, through a layered market architecture. Each layer operates on a different time horizon, and each one touches project economics differently.

Every generating or storage project in the Western Interconnection delivers energy before it reaches its Commercial Operation Date. That is not an exception or a workaround; it is how a facility is commissioned. Inverters are energised, the plant controller is tuned, ride-through settings are loaded, reactive capability is proven at the point of interconnection, and metering and telemetry are validated against live production. All of it happens in the window between first synchronisation and COD, and all of it produces energy that has to be scheduled, settled, curtailed, and accounted for.

That window is governed by two frameworks that run on different clocks and answer to different parties. The commercial framework lives in the power purchase agreement and the interconnection agreement: what test energy is worth, who can curtail it and on what notice, which milestones must be declared, and what happens to the schedule if a milestone slips. The reliability framework lives in the NERC Reliability Standards and in the regional modeling and testing requirements administered through the Transmission Planner, the Area Coordinator, and WECC. It does not care what the offtake agreement says.

The most consequential misconception in this space is that reliability compliance begins at COD. It does not. Obligations attach on registration, and a facility becomes an operating element of the system the moment it first synchronises. A project that plans its compliance program around the commercial milestone spends its entire commissioning window operating with obligations already live and no evidence being generated to demonstrate them.

The second most consequential misconception is that the model submitted during the interconnection study is the model that governs. It is not. What governs is a verified as-built representation of the equipment actually installed, with the control parameters actually loaded — and those parameters are adjusted by the vendor during commissioning, in the same weeks the developer is certifying that the as-built matches the study. That drift, between the studied baseline and the tuned plant, is where most late-stage interconnection problems originate.

This paper works through the pre-COD window for a battery storage project interconnecting to CAISO within the WECC footprint: the implementation path and its gating milestones, the metering and telemetry requirements peculiar to storage, the modeling chain and the model stack, why electromagnetic transient models are now a separate mandatory obligation rather than a study refinement, how the registration and ride-through clocks interact with a COD date, and what a defensible pre-COD test program contains. It closes with a backward timeline from COD, a failure taxonomy, and a twenty-question FAQ.

Two sentences worth internalising

Test energy is delivered by a facility that is already a registered entity with live obligations, into a market that will curtail it without compensation, under a model that has to match what was actually built.

The commercial milestone and the reliability milestone are not the same date, and neither one waits for the other.

What a CAISO-Interconnected Battery Storage Project Has to Model, Test, and Prove Between First Synchronization and Commercial Operation — and Why the Reliability Clock Starts Before the Commercial One

Storage metering is where the pre-COD window most often breaks, because the requirements are structurally different from those for a generator and are frequently designed as if they were not.

5.1 Separating Wholesale Charging From Retail Station Load

A battery facility consumes energy for two entirely different reasons. It charges the battery, which is a wholesale market transaction. It also runs auxiliary load — thermal management, controls, lighting, fire systems — which is retail station service. Those two consumptions must not be commingled in the revenue metering, and the metering configuration has to be able to distinguish them under all operating states, including the state where the battery is neither charging nor discharging and the auxiliary load is still running.

This is a design problem, not a metering technician problem. It has to be settled in the single line and the metering design during engineering. Discovering during meter validation that the current transformer placement cannot separate the two flows is a rework item on energised equipment inside the commissioning window.

5.2 Telemetry Content and Cadence

Real-time telemetry to the balancing authority for a storage resource carries more than the real and reactive power points a generator provides. State of charge and the instantaneous charge and discharge limits are operationally necessary — without them the operator cannot know what the resource can actually do in the next interval. Point lists, scan rates, and the gateway configuration must be established, mapped, and point-to-point tested before the synchronisation gate.

The telemetry design also has to survive the commissioning period itself, during which the plant controller is being tuned and the values being reported may be transiently wrong. Telemetry that reports a nonsensical state of charge or an out-of-range limit during tuning generates operator queries and, in the worst case, a hold

Test Energy Before COD in the WECC Footprint

3. What Pre-COD Test Energy Actually Is

It helps to separate the obligations by who imposes them, because the consequences of missing them are entirely different in kind.

2.1 The Commercial Clock

Power purchase agreement terms governing test energy: whether it is compensated at all, at what price, and under what conditions. In the Western markets it is common for pre-COD deliveries to be curtailable by the buyer or the balancing authority without compensation, and for the developer to bear the delivery risk entirely. This is a negotiated term, not a regional rule — read the agreement rather than assuming the market norm.
Interconnection agreement milestones: the right to test and to deliver test energy, notice obligations to the transmission provider, and the declaration mechanics for commercial operation.
Guaranteed commercial operation dates and the liquidated damages, security draws, or termination rights that attach to slipping them.
Market and registry mechanics: scheduling coordinator arrangements, resource identification, and registration in generation attribute tracking systems where environmental attributes from the test period are to be claimed. Attribute registration generally has to be in place before the energy is produced, not reconciled afterwards.

2.2 The Reliability Clock

Registration as a Generator Owner and Generator Operator, which attaches by criteria rather than by contract, and which now reaches well below the traditional Bulk Electric System thresholds for inverter-based resources.
Modeling data submittal to the Transmission Planner and Planning Coordinator, on their published schedules and in their specified formats.
Dynamic model verification and validation, now including electromagnetic transient models for inverter-based resources.
Ride-through performance requirements and the disturbance monitoring that demonstrates them.
Facility ratings, protection system maintenance, relay loadability, and voltage control coordination obligations that begin when the facility begins operating, not when it begins selling.

Why the distinction is not academic

A commercial milestone that slips costs money under a contract that both parties negotiated and both parties understand.

A reliability obligation that was live during commissioning and generated no evidence is a compliance finding, and the exposure is assessed per day per violation. The two risks are not comparable and should not sit in the same column of a project risk register.

Test energy is the electrical output of a facility that is being commissioned rather than operated. Its purpose is to exercise equipment and prove performance, and every megawatt-hour of it is a by-product of a test rather than a product being sold. Three characteristics follow from that and shape everything downstream.

First, it is uncontrolled in the market sense. The plant is being started, stopped, ramped, stepped, and deliberately disturbed on the test engineer’s schedule, not the dispatcher’s. That is why it is treated as curtailable and why the balancing authority needs visibility into it before the first synchronisation rather than after.

Second, it is the only source of the field data that the compliance program ultimately rests on. Reactive capability, voltage control loop response, ride-through behaviour, and model validation all require the plant to be operating. There is no way to generate the evidence without generating the energy, which is precisely why the reliability obligations are live during the window that developers think of as pre-operational.

Third, for a storage resource it is bidirectional. Charging energy is load. It is metered, settled, and accounted for differently from discharge, and the metering configuration that makes that distinction correctly has to be in place and validated before the resource can be certified. A storage project that treats test energy as a generation-only concept discovers the problem at the metering validation step.

For a project interconnecting to CAISO, the pre-COD window is administered through the New Resource Implementation process, which is organised into a sequence of grouped deliverables commonly referred to as buckets and administered through the ISO’s resource interconnection management system. The process is tied to the network model build cycle, which means the milestone dates are not purely a function of construction progress — they are a function of which network model build the resource lands in.

4.1 The Gating Structure

Two gates matter more than the rest, and both are hard.

The first gate is the request to synchronise. The earlier deliverable groups must be complete and accepted before the interconnection customer may request permission to synchronise to the grid, and the request itself carries a processing period on the order of two working weeks. Metering and telemetry point-to-point testing must be successfully completed before that gate opens. This is the point at which many projects discover that a metering or telemetry deficiency they had classified as a punch item is in fact a schedule-critical blocker.

The second gate is the request for commercial operation. All deliverable groups plus the metering site verification documentation must be received and accepted before the interconnection customer may request the commercial operation date, and that request carries its own processing period. The customer then files the commercial operation declaration with an effective date. Nothing about this sequence is automatic and none of it can be compressed by declaring readiness.

4.2 Trial Operations and Live Data Validation

Between the two gates sits the trial operations period, which is the real subject of this paper. Preliminary meter data validation happens here. Final validation of metering and telemetry is performed after synchronisation using actual generation data — which is to say, using test energy. The certificate confirming compliance is issued following that validation against live production.

The practical consequence is that the pre-COD window is not slack in the schedule. It is a required data-generation period with a minimum duration set by how long it takes to produce clean, validated operating data across the required conditions. Projects that plan a two-week commissioning window and a same-month COD are planning a schedule that the process cannot support.

The scheduling coordinator point

The scheduling coordinator arrangement must be in place and stable through the window. Changing scheduling coordinators mid-process, particularly before the compliance certificate is issued, creates a settlement gap that nobody recovers.

The legal entity named in the interconnection agreement must match the entity requesting market participation exactly. Entity mismatches from mid-development transfers and financing structures are a routine cause of avoidable delay.

7. The Model Stack for a BESS

Notice and Disclaimer

The single largest change in this area, and the one most likely to be missing from a project’s compliance plan, is that electromagnetic transient models are no longer an optional refinement requested by some Transmission Planners. The consolidated dynamic model verification and validation standard became effective in 2026, superseding and retiring the two predecessor standards that previously governed voltage regulator and turbine-governor model verification separately.

The obligation now has three parts that did not previously exist together.

A verified electromagnetic transient model of the facility must be provided to the Transmission Planner for inverter-based resources, covering the inverters, the collector system, auxiliary control devices, the plant controller, the step-up and main transformers, and the protective functions and limiters on both the AC and DC sides.
The electromagnetic transient model and the positive-sequence model must be cross-validated against each other under large-signal disturbances, with discrepancies resolved and documented. Submitting two models that were each developed independently and never compared does not satisfy this.
Models must be updated and resubmitted within a defined period after any change at the facility that alters dynamic response — hardware, software, firmware, control mode, or settings.

Why this lands squarely in the pre-COD window

The change triggers include settings and control-mode changes. Commissioning is a settings-change exercise by definition: the plant controller is tuned, ride-through parameters are loaded, reactive priority is configured, and ramp limits are set.

A project that freezes its model before commissioning and does not re-verify after tuning has, by the plain terms of the obligation, a model that no longer represents the in-service equipment — and it acquired that condition during the exact period when it was certifying that the as-built matched the study.

There is a narrow exception for legacy facilities where the original equipment manufacturer no longer supports an electromagnetic transient model, and for legacy facilities the Transmission Planner has not identified as applicable. Neither exception helps a new project. For anything being commissioned now, the model deliverable and the right to use it should be a procurement requirement written into the inverter and plant controller purchase, because obtaining a validated model from a vendor after the order is placed is a negotiation rather than an entitlement.

2. Two Clocks, One Project

5. Metering and Telemetry for a Storage Resource

6. The WECC Modeling Chain

The path a model takes from the field to the interconnection-wide base cases is a chain of custody with several parties in it, and understanding the chain explains why late as-built changes are so expensive.

The developer submits the steady-state and dynamic data package to the Transmission Planner, which validates it against regional criteria. The Transmission Planner passes it to the Area Coordinator, which integrates the data into the regional power flow base cases. The compiled cases go to WECC staff, which verifies that the models initialise cleanly in the interconnection-wide dynamics environment without producing numerical errors or artificial oscillations. Only after that does the representation of the facility become part of the case that everyone else studies against.

Data is prepared for the platform WECC uses to assemble the base cases and is converted for the other major planning platform, which is why model submittals must be expressible in the accepted library rather than as a proprietary artefact. A model that cannot be represented in the accepted library cannot enter the chain at all.

6.1 The Modeling Rules That Trip Projects

Generator step-up and main station transformers must be modelled explicitly. Impedance folded into the generator representation is a common submittal rejection and produces incorrect voltage and reactive results at the point of interconnection.
The full reactive footprint must initialise cleanly — capacitor banks, static compensators, reactors, and any dynamic reactive equipment. A case that will not solve with the plant’s reactive equipment in service is a case that gets returned.
The collector system must be represented as an equivalent that reproduces the real losses and impedance seen at the point of interconnection, not as a nominal placeholder.
Any equipment deviation from the configuration studied in the cluster process must be documented and carried forward, because the base case has to describe the facility that exists.

For positive-sequence work, inverter-based resources are represented by an approved library of generic models rather than by vendor-proprietary code. The stack for a battery facility is three layers, and each layer has current and superseded variants that matter.

Two further points are worth stating because they are routinely missed.

First, the first-generation wind models were retired years ago and the current representation of a Type 3 or Type 4 wind facility is a composite of the same converter and electrical control layers plus separate drivetrain, aerodynamic, pitch, and torque control modules. Any submittal still carrying first-generation wind model names is carrying a rejection.

Second, grid-forming inverter models now exist in the approved library and have been approved for interconnection-wide studies. If a project is procuring grid-forming capability — which is increasingly relevant at weak points of interconnection and for facilities with islanding requirements — the modeling approach should be settled at procurement, not discovered during the as-built submittal.

10. Ride-Through and Disturbance Monitoring

9. The Registration Clock

Registration is the step that converts a project into a compliance entity, and its threshold has moved. Inverter-based resources that individually or collectively reach an aggregate nameplate capacity of twenty megavolt-amperes or more, connected through a system delivering that capacity to a common point of connection at sixty kilovolts or above, now fall into a registration category that did not exist a few years ago and are required to register as Generator Owner and Generator Operator.

Two features of that criterion catch projects. It is an aggregate test, so several individually sub-threshold facilities sharing a common point of connection are assessed together. And it reaches well below the traditional Bulk Electric System threshold, which means a substantial population of projects that were designed, financed, and staffed on the assumption that NERC compliance did not apply to them are now inside the framework.

For a project in commissioning, the practical question is sequencing. Registration is not triggered by commercial operation. A facility that will meet the criterion should have its registration status resolved, its compliance program stood up, and its evidence-retention practice running before test energy begins — because the standards that apply on registration include obligations that can only be demonstrated with records made during the commissioning window.

Three related standards now govern inverter-based resource performance during system disturbances, and their implementation dates interact with a COD date in a way that materially affects projects reaching commercial operation now.

10.1 The Three-Part Structure

Disturbance monitoring and reporting for inverter-based resources, which took effect in 2025, requires the recording capability — sequence of events, fault recording, and dynamic disturbance recording — that everything else depends on.
Ride-through performance, effective in the fourth quarter of 2026, requires the resource to remain connected and continue delivering current through defined voltage and frequency disturbance envelopes, with momentary cessation prohibited within the no-trip region.
Post-disturbance analysis and mitigation, effective on the same date, requires unexpected changes in real power output following disturbances to be identified, analysed, and corrected.

The dependency runs one way. Ride-through performance cannot be demonstrated, and post-disturbance analysis cannot be performed credibly, without the records the monitoring standard mandates. The three should be designed as a single data architecture rather than as three separate compliance projects.

10.2 The COD Interaction That Matters Right Now

The implementation plan for the ride-through standard separates capability-based design obligations from performance-based operational obligations. For facilities on the Bulk Electric System, existing resources were given a phase-in to the effective date. Resources reaching commercial operation after the effective date are not: they must comply with the design requirements at commercial operation, with no transition period. Facilities below the Bulk Electric System threshold have a later date, and a limited exemption pathway exists for existing resources with verified hardware limitations that cannot be resolved through software or settings — an exemption that does not assist a project being built now.

For a project whose COD falls after the effective date, this means the ride-through settings loaded during commissioning are not a best-effort configuration to be optimised later. They are a compliance deliverable that has to be right at COD, verified, and evidenced. The operational demonstration follows once the monitoring capability is established, but the design obligation is immediate.

Practical sequencing consequence

Ride-through parameter verification cannot be the last item in the commissioning punch list. If it fails, the remedy may be a firmware change, a settings change requiring vendor involvement, or in the worst case a hardware limitation — and each of those triggers a model re-verification obligation on top of the schedule impact.

It should be tested early in the pre-COD window, when there is still time to do something about the answer.

The failure mode is assuming that clearing one threshold determines the others. A facility can be below the traditional Bulk Electric System definition, above the inverter-based resource registration criterion, and subject to individual modeling submittal, all at once. Screening against all three at the start of the project, and re-screening whenever the facility rating or configuration changes, is a half-day exercise that prevents a category of very expensive surprises.

4. The CAISO New Resource Implementation Path

11. Three Thresholds That Are Not the Same

Three numeric thresholds circulate in these conversations and are regularly conflated. They govern different obligations and a project can fall on different sides of each.

This document is original technical content prepared by Keentel Engineering LLC for general professional information. It is not legal advice, regulatory advice, or project-specific engineering advice, and it does not constitute a compliance determination for any facility. Registration obligations, applicable Reliability Standards, implementation dates, regional modeling requirements, market processes, and interconnection procedures vary by facility, jurisdiction, registration status, and market region, and change over time.

Commercial terms concerning test energy — including compensation, curtailment rights, milestone declarations, and attribute tracking — are contractual and vary between counterparties. Descriptions of common practice in this document are general observations and are not a substitute for reading the applicable power purchase agreement, interconnection agreement, and tariff provisions.

Standards, implementation plans, model libraries, and market processes referenced by subject in this document are current to the date of publication to the best of our knowledge and are subject to revision. The governing versions are those published by the applicable standards body, regional entity, system operator, or authority, and should be verified directly for any project decision.

Keentel Engineering LLC is an independent engineering consultancy. Reference to any standard, market process, regional entity, system operator, industry organisation, or equipment category in this document does not imply affiliation with, endorsement by, or sponsorship from any such organisation or manufacturer.

8. EMT Is a Separate and Now Mandatory Obligation

12. As-Built Verification and Material Modification

The as-built obligation exists because the base cases and the studies that depend on them must describe the facility that was actually built. Two distinct requirements are frequently conflated and should not be.

12.1 Two Different Sixty-Day Windows

One requirement runs before commercial operation: as-built documentation demonstrating that the parameterised equipment on site matches the baseline used in the study process, submitted in advance of the commercial operation milestone. The other runs after: a period of continuous, high-quality operating and telemetry data used to complete certification and to support forecasting and resource classification. They have different owners, different content, and different consequences, and a compliance plan that treats them as one item will miss one of them.

12.2 The Material Modification Question

If the as-built configuration differs from what was studied — a different inverter manufacturer, a different unit count, a materially different reactive capability, a changed collector arrangement — the transmission provider has to determine whether the change constitutes a material modification. A determination that it does can return the project to the study process, with the queue and schedule consequences that implies. A determination that it does not still requires as-built validation data demonstrating that the delivered configuration behaves acceptably.

The engineering discipline that avoids this is unglamorous: maintain a controlled record of the studied baseline, log every deviation as it occurs with its electrical significance assessed at the time, and raise material questions with the transmission provider early rather than presenting an accumulated set of changes at the as-built submittal. A single deviation raised in month four is a conversation. Eleven deviations presented in month sixteen is a restudy.

12.3 Commissioning Drift

The deviations that matter most are not hardware. Hardware changes are visible, procured, and documented. Control parameters are not: the plant controller gains, the reactive priority setting, the ramp limits, the ride-through thresholds, and the protection setpoints are adjusted by commissioning engineers and vendor technicians, often iteratively, often without a change record that anyone outside the commissioning team sees. Those parameters are the model. A plant whose hardware exactly matches the study and whose controller has been retuned three times has an as-built model that does not match the study.

13. The Pre-COD Test Program

14. Counting Backward From COD

A defensible pre-COD test program produces three things simultaneously: permission to proceed, evidence for the compliance file, and validated model parameters. Designing it to produce only the first is the common error.

Pre-synchronisation verification: protection settings against the current coordination study revision, metering and telemetry point-to-point testing, grounding and interconnection facility verification, and confirmation that the settings loaded in the inverters and the plant controller match the parameters in the submitted model.
First synchronisation and initial energisation, with the balancing authority notified and the testing plan agreed. Record the plant’s behaviour from the first energisation — the disturbance recording capability should be commissioned before, not after, the first switching operations.
Reactive capability verification at the point of interconnection across the full range, lagging and leading, at multiple real power levels, and for a storage resource in both charging and discharging states. Reactive capability is not symmetric across operating states and testing only one is testing half the plant.
Voltage control loop verification: step-response testing of the plant controller and unit-level controls, with oscillography captured at sufficient resolution to support model calibration. This is the test whose data validates the model, so it must be planned as a data-collection exercise with predefined step magnitudes and recording configuration, not performed as a tuning session that happens to be recorded.
Active power control and ramp behaviour verification, including the response to dispatch instructions and the ramp rate limits actually configured.
Ride-through parameter verification: confirmation that the voltage and frequency ride-through settings physically loaded in the equipment implement the required envelopes, that momentary cessation is not configured within the no-trip region, and that protection settings do not defeat the ride-through requirement. Perform this early.
Storage-specific tests: state-of-charge reporting accuracy, charge and discharge limit reporting, transition behaviour between charging and discharging, and auxiliary load separation verified through the metering under all operating states including idle.
Model calibration and re-verification: compare recorded field response against the simulated response, adjust model parameters to match measured behaviour, and re-verify the positive-sequence and electromagnetic transient models against each other. Then submit the corrected models.
Compile and retain the evidence package: test procedures, as-run records, recorded data, settings files as loaded, model files as submitted, and the correspondence documenting acceptance. This package is the compliance file and it is also the only defence if a disturbance later raises questions about the facility’s performance.

The following is an indicative planning sequence. Actual durations depend on the network model build cycle, the Transmission Planner’s data schedule, equipment lead times, and the project’s registration status — but the ordering is stable and the dependencies are real.ti

15. What Actually Goes Wrong

16. Keentel Interconnection and Compliance Services

Keentel Engineering supports developers, owners, and independent engineers through exactly this window — from the study-phase model that establishes the baseline to the as-built package and the evidence file that closes it out.

16.1 Interconnection Engineering and Studies

Point-of-interconnection engineering substation and collector system design, and interconnection facility design for generation and storage projects.
Load flow, short-circuit, protective coordination and selectivity, and arc-flash studies, with inverter-based resources represented as current-limited sources rather than as equivalent machines.
Harmonic, flicker, grid strength, and transient stability analysis, and electromagnetic transient modelling for control interaction, weak-grid stability, ride-through verification, and switching and overvoltage studies.
Effective grounding and ground fault overvoltage assessment, insulation coordination, and ground grid analysis.

16.2 Modeling and Model Validation

Positive-sequence and electromagnetic transient model development, cross-validation, and submittal packages prepared to the Transmission Planner’s specified format and level of detail.
Model verification and validation against staged test data and disturbance records, including step-response test design, data capture specification, and parameter calibration.
Baseline control and model management: establishing the studied configuration of record, tracking commissioning-phase deviations, and re-verifying after settings, firmware, or control-mode changes.
As-built data package preparation and support through the Transmission Planner, Area Coordinator, and interconnection-wide review chain.

16.3 NERC Compliance Support

Registration threshold screening and applicability assessment across modeling, registration, and market criteria.
Ride-through design evaluation, disturbance monitoring architecture, and post-disturbance analysis programs designed as a single data lifecycle.
Facility ratings, protection system maintenance programs, relay loadability, and voltage control coordination.
Evidence package structure and retention practice built during commissioning rather than reconstructed for an audit.

16.4 Owner’s Engineer and Commissioning Support

Pre-COD test program development, test procedure authorship traceable to the design basis and the compliance obligations, and witness support during execution.
Design review of EPC and vendor submittals, QA/QC of third-party study and model packages, and equipment specification including model and test deliverables.

Keentel Engineering holds a Florida Certificate of Authorization and maintains offices in Tampa, Austin, Sacramento, and Baltimore, supporting projects across the interconnections.

17. Frequently Asked Questions

L1-L5 Data Center Commissioning | Testing Guide

Fri, 28 Aug 2026 14:21:50 GMT

Aug 28, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

L1 Through L5 Commissioning in a Data Center

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

A data center is the only building type routinely required to prove, before a single production workload is placed on it, that it will survive the loss of its utility feed, the failure of a chiller, the failure of a UPS module, and a control system alarm storm — simultaneously, at full design load, in front of the owner. That proof is the commissioning programme, and in most projects it is organised into levels commonly labelled L1 through L5.

The levels are widely used and almost never defined the same way twice. They are not a code, not a standard, and not a certification. They are a project convention that has to be written into the commissioning plan and the contract documents, because the difference between one party's L4 and another's L5 is measured in weeks of schedule and hundreds of thousands of dollars of load bank rental.

The more consequential point is this: integrated systems testing does not fail because of how it is executed. It fails because of decisions made eighteen months earlier in design. A generator sized on steady-state kW without regard to step-load acceptance or the harmonic interaction with the UPS input will not carry the block load at L5. A single-line diagram without maintenance bypass around every critical device cannot demonstrate concurrent maintainability regardless of how the test is written. A facility with no permanent load bank connection will be tested once, expensively, and never again.

This paper works through each level in engineering detail — the specific tests, the standards behind them, and the acceptance criteria — and then works backwards to the design decisions that determine whether those tests can be passed. It closes with a defect taxonomy drawn from what commissioning programmes actually find, a section on designing for testability, and a twenty-question FAQ.

The framing that matters

Commissioning is not a construction-phase activity that verifies a finished design. It is a design-phase discipline that produces a facility capable of being verified.

Every test script at L5 should trace back to a line in the Owner's Project Requirements and a corresponding decision in the Basis of Design. If a requirement is not in the OPR, no one will test it. If a design decision is not in the BOD, no one will know what "pass" means.

Three separate numbering conventions circulate in data center work, and conflating them is the most common source of confusion in owner conversations. They measure different things and are produced by different bodies.

1. Executive Summary

Level 3 answers whether each piece of equipment can operate on its own. It is the first level at which equipment is energised, and it is the level with the highest safety exposure in the entire programme.

6.1 Protective Device Testing and Settings

This is where the design engineer's short-circuit and coordination study stops being a document and becomes a set of physical settings. Protective relays are tested by secondary injection and, where warranted, primary injection; low-voltage power circuit breaker trip units are tested across their long-time, short-time, instantaneous, and ground-fault functions; settings are entered and independently verified against the current revision of the coordination study.

The verification against the current revision is not a formality. Equipment substitutions, transformer impedance differences between the specified and the delivered unit, and cable length changes during construction all invalidate a study, and a facility running on settings from a superseded revision is a facility with an unverified selectivity claim. Ground-fault protection systems require a performance test at installation, and the test must exercise the system as installed, including the neutral and grounding arrangements that the code requires.

6.2 Equipment Start-Up

Manufacturer-authorised start-up for equipment whose warranty is conditioned on it — which is most major equipment. The start-up report is a warranty document, not a commissioning document, and both are required.
Motor rotation verification before coupling, insulation resistance, and no-load run.
UPS module start-up, DC bus verification, and battery system commissioning including an acceptance capacity test performed to the method appropriate to the battery chemistry. A battery system accepted on float voltage readings alone has not been tested.
Generator start-up, cranking and shutdown sequences, cooling and exhaust system verification, fuel system commissioning including day tank controls and transfer logic, and fuel quality verification.
Chiller, pump, and air-handling start-up with rotation, flow, and control device verification.
Point-to-point verification of every building management and electrical power monitoring point: correct tag, correct scaling, correct engineering units, correct alarm limits. A metering system whose accuracy class was specified for revenue-grade performance should be verified as installed, not accepted on datasheet.

What Each Level Actually Tests, Why the Facility Either Passes or Fails Integrated Systems Testing on Decisions Made During Design, and How to Engineer a Facility That Can Be Proven

4. L1 — Factory and Component Verification

2. Three Numbering Systems That Are Not the Same Thing

Most L1-to-L5 charts begin at the factory. The highest-leverage level begins earlier, during design, and is variously called Level 0, design-phase commissioning, or commissioning design review. Its omission is the single most common structural weakness in a commissioning programme.

Design-phase commissioning has a narrow and specific scope. It is not a peer review of the design and it is not value engineering. It asks one question repeatedly: can what is drawn be proven to satisfy what the owner asked for?

3.1 What Level 0 Actually Reviews

Owner's Project Requirements — is every availability, capacity, density, and operational requirement written in terms that can be measured? "Highly available" is not testable. "No single component failure results in loss of IT load, demonstrated at 100 percent of design load" is.
Basis of Design — does every OPR requirement have a corresponding engineering decision, with the assumptions stated? Generator step-load capability, UPS transfer criteria, thermal ride-through duration, allowable rack inlet temperature excursion.
Single-line diagram against the maintainability requirement — is there a maintenance bypass around every device that must be serviced without dropping load? Can each distribution path be de-energised in turn with the load carried on the other?
Sequences of operation — do they exist, in writing, in enough detail that a test script can be written from them? A sequence that says "generators start on loss of utility" is not a sequence. Start signal source, time delay, load block order, and failure-to-start logic are the sequence.
Testability provisions — permanent load bank connections and their capacity, test-position breakers, metering and trending points, temporary power provisions for the test programme itself.
Study executability — will the protective coordination study produce settings that can actually be entered into the specified devices, and are the devices field-adjustable to the settings the study requires?

A Level 0 finding is worth roughly a hundred times the same finding at L5

Changing a single-line diagram during design costs drafting time. Adding a maintenance bypass after the switchgear is manufactured costs a change order, a factory slot, and schedule. Discovering at integrated systems testing that concurrent maintainability cannot be demonstrated costs the commissioning date.

Level 1 answers whether the equipment was manufactured and supplied correctly. Its centre of gravity is factory acceptance testing, but the valuable work happens before the plane ticket is booked: reviewing the manufacturer's test procedure against the specification and the design intent, and deciding what will be witnessed versus what will be accepted on certified report.

4.1 Electrical Equipment

Medium-voltage switchgear — design and production test evidence per the applicable switchgear standards, including power-frequency withstand, mechanical operation, control wiring verification, and, where specified, arc-resistant construction type test evidence for the accessibility class being purchased.
Power and distribution transformers — routine tests per the applicable transformer test code: turns ratio, polarity and phase relation, winding resistance, no-load and load losses, impedance, insulation power factor, applied and induced potential. Impulse and temperature-rise tests where specified. Verify the tested unit is the unit being shipped by serial number.
Uninterruptible power supplies — efficiency at multiple load points, input current distortion, transfer performance to and from bypass, overload capability, battery runtime at design load, and the behaviour of any high-efficiency or eco-mode operating state, which is where transfer performance most often disappoints.
Generator sets — governor and voltage regulator response, block load acceptance in the steps the application requires, transient voltage and frequency dip and recovery against the declared performance class, and alternator temperature rise at rated power factor. Testing at unity power factor when the machine will run at 0.8 lagging is a common and consequential shortcut.
Switchboards, PDUs, RPPs, and busway — production testing, dielectric verification, and control and metering wiring point-to-point checks against the approved shop drawings.
Automatic transfer switches and static transfer switches — withstand and closing ratings coordinated with the upstream protective device, transfer and re-transfer timing, and the neutral switching arrangement, which must match the grounding scheme in the design.

4.2 Mechanical Equipment

Chillers, computer room air handlers, pumps, and cooling distribution units are verified against certified performance ratings at the design points and, critically, at the part-load points where the facility will actually spend its life. A chiller certified at full load and untested at 30 percent is a chiller whose behaviour at the load the facility runs on Day One is unknown.

4.3 What L1 Produces

Certified test reports, nameplate data verified against the approved submittal, serial numbers captured into the asset register, and a documented punch list of factory findings with resolution dates. The asset register created at L1 is the spine of the entire programme; if serial numbers are captured for the first time at L3, the commissioning agent is reconstructing history rather than verifying it.

Level 2 answers whether the equipment was installed correctly. It is checklist work, it is unglamorous, and it catches a disproportionate share of the defects that would otherwise appear as unexplained failures during integrated testing.

5.1 Electrical

Visual and mechanical inspection per the applicable acceptance testing specification: anchorage, alignment, clearances, arc-flash and identification labelling, and the presence of the correct devices in the correct positions.
Bolted connection torque verification to the manufacturer's values, recorded per connection, with calibrated tooling. Torque records are the most commonly falsified and most rarely audited commissioning document in the industry.
Cable insulation resistance testing, and for medium-voltage cable, field withstand or diagnostic testing appropriate to the insulation system and the cable's age, performed per the applicable field-test guide.
Phase rotation and phasing verification across every source and every transfer point. Phase relationships that are correct within each source but wrong between sources are only discovered at transfer, which is the worst possible time.
Grounding and bonding verification: grounding electrode system resistance measurement by an appropriate method, bonding continuity across the equipment ground path, and verification of the separately derived system bonding arrangement against the design.
Instrument transformer ratio and polarity verification. CT polarity errors are invisible until a protection scheme misoperates or a metering system reports impossible values.
Working space and dedicated equipment space verification against the electrical code, before the trade above installs its piping in the dedicated space.

5.2 Mechanical and Structural

Piping pressure and leak testing, flushing and cleanliness verification, and chemical treatment before equipment is put into service.
Duct leakage testing, airflow accessory installation, and containment integrity.
Vibration isolation, seismic anchorage and restraint verification where required by the building code or the owner's standard.
Refrigerant charge verification and evacuation records.

A rule worth writing into the specification

No system proceeds to Level 3 until its Level 2 checklists are complete and signed. The pressure to energise equipment ahead of installation verification is enormous and it is where most commissioning programmes lose their integrity. The cost of the rule is schedule friction; the cost of waiving it is discovering an untorqued bus joint by thermal signature at L4, after it has been energised for six weeks.

8. L5 — Integrated Systems Testing

Disputes about whether a given test belongs to Level 4 or Level 5 are common, and they are schedule disputes in disguise, because L5 requires full load bank deployment and the whole operations team while L4 does not. A workable boundary rule:

A test belongs to Level 4 if it exercises a single system and its own internal redundancy, with adjacent systems held in a normal state. A test belongs to Level 5 if its pass criterion depends on the response of a system other than the one being failed.

Under that rule, taking a UPS module offline and proving the remaining modules carry the load is Level 4. Failing the UPS and proving that the static switch transfers, the alternate path picks up, the EPMS raises the correct alarm, and the mechanical plant is unaffected is Level 5. Starting a generator on a manual signal is Level 3. Starting it on a real loss of utility with the transfer scheme live and the load real is Level 5.

Whatever boundary is chosen, it must be written into the commissioning plan before the schedule is baselined, and the load bank quantity, duration, and rental period must be derived from it. Load bank logistics are the long pole in most integrated testing programmes.

The practical consequence for an owner is that all three may appear in the same project. A facility can be designed to Uptime Tier III concurrent maintainability, carry a TIA-942 rating claim, and be commissioned through a five-level programme. None of the three substitutes for the others, and the commissioning programme is the only one of the three that produces evidence the installed equipment actually works.

6. L3 — Pre-Functional and Start-Up Testing

7. L4 — Functional Performance Testing

Level 4 answers whether each system performs as designed, exercised through its full sequence of operations and its failure modes, one system at a time.

7.1 Electrical Systems

UPS operation across every operating state: normal, on battery, on internal bypass, on maintenance bypass, and any high-efficiency mode, with transfer performance verified against the sensitivity envelope the design assumed for the IT load.
Static transfer switch operation, including transfer time and the behaviour of downstream loads during transfer.
Automatic transfer switch operation: transfer on source failure, re-transfer, time delays, engine start signal, and in-phase or closed-transition behaviour where specified. Timing verified against the standard governing the emergency or standby classification the system carries.
Generator load acceptance under load bank, at the design power factor, through the load block sequence the sequence of operations defines, with voltage and frequency transients recorded.
Load bank testing at full rated capacity for the duration the specification requires, with temperature stabilisation, so that the alternator, cabling, and cooling systems are proven at steady state and not merely at start-up.
Harmonic measurement at the distribution and service levels, verified against the limits assumed in the design. This is where the design's assumptions about non-linear load behaviour are either confirmed or shown to have been optimistic.
Thermographic survey under load, once systems are loaded and stabilised.

7.2 Mechanical Systems

Chiller performance at design and part-load conditions, with staging, sequencing, and reset strategies exercised.
Pump and valve operation, including valve authority and control stability. Hunting control loops discovered at L4 become oscillating plants at L5.
Air-side performance: CRAH and CRAC capacity, containment differential pressure, rack inlet temperature distribution, and bypass and recirculation assessment.
Economiser changeover where free cooling is provided, at the changeover conditions rather than only at design conditions.

7.3 Redundancy Demonstration

The redundancy claim is proven at L4 for each system in isolation: remove one component of an N+1 set with the system at design load and demonstrate that capacity and setpoints are maintained. For a concurrently maintainable facility, this extends to demonstrating that each distribution path can be isolated in turn, which requires that the maintenance bypasses were in the design — the point made in Section 3.

Level 5 answers whether the complete facility survives realistic failures. Systems are tested together, at full load, with the control systems live and the operations team present. It is the only level that tests the interactions, and interactions are where data centers fail.

8.1 The Load Question

Integrated testing is performed on load banks representing the design IT load, positioned to represent the design power and thermal distribution. Two design decisions determine how well this works: whether load bank connections were engineered in at the PDU or RPP level rather than improvised at the switchgear, and whether the load bank distribution can represent the density profile rather than a single aggregated block. A facility tested with all load at one end of the white space has not tested its air distribution.

8.2 The Scenario Matrix

The scenario set is project-specific but the following are close to universal for a facility with generation and mechanical redundancy:

Loss of utility, with generator start, transfer, load acceptance, and stable parallel operation; then utility restoration and re-transfer, including the load transfer back and the cooldown sequence.
Black building test — total loss of normal power with restart from a de-energised state, which exercises control power, sequencing, and every assumption about what is available during a restart.
Generator failure to start, proving the redundant generator or the load shed scheme.
UPS module and UPS system failure, proving the redundant path and the static switch behaviour under real load.
Chiller failure with standby chiller start, verifying that the mechanical system recovers before rack inlet temperatures leave the acceptable envelope.
Loss of cooling — the thermal ride-through test, measuring the actual time from cooling loss to the temperature limit at the rack inlet. This is the test most often shortened and the one whose result most directly governs operational procedure.
Loss of controls: BMS or EPMS failure, loss of control power, and network failure, verifying that systems fail to a safe and defined state rather than to an undefined one.
Cascading and multiple-failure scenarios appropriate to the facility's fault tolerance claim, and operator response drills conducted against the emergency operating procedures.

8.3 Instrumentation and Acceptance

An integrated test without instrumentation is a demonstration, not a test. Power quality recorders capable of capturing sub-cycle events at the generator terminals, UPS input and output, and critical distribution; temperature sensors at representative rack inlets; and trend logging at a resolution fine enough to see the event rather than its aftermath. Acceptance criteria must be numeric and pre-agreed: transfer times, voltage and frequency excursion limits, temperature limits and durations, and recovery times.

The interaction that catches the most projects

A UPS with a high-distortion input current characteristic, fed from a generator whose subtransient reactance was selected on cost, produces voltage distortion at the generator terminals that the UPS interprets as an unacceptable source. The UPS goes to battery on a healthy generator. Battery runs down. Load drops.

The failure occurs at L5 and its cause is a generator alternator selection and a UPS input filter decision made during design. It is not a commissioning defect; it is a design defect that commissioning found.

9. Where the L4 / L5 Boundary Actually Falls

11. What Commissioning Actually Finds

10. Designing for Testability

The following provisions cost very little at design and are difficult or impossible to add later. They are the practical output of taking commissioning seriously as an engineering discipline rather than a construction milestone.

Permanent load bank connections with adequate capacity at the level of the distribution where load must be represented, not solely at the service entrance. Include the connection in the coordination study and the arc-flash analysis.
Maintenance bypass around every device that must be serviced with the facility live, and a single-line diagram that makes each isolation sequence obvious.
Test-position or drawout construction where breakers must be exercised without de-energising the bus.
Metering and trending points sufficient to instrument the acceptance criteria: source-side and load-side of every transfer point, generator terminals, UPS input and output, and mechanical plant capacity points.
Independent, redundant control power for protection and transfer schemes, and a defined behaviour on loss of control power. Control power is the single most common undocumented dependency in a facility that otherwise has no single points of failure.
Written, complete sequences of operation issued as a design deliverable, not assembled by the controls contractor from equipment manuals.
Generator sizing based on the step-load profile and the non-linear load characteristic, not steady-state kW alone, with the alternator and the UPS input characteristic evaluated together.
Thermal ride-through capacity engineered explicitly — chilled water volume, thermal storage, or continuous cooling — with the target duration stated in the Basis of Design so that the L5 test has a pass criterion.
Selective coordination demonstrated by study where required by code for emergency, legally required standby, or critical operations power systems , using devices whose settings are field-adjustable to the values the study requires.
Grounding and bonding designed as a system across separately derived sources, transfer switches, and the white space, so that neutral switching arrangements and ground-fault sensing coordinate rather than conflict.
Space, access, and temporary power provisions for the commissioning programme itself, including load bank staging, cable routing, and heat rejection during testing.

The following defect classes recur across facilities, vendors, and geographies. They are listed by the level at which they are typically caught, which is usually later than the level at which they were introduced.

13. Reading the L1–L5 Chart Correctly

12. Turnover, Training, and Ongoing Commissioning

The commissioning programme produces a facility and a body of evidence. The evidence has to survive the handover or it will be reconstructed, expensively, the first time the facility is modified.

The systems manual: as-built single lines, sequences of operation as tested, setpoint schedules, the coordination study and the settings actually installed, and the arc-flash study aligned to the final configuration.
The complete test record: L1 factory reports, L2 checklists including torque records, L3 start-up and settings verification, L4 functional test results, and L5 scenario results with the recorded data, not merely the pass/fail sheet.
The issues log, closed out, with each finding traced to its resolution and re-test.
Operator training conducted against the emergency operating procedures and the scenarios actually tested at L5, with the operations team present during the tests rather than briefed afterwards.
Deferred and seasonal testing where design conditions could not be represented during the commissioning window — free cooling changeover being the usual example.
A re-commissioning cadence, and a rule that any modification to the electrical or mechanical topology triggers re-analysis of the coordination and arc-flash studies and re-test of the affected scenarios. IT load growth changes the load profile the facility was proven at.

Notice and Disclaimer

This document is original technical content prepared by Keentel Engineering LLC for general professional information. It is not project-specific engineering advice and does not constitute a design, a design review, a commissioning plan, or a certification for any particular facility. Commissioning level definitions described here are project conventions and vary between owners, commissioning authorities, and jurisdictions; the definitions written into a specific project's commissioning plan govern that project.

Code and standards references are provided by subject for orientation. The edition adopted by the authority having jurisdiction governs code requirements, and the current published edition of each standard governs its own requirements. Test methods, acceptance criteria, and equipment ratings must be taken from the applicable standard and the specific manufacturer's data for the equipment selected.

Keentel Engineering LLC is an independent engineering consultancy. Reference to any standard, code, certification scheme, industry organisation, or equipment category in this document does not imply affiliation with, endorsement by, or sponsorship from any such organisation, certification body, or manufacturer.

5. L2 — Installation Verification

The widely circulated L1-to-L5 summaries communicate the sequence well. Five clarifications are worth stating because each has a design or contractual consequence.

Clarification 1 — The levels are a convention, not a standard

There is no published standard that defines L1 through L5 with fixed content. Process guidance exists in the commissioning standards and in data-center-specific best practice documents, but the level numbering itself is a project convention. It must be defined in the commissioning plan and referenced from the specifications, or the parties will discover their definitions differ at the point where it costs the most.

Clarification 2 — Level 0 belongs on the chart

Design-phase commissioning is where the cheapest findings are made and where the testability of the facility is determined. A chart that starts at the factory implies that commissioning begins when procurement ends, which is the framing that produces facilities that cannot pass their own integrated test.

Clarification 3 — The levels are per system, not per facility

Progression is not a single facility-wide march from L1 to L5. Each system advances at its own pace, and integrated testing gates on the slowest one. Schedule logic that treats the levels as facility-wide phases hides the actual critical path, which is usually one late-delivered switchboard or one incomplete controls integration.

Clarification 4 — The redundancy tests at L4 and L5 are different tests

Removing a redundant component and showing the system holds is a Level 4 activity. Failing a system and showing the rest of the facility responds correctly is Level 5. Charts that list "redundancy testing" at L4 and "UPS failure" at L5 without stating the distinction leave the boundary to be negotiated during execution.

Clarification 5 — There is a level after L5

Post-occupancy activity — seasonal testing, IT load migration verification, re-commissioning after modification, and ongoing commissioning against trended performance — is where the value of the programme is either preserved or lost. A facility proven once at handover and never re-proven is a facility whose resilience claim expires quietly.

14. Keentel Data Center Engineering Services

Keentel Engineering provides the electrical engineering that determines whether a data center can be commissioned successfully — the design work upstream of the test, and the technical support during it. Our data center practice covers the following.

14.1 Power Delivery and Interconnection

Utility service planning and large-load interconnection: load characterisation, service capacity studies, coordination with the serving utility and, where applicable, the transmission provider and ISO or RTO large-load interconnection process.
Point-of-interconnection engineering, substation design, and medium-voltage collection and distribution design for campus and multi-hall developments.
On-site generation interconnection, including behind-the-meter generation, parallel operation, and export arrangements where the facility participates in grid programmes.

14.2 Facility Electrical Design

Distribution topology development: N+1, 2N, 2(N+1), and block-redundant architectures evaluated against the availability and concurrent maintainability requirements in the OPR, with the maintenance and isolation sequences designed in rather than discovered later.
Medium- and low-voltage distribution design, UPS and battery system design, generator plant design including paralleling schemes and fuel systems, and white-space power distribution.
Grounding, bonding, and lightning protection design across separately derived sources, transfer equipment, and the white space.

14.3 Power System Studies

Short-circuit, protective device coordination, and selective coordination studies where required for emergency, legally required standby, and critical operations power systems .
Arc-flash incident energy analysis and labelling, with maintainability and equipment selection evaluated together rather than as a post-design compliance exercise.
Load flow, motor starting, generator step-load and stability analysis, and harmonic and power quality studies, including the generator-to-UPS interaction that governs integrated test outcomes.
Electromagnetic transient modelling where the interconnecting utility or the equipment configuration requires it.

14.4 Commissioning and Owner's Engineer Support

Design-phase commissioning review: OPR and BOD verification, single-line review against the maintainability requirement, sequence-of-operation completeness, and testability provisions.
Commissioning specification and plan development, level definitions, and acceptance criteria written to be measurable.
Test script development traceable to the design basis, factory acceptance test procedure review and witnessing, and technical support during functional and integrated testing.
Design review of EPC and contractor submittals, QA/QC of third-party design packages, and turnover documentation review.

Keentel Engineering holds a Florida Certificate of Authorization and maintains offices in Tampa, Austin, Sacramento, and Baltimore, supporting projects across the interconnections.

15. Frequently Asked Questions

3. Level 0 — The Level Most Programs Omit

BESS for Data Centers | Power Quality Engineering Guide

Fri, 28 Aug 2026 08:18:02 GMT

Aug 28, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Adding a BESS to a Data Center Campus

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

Battery energy storage has moved from the periphery of data center design to its centre. Storage is now being deployed on data center campuses to bridge outages, shave peaks, defer utility infrastructure, provide the load flexibility that increasingly determines how quickly a large load can be interconnected at all, and in some markets to participate in wholesale services. The commercial case is well understood. The electrical consequences are frequently not.

The framing that causes the most trouble is treating storage as a container of energy. A BESS is a power-electronic converter with a battery attached. Everything the grid experiences — harmonic current injection, reactive support, ramp behaviour, fault current contribution, ride-through, stability under weak-grid conditions — is determined by the power conversion system and its control tuning, not by the cells. The energy capacity sets how long it can do something. The inverter sets what it does.

On a data center campus the analysis is harder than on a standalone storage site, for one reason: the site is already an unusually distorted electrical environment before the BESS arrives. A large data center is a concentration of rectifier front ends — UPS systems, IT power supplies, variable-speed drives on the mechanical plant — sitting behind a long medium-voltage cable system with substantial distributed capacitance. Adding a converter of comparable rating to that environment is not a superposition problem that can be waved through; harmonic currents from different sources can cancel or reinforce depending on phase, and the network the BESS is injecting into may already have a parallel resonance sitting near a harmonic order the converter produces.

This paper works through what actually changes when a BESS is added at a data center campus: the correct harmonic compliance framework and why the metric is total demand distortion at the point of common coupling rather than voltage total harmonic distortion at a panel, how network resonance is created and excited, what governs voltage fluctuation and flicker, why grid strength and control interaction dominate the analysis on weak interconnections, what protection and overvoltage behaviour changes, and the complete study set an owner should expect. It closes with a mitigation toolkit, a compliance measurement protocol, and a twenty-question FAQ.

The sentence the source graphic gets right

The battery may be the asset, but the question that determines whether the project is buildable is how the grid — and the campus’s own power quality — respond once it is connected.

That question is answered by studies performed during design, using validated manufacturer models of the specific converter being purchased. It is not answered by a datasheet line stating that total harmonic distortion is below five percent.

The design implications differ by use case, and the use case must be settled before the electrical design begins, because it determines the converter rating, the duty cycle, the point of connection, and the interconnection pathway.

Bridging and ride-through. Lithium storage replacing or augmenting conventional UPS energy storage, extending autonomy and reducing the frequency of generator starts. Connection is typically at low voltage within the critical power path.
Peak shaving and demand management. Reducing billing demand and deferring utility service upgrades. Connection is typically at the campus medium-voltage bus.
Load flexibility as interconnection currency. This is the fastest-growing driver. Where utilities and system operators are constrained on large-load interconnection, the ability to curtail or shift campus demand on request — without dropping IT load — is increasingly what determines the energisation date. Storage is the mechanism that makes a firm IT load look flexible from the utility’s side of the meter.
Generator reduction. Displacing some engine-generator capacity or runtime, with consequences for air permitting, fuel storage, and the resilience case that has to be demonstrated during commissioning.
Wholesale market participation. Frequency response, regulation, or capacity, which brings the facility into a different regulatory and performance regime and imposes telemetry, model, and ride-through obligations.

Each of these places the BESS at a different point in the single line and subjects it to a different interconnection framework. Storage inside the critical power path at low voltage is a facility design problem. Storage on the medium-voltage campus bus operating in parallel with the utility is an interconnection problem, and the study obligations, model requirements, and protection requirements change accordingly.

1. Executive Summary

Resonance is the most misunderstood item on every BESS power quality summary. A converter does not create resonance. Resonance is a property of the network — the interaction between system and transformer inductance and the capacitance of cables, filters, and any power factor correction equipment. What the converter does is inject current at frequencies that may coincide with a resonant condition that already exists, at which point a modest injected current produces a large voltage.

6.1 Where the Capacitance Comes From on a Data Center Campus

Medium-voltage cable. Campus distribution is cable, not overhead line, and often a great deal of it. Distributed cable capacitance is frequently the dominant capacitive element on a large campus and it grows with every phase of build-out.
Converter output filter capacitance, from both the storage converter and any other inverter-based resource on site.
Power factor correction and harmonic filter banks, where fitted.
Surge capacitors and instrument transformer burden, minor individually and occasionally relevant in aggregate.

6.2 The Consequence of Getting It Wrong

A parallel resonance near a harmonic order the converter produces amplifies that order, producing voltage distortion far above what the injected current would suggest, overheating capacitors and transformers, and in the worst case causing protective device operation on a system with no fault. Because campus cable capacitance grows as phases are built, the resonant frequency migrates across the build-out. A study performed on Phase 1 does not describe the campus at Phase 3.

The study that answers this is a frequency scan

A harmonic impedance scan across the frequency range of interest, performed at the point of connection for each credible network configuration — each phase of build-out, each transformer and feeder switching state, minimum and maximum load damping — identifies where the resonant peaks sit relative to the orders the converter produces.

It is inexpensive during design and it is the only way to know whether a filter is needed, what it should be tuned to, and whether a proposed filter will move a resonance onto a worse order rather than away from one.

What Battery Storage Actually Does to Power Quality at the Point of Interconnection, Which Studies Have to Prove It, and Why the Inverter — Not the Battery — Decides the Outcome

4. Harmonics: The Right Metric at the Right Point

2. Why Storage Is Appearing on Data Center Campuses

The power conversion system determines grid behaviour through a small number of design and tuning choices, and the electrical engineer needs the manufacturer’s specifics on each of them before any study can be run.

Switching frequency and modulation strategy, which set where converter-produced current distortion appears in the frequency spectrum. Modern converters push the dominant switching content into the kilohertz range, which is a different problem from the low-order harmonics produced by the six-pulse rectifiers that shaped the classic distortion literature.
Output filter topology — typically an inductor-capacitor-inductor arrangement — whose capacitance is itself a network element that participates in resonance, and whose damping determines behaviour at the filter’s own resonant frequency.
Current control bandwidth and the phase-locked loop tuning, which govern stability margin against network impedance and are the mechanism behind converter-network control interactions on weak systems.
Reactive capability and the reactive priority setting, which determine what voltage support is available and whether the converter gives up real power to provide it.
Ramp rate limits and their configurability, which govern the voltage fluctuation the site imposes on the network.
Fault current contribution and duration, typically limited to a small multiple of rated current, which is the single most consequential difference between an inverter-based resource and a synchronous machine for protection design.
Ride-through configuration and the priority between ride-through obligations and equipment self-protection, which is where a converter that meets its datasheet can still trip in a way the interconnecting entity considers non-compliant.

The model is the deliverable

Studies require validated converter models — a positive-sequence model for load flow and stability work, and an electromagnetic transient model for control interaction, weak-grid, and harmonic behaviour.

Generic library models produce answers that are wrong in the specific ways that matter, because the behaviour of interest is a property of the manufacturer’s control code. Model availability, quality, and the right to use the model should be procurement requirements written into the converter specification, not requests made after the order is placed.

Harmonic compliance conversations go wrong in three predictable ways: the wrong metric, the wrong measurement point, and the wrong assumption about who owns the limit.

4.1 Current Distortion Is Measured as Total Demand Distortion

The governing North American framework expresses current distortion limits as total demand distortion, referenced to the maximum demand load current at the point of common coupling, rather than as total harmonic distortion referenced to the instantaneous fundamental. The distinction is not academic. A converter operating at ten percent of rating can show an alarming current total harmonic distortion figure while contributing a trivial amount of actual distorting current. Judging a project by that number produces both false alarms and false comfort.

The allowable total demand distortion is not a single value. It scales with the ratio of available short-circuit current at the point of common coupling to the maximum demand load current — a stiffer connection relative to the load is permitted more distortion, because the resulting voltage distortion is smaller. Individual harmonic orders carry their own limits, tightening as order increases, with even-order harmonics held to a fraction of the odd-order limits.

4.2 The Point of Common Coupling Is a Specific Place

Limits apply at the point of common coupling — the point where other customers are or could be served — not at a panel inside the facility, not at the converter terminals, and not at the low-voltage side of the campus transformer. Internal distortion is a facility engineering matter governed by equipment withstand and transformer heating, not by the interconnection limits. Conflating the two produces expensive filter specifications aimed at a problem the utility never had.

4.3 Voltage Distortion Limits Belong to the Utility

The framework assigns current limits to the customer and voltage limits to the system operator, with the voltage limits tightening as system voltage rises. In practice this becomes a shared responsibility: the customer’s current injection acting on the utility’s system impedance produces the voltage distortion the utility must hold. This is why background distortion measurement before energisation matters. A site that arrives at a point of common coupling already near the voltage distortion limit has far less headroom than the current-limit table alone suggests, and discovering that after commissioning is a mitigation project rather than a design decision.

4.4 Above the Classic Spectrum

Converter switching content in the range above the traditional harmonic measurement band — the supraharmonic region — is not addressed by the classic limit tables and is increasingly the source of real problems: interference with power line communication and metering, audible noise in magnetics, and interaction between converters from different manufacturers sharing a bus. A data center campus with a large storage converter, dozens of UPS modules, and hundreds of drives is exactly the environment where this appears. It should be measured, not assumed away.

Standalone storage projects are analysed against a network. A data center campus adds a large, non-linear, dynamically varying load between the converter and the point of common coupling, and that changes the analysis in four specific ways.

5.1 Superposition Does Not Hold Naively

Harmonic currents from the storage converter and from the facility’s rectifier loads combine as phasors, not as magnitudes. Depending on the phase relationship at each order, they can partially cancel or reinforce. A study that adds distortion budgets arithmetically will be wrong in both directions, and the direction it is wrong in is not predictable without modelling the actual sources with their phase information.

5.2 The Facility Load Varies and So Does the Answer

IT load, mechanical load, and storage dispatch all vary independently. Compliance is not a single operating point. The harmonic study must evaluate the credible combinations — minimum facility load with the converter at full charge, full facility load with the converter idle, converter at full discharge during a demand response event — because the worst case for total demand distortion is often minimum demand rather than maximum, and the worst case for resonance is a specific capacitance and load damping combination that occurs at neither extreme.

5.3 Transformer Heating Is a Separate Calculation

Harmonic current increases transformer losses disproportionately, because eddy current loss rises with the square of harmonic order. The relevant assessment is a harmonic loss factor calculation against the transformer’s capability, or specification of a transformer rated for the harmonic duty. Data center campus transformers see this loading continuously, not intermittently, and a transformer selected on kVA alone in a high-distortion environment is a transformer with an unstated derate.

5.4 The Critical Load Is Sensitive to the Same Things

Voltage distortion and fluctuation at the campus bus propagate to the UPS input. UPS systems evaluate their source and will transfer to battery if input quality falls outside their acceptance window. A storage converter whose behaviour degrades campus voltage quality can therefore cause the facility’s own critical power systems to abandon the utility source — the same class of failure discussed in the commissioning literature as a generator-to-UPS interaction, arriving through a different door.

8. Grid Strength and Control Interaction

Most storage converters in service today are grid-following: they synchronise to a measured voltage and inject a controlled current. They require a stable external voltage reference and they contribute nothing to system inertia or to establishing voltage on a de-energised network.

Grid-forming control instead regulates voltage magnitude and angle behind an impedance, presenting a voltage source to the network. The practical consequences relevant to a data center campus are substantial: grid-forming converters can operate at very low grid strength where grid-following control becomes unstable, can contribute to frequency stability, and can establish voltage on an islanded network, which makes campus islanding and black start technically feasible in a way grid-following equipment does not.

The trade-offs are real. Grid-forming operation imposes different current headroom requirements during transients, the control mode interacts with the protection design, and interconnection requirements and model validation practice for grid-forming resources are still maturing. The decision should be driven by the grid strength at the interconnection and by whether campus islanding is an operational requirement, and it should be made early, because it affects converter procurement, protection design, and the study scope.

6. Resonance: The Network Creates It, the BESS Excites It

7. Voltage Fluctuation, Flicker, and Ramp Rate

Voltage fluctuation from a storage system is a power-change problem, not a harmonic problem, and it is straightforward to estimate. A change in the real and reactive power exchanged at a bus produces a voltage change governed by the change in apparent power relative to the short-circuit strength at that bus and the resistance-to-reactance ratio of the source impedance. Two consequences follow directly.

First, the magnitude of the fluctuation is set as much by the strength of the interconnection as by the size of the converter. The same converter produces a small voltage step on a strong system and a visible one on a weak system. Second, because the converter’s reactive capability is fast and controllable, much of the fluctuation is correctable by the converter itself if the control mode and the reactive priority are configured to do so. A storage system is more often the remedy for voltage fluctuation on a campus than the cause of it.

Flicker — the perceptibility of rapid voltage variation — is assessed by a standardised measurement method producing short-term and long-term severity indices, evaluated against planning levels the system operator sets. The relevant question for a data center campus is rarely the storage converter operating on a smooth dispatch signal; it is fast events. Step changes during demand response dispatch, transitions between charge and discharge, and coincidence with mechanical plant starting are the cases that need evaluation.

The design controls are ramp rate limiting configured in the converter, reactive support enabled with an appropriate priority, and, where the interconnection is genuinely weak, dynamic reactive equipment sized from the study rather than from a rule of thumb.

The ratio of available short-circuit strength at the point of interconnection to the rating of the inverter-based resource is the single most predictive number for whether a storage project will behave. A high ratio means the network holds voltage and the converter’s controls have ample stability margin. As the ratio falls, the converter’s own current injection begins to move the voltage it is measuring, the phase-locked loop and current controllers are operating against a network impedance comparable to their own control impedance, and stability margin erodes.

For a campus with a large storage converter, other on-site inverter-based generation, and a long radial supply, the effective strength seen by the converter can be considerably lower than a headline short-circuit figure at the utility substation suggests. Where multiple inverter-based resources share an interconnection, the aggregate matters, and system operators in the most affected regions have developed weighted and site-specific formulations precisely because the simple ratio overstates strength when converters interact.

The failure modes at low grid strength are not gradual degradation. They are oscillatory instability at frequencies that positive-sequence simulation cannot represent, unexpected tripping during and after nearby faults, and sustained interactions between converters from different manufacturers that neither exhibits alone. Sub-synchronous control interaction with series-compensated transmission is the best documented example and is not the only one.

This is why electromagnetic transient modelling is a requirement rather than a refinement at weak interconnections. Positive-sequence stability tools cannot represent converter control dynamics at the timescales where these phenomena live, and a study set that stops at load flow and short circuit has not addressed the risk that actually threatens the project.

9. Grid-Following and Grid-Forming Behaviour

11. The Study Set

10. Protection, Fault Behaviour, and Overvoltage

10.1 Fault Current Contribution

An inverter-based resource contributes fault current limited by its control to a small multiple of rated current, for a controlled duration, with a waveform that is not the decaying symmetrical contribution a synchronous machine produces. Overcurrent protection philosophies built around the assumption of substantial fault current from every source do not translate. Where the storage converter is a significant portion of the source capacity at a campus bus, the coordination study must model the converter’s actual current-limited contribution rather than an equivalent machine, and directional, differential, or communication-assisted schemes may be required where overcurrent alone cannot discriminate.

10.2 Ground Fault Overvoltage and Effective Grounding

When a converter and its interconnection transformer remain energised on a portion of the system that has been separated from the utility source by an upstream device operating for a ground fault, the grounding arrangement of the interconnection transformer determines whether the unfaulted phases rise to a level the system’s equipment insulation and surge arresters can tolerate. This is a design decision — winding configuration, grounding transformer, neutral treatment — that must be settled against the utility’s grounding practice, and it is a common cause of late-stage interconnection redesign when it is deferred.

10.3 Load Rejection and Transient Overvoltage

A converter charging at full rate that suddenly loses its source, or discharging into a load that disappears, produces a transient the design must bound. Combined with transformer energisation inrush, sympathetic inrush between adjacent transformers, and the interaction with surge arrester duty, this belongs in the transient study scope rather than in the assumption that converter controls will handle it.

10.4 Islanding and Interconnection Compliance

Unintentional islanding detection, ride-through obligations during voltage and frequency excursions, and disturbance monitoring and reporting obligations attach to the resource according to its size, voltage level, and the interconnection framework it falls under. The requirements differ materially between a distribution-connected resource and a transmission-connected one, and between market regions. Establishing which framework applies is a first-week question, not a permitting-phase question, because it drives model requirements, telemetry, protection, and testing scope.

The following is the study scope an owner should expect for a storage system interconnected at a data center campus. Scope varies with size, voltage, and jurisdiction, but a proposal that omits several of these is a proposal that has deferred risk rather than priced it.

13. Measurement and Compliance Verification

12. The Mitigation Toolkit

Mitigation should follow from the studies, in order of cost and intrusiveness. Specifying a filter before running a frequency scan is the most common and most expensive sequencing error in this work.

Notice and Disclaimer

This document is original technical content prepared by Keentel Engineering LLC for general professional information. It is not project-specific engineering advice and does not constitute a design, a study, an interconnection application, or a compliance determination for any particular facility. Study scope, interconnection requirements, and power quality obligations vary by system operator, utility, voltage level, resource size, and jurisdiction, and the requirements applicable to a specific project govern that project.

Standards and regulatory references are provided by subject for orientation. The current published edition of each standard and the requirements adopted by the applicable authority, utility, or system operator govern. Equipment behaviour, ratings, and model characteristics must be taken from the specific manufacturer’s validated data for the equipment selected.

Keentel Engineering LLC is an independent engineering consultancy. Reference to any standard, code, market rule, industry organisation, or equipment category in this document does not imply affiliation with, endorsement by, or sponsorship from any such organisation or manufacturer.

5. The Data Center Complication

A compliance claim that rests on simulation alone is incomplete. The verification sequence that holds up is straightforward and is frequently skipped.

Measure background power quality at the point of common coupling before the storage system is energised, using instrumentation of the appropriate measurement class, over a period long enough to capture the site’s weekly load cycle. Without this baseline, no post-energisation measurement can distinguish the storage system’s contribution from what was already there.
Record the facility’s own distortion contribution before energisation as well. On a data center campus this is substantial and it is the owner’s, not the utility’s.
After energisation, measure across the operating envelope — charge, discharge, idle, and transition — at the load conditions the study identified as governing, not merely at a convenient one.
Evaluate against the limits using the statistical basis the standard specifies rather than instantaneous peaks. Compliance frameworks are written around percentile values of aggregated measurement intervals, and a single high sample is not a violation any more than a single low sample is a pass.
Repeat the assessment at each phase of campus build-out. Cable capacitance, load, and short-circuit strength all change, and the resonance condition moves with them.
Retain the measurements. When a power quality dispute arises with the utility or with an equipment vendor, the party holding a defensible baseline and a documented commissioning survey is in a completely different position from the party holding a simulation report.

14. Reading the Graphic Correctly

The four-panel summaries of BESS power quality effects that circulate widely are directionally useful and technically loose in ways that matter. Five clarifications.

Clarification 1 — The metric is total demand distortion, not THD

Stating that a storage system "increases THD levels" points at the wrong number. Interconnection compliance for current is assessed as total demand distortion at the point of common coupling, referenced to maximum demand load current. Voltage total harmonic distortion is a separate limit belonging to the system operator. Using the terms interchangeably produces both unnecessary filter purchases and missed non-compliance.

Clarification 2 — The converter does not create resonance

Resonance is created by network inductance and capacitance. The converter excites a condition that already exists. This matters practically because it locates the fix: the answer is a frequency scan and, if required, a change to the network — filter, reactance, or point of connection — not a search for a better-behaved converter.

Clarification 3 — Storage is usually the cure for voltage fluctuation, not the cause

A converter with fast reactive control and configurable ramp limits is one of the most effective voltage stabilisation devices available at a campus bus. Presenting voltage fluctuation solely as a risk of adding storage inverts the more common engineering reality, where the storage system is dispatched to correct fluctuation caused by the facility’s own load.

Clarification 4 — The dominant risks are missing from the four panels

Harmonics, voltage fluctuation, and resonance are the classic list. On modern projects the risks that actually stop or delay work are grid strength and converter control interaction, ground fault overvoltage and the interconnection transformer grounding arrangement, ride-through compliance and disturbance monitoring obligations, and model availability and validation. None of these appear on a four-panel summary and all of them are schedule-critical.

Clarification 5 — Compliance is a set of operating points

Every one of these effects varies with dispatch, facility load, network configuration, and build phase. A single-case study and a single commissioning measurement are not evidence of compliance; they are evidence of one condition. The study set and the measurement protocol both have to be built around the envelope.

15. Keentel Data Center and Storage Engineering Services

3. The Inverter Is the Asset

16. Frequently Asked Questions

Keentel Engineering provides the interconnection engineering , power system studies, and facility electrical design that determine whether storage added to a data center campus behaves as intended and can be approved. Our practice in this area covers the following.

15.1 Interconnection and Point-of-Interconnection Engineering

Large-load and generation interconnection support: load and resource characterisation, application-stage technical packages, coordination with the serving utility, transmission provider, and system operator, and management of study-phase technical questions.
Point-of-interconnection design, substation design, and medium-voltage collection and campus distribution design for phased data center developments.
Interconnection transformer configuration and effective grounding assessment, settled early rather than discovered during utility review.

15.2 Power System Studies

Load flow, short circuit, protective coordination and selectivity, and arc-flash analysis, with inverter-based resources modelled as current-limited sources rather than as equivalent machines.
Harmonic frequency scan and injection studies across build-out phases and operating combinations, with compliance assessed at the correct point using the correct metric.
Flicker and voltage fluctuation assessment, grid strength evaluation, and transient stability analysis.
Electromagnetic transient modelling for control interaction, weak-grid stability, ride-through verification, and switching and overvoltage transients, using validated manufacturer models.
Transformer harmonic loading assessment, insulation coordination, and ground grid and step-and-touch analysis.

15.3 Facility Electrical Design and Owner’s Engineer Support

Campus distribution topology, UPS and generator plant integration with storage, and white-space power distribution designed against the availability and maintainability requirements.
Converter and equipment specification including model deliverable requirements, ride-through and control configuration, and factory test scope.
Design review of EPC and vendor submittals, QA/QC of third-party study packages, commissioning specification and test script support, and power quality baseline and verification survey scoping.
NERC compliance support for registered resources, including disturbance monitoring, ride-through, and model-related obligations.

Keentel Engineering holds a Florida Certificate of Authorization and maintains offices in Tampa, Austin, Sacramento, and Baltimore, supporting projects across the interconnections.

PSCAD Models for Inverter OEMs | EMT Guide

Thu, 27 Aug 2026 14:15:05 GMT

Aug 27, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

PSCAD Models for Inverter OEMs: What Manufacturers Need to Know About EMT Model Development

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

If you manufacture grid-connected power electronics — solar inverters, battery inverters, wind converters, STATCOMs, electrolyser rectifiers, EV charging power units, or data centre power systems — you will be asked for a PSCAD model. Not by a regulator, but by your own customers, because their interconnection studies cannot proceed without one.

Most OEMs encounter this requirement for the first time when a sale is already in motion. The developer's consultant asks for the EMT model, the OEM has nothing to send, and the project stalls. Sometimes the developer switches to a vendor who has one.

This article explains what an EMT model is, why the requirement exists, what a manufacturer-grade PSCAD model actually contains, what data is needed to build one, how long it takes, and where these projects go wrong. It is written for OEM engineering and product managers who need to understand the scope before committing to it.

The requirement is not new, but it has hardened considerably.

ERCOT began requiring EMT models for all new inverter-based resource interconnections in 2016, initially driven by subsynchronous resonance and subsynchronous control interaction concerns. What started as a regional response to a narrow stability problem has become a general expectation across North America.

Three things drove the expansion.

Real disturbances with real consequences

A sequence of events across the Western and Texas Interconnections showed inverter-based resources tripping or entering momentary cessation during faults that they should have ridden through. The June 2022 Odessa Disturbance in Texas is the reference case: a single transmission fault resulted in the unexpected loss of roughly 1,711 MW of solar PV output. Post-event analysis found that positive-sequence models had not predicted the observed behaviour, because the behaviour originated in fast inner-loop control and protection logic that RMS models do not represent.

Regulatory response

FERC Order No. 901, issued in October 2023, directed NERC to close reliability gaps for inverter-based resources across data sharing, model quality, model validation and performance. That directive produced the PRC-028 / PRC-029 / PRC-030 standards family and NERC Project 2022-04 on EMT modelling. PRC-029-1, approved by FERC in Order 909 in July 2025, establishes performance-based frequency and voltage ride-through requirements for IBRs and adopts the IEEE 2800-2022 performance curves. It becomes effective 1 October 2026 for Bulk Electric System facilities and 1 January 2027 for non-BES facilities. The applicability threshold dropped from 75 MVA to 20 MVA.

Falling system strength

As synchronous generation retires, short-circuit ratios at interconnection points fall. Converter control loops that are stable at SCR 5 can become unstable at SCR 2. Positive-sequence simulation cannot reliably predict this. EMT simulation can.

The net effect for an OEM: your equipment cannot be studied without an EMT model, and increasingly it cannot be sold without one either.

Who gets asked

Manufacturers of utility-scale PV and BESS inverters
Wind turbine converter suppliers
STATCOM and SVC vendors
HVDC and MVDC converter suppliers
Electrolyser and industrial rectifier manufacturers
Data centre power system suppliers, as large-load ride-through expectations tighten
EV charging power unit manufacturers at megawatt scale
Plant controller (PPC) vendors, whose control layer must be represented alongside the converter

Summary

Model quality tests exist because reviewers were receiving models that would not run. They are basic in concept and unforgiving in practice.

ERCOT introduced model quality test requirements effective May 2020, initially for PSS®E dynamic models, and requires MQT reports for all new and updated models — PSS®E and PSCAD. Since 2024, PGRR-109 has required IBRs to submit an as-built model alongside the quarterly stability assessment model and overlay the MQT plots before commissioning. ERCOT's NOGRR-245 updated voltage and frequency ride-through requirements to align with IEEE 2800, and ride-through capability must be demonstrated in models.

Typical test content across ISOs:

A model that fails MQT is returned for retuning and resubmission. For an OEM, that means the customer's schedule slips and the OEM is visibly responsible.

Power Systems Modeling & NERC Compliance

3. Model types: black box, white box, and real code

1. Why OEMs are suddenly being asked for PSCAD models

Power system simulation operates in two domains, and the distinction matters because ISOs generally want models in both.

Positive-sequence (RMS / phasor domain)

Tools: PSS®E, PSLF, TSAT, PowerFactory RMS. Typical timestep: 4–10 milliseconds. The network is represented as balanced positive-sequence phasors at fundamental frequency. Fast switching, harmonics, and unbalance are not represented. These models run large networks quickly, which is why they remain the backbone of planning studies.

Electromagnetic transient (EMT)

Tools: PSCAD/EMTDC, EMTP-RV, PowerFactory EMT, RSCAD. Typical timestep: 1–20 microseconds — roughly a thousand times finer. Instantaneous three-phase quantities are solved, so switching behaviour, unbalance, harmonics, DC offset, saturation and control-loop dynamics down to the current regulator are all represented.

An inverter's response to a fault is governed by control loops with bandwidths in the hundreds of hertz to low kilohertz. A 4 ms timestep cannot see them. That is the entire reason EMT models are required: the phenomena that caused the disturbances are invisible in the phasor domain.

Where EMT is specifically required:

Weak grid and low-SCR stability assessment
Subsynchronous resonance and subsynchronous control interaction
Phase angle jump response
Unbalanced fault response and negative-sequence current injection
Harmonic interaction and control instability between adjacent plants
Detailed ride-through verification against IEEE 2800 curves
Series-compensated line applications
Grid-forming behaviour and islanding

Where positive-sequence remains sufficient: bulk transient stability, large-area planning, most contingency screening.

Because the two must agree, cross-validation between the EMT and RMS models is now a formal deliverable in its own right — and often the item that exposes errors in both.

OEMs face a genuine tension. The model must be detailed enough to reproduce controller behaviour, but the controller is the product. Three approaches manage that tension differently.

Black box (compiled, encrypted)

The control logic is implemented in the EMT tool, then compiled into a library and encrypted. Third parties can run the model, set parameters and view terminal quantities, but cannot inspect or extract the control code.

Protects IP: strongly
Fidelity: high, limited by the accuracy of the implementation
Accepted by ISOs: yes, this is the standard commercial form
Maintenance: must be rebuilt when firmware changes

This is what the overwhelming majority of OEM models are, and what most manufacturers should build.

White box (open source)

Control logic visible and editable. Rare in commercial practice — used internally, in research, or where a utility contractually demands full visibility.

Protects IP: no
Fidelity: high
Use case: internal development, academic collaboration

Real code (firmware-derived)

The production controller source — typically C running on a DSP, sometimes with an FPGA layer — is converted into the EMT environment rather than re-implemented from block diagrams. The simulated controller executes the same logic as the physical controller.

Protects IP: yes, when compiled and encrypted
Fidelity: highest available in software
Cost and effort: substantially higher
Advantage: tracks firmware releases with much lower revalidation effort
Requirement: sustained access to firmware source and the build toolchain

Real-code models are becoming the expectation in some markets and for some equipment classes. For most OEMs entering the space, a well-built black-box model is the correct first deliverable, with real-code conversion considered later if market pressure justifies it.

"A PSCAD model" is not a single artefact. A model package that will survive ISO review has several layers, and OEMs consistently underestimate the ones below the surface.

4.1 Power stage

Converter topology and switching representation — two-level, three-level, NPC, ANPC, modular multilevel as applicable
Modulation scheme, carrier arrangement, dead time
DC link capacitance, pre-charge circuit, DC bus operating window
AC filter topology and component values, including damping
Interface transformer: rating, vector group, impedance, X/R, tap range, and saturation characteristic where relevant
Internal impedances between modules, cabinet and point of connection
DC-side source or load representation — battery, PV array, wind generator, rectifier load — with its own dynamics

An important practical point: full switching representation is expensive to simulate. Many production models use an averaged or hybrid switching representation that reproduces terminal behaviour accurately at a manageable timestep. Whether that is acceptable depends on the study type, and the choice must be documented, not silently made.

4.2 Control system

This is where the effort concentrates.

Phase-locked loop: structure, bandwidth, and behaviour under unbalanced and distorted supply
Inner current regulator: reference frame, bandwidth, decoupling, anti-windup
Outer loops: DC voltage, active power, reactive power, AC voltage, power factor
Current limiting strategy and active/reactive priority under fault
Negative-sequence current injection behaviour
Ride-through logic: momentary cessation behaviour, blocking and de-blocking thresholds, recovery ramp rates
Autonomous grid support functions: volt-VAR, volt-watt, frequency-watt
Sample rates, computation delays, actuation delays, measurement filter time constants
Start-up, synchronisation, soft-start, pre-charge, shutdown sequences

Delays and filter time constants deserve special mention. They are routinely omitted from OEM documentation and routinely decisive in weak-grid stability results. A model with correct gains and wrong delays will look right in a strong grid and be wrong where it matters.

4.3 Protection

Protection is not optional detail. It determines whether the model trips, and tripping behaviour is precisely what ISOs are trying to assess.

AC and DC over/under voltage and over/under current, with pickup levels and time delays
Over and under frequency
Phase jump, RoCoF and vector shift protection, and whether these are user-defeatable
Thermal, ground fault, insulation monitoring
Anti-islanding scheme where applicable

PRC-029-1 recommends disabling phase-jump protection that trips unnecessarily during permissible phase angle excursions. A model that does not represent the protection cannot demonstrate compliance either way.

4.4 Plant controller interface

Most ISOs require that the model accept external setpoints in the way the real plant does. SPP's PSCAD model requirements, for example, specify that the plant power controller accept an external active power setpoint and a voltage setpoint, implement a settable voltage droop, and that the model initialise to the setpoints specified in the PPC. Communication latency and setpoint update rate must be represented.

4.5 Aggregation

A plant contains many inverters. Simulating each one individually is impractical. The model must support a documented aggregation approach — typically a single-unit equivalent scaled to plant rating, sometimes with a small number of representative units where diversity matters. The aggregation basis and its limitations belong in the documentation.

4.6 Numerical behaviour

The least visible layer, and the one that most often causes rejection.

Timestep: models are generally expected to run at 10–20 µs. Some ISOs accept smaller timesteps where the penalty is not significant, but a model that only runs at 1 µs will be unwelcome in a large study case.
Initialisation: the model must reach steady state cleanly from a flat start and initialise to the specified setpoints, without manual intervention.
Snapshot capability: studies rely on saving and restoring a converged state. A model that cannot snapshot forces every run to re-initialise.
Multi-instance: dozens of instances in one case, each with independent parameters and no shared-state collisions.
Parallel and multi-core execution: required for large study cases.
Compiler compatibility: typically Intel Fortran, against the PSCAD versions the ISO and the consultant are running.

4.7 Documentation and packaging

Encrypted compiled library, installer, and licence file where used
User manual: installation, parameters with ranges and units, control modes, aggregation guidance
Parameter data sheet mapping model parameters to real product settings
Validation report
Model quality test conformance record
Release note recording model version, firmware version modelled, and known limitations

7. The PSS®E companion model

Conventional inverters are grid-following: they synchronise to a measured voltage through a PLL and inject current. This works while the grid is strong. As system strength falls, PLL-based control approaches its stability limit.

Grid-forming control instead regulates voltage magnitude and angle directly, presenting as a voltage source behind an impedance. Grid-forming inverters can support islanded operation, contribute to system strength, and provide inertial response.

For OEMs, three implications:

Tender requirements are shifting. Grid-forming capability is increasingly requested for BESS, microgrid and backup power applications, and is being specified in some interconnection requirements.
Grid-forming models require different validation. The relevant tests are weak-grid and no-grid cases, black start, load steps in islanded operation, and parallel operation with other grid-forming units — not the standard grid-following test matrix.
Dual-mode products need dual-mode models. If the product can operate in either mode, the model must represent both, including the transition.

Modelling a grid-forming mode is not a small increment on a grid-following model. Budget for it separately.

2. EMT versus positive-sequence: why both exist

5. Model quality testing: what reviewers actually run

6. Validation: the part that determines everything

A model that has not been validated against hardware is an assertion, not evidence. Validation is also, consistently, the item that determines whether a modelling project runs to schedule.

What is needed

The distinction that matters most: test certificates are not validation data. A UL 1741 SB or IEC 61851 certificate states that a test was passed. It does not contain the waveform. Validation requires raw records — COMTRADE or equivalent — showing the measured response.

Required records typically include:

Low and high voltage ride-through, balanced and unbalanced, across the applicable envelope
Frequency ride-through and RoCoF events
Phase angle jump
Reference step responses: active power, reactive power, AC voltage, DC voltage
Weak grid stability tests with the tested SCR stated
Harmonic emission and power quality records
Hardware-in-the-loop results, where available
Field disturbance records from operating installations, where available

Sources of validation data

Type test campaigns. UL 1741 SA/SB, IEEE 1547.1, IEC 61400-21, IEC 62116 and equivalents generate exactly the disturbance responses needed — but only if raw captures were retained. Many were not.
Hardware-in-the-loop. The most efficient source. A real controller running against a simulated grid can produce any test case on demand.
Factory test records. Useful for control response, usually not for ride-through.
Field records. Highest credibility for post-event validation, but opportunistic.

When the data does not exist

This is common, particularly for products that were certified before EMT modelling became a commercial requirement. Options, in descending order of preference:

Run a targeted HIL campaign against the specific test cases the model needs
Run a supplementary laboratory campaign at a grid simulator facility
Validate against a subset and document the unvalidated envelope explicitly
Proceed on documented assumptions, accepting reduced defensibility under review

The last option is legitimate but should be a conscious decision. Every assumption belongs in an assumptions register reproduced in the validation report. Reviewers do not object to assumptions; they object to discovering them later.

Most OEMs asking for a PSCAD model actually need two models.

Transmission providers run planning studies in the phasor domain and detailed studies in EMT. Both need to reflect the same equipment, and the requirement to show they agree is now explicit — ERCOT's MQT includes benchmarking PSCAD against PSS®E, and NERC's EMT modelling work formalises cross-validation practice.

Two routes exist for the positive-sequence model:

Generic library models. The WECC/IEEE second-generation renewable energy models — REGC_A/B/C, REEC_A/B/C/D, REPC_A/B, and their equivalents — parameterised to represent the product. Fast, widely accepted, and adequate for many products. The limitation is real: generic models cannot represent behaviour their structure does not contain.

User-written models. A custom compiled dynamic model. Higher fidelity, higher effort, higher maintenance burden.

Whichever route is taken, developing the EMT and RMS models in parallel is materially more efficient than sequentially, because cross-validation failures usually indicate an error in one of them, and finding that error early is cheaper than finding it at submission.

8. Grid-forming: the emerging second requirement

10. Where these projects go wrong

9. Schedule: what actually drives it

A manufacturer-grade PSCAD model for a single product, grid-following, black box, with a companion positive-sequence model, is realistically a 650 to 900 hour engagement. On a standard programme that is 18 to 24 weeks. Compressed with parallel workstreams and a dedicated team, 12 weeks is achievable. A project-specific model for one named study can be delivered in around 5 weeks.

What sets the position within that range is rarely engineering difficulty. It is:

Documentation quality. Complete control block diagrams with gains, time constants and per-unit bases put a project at the fast end. Marketing datasheets put it at the slow end.
Whether documentation matches shipping firmware. The most expensive discovery in this work is that the block diagrams describe a control structure that was superseded two firmware revisions ago.
Validation data availability. Raw waveforms available: fast. Certificates only: add substantial time or a test campaign.
OEM responsiveness. Query turnaround is the single largest schedule variable in practice. A 48-hour response cadence and a 10-day response cadence produce very different delivery dates from identical scopes.
Number of variants. Each hardware variant and each firmware baseline is a separate modelling exercise, not a parameter change.

Accelerating delivery

Two techniques genuinely compress the critical path:

Interim model release. A numerically stable, functionally complete model can be issued before formal validation and packaging are complete. The customer begins preliminary study work weeks before formal release, under a clear limitation statement, and receives the validated model when it is ready. From the customer's perspective the schedule halves.

Parallel workstream execution. Power stage and control development run concurrently; documentation is drafted alongside validation rather than after it; the positive-sequence model is developed in parallel with the EMT model throughout.

Technical

Patterns that recur across OEM engagements.

Treating it as a documentation exercise. It is a development project with its own validation, version control and release management.

Underestimating numerical hardening. A model that reproduces every waveform correctly but cannot snapshot, cannot run multi-instance, or requires a 1 µs timestep will be rejected. This layer routinely consumes 15 to 20 percent of total effort.

Certificates instead of waveforms. Discovering at week eight that no raw records exist.

No version control. A model released without recording which firmware version it represents becomes unmaintainable within two release cycles.

No maintenance plan. Every firmware release touching grid-facing control or protection invalidates the released model. An OEM with a two-year-old model and four firmware releases since has a model that misrepresents its own product — a genuine liability once customers have submitted it to regulators.

Modelling one variant and issuing it for all. Different hardware ratings and different firmware baselines require separate models or, at minimum, a documented validity statement.

Firmware and modelling teams not talking. The model is derived from control logic. If the firmware team is not directly engaged, the model is being built from documentation, and documentation drifts.

Frequently Asked Questions

4. What a manufacturer-grade PSCAD model actually contains

11. Maintenance and versioning

The deliverable is not one-time. A sustainable arrangement includes:

A defined trigger for revalidation — typically any firmware change affecting control, protection or grid-facing behaviour
Version numbering that ties model version to firmware version explicitly
A release note recording what changed and what was revalidated
Migration across PSCAD major versions as they are released
First-line support for customers and their consultants running the model

Per-release revalidation of a black-box model is typically 40 to 80 hours. An annual retainer is usually more economical than release-by-release engagement, and it avoids the situation where a customer needs an updated model faster than a procurement cycle can produce one.

General

Working with Keentel Engineering

Data and validation

Regulatory

Commercial and process

Keentel Engineering provides power systems engineering and NERC compliance services to utilities, developers, EPCs, OEMs and public agencies across the United States. Our modelling practice covers:

PSCAD EMT model development for inverter-based resources, including black-box packaging and encryption
PSS®E, PSLF, TSAT and PowerFactory dynamic model development and user-defined models
EMT-to-RMS cross-validation and benchmarking
ERCOT Model Quality Test preparation and submission support
PRC-029-1 applicability assessment and ride-through design evaluation
MOD-026 / MOD-027 model verification and parameter verification reporting
Weak-grid, low-SCR and subsynchronous studies
Model maintenance and revalidation across firmware and software releases

If you manufacture grid-connected power electronics and need a model your customers can submit, we can scope it from your existing documentation and tell you what is missing before you commit to anything.

Keentel Engineering

Head Office: 400 N Ashley Dr STE #2600, Tampa, FL 33602 — (813) 389-7871

Austin: 5900 Balcones Drive STE 100, Austin, TX 78731 — (512) 591-0752

Sacramento: 1401 21st St Ste R, Sacramento, CA 95811 — (916) 913-4524

Baltimore: 306 W Redwood St STE 200, Baltimore, MD 21201 — (410) 225-2181

contact@keentelengineering.com | keentelengineering.com

This article is provided for general information and does not constitute legal, regulatory or engineering advice for any specific project. Regulatory effective dates, thresholds and ISO requirements are subject to change; verify current requirements with the applicable authority. Information reflects Keentel Engineering's understanding at the date of publication.

12 Electrical Safety Rules Every Engineer Must Know

Thu, 27 Aug 2026 10:27:03 GMT

Aug 27, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Twelve Electrical Safety Rules Every Engineer Must Know

And the OSHA, NFPA 70E, NESC, NEC and IEEE requirements that sit behind each one

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

The single most common source of confusion in electrical safety is treating "the code" as one document. In the United States there are at least five separate instruments in play, they have different legal force, and they apply to different installations. Before discussing any individual rule, it is worth being precise about which document is doing the work.

1. Always De-Energize — And Understand the Only Two Exceptions

OSHA 1910.333(a)(1) states the requirement plainly: live parts to which an employee may be exposed shall be de-energized before the employee works on or near them. It then provides two, and only two, permissible bases for working energized. The first is that de-energizing introduces additional or increased hazards — the classic examples are shutdown of life-support equipment, deactivation of emergency alarm systems, and shutdown of hazardous location ventilation. The second is that de-energizing is infeasible due to equipment design or operational limitations, which covers testing that can only be performed on an energized circuit and troubleshooting that cannot be done on a dead system.

Neither exception is a schedule exception. Production pressure, a customer who does not want an outage, a commissioning milestone, and the cost of a shutdown are not, on their own, legal justifications. NFPA 70E 130.2(A) mirrors this and adds the explicit clarification that additional hazards or infeasibility must be demonstrated, not asserted. Where energized work is genuinely justified, NFPA 70E 130.2(B) requires an Energized Electrical Work Permit documenting the circuit, the justification, the results of the shock and arc flash risk assessments, the PPE required, the means of restricting access, and the approving signatures.

The test to apply on site is simple. If someone cannot articulate which of the two exceptions applies, and cannot produce a signed permit that says so, the work is not authorized. That conversation is easier to have before the crew is mobilized than at the switchgear door.

Design lever: The number of energized work permits a facility issues is an engineering outcome, not a discipline outcome. Redundant feeds, properly placed isolation devices, drawout breakers, load-break switching, and dual-source designs are what make de-energizing feasible. Facilities that write permits every week usually have a single-line diagram problem, not a culture problem.

2. Lockout/Tagout — Know Which Rule You Are Under

Three OSHA provisions address the control of hazardous electrical energy and they are not interchangeable. 1910.147 is the general lockout/tagout standard for servicing and maintenance of machines and equipment. 1910.333(b) covers lockout and tagging for the specific case of working on de-energized electrical circuits. 1910.269(d) applies to generation, transmission and distribution work and includes its own hazardous energy control requirements, alongside 1910.269(m) for de-energizing lines and equipment for employee protection.

The sequence that satisfies all three is the same in structure: identify every energy source feeding the equipment, notify affected personnel, shut down using the normal operating sequence, isolate each source with a disconnecting means, apply a lock and a tag to each isolating device, release or restrain all stored energy, and then verify isolation by test. Each authorized employee applies their own lock.

Group lockout devices and lockbox arrangements are permitted, but they must preserve individual control — no worker should be relying on someone else remembering to keep them safe.

The step most commonly compressed is stored energy release. Capacitor banks, filter capacitors in drives, station battery systems, UPS units, DC buses in battery energy storage systems, charged cables, and springs in stored-energy breaker mechanisms all hold energy after the upstream source is open. Isolation of the AC source does not address any of them.

Backfeed deserves separate attention. Standby generators, parallel inverters, photovoltaic arrays energized by daylight regardless of any switch position, control power transformers fed from a different bus, and instrument transformer secondaries can all re-energize equipment that appears isolated. A lockout that traces only the primary feed is incomplete.

Design lever: Complete, current single-line diagrams with every source, tie and backfeed path shown are the difference between a lockout that works and one that misses a circuit. Where drawings are stale, a field verification and as-built update is a safety deliverable, not a documentation task.

3. Verify Before You Touch — Live, Dead, Live

NFPA 70E 120.6 establishes the process for establishing an electrically safe work condition, and the verification step is specific: an adequately rated portable test instrument shall be used to test each phase conductor or circuit part to verify it is de-energized, and the test instrument shall be verified as operating satisfactorily on a known voltage source before and after the test. This is the live-dead-live sequence. Testing the meter on a known source first proves it works; testing the circuit proves the circuit is dead; testing the known source again proves the meter did not fail between the first two steps.

The failure mode this guards against is not exotic. A blown internal fuse, a broken lead, a depleted battery, or a meter left in the wrong function will all display zero volts on a fully energized 480 V bus. Without the closing verification, zero on the display is indistinguishable from a dead instrument.

Instrument rating matters as much as sequence. Test instruments used on circuits above 50 V must be rated for the voltage and, critically, for the measurement category of the location. IEC 61010-1 category ratings reflect available fault energy, not just voltage: CAT III for distribution-level and fixed installation work, CAT IV for the service entrance and outdoor supply. A CAT II 1000 V meter used at a service entrance is rated for the voltage and unrated for the transient energy present there. NFPA 70E 110.4(A) also requires that instruments and leads be visually inspected for damage before each use.

Where opening the enclosure is itself the hazardous step, NFPA 70E 120.6(7) permits a permanently mounted absence-of-voltage tester listed to UL 1436 to satisfy the verification, eliminating the need to open the door to test. On new switchgear and MCC procurements this is a low-cost specification item with a direct risk reduction.

Design lever: Specify permanently mounted absence-of-voltage testers and voltage indicators on new medium-voltage and 480 V equipment. Specify infrared inspection windows so thermographic surveys do not require opening doors. Both remove routine energized exposure from the maintenance program permanently.

4. Use PPE — But Know Which Selection Method You Are Using

NFPA 70E permits two mutually exclusive methods for selecting arc flash PPE, and mixing them is a documented error. The incident energy analysis method under 130.5(G) uses calculated incident energy in cal/cm², typically from IEEE 1584-2018, and selects arc-rated clothing and equipment with an arc rating equal to or greater than that value at the working distance. The arc flash PPE category method under 130.7(C)(15) uses tables that specify a category based on equipment type, voltage, available fault current and clearing time, provided the installation falls within the stated parameters. The two methods shall not be used on the same piece of equipment for the same task.

Shock PPE is a separate assessment governed by 130.4. Rubber insulating gloves meeting ASTM D120 and OSHA 1910.137, with leather protectors, are selected by class for the nominal voltage. Under 1910.137(c)(2)(viii), rubber insulating gloves must be electrically tested before first issue and at intervals not exceeding six months thereafter; sleeves at twelve months. Gloves that have exceeded the test interval are not PPE, regardless of appearance.

Arc-rated clothing must meet ASTM F1506 and the entire system matters. A 12 cal/cm² arc-rated shirt worn over a polyester undershirt is not a 12 cal/cm² system — meltable synthetic fabrics underneath can ignite or melt to skin. Arc-rated face shields with balaclavas, or arc flash suit hoods above 12 cal/cm², protect the head and neck; safety glasses alone do not. Dielectric footwear provides secondary protection only and is never the primary shock control.

One point deserves emphasis because it is routinely misunderstood. Arc-rated PPE is designed to reduce the severity of a second-degree burn to a survivable level at the calculated working distance. It is not designed to make an arc flash a non-event, and it does not protect against the pressure wave, molten metal spray, or blunt trauma from equipment failure. PPE is the last control in the hierarchy, not the first.

Design lever: Incident energy is a calculated result, and it is calculable in advance. Fault current sources, protective device settings, clearing times and working distances are all engineering variables that Keentel adjusts during coordination studies to bring incident energy down before PPE is ever selected.

5. Use the Right Tools — Rated, Tested and Undamaged

OSHA 1910.335(a)(2)(i) requires insulated tools or handling equipment when working near exposed energized conductors where contact is possible, and requires that the insulating capability be appropriate for the voltage. Insulated hand tools for work up to 1000 V AC are manufactured and tested to ASTM F1505 and marked with the double-triangle 1000 V symbol. Tools that are merely dipped in plastic, comfort-gripped, or wrapped in tape are not insulated tools.

Live-line tools for medium and high voltage work fall under ASTM F711 for fiberglass-reinforced plastic rod and tube. OSHA 1910.269(j) requires that live-line tools be wiped clean and visually inspected before use each day, and removed from service and tested if any defect or contamination that could adversely affect the insulating qualities is found. Surface contamination and moisture absorption are the mechanisms that matter; a hot stick that has been left leaning against a fence in the rain is not a hot stick.

Insulated blankets, line hose, and covers under ASTM F479 and F478 provide secondary protection where accidental contact with adjacent parts is credible. They are not a substitute for de-energizing and are inspected before each use.

Two further practices belong here. First, non-conductive ladders — fiberglass, not aluminum — near any potentially energized circuit. Second, torque tools calibrated and used to the manufacturer specification on every electrical termination. Loose connections are the leading cause of thermal failure and one of the leading initiators of arc faults in distribution equipment; a calibrated torque wrench is a safety tool.

6. Keep the Work Area Safe — Illumination, Barriers and Boundaries

OSHA 1910.333(c)(4) prohibits employees from entering spaces containing exposed energized parts unless illumination is provided that allows the work to be performed safely, and prohibits reaching blindly into areas that may contain energized parts. Poor lighting inside a switchgear lineup is a recognized hazard, not an inconvenience.

Access control is governed by NFPA 70E 130.7(E), which requires safety signs, barricades and, where the situation warrants, an attendant to prevent unqualified persons from entering the arc flash boundary or approaching energized parts. Conductive barricade tape is prohibited near energized conductors. The arc flash boundary is the distance at which incident energy falls to 1.2 cal/cm² — the onset of a second-degree burn on bare skin — and inside it, anyone present requires arc-rated protection, including observers, apprentices and the project manager who wandered over to watch.

Wet conditions materially increase shock severity by lowering contact resistance, and 1910.333(c)(2) addresses conductive materials and equipment handled near energized parts. Standing water, damp concrete, and wet PPE all change the risk picture. In outdoor substation work, the same conditions raise step and touch potential exposure.

Housekeeping is not cosmetic. Egress from in front of switchgear during an arc fault is measured in fractions of a second, and equipment stored in the working space, in violation of NEC 110.26 dedicated space and working clearance requirements, becomes a trip hazard at exactly the wrong moment.

7. Follow Clearances — And Use the Correct Table

Three distinct sets of distances get called "clearance" and they are not the same thing.

Design clearances are the permanent physical separations built into an installation: NESC Rule 232 for vertical clearances of overhead conductors above ground, roadways and buildings, and NESC Rule 234 for clearances to buildings, signs and other installations. For premises wiring, NEC 110.26 establishes working space depth, width and headroom in front of electrical equipment, and NEC 110.34 does the same for equipment over 1000 V. These are installation requirements checked at design and inspection.

Approach boundaries are work practice distances from NFPA 70E 130.4. The limited approach boundary is the distance beyond which unqualified persons may not approach without an escort. The restricted approach boundary is the distance within which only qualified persons using shock PPE and working under a documented plan may go. Representative AC values from Table 130.4(E)(a) illustrate the scale, though the current edition governs and must be consulted directly.

Minimum approach distances for qualified line workers are separate again, from OSHA 1910.269(l) and its accompanying tables, and from 1910.333(c)(3) for general industry work near overhead lines. For work near overhead lines by unqualified persons and for vehicle and equipment operation near lines, 1910.333(c)(3)(iii) sets the familiar 10-foot minimum for lines up to 50 kV, increasing above that voltage.

Crane, aerial lift and material handling operations near energized lines are the highest-consequence application of this rule. The controlling requirement is a documented determination of line voltage, the corresponding minimum clearance, and a physical means of maintaining it — a dedicated spotter, range limiting, or de-energizing and grounding the line.

The Regulatory Stack: Five Documents, Five Different Jobs

Why the Poster on the Wall Is Not a Safety Program

The economics are not close

A single arc flash injury carries direct medical costs frequently exceeding one million dollars, plus outage cost, equipment replacement, OSHA penalties, litigation and insurance consequences. A facility-wide short circuit, coordination and arc flash study is a small fraction of that figure, and it is performed once with periodic updates.

The study also produces settings that improve selectivity, a verified single-line diagram, and documentation that survives an incident investigation. There is n

A Practical Sequence for Facilities Starting from Zero

Field-verify the single-line diagram, including every source, tie, backfeed path and disconnecting means. Nothing downstream is valid without this.
Obtain utility available fault current and clearing data at the point of service, and equipment nameplate and cable data throughout.
Perform the short circuit study and confirm equipment interrupting and withstand ratings are adequate for present and projected fault duty.
Perform the protective device coordination study, resolving selectivity against clearing time — recognizing that slower is more selective and more dangerous.
Perform the arc flash incident energy analysis to IEEE 1584-2018 and identify locations above practical PPE limits.
Implement mitigation where incident energy is excessive: settings changes, maintenance mode switches, arc energy reduction, equipment replacement.
Produce and apply labels compliant with NEC 110.16 and NFPA 70E 130.5(H), with study date and review interval.
Write the electrical safety program, EEWP process and job safety plan templates around the results, and train qualified persons on the actual equipment.
Establish the maintenance program to NFPA 70B, and PRC-005 where applicable, so the clearing times the study assumed remain true.
Set the review trigger: five years maximum, and immediately on any change to sources, protective devices or major equipment.

Representative values shown for illustration of the concept only. The current edition of NFPA 70E Table 130.4(E)(a), and the applicable OSHA tables for line work, govern in all cases.

8. Check Equipment — The "Normal Operating Condition" Test

NFPA 70E 130.2(A)(4) contains one of the most operationally important provisions in the standard, and it is frequently overlooked. Equipment is considered to be in a normal operating condition only when all five of the following are true: the equipment is properly installed; the equipment is properly maintained; all equipment doors are closed and secured; all covers are in place and secured; and there is no evidence of impending failure.

This matters because the risk assessment for tasks like operating a circuit breaker or a disconnect depends on the equipment being in normal operating condition. A switchgear lineup with a cracked bushing, an open panel cover, evidence of overheating or arcing tracks, water intrusion, corrosion, or no maintenance record is not in normal operating condition, and the likelihood of an arc flash event during operation increases accordingly. The task that was previously assessed as low risk is no longer that task.

"Properly maintained" is a defensible claim only where records exist. NFPA 70B, now issued as a standard rather than a recommended practice, establishes the electrical equipment maintenance program framework — condition assessment, intervals, testing and documentation. For protection systems on the bulk electric system, NERC PRC-005 imposes mandatory maintenance and testing intervals with auditable evidence. A protective relay that has not been tested to its PRC-005 interval cannot be assumed to clear a fault in the time the arc flash study used.

Inspection findings need a route to resolution. "Report any defects immediately" only functions where there is a defined recipient, a defined authority to remove equipment from service, and a work order system that closes the loop. Otherwise defects are reported into a void and the same finding appears in next year report.

Design lever: Arc flash results depend on protective device clearing time. If relay and breaker maintenance lapses, the calculated incident energy on every label in the facility silently becomes optimistic. Maintenance program integrity is an input to the study, not a separate topic.

9. Do Not Overload — Design Margin Is a Safety Margin

Overload is usually presented as a housekeeping rule about extension cords. It is more usefully understood as the field-level expression of a design constraint. NEC 210.19 and 215.2 establish conductor sizing for branch circuits and feeders, and NEC 210.20(A) and 215.3 require that overcurrent devices supplying continuous loads be sized at not less than 125 percent of the continuous load — the familiar 80 percent rule expressed from the other direction.

Ampacity is not a single number. NEC 310.16 values are subject to ambient temperature correction under 310.15(B) and adjustment for more than three current-carrying conductors in a raceway. Terminal temperature limitations under 110.14(C) frequently control the final answer, and a conductor selected from the 90°C column but landed on 75°C-rated terminations is limited by the terminations. For underground and duct bank installations, IEEE 835 and Neher-McGrath methods and soil thermal resistivity determine real capacity, often well below table values.

For transformers, IEEE C57.91 governs loading above nameplate and the relationship between load, temperature and insulation life. Sustained overload does not trip anything; it consumes insulation life and eventually produces a failure that is both an outage and an arc flash source.

Flexible cords and temporary power on construction sites carry their own requirements under NEC Article 400 and OSHA 1926 Subpart K, including GFCI protection or an assured equipment grounding conductor program. Daisy-chained relocatable power taps and cords run through doorways or under carpet remain among the most common findings in facility inspections.

10. Use Proper Labeling — The Label Is an Engineering Deliverable

NEC 110.16(A) requires field-applied arc flash warning labels on switchboards, switchgear, panelboards, industrial control panels, meter socket enclosures and motor control centers in other than dwelling units, where the equipment is likely to require examination, adjustment, servicing or maintenance while energized. NEC 110.16(B) adds a requirement for service equipment rated 1200 A or more to carry a label with specific arc flash information.

The content requirement is in NFPA 70E 130.5(H). The label shall contain the nominal system voltage, the arc flash boundary, and at least one of the following: available incident energy and the corresponding working distance; minimum arc rating of clothing; site-specific level of PPE; or the arc flash PPE category from the tables, where that method is used. Labels shall be legible, durable, and consistent with ANSI Z535.4 formatting. Where the incident energy analysis method is used, the label must state the date of the study and the review interval.

NFPA 70E 130.5 requires the arc flash risk assessment to be reviewed at intervals not to exceed five years, and whenever a major modification or renovation occurs, or when changes in overcurrent protective devices or upstream system capacity render the results invalid. A utility service upgrade, a new transformer, a generator addition, or a relay setting change all potentially invalidate every downstream label.

Beyond arc flash, the labeling regime includes NEC 408.4 circuit directories and source identification for panelboards, NEC 110.22 marking of each disconnecting means with its purpose, NEC 110.21(B) requirements for field-applied hazard markings, and equipment nameplate and circuit identification consistent with the single-line diagram. A lockout is only as good as the ability to identify the correct disconnect, and a mislabeled panel is a shock hazard with a delay fuse on it.

Design lever: Labels are the visible output of the short circuit, coordination and arc flash study chain. A facility with unlabeled or expired equipment does not have a labeling problem; it has an unfinished or outdated power system study.

11. Respect Energy — Assume Live Until Proven Dead

The instinct behind this rule is correct and worth restating in engineering terms: electricity provides no sensory warning, and several energy sources present in modern facilities are not intuitive.

Instrument transformer secondaries are the classic example. A current transformer secondary must never be open-circuited while the primary is energized — the CT will attempt to drive its magnetizing current through an open circuit and can develop lethal secondary voltages and destroy itself. Shorting blocks exist for this reason and their use is not optional. Voltage transformer secondaries can backfeed the primary if energized from the secondary side, re-energizing a "dead" bus at full primary voltage.

Capacitive coupling and induction on de-energized conductors running parallel to energized circuits — a de-energized transmission circuit on a double-circuit structure, a de-energized cable in a shared duct bank — can produce dangerous voltages and currents on a line that has been properly isolated. This is why protective grounding under OSHA 1910.269(n) exists: grounds are applied to provide a path that limits the voltage a worker can be exposed to, and they must be rated for the available fault current, applied in the correct sequence, and placed so the worker is within the protected zone.

DC systems are increasingly the exposure. Station batteries in substations, DC buses in battery energy storage systems, photovoltaic array DC that is energized whenever there is light regardless of any switch position, and drive DC links with capacitor storage all persist after AC isolation. DC arc faults do not have a natural current zero and behave differently from AC arcs; DC arc flash analysis requires a different methodology than IEEE 1584, which is an AC method.

Step and touch potential in substation and line work is the failure mode that kills people who never touched a conductor. During a ground fault, the earth surface develops a voltage gradient; the difference between the potential at a person hand contact point and their feet, or between their two feet, drives current through the body. IEEE 80 is the governing design methodology, and an adequate ground grid, surface layer resistivity and equipment bonding are what keep those potentials below tolerable limits.

Design lever: Grounding design under IEEE 80, protective grounding procedures, and DC-specific arc flash analysis for BESS and PV installations are engineering scopes, not field procedures. Where they are missing, no amount of field discipline substitutes.

12. Be Trained and Alert — Qualified Means Something Specific

OSHA 1910.399 defines a qualified person as one who has received training in and has demonstrated skills and knowledge in the construction and operation of electric equipment and installations and the hazards involved. NFPA 70E 110.6 expands this: a qualified person must be able to distinguish exposed energized parts, determine nominal voltage, know the approach distances and the corresponding voltages, understand the process of risk assessment and hazard elimination, and be able to select and use the appropriate PPE. Qualification is task-specific and equipment-specific. A person qualified on 480 V motor control is not thereby qualified on 15 kV switchgear.

Retraining is required at intervals not to exceed three years under NFPA 70E 110.6(D)(3), and additionally when work practices change, when new equipment or procedures are introduced, or when supervision or annual inspection indicates the employee is not complying with safe work practices.

The job briefing requirement under OSHA 1910.269(c) is one of the highest-value and lowest-cost controls available. Before each job, the person in charge covers the hazards, work procedures, special precautions, energy source controls and PPE requirements. For repetitive or similar work during the same day, an abbreviated briefing is permitted, but a briefing at the start of each shift is required. NFPA 70E 110.5(C) requires an equivalent briefing and, for work within the limited approach boundary or the arc flash boundary, a documented job safety plan.

Distraction and complacency are named on every poster because they are real, but the engineering-adjacent point is that the most dangerous condition is a routine task performed on equipment whose condition has changed. The crew is not less alert; the equipment is different from the last time. That is precisely what Rule 8 exists to catch.

The practical implication is direct. If you are working inside a commercial or industrial facility, your work practice obligations flow from 1910.331–.335 and your PPE and boundary decisions from NFPA 70E, while the installation itself was built to the NEC. If you are working in a substation, on a transmission line, or inside a generating station, 1910.269 governs your work practices and the installation was built to the NESC. Applying NEC working-space rules to a 138 kV switchyard, or NESC clearances to a 480 V motor control center, produces answers that are not merely conservative or liberal — they are the wrong answers to the wrong question.

A correction worth making

Safety graphics frequently state that clearance from overhead power lines is "per NEC/IEC." In the United States this is incorrect. Design clearances for overhead supply conductors come from the NESC (ANSI C2). Minimum approach distances for workers come from OSHA 1910.333(c)(3) for general industry and OSHA 1910.269 Table R-3 through Table R-8 for qualified line work. The NEC explicitly excludes utility-owned outdoor supply installations from its scope.

This is not pedantry. The distances differ materially, and citing the wrong document in a job hazard analysis will not survive an incident investigation.

In an Emergency

Closing

Where Safety Is Actually Decided: The Design Phase

The response sequence taught on posters is correct and worth stating with the reasoning attached. Do not touch a victim who may still be in contact with an energized conductor — rescuer electrocution is a documented and recurring outcome. De-energize the source if it can be done immediately and safely; otherwise use an approved insulated contact-release tool rated for the voltage. Call for emergency medical response immediately; electrical contact injuries include cardiac arrhythmia and deep tissue damage that are not visible externally and require evaluation even where the victim appears uninjured.

NFPA 70E 110.6(C) requires emergency response training for employees exposed to shock hazards, including first aid, CPR and AED use, with annual verification of CPR and AED competence and annual verification of contact release training. OSHA 1910.269(b) sets first aid and CPR availability requirements for line work. Preserve the scene and equipment condition for investigation, and report the incident through the established process — the corrective action that prevents the next event depends on accurate information from this one.

Frequently Asked Questions

Twelve rules on a wall are a useful reminder for people who already understand what sits behind them. They are a poor substitute for the understanding itself. Every rule in this article is downstream of a decision an engineer made — how the system was configured, what the protective devices were set to, whether the isolation points exist, whether the study was performed, whether the labels are current, whether the maintenance program preserved the assumptions the study relied on.

The most reliable way to keep people safe around electrical systems is to design and maintain systems where the safe method is also the convenient one: where de-energizing does not require an outage negotiation, where verification does not require opening a door, where incident energy is low enough that ordinary arc-rated clothing is sufficient, and where the labels tell the truth. That is engineering work, and it happens before anyone is standing in front of the equipment.

Keentel Engineering — Power System Studies

Short circuit, coordination and arc flash analysis (IEEE 1584-2018) · Grounding studies (IEEE 80) · Cable ampacity and duct bank thermal analysis · Relay setting development and verification · Arc flash labeling programs · Protection system maintenance program support (NERC PRC-005) · Owner engineer services.

Tampa, FL · Austin, TX · Sacramento, CA · Baltimore, MD | 813-389-7871 | contact@keentelengineering.com | FL Firm Reg. No. 36853

Walk into almost any electrical contractor trailer, substation control building or plant maintenance shop and you will find a laminated poster listing ten or twelve electrical safety rules. De-energize. Lock out. Test before you touch. Wear your PPE. The rules are correct, they are memorable, and they are not sufficient. A poster tells a worker what to do. It does not tell them how the requirement is written, when an exception legitimately applies, who is authorized to approve that exception, or what engineering work has to happen months earlier so that the rule can actually be followed on the day.

That gap is where people get hurt. A technician who has memorized "always de-energize" but has never seen an energized electrical work permit will improvise when a plant manager says the line cannot come down. A crew that knows to "wear PPE" but works from an arc flash label produced from a study that was never updated after a transformer replacement is wearing the wrong PPE with complete confidence. A field engineer who was taught that clearances come from the National Electrical Code will apply the wrong clearance table entirely when working near an overhead distribution circuit, because the NEC does not govern utility overhead lines.

This article takes the twelve rules that appear on those posters and does what the poster cannot: connects each one to the governing requirement, explains the failure mode it is guarding against, and identifies the engineering decision that determines whether the rule is easy or nearly impossible to follow in the field. Electrical safety is not primarily a behavioral problem. It is a design problem with a behavioral layer on top.

Who this is written for

Field engineers, commissioning engineers and construction managers working on energized or recently energized systems.

EHS managers and safety directors who need to understand the technical basis for the rules they enforce.

Owners and developers who are deciding how much engineering to fund before a facility goes into service.

Anyone signing an energized electrical work permit — the signature carries the technical judgment.

The Twelve Rules, Read as an Engineer Would Read Them

Every rule above assumes an installation someone already built and a study someone already performed. The severity of the hazard a worker faces was fixed long before they arrived. Incident energy, available fault current, clearing time, isolation flexibility, approach requirements and step and touch potential are all determined at design, procurement and settings. The field crew inherits the number; they do not set it.

That is the argument for treating power system studies as a safety investment rather than a compliance expense. The table below maps the common hazards back to the engineering control that reduces them and the deliverable that documents it.

PV Reverse Current & Mismatch | Solar DC Design Guide

Thu, 27 Aug 2026 10:05:18 GMT

Aug 27, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

PV Array Voltage Mismatch and Reverse Current

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

Every utility-scale and commercial PV array is a collection of independent current sources forced to share a single voltage. That one architectural fact — parallel strings tied to a common DC bus and operated at a single maximum power point — is the origin of nearly every DC-side reliability problem the industry deals with: mismatch loss, hot spots, combiner fires, welded connectors, blown string fuses that do not clear, and inverter-level harvest deficits that show up in production reports as an unexplained one to four percent.

Mismatch and reverse current are frequently discussed as if they were the same phenomenon. They are not. Mismatch is an energy problem: strings that cannot reach their individual maximum power points give up yield. Reverse current is a safety problem: a string whose operating voltage falls below the DC bus voltage stops being a source and becomes a load, and the healthy strings in parallel with it push current backwards through it. The first costs money. The second starts fires.

The transition between the two is not gradual and it is not symmetric. Light soiling and partial shading almost never produce destructive reverse current — they reduce current, not voltage, and the string simply operates off-peak. What produces destructive reverse current is anything that collapses string voltage: a string built with fewer modules than its neighbors, a module with multiple bypass diodes conducting or shorted, a cracked-cell cluster that opens a substring, a ground fault on a floating array, or the single most damaging commissioning error in the industry — a string landed with reversed polarity.

This paper works through the physics, the arithmetic, the code and standards framework, and the practical design decisions: how to size a string fuse inside a constraint window that modern high-current modules have made uncomfortably narrow, when a fuse will and will not clear, when fuseless architecture is legitimate, and what instrumentation actually detects the failure before it becomes a claim. It closes with a commissioning protocol, an O&M trending regime, and a twenty-question FAQ.

The single most important idea in this paper

A string fuse in a PV array is back-fed. Its available fault current does not come from a utility source with thousands of amps behind it — it comes from the handful of other strings on the same combiner bus, and it is limited to roughly 1.25 times short-circuit current per contributing string.

If there are not enough parallel strings, the available backfeed current can land in the fuse's non-clearing region: high enough to run the element hot indefinitely, not high enough to open it. That is a combiner fire waiting for a hot afternoon, and it is a design decision, not a maintenance failure.

A photovoltaic module is a current source with a voltage ceiling. Short-circuit current scales almost linearly with irradiance and drifts only slightly with temperature. Open-circuit voltage is nearly independent of irradiance above a few hundred watts per square metre and falls with cell temperature at roughly 0.24 to 0.30 percent per degree Celsius depending on cell technology. Series-connecting modules into a string adds voltage; paralleling strings adds current.

When strings are paralleled on a combiner bus, they no longer have individual operating points. Kirchhoff makes the bus voltage common to all of them. The inverter's maximum power point tracker sweeps that single bus voltage looking for the peak of the aggregate power curve, and it finds a compromise: a voltage that is optimal for the population average and suboptimal for every individual string that deviates from it.

For a healthy array this is a small, well-behaved loss. Modules are binned, strings are the same length, and the spread in string maximum power voltage is a percent or two. The MPPT settles near the middle of the distribution and each string operates within a few tenths of a percent of its own peak. Standard energy models capture this as a module and string mismatch loss of roughly one to three percent, and it is the least interesting number in the loss table.

The behaviour changes character when a string deviates far enough that the common bus voltage lies outside its useful operating range. A string whose open-circuit voltage is below the bus voltage has no operating point on its own I-V curve at that voltage. It is driven into the fourth quadrant — positive current in the reverse direction, or equivalently negative power — and it dissipates energy instead of producing it.

2.1 The Fourth-Quadrant Condition Stated Precisely

Reverse current flows in a string when the DC bus voltage exceeds the string's open-circuit voltage at the prevailing cell temperature and irradiance. That is the condition. Everything else — shading, cracking, diode failure, wrong module count — matters only insofar as it drives a string toward that inequality.

This is why the distinction between current-limiting faults and voltage-collapsing faults is the central diagnostic question on the DC side. Soiling reduces Isc; the string still develops nearly full Voc and stays a source. A shorted bypass diode removes the voltage of one substring; on a twenty-eight-module string that is a two to four percent voltage deficit — a yield problem, not a reversal. But a string wired with twenty-four modules where its neighbours have twenty-eight is fourteen percent short, and at high cell temperature on a clear day that string can sit below bus voltage for most of the production window, every day, for years.

Diagnostic shorthand

Current-type faults (soiling, uniform shading, glass haze, encapsulant browning): string current falls, string voltage nearly intact, no reversal, energy loss only.

Voltage-type faults (missing modules, conducting or shorted bypass diodes, open substrings from cracked cells, ground faults, reversed polarity): string voltage collapses, reversal becomes possible, thermal and fire risk enters the picture.

The remote monitoring alarm that matters is not "low string current" — it is "string current sign reversal" and "string voltage deviation from combiner median."

1. Executive Summary

String fuse selection is over-constrained. Four independent requirements must be satisfied simultaneously, and on high-current modules in hot climates they are not always all satisfiable — which is itself a design finding that should be reported rather than papered over.

6.1 Constraint One — Do Not Nuisance-Operate

The National Electrical Code establishes maximum PV source circuit current as 125 percent of module short-circuit current, and then requires the overcurrent device to be rated at not less than 125 percent of that value. The compound result is the familiar 156 percent multiplier. A 13.9 A module therefore requires a fuse of at least 21.7 A, rounding up to the next standard size.

6.2 Constraint Two — Do Not Exceed Module Withstand

The fuse rating must not exceed the nameplate maximum series fuse rating. With a 25 A nameplate and a 21.7 A floor, the entire selectable window for the example module is a 25 A fuse. There is no engineering margin left in either direction.

6.3 Constraint Three — Derate for Enclosure Temperature

PV fuses are qualified at a reference ambient, but combiner enclosures in desert and high-plains sites routinely reach 60 to 75 degrees Celsius internally under full sun with the array producing. Manufacturers publish derating curves; a common figure is 70 to 80 percent of nominal rating at 60 degrees Celsius and lower still above that. A 25 A fuse in a 70 degree enclosure may only be good for 18 to 20 A continuous — below the 21.7 A the code requires.

That contradiction cannot be resolved by ignoring it. The available responses are a larger fuse (blocked by the module rating), a cooler enclosure (ventilated or shaded combiners, light-coloured enclosures, thermal modelling of the installed condition), a different fuse series with a better temperature characteristic, fewer strings per combiner to reduce internal heat load, or a fuseless architecture. Choosing among those is a real engineering decision with cost and reliability consequences on both sides.

6.4 Constraint Four — Be Able to Clear

The available backfeed current from Section 4 must be high enough, at the irradiance levels where faults actually occur, to drive the fuse into its clearing region within a time short enough to protect the compromised string. Working backwards from a target of roughly three times nominal fuse current for reliable operation and the 156 percent floor, the minimum string count per protected group falls in the range of five to six. Below that, a fuse is decoration.

Where Backfeed Actually Comes From, Why String Fuses Sometimes Do Not Clear, and How to Design, Commission, and Monitor a DC Collection System That Does Not Burn Itself Down

4. Quantifying Reverse Current

2. Why Parallel Strings Share a Voltage

Mismatch sources should be classified by whether they are designed in, built in, or developed over life, because the mitigation differs completely across those three categories.

3.1 Design-Origin Mismatch

Unequal string lengths. Terrain breaks, setback constraints, and racking table geometry tempt designers into partial strings. A short string on a shared MPPT is a permanent voltage deficit and a permanent reverse-current candidate.
Mixed module models, bins, or vintages on a shared MPPT. Repowering and warranty replacement are the usual culprits. Two modules from different production years can differ by several percent in Vmp.
Mixed orientation or tilt on a shared MPPT. East and west sub-arrays have different irradiance profiles and different cell temperatures, therefore different Vmp trajectories through the day.
Bifacial rear-irradiance non-uniformity. Edge rows, rows over disturbed or vegetated ground, rows adjacent to access roads, and rows near equipment pads see materially different rear-side gain. Mixing edge-row and interior-row strings on one MPPT builds mismatch into the plant.
Excessive strings per fuse group, or too few. Both directions create problems — the first exceeds module reverse-current withstand, the second leaves the fuse unable to clear.

3.2 Construction-Origin Mismatch

Reversed polarity strings. A string landed backwards presents a negative voltage to a positive bus. The reverse current is immediate and severe, limited only by the parallel strings and the string's own series resistance.
Cross-mated connectors. Mating a connector from one manufacturer to a visually similar connector from another produces a joint with poor contact pressure and a high-resistance interface that degrades under thermal cycling. This remains one of the leading root causes of DC-side fires.
Under-torqued or over-torqued terminations at fuse holders, bus bars, and lugs.
Modules damaged in handling — walked on, dropped on a corner, or torqued during racking — producing cracks that are electrically invisible at commissioning and open circuits after two winters of thermal cycling.
Missing or backwards module-level jumper in a string, producing a string one module short that nobody counts.

3.3 Life-Origin Mismatch

Cell cracking propagation. Microcracks from transport and installation grow under thermal and mechanical cycling until they isolate cell regions, forcing bypass diodes into conduction.
Bypass diode failure. Diodes fail short (removing substring voltage permanently and quietly) or open (removing hot-spot protection and setting up a far worse failure).
Potential-induced degradation. PID reduces shunt resistance, hitting fill factor first and voltage at low irradiance, and it is spatially non-uniform — worst at the string ends with the highest potential to ground.
Non-uniform soiling and vegetation encroachment. A row shaded at its base by unmanaged vegetation drives bypass diodes into conduction for hours a day.
Junction box and connector resistance growth from moisture ingress and thermal cycling.
Ground faults developing in cable insulation abraded at racking edges or damaged by rodents and mowing equipment.

The magnitude of reverse current into a compromised string is bounded by what the parallel strings can supply. In a combiner with N parallel strings, if one string is fully compromised, the remaining N minus one strings can each deliver up to their short-circuit current into it. Applying the standard 125 percent irradiance-enhancement factor that accounts for cloud-edge focusing and above-STC irradiance events:

Ireverse, max = (N − 1) × 1.25 × ISC

This is the number that must be compared against the module's reverse-current withstand. It is also, and this is the point routinely missed, the available fault current for the string fuse. There is no other source. A grid-tied transformerless inverter does not sustain reverse current into the array; its DC-link capacitance contributes a brief transient at most and its bleed circuits discharge the bus when it is not operating.

4.1 Worked Example

Take a modern n-type bifacial module: 585 W, Isc of 13.9 A at STC front-side, maximum series fuse rating of 25 A on the nameplate. Consider three combiner configurations.

Every module nameplate carries a maximum series fuse rating, sometimes labelled maximum overcurrent protective device rating. It is not an arbitrary number and it is not the same as the module's current rating. It is derived from the module's demonstrated ability to survive reverse current without ignition, established by the reverse-current overload test in the module safety standard — MST 26 in IEC 61730-2, carried through into the harmonised UL 61730 series that replaced UL 1703 in North America.

In that test the module is subjected to a reverse current of 1.35 times the declared maximum overcurrent rating for a defined duration while wrapped in tissue paper, and the acceptance criterion is that nothing ignites. It is a fire-safety qualification, not a performance rating. Passing it does not mean the module is undamaged; a module that has taken sustained reverse current should be treated as compromised and replaced regardless of whether it still produces power.

Two ratings, two different jobs

Maximum series fuse rating (typically 15 A to 30 A): the largest overcurrent device allowed upstream of the module. It caps fuse selection from above.

Short-circuit current Isc (typically 11 A to 19 A for current-generation large-format modules): drives the minimum fuse rating from below via the code multiplier.

As Isc has climbed with larger wafer formats while maximum series fuse ratings have not climbed proportionally, the gap between the floor and the ceiling has narrowed. On some modules the usable fuse window is now a single standard rating.

8. Code and Standards Framework

9.1 Cell-Level Hot Spots

When reverse current is forced through a string, it must pass through every cell in series. A cell that is cracked, shaded, or mismatched relative to its neighbours cannot carry that current at forward bias, so it goes into reverse bias. The reverse-bias voltage across that one cell can reach tens of volts before breakdown, and the product of that voltage and the string current is dissipated in a few square centimetres of silicon.

Localised temperatures well above 150 degrees Celsius are readily reached. The consequences progress from encapsulant browning and delamination, through solder joint reflow and interconnect ribbon failure, to backsheet scorching and, in the worst case, ignition. Bypass diodes are the designed defence — they clamp substring reverse voltage to roughly a volt — but a diode that has failed open removes that defence entirely, and a diode that has failed short removes substring voltage and pushes the string closer to the reversal condition in the first place.

9.2 Bypass Diode Thermal Failure

A conducting bypass diode carries full string current, and in a reverse-current event it carries the backfeed current from all parallel strings. Junction box thermal design assumes intermittent conduction under transient shading, not sustained conduction under a permanent voltage-collapse fault. Sustained conduction drives junction temperature up, forward voltage down, and current concentration higher, which is a positive-feedback path to junction failure. Once the diode fails open, the substring it was protecting is exposed.

9.3 Connector and Termination Failure

Reverse current does not stop at the module. It flows through every connector, splice, fuse holder, and lug in the string path. A joint with elevated contact resistance dissipates I-squared-R at a point with almost no thermal mass. Cross-mated connectors, connectors field-crimped without the manufacturer's tooling, and connectors mated with contamination on the pins are the usual sites. The failure signature — a melted connector shell with copper conductor visible and the racking scorched beneath — is one of the most common DC-side incident photographs in the industry.

9.4 DC Arcing

Direct current does not have a natural zero crossing. An arc established in a DC circuit at 1000 or 1500 V sustains itself as long as the source can supply current and the gap geometry permits. Series arcs (a degraded connection in the string path) and parallel arcs (insulation breakdown between conductors or to ground) both become plausible once thermal degradation has begun. Arc-fault detection is the mitigating control, but it is a detection layer over a design problem, not a substitute for correct protection sizing.

9.5 Harvest Loss and Inverter Behaviour

The energy consequence is often larger than operators assume, because it is not confined to the affected string. When a badly mismatched string is present on an MPPT, the tracker's aggregate power curve develops local maxima. A tracker that settles on a local rather than global maximum can leave several percent of the MPPT's output on the table across every string it serves, not just the compromised one. Global scan intervals, sweep behaviour, and the inverter's handling of multi-peak curves become material — and vary considerably between manufacturers and firmware versions.

3. Taxonomy of Mismatch

6. Fuse Selection: A Four-Sided Constraint Window

7. Fused and Fuseless Architecture

Both the North American and IEC frameworks recognise that string overcurrent protection is not always required. The logic is identical in both: if the worst-case reverse current the parallel strings can deliver is below what the module can withstand, there is nothing for a fuse to protect against.

The IEC array design standard expresses this as a direct comparison — protection may be omitted where the number of parallel strings minus one, multiplied by module short-circuit current, does not exceed the module's maximum overcurrent rating. The NEC reaches the same place through the overcurrent protection article, where the requirement attaches to circuits that could be subject to currents exceeding conductor ampacity from parallel sources, with the familiar allowance for arrangements with a small number of source circuits and no external sources.

In practice, with modules in the 13 to 19 A Isc range and 20 to 30 A maximum series fuse ratings, genuine fuseless operation is limited to two, occasionally three, parallel strings per protected group. That is a common architecture in residential and small commercial work and in string-inverter plants where each MPPT input takes two strings directly. It is not a utility-scale central-inverter combiner architecture.

7.1 The Awkward Middle

Between roughly three and five strings per group, neither architecture is comfortable. Reverse current exceeds module withstand, so fuseless is not defensible; but available backfeed current is too low for confident fuse clearing. Designs that land here should be restructured — either consolidate to a higher string count per fuse group or split to a genuinely fuseless two-string arrangement — rather than accepting a protection scheme that only works on paper.

7.2 One Pole or Two

Whether to fuse both polarities depends on the array's earthing architecture. In a functionally grounded system where the reference is established through the inverter, a single fuse per string in one polarity is the common approach. In a genuinely floating array, a single ground fault does not by itself produce current, but a second fault in the opposite polarity elsewhere in the array creates a loop that a single-pole fuse arrangement may not interrupt at both ends. The decision should follow from the inverter's earthing topology, the ground-fault detection scheme, and the applicable code edition, and it should be documented in the design basis rather than inherited from a combiner vendor's standard product.

The following are the provisions that a DC collection design review should actually be able to cite. Editions differ by jurisdiction and by the code cycle adopted locally; the design basis should name the specific edition in force.

9. What Actually Fails, and How

11. Detection and Measurement

10. Reading the Standard Diagram Correctly

The one-line and block-diagram treatments of this topic that circulate widely — including the array diagram that prompted this paper — communicate the shape of the problem well and get several important details wrong. Four clarifications are worth making explicitly, because the errors have design consequences.

Clarification 1 — The inverter is not the source of reverse current

Diagrams commonly draw an arrow from the inverter back into the array, labelled reverse flow. In a grid-tied transformerless PV inverter this is not the sustained mechanism. The inverter presents a controlled DC bus voltage; it does not push power back into the array. The source of sustained reverse current is the healthy parallel strings on the same bus, and the correct arrow runs laterally from string to string through the combiner bus, not backwards from the inverter. This matters because it determines where the protection goes and what the available fault current is.

Clarification 2 — Mismatch and reverse current are different regimes

Showing soiled or lightly shaded modules and destructive backfeed in the same picture implies a continuum that does not exist. Reverse current requires the string voltage to fall below the bus voltage. Most soiling and most partial shading do not achieve that; they cost energy and nothing else. Treating every mismatch alarm as a fire risk produces alarm fatigue, and alarm fatigue is why the one that mattered got closed without a truck roll.

Clarification 3 — A fuse symbol is not protection

Every string in every diagram has a fuse symbol on it, which implies the problem is solved. As Section 6 shows, whether that fuse can clear depends on the number of parallel strings, the prevailing irradiance, and the enclosure temperature. The symbol carries none of that. A protection review must show the arithmetic.

Clarification 4 — Measurement belongs at the combiner, and it must be signed

Reverse-current measurement drawn at the end of a string is conceptually right but operationally impractical. The instrumentation that exists in real plants is at the combiner: a current transducer per string on the positive or negative leg. What determines whether it is useful is whether it reports signed current. A sensor that reports magnitude only cannot distinguish a string producing 8 A from a string absorbing 8 A, and the second condition is the one that burns.

Specification language that changes outcomes

Require string-level current monitoring with signed output and explicit negative-current alarming, not magnitude-only measurement.

Require accuracy sufficient to resolve a 5 percent deviation from combiner median at irradiance above 600 W/m², which in practice means better than ±1 percent of full scale.

Require combiner internal temperature monitoring with alarming, so that fuse-holder heating is visible before it becomes an incident.

11.1 String Current Trending

The workhorse method. Compare each string's current to the median of its combiner group, filtered to irradiance above roughly 600 W/m² and to periods without transient shading. A persistent deviation beyond five percent over multiple days is a work order. A negative reading at any irradiance is an immediate dispatch. The filtering matters — unfiltered comparison at low irradiance generates enough false positives to make the alarm useless.

11.2 String Voltage and Open-Circuit Verification

Because reversal is a voltage phenomenon, string voltage is the more direct indicator, but it is rarely instrumented per string. Where it is not, periodic manual open-circuit voltage measurement per string, temperature-corrected using the module's temperature coefficient and compared to the combiner group, remains the fastest field screen for a voltage-type fault.

11.3 I-V Curve Tracing

The definitive diagnostic. A traced curve translated to standard test conditions and compared against the module datasheet distinguishes among fault classes by curve shape:

Reduced Isc with normal Voc and fill factor: uniform soiling or irradiance-type loss.
Reduced Voc with normal Isc: modules missing from the string, or shorted bypass diodes.
A distinct step or shoulder in the curve: one or more bypass diodes in conduction, indicating shading or an open substring.
Reduced fill factor with a shallow slope near Voc: series resistance growth — connectors, solder bonds, terminations.
Reduced fill factor with a steep slope near Isc: shunt resistance loss, characteristic of PID or cell damage.

Full-population curve tracing at commissioning is the highest-value quality step available on the DC side, and it is the one most often cut for schedule.

11.4 Infrared Thermography

Aerial infrared survey has made full-plant thermal inspection economical. Interpretation follows recognisable patterns: a single hot cell indicates a cell-level defect under reverse bias; a hot substring band indicates a conducting bypass diode; a uniformly hot module indicates the module is short-circuited or carrying reverse current; a hot junction box indicates a diode or termination problem. Temperature differentials above roughly 20 kelvin against neighbouring cells warrant intervention. Surveys should be flown at irradiance above 600 W/m², with the plant producing, and reported per the applicable thermography technical specification.

11.5 Electroluminescence Imaging

For root-cause work on modules already flagged, electroluminescence reveals cracks, inactive cell regions, and interconnect failures that are invisible in both visual and thermal inspection. It is a night-time or enclosure method and is used on samples, not populations.

11.6 Insulation Resistance and Ground-Fault Detection

Ground faults on a floating array are a voltage-collapse mechanism and a reverse-current path. Insulation resistance testing per the commissioning standard, plus a functioning ground-fault detection and interruption or residual-current monitoring scheme at the inverter, closes the loop. Ground-fault detectors that have been bypassed to keep a plant online — which happens more than the industry likes to acknowledge — remove the only automatic defence against this class of fault.

13. Commissioning Protocol

The table shows the real design tension. Too few strings and the fuse cannot clear. Too many and the module is exposed to reverse current an order of magnitude above its rating during the fuse's clearing time. The workable band for a fused combiner with contemporary high-current modules is typically six to twelve strings per protected group, and it should be verified arithmetically for the specific module and fuse, not assumed from a standard combiner catalogue.

4.2 The Irradiance Problem

Fault current in a PV array scales with irradiance. A fault that occurs at 1000 W/m² may clear in under a second. The same fault at 300 W/m² presents roughly thirty percent of the current, which for most PV fuse time-current curves is deep in the non-clearing region. The fuse element runs at elevated temperature, the fuse holder's spring contacts anneal, contact resistance rises, and the assembly enters a thermal runaway that ends in an arcing fault inside a sealed polymer enclosure.

This is the mechanism behind a meaningful share of combiner box fires, and it is why a design review that stops at "the fuse rating is below the module maximum series fuse rating" has not actually reviewed the protection.

12. Design Controls

The following controls, applied at design stage, eliminate most of the failure modes above at negligible cost. Applied after commissioning, the same items cost ten to a hundred times more.

Build every string on a shared MPPT to identical module count. Where terrain forces a short string, give it a dedicated MPPT input or dedicated inverter, or omit it.
Bin modules and keep a single bin within a combiner group. Record the binning in the as-built so that future replacements can respect it.
Never mix module models or production vintages on a shared MPPT, including during warranty replacement. Maintain an attic stock of the original model for the life of the plant, or plan replacements at whole-MPPT granularity.
Separate orientations and tilts onto separate MPPTs.
For bifacial plants, group edge rows separately from interior rows where the layout permits, and specify uniform ground treatment beneath the array.
Compute the four-sided fuse constraint window explicitly for the specific module and fuse series, at the actual expected combiner internal temperature, and document it. If the window is empty, change the architecture.
Set strings per protected group to satisfy both the module withstand limit and the fuse clearing requirement — typically six to twelve for current module technology.
Specify combiner enclosures with thermal performance appropriate to the site, not the catalogue default: ventilation or thermal management, light exterior finish, and where practical shading from the array itself.
Specify signed string-level current monitoring with negative-current alarming and combiner internal temperature monitoring.
Specify a single connector manufacturer and part number across the entire DC system, prohibit cross-mating in the specification, and require manufacturer tooling and installer certification for any field terminations.
Specify arc-fault detection where it is code-required and evaluate it where it is not, particularly on plants with long service lives and limited site attendance.
Perform DC arc-flash analysis for combiner and inverter DC compartment work, recognising that the AC incident-energy standard does not apply.

Notice and Disclaimer

This document is original technical content prepared by Keentel Engineering LLC for general professional information. It is not project-specific engineering advice and does not constitute a design, a design review, or a certification for any particular installation. Code citations are provided by article and subject; the edition adopted by the authority having jurisdiction governs, and provisions change between code cycles. Standards references are provided for orientation; the current published edition of each standard governs its own requirements.

Product ratings, module characteristics, and fuse performance data cited in examples are representative of current commercial equipment and are used for illustration only. Actual design must use the specific manufacturer data for the equipment selected, at the environmental conditions of the installed location.

Keentel Engineering LLC is an independent engineering consultancy. Reference to any standard, code, industry organisation, or equipment category in this document does not imply affiliation with, endorsement by, or sponsorship from any such organisation or manufacturer.

5. Module Withstand and the Maximum Series Fuse Rating

The commissioning sequence below is ordered so that the destructive errors are caught before the array is energised onto a common bus.

Verify polarity on every string at the combiner before landing conductors on the bus. This is non-negotiable and it is the single highest-value step in the sequence.
Measure open-circuit voltage on every string with the combiner bus open. Temperature-correct and confirm each string is within roughly two percent of the group mean. Any outlier is a module count or connection error and must be resolved before proceeding.
Measure short-circuit or operating current on every string and confirm consistency within the group.
Perform insulation resistance testing per the commissioning standard on every source circuit.
Trace I-V curves on the string population — full population where the contract permits, otherwise a statistically defensible sample with mandatory full-population tracing on any combiner showing an outlier.
Verify fuse ratings installed against the design calculation, physically, per combiner. Substitutions made in the field to keep a schedule are common and are exactly the failure this step exists to catch.
Verify termination torque on every fuse holder, lug, and bus connection, and record it.
Confirm the monitoring system reports signed string current and that negative-current alarms are configured and tested by injection or simulation.
Perform a baseline aerial infrared survey with the plant producing at above 600 W/m², and archive it as the reference against which all future surveys are compared.
Archive the commissioning dataset — string Voc, string current, I-V curves, insulation resistance, torque records, IR imagery — as the plant's electrical birth certificate. Its absence is the most common obstacle to a successful serial-defect warranty claim.

14. Operations, Trending, and Warranty

The operational regime that keeps this class of failure out of the incident log is unglamorous and consists of four habits.

First, trend string current against combiner median continuously, with irradiance filtering, and treat sign reversal as a dispatch condition rather than an alarm to be acknowledged. Second, repeat the aerial infrared survey annually and compare against the commissioning baseline rather than against an absolute threshold — the rate of appearance of new thermal anomalies is far more informative than the count. Third, sample I-V curves annually on a rotating basis so that the whole plant is traced over a three to five year cycle, which surfaces gradual degradation before it becomes mismatch. Fourth, treat any module that has carried reverse current as scrap, not as a candidate for return to service.

On the commercial side, this data has a second life. Serial defect claims against module and connector manufacturers succeed or fail on the quality of the baseline and the trend record. An owner who can demonstrate a defect population with commissioning I-V curves, sequential thermal surveys, and electroluminescence on failed samples is in a materially different negotiating position from an owner presenting photographs of burnt connectors.

15. How Keentel Supports This Work

16. Frequently Asked Questions

Keentel Engineering provides the electrical design and engineering analysis behind DC collection systems that behave predictably over a twenty-five to thirty-five year service life. On PV and PV-plus-storage projects our involvement typically covers:

DC collection system design and review — string sizing, combiner architecture, conductor sizing, and the explicit fuse constraint analysis described in Section 6, documented to a defensible design basis.
Protection coordination across the DC collection system, inverter, and MV collection, including DC arc-flash analysis for maintenance work planning.
Owner's engineer services — specification development, design review of EPC submittals, factory and site witness testing, and commissioning oversight including I-V curve and thermography programme specification.
Interconnection engineering and power system studies for the AC side of the same plants: point-of-interconnection studies, short-circuit and load flow, harmonic and flicker analysis, and EMT modelling where the interconnecting utility requires it.
NERC compliance support for registered generating facilities, and QA/QC review of third-party design packages.

The firm holds a Florida Certificate of Authorization and maintains offices in Tampa, Austin, Sacramento, and Baltimore, supporting projects across the interconnections.

PGRR144, PSS®E v36 & ERCOT Batch Zero Guide

Wed, 26 Aug 2026 12:40:53 GMT

Aug 26, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

PGRR144, PSS®E Version 36, and Batch Zero: A Complete Guide to the 2025 to 2026 Transition Reshaping ERCOT Load and Generation Modeling

By the Keentle Engineering Modeling Practice

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

If you are involved in interconnecting a data center, a battery storage project, a solar or wind farm, or any large industrial facility in the ERCOT footprint, the eighteen months between mid-2025 and late-2026 have delivered more meaningful change than any comparable stretch in the last decade. Three intersecting developments — Planning Guide Revision Request 144 (PGRR144), the mandatory transition from Siemens PSS®E version 35 to version 36, and the introduction of the Batch Zero framework for large loads under PGRR145 and NPRR1325 — have collectively rewritten what a compliant model submission looks like, who has to submit one, and how those submissions get processed.

Some of these changes have been hard deadlines. The June 1, 2026 date for PSS®E v36-compatible dynamic model submissions is one that has already passed. Some are process transformations, like the Batch Zero framework approved by the Public Utility Commission of Texas on June 18, 2026, which fundamentally changes how large-load interconnections are studied. And some are content transformations, like PGRR144's move away from composite load models toward EPRI User-Defined Models plus Composite Motor Load Dynamics representations for Large Electronic Loads.

For project developers, Interconnecting Entities, Resource Entities, and Large Load owners, the practical question is: what has actually changed, what do I have to do differently, and by when? This blog is Keentle Engineering's answer, written in the level of technical depth that project engineers and modeling consultants actually need to make decisions. We cover the regulatory context, the modeling mechanics, the OEM readiness realities, the interaction between the changes, and the schedule and cost implications. At the end, we address twelve of the questions we hear most often from clients navigating this landscape.

The blog is organized in six substantive sections followed by a comprehensive FAQ. Each section stands alone, so if you already understand the PGRR144 framework and just need the PSS®E v36 mechanics, jump to Section 3. If you have generation experience and are learning the load-side rules for the first time, Section 2 is where to start.

Section 1 — Setting the Stage: Why ERCOT Changed the Rules
Section 2 — PGRR144: Formal Dynamic Modeling for Large Loads
Section 3 — The PSS®E v35 to v36 Transition: Dates, Mechanics, OEM Readiness
Section 4 — Batch Zero: What It Is, How It Works, and Why It Matters
Section 5 — The Timeline: Every Date That Matters
Section 6 — Practical Implementation: What This Means for Your Project
Frequently Asked Questions — Twelve in-depth Q&As addressing the most common client concerns

Why This Blog Matters Right Now

Because these three initiatives overlap in time and interact in effect, a consolidated timeline is one of the most useful tools for developers trying to understand where they are in the process. The timeline below captures the dates that have already occurred and the dates that are upcoming as of the writing of this blog in August 2026.

5.1 Historical Timeline of the Key Events

Section 6: Practical Implementation — What This Means for Your Project

5.2 Forward-Looking Timeline

Section 2: PGRR144 — Formal Dynamic Modeling for Large Loads

PGRR144 formalizes dynamic model submission, validation, and review requirements for Large Loads. It is the first ERCOT rule to treat the load side with the same modeling rigor long applied to inverter-based generation, and as of the July 2026 ROS meeting the proposed language was approved and PGRR144 was progressing through the ERCOT stakeholder process toward final effectiveness.

2.1 What PGRR144 Actually Covers

The rule applies to Large Loads at or above 75 MW, with escalating requirements based on how much power-electronic content the load contains. Three broad categories emerge from the rule text and from ERCOT's Large Load Working Group discussions:

Section 3: The PSS®E v35 to v36 Transition — Dates, Mechanics, and OEM Readiness

This status snapshot will age quickly. Developers should confirm current OEM readiness with their equipment vendor directly rather than relying on point-in-time summaries. The key operational point is that ERCOT's flexibility on the June 1 deadline is calibrated to OEM readiness, not to developer preference. If your OEM has released a v36 DLL and you have not submitted, ERCOT expects you to submit. If your OEM has not released a v36 DLL, ERCOT is working through the gap with you — but that flexibility is provisional and requires documented engagement.

3.5 What You Have to Do Differently for v36 Submissions

The mechanics of a v36 submission are similar to a v35 submission with a small number of important differences:

Model files: v36 UDM DLL compiled as VINDP, plus the associated .raw or .sav case file, .dyr dynamics file, and Excel Dynamic Model Template.
File naming: Keentle Engineering recommends including 'v36' in the zip filename to distinguish v36 packages from any legacy v35 files that may still be in circulation. Our convention is (SITECODE)_DYNAMIC_v36_(YYYY-MM-DD).zip.
RIOO-RS subject line: ERCOT has requested that entities include 'v36' in the submission subject line to help prioritize reviews during the transition window.
MQT report: A Model Quality Test report must accompany the v36 submission per Planning Guide Section 6.2(5)(c). Simply recompiling a DLL and resubmitting without an MQT overlay is insufficient. If both v35 and v36 models are submitted (which may be required during the base-case transition), the MQT should overlay both versions on the same axes to confirm consistent response.
PSCAD models: The v36 transition is a PSS/E-specific change. PSCAD models are not affected, but reviewers will expect the combined PSS/E-PSCAD-TSAT overlay plot to use the v36 PSS/E model.

3.6 Grid-Forming Models and the v36 Timing

One important nuance in the v36 transition concerns grid-forming (GFM) inverter models used in Advanced Grid Support ESR (AGS-ESR) submissions. There are no grid-forming generic models available in PSS/E v35 — the v35 standard library predates GFM standardization. ERCOT has allowed temporary flexibility around v35 UDM use for AGS-ESR FIS and additional studies where a v35 UDM is the only available representation of the equipment. But the expectation is that as v36 GFM generic models mature, submissions will migrate to those generics or to v36 UDMs, with reasonable correspondence between v35 UDM response and later v36 generic model response.

For AGS-ESR developers, the practical guidance is to build the FIS-stage submission using whatever combination of v35 UDM and v36 v36-compatible UDM is available at the time, but to plan explicitly for a re-submission to v36 GFM generic or v36 UDM as the ecosystem matures. Keentle Engineering has structured multiple AGS-ESR engagements as two-phase modeling projects for exactly this reason.

Section 1: Setting the Stage — Why ERCOT Changed the Rules

Closing Thoughts

The eighteen-month window between mid-2025 and late-2026 has reshaped the ERCOT modeling landscape more than any comparable period in recent memory. PGRR144 has extended the dynamic modeling discipline that has long governed inverter-based generation to the Large Load side of the grid, filling a reliability gap that had produced real operational events. The PSS®E v36 transition has resolved a decade-long version-compatibility treadmill through the introduction of Version Independent DLLs. And Batch Zero has restructured the large-load interconnection process to handle the unprecedented queue growth that no per-project workflow could scale to.

For developers and operators navigating this landscape, the practical takeaway is that the level of modeling discipline required for successful ERCOT interconnection has stepped up significantly, and the coordination overhead across OEMs, ERCOT reviewers, and internal project teams has grown accordingly. The organizations that will succeed in this environment are the ones that treat modeling as an integrated project workstream from the beginning, not as a compliance afterthought at the end.

Keentle Engineering has built its practice around exactly this integrated approach. We work with Resource Entities, Interconnecting Entities, Large Load developers, OEMs, and asset managers to deliver dynamic models, MQT reports, verification reports, PSCAD Converter Model Validations, and Batch Zero submission packages that meet ERCOT's rigor on first review. Our team combines deep simulation expertise across PSS/E, TSAT, and PSCAD with practical knowledge of the ERCOT Planning Guide, Nodal Operating Guide, PGRR144, PGRR145, and the ongoing v36 transition.

If you are working on a project affected by any of the changes discussed in this blog, we invite you to reach out for a no-commitment scoping conversation. Whether you are managing an existing generation portfolio through the v36 migration or preparing a first-of-kind LEL submission for Batch Zero, our team is structured to support engagements ranging from single targeted consultations through multi-year modeling program management.

How to Read This BlogNew Title

The three-category structure matters because it determines the size of the modeling deliverable a developer is on the hook for. A conventional industrial load at 100 MW may fall into the LL category and require a comparatively straightforward composite-model-plus-MQT submission. A 250 MW hyperscale data center with GPU training racks will fall into the LCL category and require the full six-artifact submission described in Section 6.1 of this blog. The difference in modeling effort is roughly an order of magnitude.

2.2 The Deprecation of Composite Load Models for LELs

The single most consequential technical change in PGRR144 is that composite load models are explicitly deprecated for Large Electronic Loads. This is not a soft recommendation; ERCOT has been direct at the Large Load Working Group that composite representations will not be accepted as the primary dynamic model for LELs. The rationale reflects everything we know about power-electronic load behavior.

A composite load model — the standard ZIP plus induction motor representation — assumes that when voltage drops, the load's active power draw drops in a mathematically well-behaved way, and the induction motor component may stall or trip based on stall protection curves. For an inverter-fed UPS driving computational load, none of that behavior applies. The UPS rectifier isolates the DC bus from AC voltage disturbances until either the UPS ride-through envelope is exceeded (triggering an inverter output shutdown to protect the DC bus) or the AC voltage recovers within the ride-through window. The tripping is binary: full load one millisecond, zero load the next.

PGRR144 requires that this binary tripping behavior — the actual physics of what happens at a data center during a voltage disturbance — be represented in the model. That is what the EPRI User-Defined Model framework provides, and it is why PGRR144 recommends the EPRI UDM in combination with a Composite Motor Load Dynamics (CMLD) representation for the cooling motor portion of the facility.

THE CORE MODEL ARCHITECTURE FOR LELs

For LELs, the model is not a static motor equivalent with a slightly modified stall curve. It is a two-part representation combining an EPRI User-Defined Model for the power-electronic UPS-fed load with a Composite Motor Load Dynamics model for the induction motors that drive chillers, cooling towers, and computer-room air handlers. This is a genuine departure from prior practice, and it requires modeling skill that most load-side project teams have never needed to develop.

2.3 The Converter Model Validation (CMV) Exercise

PGRR144 introduces a new artifact called the Converter Model Validation report, applicable specifically to LELs. It is a hardware-benchmark exercise, analogous in spirit to the Unit Model Validation report long required for inverter-based generation. The purpose is to demonstrate that the submitted PSCAD model accurately reflects the physical converter behavior under voltage disturbances and in the sub-synchronous frequency range where computational-load switching can drive real reliability issues.

Practically, the CMV requires:

Bench-test measurement of the UPS converter response to a defined set of voltage disturbances, typically obtained from the UPS OEM.
Small-signal frequency sweep measurement over the 5 to 55 Hz range, which is the sub-synchronous band where computational-load switching interactions are known to occur.
PSCAD simulation of the same disturbances and frequency sweeps.
Side-by-side comparison of the measured versus simulated response, with documented explanations for any residual differences.
A summary attestation from a qualified engineer confirming the PSCAD model represents the physical converter within acceptable tolerances.

The CMV is the point in the PGRR144 submission workflow where OEM cooperation is most critical. UPS vendors have historically not been asked to provide dynamic model documentation of this depth, and most UPS OEMs are still building out the internal capability to support CMV requests. Developers should not wait until the modeling phase to raise this with their UPS vendor — the CMV should be a contract requirement negotiated at UPS purchase, not an afterthought.

2.4 The Voltage Ride Through (VRT) Dimension and NOGRR282

PGRR144 does not stand alone. It coordinates with Nodal Operating Guide Revision Request 282 (NOGRR282), which establishes voltage ride-through obligations for Large Loads. Where NOGRR282 applies, the PSS®E and PSCAD models submitted under PGRR144 must explicitly demonstrate ride-through capability across the NOGRR282 envelope, and the test results are pass/fail: a failure means the load cannot interconnect on the terms proposed.

For loads not subject to NOGRR282 compliance, voltage disturbance testing under PGRR144 is still performed — but the results are informational rather than pass/fail. This distinction matters. At the April 23, 2026 LLWG meeting, ERCOT explicitly clarified that voltage disturbance testing under PGRR144 for non-NOGRR282 loads is intended only to validate dynamic model performance under disturbance conditions, not to impose new compliance requirements. The point is to build ERCOT's understanding of aggregated load behavior, not to disqualify individual projects.

Even in the informational case, Keentle Engineering recommends the full VRT test catalog. Operational reality will inevitably subject the facility to voltage disturbances, and the informational test results become the baseline against which future firmware or configuration changes are compared. A well-documented pre-operation VRT baseline saves substantial trouble the first time a real disturbance produces an unexpected load response.

2.5 The Large Computational Load (LCL) Power Variation Limit

PGRR144 introduces a specific limit on active power variation for Large Computational Loads: 10 megawatts peak-to-peak measured over a rolling five-second interval. At the June 19, 2026 LLWG meeting, ERCOT presented proposed evaluation language focusing on oscillatory frequency components between 0.1 Hz and 55 Hz, capturing both inter-area and sub-synchronous oscillations.

The limit exists because computational workloads — particularly AI training workloads that run tens of thousands of GPUs in synchronized training cycles — can produce coordinated power draw swings of a magnitude and frequency that can interact adversely with the transmission system's inertia and damping characteristics. A single 200 MW GPU cluster starting or stopping a training batch is not a grid problem. Fifty synchronized 200 MW clusters cycling at a resonant frequency of the surrounding transmission network is a grid problem, and it is exactly the sort of emerging behavior that ERCOT is trying to get ahead of.

For developers, the practical implication is that the workload management system for the facility must respect the 10 MW / 5-second limit, and the model must demonstrate that respect. Keentle Engineering has developed workflow-simulation harnesses that verify this compliance during the MQT phase, so that any workload management defects are caught during modeling rather than during operation.

LCL COMPLIANCE IS A FACILITY-LEVEL OBLIGATION

The 10 MW / 5-second LCL power variation limit is enforced at the facility level, not the rack level or the individual training cluster level. If your data center has multiple training clusters that can synchronize their workload starts and stops, the aggregate synchronized behavior is what matters. Workload orchestration systems that treat training clusters independently, without an aggregate rate-of-change limit, will not demonstrate LCL compliance no matter how well-behaved each cluster is in isolation.

2.6 The Load Information Form (LIF)

Alongside the technical models, PGRR144 requires a fully populated Load Information Form. The LIF is the load-side analog of the generation-side data submissions that Resource Entities have long provided. It documents:

Facility electrical characteristics: total load, breakdown of power-electronic vs. motor content, on-site generation if any.
Tables A, B, and C: protection settings, ride-through behavior parameters, and interconnection point electrical data.
Survey Questions 39 through 50: UPS operating modes, backup power switching sequence, on-site generation dispatch philosophy during grid events, demand-side management participation, and computational-load transfer capability to off-ERCOT infrastructure.

The LIF is more than a bureaucratic checklist. ERCOT uses the LIF responses to build assumptions about how the facility will behave operationally, and those assumptions feed forward into both the dynamic modeling and the transmission planning studies. LIFs that are hand-waved through — with 'to be determined' entries or placeholder values — produce dynamic models that carry those uncertainties forward, and produce reviewer questions that must be resolved before the submission can be accepted.

Keentle Engineering's LIF workflow starts with a facility characterization workshop that produces documented answers to the survey questions before any modeling work begins. This front-loads the coordination overhead with the UPS vendor, cooling equipment vendor, and workload management team, so that modeling proceeds on solid assumptions rather than on placeholders that must be revisited.

The August 2026 classification notices are the near-term milestone that developers should be watching. Classification determines which projects are formally included in Batch Zero and what their category is within the batch. This has direct implications for financing conversations, offtake negotiations, and equipment procurement timelines. Developers who submitted qualifying requests before the batch cutoff should be preparing internally for classification notice receipt and the follow-on modeling coordination that classification triggers.

4.3 How Batch Studies Actually Work

The mechanical difference between a batch study and a project-by-project study is worth understanding in detail, because it changes the assumptions every developer should be making about their project's schedule and cost allocation.

In a project-by-project study, each interconnection request was studied against a base system model that reflected the status quo plus previously-committed transmission upgrades. The study identified the additional upgrades required to accommodate the specific project, and those upgrades were assigned to the project as its cost responsibility. If a second, nearby project submitted an interconnection request during the study period, the interaction between the two projects could not be evaluated without a restudy — and restudies were common enough that project schedules became unpredictable.

In a batch study, all qualifying projects within the batch are studied simultaneously against a base system that reflects the status quo. The study identifies the transmission upgrades required to accommodate the entire batch — not project by project — and those upgrades are allocated across the batch participants using a defined methodology. Individual projects do not receive individual upgrade assignments; they receive a share of the batch-wide upgrade plan.

THE DEVELOPER MENTAL MODEL HAS TO CHANGE

This is a fundamental change in how developers should think about their project. You no longer control your interconnection timeline as a solo actor. Your project's classification, transmission upgrade allocation, and operational date are functions of the batch as a whole. This has substantial implications for how you negotiate with lenders, offtakers, and equipment vendors — none of whom are accustomed to interconnection schedules being tied to a batch cadence.

4.4 The Interaction Between Batch Zero and PGRR144

Batch Zero and PGRR144 do not exist in isolation. Every Large Load in Batch Zero must satisfy PGRR144's dynamic modeling requirements to be studied in the batch. If the model submission is incomplete or non-compliant, the project cannot be evaluated in the batch study — and there is no mechanism to defer that project to Batch 1 without restarting the interconnection request. Modeling compliance is a prerequisite for batch participation, not something that can be addressed after the batch study is complete.

At the July 2026 ERCOT TAC meeting, ERCOT clarified that the PGRR144 process will include Model Quality Testing (MQT) to validate submitted Large Load models and assess compliance with applicable NOGRR282 requirements as part of the batch study workflow. Actual operational performance verification will be performed once PMU data becomes available at the operational facility. This creates a two-stage validation chain: pre-operational validation through MQT and CMV as part of the batch study, and post-operational validation through PMU data comparison after energization.

For developers, the immediate implication is that PGRR144 compliance is time-critical. Batch Zero classification is expected in August 2026, and the final transmission plan is expected in Fall 2027. That gives developers roughly twelve to fifteen months to complete their PGRR144 model submissions in a form that can be evaluated in the batch study. Facilities that treat modeling as a downstream activity — something to be worked out once the project is 'real' — will find themselves outside the batch study timeline and losing their place in the queue.

4.5 The Queue Diversity Insight

One of the most important observations from the LLWG meetings in 2026 is that the interconnection queue and the actual coincident load are very different quantities. At the April 23, 2026 LLWG meeting, ERCOT noted that approximately 9 GW of large load was approved to energize but that the observed non-simultaneous monthly peak consumption was approximately 4 GW, with a simultaneous peak of approximately 3.7 GW in March 2026. In other words, the actual coincident large-load draw was less than half of the approved-to-energize quantity, reflecting diversity in usage patterns across facilities.

This matters for developers because it informs how ERCOT thinks about batch study assumptions. The system does not need to be planned for every large load to be at 100 percent utilization simultaneously — that scenario is not physically realistic. But it also means that ERCOT is scrutinizing load profiles, workload diversity, and coincidence assumptions during the batch study. A developer whose load profile assumes 100 percent utilization 24/7 will face different treatment than one whose profile reflects realistic diurnal and workload-driven variation.

The Load Information Form is the primary place where this profile information is documented, which is one more reason why the LIF cannot be treated as a paperwork exercise. The survey responses about workload management, demand-side management participation, and computational-load transfer capability directly feed the coincidence assumptions ERCOT uses in the batch study.

4.6 What Happens After Batch Zero

Applications for Batch 1 are expected to open in Summer 2027. The precise cadence of subsequent batches has not been formally set, but the general expectation is that batches will occur on an annual or semi-annual basis, cycling through submission windows, study periods, transmission planning, and final approval on a repeatable schedule. Developers whose projects miss Batch Zero can plan for Batch 1, but the timing of study completion and transmission plan finalization will follow the batch schedule, not the project's individual schedule.

The batch cadence has second-order implications for OEM procurement and equipment lead times. Data center inverters, UPS units, and computational hardware often have twelve to eighteen month lead times, and the interaction between order-to-delivery timing and the batch study timing can be tight. Keentle Engineering has begun helping clients with integrated schedule planning that reconciles the batch study calendar with equipment procurement calendars — a coordination problem that did not really exist in the pre-batch era.

Before diving into what PGRR144, the v36 transition, and Batch Zero require, it is worth understanding why these changes happened when they did. Regulatory frameworks do not update themselves. Each of these three initiatives was driven by specific reliability concerns and specific operational realities that ERCOT could no longer address using its existing tools.

1.1 The Load Growth That Forced Everyone's Hand

ERCOT has grown accustomed to being one of the fastest-growing grids in North America, but the load additions that have appeared in the interconnection queue since 2023 are unprecedented. At the March 13, 2026 Large Load Working Group meeting, ERCOT reported that 137 new large-load interconnection requests totaling approximately 140,000 MW had been submitted and were not yet reflected in current queue charts. When those requests are included, the total large-load queue moves from around 238,000 MW to nearly 380,000 MW. To put that in context, ERCOT's all-time system peak has never exceeded approximately 85,000 MW. The queue represents multiple times the current system capacity.

Not all of that will build. ERCOT's own data indicates that the majority of interconnection requests expect to be operational by 2030, but historical attrition rates in interconnection queues are substantial. Even so, the volume is real enough — and concentrated enough in specific technology categories — that the pre-2025 project-by-project interconnection process was buckling under the load. ERCOT's original interconnection process, designed when new large-load connections were episodic industrial expansions, could not scale to hundreds of near-simultaneous data center and computational load requests.

1.2 The Reliability Concerns That Drove PGRR144

The volume of new load is one problem. The nature of that load is another. The interconnection queue is dominated by facilities whose electrical behavior is fundamentally different from the industrial and commercial loads ERCOT has traditionally served:

Hyperscale data centers: hundreds of megawatts of power-electronic-fed IT load behind uninterruptible power supplies, with cooling loads that ramp on timescales of seconds and IT loads that can shift among data centers on timescales of minutes.
AI training clusters: even higher power density than traditional data centers, with training workloads that can swing tens of megawatts within seconds as GPU jobs start and stop.
Cryptocurrency mining facilities: power-electronic loads with the ability to reduce demand on operator command, which is useful for demand response but adds control-system dependencies that must be modeled.
Electrolyzers for hydrogen production: DC-coupled loads with converter-driven behavior similar in some respects to inverter-based generation.

The traditional dynamic modeling approach for loads used composite models — ZIP-load approximations, static motor equivalents, and simplified induction motor models. These worked adequately when loads were largely resistive and motor-driven. They do not work for facilities where the dominant behavior is determined by power-electronic firmware.

ERCOT has observed real reliability events tied to this modeling gap. Multiple system disturbances since 2024 have involved large blocks of computational load tripping in response to voltage disturbances that composite load models would not have predicted. The interim Voltage Ride Through (VRT) assessment that ERCOT Operations Stability Analysis presented at the May 21, 2026 Large Load Working Group meeting identified four major large-load trip groups capable of causing tripping above 3,200 MW during severe disturbances. That is a system-shaping quantity of load that could disappear during a fault, and it was not visible in any planning study.

WHY PGRR144 EXISTS

PGRR144 is not a bureaucratic elaboration of existing rules. It is a direct response to observed reliability events. When ERCOT identified four large-load trip groups totaling more than 3,200 MW of at-risk demand, the case for formalizing dynamic modeling for the load side became overwhelming. The rule filled a modeling gap that had grown into a reliability gap.

1.3 Why the v36 Transition Happened in Parallel

The PSS®E v35 to v36 transition is not tied directly to PGRR144, but its timing is not coincidental. ERCOT relies on Siemens PTI PSS®E as the primary planning tool for the Dynamics Working Group base cases. Historically, ERCOT has stayed one full version behind the current Siemens release to avoid the disruption of forcing OEMs and consultants to recompile User-Defined Models against every new version.

Version 36, released by Siemens in 2024, introduced a feature that changed that calculus: Version Independent (VINDP) dynamic linked libraries. VINDP DLLs compiled against v36 can be used in all future versions of PSS®E without recompilation. That resolves one of the most persistent operational headaches in ERCOT's modeling ecosystem — the perennial coordination exercise of getting OEMs to release new DLLs every time PSS®E advances a version. With VINDP, the OEM compiles once against v36 and the same DLL works forever.

From ERCOT's perspective, the sooner the ecosystem transitions to v36, the sooner the version-compatibility treadmill ends. The May 2, 2025 market notice W-A050225-01 set the June 1, 2026 deadline for v36-compatible submissions, and while OEM readiness has been uneven, the direction is unambiguous. Every model built or refreshed from mid-2026 forward should be a v36 VINDP DLL.

1.4 Why Batch Zero Was the Logical Next Step

The volume challenge in Section 1.1 and the modeling challenge in Section 1.2 converge on a procedural problem: how does ERCOT actually study 140,000 MW of new interconnection requests? The traditional project-by-project approach, in which each interconnection was studied against a static system model and the transmission upgrades required were assigned to that specific project, produced two failure modes when applied at scale.

First, it generated a growing backlog. Studies took months, and the queue grew faster than studies could clear. Second, it triggered repeated restudies. A large-load interconnection studied in isolation would identify certain transmission upgrades. Six months later, a second interconnection nearby would trigger a restudy that assigned different upgrades to different projects. Neither project could commit to a cost or schedule until the interaction between them was resolved, and every additional interconnection request compounded the problem.

The Batch Zero framework, approved by PUCT on June 18, 2026 under PGRR145 and NPRR1325, addresses both failure modes by grouping qualifying loads (75 MW or greater) into a single system-wide study. Instead of studying each project against a fixed system, all qualifying projects are studied together against a jointly-planned system. The transmission upgrades needed to serve the batch as a whole are identified, and the cost of those upgrades is allocated across the batch participants.

Batch Zero is the first batch under this framework. ERCOT expects to notify Batch Zero applicants of their project classification in August 2026, and a final transmission plan covering the entire batch of projects across Texas is expected to be published in Fall 2027. Applications for Batch 1 are expected to open in Summer 2027. The batch cadence is expected to become the ongoing rhythm of large-load interconnection in ERCOT going forward.

SCHEDULE IMPLICATION FOR DEVELOPERS

The batch approach means that a large-load developer no longer controls their own study timeline in the way they once did. Batch Zero participants receive their classification and their allocated transmission upgrades on the batch's schedule, not on individual project schedules. This has substantial implications for financing, offtake negotiation, and equipment procurement. Developers used to project-by-project timelines should adjust their expectations accordingly.

The PSS®E v36 transition is the most concrete of the three changes covered in this blog. It has a specific market notice, a specific deadline, and a specific technical mechanism that resolves a long-standing pain point in the ERCOT modeling ecosystem. It is also the change that has produced the most confusion, because OEM readiness has been uneven and the deadline has already passed.

3.1 The Market Notice and the Deadline

On May 2, 2025, ERCOT issued market notice W-A050225-01: 'Requirement to submit dynamic models compatible with PSS®E version 36 prior to June 1, 2026.' The notice was unambiguous in its language. Interconnecting Entities and Resource Entities were required to submit dynamic models fully compatible with PSS®E version 36 prior to June 1, 2026, so that ERCOT could complete its transition of Dynamics Working Group base cases to v36.

The deadline has passed. As of the June 2026 IBRWG and RIWG meetings, ERCOT confirmed that entities with outstanding v36 submissions should coordinate directly with ERCOT to close remaining gaps. The rule was firm in principle, but ERCOT has recognized the practical dependency on OEM DLL availability and is working through gaps case-by-case rather than blocking projects wholesale.

THE DEADLINE HAS PASSED — HERE'S WHAT THAT MEANS

The June 1, 2026 deadline was a milestone, not a guillotine. ERCOT is closing gaps case-by-case for entities whose OEMs are still working on v36 DLL delivery. But this flexibility is not a substitute for action — if your project has outstanding v36 obligations, engage ERCOT proactively and document the OEM release schedule you are depending on.

3.2 What Version 36 Actually Delivers: Version Independent DLLs

The reason ERCOT prioritized the transition to v36 — over any of the intermediate versions Siemens has released — is a single feature: Version Independent (VINDP) dynamic linked libraries. Understanding VINDP is essential to understanding why the transition matters and why the effort of a one-time re-compile is worth it.

Historically, when Siemens released a new major version of PSS®E, existing UDMs stopped working. The OEM had to recompile the DLL against the new version's development kit, redistribute it to modeling consultants and asset owners, and coordinate with each ISO or utility that used the model. This produced a recurring version-compatibility cycle that consumed OEM engineering time and created windows during which certain models were simply not available on certain platforms.

Version 36 changes this by introducing a formal API layer between the UDM code and the PSS®E internal data structures. UDMs compiled against v36 access PSS®E bus voltage, model CON parameters, ICON values, and other internal data through a documented set of Python-based pssdm APIs — the COMON4.INS common block that older UDMs used is no longer accessible from v36 onwards. In exchange for this abstraction layer, VINDP DLLs compiled against v36 will run in all future versions of PSS®E without recompilation. The version-compatibility treadmill ends.

3.3 Converting Existing UDMs to VINDP

Siemens provides a Python-based UDM converter as part of the Environment Manager (EM) version 10.0 and higher. The converter automates most of the mechanical transformation from COMON4.INS-based UDM source code to the VINDP API form. For most OEMs, running the converter is a relatively straightforward exercise; the trickier work is in the edge cases where UDM code depended on specific data structures that do not map cleanly to the API.

Once the source code is in VINDP form, the OEM compiles it against v36 to produce the new DLL. The resulting DLL can be used in v36 and all future versions of PSS®E without further recompilation. Keentle Engineering has coordinated dozens of these conversions on behalf of clients and has developed diagnostic procedures for the recurring failure modes: undocumented common-block accesses, model initialization code that assumed specific memory layout, and protection models that used raw pointer arithmetic instead of PSS®E's abstraction layer.

3.4 OEM Readiness as of Mid-2026

The transition is only useful to the extent that OEMs have released v36-compatible DLLs. ERCOT has publicly tracked OEM readiness at working group meetings, and as of the June 2026 IBRWG and RIWG meetings the status was:

Section 4: Batch Zero — What It Is, How It Works, and Why It Matters

Batch Zero is the third leg of the 2025 to 2026 transition. Where PGRR144 changes what a Large Load has to submit and the v36 transition changes what tool the submission runs on, Batch Zero changes how ERCOT processes the submission and produces the transmission plan around it. It is a structural reform of the large-load interconnection process, and it addresses the scaling problem that the pre-2025 process could no longer manage.

4.1 The Approval and the Framework

The Batch Zero framework was formally approved by the Public Utility Commission of Texas on June 18, 2026, through two coordinated ERCOT revision requests: NPRR1325 (which modifies the Nodal Protocols to define the batch process) and PGRR145 (which aligns the Planning Guide to support batch studies). PGRR145 should not be confused with PGRR144 — the former is the batch framework, the latter is the dynamic modeling framework. The two rules were developed in parallel and are complementary: PGRR144 defines what each large load must submit, and PGRR145 defines how ERCOT will study the resulting submissions in coordinated batches.

The batch framework applies to loads at or above 75 MW. Loads below that threshold continue to interconnect through the traditional process. Above the threshold, qualifying loads are grouped into a single system-wide batch study rather than being studied one at a time. Batch Zero is the first batch under this framework and covers loads that submitted qualifying interconnection requests by the batch cutoff date.

4.2 The Batch Zero Timeline

Section 5: The Timeline — Every Date That Matters

Regulatory frameworks are only useful to the extent that project teams can translate them into concrete workstreams. This section addresses the practical implementation of PGRR144 and the v36 transition for the two main developer types this blog is written for: Large Load developers navigating PGRR144 for the first time, and Resource Entities managing the v36 transition on an existing generation portfolio.

6.1 The Six-Artifact PGRR144 Submission for a Large Electronic Load

A PGRR144 submission for a Large Electronic Load is not a single document. It is a coordinated package of six artifacts that must be internally consistent and cross-referenced. In Keentle Engineering's practice, the six artifacts are:

PSS®E v36 dynamic model — built using EPRI UDM v4 for the power-electronic UPS/IT load portion, combined with a Composite Motor Load Dynamics (CMLD) representation for the cooling motor load. Compiled as a Version Independent (VINDP) DLL. Includes the full model package: .raw or .sav case file, .dyr dynamics file, .dll for the UDM, Excel Dynamic Model Template, and Python initialization scripts for switched shunts and transformer taps if any are present.
PSCAD electromagnetic transient model — detailed converter representation embedded in the ERCOT PSCAD Template. Not built from PSCAD master library standard blocks; instead built from UPS OEM source-code-based converter models where available, or from a documented equivalent representation validated through the Converter Model Validation exercise. Includes the completed PSCAD Guideline Checksheet.
TSAT model — required where the PSS®E submission uses a UDM. Includes bus-number and equipment-name variants with 100 percent identical results across formats, plus .raw, .dyr, .dll, .tudm files and associated .mon, .dat, and .swi files.
Model Quality Test report — combined overlay plot of PSS/E, PSCAD, and TSAT responses on the same axes for each required test scenario. Test scenarios include flat-start initialization, voltage disturbance testing per NOGRR282 envelope where applicable, sub-synchronous response testing over 5-55 Hz, and — for LCLs — power variation testing against the 10 MW peak-to-peak per 5-second limit.
Converter Model Validation (CMV) report — hardware-benchmark exercise comparing the PSCAD model against actual UPS converter measurements. Requires OEM cooperation for bench-test data. Documents any residual differences between measured and simulated response with engineering explanations.
Load Information Form (LIF) — fully populated with Tables A, B, and C for facility electrical characteristics, protection settings, and interconnection point data, plus Survey Questions 39 through 50 addressing UPS operating modes, on-site generation dispatch philosophy, demand-side management participation, and computational-load transfer capability.

Missing any one of the six artifacts triggers a rejection or an incomplete-submission finding. The order of assembly matters as well: the LIF should be finalized first because it drives assumptions used in the models, the CMV depends on OEM bench-test data that has long lead times, and the MQT overlay depends on all three model platforms being consistent with each other.

6.2 The v36 Migration for an Existing Generation Portfolio

For a Resource Entity managing a fleet of existing generation projects, the v36 migration is a fundamentally different exercise. The models exist; the question is how to update them efficiently. Keentle Engineering's fleet migration workflow proceeds in five stages:

Portfolio inventory — catalog every dynamic model currently on file with ERCOT for every site in the portfolio. Identify which are UDMs (requiring OEM recompilation), which are standard library models (which automatically transition to v36), and which are legacy models that may need updating for other reasons.
OEM coordination — for every UDM in the inventory, identify the OEM and current v36 release status. For OEMs that have released v36 VINDP DLLs, request the DLL and any updated documentation. For OEMs still working on v36 releases, obtain a documented release schedule and share it with ERCOT to preserve deadline flexibility.
Recompilation and testing — receive the v36 VINDP DLL from each OEM, integrate it into the site model, and run the DMVIEW MQT catalog to confirm consistent response with the pre-migration v35 model. Any material response differences require OEM engagement to resolve before the submission proceeds.
MQT overlay preparation — build combined MQT reports overlaying v35 and v36 PSS/E responses (plus PSCAD and TSAT for IBRs) on the same axes for each required test scenario. Document any acceptable differences and explain their source.
RIOO-RS submission — assemble the submission package with (SITECODE)_DYNAMIC_v36_(YYYY-MM-DD).zip and (SITECODE)_PSCAD_(YYYY-MM-DD).zip naming, include 'v36' in the submission subject line, and upload as an Attachment-Only Change Request. Coordinate with the TSP on the change notification.

Portfolio-wide v36 migrations typically complete in six to twelve weeks depending on OEM responsiveness. The bottleneck is almost always OEM coordination, not the modeling work itself. Fleet operators that have not begun this exercise should start immediately — the June 1, 2026 deadline has passed and ERCOT's case-by-case flexibility is not indefinite.

6.3 Schedule and Cost Realities

Both PGRR144 and the v36 transition involve real time and cost. Directional ranges from Keentle Engineering's experience across dozens of engagements:

These ranges assume reasonable OEM cooperation and no unexpected model defects. Real-world engagements can compress or extend based on OEM responsiveness, the maturity of the client's existing documentation, and the complexity of the facility architecture. The single largest cost driver, particularly for first-of-kind LEL projects, is the coordination overhead of aligning the ERCOT reviewer, the UPS OEM, the cooling equipment vendor, and the developer's workload management team on consistent modeling assumptions.

6.4 Common First-Time Developer Mistakes

Because the load-side developer community is largely new to ERCOT dynamic modeling, several failure patterns recur. Keentle Engineering has observed each of these in the field:

Treating the LIF as paperwork. The LIF drives dynamic model assumptions and batch study coincidence factors. Placeholder responses produce placeholder-quality studies.
Not negotiating CMV support at UPS purchase. Retrofitting a Converter Model Validation obligation onto an existing UPS supply agreement is difficult and expensive; building it into the original purchase is straightforward.
Assuming composite models will suffice. Composite load models are explicitly deprecated for LELs under PGRR144. Attempting to submit one triggers a rejection and a full re-do.
Underestimating the OEM coordination timeline. UPS vendors and cooling equipment vendors have historically not been asked to support ERCOT-style modeling. Their internal capability is uneven and their response times are longer than IBR OEMs.
Sequencing modeling after equipment procurement. The modeling process reveals design-basis constraints that can inform equipment specification. Doing modeling in parallel with equipment specification produces better outcomes than doing it after equipment is ordered.
Missing the Batch Zero preparation window. Batch Zero classification arrives in August 2026; modeling for the batch study needs to be underway before that. Waiting for classification to begin modeling loses months of runway.

Frequently Asked Questions

Two dates in this timeline deserve particular attention. The August 2026 Batch Zero classification notice is the near-term operational trigger for many developers — until classification is received, projects cannot fully plan their modeling submission timelines. And the Fall 2027 transmission plan publication is the point at which Batch Zero developers will know the actual transmission upgrade allocation for their projects, which drives cost, timeline, and financing conversations.

5.3 Where You Are on This Timeline Determines What You Do Next

The single most useful diagnostic question for a developer is: where does my project sit on this timeline? The answer determines what actions are most valuable right now.

PFE Compliance for BESS Projects

Tue, 25 Aug 2026 11:37:55 GMT

Aug 25, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Navigating Prohibited Foreign Entity Compliance for Energy Storage Projects

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

Battery energy storage came out of the One Big Beautiful Bill Act (OBBBA, Pub. L. 119-21, enacted July 4, 2025) in better shape than wind and solar. Storage kept its long-runway Section 48E investment t ax credit while wind and solar were pushed onto a hard beginning-of-construction cliff. But storage did not escape the n ew prohibited foreign entity (PFE) regime — and in several respects storage carries the heaviest burden of any technology under it.

Three facts define the 2026–2027 planning environment for storage developers:

Storage faces the highest material assistance thresholds of any generation or storage technology. An energy storage technology (EST) beginning construction in 2026 must demonstrate that at least 55% of its manufactured-product direct costs are not attributable to prohibited foreign entities. Generation facilities need only 40%. The storage threshold climbs to 75% for construction starting in 2030 or later.
The battery supply chain is the most PFE-concentrated supply chain in the industry. Lithium iron phosphate cell manufacturing, cathode and anode active material production, separator and electrolyte supply, and midstream mineral refining are all heavily concentrated in a covered nation. Several of the largest global cell and integrated-system suppliers are named by statute as specified foreign entities.
Storage cannot hedge into the production tax credit. Energy storage technology is eligible for Section 48E only. Generation projects worried about the Section 48E ten-year recapture exposure for prohibited payments can elect Section 45Y instead. Storage has no such escape hatch — the ITC recapture rule applies, and there is no alternative credit to fall back on.

IRS Notice 2026-15 , released February 12, 2026, gave the market its first workable framework for calculating the material assistance cost ratio (MACR). It is, on balance, favorable — particularly for distributed fleets and for developers who can obtain supplier certifications. But it deliberately deferred the harder questions: what makes an entity a specified foreign entity, how ownership is traced, and what “effective control” means in a long-term service agreement or a software license. Treasury has committed to proposed regulations and to PFE-specific safe harbor tables by December 31, 2026.

This brief explains the statutory architecture, walks the MACR calculation as an engineer and a cost estimator would actually build it, translates the rules into procurement and contracting practice, and identifies the documentation package a project needs to survive diligence, a credit transfer, and an eventual examination.

The OBBBA restructured the technology-neutral credits along two axes: when the credit is available , and who and what may be inside the project .

On the first axis, storage did well. Wind and solar facilities must begin construction by July 4, 2026, or be placed in service by December 31, 2027. Energy storage technology under Section 48E was not subjected to that acceleration; storage remains on the longer phase-down curve, with the credit stepping down for projects beginning construction in 2034 and later and terminating thereafter.

On the second axis, storage did poorly — not because Congress wrote harsher rules for storage in every respect, but because the statute's percentage thresholds are highest for storage while the underlying supply chain is least diversified. That is the central tension of PFE compliance for battery energy storage systems (BESS).

Consider what a grid-scale BESS actually is, from a cost perspective. In a typical utility-scale AC-block or DC-block procurement, the battery modules and racks — cells, cell-to-pack hardware, module-level electronics — dominate the manufactured-product cost stack. The enclosure, thermal management system, fire detection and suppression, battery management system, DC combiners and disconnects, power conversion system, medium-voltage transformer, MV switchgear, auxiliary transformer, controls and the energy management system, AC and DC cabling, and the balance-of-plant steel each carry meaningful but individually smaller shares.

Now overlay the material assistance test. It does not ask whether the project contains PFE content. It asks w hat fraction of direct cost is not PFE-a ttributable — and whether that fraction clears the year's threshold. A project can contain Chinese-manufactured content and still qualify. What it cannot do is let that content dominate the cost stack. When the single largest line item in your bill of materials is the item most likely to be PFE-produced, the arithmetic gets difficult quickly.

PFE compliance for storage is not a tax-department exercise bolted onto an otherwise normal project. It is a procurement and engineering design constraint that has to be embedded at the specification stage, before the supply agreement is signed and before construction starts.

Executive Summary

Notice 2026-15 (issued February 12, 2026) provides interim guidance and previews forthcoming proposed regulations. Taxpayers may rely on it f or any Section 45Y or 48E qualified facility or EST beginning construction after December 31, 2025, until 60 days after publication of the forthcoming PFE safe harbor tables. Comments were due March 30, 2 026.

5.1 The three-step Clean Electricity MACR

Step 1 — Identify

Determine the manufactured products (MPs) and manufactured product components (MPCs) incorporated into the EST. The regime requires looking two levels up the supply chain: the major equipment assemblies, and the components that go directly into them.

Step 2 — Cost

Determine the direct costs for each MP and MPC. Critically, the definition depends on whether the taxpayer produced or acquired the item:

If the taxpay er produces the MP, direct costs include direct material and labor costs under Treas. Reg. § 1.263A-1(e)(2)(i)(A) and (B).
If the taxpayer acquires the MP, direct costs are simply the taxpayer's acquisition cost for the MP and its incorporated MPCs.

For a typical storage developer buying integrated AC or DC blocks, this means acquisition cost is the starting point — but the developer still has to disaggregate what sits inside that acquisition cost to identify the MPCs and their PFE status.

Step 3 — Subtract

Reduce total direct costs by the costs attributable to PFE-produced MPs and MPCs, and divide by total direct costs.

A favorable and important nuance: if an MP is PFE-produced but contains MPCs that are not PFE-produced, only the PFE-produced MPCs count against the MACR. The taxpayer retains credit for non-PFE content inside a PFE-produced assembly. This is not an all-or-nothing rule at the assembly level, and it materially changes procurement strategy — a non-PFE cell inside a PFE-assembled module is not wasted.

5.2 The three interim safe harbors

None is mandatory. They can, in appropriate cases, be combined.

Identification Safe Harbor

Permits the taxpayer to use the domestic content safe harbor tables in existing IRS guidance (Notice 2025-08 sections 5.05, 5.06, 6.02 and 7.02; Notice 2024-41 § 3.02 for hydropower and pumped storage; Notice 2023-38 § 3.04 for offshore wind) as an exclusive list of MPs and MPCs. Items not on the list are disregarded. This converts an open-ended supply chain investigation into a bounded, standardized component list — an enormous practical benefit. It is available only for listed technologies.

Cost Percentage Safe Harbor

Available where the Identification Safe Harbor applies. Substitutes the Assigned Cost Percentages from the domestic content tables for fact-specific direct cost determinations:

MACR = (Total Percentage − Total PFE Percentage) ÷ Total Percentage

where Total Percentage is the sum of Assigned Cost Percentages for all listed MPs and MPCs incorporated, and Total PFE Percentage is the sum for those that are PFE-produced.

Two features of this safe harbor deserve emphasis:

Steel and iron components listed on the tables are disregarded. You cannot improve your MACR under this method by buying domestic racking, foundations or structural steel. That is a real trap for developers accustomed to the domestic content framework, where steel and iron are a separate and highly useful test.
The “Production” assigned cost percentage is counted as PFE-produced if the MP itself is PFE-produced. This differs from the domestic content rules and, in mixed-source situations, generally shifts the arithmetic in the taxpayer's favor relative to a strict all-or-nothing approach.

Certification Safe Harbor

Permits reliance on supplier certifications to establish (i) whether an item is PFE-produced or PFE-sourced, and (ii) the direct costs attributable to PFE-produced items — provided the taxpayer does not know or have reason to know the certification is inaccurate.

This is the workhorse provision for storage. Two clarifications make it usable:

Onl y the manufacturer's PFE status is relevant, not the supplier's or distributor's. A U.S. distributor selling a PFE-manufactured product does not launder the product's status; conversely, a supplier's own corporate structure is not the question.
The manufacturer's PFE status is generally tested in the year the MP or MPC is purchased. And if the taxpayer had no prior knowledge, a valid certification protects the taxpayer even if the manufacturer is later determined to be a PFE. T his is meaningful protection against mid-project list additions.

5.3 Allocation methods

The 10% de minimis allocation. For qualified facilities and ESTs, taxpayers may assign MPs or MPCs of the same type across facilities or ESTs placed in service in the same taxable year without project-specific trackin g, so long as the assigned items represent less than 10% of total direct costs of the facility or EST. For portfolio developers moving common components — controls, auxiliary equipment, cabling, protection devices — across sites, this removes a very large tracking burden.

The sub-1 MW BESS allocation method. For ESTs with capacity less than 1 MW(AC) placed in service in the same taxable year, taxpayers may compute an average direct cost for each MP/MPC type and a PFE Production Percentage based on the proportion of those items that were PFE-produced across a specified period, rather than tracking each unit individually.

Specified periods must be at least one whole calendar day, be gin on the first day of the taxable year for the first period, be contiguous if shorter than a full taxable year, collectively cover every day of the taxable year, and not exceed the taxable year. Within those constraints there is substantial flexibility — which is exactly what a developer transitioning suppliers mid-year needs.

For commercial and industrial and community-scale storage fleets, this method is transformative. It also relieves the pathological result where an individual small unit fails the threshold because of how components happened to be blended into that specific enclosure.

5.4 Substantiation

A taxpayer using any safe harbor must attach a statement to the applicable form on which the credit is claimed, filed with the annual return for the first taxable year in which the credit is claimed, identifying the specific safe harbor used, the specific safe harbor tables applied (if applicable), and how the safe harbor was applied — for example, whether it was used to identify MPs and MPCs, to determine direct costs, or both.

This is not a checkbox. It is a positive disclosure that frames the examination, and it needs to be supported by a calculation package that can be reconstructed years later.

5.5 Domestic content is not MACR

Because Notice 2026-15 borrows the domestic content tables, there is a strong temptation to treat the domestic content analysis and the MACR analysis as one workstream. They are not the same, and treating them as interchangeable will produce wrong answers.

A Korean-manufactured cell, a Japanese separator, a German inverter, a Canadian transformer — none of these help your domestic content position, and all of them help your MACR. That asymmetry should drive procurement.

Section 48E, the Material Assistance Cost Ratio, and What BESS Developers Must Get Right in 2026–2027

3. Effective Dates: What Applies When

1. Why This Matters More for Storage Than for Anything Else

OBBBA added two new definitional paragraphs to the Internal Revenue Code: Section 7701(a)(51), defining prohibited foreign entity, and Section 7701(a)(52), defining material assistance from a prohibited foreign entity. Together with the credit-specific operative provisions in Sections 45Y, 48E and 45X, they create three independent tests. Failing any one of them is fatal to the credit. Passing two does not help you if you fail the third.

Test 1 — The entity test: Is the taxpayer itself a PFE?

A prohibited foreign entity is either a specified foreign entity (SFE) or a foreign-influenced entity (FIE).

Specified foreign entity

Specified foreign entity captures five categories:

A foreign e ntity of concern as described in specified subparagraphs of Section 9901(8) of the FY2021 National Defense Authorization Act;
An entity identified as a Chinese military company operating in the United States under Section 1260H of the FY2021 NDAA;
An entity on certain lists required under the Uyghur Forced Labor Prevention Act (Pub. L. 117-78);
An entity specified under Section 154(b) of the FY2024 NDAA — which names, by statute, Contemporary Amperex Technology Co. Ltd. (CATL), BYD Co. Ltd., Envision Energy Ltd., EVE Energy Co. Ltd., Gotion High-Tech Co. Ltd., Hithium Energy Storage Technology Co. Ltd., and any successor to those entities;
A foreign-controlled entity — the government of a covered nation (including sub-national government), a citizen/national/resident of a covered nation, an entity organized under the laws of or with its principal place of business in a cov ered nation, or an entity controlled by any of the foregoing. “Covered nation” means China, Russia, Iran and North Korea. Control generally means more than 50% by vote, value, profits or capital interest, or beneficial ownership, with Section 318 attribution rules applying (excluding downward attribution under 318(a)(3)).

For storage developers, the fourth bullet is the one to internalize. Six of the most significant names in global battery and integrated-BESS supply are specified foreign entities by act of Congress, not by administrative listing. There is no fact-based argument to be made about them.

Foreign-influenced entity

Foreign-influenced entity captures entities under SFE influence. An entity is an FIE if, during the taxable year:

An SFE has direct authority to appoint a covered officer — a board member, or an executive-level officer such as president, CEO, COO, CFO, general counsel or senior vice president;
A single SFE owns at least 25% of the entity;
One or more SFEs own at least 40% in the aggregate;
At least 15% of the entity's debt is held in the aggregate by one or more SFEs; or
The entity made a payment to an SFE in the preceding taxable year under a contract, agreement or other arrangement that entitles the SFE (or a related entity) to exercise effective control over the qualified facility, energy storage technology, or production of eligible components.

Status is generally tested as of the last day of the taxable year. Pu blicly traded entities receive modified treatment on the ownership and debt prongs — the ownership prongs are keyed to beneficial ownership reported under Rule 13d-3 of the Securities Exchange Act of 1934 (or an equivalent foreign rule), an d the ownership prongs generally do not apply to a publicly traded entity or its 80%-controlled subsidiaries. Publicly traded entities remain fully subject to the effective control payment rules and the material assistance rules. The public-company relief is narrower than it first appears.

Test 2 — The payment test: Has the taxpayer made a prohibited payment to an SFE?

This is the “effective control” prong, and for storage it is the most under-appreciated risk in the entire regime.

Under Section 7701(a)(51)(D)(i i), effective control means contractual arrangements giving a counterparty specific authority over key aspects of production or operation — authority not otherwise captured by the ownership, officer or debt tests. The statute describes categories of contracts and terms treated as conferring effective control, including certain intellectual property licensing arrangements, subject to a bona fide purchase-or-sale-of-IP exception. Notice 2026-15 confirmed that the IP effective control provisions are read as a disjunctive list, and that they apply to contracts, agreements or arrangements entered into or modified on or after July 4, 2025.

Any payment under such an arrangement — regardless of size — can trigger the rule.

Why this matters disproportionately for storage: a grid-scale BESS is not a set-and-forget asset. It typically comes with a long-term service agreement (LTSA), a capacity maintenance or augmentation agreement, a warranty administered by the OEM, an OEM-supplied or OEM-licensed battery management and energy management software stack, remote monitoring and diagnostics, firmware update rights, and in many cases OEM involvement in di spatch envelope enforcement or operating-mode approval as a warranty condition. Every one of those creates an ongoing payment stream to the equipment supplier across the entire operating life — precisely the period during which the recapture rule is live.

A solar project pays its module supplier once. A storage project pays its integrator every year for fifteen years. The contractual architecture of that relationship is where PFE compliance for storage is won or lost.

Test 3 — The material assistance test: Does the project itself contain too much PFE content?

This is the MACR test, and it is the subject of Section 4 below

Two practical consequences follow.

Firs t, a project that began construction on or before December 31, 2025 is not subject to the material assistance test at all. The entity tests still apply to the taxpayer, but the MACR analysis does not. This is why the safe-harbor procurement activity of late 2025 was so intense, and why establishing and documenting a 2025 construction start remains one of the most valuable positions a storage portfolio can hold.

Second, the delayed effective date on the payment prong and the recapture rule is a planning window, not a reprieve. Contracts signed today will still be in force in 2028. An LTSA executed in 2026 with a supplier that is or becom es an SFE, containing terms that confer effective control, will start producing recapture exposure the moment the rule switches on. The drafting has to happen now.

Getting the timing right is essential, because different prongs switch on at different moments.

4.1 The formula

For a qualified facility or energy storage technology, the Clean Electricity MACR is:

MACR = (A − B) ÷ A

where A = total direct costs of all manufactured products and manufactured product components incorporated into the qualified facility or EST at the time of completion, and B = the portion of those direct costs attributable to items mined, produced or manufactured by a prohibited foreign entity.

The MACR measures the non-PFE share. Higher is better. The project qualifies if the MACR meets or exceeds the applicable threshold. This is the single most common point of confusion in the market — the threshold is a floor o n clean content, not a ceiling on PFE content. A 2026 storage project with a MACR of 55.0% passes; a MACR of 54.9% fails, and the entire Section 48E credit is lost.

Note also the cliff structure. There is no partial credit, no proportional reduction, no cure period. The test is binary and it is applied to the whole energy storage technology.

The year of beginning of construction fixes the threshold for the life of the project. A project that establishes a 2026 construction start is tested at 55% even if it is placed in service in 2029. This makes beginning of construction the single highest-leverage variable in storage PFE planning — see Section 6.

4.3 For reference — Section 45X eligible component thresholds

Relevant to storage developers because it shapes what your domestic and allied suppliers can economically build:

7. Section 48E-Specific Exposure: Recapture and Transferability

Disclaimer

8.1 Specification-stage decisions

Set the MACR target above the threshold with margin. A 2026 project should target well above 55%, not 55.5%. Cost estimates move, scope changes, change orders happen, and the test is applied to actual direct costs at completion.
Model the cost stack before selecting the technology. Run the MACR under both the direct cost method and the Cost Percentage Safe Harbor. They give different answers, and the better answer sometimes points to a different equipment configuration. Method selection is a substantive planning decision, not a formality.
Recognize where substitution is realistic. Power conversion systems, medium-voltage transformers, MV switchgear, protection relays, SCADA and EMS hardware, HVAC, fire detection and suppression, and cabling all have credible non-PFE supply. Cells are the hard problem. Non-covered-nation cell capacity — Korean, Japanese, U.S., European, and increasingly Southeast Asian and Indian — is the pivot point of the entire analysis.
Do not assume steel helps. Under the Cost Percentage Safe Harbor it is disregarded.
Disaggregate the integrator. A si ngle line item for “DC block, delivered” is not an answer to Step 1. The integrator has to open the box: cells, modules, racks, BMS, thermal system, fire system, enclosure, DC electrical, each with manufacturer identity and cost.

8.2 Supply agreement provisions

PFE representations and warranties, given at signing, at each delivery, and at placed-in-service — covering the manufacturer's status under each of the five SFE categories and under the FIE ownership, debt, officer and effective-control prongs.
Certification obligations in a form that satisfies the Certification Safe Harbor: identifying manufacturer, place of manufacture, PFE status, and direct cost attribution at the MP and MPC level, in a format the taxpayer can attach to its calculation package.
Cost disclosure and disaggregation rights sufficient to build the MACR, with confidentiality protection for the supplier's commercially sensitive pricing.
Audit and inspection rights, including factory access and the right to witness FAT — which the owner's engineer should be exercising in any event.
Change-in-status provisions: notice obligations if the manufacturer or any upstream owner is added to a listed category; a right to substitute equipment; and price protection for the substitution.
Delivery timing and title transfer terms aligned to the BOC strategy and to the taxable year in which the manufacturer's status is tested.
Indemnification for credit loss attrib utable to a breach, sized realistically against the credit at risk rather than capped at contract value.

8.3 Structuring to avoid effective control

The goal is a commercial relationship that gives the supplier no authority over key aspects of the project's operation. Areas to scrutinize:

Operating envelope and dispatch. Warranty terms that condition coverage on operating within defined limits are ordinary and appropriate. Terms that give the OEM approval rights over dispatch decisions, market participation strategy, or state-of-charge management move toward control.
Software and IP licensing. Perpetual, non-conditional licenses look different from ongoing royalty arrangements with continuing supplier authority. The statute's bona fide purchase-or-sale-of-IP exception is the relevant path, and its scope is one of the open questions Treasury has been asked to clarify.
O&M and augmentation. Who decides when to augment, what to install, and on what schedule? Owner-controlled augmentation with supplier supply obligations is a different posture than supplier-controlled augmentation.
Data and remote access. Remote monitoring is normal. Remote control authority — the ability to change setpoints, curtail, or place the asset in a mode — is a control indicator.
Personnel and key-decision rights. Any right of the supplier to designate operating personnel, or to approve the O&M contractor, warrants review.

None of this means an owner must forgo OEM service. It means the allocation of authority in the LTSA has to be drafted deliberately, with the effective control test in view, and with tax counsel reviewing the operative provisions rather than only the tax representations.

2. The Statutory Architecture: Three Independent Tests

5. Notice 2026-15: How the Calculation Actually Works

6. Beginning of Construction: The Highest-Leverage Variable

Because the MACR threshold is fixed by the calendar year in which construction begins, and because the material assistance rules do not apply at all to projects that began construction on or before December 31, 2025, beginning of construction (BOC) is where storage developers should concentrate effort.

6.1 The tests available to storage

Storage was not covered by Notice 2025-42, which restricted the 5% safe harbor for wind and for solar facilities above 1.5 MW(AC). Beginning of construction for energy storage continues to be governed by the pre-IRA framework — Notices 2013-29, 2018-59 and 2022-61 — under which both methods remain available:

Physical Work Test. Physical work of a significant nature, on-site or off-site, with off-site work performed under a binding written contract and the resulting property incorporated into the project. For storage, off-site physical work on custom-manufactured items — a project-specific main power transformer, custom MV switchgear lineups, purpose-built enclosures, custom collection cable — is the usual route. The work must be of a significant nature, judged qualitatively, not by cost.
5% Safe Harbor. Paying or incurring 5% or more of total project cost. For accrual-method taxpayers, the 3½-month rule and the economic performance rules govern when costs are treated as incurred, which is why safe-harbor equipment procurement is typically structured around delivery timing and title transfer.

Notice 2026-15 resolved a question that had been open through 2025. For purposes of the prohibited foreign entity rules under Section 7701(a)(51) and the material assistance rules under Section 7701(a)(52), the notice confirms that Notice 2025-42 does not apply. Instead, consistent with the statute, the principles of Notices 2013-29 and 2018-59 — together with subsequently issued guidance clarifying, modifying or updating them, as in effect on January 1, 2025 — govern beginning of construction for PFE and material assistance purposes.

This is a significant clarification with two consequences. First, it removes the concern that FEOC-specific BOC guidance might retroactively narrow the tests. Second, and less obviously, it means the BOC test for PFE purposes is frozen to a January 1, 2025 baseline. Guidance issued after that date does not move the PFE beginning-of-construction line, which insulates the material assistance analysis from the ongoing turbulence in wind and solar BOC guidance.

Both methods are subject to the continuity requirement, with the four-year continuity safe harbor as the practical anchor. A taxpayer using the Physical Work Test satisfies continuity through a continuous program of construction; a taxpayer using the 5% safe harbor through continuous efforts, which expressly allows permits and payments to count.

6.2 The Notice 2025-42 litigation — relevant context, indirect effect

On June 6, 2026, the U.S. District Court for the District of Columbia vacated Notice 2025-42 in its entirety and remanded to Treasury and the IRS, holding the notice arbitrary and capricious under the Administrative Procedure Act — principally for failing to give a reasoned explanation for treating wind and large solar differently from other technologies under a technology-neutral credit, and for failing to account for more than a decade of reliance on the dual-test framework. The vacatur was granted nationwide. An appeal and a possible stay were anticipated, and the court itself acknowledged appellate timing would extend past the July 4, 2026 wind/solar deadline.

For storage, the direct effect is nil — Notice 2025-42 never appli ed. The indirect effects are worth tracking:

The court's reasoning cuts against technology-by-technology divergence in BOC rules. That is helpful to storage's continued access to the 5% safe harbor, but it also means Treasury could respond on remand by restricting the 5% safe harbor across the board with a better-reasoned record.
The vacatur does not disturb the PFE beginning-of-construction position. Notice 2026-15 confirmed that Notice 2025-42 does not apply for Section 7701(a)(51) and (a)(52) purposes, and that the Notice 2013-29 / 2018-59 principles as in effect January 1, 2025 govern. The litigation is about the wind and solar credit deadline, not about material assistance.
What the vacatur does not change: material assistance and FEOC restrictions, domestic content, prevailing wage and apprenticeship, post-BOC continuity, transferability and elective pay, and general credit eligibility all remain exactly as they were.

Practical posture: run both tracks. Anchor on physical work of a significant nature with an independently sufficient evidentiary record, and treat safe-harbor procurement as reinforcing optionality rather than as the sole foundation.

6.3 What a defensible BOC file contains

Executed binding written contracts, with the binding-nature analysis documented (damages provisions, limitation clauses, enforceability under governing law).
Manufacturer work orders, production schedules, shop travelers, serialized progress photographs, and factory acceptance test scheduling for off-site custom work.
For on-site work: geotechnical completion, foundation excavation and rebar placement, trenching for the MV collection system, grounding grid installation, with dated field reports and daily logs.
For 5% safe harbor: invoices, proof of payment, title transfer documentation, delivery records, storage arrangements, and a cost model showing the 5% computation against total expected project cost, with headroom against cost escalation.
A continuity narrative maintained contemporaneously — not reconstructed later.
Project definition and single-project aggregation analysis. Which units constitute the project matters for both BOC and for the MACR test, since the MACR is applied at the energy storage technology level.

7.1 The ten-year recapture rule

Under the OBBBA amendments to Section 50(a), a taxpayer that claims the Section 48E credit and subsequently makes an applicable payment to an SFE under a contract, agreement or arrangement conferring effective control, at any point during the ten years after the property is placed in service, faces 100% recapture of the previously claimed credit — the credit is retroactively reduced to zero in the year the payment occurs, with carryback and carryover adjustments. The rule applies to taxpayers allowed a Section 48E credit for taxable years beginning after July 4, 2027.

Three observations for storage:

First, ten years is longer than most storage LTSAs are short. The standard product in the market is a fifteen- to twenty-year service and augmentation arrangement. The recapture window sits entirely inside the service term.

Second, storage has no PTC alternative. Generation developers concerned about ITC recapture can pivot to Section 45Y, which has no analogous payment-based recapture. Energy storage technology is not eligible for Section 45Y. For storage, the Section 48E ITC is the only credit, and the recapture rule comes with it.

Third, SFE status is not static. The Section 154(b) list is statutory, but the Section 1260H and UFLPA lists are administratively maintained and can be expanded. A supplier that is not an SFE when the supply agreement is signed may become one during the recapture window. The contract has to contemplate that.

7.2 Credit transfer under Section 6418

A taxpayer transferring Section 48E credits under Section 6418 may not transfer to a specified foreign entity. The statutory restriction names SFEs specifically, not FIEs.

In practice, the transfer market has moved well beyond the statutory minimum. Buyers and their advisors are demanding:

Seller representations as to non-PFE status of the seller and relevant upstream owners, tested as of the last day of the relevant taxable year;
The MACR calculation package, with supporting supplier certifications;
Representations that no effective control payment has been made and that no contract in force confers effective control;
Covenants restricting the entry into or modification of contracts that could confer effective control during the recapture window;
Indemnification for recapture and disallowance, often backstopped by tax insurance;
Ongoing reporting during the recapture period.

PFE compliance quality is now priced. A project with a clean, well-documented MACR package and a well-drafted supply and service agreement transacts at a tighter discount than one with a defensible but thinly documented position. The engineering documentation is a financial asset.

8. Procurement and Contracting Playbook

10. Open Questions Heading Into 2027

9. The Documentation Package

PFE compliance produces a deliverable, and the deliverable outlives the project's construction phase by a decade. A complete package includes:

BOM decomposition to the MP and MPC level, mapped to the applicable safe harbor tables where the Identification Safe Harbor is used.
Cost model reconciling total direct costs, distinguishing produced from acquired items, and tying to the accounting records and eventual ITC basis.
MACR calculation memorandum — method selected, safe harbors elected, calculations shown, sensitivity to change orders documented.
Supplier certification register — every certification, with dates, scope, the manufacturer's identity, and the knowledge/reason-to-know analysis supporting reliance.
Entity status file — the taxpayer's own ownership, debt and officer analysis as of the last day of each relevant taxable year, including upstream tracing to the extent practicable.
Contract review memorandum — effective control analysis for every agreement with a supplier, licensor or service provider, refreshed on each amendment.
Beginning of construction file — as described in Section 6.3.
Placed-in-service reconciliation — confirming the as-built configuration matches the modeled configuration, with change orders reflected in the final MACR.
The return statement — prepared and reconciled to the calculation package.
A monitoring protocol for the recapture window — periodic re-screening of suppliers against listed categories, and a payment review process before any new or amended agreement with an equipment supplier.

Entity status

Notice 2026-15 was deliberately partial. The following remain unresolved, and each carries planning implications:

PFE and SFE status determinations. Treasury deferred the definitional rules to proposed regulations. Ownership attribution through complex structures, the treatment of minority foreign investors, and whether a certification safe harbor will be extended to entity status (as it was to material assistance) are all open.
Whether a “no-look-through” approach is adopted. Commentators have noted signals in other Treasury guidance suggesting a domestic-blocker-style simplification could be considered. Adoption would substantially reduce diligence burden in project finance structures.
“Effective control” and the definition of a licensing agreement. The most consequential open item for storage, given the LTSA and software licensing exposure described above.
Anti-circumvention and anti-abuse rules under Section 7701(a)(51)(D) and (K), which Treasury has said are coming.
The PFE safe harbor tables, statutorily due December 31, 2026. Whether Treasury issues PFE-specific tables or continues to use the domestic content tables for both purposes is unknown. Either way, positions taken under the interim framework will need to be re-tested against the published tables. Projects beginning construction more than 60 days before the tables are published may continue to rely on the domestic content tables for the Clean Electricity MACR.
Critical minerals thresholds, due by December 31, 2027.
Qualified interconnection property. How interconnection costs included in ITC basis interact with the MACR is not fully settled — a live question for storage projects with substantial POI substation, MV/HV transformer and transmission line scope.
Incremental capacity and additions. How the material assistance rules apply to capacity additions and augmentation at an existing facility is not fully addressed.
The BOC litigation trajectory and whether Treasury reissues BOC guidance on remand with broader technology scope.

Frequently Asked Questions

4. The Material Assistance Cost Ratio: Thresholds and Formula

The critical minerals schedule is worth noting: the threshold is zero through 2029, then ramps. Treasury is directed to issue critical-minerals threshold guidance by December 31, 2027.

11. How Keentel Engineering Supports PFE Compliance

PFE compliance sits at the intersection of tax law, procurement, and electrical engineering. Tax counsel owns the legal conclusions. What the project needs alongside counsel is an engineering organization that can decompose the plant into its manufactured products and components, price them defensibly, evaluate substitution options against real electrical performance requirements, and produce documentation that survives diligence.

Keentel Engineering supports storage developers, sponsors and investors with:

Bill of materials decomposition and MP/MPC mapping for AC-block, DC-block and modular BESS architectures, aligned to the applicable safe harbor tables.
MACR modeling and sensitivity analysis under both the direct cost method and the Cost Percentage Safe Harbor, including scenario testing across supplier configurations and construction-start years.
Equipment substitution engineering — evaluating non-PFE alternatives for power conversion systems, MV transformers, switchgear, protection and control, and auxiliary systems against short-circuit duty, protection coordination, grounding, arc-flash, harmonic and ride-through requirements, so that a compliance-driven substitution does not create an electrical performance or code compliance problem.
Owner's engineer scope : specification development with PFE provisions embedded, technical bid evaluation, factory audits and FAT witnessing, and supplier documentation review.
Beginning of construction technical documentation — physical work evidence packages, off-site custom manufacturing verification, and construction continuity records.
Interconnection and POI engineering for storage projects, including system impact study support, protection and control design, and substation design from 30/60/90 to IFC.
Diligence support for tax equity investors, credit transferees and lenders — independent technical review of the MACR package, BOM substantiation and BOC evidence.

Our four service lines — Electrical Design, Power System Studies , NERC Compliance, and QA/QC — cover the technical scope from POI to the battery terminal.

Scope and applicability

Primary Authority

MACR mechanics

Safe harbors

Beginning of construction

Recapture, transferability and contracting

Practical planning

The analysis in this brief rests on the following primary sources. Readers should confirm the current status of each, as the regime is actively developing.

Statute

One, Big, Beautiful Bill Act, Pub. L. No. 119-21, 139 Stat. 72 (July 4, 2025) — §§ 70512, 70513 and related provisions.
IRC § 7701(a)(51) — definitions of prohibited foreign entity, specified foreign entity, foreign-influenced entity, foreign-controlled entity, effective control, covered nation; control and attribution rules; publicly traded entity exceptions.
IRC § 7701(a)(52) — material assistance from a prohibited foreign entity; material assistance cost ratio; threshold percentages; safe harbor table authority; certification requirements.
IRC § 48E — clean electricity investment credit, including energy storage technology; PFE restrictions.
IRC § 45Y — clean electricity production credit; PFE restrictions.
IRC § 45X — advanced manufacturing production credit; eligible components; PFE restrictions.
IRC § 50(a) — recapture, as amended to add the ten-year prohibited-payment recapture applicable to § 48E.
IRC § 6418 — transfer of credits; prohibition on transfer to a specified foreign entity.
Treas. Reg. § 1.263A-1(e)(2)(i)(A), (B) — direct material and direct labor cost definitions incorporated by reference into the MACR.
NDAA FY2021, Pub. L. 116-283 — § 9901(8) (foreign entity of concern) and § 1260H (Chinese military companies).
Uyghur Forced Labor Prevention Act, Pub. L. 117-78, § 2(d)(2)(B) — entity lists.
NDAA FY2024, Pub. L. 118-31, § 154(b) — statutorily named battery entities.
10 U.S.C. § 4872(f)(2) — definition of covered nation.

IRS guidance

Notice 2026-15 (Feb. 12, 2026) — Guidance to Apply Interim Safe Harbors for Purposes of Determining a Taxpayer's Material Assistance from a Prohibited Foreign Entity; Other Prohibited Foreign Entity Guidance. The controlling interim guidance. Available at irs.gov/pub/irs-drop/n-26-15.pdf
Notice 2025-08 — domestic content elective safe harbor; source of the safe harbor tables incorporated by Notice 2026-15 (§§ 5.05, 5.06, 6.02, 7.02).
Notice 2024-41 — domestic content; § 3.02 tables for hydropower and pumped hydropower storage.
Notice 2023-38 — domestic content; § 3.04 for offshore wind.
Notice 2025-42 (Aug. 15, 2025) — beginning of construction for wind and solar. Vacated June 6, 2026. Confirmed by Notice 2026-15 as inapplicable for PFE and material assistance purposes.
Notices 2013-29, 2018-59 and 2022-61 — the beginning-of-construction framework that governs energy storage, and that governs beginning of construction for PFE and material assistance purposes as in effect January 1, 2025.

Executive action and litigation

Executive Order 14315, Ending Market Distorting Subsidies for Unreliable, Foreign-Controlled Energy Sources (July 7, 2025).
Oregon Environmental Council v. IRS, No. 1:25-cv-4400 (D.D.C. June 6, 2026) — vacating Notice 2025-42 nati onwide and remanding to Treasury and the IRS.

Pending

Proposed regulations on PFE status, effective control, and the MACR — announced in Notice 2026-15; comments were due March 30, 2026.
PFE safe harbor tables — statutorily due December 31, 2026.
Applicable critical minerals threshold guidance — statutorily due December 31, 2027.

This document is provided by Keentel Engineering LLC for general informational and educational purposes. It addresses engineering, procurement and documentation practices associated with federal clean energy tax credit co mpliance. It is not tax advice, legal advice, or an opinion on which any taxpayer may rely. Keentel Engineering is not a law firm or an accounting firm. Determinations of prohibited foreign entity status, material assistance cost ratio compliance, beginning of construction, credit eligibility, recapture exposure and transferability must be made by qualified tax counsel and accounting advisors based on the specific facts of each taxpayer and project.

The prohibited foreign entity regime is in an early and actively developing stage of implementation. Treasury and the IRS have stated their intention to issue proposed regulations and additional guidance, including safe harbor tables. Statutory listings may be amended and administrative lists may be expanded. Litigation affecting related guidance is ongoing. Readers should confirm the current state of the law before acting.

Keentel Engineering LLC is an independent engineering consultancy and is not affiliated with, endorsed by, or acting on behalf of any government agency, standards body, utility, transmission provider, equipment manufacturer, software vendor, or other organization referenced in thi s document. All third-party names are used for identification purposes only.

Transmission Structure Height Calculation: NESC Guide

Sat, 22 Aug 2026 14:46:37 GMT

Aug 22, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Transmission Structure Height Is Not a Function of Voltage

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

Three things, and they matter

Higher voltage does allow lower current for the same power , an d I²R loss does fall accordingly. That is the entire economic case for transmission at high voltag e, and the graphic states it correctly.

Higher voltage does require more clearance. Not as much as the graphic implies, and not in the way it implies, but the direction is right and the NESC quantifies it.

Taller structures do, on average, carry higher voltages. The correlation is real. It is the causation that is wrong — and the difference between those two statements is the difference between a poster and a design.

Before going further, fix the ladder. These are the AC transmission voltages actually in US service, with their ANSI C84.1-2020 maximum system voltages:

1. What the graphic gets right

Same crossing. Same terrain. Different voltage, and — critically — a different span.

Step 1 — Required ground clearance

Table 232-1 base: 18.5 ft. Rule 232C1 adder: phase-to-ground = 500 / √3 = 288.7 kV; (288.7 − 22) × 0.4 in = 106.7 in = 8.89 ft

Required clearance = 27.4 ft , plus a 3 ft buffer = 30.4 ft design clearance.

A note on method at EHV. Rule 232C1's linear adder is available for voltages between 22 and 470 kV, applied phase-to-g round — so it remains nominally available even at 765 kV, where phase-to-ground is 441.7 kV. But NESC Rule 232D, "Alternate clearances for voltages exceeding 98 kV ac to ground or 139 kV dc to ground," offers a switching-surge-based method that EHV designers normally use instead, because the linear adder is conservative at these voltages. Rule 232D requires the maximum switching-surge factor — the switching-surge level for circuit breakers corresponding to 98% probability, or the maximum anticipated level from other means, whichever is greater — and reference heights from Table 232-3. Above 470 kV, Rule 232D is mandatory, not optional. Work the actual numbers from the printed code.

Step 2 — Maximum final sag

500 kV lines run long spans. Take a 1,400 ft ruling span with a three-conductor bundle, governing final sag 45 ft.

Step 3 — Build the stack

Illustrative design guidance — sag, insulator and shield-wire values are example inputs, not lookup values.

AEP publishes 500 kV single-circuit at 120 ft hilly / 135 ft flat; CPUC gives 80–200 ft. Our 110 ft sits inside both.

The NESC and IEEE Version

Transmission Line Design | Technical Guide, FAQ and Case Studies | August 2026

Three things about that list that matter more than they look.

The EHV boundary sits between 230 and 345 kV. A NSI C84.1 defines HV as ≥100 kV and ≤230 kV, and EHV as >230 kV and <1000 kV. So 230 kV is HV, not EHV — a distinction that gets written wrong constantly, and one that matters because "EHV" triggers diffe rent design conventions, different clearance methods and different equipment standards. By the same definition, 69 kV is formally medium voltage, which the industry universally ignores but which is worth knowing before you put it in a document that will be reviewed.

The ladder is not evenly spaced, and it was not designed. The gaps between classes are historical — regional utility choices, equipment availability at the time of build-out, and interconnection decisions made decades apart. 138 kV and 161 kV serve overlapping roles in different parts of the country for no reason more principled than who built first. Do not look for a rule in the spacing; there isn't one.

What a voltage class actually governs. The nominal voltage sets the maximum system voltage, and the maximum system voltage sets the insulation coordination — BIL, insulator string length, air clearances, equipment ratings. That chain is real and it does push structure dimensions upward. What it does not do is set structure height directly, which is the subject of the rest of this article.

4. What actually determines height

2. US transmission voltage classes

NESC Rule 232C1 — Voltages exceeding 22 kV — states that for voltages between 22 and 470 kV, the clearance from Table 232-1 "shall be increased at the rate of 10 mm (0.4 in) per kilovolt in excess of 22 kV." For effectively grounded circuits, that adder is applied to th e phase-to-ground voltage , not phase-to-phase.

So for a supply conductor over a road subject to truck traffic, starting from the Table 232-1 base of 18.5 ft:

Read the middle column. Going from 115 kV to 500 kV — a 4.3× increase in vol tage — adds 7.4 feet of required ground clearance. Going all the way to 765 kV adds 12.5 feet over the 115 kV case.

Now compare that against what utilities actually publish for structure height at a single volt age class:

Georgia Transmission gives 115 kV single poles 80–120 ft roadside and 60–80 ft cross-country — and gives 230 kV single poles the identical 80–120 ft roadside range. Two voltage classes, one number.
CPUC/Aspen gives 220 kV single-circuit tubular steel poles 70–200 ft, and 500 kV single-circuit lattice 80–200 ft. A 2.3× voltage ratio, and the ranges alm ost entirely overlap.
AEP gives 765 kV single-circuit 135 ft in hilly terrain and 150 ft in flat, 500 kV single-circuit 120/135, 345 kV single-circuit 110/125 — and 345 kV double-circuit 160/175 ft.

That last line is the whole argument in one row. AEP's 345 kV double-circuit structures are taller than its 765 kV single-circuit structures. Circuit configuration beats voltage, decisively.

And the arithmetic: the entire voltage contribution across the 115 kV to 500 kV span is 7.4 ft, against a within-class published spread of 130 ft at 220 kV. Voltage accounts for something like 6% of the variation. It is a real term. It is not the term.

One counter-point, stated for honesty: the same CPUC fact sheet that publishes those overlapping ranges also says "higher voltages require greater phase separation… resulting in taller structures with wider cross arms." Voltage is not irrelevant. It is secondary, and the graphic inverts the hierarchy.

5. Worked example A — 115 kV single-circuit, road crossing

Height is a stack. Build it from the ground up.

Required ground clearance — from NESC Table 232-1 for the crossing type, plus the Rule 232C1 voltage adder, plus altitude correction where it applies, plus whatever margin the utility's own standard adds on top.
Maximum final sag at the governing condition — usually the single largest term, and set by span length, conductor type and tension, and the loading district.
Insulator string length — the conductor attaches at the bottom of the string; the crossarm sits at the top of it.
Shield wire position — set by the shielding angle needed for lightning performance, which puts the overhead ground wire above and usually inboard of the outer phases.
Terrain and profile — the controlling point is wherever the ground rises under the span, not necessarily midspan.

Of these , span length is the dominant variable, and it interacts with terrain in a way that surprises people.

AEP's flat-terrain structures are taller than its hilly-terrain structures — 150 ft vs 135 ft at 765 kV. That looks backwards until you see the mechanism. Flat ground permits longer spans; longer spans mean more sag; more sag means more height. Terrain does not act on height directly. It acts through span, and it can act in either direction.

The condition the clearance is checked at

NESC Rule 232A specifies that the vertical clearances of Rule 232B1 apply under the following conductor temperature and loading conditions, whichever produces the largest final sag:

50 °C (120 °F), no wind displacement
The maximum conductor temperature for which the line is designed to operate, if greater than 50 °C, with no wind displacement
0 °C (32 °F), no wind displacement, with the radial thickness of ice for the loading district

Three things to take from that.

It is final sag — long-term, after creep and after the conductor has taken its permanent set. Not initial sag, and emphatically not stringing sag.

It is the largest of three conditions, not a single design case. A line in a Heavy loading district may be governed by the iced condition; a line operating at 100 °C may be governed by the high-temperature condition. You check all three.

And it is at maximum operating temperature , not at the temperature the line happened to be at when someone surveyed it. This is the single most common way an as-built line turns out to be non-compliant while every field measurement looked fine — see Case Study 1.

Design case. 115 kV single-circuit tangent structure, effectively grounded, crossing a two-lane road subject to truck traffic. Cross-country alignment, 800 ft ruling span, Medium loading district.

Step 1 — Required ground clearance

Table 232-1 base for open supply conductors over a road subject to truck traffic: 18.5 ft

Rule 232C1 adder: phase-to-ground = 115 / √3 = 66.4 kV; (66.4 − 22) × 0.4 in = 17.8 in = 1.48 ft

Required clearance = 20.0 ft. Utility standards commonly add a design buffer for survey tolerance, future road grading and conductor creep beyond the design assu mption; carry 2 ft, giving a 22.0 ft design clearance at the controlling point.

Step 2 — Maximum final sag

From the sag-tension calculation for the actual conductor, ruling span, tension limits and loading district, checked against all three Rule 232A conditions. For this example, take the governing final sag at maximum operating temperature as 22 ft.

This number is not a lookup. It is the output of a sag-tension run, and it will move by several feet for the same span if you change conductor, tension limit, or loading district.

Step 3 — Build the stack

Sanity check against published data. Georgia Transmission gives 115 kV single poles 60–80 ft cross-country and 80–120 ft roadside. Our 59 ft lands right at the bottom of the cross-count ry band, which is what you would expect for a level crossing with no distribution underbuild and no roadside constraints. Roadside structures run taller because they carry underbuild, cross the road at unfavourable angles, and are set for future widening.

7. "Wider arms = more capacity" is backwards

Closing

The poster is a teaching aid, and as a teaching aid it is defensible. Higher voltage really does mean lower current, lower loss and — on average — bigger structures. Someone seeing a transmission line for the first time is better off with that mental model than with none.

The trouble starts when it leaves the classroom. A height-by-voltage table looks exactly like a design table, and it is not one. It contains no span, no conductor, no loading district, no ground profile and no crossing type — which is to say it contains none of the variables that actually set the number it reports. Put it in a cost model and it becomes a liability, as Case Study 2 shows. Put its second claim into an uprate study and it points the work in the wrong direction entirely, as Case Study 3 shows.

The honest version is less quotable and more useful: structure height is the sum of a required clearance, a computed sag, an insulator string and a shield wire offset — and span and terrain move it more than voltage does.

At Keentel Engineering, transmission line design, NESC clearance and sag-tension analysis, structure loading, and power system studies are delivered together, because the questions above do not stay in one discipline — a capacity study becomes a clearance study becomes a foundation study. If you have a line to route, a clearance survey to reconcile against Rule 232A, a corridor to uprate, or a typical-heights table you are not sure about, we would be glad to look at it.

KEENTEL ENGINEERING

Tampa, FL · Austin, TX · Sacramento, CA · Baltimore, MD

keentelengineering.com · contact@keentelengineering.com · 813-389-7871

Engineering note: NESC rule numbers and values are given for the 2023 edition (IEEE C2-2023) and were verified against published sources at time of writing; confirm against the edition adopted in your jurisdiction and against the printed code, particularly for Table 232-1 rows and the Rule 232D method, which should be worked from the code text rather than any secondary source. Sag values, insulator string lengths and shield wire offsets in the worked examples are illustrative design inputs, not lookup values. Nothing here substitutes for a project-specific engineering design.

The graphic says nothing about what the structure has to survive, which is where most of the steel actually goes.

NESC Section 25 — loading districts

Rule 250B, combined ice and wind district loading, defines three districts:

This article builds the whole thing on NESC (IEEE C2-2023) and IEEE basis, with two worked height stacks at 115 kV and 500 kV, and corrects the second claim in that key-points box, which is not a simplification but simply backwards.

3. What voltage actually contributes to height

6. Worked example B — 500 kV single-circuit, same road crossing

Step 4 — Now compare the two examples

The structure is 51 ft taller . The voltage adder accounts for 7.4 ft of that — about 15%. The largest single contributor is sag, and sag came from choosing a 1,400 ft span instead of an 800 ft span.

Hold the voltage constant and change the span, and you move the height more than changing the voltage does. That is the finding, and it is why the lookup table on the poster cannot be used as a design tool.

Figure 1 , on the following page, summarizes the whole argument: the height stack for both examples, the Rule 232C1 adder by voltage class, the published height ranges, and the capacity question addressed next.

FIGURE 1 — Where transmission structure height actually comes from: the height stack, the NESC Rule 232C1 voltage adder, published height ranges, and the capacity question

Compare rows two and three. Same voltage. Same number of circuits. Same structure geometry. Going from a two-conductor bundle to a three-conductor bundle raises SIL by 37% — because a larger bundle behaves like a larger effective conductor radius, which lowers Zc and raises both SIL and thermal rating at once. That is what buys capacity. Not arm width.

The line-length dimension

One more piece, because it determines which limit you are actually against. The St. Clair curve — H. P. St. Clair, Practical Concepts in Capability and Performance of Transmission Lines , AIEE 1953, given its analytical basis by Dunlop, Gutman and Marchenko in IEEE Trans. PAS, 1979 — divides lines into three regimes:

Short lines: thermally limited. The conductor's temperature rating governs.
Medium lines: voltage-drop limited. Typically a 5% drop criterion.
Long lines: stability limited. MISO uses an angular displacement criterion of 44.5°; the stability regime typically begins around 300–500 km.

A short 345 kV line and a long 345 kV line have very different usable capacities with identical structures. Capacity is a system property, not a tower property.

The second claim in the key-points box is not a simplification. It is wrong, and it is wrong in a direction that can be demonstrated.

What actually sets phase spacing

Four things, none of which is capacity.

Electrical clearance between phases und er NESC Rule 235, "Clearance for wires, conductors, or cables carried on the same supporting structure."
Insulator swing. Suspension strings blow out under transverse wind, carrying the conductor toward the structure. The arm must be long enough that the conductor maintains clearance to the steel at the design swing angle — RUS 1724E-200 tabulates swing angle values for exactly this purpose.
Galloping. Conductors with asymmetric ice accretion oscillate in large low-frequency ellipses. Phase spacing must keep those ellipses from intersecting. The standard treatment is the Lissajous-ellipse method — CIGRE Technical Brochure 322, State of the Art of Conductor Galloping (WG B2.11, 2007).
Switching surge withstand at EHV, which is the same physics driving Rule 232D.

Every one of those is a clearance requirement. N one of them is a current-carrying or power-transfer requirement.

What actually sets capacity

Thermal rating — the current at which the conductor reaches its maximum allowable temperature, calculated per IEEE Std 738-2023, IEEE Standard for Calculating the Current-Temperature Relationship of Bare Overhead Conductors. It is a function of conductor size, material, surface condition, and the ambient conditions assumed — wind speed, air temperature, solar radiation. Phase spacing does not appear in it.

Surge impedance loading (SIL) — the natural loading of the line:

SIL = V(L-L)² / Zc where Zc = √(L / C)

SIL depends on line voltage and the distributed inductance and capacitance per unit length — and, as ERCOT puts it, "not the length of the line." Typical overhead surge impedance runs roughly 200–400 Ω.

And now the part that makes the claim backwards

Move the phases further apart and the loop area increases, so distributed inductance L rises. At the same time the conductors are further from each other and from ground, so distributed capacitance C falls. Both changes push Zc = √(L/C) up — and SIL = V²/Zc therefore goes down.

Wider arms slightly reduce surge impedance loading. They do not increase it. The claim is not merely unsupported; it has the sign wrong.

What does raise capacity: bundling

Here is the clean proof, from published ERCOT/MISO data:

8. Loading, strength and structure type

The constant K is added to the resultant of the per-linear-foot weight, wind and ice loads.

Beyond the district case, Rule 250C covers extreme wind loading and Rule 250D covers extreme ice with concurrent wind. Rule 250C carries an exclusion for structures under 60 ft — an exclusion that several utilities and RTOs explicitly override in their own criteria. PJM, for one, states that the provision permitting exclusion of structures less than 60 ft from Rule 250C shall not apply. If you are designing to a transmission owner's criteria, read their overrides before you read the code.

The 2023 NESC replaced the extreme-wind speed maps in Figures 250-2(a) and 250-2(b) and reformatted Rule 250C1b. If you are carrying wind speeds forward from a 2017-edition design, check them.

Grades of construction and strength

Section 24 covers grades of construction — Grade B at crossings of limited-access highways, railroads and navigable waterways; Grade C elsewhere. Rule 261 carries the strength requirements, and Table 261-1 gives the strength factors: the permitted load is the designated strength of the structure multiplied by the applicable NESC strength factor.

Structure types and the ASCE layer

US practice uses lattice steel towers, tubular steel poles (single-circuit, double-circuit and H-frame), wood H-frames, concrete poles and guyed structures. The NESC sets the safety floor; the structural design comes from ASCE:

ASCE Manual of Practice 74 (2020, 4th ed.) — Guidelines for Electrical Transmission Line Structural Loading
ASCE/SEI 10-15 — Design of Latticed Steel Transmission Structures
ASCE/SEI 48-19 — Design of Steel Transmission Pole Structures
ASCE 7 wind and ice mapping, which utility practice commonly pairs with MOP 74 for construction and maintenance load cases alongside NESC 250C/250D

And RUS Bulletin 1724E-200, Desi gn Manual for High Voltage Transmission Lines , remains the most complete free US reference for the whole chain — clearances in Chapter 4, ROW in Chapter 5, conductor-to-conductor and galloping in Chapter 6, insulator swing in Chapter 7, insulation in Chapter 8, loadings in Chapter 11, structures in Chapter 13, guyed structures in Chapter 14.

Right-of-way

Height's companion variable, and also not a voltage lookup — though it correlates more strongly. AEP publishes 200 ft for 765 kV single-circuit, 175–200 ft for 500 kV, and 150 ft for 345 kV single- or double-circuit. Georgia Transmission publishes roughly 100 ft for 115 kV, about 100 ft for 230 kV single pole and 125 ft for 230 kV H-frame, and 150–180 ft for 500 kV lattice. RUS 1724E-200 Table 5-3 is the canonical citable table.

10. Anonymized case studies

9. Eight things worth correcting

1. Structure height is computed, not looked up. Ground clearance plus final sag plus insulator length plus shield wire position, driven mainly by span and terrain.

2. The voltage contribution is small and quantifiable. NESC Rule 232C1: 0.4 in per kV above 22 kV, applied phase-to-ground. From 115 kV to 500 kV that is 7.4 ft, against within-class published height ranges of 130 ft.

3. "Wider arms = more capacity" is backwards. Wider spacing raises Zc and therefore lowers SIL. Capacity comes from conductor size and bundling.

4. A height quoted without a span is not a design value. Span, conductor, tension and loading district all have to travel with the number.

5. One height does not apply to a whole line. The controlling ground point moves with the profile, and so does the required structure height.

6. 230 kV is HV, not EHV. ANSI C84.1 puts the EHV boundary above 230 kV.

7. Clearance is checked at final sag under the worst of three Rule 232A conditions — including maximum operating temperature — not at the sag you can measure on a mild day.

8. Circuit configuration outranks voltage. AEP's 345 kV double-circuit structures are taller than its 765 kV single-circuit structures.

12. Standards reference

The following are anonymized and composited from typical engagements. Figures are representative of the scenarios described and have been rounded and adjusted; they do not identify any specific client or facility.

Case Study 1 — 138 kV rebuild: the survey that proved nothing

Situation. A utility completed a 12-mile 138 kV rebuild. The construction survey confirmed ground clearance at every road crossing with margin. Two summers later, during a regional heat event with the line loaded near its rating, a contractor working under the line reported clearance that looked visibly short at one crossing.

What the analysis found

The as-built survey had been performed in October at an ambient of about 55 °F, with the line lightly loaded. Measured conductor temperature was on the order of 60 °F. At that condition the clearance at the crossing was 24 ft — comfortably above the 20.4 ft required for 138 kV over a road subject to truck traffic.

The line's design maximum operating temperature was 212 °F. Und er NESC Rule 232A, clearance must be verified at whichever of the three specified conditions produces the largest final sag — and for this line that was condition 2, the maximum operating temperature. Modelling the conductor at 212 °F final sag put clearance at the crossing at 19.1 ft, about 1.3 ft below the requirem ent.

Nothing had changed in the field. The survey had simply never been converted to the code condition.

What was done

The survey data was reprocessed properly: conductor positions were used to back-calculate tension at the measured temperature, and sag was then modelled at all three Rule 232A conditions across the whole rebuild. Four crossings were found short at maximum operating temperature; a fifth was short under the iced condition, which the utility had not been checking at all because the line was in a Light loading district and the team had assumed ice was not a case. It is a case — Light district specifies no radial ice, but the 0 °C condition still applies and can govern on short spans with high tension.

Two structures were raised, one span was re-tensioned, and the utility's survey procedure was rewritten to require the conversion to Rule 232A conditions as a deliverable rather than a subsequent analysis someone might or might not run.

Outcome

No incident. Four crossings corrected. The procedural change is the durable part — the utility now cannot close out a line survey without the clearance-at-governing-condition calculation attached.

Engineering takeaway

A clearance survey is a measurement of where the conductor was, not a demonstration that the line is compliant. Compliance lives at the Rule 232A governing condition, and getting from one to the other is a calculation that has to be somebody's named deliverable.

Case Study 2 — 230 kV greenfield: the height table that came from a poster

Situation. A developer's preliminary design for a 230 kV single-circuit line set a uniform 115 ft structure height across a 20-mile alignment, citing an internal "typical heights by voltage" table. Structure count, foundation quantities and the cost estimate were all built on that number, and the estimate went into a bid.

What the analysis found

The 115 ft figure had no derivation behind it. Traced back, it originated from a general-reference table of the same kind as the poster this article is about — a plausible midpoint presented as a design value.

Running the actual profile told a different story. On the flat southern third of the alignment, the terrain permitted 1,10 0 ft spans, and at that span the governing final sag required attachment heights that put structures at 130 to 140 ft. On the rolling northern section, ridge-to-ridge geometry gave shorter effective spans and the controlling ground points fell away under the conductor, so 90 to 100 ft structures were sufficient — and in three locations, 85 ft.

The uniform 115 ft assumption was simultaneously too s hort over a third of the line and too tall over another third. The too-short third was the expensive problem: it would have surfaced during detailed design, after the bid.

What was done

The alignment was resegmented into four span-and-height regimes driven by the ground profile rather than the voltage. Structure heights were derived from the clearance-plus-sag stack in each regime. Total steel tonnage rose about 6% against the original estimate; structure count fell, because the taller structures in the flat section carried longer spans. The developer's internal typical-heights table was retired and replaced with a one-page procedure that starts from the profile.

Outcome

The corrected quantities went into the bid. The 6% steel increase was a known number before award rather than a change order after it.

Engineering takeaway

A typical-heights-by-voltage table is a communication aid, not an estimating basis. The moment it enters a cost model it stops being a simplification and starts being a liability — and the error is not symmetric, because the too-short segments cost far more to fix than the too-tall segments save.

Case Study 3 — 345 kV uprate: the request for wider arms

Situation. A transmission owner facing a congested 345 kV corridor asked for a study on increasing transfer capability. The initiating request, written by a planner working from a general reference, proposed "widening the crossarms to increase the line's power transfer capacity" as one of the options to be evaluated.

What the analysis found

The option was evaluated, and it does not work — in fact it moves the wrong way.

Widening phase spacing increases the distributed inductance of the line and decreases its distributed capacitance. Surge impedance Zc = √(L/C) therefore rises , and since SIL = V²/Zc, surge impedance loa ding falls. The modelled effect of the proposed spacing increase was a reduction in SIL of a little under 2%. Thermal rating was unchanged, because conductor thermal rating under IEEE Std 738 is a function of the conductor and the ambient conditions, not of the geometry of the structure holding it.

The study also established which limit the line was actually against. At its length, the corridor sat in the voltage-drop-limited regime of the St. Clair curve — not the thermal regime — so a pure thermal uprate would not have delivered the transfer capability the planner wanted either.

What was done

Three options were carried forward: adding a third sub-conductor to the existing two-conductor bundle; reconductoring to a high-temperature low-sag conductor; and series compensation to address the voltage-drop limit directly.

Bundling came out ahead on capability per dollar. Published comparisons of 345 kV double-circuit config urations show the move from a two-conductor to a three-conductor bundle raising SIL by roughly 37% at identical voltage and circuit count — the mechanism being a larger effective conductor radius, which lowers Zc and raises both SIL and thermal rating together.

The catch, and the reason this was a structural study and not only an electrical one: a third sub-conductor adds weight and wind area, which changes sag, which changes clearance, and changes structure and foundation loading. The uprate was feasible, but it was a transmission line design problem, not a conductor swap.

Outcome

The corridor was uprated by bundling with structure reinforcement at 40% of the structures and clearance correction at eleven crossings. Arm width was never touched.

Engineering takeaway

Capacity lives in the conductor and the system; clearance lives in the geometry. When someone proposes a geometric change to solve a capacity problem, the first question is which of the two they have actually confused — and the second is which limit the line is really against, because a thermal fix does nothing for a stability-limited or voltage-drop-limited corridor.

11. Frequently asked questions

Gas-Insulated Substations (GIS): Design & Safety Guide

Sat, 22 Aug 2026 09:59:29 GMT

Aug 22, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Gas-Insulated Substations: What the Component Diagrams Leave Out

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

Credit first, because the framework is sound.

A GIS bay really is the same functional set as an air-insulated bay — bus, breaker, disconnectors, earthing switches, instrument transformers — with two differences: the live parts sit inside grounded aluminum enclosures filled with an insulating gas at a few bar, and the dielectric strength of that arrangement lets you shrink the phase-to-phase and phase-to-ground clearances by roughly an order of magnitude.

The diagram is also right that:

The circuit breaker interrupts load and fault current, and in a conventional GIS the same gas that insulates also quenches the arc.
The disconnector is a no-load device. It is not rated to break load current, and operating one under load is a well-known way to destroy a GIS bay.
The combined disconnector/earthing switch (DS/ES) saves length and gas volume by putting two functions in one enclosure — the single most common space-saving move in modern GIS design.
The make-proof earthing switch is a real and distinct device class, rated to close onto a fault without welding shut.
CTs and VTs feed protection and metering.

All correct. Now the parts that are not.

T he diagram states that the disconnector "provides visible isolation of equipment." In a GIS, that is exactly backwards — and the standard that governs disconnectors says so, clearly, in a sentence worth reading twice.

IEC 62271-102, subclause 5.104.3.1, "Indication of position":

"It shall be possible to know the operating position of the disconnector or earthing switch. For the open position this requirement is met if one of the following conditions is fulfilled: — the isolating distance or gap is visible; — the position of each movable contact ensuring the isolating distance or gap is indicated by a reliable visual position indicating device."

Read the structure. The standard gives two alternative ways to satisfy a single requirement — knowing the position . Air-insulated switchgear satisfies it the first way: you look up and see the gap. A GIS disconnector is a moving contact inside an opaque, grounded, gas-filled aluminum enclosure. It cannot satisfy the first condition and must satisfy the second.

Visible isolation is not a property of a GIS disconnector. It is the property GIS gives up in exchange for everything else it provides.

What replaces it

This is not a weakness — it is a different, and arguably more rigorous, safety architecture. Four elements do the work that a visible gap does in AIS.

A position indicator that is mechanically coupled, not inferred. The normative annex of IEC 62271-102 does not treat "reliable" as a marketing word. The kinematic chain between the moving contact and the indicator must be a continuous mechanical connection to ensure a positively driven operation , must have sufficient mechanical strength, and must indicate open or closed only after the moving contacts have actually reached that position. A strain-limiting device is not permitted anywhere in that chain. The indicator is not a lamp driven by an auxiliary switch — it is a mechanical extension of the contact itself, type-tested to prove it.

Interlocking. Mechanical and electrical interlocks between breaker, disconnector and earthing switch prevent the sequence errors that cause GIS failures. Manual operation is possible only under electrical interlock release. The feeder earthing switch is interlocked with its breaker and disconnector.

Voltage presence verification. Interlocking the earthing switch against line VT secondary voltage through undervoltage relay contacts proves the circuit is dead before the earthing switch can close.

The integrated earthing switch itself. In AIS, the visible gap is the proof of the safe state. In GIS, the applied earth is the proof. That is a stronger guarantee, because an earthing switch does not merely separate you from the source — it holds the conductor at ground potential against induced and capacitively coupled voltage.

Some GIS designs offer inspection windows, and some utilities specify them or add camera monitoring of the viewport. Neither IEC 62271-102 nor IEC 62271-203 requires them. They are a specification option, not a code baseline, and they should never be the sole basis of a switching procedure.

The practical consequence. If your lockout/tagout or switching procedure contains the phrase "verify visible break," it was written for air-insulated equipment and it cannot be executed on a GIS. It must be rewritten around position indication, interlock status, applied earths, and voltage verification. See Case Study 2 — this is not hypothetical.

1. What the diagram gets right

The decision is rarely about the equipment. It is about the site.

1. GIS discon nectors do not provide visible isolation. IEC 62271-102 5.104.3.1 offers a visible gap or a reliable visual position indicating device; GIS ne cessarily uses the second.

2. ES is the earthing switch, not an earthing transformer.

3. A panel titled "DS – Circuit Breaker " containing disconnector text is a mislabel.

4. The permissible leakage rate for SF₆ GIS unde r IEC 62271-203 Ed. 3.0 is 0.1% per year per gas compartment, not the older 0.5%.

5. Cite current editions: C37.122-2021, C37.122.2-2022, C37.122.3-2024, C37.122.6-2013, C37.122.7-2021, C37.20.9-2025, IEC 62271-203:2022 Ed. 3.0. The IEC document was explicitly extended in 2022 to cover alternative gases alongside SF₆ — which is why it is the right reference for a non-SF₆ specification.

6. Hitachi EconiQ is fluoronitrile-based, not clean air. Only Siemens blue is vacuum plus clean air. Vendors' marketing names do not map to a single techno logy.

7. 420 kV vacuum-plus-clean-air GIS is a pilot, not a product. 420 kV C4-FN GIL has been in service since 2017.

8. SF₆ is an asphyxiant that pools. Five times heavier than air, in trenches and cable basements. The building design has to address it.

Historically only about 2 to 5% of new US substations were built as GIS, with the cost case strengthening as voltage rises. That share is climbing, driven by urban load growth, land constraints and resiliency requirements.

Seismic qualification. IEEE Std 693-2018 (with amendment 693a-2024) is the US recommended practice for seismic design of substations, defining Low, Moderate and High qualification levels. But the standard written specifically for this equipment is IEC 62271-207, "Seismic qualification for gas-insulated switchgear assemblies for rated voltages above 52 kV." It uses the same three levels, harmonized with IEEE 693, and permits site-specific spectra. On a GIS project, specify both.

Arc flash — and a trap. IEEE Std 1584-2018 applies only from 208 V to 15 kV. Above 15 kV it is outside its range of validity and should not be used to calculate incident energy. For enclosed switchgear at 52 kV and below, IEEE C37.20.7-2024, "Recommended Practice for Testing Switchgear Rated Up to 52 kV for Internal Arcing Faults," governs arc-resistant construction testing. Internal Arc Classification with accessibility types — the IAC/A/B and F/L/R designations — is defined in IEC 62271-200 for MV switchgear at 52 kV and below, not in IEC 62271-203. NFPA 70E requires that an assessment be performed; it does not prescribe the method above 15 kV.

Figure 1, on the following page, summarizes the whole picture: the bay and its gas zones, the isolation question, what the diagrams omit, the two SF₆ schedules, and the alternatives compared.

FIGURE 1 — Gas-insulated substations: the bay and its gas zones, the isolation question, what the diagrams omit, the SF₆ schedules, and the alternatives

Substation Design | Technical Guide, FAQ and Case Studies | August 2026

3. Four more corrections

5. The SF₆ question

2. The error that matters: "visible isolation"

Converting non-SF₆ equipment to SF₆ is also prohibited from the applicable date. Replacement parts for existing equipment are exempt. Annual reports are due to CARB by 1 June.

Note the breakpoints: 38 / 145 / 245 kV, with sub-thresholds at 25 kA and 63 kA. They are not the same as the EU's. Do not blend the two schedules — a specificat ion written against the wrong set of numbers will be wrong at exactly the voltage classes where it matters.

5.4 The EU schedule, for anyone with a global supply chain

Regulation (EU) 2024/573 prohibits placing on the market:

4. What the diagrams leave out entirely

Everything above is about labels. This section is about engineering, and it is where GIS actually differs from a metal box full of familiar devices.

4.1 Gas compartments are a design decision, not a detail

A GIS is not one volume of gas. It is divided into gas compartments separated by gas-tight barrier insulators, and the partitioning scheme determines what has to be de-gassed, opened, evacuated and refilled to work on any given component — and therefore how much of the substation goes out of service for any given job.

Segment too coarsely and a single leak or a single component failure takes an entire bus section out. Segment too finely and you add barrier insulators, gas zones, density monitors, cost and leak paths. This is one of the highest-leverage decisions in GIS specification, it is invisible on the single-line diagram, and it is almost never discussed in component overviews. See Case Study 3.

4.2 Density monitoring, not pressure gauges

GIS uses temperature-compensated density monitors, not pressure gauges. The reason is simple: pressure in a sealed compartment swings with ambient temperature through the day while the quantity of gas is unchanged. A pressure reading cannot distinguish a cold morning from a slow leak. Density monitors measure pressure and temperature together and compute density against manufacturer correction curves.

Each gas zone gets a two-stage scheme: stage 1 alarm (top up at the next opportunity) and stage 2 lockout (block operation or trip and isolate). Setpoints are compartment- and manufacturer-specific and should never be copied from a generic table.

The permissible leakage rate has tightened. IEC 62271-203 Edition 3.0 (2022) states in its foreword that "the tightness requirements for type tests for gasses with GWP > 1 000 has been reduced from 0,5 % to 0,1 % per year per gas compartment." If your specification still calls for 0.5% per year on SF₆ equipment, it is citing the previous edition.

4.3 Very fast transient overvoltages (VFTO)

This is a GIS-specific phenomenon with no real air-insulated equivalent. When a GIS disconnector switches a small capacitive current, the restrikes across the slowly moving contacts launch travelling waves between the inner conductor and the enclosure. Because the geometry is a low-loss coaxial line, those waves bounce.

The result is a tra nsient with a rise time in the nanosecond range — roughly 10 to 1,000 ns depending on gas pressure — and a peak typically in the range of 1.4 to 2.0 per unit of rated voltage. Standard lightning impulse testing does not represent this waveform, because the rise time is two orders of magnitude faster.

VFTO matters for insulation coordination inside the GIS, for transformer windings connected directly to a GIS without intervening cable, and for the secondary systems that see the transient couple out. IEC 62271-102 Annex F addresses disconnector switching requirements at 72.5 kV and above; IEC 60071-1 covers insulation coordination; CIGRE has treated it extensively in WG A3.22, A3.28 and C4.306 and Technical Brochures 362, 400 and 456.

4.4 Enclosure currents and transient enclosure voltage

The aluminum enclosure of a GIS is not passive. It carries an induced return current that can reach up to about 90% of the conductor current, flowing in the opposite direction. That is a feature, not a problem — the near-cancellation of the magnetic field is one reason GIS has such low external field — but it only works if the enclosure is electrically continuous.

Every module must be bonde d, by flange connections or external shunts, into a continuous enclosure. Where that continuity is broken — at discontinuities, at expansion joints, at air-to-gas bushings — travelling waves can produce transient enclosure voltage (TEV). TEV is generally not a direct shock hazard to personnel, but it will produce electrostatic sparks if the multipoint grounding system is not correctly installed, and those sparks are both a personnel startle hazard and a nuisance-trip source for nearby secondary wiring.

GIS grounding is a multipoint scheme by design, and it interacts with — but is not the same as — the station ground grid analysis under IEEE Std 80.

4.5 Partial discharge monitoring

GIS is unusually intolerant of small internal defects. A free metallic particle a few millimetres long, a protrusion on a conductor, a floating electrode, a void or delamination in a cast spacer — any of these can initiate a flashover in a system whose dielectric margin is engineered tight.

The industry answe r is UHF partial discharge monitoring, using sensors coupled into the enclosure. Per IEC TS 62478, the UHF band is nominally 300 MHz to 3 GHz ; commercial GIS instruments typically work in a sub-band of that. Sensitivity verification is covered by CIGRE TB 654 (WG D1.25), and the more recent CIGRE TB 933 (WG D1.66) addresses requirements and application of UHF PD monitoring systems for gas-insulated systems.

PD monitoring is now routine on new transmission-class GIS and is one of the strongest arguments for the technology: an internal defect that would be invisible in an AIS yard is detectable, trendable and locatable in a GIS.

4.6 Decomposition byproducts and personnel safety

SF₆ itself is chemically inert and non-toxic. Arced SF₆ is neither.

The bypro ducts of arcing and of arcing against silica-filled epoxy spacers include SOF₂ (thionyl fluoride), SO₂F₂ (sulfuryl fluoride), SF₄ (sulfur tetrafluoride), S₂F₁₀ (disulfur decafluoride), HF (hydrogen fluoride), SiF₄ (silicon tetrafluoride), SO₂, and solid metal fluorides as a fine powder. EPA's own guidance identifies S₂F₁₀ as the most toxic of these, on the order of 43 times more toxic than the others.

OSHA permissible exposure limits apply to several of these directly — S₂F₁₀ at a 0.025 ppm TWA, SF₄ at a 0.1 ppm ceiling, HF at 3 ppm TWA, SO₂ at 2 ppm TWA. Protective clothing and an approved respirator are required where decomposition products may be present, and the solid fluoride powder must be handled as a contaminant, not swept up.

There is no SF₆-specific OSHA rule, but SF₆ does carry an OSHA PEL of 1,000 ppm — and the practical hazard is not toxicity, it is asphyxiation. SF₆ is roughly five times heavier than air. It pools in cable basements, trenches, pits and the low corners of GIS buildings. Confined-space procedures and oxygen monitoring belong in the design of the building, not just in the O&M manual.

IEEE C37.122.3-2024 is the current gas-handling guide. Note the edition — it superseded the 2011 version, and specifications still calling for C37.122.3-2011 are two revisions behind on a document that has been substantially updated as alternative gases entered service.

Its atmospheri c lifetime is approximately 3,200 years. A kilogram released today is still working in the year 5200.

Which number you use depends on the regime. The US EPA Greenhouse Gas Reporting Program uses AR4 (22,800); the US national GHG inventory uses AR5 (23,500) under UNFCCC rules; EU Regulation 2024/573 uses AR6 (24,300), which is the figure that now appears on mandatory F-gas equipment labels.

5.2 What US federal law actually requires

Less than most people assume, and it is worth being precise because vendors are not always.

40 CFR Part 98, Subpart DD — "Electrical Transmission and Distribution Equipment Use" — requires electric power systems to report emissions of fluorinated GHGs from transmission and distribution equipment, where the facility meets the general 25,000 metric ton CO₂e per year threshold. Reported data includes nameplate capacity of insulating gas, gas acquired and disbursed, and transmission and distribution line miles.

There is no federal EPA ban on SF₆ in electrical equipment. Subpart DD establishes reporting requirements only — no operational bans, no emission limits. And the voluntary EPA SF₆ Emission Reduction Partnership (later the Electric Power Systems Partnership) is no longer active; its resources are archived.

If a vendor tells you federal law is forcing your hand, it is not. State law and, increasingly, corporate ESG commitments are.

5.3 California CARB — the schedule that is actually binding

California adopted its Regulation for Reducing Sulfur Hexafluoride Emissions from Gas Insulated Switchgear in October 2020, effective January 2022, codified at 17 CCR §§95350–95359.1. Section 95352 sets a phase-out: from the dates below, no person may acquire SF₆ gas-insulated equipment for use in California.

38 kV and below (Table 1):

Every popular GIS diagram says the same thing: the equipment uses SF₆ for insulation and arc quenching. That was a complete answer in 2015. In 2026 it is the beginning of a procurement conversation, and getting it wrong can strand a substation design.

5.1 Why the pressure exists

SF₆ is the most potent greenhouse gas the IPCC has assessed. Its 100-year global warming potential is:

Above 38 kV (Table 2):

Plus a date that catches everyone by surprise: from 1 January 2035, only reclaimed or recycled SF₆ may be used for maintenance or servicing of electrical switchgear.

Derogations exist under Articles 13(11) and 13(12) where suitable alternatives are unavailable or technically infeasible. Repair and servicing of existing equipment is not caught by the placing-on-the-market prohibitions.

Why does this matter to a US project? Because the manufacturers are global. Product lines are being redesigned to the EU dates, and the SF₆ variant of a given GIS family may be discontinued well before your local regulation requires it.

5.5 The alternatives, honestly assessed

Three technologies are in commercial play, and they are not equivalent.

Vacuum interrupter plus clean air. Siemens Energy's blue portfolio uses a vacuum interrupter for switching and clean air — nitrogen and oxygen only — for insulation. GWP of zero, no fluorinated gas at all. This is the cleanest answer where it is available. Commercially available at 72.5 kV (8VM3, 8VM1) and 145 kV (8VN1). Be precise about the upper end: 420 kV blue GIS is not in commercial service. It is an EU LIFE demonstration project launched October 2024 running to September 2028, with a single pilot bay at an ELIA substation in Belgium.

Fluoronitrile (C4-FN) mixtures. C4-FN — perfluoroisobutyronitrile, CAS 42532-60-5 — is used diluted in CO₂ and O₂ because it liquefies at −4.7 °C pure. Typical HV mixtures run 3.5–5% C4-FN with 10–13% O₂, balance CO₂, rated to −25 °C or −30 °C. GE Vernova markets it as g3, with a mixture GWP around 327 versus SF₆'s 24,300 — roughly a 99% reduction. Hitachi Energy's EconiQ is also C4-FN based, including the ELK-3 420 kV GIS at 63 kA. Track record at the top end: the world's first SF₆-free 420 kV g3 GIL was energized at National Grid's Sellindge substation in April 2017; a 420 kV g3 circuit breaker for GIS was unveiled in August 2024.

The practical summary today: vacuum plus clean air is proven and commercial to 145 kV; C4-FN mixtures are the only route currently offered at 420 kV and above.

5.6 The question nobody is asking loudly enough

C4-FN is a PFAS. Any molecule with more than two carbon atoms bonded to fluorine falls inside the definition, and C₄F₇N does.

The ECHA universal PFAS restriction is live: a narrowed proposal was published in August 2025, committee opinions are expected at the end of 2026, and a Commission decision is possible in 2027. It covers 21 sectors including energy, with sector derogations of five to thirteen and a half years.

And the supply chain has already moved. 3M announced in December 2022 that it would exit all PFAS manufacturing by the end of 2025 — which covers the Novec line, including Novec 4710 (C4-FN) and Novec 5110 (C5-FK), the additives in the leading fluorinated alternatives.

None of this means C4-FN equipment is a bad choice today. It does mean that a utility replacing an SF₆ fleet with a fluorinated alternative is making a second gas-transition decision, not a final one, and should say so in its asset strategy. Where the voltage class allows it, vacuum plus clean air avoids the question entirely.

9. Anonymized case studies

Closing

Frequently Asked Questions

There is a widely shared infographic that labels the parts of a GIS bay — bus, circuit breaker, disconnector, combined disconnector/earthing switch, CT, VT — and gives each a two-line description. As an orientation slide it does its job. Most engineers who have never stood next to a GIS could point at the right pieces after reading it.

But one of its statements is not merely imprecise. It describes the safety property that GIS does not have , and it describes it as the reason you would use the equipment. If that sentence made it into a switching procedure, it would put someone at risk.

This article does three things. It corrects what the popular diagrams get wrong, on IEEE C37.122-series and IEC 62271 basis. It covers the GIS-specific engineering the diagrams omit entirely — gas zoning, density monitoring, very fast transients, enclosure currents, partial discharge, decomposition byproducts. And it deals with the question that now dominates every GIS procurement in North America: what happens when you cannot buy SF₆ equipment any more.

"ES = Earthing transformer." The diagram's legend expands ES as earthing transformer . ES is the earthing switch — the device drawn in the picture. An earthing transformer is an entirely different animal: a zig-zag or wye-delta transformer that creates a neutral point on a delta or ungrounded system so it can be resistance- or reactance-grounded. Putting one term where the other belongs is the kind of error that survives into a bill of materials.

A box titled "DS - Circuit Breaker." It contains the disconnector's description — visible isolation, no-load operation, maintenance safety — duplicating the DS box two panels over. It is a mislabel, not a device.

Two identical CT panels. "CT - Current Transformer" and "CT - Current" carry the same three bullets. Space that could have described gas zoning or interlocking.

And the VT is drawn without an isolating disconnector. In practice, GIS VT compartments are frequently separated by a disconnector or removable link so the VT can be isolated for testing without de-gassing the bus. Whether you need one is a design decision, but it belongs on the diagram.

6. GIS or AIS?

Illustrative design guidance — not a substitute for a project-specific engineering design or a qualified safety procedure review.

7. Eight things worth correcting

The component diagrams are not wrong to exist. Somebody has to explain what a disconnector does before anyone can specify one. But they describe a GIS as a familiar substation with the air replaced, and it is not that. It is a system whose safety architecture, failure modes, maintenance philosophy and regulatory exposure are all different, and every one of those differences shows up in a document — a switching procedure, a specification, a gas handling plan, a procurement schedule — long before it shows up in a fault.

The three case studies above are the same story told three ways: a GIS was treated as an AIS substitution, and the substitution held right up until it did not. A regulatory date, a procedure step, a mechanical detail. None of them were equipment failures.

At Keentel Engineering, s ubstation design power system studies, and owner's engineering are delivered together, because the questions above cross all three. If you are specifying a GIS, converting from AIS, writing a switching procedure for equipment you have not operated before, or trying to work out what your SF₆ exposure looks like in 2033, we would be glad to look at it.

10. Standards reference

Case Study 1 — California, 230 kV: a phase-out date discovered at procurement

Situation. A California transmission owner was midway through detailed design of a 230 kV GIS addition at an existing urban substation. The design basis, written two years earlier, specified conventional SF₆ GIS with a 63 kA short-circuit rating. Long-lead procurement was scheduled to begin the following quarter.

What the analysis found. Under 17 CCR §95352 Table 2, equipment in the 145 kV to 245 kV band rated below 63 kA could no longer be acquired in California after 1 January 2027, and equipment at 63 kA and above after 1 January 2031. The project's 63 kA rating placed it in the later bracket — but only just, and the rating had been selected with margin rather than from a hard system requirement.

The short-circuit study was revisited. Available fault duty at the bus was well under 50 kA, and the 63 kA specification was inherited f rom a template. Had it been trimmed to a standard 50 kA rating, the project would have fallen into the 2027 b racket and the SF₆ design would have been unpurchasable before the second phase of the build.

Separately, the manufacturer's roadmap indicated the SF₆ variant of the chosen 245 kV family was scheduled for discontinuation in line with the EU 2032 date, with engineering support winding down earlier.

What was done. The design was converted to a C4-FN mixture GIS at the same voltage and current ratings. Three things had to be reworked: enclosure dimensions grew modestly, which propagated into the building layout and the crane coverage; the gas handling plan, spares strategy and technician training were rewritten around a mixture rather than a single gas; and the asset strategy was updated to record explicitly that C4-FN is a PFAS and that a second gas transition may be required within the asset's life.

Outcome. The change added roughly four months to the design phase and avoided a procurement dead end that would have surfaced during construction of phase two. The 63 kA rating was retained — but as a documented decision with a stated basis, not an inherited default.

Engineering takeaway. On a GIS project, the regulatory schedule is a design input with the same standing as the fault duty. And check whether an inherited rating is putting you in a different regulatory bracket than the system actually requires — the CARB thresholds sit at 25 kA and 63 kA, and margin you did not need can move your deadline by four years in either direction.

Case Study 2 — Industrial plant, 115 kV: the procedure that could not be performed

Situation. A process plant replaced a congested outdoor 115 kV air-insulated switchyard with an indoor GIS. Commissioning went well. Six weeks into operation, a maintenance crew preparing to work on a feeder cable stopped work and escalated, because they could not complete step 7 of the switching procedure.

What the analysis found. The plant's high-voltage switching and lockout/tagout procedure — carried over from the AIS installation with the equipment names updated — required the switching operator to " visually verify the open disconnect blades before applying grounds."

There were no blades to see. The disconnector was inside a grounded aluminum enclosure. The crew, correctly, refused to proceed on the basis of a step they could not perform, and the escalation was the system working as intended.

Review found the deeper problem. The procedure had been treated as a document to be find-and-replaced rather than rewritten, and the same assumption ran through several other documents: the operator training material, the permit-to-work form, and a contractor safety briefing that instructed visiting crews to "confirm visible isolation."

What was done. The switching procedure was rewritten from the equipment up, around the four things that actually establish the safe state on a GIS: the mechanically coupled position indicator and what a compliant indicator is; interlock status and the electrical release conditions; applied earthing switches as the physical proof; and voltage verification through the VT undervoltage interlock. Operator training was rebuilt around position indication and interlocking rather than visual confirmation. The local control cabinet mimic was re-labelled to make position indication unambiguous at a glance.

The plant also added a step that has no AIS equivalent — confirming the gas density status of the relevant compartments before any operation, because a compartment in stage-2 lockout will not operate and the crew needs to know that before they are standing at the panel rather than after.

Outcome. No injury and no incident. The cost was a procedure rewrite and a retraining cycle. The near-miss report is now part of the plant's onboarding for the GIS.

Engineering takeaway. The most dangerous document on an AIS-to-GIS conversion is the procedure that was updated instead of rewritten. "Visible isolation" is not a phrase that can be find-and-replaced — the safety architecture underneath it is different, and every document that assumed a visible break has to be rebuilt on position indication, interlocks and applied earths.

Case Study 3 — Utility, 138 kV: one leak, one bus, one long outage

Situation. A utility's 138 kV GIS lost an entire bus section for eleven days after a slow gas leak developed at a flange on a single bay. The equipment performed exactly as designed. The design was the problem.

What the analysis found. The GIS had been procured to the manufacturer's standard gas-compartment arrangement, which placed the entire main bus run across five bays in a single gas zone. There was no barrier insulator between bays on the bus.

When the density monitor for that zone reached its stage-2 setpoint, the protection scheme correctly blocked operation and the bus section was taken out of service. Repair required de-gassing the whole zone — recovering the gas, opening the enclosure, replacing the flange seal, evacuating, refilling to the correct density, and completing dielectric and moisture verification before re-energizing. Eleven days, five bays, one seal.

The specification had said nothing about compartment segmentation. It specified voltage, current, short-circuit rating, ambient, seismic level and interface details, and left the gas zoning to the vendor's standard offering — where it is optimized for the vendor's cost, not the utility's outage exposure.

What was done. Two changes, one immediate and one structural.

Immediately, the utility's GIS specification template was rewritten to require the gas compartment scheme as a submittal, evaluated against a stated outage philosophy: no single gas compartment may contain more than one bay's bus section, and the compartment boundary must permit work on any single bay without de-gassing an adjacent one. The number of density monitors, alarm points and SCADA mappings rose accordingly, and those costs were accepted explicitly against an avoided-outage case rather than absorbed as a surprise.

Structurally, the existing installation was retrofitted at the next planned outage with barrier insulators between bays on the bus run, converting one gas zone into four. The retrofit was performed during a scheduled outage that was already planned for control system work.

The utility also took the opportunity to correct a specification item it had been carrying for years: the leakage requirement still read 0.5% per year per gas compartment, the older figure. IEC 62271-203 Edition 3.0 reduced the type-test requirement to 0.1% per year per compartment for gases with GWP above 1,000. The template now cites the current edition.

Outcome. The same failure today takes one bay out, not five. The eleven-day outage became the business case for a specification change that will apply to every GIS the utility buys.

Engineering takeaway. Gas compartment segmentation is an availability decision disguised as a mechanical detail, and if you do not specify it, someone else decides it for you. Work it backwards from the outages you have to survive, put it in the specification as a submittal, and price the density monitors against the outage they prevent.

AC Cable Sizing for Solar PV Plants: NEC Guide

Sat, 22 Aug 2026 09:25:20 GMT

Aug 22, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Sizing AC Cables in a Utility-Scale Solar PV Plant

Any AC conductor in a PV plant has to pass four independent tests. Fail any one and the conductor is wrong, no matter how comfortably it passes the other three.

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

The US / NEC Method

Power Systems Design | Technical Guide, FAQ and Case Studies | August 2026

6. Step 5 — Short-circuit withstand: the US method

The popular graphic covers the first three. It omits the fourth almost entirely, and inside the first it omits the single most important number in US PV conductor sizing: 125%.

NEC 690.8(B)(1) requires that PV circuit conductors have an ampacity, before the application of adjustment and correction factors, of not less than 125% of the maximum current determined in 690.8(A). This is the continuous-duty multiplier, and it exists because a PV plant runs at full output for hours at a time — unlike almost any other load on a distribution system.

Skip it and every conductor in your plant is undersized by 25%. There is no derating stack, no cable-tray factor, and no clever conductor material that recovers from that.

Edition note. 690.8(A)(3) is the inverter-output-circuit current rule in the 2017, 2020 and 2023 NEC. The 2026 NEC restructured 690.8(A), and the same rule now sits at 690.8(A)(1)(c). Cite the subsection with its edition, or cite the rule by name.

2. Step 1 — Design current: use the nameplate, not the power formula

The graphic uses the IEC adiabatic criterion, I²t ≤ k²S² , and then makes the most quantifiable error on the sheet.

The graphic states: "k = Constant (115 for Cu, 143 for Al)." Both numbers are copper. Per IEC 60364-4-43, Table 43A: copper PVC = 115, copper XLPE/EPR = 143, aluminium PVC = 76, aluminium XLPE/EPR = 94. The aluminium value is not 143 and does not appear on the sheet at all.

The example compounds it. The cable is specified as XLPE, 90 °C — but t he calculation uses 115, the PVC value. Correcting to k = 143 changes the answer from S ≥ 217 mm² to S ≥ 175 mm², a 24% overstatement of the required area. Conservative, so nobody gets hurt; wrong, so nobody should copy it.

The North American formula

US practice uses the ICEA P-32-382 form of the Onderdonk equation:

( I / A )² × t = K × log₁₀ [ ( T₂ + C ) / ( T₁ + C ) ]

with A in circular mils , I in amperes, t in seconds, and:

There is a well-circulated infographic on AC cable sizing for PV plants. It lays out four checks — ampacity, voltage drop, derating, short-circuit withstand — and works an example from a 350 kW inverter to a 300 mm² copper cable. The framework is right. The sequencing is right. It is a better starting point than most.

It is also written to IEC practice, and if you carry it onto a US project it will produce conductors that are the wrong material, the wrong size, and non-compliant for reasons the graphic never mentions.

This article rebuilds the whole metho d on NEC 2023 (with 2026 changes flagged), in AWG and kcmil, with aluminum conductors, at real US inverter voltages — and works two complete examples: a 240 A string-inverter AC output circuit and a 2,500 kVA central inverter block. Along the way we will be specific about what the popular version gets wrong, because two of its errors are the kind that survive design review and show up as a re-pull.

11. Anonymized case studies

Conductor material. US utility-scale PV AC collection is overwhelmingly aluminum , not copper. Aluminum is roughly twice as conductive per pound, and at commodity ratios near 4:1 the cost case is not close. AA-8000 series compact-stranded aluminum has been the code-recognized standard since 1972. A copper-only worked example is not representative of how these plants are actually built.

Conductor listing voltage

Most building wire is listed to 600 V. At 630, 660 or 690 V inverter output — all common on 1500 Vdc central inverters — you are above that and need conductors and terminations listed for 1000 V. Note that the NEC's low-voltage boundary moved from 600 V to 1000 V ac / 1500 V dc across the 2014–2017 cycles, and in the 2023 NEC the over-1000 V equipment rules live in the new Article 495. So 690 V does not put you into the medium-voltage chapters — but it does put you outside standard 600 V-listed wire. These are two different constraints and they are frequently conflated.

Article 691 applicability. NEC Article 691, "Large-Scale Photovoltaic (PV) Electric Supply Stations," applies to facilities with a generating capacity of not less than 5,000 kW and not under exclusive utility control. Generating capacity here is read as the sum of inverter continuous AC output at 40 °C, not DC nameplate. Above that threshold, engineered designs stamped by a licensed PE can substitute for certain prescriptive requirements — which changes what you have to prove and to whom.

Medium-voltage collector is a different article. The 34.5 kV collector system is governed by Article 315 in the 2023 NEC (Article 311 in 2020), with its own ampacity tables and its own 40 °C-based correc tion table. Do not carry Table 310.16 or Table 310.15(B)(1) into the collector design.

Conductor type on the AC side. XHHW-2 aluminum is the workhorse — 90 °C wet and dry, UL 44, tray-rated at 1/0 AWG and larger. PV Wire and USE-2 are DC-side conductors un der NEC 690.31(C); they do not belong on an AC circuit.

Terminations and torque

Aluminum terminations require listed connectors, an oxide-inhibiting compound where specified by the listing, and calibrated torque per 110.14(D). More aluminum conductor failures trace to the termination than to the conductor.

10. Design checks that catch real problems

1. IS 10262 is a concrete standard. The reference list cites "IS 10262" among the derating factor standards. *IS 10262 is the Bureau of Indian Standards guideline for Concrete Mix Proportioning. It has nothing to do with cables. The Indian cable standards are IS 1554 (PVC), IS 7098 (XLPE), and IS 3961* (current ratings). On a US project none of them apply anyway.

2. The k values are both copper. "115 for Cu, 143 for Al" — 115 is copper PVC, 143 is copper XLPE, aluminium XLPE is 94. Correcting the worked example from k = 115 to the appropriate XLPE value moves the answer from 217 mm² to 175 mm².

3. The derating product is 0.566, not 0.60.

4. Installation method is not a derating factor. It selects the table.

5. Soil thermal resistivity does not apply to a cable in a tray.

6. The 125% continuous-duty multiplier is missing. NEC 690.8(B)(1). This is the largest single omission for a US reader.

7. Terminal temperature limitation is missing. NEC 110.14(C). Reading 90 °C ampacity straight to a 75 °C lug is the most common ampacity error in US practice, and it is invisible until the thermal scan.

8. Design current should come off the nameplate. NEC 690.8(A)(3). And the PF = 1.0 assumption hides the reactive-power obligation most US plants carry.

9. Voltage drop is not a US code check. It is an informational note. Which means the real target is not 1.5% — it is whatever the 25-year loss econo mics and the inverter operating window justify.

What the graphic gets right, and should get credit for: the four-check framework in the correct order, t he full voltage drop formula including reactance, the reminder that ampacity tables are reference-only and manufacturer data governs, and the closing line — good cable sizing is about safety, efficiency, reliability and economy, not just current rating. That is exactly right, and it is the part most engineers skip.

Case Study 1 — Desert Southwest, 200 MWac: the 30 °C assumption

Situation. An EPC submitted the AC cable schedule for a 200 MWac single-axis-tracker project in southern Arizona. Conductor sizes had been carried over from a prior project of similar block size in the upper Midwest.

What the analysis found

Every inverter output circuit had been sized against Table 310.16 wit h no ambient correction applied. The 30 °C table basis had been treated as the site condition. Recorded design ambient at the site was 45 °C in the conduit environment, and above-grade conduit in direc t sun runs hotter still.

Applying Table 310.15(B)(1) at the 41–45 °C band — a 0.87 factor on the 90 °C column — moved a substantial fraction of the blocks below their required post-derating ampacity. Roughly a third of the AC circuits needed one size up. The conductors would not have failed immediately; they would have run continuously above their rated insulation temperature every summer afternoon, aging the insulation on a curve nobody was tracking.

What was done

The schedule was rebuilt with the correct ambient band, and the design ambient was pinned in the basis of design so it could not be silently inherited again. Circuits were reviewed for the 110.14(C) ceiling at the same time, which caught a separate set of blocks where the 90 °C column had been used at 75 °C lugs. Conductor cost rose about 4% on the AC scope. The change was made in design, before procurement.

Engineering takeaway

An ampacity table is a statement about a thermal environment, not about a conductor. A conductor schedule that travels between climates without its ambient assumption travelling with it is a latent defect.

Case Study 2 — Northeast, 20 MWac: the 90 °C conductor at the 75 °C lug

Situation. A 20 MWac project reached AHJ review with a complete, internally consistent AC design. The inspector flagged the inverter output conductors.

What the analysis found

The designer had specified XHHW-2 aluminum and sized every ci rcuit from the 90 °C column of Table 310.16 — correctly noting that XHHW-2 is a 90 °C wet-and-dry conductor. But all the AC combiner breakers, inverter terminals and switchgear lugs were listed for 75 °C terminations. Under NEC 110.14(C)(1)(b), circuits over 100 A must be sized from the 75 °C column.

The gap between the two columns is roughly 13% at these sizes — a 750 kcmil alu minum conductor is 435 A at 90 °C and 385 A at 75 °C. Across the plant, that put a meaningful number of circuits one size undersized against the termination limit.

What was done

The circuits were resize d against the 75 °C column, with the 90 °C column retained only as the derating basis. Because conductors were upsized for a code requirement under 110.14(C) rather than for voltage drop, 250.122(B) did not require the EGCs to be increased proportionally — a distinction worth several thousand dollars of copper on a plant this size, and one that is frequently applied incorrectly in the conservative di rection.

Outcome

The design cleared review on resubmittal. Because the issue was caught before the conductors were pulled, the cost was engineering hours rather than a re-pull.

Engineering takeaway

The conductor's temperature rating and the circuit's usable ampacity are two different things, and the equipment decides which one you get. If you cannot name the termination temperature rating of every device on a circuit, you cannot yet state its ampacity.

Case Study 3 — ERCOT, 150 MWac: sized in kilowatts, operated in kilovolt-amperes

Situation. A 150 MWac project completed detailed design with the AC collection sized from block real power at unity power factor. Late in the process the interconnection agreement was finalized, requiring the plant to maintain 0.95 leading to 0.95 lagging power factor at the point of interconnection across its operating range.

What the analysis found

The design current for every inverter output circuit had been computed as P / (√3 × V × 1.0). Meeting a 0. 95 power factor obligation at full real power output requires the inverters to be sized and operated on kVA, not kW — which raised continuous output current by a factor of 1/0.95, or 5.3%, above the value the cable schedule had been built on.

On its own, 5.3% is inside most conductors' margin. Stacked on top of the 125% continuous multiplier and a summer ambient corr ection, it pushed a set of circuits past their limit — and it also degraded voltage drop, because the reactance term that vanishes at unity power factor becomes significant at 0.95. On the longest runs, calculated drop went from about 1.2% to about 1.5%, tightening the inverter's AC operating window exactly when it was being asked to produce reactive power.

What was done

Design current was re-established from inverter nameplate continuous output current, per NEC 690.8(A)(3), rather than from a power calculation. Voltage drop was recomputed with the full R cos φ + X sin φ expression at 0.95 PF. Circuits on the longest runs were upsized one size — and because that upsize was for voltage drop rather than for derating, 250.122(B) did require the EGCs to be increased proportionally to circular mil area.

Outcome

The revision touched approximately 15% of the AC circuits and was absorbed in design. The reactive-power obligation was verified against the corrected operating window before the interconnection studies were finalized.

Engineering takeaway

An inverter is a kVA machine, and the interconnection agreement decides how much of that kVA is real power. Sizing conductors from kW at unity power factor builds a plant that is correct only on the days it is not asked to support voltage.

1. The four checks — and the one that is missing

4. Step 3 — Correction and adjustment: what counts, and what does no

The graphic starts with I = P / (√3 × V × PF). For a motor or a general load, fine. For a PV inverter, the NEC tells you not to.

NEC 690.8(A)(3) (2023): for the inverter output circuit, the maximum current shall be the inverter continuous output current rating. That is a value on the nameplate. You look it up; you do not derive it.

This is not pedantry. Three things go wrong when you compute it instead.

The power factor assumption is usually wrong

The graphic assumes PF = 1.0. Most US interconnection agreements require the plant to operate across a 0.95 leading to 0.95 lagging power factor range for voltage support. An inverter is a kVA-limited device: at 0.95 PF it still pushes its full rated current, it just delivers less real power. If you size the conductor from kW at unity power factor and the plant is later required to run at 0.95, your design curre nt was 5.3% low before you st arted. See Case Study 3.

The AC voltage is not a single number

A 4,200 kVA SMA Sunny Central UP-US is nominally 630 V but opera tes across 504–756 V. Current at the low end of that window is materially higher than current at nominal. The nameplate continu ous output current rating already accounts for this. Your formula does not.

Inverters derate on temperature

Continuous output at 50 °C ambient is not continuous output at 25 °C. The nameplate rating is defined; a back-calculation from DC nameplate is not.

Get the rating from the datasheet. If the datasheet is not final, do not finalize the cable schedule.

3. Step 2 — Ampacity: two tests, then a ceiling

The ceiling almost everyone forgets: NEC 110.14(C)

This is the most consequential difference between the graphic and US practice.

The graphic reads ampacity straight off a 90 °C XLPE table and derates from there. In the US you cannot do that, because NEC 110.14(C) requires conductor ampacity to be selected so as not to exceed the lowest temperature rating of any connected termination, conductor, or device.

Circuits rated 100 A or less, or marked for 14 AWG–1 AWG → 60 °C column, unless the equipment is listed and identified for higher
Circuits rated over 100 A, or marked for conductors larger than 1 AWG → 75 °C column

Virtually all utility-scale PV AC equipment — breakers, AC combiners, inverter terminals, switchgear lugs — is listed for 75 °C terminations. So:

You may use the 90 °C column as the basis for applying correction and adjustment factors. You may not end up above the 75 °C ampacity.

XHHW-2 aluminum is a 90 °C wet-and-dry conductor. You buy it for the derating headroom, not for the ampacity. A 750 kcmil aluminum condu ctor is a 435 A conductor for derating math and a 385 A conductor at the lug.

This single rule is why US conductor schedules run larger than an IEC-trained engineer expects, and why "but the cable is rated 90 °C" is the most common losing argument in a US design review.

The graphic multiplies five factors together:

Three separate problems, in increasing order of seriousness.

The arithmetic is wrong. 0.91 × 0.91 × 0.80 × 0.90 × 0.95 = 0.566, not 0.60 . A 6% error in the conservative direction, presented as an equality.

Installation method is not a multiplier. In-tray versus in-conduit versus direct-buried determines which table and which column you read. It is not a coefficient you apply on top of a value you already read from the correct table. Applying 0.91 for "in cable tray" to a tray ampacity is double-counting the same physics.

Soil thermal resistivity does not apply to cable in a tray. The example route is explicitly "In Cable Tray." Soil rho is a buried-cable parameter. This factor is simply from a different problem.

And the same double-count logic applies to spacing versus grouping: IEC 60364-5-52 grouping factors are already tabulated by arrangement, including touching. Multiplying a grouping factor by a separate "touching" factor applies the same derate twice.

What the NEC actually gives you

Two factors, and they are cleanly separated. Ambient correction — Table 310.15(B)(1) (for the 30 °C-based tables):

Two traps here. The NEC works in bands, not point values — 44 °C and 41 °C get the same factor. And there are two correction tables: 310.15(B)(1) for the 30 °C-based tables (310.16) and 310.15(B)(2) for the 40 °C-based free-air and medium-volt age tables. They are routinely confused online. Using the wrong one on a desert site is a real error with a real magnitude.

Adjustment for conductor count — 310.15(C)(1) , applies to more than three cur rent-carrying conductors:

The graphic frames voltage drop as a compliance test with a "Typical Limit: 1% to 1.5%" and a green checkmark. In the US that framing is wrong, and the correction is worth understanding because it changes what you optimize.

The NEC does not mandate a voltage drop limit for feeders or branch circuits. The familiar 3% and 5% figures appear in 210.19(A) Informational Note and 215.2(A) Informational Note. Per NEC 90.5(C), informational notes are explanatory and are not enforceable as requirements. No inspector can red-tag a feeder for 4% drop.

Voltage drop becomes enforceable only in specific places — through 110.3(B) where a listing or manufacturer's instructions specify a minimum operating voltage, and through 695.7 for fire pumps.

So why do we still care? Two real reasons, neither of which is compliance.

Inverter operating window. The inverter must stay inside its AC voltage range at full output, and the plant must meet its reactive-power obligation at the point of interconnection. Excessive drop between the inverter and the LV winding of the pad-mount eats into that headroom.

Energy yield. Voltage drop is I²R loss, and in a PV plant that loss runs for 25 years. Design practice target s around 1% total AC cabling loss, with most designs aiming to stay under 1.5%.

In a US PV plant, the code sets the minimum conductor; the loss economics usually set the actual conductor, and it is frequently one or two sizes larger. That is an engineering-economics decision, not a code compliance box.

And do not drop the reactance term

The graphic's formula is correct and better than most:

ΔV = √3 × I × L × ( R cos φ + X sin φ )

But its worked example sets PF = 1.0, which makes sin φ = 0 and quietly deletes the reactance term. At 0.95 PF — the condition your interconnection agreeme nt is likely to require — the reactance term is real. In Example A below, including it takes voltage drop from 1.18% to 1.47%, a 25% increase, on the same conductor at the same current.

Use NEC Chapter 9, Table 9 for AC resistance and reactance at 75 °C in your actu al raceway type. Values differ between PVC, aluminum and steel conduit, and steel is materially worse.

For a three-phase, three-wire inverter output circuit in its own conduit, there are exactly three current-carrying conducto rs and no adjustment applies. The EGC is not a current-carrying conductor.

Cable tray is its own rule set

If you are in tray — and in utility-scale PV you often are — the governing section is NEC 392.80, not 310.15(C)(1).

392.80(A)(1), multiconductor cables: base ampacity from Table 310.16. 310.15(C)(1) adjustment applies only to cables containin g more than three current-carrying conductors, counted per cable — not per tray population. A solid unventilated cover longer than 6 ft imposes 95% .
392.80(A)(2), single-conductor cables: base ampac ity from Table 310.17 (free air), then in uncovered tray, 75% for 600 kcmil and larger, 65% for 1/0 AWG throug h 500 kcmil. Single layer, uncovered, with one conductor diameter of maintained spacing, 1/0 and larg er: full free-air ampacity.

The tray factors replace the conduit-fill adjustment. Do not stack both. T his is the tray version of the same double-counting error the graphic makes.

5. Step 4 — Voltage drop: not a code check in the United States

Typical temperature s: T₁ = 90 °C (XLPE/EPR continuous rating), T₂ = 250 °C (XLPE/EPR short-circuit limit).

Both forms are adiabatic — they assume no heat esca pes into the insulation. That is conservative below roughly 5–10 seconds and increasingly inaccurate above it. Neither should be used for long-duration backup clearing times without a thermal model.

Where the fault current actually comes from

The graphic gives a 25 kA prospective fault current with no source identified, and this matters more in a PV plant than almost anywhere else.

A grid-following PV inverter is a current-limited source. IEEE PES-TR67.r1 models inverter fault contribution at approximately 1.2 per unit; Sandia puts grid-following inverters at 1.2–1.5 p.u. and grid-forming inverters at 2–3 p.u. A 2,400 A inverter contributes on the order of 2,900 A into a bolted fault — not 25 kA.

The 25 kA comes from the grid, backfeeding through the medium-voltage step-up transformer. So the fault current on your inverter output circuit is set by the utility sou rce impedance and the GSU impedance, not by the inverter. Two consequences:

Get the number from th e system short-circuit study, not from the inverter datasheet.
Use the actual protective device clearing time for that fault path, including relay time plus breaker interrupting time. The graphic's t = 1 s is a placeho lder; real clearing on a modern AC combiner breaker is closer to 0.1–0.3 s, and the required area scales with √t.

7. Worked Example A — 240 A string-inverter AC output circuit

System. Three-phase, 600 V AC, string inverter with a nameplate continuous output current rating of 240 A. XHHW-2 aluminum in PVC conduit, one circuit per conduit, 400 ft one-way, desert site with 45 °C design ambient, 75 °C-rated terminations at both ends.

Step 1 — Design current

240 A, read from the nameplate per NEC 690.8(A)(3). Not calculated.

Step 2 — Ampacity, test one: 690.8(B)(1)

Required ampacity before correction and adjustment = 1.25 × 240 = 300 A. Because terminations are 75 °C and the circuit is over 100 A, this test is read against the 75 °C column:

500 kcmil Al, Table 310.16, 75 °C = 310 A ≥ 300 A — passes.

Step 3 — Ampacity, test two: 690.8(B)(2)

Required ampacity after correction and adjustment ≥ 240 A. Here we may use the 90 °C column as the derating basis, because the conductor is XHHW-2.

500 kcmil Al, 90 °C = 350 A
Ambient 45 °C, Table 310.15(B)(1), 41–45 °C band, 90 °C conductor: × 0.87
Conduit fill: three current-carrying conductors → no adjustment

350 × 0.87 = 304.5 A ≥ 240 A — passes.

Now apply the 110.14(C) ceiling. The corrected value, 304.5 A, is compared against the 75 °C termination limit of 310 A. It is lower, so the effective ampacity is 304.5 A. Had the ambient been milder, the answer would have been capped at 310 A regardless of what the 90 °C column said.

Step 4 — Overcurrent protection

The OCPD must be at least 125% of continuous current (300 A) and must not exceed conductor ampacity (304.5 A). A 300 A device satisfies both. Per 240.4(B), the next standard size up would be permitted where the ampacity does not land on a standard rating, up to 800 A — not needed here.

Step 5 — Voltage drop

500 kcmil aluminum in PVC conduit, NEC Chapter 9 Table 9: R ≈ 0.0424 Ω/1000 ft, X ≈ 0.041 Ω/1000 ft at 75 °C. (Sanity check from first principles: ρ_Al ≈ 17.0 Ω·cmil/ft at 20 °C gives 0.0340 Ω/kft, corrected to 75 °C gives 0.0415, plus stranding and skin effect ≈ 0.042. The table value is sound.)

At unity power factor: ΔV = √3 × 240 × 0.400 × 0.0424 = 7.05 V = 1.18%

At 0.95 power factor, with the reactance term included: ΔV = √3 × 240 × 0.400 × (0.0424 × 0.95 + 0.041 × 0.312) = 8.83 V = 1.47%

Both acceptable. Note that ignoring reactance would have understated the answer by 25%.

Step 6 — Short-circuit withstand

Available fault current at the AC combiner from the system study: 25 kA, clearing in 0.2 s. Aluminum, T₁ = 90 °C, T₂ = 250 °C:

(I/A)² × 0.2 = 0.0125 × log₁₀(478/318) = 0.0022125 → A ≥ 25,000 / √(0.0022125 / 0.2) = 237,700 cmil ≈ 238 kcmil

500 kcmil provides 500,000 cmil — better than 2× margin.

Cross-check. 237,700 cmil is 120.4 mm². The IEC method on the same problem gives S ≥ 25,000 × √0.2 / 94 = 118.9 mm². The two methods agree within 1.3% — which is what you would expect, since they are the same adiabatic physics expressed in different units. If your ICEA and IEC answers disagree by more than a few percent, you have used the wrong k or the wrong constants.

Step 7 — Equipment grounding conductor

Per Table 250.122, a 300 A OCPD requires 4 AWG copper or 2 AWG aluminum.

If the ungrounded conductors are la ter upsized for voltage drop, 250.122(B) requires the EGC to be increased proportionally to the increase in circular mil area — not by AWG steps, and not proportionally to ampacity. Note the nuance: upsizing to compensate for ambient or fill derating under 310.15(B)/(C) does not trigger 250.122(B). Upsizing for vo ltage drop does.

Result

8. Worked Example B — 2,500 kVA central inverter block

System. 2,500 kVA central inverter, 600 V AC, feeding the LV winding of a pad-mounted step-up transformer. XHHW-2 aluminum, parallel sets in individual PVC conduits, 40 °C design ambient, 75 °C terminations.

Nameplate continuous output current: 2,500,000 / (√3 × 600) = 2,406 A
690.8(B)(1): 1.25 × 2,406 = 3,007 A required before derating
Conductor: 750 kcmil Al — Table 310.16: 75 °C = 385 A, 90 °C = 435 A
Derating: 40 °C ambient, 36–40 band, 90 °C conductor → 435 × 0.91 = 395.9 A. Three current-carrying conductors per conduit → no fill adjustment.
110.14(C) ceiling: 395.9 A exceeds the 75 °C limit of 385 A, so the effective ampacity is 385 A per conductor.
Sets required: 3,007 / 385 = 7.81 → 8 parallel sets
Verify: 8 × 385 = 3,080 A ≥ 3,007 A and ≥ 2,406 A
Overcurrent protection: a 3,000 A device. Per 240.4(C), for OCPDs rated over 800 A the conductor ampacity must be equal to or greater than the device rating — there is no next-size-up allowance. 3,080 A ≥ 3,000 A.

Parallel conductor rules — NEC 310.10(G). Conductors 1/0 AWG and larger may be paralleled, but every set must be the same length, material, size, insulation type, and terminated in the same manner. On a 2,500 kVA block that is eight sets of three conductors plus EGCs — 24 phase conductors that must all match. Length mismatch between sets causes current sharing imbalance, which is a thermal problem long before it is an electrical one.

Reality check. A published US design for a 2,500 kVA inverter block la nded on nine sets of 750 kcmil aluminum at 75 °C. Our eight-set answer differs because of ambient and voltage assumptions — which is exactly the point. The method is deterministic; the inputs are where projects diverge. Document your ambient, your terminations, and your inverter nameplate, and th e review is short.

Figure 1 , on the following page, summarizes the whole method: the circuit and which rule governs each segment, the three ampacity rules, the eight-step sequence, both worked examples, the NEC tables, and the errors to avoid.

FIGURE 1 — AC cable sizing for utility-scale solar PV, US / NEC method: circuit segments, the three ampacity rules, sizing sequence, worked examples and tables

12. Frequently asked questions

13. Standards reference

Closing

The four-check framework in that infographic is sound, and anyone using it is ahead of an engineer who sizes conductors by ampacity alone. The problem is not the method. The problem is that the method was assembled from one code family and is being applied in another, and the two do not disagree politely — they disagree by 25% on the continuous-duty multiplier and by 13% at the lug.

On a US utility-scale PV project the conductor schedule is one of the largest single material line items on the AC scope, and it is also one of the least forgiving. Undersize it and you get insulation aging you cannot see and losses you pay for every day for 25 years. Oversize it uniformly and you have spent money that had better returns elsewhere in the plant.

At Keentel Engineering , AC and DC collection design, cable schedules, short-circuit and arc-flash studies, and POI interconnection engineering are delivered together, because they are the same problem viewed from different angles. If you have a cable schedule to review, an ambient assumption you inherited, or an interconnection requirement that just changed your design current, we would be glad to look at it.

KEENTEL ENGINEERING

Tam pa, FL · Austin, TX · Sacramento, CA · Baltimore, MD

keentelengineering.com · contact@keentelengineering.com · 813-389-7871

Engineering note: NEC section numbers are given for the 2023 edition unless otherwise stated, with 2020 and 2026 differences flagged where they exist. Confirm against the edition adopted in your jurisdiction. Ampacity values are from NEC Table 310.16 and are reproduced for illustration; design from the code book. Nothing here substitutes for a project-specific engineering design.

9. Nine things worth correcting in the popular guides

US ampacity is a three-part exercise, and the popular guide does only the middle part.

The two-part 690.8(B) test

690.8(B)(1) — ampacity, before adjustment and correction factors, ≥ 125% × maximum current
690.8(B)(2) — ampacity, after adjustment and correction factors, ≥ maximum current

You compute both and take the larger conductor. On a hot site with heavy derating, (B)(2) can govern. On a cool site in a single conduit, (B)(1) governs. Neither one alone is the answer.

Which table

Table 310.16 — "Ampacities of Insulated Conductors with Not More Than Three Current-Carrying Conductors in Raceway, Cable, or Earth (Directly Buried)," based on 30 °C (86 °F) ambient.

Note the renumbering: Article 310 was reorganized in the 2020 NEC. What older references call Table 310.15(B)(16) is now Table 310.16; 310.15(B)(17) is now 310.17 (free air). Medium-voltage conductors moved out of Article 310 entirely — to Article 311 in 2020, and then to Article 315, "Medium Voltage Conductors and Cable," in the 2023 NEC. Your collector system lives there, not in 310.

Working values for utility-scale PV, Table 310.16, three current-carrying conductors, 30 °C:

NERC Data Center Rules

Sat, 22 Aug 2026 08:26:01 GMT

Aug 21, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

The Grid Is Coming for Your Data Center

Inside NERC's Large Loads Action Plan — and what computational load operators must do before December 31, 2026

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

For twenty years, NERC regulated the supply side. Generators, transmission owners, balancing authorities, reliability coordinators — the entities that make and move electricity — carried mandatory, federally enforceable obligations. The entities that consume electricity did not. Load was something the grid planned around, not something the grid regulated.

On July 16, 2026, the Federal Energy Regulatory Commission issued an order in Docket No. RD26-7-000 directing NERC to file registry criteria and new Reliability Standards for computational loads by December 31, 2026, with a follow-on work plan due March 1, 2027. NERC had already been moving that way under its Large Loads Action Plan. FERC converted a voluntary schedule into a binding one — and pulled it forward. NERC's own filing had contemplated registry criteria in Q1 2027 with standards drafted across 2027. FERC compressed that by roughly a year.

As of August 19, 2026, two things sit in open comment periods closing September 18:

Proposed Rules of Procedure revisions creating two new registered entity types — Computational Load Owner and Computational Load Operator — via changes to Appendices 2, 5A and 5B
Three proposed foundational Reliability Standards — CLO-001-1, CLO-002-1 and CLO-003-1 — plus limited conforming updates to FAC-001-5 and FAC-002-5

If your facility clears the thresholds, it becomes a NERC registered entity: mandatory requirements, compliance audits, and civil penalty exposure at $1,584,648 per violation per day.

This is the most consequential regulatory development for the data center sector since interconnection queue reform. Here is how it happened, what the primary documents actually say, and what to do about it.

Regulators do not create new registered entity classes on theory. NERC built this case over three years across incident reviews, a mandatory industry survey, and two technical white papers.

The events

July 10, 2024 — Eastern Interconnection. A lightning arrestor failure on a 230 kV line triggered a sequence of faults over 82 seconds in the Virginia data center corridor. Voltage depressed to between 0.25 and 0.40 per unit. Approximately 1,500 MW of data center load disconnected — and critically, not because a utility shed it. It disconnected because customer-side protection and controls decided, autonomously and nearly simultaneously, to drop the grid and transfer to backup power.

Frequency rose to 60.053 Hz and took roughly four minutes to settle.

That is a survivable outcome in the Eastern Interconnection, which is enormous. NERC's analysis noted that an equivalent 1,500 MW loss in ERCOT would produce a frequency excursion on the order of 235 mHz rather than the 53 mHz observed. The event was a warning shot fired in the interconnection best equipped to absorb it.

January 2023 – September 2025 — ERCOT. NERC's January 2026 incident review catalogued 26 ride-through events involving indirect load loss greater than 100 MW from crypto mining facilities, concentrated in Central Texas, Far West Texas, the Panhandle and the North Zone. Facilities shed between 17% and 95% of pre-disturbance consumption; three-phase faults averaged around 65%.

The causes were unglamorous and fixable: constant-power electronic supplies that collapse within milliseconds of a voltage sag; facilities shedding significant load when single-phase voltage fell below roughly 0.7 per unit, tracking the ITIC curve dropout boundary; single-phase depressions triggering neutral overcurrent protection across an entire site; cooling variable-frequency drives with undervoltage trip thresholds set as high as 0.96 per unit — one of which shut down an entire facility on a 50-millisecond voltage depression during a normally cleared fault; and wye-wye transformer configurations that passed sags straight through.

NERC's gap assessment puts the behavioral finding more bluntly than any incident report does. These facilities, it says, "have demonstrated tendencies to disconnect for faults outside of their zone of protection — or more simply for faults on lines they are not connected to — which is unexpected. This type of behavior is not modeled in operational or operations planning analyses."

What Changed, and When It Lands

If you are siting, this matters. On nearly every measure, TRE (ERCOT) and ReliabilityFirst are the most prepared and NPCC is the least. NPCC scored 2.0% on design criteria, 2.0% on commissioning, 0.0% on modeling-update processes, 2.0% on operating protocols, 0.0% on disconnect/reconnect limits, and 7.1% on pre-energization coordination. ERCOT scored 30.0%, 16.7%, 20.0%, 23.3%, 15.0% and above on the same measures.

The corollary: mature framework means more will be asked of you up front, and fewer surprises later. Immature framework means an easier interconnection and a much higher chance that requirements arrive retroactively.

NERC describes its own effort as three concurrent workstreams: Registration, Reliability Standards, and Technical Insights. It is worth using NERC's framing, because the three run on different clocks and carry different legal weight.

Part Three: The Regulatory Machinery

Here is a fact that should reorganize how you think about your relationship with the grid.

NERC asked utilities to submit the dynamic models they use to represent data centers in grid stability studies. The responses covered 33,282 MW of currently operational data center load. For 10,536 MW, the utility answered "None" when asked what protection-device model represented the facility. For another 14,968 MW, the answer was blank.

That means 25,504 MW — roughly 77% of the operational data center fleet those utilities serve — is represented in the models that determine whether the grid stays up as a static block of impedance. No representation of protection. No representation of the facility's ability to trip. No representation of what happens when voltage sags.

Only 137 MW — four tenths of one percent — used a modern power-electronic load model.

NERC's own summary: "Roughly three quarters of all data center load models are insufficient to represent data center dynamic behavior."

The planners deciding whether your region needs another transmission line, and the operators deciding what contingencies to hold reserves against, largely do not know your facility can disappear. That is the problem NERC is now solving, and it is solving it by making data centers regulated entities.

Part Two: How Unprepared Everyone Is

In September 2025, NERC issued a Level 2 Alert — a mandatory data request to registered entities on large load interconnection, study, commissioning and operations. It ran 105 questions with a mandatory data spreadsheet, with responses due January 28, 2026. NERC published the aggregated results in March 2026.

If you read one document in this corpus, read that one. It is the empirical foundation for everything that followed, and its findings are not what a data center operator would expect.

The demand picture

And the finding that should stop you: 37% of Transmission Planners and Planning Coordinators reported having large loads on their system whose transient dynamic performance — including voltage and frequency ride-through characteristics — is simply unknown. In ReliabilityFirst the figure is 57.1%. Among Planning Coordinators that are not also Transmission Planners, it is two-thirds.

NERC also found that "nearly two-thirds of all submitting entities indicated a lack of experience in integrating modern data centers or other large loads."

The data problem, from the utility side

The physics

Composition. IT equipment is 60–95% of a data center's demand, essentially all of it behind power electronic converters. Cooling adds VFDs and inverters. Crypto runs on ASICs. Hydrogen electrolysis — also in scope for the broader large-load discussion — can be over 85% converter load. Legacy industrial load was dominated by induction motors with mechanical inertia and graceful degradation. This class has neither.

Ramp rate. NERC documented AI training facilities changing demand at 1.9 per unit per second sustained for roughly 250 milliseconds, measured on a 50 MW block inside a 200 MW facility. One North American data center ramped 450 MW down to 40 MW in 36 seconds, held near 7 MW for four hours, then walked back to 450 MW over minutes. A crypto facility dropped 298 MW in 25 seconds after a telecom failure, with residual oscillations of roughly 25 MW peak to peak.

The AI training profile. The gap assessment reproduces a modelled AI training load shape with real numbers: 32 ms rise from base to peak, a 736 ms plateau, a 32 ms fall, and 100 ms at base before the cycle repeats. That is a square wave with millisecond edges, cycling continuously. NERC notes that the best available positive-sequence load model — EPRI's aggregated EV-charger model — explicitly cannot replicate cyclic load injection of this kind.

Oscillation. Large loads can source forced oscillations across the 0.1 Hz to 30 Hz range, either from unintended control interactions or as a byproduct of the workload itself. A documented real-world case involved 14.7 Hz oscillations emerging from a data center. Anything above 5 Hz requires EMT modeling to evaluate — and nearly no utility performs EMT studies when integrating a large load.

Subharmonics. Prior IEEE work suggests subharmonic voltages should stay below 0.1% of nominal. NERC observes that "some AI workloads may produce subharmonics that are tens or hundreds of megawatts in magnitude," and that reducing them "is sometimes beyond the capability of traditional solutions such as static VAR compensators." Battery storage with grid-forming inverters can cut subharmonic current magnitude by roughly 70%.

Voltage sensitivity. Constant-power supplies draw more current as voltage falls — the opposite of the impedance behavior grid stability analysis historically assumed. Combined with protection designed to protect silicon rather than support the grid, you get load that vanishes precisely when the system most needs it.

Reactive behavior. Active front-end electronics operate near unity displacement power factor, but their input filters generate reactive power, making large facilities net capacitive. That forces nearby synchronous generators into reactive absorption, raising transfer angles and eroding angular stability margin.

Scale. As of April 2025, ERCOT's interconnection queue held roughly 136 GW of large load with energization dates through 2030 — against a system historic peak near 85 GW.

Part Four: Registration

The threshold test

NERC's April 2026 posting proposed a three-part conjunctive test. An entity registers if it:

Is an end-user, or an entity hosting end-users, receiving electric power for Computational Load — "load comprised of electric power demand from information technology equipment, such as servers, storage, and networking hardware"; and
Hosts 1 MW or greater of computational load; and
Contributes to an aggregate connected load capability greater than or equal to 20 MW at a single point of interconnection to the BPS at a voltage greater than or equal to 60 kV

Verify these against the currently posted Appendix 5B redlines. These values come from the April posting; the August 19 revision restructures the function into Owner and Operator roles, and thresholds are exactly the kind of parameter that moves between postings.

The 20 MW figure is not arbitrary. The Level 2 report found that "the 20 MW and above threshold captures roughly 90–95% of electrical size of all large load facilities," and that TO-defined thresholds clustered above 20 MW while DP thresholds fell below it.

Part One: The Evidence Base

"Forecasts collected from the worksheet information indicate that load may increase by 300 GW between 2028 and 2030, with the bulk of this increase represented by data centers."

Use that number with NERC's own caveat attached: it says confidence in the 2028-and-beyond values "is questionable due to the differences associated with entity interpretations on the word 'forecast'" — some respondents gave maximum projected load, others a 50/50 confidence value. NERC treats its own headline number as evidence of a forecasting gap, not as a forecast.

The composition shift is starker than the growth. Of in-service large load at end 2025, 49% was industrial and 42% was data center or other computational. By end 2027, on the same respondents' projections, computational load is 90% and industrial load is 8%. Within the computational slice, AI data centers go from 3% to 17%, and facilities combining traditional, AI and crypto workloads go from 24% to 47%. Crypto falls from 14% to 8% of a much larger base — diluted, not displaced.

Geographically, growth concentrates in the East North Central, South Atlantic and West South Central census zones — the Ohio Valley, Northern Virginia, Atlanta, and the Dallas–Fort Worth/central Texas corridor.

Two things you would not expect

Sources

Part Five: The Alert and the Standards

The framing NERC has adopted

Across the gap assessment, one comparison recurs deliberately: "it is arguable that these large loads can operate as controllable Bulk Electric System elements that can significantly ramp up and down with high oscillatory behavior… Their size, controllability, and variability introduce potential reliability risks to the BPS comparable to those of a generator but with the added complexity of being on the demand side."

Elsewhere: "the sudden addition or loss of large loads can have equivalent reliability impacts as generation contingencies." And: "emerging large loads are comparable to the size of significantly large generators… which create challenges for having a third party be responsible for their performance."

That last clause is the whole argument for registration, compressed into a sentence.

The readiness picture

These are percentages of responding utilities that have the practice in question.

What "nearly all" means

The report's most damning findings are qualitative, because the answer was near-unanimous:

"Most responses indicated that no model verification is performed against on-site settings to verify the as-built model of the load facility."
"Nearly all entities do not validate performance after energization."
"Almost all of the entities indicated that they did not have a process to review performance and modeling during a multi-year load growth for large load projects." Phases 2 through 6 of your campus are, in practice, unstudied.
"Nearly all entities indicated that no EMT study was performed when integrating large loads."
On post-disturbance recovery requirements, protection coordination requirements, and oscillation mitigation: "Nearly all entities responding to this question indicated that there is not a process to coordinate" — three separate questions, three identical answers.
"Nearly all entities indicated that large loads are not included in automatic load-shedding programs."

And on the model that most confident planners rely on: "the composite load model's structure is unable to capture the key features of computational load performance." NERC observed active disagreement — in one region "most of the TPs claimed that their models were accurate enough, but the PC for that area stated that most models are not" when compared against real system events.

The numbers utilities are actually using

This thread is the clearest explanation of why registration is happening.

Utilities largely cannot get models from you. One TO in a high-penetration area reported that "out of five entities contacted for a dynamic model, only one was able to provide a dynamic model."
Utilities say they have no authority to ask. "Some entities reported that they could not gather this information because they lacked the authority to do so." And: "if an entity was unable to provide a model, a study could not be performed."
Utilities mostly did not even try. "Roughly all (~85–90%) respondents indicated that they did not coordinate with large load entities when responding to this Alert" — which NERC flagged as "counter to the expectation."
Contact channels are broken. Utilities reported "erroneous contact information" and no response "through their typical communication methods."
Recorders exist but data doesn't flow. "Many indicated that these recordings are not accessible for the utility but that the device is present"; where the recorder is owned by the load, "their only method was a formal information request that may not be successful."
Your peers are on record as exhausted. "Some reported that their large load entities are fatigued by the multitude of data requests."
Some utilities have given up and gone to orbit. "A few entities use alternative methods like satellite imagery to validate the data center's progress for their demand projections."

The gap assessment adds the commercial explanation, and names names: "Data center operators are inherently protective of their operational data due to commercial sensitivity, cyber security concerns, and competitive pressures… Operators of cryptocurrency mining facilities in particular may avoid sharing detailed operational data due to market competitiveness and strategic responses to electricity price fluctuations."

It also identifies the structural problem behind speculative development: "The large load developer can be different than the entity that owns or operates the computational equipment. The developer might not have information on the end-user of the facility until late in the interconnection process. This can lead to gaps in the utility's ability to accurately study the proposed facility."

Where requirements exist, the Level 2 report captures what they look like — useful benchmarks for negotiating your own interconnection agreement:

"Large load" definitions ranged from 100 kW to 400 MW, with clear clustering at 20, 50 and 75 MW. NERC's finding: "the 20 MW and above threshold captures roughly 90–95% of electrical size of all large load facilities." That sentence is why the proposed registration threshold is 20 MW.
Ramp limits, where imposed, ranged 8 MW/min to 300 MW/min, with most between 10 and 30 MW/min. NERC notes this "is not a normal requirement for load."
Oscillation limits: "requirements are set to limit the peak-to-peak measurement of the active power oscillation to 25 MW," and to "not inject at known system modes." Some entities set a minimum damping ratio.
Maximum load loss from a single contingency: values "ranged from 300 MW to 2,600 MW," with a plurality at 300 MW. One ISO/RTO structure uses two tiers — "1,000 MW per Point of Interconnection" for TPL-001 contingency purposes and "2,600 MW of aggregate loss" as a frequency-stability system operating limit.
Disturbance monitoring is sometimes triggered by a "75 MW or more behind the delivery point" threshold. ERCOT's Nodal Operating Guide Section 6.1 requires disturbance monitoring equipment for large loads at 75 MW or above on request — and NERC notes "since this requirement does not exist in NERC standards, there is a potential gap."
34.5% of utilities with large load interconnection requirements require no electrical recording device at all. In ReliabilityFirst that figure is 76.9%.

Some utilities are running data center interconnection through generator interconnection machinery: "Some entities explained that they mirror the Small Generator Interconnection Agreement or Large Generator Interconnection Agreement processes to study these loads due to their complexity."

And at least one grid operator already has a direct kill switch: "One entity mentioned that the RC has telemetry set up so that they can issue curtailment commands directly to the load."

Regional readiness varies enormously

The sequence

That it "is intended to supplement, rather than replace, any existing or new Reliability Standards." Guidelines still matter: they establish what a reasonable operator should have known.

May 4, 2026 — Level 3 Essential Action Alert. NERC's most severe alert level, requiring Board approval, used sparingly. Acknowledgement due May 11; full responses due August 3, 2026.July 16, 2026 — FERC Order, RD26-7-000. Under Section 215 of the Federal Power Act, FERC directed NERC to file new or modified Reliability Standards, Glossary changes, and Rules of Procedure revisions including registry criteria by December 31, 2026, and a work plan for further revisions by March 1, 2027. FERC cited data center consumption projections of 6.7%–12% of U.S. electricity by 2028, the July 2024 Virginia event, and the ERCOT record.

2025 – Q1 2026 — Technical foundation. The Large Loads Task Force was promoted to the Large Loads Working Group by the RSTC in December 2025. It produced Characteristics and Risks of Emerging Large Loads and Assessment of Gaps in Existing Practices, Requirements, and Reliability Standards for Emerging Large Loads , both March 2026. The gap assessment's contributor list runs to ERCOT, EPRI, AEP, Dominion, Duke, Exelon, Hydro-Québec, IESO, MISO, NYISO, SPP, PNNL, Oak Ridge, MITRE and others — with one named hyperscaler participant, from Microsoft. That imbalance is itself worth noticing.

November 21, 2025. NERC filed comments in FERC's ANOPR on large load interconnection reform, Docket RM26-4.March 16, 2026. NERC launched Project 2026-02. The SAR was posted for comment April 1 with comments due April 30; the Standards Committee approved drafting on May 20, 2026.

March 20, 2026. NERC supplemented its accelerated Large Loads Action Plan filing in RM26-4.

April 30 / May 2026 — Reliability Guideline. Risk Mitigation for Emerging Large Loads . Non-binding, and NERC is explicitAugust 19, 2026. Registry criteria and three draft standards posted. Comments close September 18.

On the speed

This is extraordinarily fast for NERC. Standards development normally runs three to five years through multiple ballot rounds. NERC is attempting registry criteria plus three new standards in roughly ten months — and FERC pulled the registry timeline forward from NERC's own Q1 2027 target.

NERC's answer to the process objection is deliberate: the proposed CLO standards are built from requirements already existing in other Reliability Standards that were previously compliance-vetted and industry-supported. NERC is not inventing obligations; it is porting proven ones onto a new functional entity.

It also means the genuinely novel questions — enforceable ride-through curves for load, model quality standards, ramp-rate limits — are largely deferred to Phase 2. Phase 1 establishes the entity and the plumbing. Anyone reading Phase 1 and concluding this is manageable should understand they are reading the easy half.

Owner versus Operator

The April posting proposed a single "Computational Load Entity." The August 19 posting splits it into Computational Load Owner and Computational Load Operator, revising Appendices 2, 5A and 5B.

That split responds to the loudest objection to the April draft: in a colocation or build-to-suit arrangement, who is the regulated party? It mirrors the Generator Owner / Generator Operator architecture, letting facility obligations (models, protection settings, as-built configuration) be allocated separately from operational obligations (real-time communication, responding to operating instructions, data provision).

Whether it resolves multi-tenant colocation — where the landlord owns the electrical infrastructure and dozens of tenants independently drive the ramps — is the single most important thing to comment on before September 18.

Three ways NERC could have drawn the line

The gap assessment contains a passage that almost no summary covers, and it explains why the criteria look the way they do. NERC considered three candidate scopes for registration:

All large loads register
Power electronic-interfaced loads register
Loads with a significant amount of IT equipment load register

Its own observation: "registration of all large loads or registration of power electronic-interfaced loads would include other large loads in addition to data centers or cryptocurrency mining facilities." Option 2 would sweep in every VFD-driven industrial motor load in North America. Option 3 — significant IT equipment load — is the only one that targets data centers and crypto without capturing manufacturing. That is the option the proposed criteria implement, which is why the test keys on IT equipment specifically rather than on power electronics or megawatts alone.

The fourteen functions

Recommendation 1 of the gap assessment is the closest thing that exists to a job description for a registered computational load entity. NERC recommends registering an entity type able to do the following:

Provide accurate short-term demand forecast data
Provide sufficient data to RC, TOP and BA for short-term demand forecasts and operating plans
Provide accurate load model data and ongoing model updates
Inform the TO of pertinent changes to load characteristics before and after energization
Perform real-time and operations-planning coordination with TOP, BA and RC
Comply with operating instructions from TOP, BA and RC, and ensure appropriate training for receiving those instructions
Ensure compliance with disturbance ride-through requirements
Ensure compliance with ramp rate requirements (down ramp and up ramp)
Provide accurate dynamic model information to support interconnection and transmission planning studies, as early as possible
Provide accurate information for interconnection studies
Coordinate to establish a comprehensive commissioning process ensuring operational readiness
Adhere to operating and communication protocols
Ensure compliance with performance requirements to minimize and mitigate unintentional power oscillation interaction during normal and post-event conditions
Provide the utility with a site vulnerability/risk assessment and mitigate issues discovered
Notify the utility, close to real time, of security breaches

Read that list as the medium-term destination. Phase 1 will not implement all of it. Phase 2 is scoped to.

Why NERC chose standards over interconnection requirements

The gap assessment gives four reasons, and they are worth understanding because they define how much leverage contract negotiation will retain:

"interconnecting utilities generally do not have the authority to assess performance-based financial penalties except under certain very limited circumstances"
"NERC standards provide consistent performance requirements across the BPS, while interconnection requirements can vary between TOs or regions"
"if large loads… do not have NERC standards directly applicable to them, the existing registered entities will then be responsible for the performance of the large loads. While this may have been appropriate for historical loads of significant size, emerging large loads are comparable to the size of significantly large generators… which create challenges for having a third party be responsible for their performance"
"enforcement of interconnection procedure/agreement terms and conditions may vary, while enforcement mechanisms that exist for the NERC standards provide clarity"

NERC also left itself an explicit exit: "if interconnection requirements are seen as being more appropriate for addressing specific gaps, the LLWG's opinion in this area may change and this may reduce the need for registration of these new entities." That sentence is the strongest textual hook available to anyone arguing against registration in the comment record. It is also, at this point, a narrow one — FERC has since ordered the registry criteria filed.

The Load-Serving Entity backstory

The gap assessment returns four times to a piece of history worth knowing: the Load-Serving Entity function was removed from the NERC Rules of Procedure roughly ten years ago under risk-based registration reform. The paper repeatedly asks "whether the BPS impacts associated with large loads might warrant reconsidering reactivation of the LSE function."

NERC ultimately went a different route — a purpose-built computational load function rather than a revived LSE. But the LSE thread explains the structural diagnosis: "neither large loads nor LSEs are registered entity functions under the NERC registry criteria," and therefore "the existing Reliability Standards and interconnection requirements do not adequately mitigate large loads' reliability impacts on the BPS."

What registration actually means

Mandatory, enforceable requirements applying to you directly by function, as a matter of federal law
CMEP exposure — audits, spot checks, self-certifications, self-reporting, mitigation plans
Civil penalties up to $1,584,648 per violation per day (the 2025 level, carried into 2026 after OMB cancelled the annual inflation adjustment). Penalties scale with reliability harm under FERC's Penalty Guidelines; a documented compliance program is a genuine mitigating factor
Section 1600 data requests. Registered entities must respond to NERC information requests and can be required to develop corrective action plans. Today you cannot be compelled — which the gap assessment identifies as actively impairing event analysis: "Multiple events in the Eastern Interconnection and ERCOT have involved large loads reducing significant amounts of load during voltage disturbances… The significant gaps… can impair NERC and the Regional Entities from obtaining the needed information directly from the large load entities to perform thorough root-cause analysis"
A standing compliance function — evidence retention, documented procedures, designated contacts, training records
Public enforcement. Notices of Penalty are published. Expect registration status and compliance representations to migrate into colocation agreements, ESAs and financing covenants.

Registration of new computational load entities begins only after FERC approves the criteria. The filing deadline is December 31, 2026; approval and implementation follow.

Part Six: What Phase 2 Actually Reaches

The Level 3 Alert is your preview

The seven Essential Actions were addressed to registered entities — TPs, PCs, TOs, TOPs, RCs and BAs. Not to you. NERC says so directly in its own primer: the August 3 deadline "only applies to currently registered entities, not the new entities that may be incorporated under the new registration criteria. However, computational load entities may see elements within the alert flow down to them."

Every action requires something from your facility:

Responses were due August 3, 2026. Your interconnecting utility has already told NERC what it will require of you. Asking to see that response is the highest-value phone call available to you this quarter.

The three standards

Project 2026-02 proposes three foundational standards, posted August 19 for comment through September 18. Registered Ballot Body members had to join the ballot pool by September 3, 2026 to vote.

CLO-001-1 — Computational Load Interconnection, Studies, and Modeling. Facility data, dynamic models, load composition, ramp characteristics, protective device settings, on-site generation; keeping it current across the lifecycle; supporting TP and PC studies.

CLO-002-1 — Computational Load Operational Data and Communications. Interpersonal Communication capability with TOP, RC and BA; operational data provision; reachability and responsiveness during events. This is the requirement class that converts a data center from an unreachable customer into an operational counterparty.

CLO-003-1 — Computational Load Protection Coordination and Disturbance Monitoring. Protection coordination with the transmission system, relay settings sharing, disturbance monitoring and data provision.

Plus limited conforming updates to FAC-001-5 and FAC-002-5.

The requirement text is in the posted drafts. The scope descriptions above reflect NERC's own characterization and the underlying SAR; read the drafts.

Why protection coordination is its own standard

The gap assessment explains the mechanism, and it is the clearest statement of a problem most operators do not know they have:

> "TOs set relay pickups, breaker clearing times, and reclosing schemes on their side of the meter. However, the TOs historically did not need to share those curves with the large load facility owner and operator who program uninterruptible power supply (UPS) and variable frequency drive disturbance ride-through logic. Therefore, the two protection schemes operate independently of one another. Without that visibility and communication, a normally cleared transmission fault can trigger protective systems on the load side even when it is cleared appropriately by the utility."

Two independent protection philosophies, no shared information, and a fault the grid handled correctly still takes your facility offline. CLO-003-1 exists to close that loop.

The paper goes further than most operators would expect into facility design: "Certain design considerations (e.g., the facility's transfer to backup power supply and the return to utility during/after voltage disturbances) for large loads can significantly impact the characteristics and risks of that large load as seen by the grid… This consideration could also extend to decisions regarding how data center compute is implemented from a software perspective."

NERC is describing workload orchestration as a reliability-relevant design decision. It also calls out UPS eco-mode versus double-conversion mode by name as a modeling-critical parameter, along with balance-of-plant loads and "the thresholds that trigger grid disconnection and transfer of load to local backup" — noting such information "is not readily available for inclusion in models."

Part Seven: The Security Question

The gap assessment ranks its risks by likelihood and impact. That ranking is the best available predictor of Phase 2 scope — and it contains a correction to the conventional wisdom.

Rated high impact and unmitigable by existing registered entities

These are the risks NERC says cannot be fixed without a registered load entity, which makes them the Phase 2 core:

Transmission adequacy · short-term demand forecasting · balancing and reserves · real-time and operations-planning coordination · proper integration and operational readiness · frequency stability (high likelihood in ERCOT specifically; low in the Eastern and Western Interconnections) · voltage stability · angular stability · oscillations · cyber security of large load · physical security

Ride-through: the real fight

There is currently no enforceable NERC standard establishing minimum voltage or frequency ride-through for load. PRC-024 and PRC-029 address generators only. The gap assessment is precise about why the obvious external standards do not fill the hole:

IEEE 2800 " applies to generation resources only and does not address load facilities. " It is the model of what does not exist for load.
IEEE 1668 provides voltage sag ride-through testing practice for end-use equipment rated below 1,000 V — the Tennessee Valley Authority uses it to assess data center sag performance — but "it does not establish minimum ride-through performance requirements for large loads connecting to the BPS."
The ITIC curve is used by operators to guess when large electronic loads will drop, but "may not be applicable for grid studies if there is a UPS between the grid and the power supplies. UPS settings or other site-level protections may impact the disturbance ride-through characteristics of the load."
SEMI F47 — the canonical semiconductor-fab sag immunity standard, and the obvious analogue — appears nowhere in either document.

So the ride-through requirement, when it comes, will be purpose-built. ERCOT, SPP and Dominion Energy are already developing large load ride-through requirements, and ERCOT's NOGRR282 is the only named ride-through rule anywhere in the Level 2 record. Expect regional rules to arrive first and NERC to harmonize afterward.

One technical tension worth understanding, because it will shape how the requirement is written: "Requiring large loads to remain connected during disturbances… may mitigate rotor-angle or frequency stability problems… However, a large load maintaining full consumption during undervoltage conditions will increase the dynamic reactive power support necessary to maintain voltage stability." Ride-through is not free to the grid either. The requirement will be a negotiated curve, not a mandate to never trip.

Ramp rates and oscillations

Ramp rate compliance appears explicitly in Recommendation 1's function list — "down ramp and up ramp." Utilities already impose limits between 8 and 300 MW/min where they impose any.

On oscillations, NERC is candid that the science is incomplete: "there are no guidelines or limits by which a system owner/operator can identify that a given forced oscillation is a reliability risk," and "there is no established practice for determining whether the oscillation is a risk to reliability and what level of urgency is associated with the issue." But the direction is clear: "Oscillations need to be mitigated before they occur so that they do not occur in real time," and the risk "needs to be considered on an interconnection-wide basis."

There is also a lever available today that most operators do not know exists: "the TOP has the ability to deny the large load permission to operate if it is causing oscillation issues."

Harmonics: probably not coming

This corrects a common assumption, including one I made in an earlier version of this article. NERC rates harmonics low likelihood, low impact, and says directly: "It is not a significant impact to BPS reliability, so NERC-enforceable requirements are not necessary. This can be adequately addressed with existing registered entities' policies."

IEEE 519 continues to govern at the point of common coupling, with distortion management framed as a joint responsibility — you limit current distortion, the grid operator limits voltage distortion. Two caveats matter: IEEE 519 "does not provide universally applicable limits for interharmonics," and if on-site generation exceeds 10% of annual average load demand and includes inverter-based or distributed resources, IEEE 519 stops applying and IEEE 2800 or IEEE 1547 governs instead.

NERC also warns against treating power quality compliance as a reliability defense: "Power quality violations are not sufficient to indicate an imminent reliability risk, and power quality standards are not intended to be used as a basis for meeting reliability requirements."

Voltage fluctuation and flicker are likewise rated low/low. System restoration is rated low/low, because "large loads will not usually be included in restoration plans."

Modeling

The Level 3 Alert already names PERC1 as the minimum baseline model. The gap assessment describes it as "the most state-of-the-art model for large power-electronic loads," developed after the paper was drafted, and notes "some of the gaps may be reduced with the new PERC1 model." NERC's Load Modeling Working Group is running a Data Center Load Modeling Workshop on September 15, 2026.

Two structural obstacles are worth knowing. The best available positive-sequence alternative — EPRI's aggregated model — "is not available as a library model in any simulation tool other than GE PSLF." And model validation cannot simply copy the generator playbook: "Staged tests are generally not possible in a load facility like the exciter step response tests that can be conducted in power plants."

Load shedding and segmentability

A quiet requirement with real design consequences. NERC wants large loads to be divisible: "the manual or UFLS shedding of the entirety of a very large load could lead to over-frequency or overvoltage. The ability to partially shed large loads, either manually or as part of a UFLS scheme, may not be part of existing utility practices."

Manual load shed obligations are rated high impact precisely because "for a situation in which a TOP or DP needs to shed a portion of a large load but the particular load is not able to be segmented, other issues, such as voltage instability, could be caused by the shedding of the entirety of the load."

There is a registration wrinkle here too: if segmentation requires UFLS relays inside your facility, "a non-registered entity owning UFLS equipment might be a risk. Therefore, the load entity might need to be registered as a UFLS-Only DP."

A "largest credible load loss"

Every Balancing Authority submits its largest credible generator loss under BAL-003, feeding directly into interconnection frequency response obligations. NERC observes: "If the largest credible load loss is not analyzed on a recurring basis, there could be a gap in preparing for the load loss/reduction events."

A "largest credible load loss" submission analogous to the generator N-1 is a plausible Phase 2 outcome — and it would make your facility a named contingency in the operator's reserve math.

Part Eight: Six Things Worth Watching

This is the quietest thread in the corpus and potentially the most disruptive, and it is more developed than most coverage suggests.

Cyber security of large load is rated high impact and explicitly not mitigable by existing registered entities: "Existing registered entities do not have adequate visibility of the large load asset owner's security protocols and practices to be able to appropriately provide requirements to mitigate the cyber security risks."

NERC names three specific threat scenarios:

Coordinated destabilization. "Adversaries could coordinate cyber attacks that might intentionally overload the grid or initiate simultaneous disconnections that might destabilize the system."
Workload manipulation — the data-center-specific threat. "Attackers may be able to manipulate workloads, potentially changing the demand of the load. These manipulations fall outside traditional operational parameters… data center security protocols are not reviewed or vetted via NERC standards."
Insider and bad-actor control. "Bad-actor control of a large load could result in excessive ramp rate, oscillations, or other performance aberrations that could impact the BPS. Furthermore, no Reliability Standards provide requirements to prevent an insider threat from accessing operational systems of large loads."

The specific CIP gaps identified: CIP-004 (personnel vetting, identity verification, criminal background checks for access to critical cyber assets), CIP-006 (physical security of BES cyber systems), CIP-007 (access control), CIP-008 (incident reporting and response), and CIP-014 (physical security of critical transmission assets). None applies to large loads today. Neither does EOP-004: "Large load asset owners have no such mandate related to event reporting for the purpose of grid integrity, reliability, and resilience."

Nor do the personnel standards. "PER-003 requires RC, TOP, and BA staff performing reliability-related tasks to have NERC certifications… no such standards exist applicable to large loads."

On physical security, NERC proposes a boundary that will end up in contracts: "the data center will have to be responsible for the security inside the data center and the TO might be responsible for the security outside of the data center (e.g., the substation)."

All of this routes to the NERC Security Working Group, "including potential SARs." Nothing in Phase 1 imposes CIP obligations. But two of the fourteen functions in Recommendation 1 are already security functions — site vulnerability assessment, and near-real-time breach notification. The architecture is being built.

Notably, NERC is less worried about the communication link than about the facility: cyber security of communications between large load and utility is rated low/low, on the reasoning that "data centers already have security practices" and "usually have tiered cyber security protection practices — although the security practices may not be consistent across the industry."

Part Nine: What To Do Now

1. The colocation allocation problem is not solved. The Owner/Operator split is progress, but a multi-tenant facility where the landlord owns the electrical infrastructure and dozens of tenants independently control the workloads driving the ramps does not decompose cleanly into two roles. NERC has already identified the underlying problem — "the large load developer can be different than the entity that owns or operates the computational equipment" — without solving it. Expect this to consume the September comment record, and expect the answer to reshape colocation contracts industry-wide. A related flag from the gap assessment: requirements for large loads "may need to be coordinated with NERC requirements for generation, such as in the case of a generator and a large load facility bein g co-located with a common point of interconnection."

2. The thresholds will drive structuring behavior. A 20 MW aggregate threshold at a single ≥60 kV interconnection invites multiple smaller interconnections, sub-60 kV service, and behind-the-meter configurations. NERC has run this play before with inverter-based resource registration. Watch for anti-fragmentation language.

3. Utilities already have a lever, and NERC told them to use it. Three separate times, the gap assessment recommends: "the TO could require that delays in the provision of necessary data by load entities result in delayed energization of load or delayed ramping of load." No registration required. If your data package is late, your energization date is the collateral. This is the single most actionable sentence in the corpus, and it is aimed at your counterparty, not at you.

4. The "qualified change" trap. NERC flags that "PCs may not have adequate knowledge of data centers to understand what should be considered a qualified change… This could lead to situations in which a large load performs a configuration change that significantly impacts the performance of the facility but does not meet the PC's definition of qualified change." Essential Action 3 tells PCs to fix that. Once they do, hardware refreshes, control setting changes and workload repurposing become restudy triggers. Build restudy lead time into deployment planning.

5. Behind-the-meter generation is genuinely unresolved. NERC says plainly that "it is unclear if and how utilities should consider information about behind-the-meter generation in load forecasts and resource adequacy studies," and that collocated generator and battery dynamics cannot be captured by current models. There is no FERC co-location docket analysis here, no capacity accreditation treatment, no position on whether co-located load is grid-connected. Do not read the silence as a safe harbor; read it as an unwritten chapter.

6. Speed versus durability. A December 31 filing produced on this schedule will be tested — in the ballot process, potentially on rehearing, and certainly in the first enforcement action. NERC's own gap assessment carries a disclaimer that "some of the requirements, practices, and modeling at the time of the publishing of this paper may be different than discussed in this paper." A standard drafted fast and challenged later is not a standard you can plan around with confidence.

Frequently Asked Questions

In the next 30 days

Run the threshold test against every North American site and every pipeline project. Read the posted Appendix 5B redlines, not summaries.
Register on the ERO Portal at eroportal.nerc.net, then request addition to the NERC-info listserv, Project 2026-02, the Large Loads Working Group and the Load Modeling Working Group.
Create a Standards Balloting System account at sbs.nerc.net and request voter credentials to join the Registered Ballot Body. Critical detail most operators miss: only one person per entity per segment can join, and computational loads fall in Segment 7 — Large Electricity End Users. Decide internally who that person is before someone else in your company claims it, or before you find you have no seat at all.
File comments. Registry criteria to ROPcomments@nerc.net; standards through SBS. Comment specifically on colocation allocation if it affects you.
Ask your interconnecting utility for its August 3 Level 3 Alert response. It tells you exactly what will be asked of you.

In the next 90 days

Inventory your data. Can you produce today: a dynamic model of your facility, IT vs. non-IT load split, ramp rate profile, protective device settings, on-site generation configuration, UPS operating mode and transfer thresholds, and as-built electrical documentation? Every gap becomes a compliance gap on the day registration takes effect. Given that only one in five facilities could produce a dynamic model when asked, this is where most operators actually stand.
Audit your protection settings against the ERCOT diagnostic list: cooling VFD undervoltage thresholds, neutral overcurrent behavior under single-phase depressions, transformer configuration, and transfer-to-backup logic. A 0.96 p.u. VFD setting is a facility-wide trip waiting for a 50-millisecond sag.
Determine whether your facility is segmentable and at what granularity. Partial-shed capability is a live requirement direction, and inability to segment is what NERC cites as making manual load shed a high-impact risk.
Attend the LMWG Data Center Load Modeling Workshop on September 15, 2026.
Establish an operational contact path with your TOP, BA and RC, staffed on the availability basis they operate on — not a daytime account-management channel. NERC has documented "erroneous contact information" as a live reliability finding.

In the next year

Stand up a NERC compliance function. Named owner, evidence retention, documented procedures. Do not let it land on facilities management by default.
Revisit your contracts. Allocate registration and compliance responsibility explicitly in colocation agreements, ESAs and build-to-suit arrangements. Address who holds the Segment 7 seat, who provides models, who carries penalty risk, and who owns the physical security boundary at the substation fence. Silence will not resolve in your favor.
Build Phase 2 assumptions into procurement. If you are specifying UPS systems, transformers or cooling infrastructure with a ten-year life, assume ride-through and ramp-rate requirements are coming.
Consider NERC membership to participate in governance, not just standards balloting.
Watch the March 1, 2027 informational filing — NERC's Phase 2 roadmap and your earliest reliable read on substantive performance requirements.

NERC is doing something it has not done in its history as the federally certified Electric Reliability Organization: extending mandatory reliability obligations to electricity consumers. It is doing so under a FERC deadline, on a schedule that compresses a multi-year process into months.

The justification is in the documents. Gigawatt-scale load has disappeared from the grid in seconds because of protection settings no grid operator could see. Three quarters of the operational data center fleet is modeled as inert impedance. Thirty-seven percent of planners have large loads on their system whose dynamic behavior they cannot characterize at all. Utilities report they lack the authority to ask you for the data, and one in five of you could supply a model when asked.

Every one of those is a fixable problem, and none of them gets fixed by a customer relationship. That is the case for registration, and it is a strong one.

The comment window closes September 18, 2026. The filing lands December 31. Phase 2 scoping begins March 2027.

The operators who come out of this well will be the ones who spent 2026 getting their facility data, protection settings, model representation and operational relationships in order — rather than waiting to find out whether the rules would really apply to them.

Primary NERC documents

NERC Large Loads Action Plan
Large Load Action Plan: 2026–2027 Timeline
Aggregated Report — Level 2 Alert: Large Load Interconnection, Study, Commissioning and Operations
Assessment of Gaps in Existing Practices, Requirements, and Reliability Standards for Emerging Large Loads
White Paper: Characteristics and Risks of Emerging Large Loads
Essential Action to Industry — Level 3 Computational Load Alert
Reliability Guideline: Risk Mitigation for Emerging Large Loads
Primer for Computational Loads
Checklist for Computational Load Entities
Incident Review: Voltage-Sensitive Crypto Load Reductions
Incident Review: Considering Simultaneous Voltage-Sensitive Load Reductions
Large Loads Frequently Asked Questions

Registration and standards

Computational Load Entity — Summary of Changes, April 2026 Posting
Proposed Changes to Rules of Procedure
Project 2026-02 Computational Loads
Standard Authorization Request — Computational Load Alignment Phase 1
ERO Enterprise 101 Informational Package

Regulatory

FERC Order on Computational Loads, Docket RD26-7-000 (July 16, 2026)
NERC Comments to FERC, Docket RM26-4 (March 2026)

Working groups and events

Large Loads Working Group (LLWG)
Load Modeling Working Group (LMWG)
LMWG Workshop on Data Center Load Modeling, September 15, 2026

Commentary

Willkie: FERC Orders New Reliability Standards for Data Centers and Other Computational Loads
Steptoe: NERC Releases Proposed Registration Requirements for "Computational Load" Customers
Morgan Lewis: NERC Alert Moves Data Centers From Emerging Risks to Planning Obligations
American Public Power Association: FERC Sets Year-End Deadline for NERC

POI Interconnection for Data Centers & Large Loads | 2026 Guide

Thu, 20 Aug 2026 22:16:57 GMT

Aug 20, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

The Complete 2026 Guide to POI Interconnection for Large Loads and Data Centers

PART 1 — WHAT CHANGED, AND WHY IT MATTERS TO YOUR PROJECT

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Prepared by Keentel Engineering | Power System & Substation Engineering | Tampa · Austin · Sacramento · Baltimore

The most expensive mistake in this business is assuming the process you learned in one region applies in another. It does not. The seven ISO/RTOs sit at radically different levels of maturity — one has a fully tariffed, RTO-run large load study process; two have no load process at all and are being ordered to build one.

3.1 The single most important structural fact

In most of the United States, the RTO does not interconnect your load. The transmission owner does.

Generator interconnection is FERC-jurisdictional and RTO-run: a defined queue, defined study phases, a pro forma interconnection agreement. Load interconnection, historically, is a retail matter — state-jurisdictional, executed by the local transmission owner or electric distribution company under a state-approved tariff. The RTO's involvement was limited to load forecasting and regional reliability planning.

That model held while the biggest new customer was a 30 MW factory. It broke when the biggest new customer became a 1,000 MW AI campus that can be built in two years on a grid where transmission takes seven to ten.

Here is where each region stands today:

1.1 The event that rewrote the rulebook

On July 10, 2024, a lightning arrester failed on a 230 kV line in Northern Virginia. What followed was six faults in 82 seconds. No utility equipment tripped the load. Instead, approximately 1,500 MW of data center load disconnected itself — voltage at those sites dipped to roughly 0.25–0.40 per unit, and customer-side protection and control logic took the facilities off the grid. About 1,260 MW stayed off for hours.

NERC's incident review traced the sustained loss to an interaction nobody had modeled: the automatic reclosing sequence on the faulted line produced repeated voltage dips, and the data centers' UPS control schemes counted those dips. Three disturbances inside one minute, and the facilities transferred to backup power and stayed there.

That single event is the most cited technical fact in the entire large-load reliability debate. It is why your ride-through settings are now a negotiated term in your interconnection agreement, why the grid operator wants a dynamic model of your load, and why "we'll figure out the electrical details later" is no longer a viable project strategy.

Source: NERC, Incident Review: Simultaneous Voltage-Sensitive Load Loss (January 8, 2025).

1.2 The regulatory cascade, 2025 to today

1.3 The five things FERC told every RTO to fix

Every one of the six show-cause orders issued June 18, 2026 demands the same five categories of reform. Understanding these tells you where every region is heading, regardless of where it is today:

1. A real application and study process. Rolling acceptance of requests, a non-refundable application fee, escalating readiness milestones (deposits and site control), and studies completed in 60 to 90 days. Studies must evaluate rapid ramp-up, demand variability, cumulative regional impact, and alternative transmission technologies — not just the traditional "build a bigger wire" answer.

2. Cost transparency and protection against cost shifting. A publicly searchable database of large load additions by pricing zone. Published network upgrade cost estimates. Pro forma cost recovery agreements with minimum customer contributions and financial security.

3. Clear rules for co-location and behind-the-meter generation. Three new transmission services — interim non-firm network service during construction, Firm Contract Demand, and Non-Firm Contract Demand. Ancillary services charged on gross, not net, demand.

4. Real services for flexible load. Loads that accept operational limits get faster access and smaller network upgrades. This is the single largest schedule lever available to a developer.

5. Fast-track interconnection for generation electrically proximate to a large load. Generation within roughly two substations of the load gets an accelerated path, including load-limited service with no injection to the transmission system.

The strategic read: the era of a data center being a passive, anonymous, price-taking block of demand is over. In every region in the country, a large load is becoming a studied, modeled, telemetered, curtailable, cost-responsible grid asset — much closer to how a generator is treated. Plan your project accordingly.

1.4 FERC's proposed definitions (the emerging national baseline)

Large Load — a new commercial or industrial customer at a single site with peak load ≥50 MW, interconnecting above 69 kV, not part of a co-location arrangement.
Co-Located Load — end-use load physically connected to an existing or planned generating facility, on the interconnection customer's side of the POI.
Eligible Load — co-located load and/or load served by behind-the-meter generation.
Flexible Large Load — a large load willing to limit energy withdrawals under specified system conditions.
Electrically Proximate Large Load — load no more than two substations from the generating facility.
Computational Load (FERC/NERC, RD26-7) — power demand from IT equipment: servers, storage, networking. Covers data centers, AI training facilities, and cryptocurrency mining.

3.2 SPP — the most developed framework in the country

SPP's High Impact Large Load (HILL) process, approved by FERC on January 14, 2026 (Docket ER26-247, 194 FERC ¶ 61,031) and effective January 15, 2026, is the model everyone else is being measured against. FERC explicitly held SPP up as best practice for "bring your own generation."

HILL threshold: a new commercial or industrial load that is ≥10 MW connected at ≤69 kV, or ≥50 MW connected above 69 kV. Electric storage resources are excluded.

Four pathways:

How to Use This Guide

PART 6 — WHAT AN ELECTRICAL POWER ENGINEERING FIRM ACTUALLY DELIVERS

Regional mechanics differ, but the arc is the same everywhere. Here is the generic ten-stage sequence, with the deliverable that gates each stage.

Stage 1 — Site screening and POI feasibility (4–8 weeks). You have three candidate sites. Your engineer screens each against nearby transmission: available capacity, distance to the nearest suitable substation, existing line ratings, known constrained corridors, and the transmission owner's published interconnection requirements (the FAC-001 document). Deliverable: a POI screening memo ranking sites by expected upgrade scope, cost, and schedule. This is the highest-leverage $50,000 you will ever spend, because it is the only point at which changing your answer is free.

Stage 2 — Pre-application coordination. Meet with the transmission owner and, where applicable, the RTO. Confirm the POI, confirm which tariff or process applies, confirm the study deposit, confirm the data package. In NYISO this is formalized (pre-filing coordination with the CTO to define the POI). Everywhere else it is informal and skipping it is a classic error.

Stage 3 — Application and data package (4–8 weeks). The interconnection request itself, plus the technical data package: load characteristics, one-line, model files, site data, ramp schedule, site control evidence, and the deposit. Incomplete applications are the single largest source of avoidable schedule loss. Deficiency cure periods are typically 10 to 15 days; miss one and you fall to the next batch or cycle.

Stage 4 — Scoping call and study agreement. The clock usually starts here, not at submission. SPP: scoping call within 10 days, which starts the 90-day HDPS. NYISO: scoping call about two weeks after the request is deemed complete, then a System Impact Study Agreement within 30 days with a $150,000 deposit.

Stage 5 — System Impact / Delivery Point Study. Steady-state thermal and voltage, short circuit, stability, and increasingly EMT screening and dynamic load model review. Output: the list of overloaded facilities, the network upgrades required, and — critically — the MW at which each constraint binds. That last number tells you what you can energize without upgrades, which is often the difference between a 2028 and a 2032 project.

Stage 6 — Facilities Study and cost estimate. Scope, single-line, general arrangement, schedule, and a Class 3 or Class 2 cost estimate for the transmission owner's interconnection facilities and network upgrades. Non-binding, typically ±30% early and ±15% later.

Stage 7 — Interconnection agreement and security. The agreement defines the POI, the ownership boundary, cost responsibility, the ride-through and operating requirements, telemetry obligations, curtailment obligations, milestones, and the consequences of missing them. Negotiate the technical operating requirements here, not at commissioning. Financial security is posted; in most current regimes a substantial portion is non-refundable.

Stage 8 — Detailed design (30% → 60% → 90% → IFC). Your engineering firm's core production phase. Long-lead equipment specifications must be issued for procurement at 60%, because transformer and breaker lead times, not design, drive the schedule.

Stage 9 — Construction, testing and commissioning. Factory acceptance tests, site acceptance tests, relay commissioning, CT/PT ratio and polarity verification, grounding verification per IEEE 81, protection end-to-end testing, SCADA point-to-point verification, and — new under NERC's Level 3 Alert — full-load / no-load and ±10% voltage testing with as-built model verification.

Stage 10 — Energization, model attestation and ongoing compliance. Staged energization per your load commissioning plan, as-built model submission with attestation that parameters represent installed equipment, telemetry cutover, and entry into ongoing obligations: forecasting, curtailment availability, disturbance data sharing, and periodic model validation.

PART 5 — THE POI INTERCONNECTION PROCESS, END TO END

Practical takeaway: at 50 MW you are a large load essentially everywhere. At 20 MW you trigger NERC modeling and data expectations. Below 10 MW you are generally in ordinary retail service territory — but if you have any intention of phasing up, design and disclose to your ultimate MW from day one. Retroactively pulling a project into the large-load regime after energization is the most expensive way to discover these rules.

There is no single national definition. There are at least eight in active use, and they do not match. If you are running a multi-region site selection, build this table into your model.

6.1 The stage-gate deliverable matrix

This is industry practice, not a published rule, but it is near-universal in U.S. utility and EPC contracting, and cost estimate classes map to it via AACE.

Two practitioner rules worth putting on the wall:

At 30%, the index must already be complete even though the sheets are not. The index is the contractual definition of scope.
Nothing should appear at IFC that the owner and reviewers have not already seen. IFC is a sealing exercise, not a design exercise.

6.2 The drawings, one by one

One-Line Diagram (SLD / key single line)

What it is: the single-per-phase representation of the POI through your medium-voltage distribution. Why the TO and ISO need it: it is the primary artifact against which the transmission owner builds its short-circuit model, decides breaker configuration and the ownership boundary, and confirms your facility cannot island or backfeed.

Content that will get your SLD bounced back if missing:

POI location and the ownership/maintenance boundary line, explicitly drawn
All breakers, switches, motor-operated air break switches and disconnects, with device numbering and normal position
Transformer MVA ratings (ONAN/ONAF/ONAF), impedance (%Z on a stated base), winding connection, and neutral grounding method — note that some utilities specify delta on the transformer high side
BIL by voltage class, bus continuous ratings, breaker interrupting ratings
CT and PT locations, ratios, and accuracy classes
Surge arrester locations and ratings
All on-site generation, BESS, UPS, and standby gensets — including the transfer scheme
Reactive devices: capacitor banks (size, kV, number of steps), harmonic filters, STATCOM/SVC
Phasing and grounding of station service

Three-line diagram

Per-phase (A/B/C) plus neutral, showing every CT secondary circuit, every PT secondary, polarity marks, wye/delta secondary connections, relay input terminals and metering circuits.

Why it matters: this is the drawing the utility protection engineer actually checks. Wrong CT polarity, or a wye-versus-broken-delta PT error on a directional or differential element, is the single most common cause of a failed commissioning test. It is also the basis for CT saturation checks and burden calculations.

Site plan / plot plan / general arrangement

Property boundary and legal description from an ALTA/NSPS survey, easements, right-of-way strips
POI coordinates to latitude/longitude survey accuracy. Some TOs require considerably more: final yard elevation at the termination structure, conductor and shield-wire attachment point elevations, structure orientation, and line terminal connector type
Access roads, gates, turning radii for transformer delivery, crane pads
Fence line, security, oil containment and SPCC berms, stormwater
Setbacks from wetlands and floodplain, and the site's relationship to the transmission corridor
Acreage: a 500 MW campus switchyard at 230 kV commonly occupies 8 to 20 acres depending on bus configuration and AIS versus GIS. GIS trades roughly 70–80% of the footprint for cost

Substation physical layout, sections, elevations, bus configuration

Bus configuration is a reliability and cost decision your consultant should document with a trade study. Common relative installed cost benchmarks:

7.1 Steady-state load flow (N-0, N-1, N-1-1)

Answers: can the system serve your load at each ramp step without thermal overload or voltage violation, and what network upgrades does that trigger?

Base cases: summer peak, winter peak, light load, shoulder — each at Phase 1 / Phase 2 / Phase 3 MW
N-0 (all elements in), N-1 (single element out), N-1-1 (one element out for maintenance plus a contingency). N-1-1 is where large loads usually fail, and it is the standard driver of network upgrades
Criteria: ANSI C84.1 Range A/B at utilization; TO planning criteria (typically 0.95–1.05 pu normal, 0.90–1.05 emergency); NERC TPL-001-5.1 performance categories P0 through P7
Outputs: thermal loading tables, voltage profiles, contingency ranking, and — the single most commercially valuable output — a list of overloaded facilities with the MW headroom at which each binds. That is the number that tells you what you can energize before upgrades are complete
Also covered: voltage drop across your ramp, and motor-start dips (some TOs limit motor-starting dip to 3% on the transmission system)

Tools: PSS/E, PSLF, PowerWorld Simulator, TARA.

7.2 Short circuit, fault duty, breaker duty

Answers: do existing TO breakers remain within rating once your transformers, on-site generation and BESS are added? What fault duty must you specify to?

Three-phase, single-line-to-ground, line-to-line-to-ground and line-to-line faults; momentary (close-and-latch) and interrupting duty
ANSI/IEEE C37.010 / C37.04 / C37.06 methodology (X/R correction, E/X with decrement, contact parting time) versus IEC 60909. U.S. TOs use ANSI — specify which method in your scope
Breaker continuous rating commonly set at 125% of maximum transformer rating
Output includes Z1/Z0 at the POI, X/R ratio, and updated fault duty for the TO's protection settings

A frequently overlooked risk: large medium-voltage BESS installations behind your POI can push nearby buses over breaker duty. TOs care primarily about whether you break their breakers. Find out early.

Tools: ASPEN OneLiner, CAPE, ETAP, SKM PowerTools, EasyPower, PSS/E short circuit module, PowerFactory.

7.3 Transient stability, voltage stability, PV/QV

Answers: does your load's abrupt disconnection or reconnection destabilize the area? Is there enough reactive margin?

Transient stability: three-phase and single-line-to-ground faults with normal and delayed clearing, per TPL-001 P1 through P7
The genuinely new case: loss of the large load itself as a contingency. A 500 MW instantaneous load rejection is a generation-surplus and over-frequency event, and a voltage-rise event
PV curves for the import path into the load pocket, determining the MW transfer limit and margin
QV curves at the POI bus, giving reactive margin in MVAr and the sizing basis for capacitor banks and STATCOM
Small-signal and modal analysis where AI training oscillations sit near inter-area modes

One finding that inverts a century of load-modeling intuition, and belongs in every developer briefing: NERC found that large loads with active power electronics operate at roughly unity displacement power factor and their filters produce reactive power — making them a capacitive, leading-power-factor load, which worsens rotor angle stability by forcing nearby synchronous machines to absorb VArs. The classical assumption that load is inductive and helps damping does not hold for a modern AI campus.

7.4 Dynamic load modeling — and why your load is hard to model

The core problem. The classic composite load model (CMLD / CMPLDW — three-phase motors A/B/C, single-phase motor D, electronic load, static load, distribution equivalent, plus DER_A) was built to represent residential and commercial feeders dominated by induction motors. A 500 MW AI campus is 80–90% power-electronic rectifier load behind UPS, plus a large VFD-driven chiller block, plus fast control loops that are vendor intellectual property.

What your load actually does, from published measurements:

One 50 MW block of a 200 MW AI data center changed demand at 1.9 per unit per second for about 250 milliseconds
A crypto facility ramped down 298 MW in 25 seconds
A data center went from roughly 450 MW to 40 MW in 36 seconds
Large AI clusters have been observed swinging 35–70 MW or more within a minute
AI training shows "jittery" oscillation between compute-intensive and network-intensive phases within a single training step — sub-second and quasi-periodic
AI training facilities frequently omit UPS for the IT load entirely, relying on checkpoint and restore. That means no ride-through buffer at all

What your engineer must deliver:

A site-specific dynamic load model — either a parameterized CMLD/CMPLDW, or a user-defined model. Several TOs now explicitly require "composite load model or user-defined dynamic model in PSS/E format," plus the NERC data center information questionnaire
The model must include protection and control behavior: disconnection and real-power reduction thresholds for voltage and frequency events, reconnection thresholds (level and duration), and the disturbance-counting logic if present
Load composition breakdown — IT equipment, motors, VFDs, power electronics, lighting. NERC's Level 3 Alert requires IT versus non-IT (cooling) percentages at multiple load levels
Both a positive-sequence phasor-domain model and an EMT model, which must reasonably match each other in simulated behavior and parameterization
Model verification (structure and parameters match the installed equipment) and model validation (simulation versus measurement), with post-commissioning validation using high-resolution fault records
An attestation that as-built parameters represent installed equipment, not library defaults. NERC calls this out specifically, because default-parameter submissions are common and worthless

PERC1 — the load model format SPP prefers and NERC's Level 3 Alert names ("the PERC1 model or equivalent-or-better") — is emerging as the standard representation for computational load. Expect it to spread.

7.5 EMT / PSCAD studies — when they are required

Screening triggers (NERC's reliability guideline gives the cleanest published list):

Phasor-domain simulations fail to converge
Large power electronic devices present
Weak grid: short-circuit ratio ≤ 2
Series capacitors present on the network
Large turbine generators with load fluctuations in the 5–60 Hz range — i.e., torsional and sub-synchronous resonance interaction risk from AI load oscillation

SPP's HILL process makes it explicit and quantitative: EMT screening in the base study using SCR, weighted SCR, composite SCR and critical clearing time, with thresholds of SCR/WSCR/CSCR ≥ 6.0 and CCT ≥ 0.15 s. Fail the screen and you owe a detailed PSCAD study. ERCOT went further: PGRR144 requires PSCAD models for all Large Loads, hardware-benchmarked, with generic library blocks not acceptable.

What EMT answers that phasor-domain simulation cannot:

Sub-synchronous oscillation and sub-synchronous control interaction (SSO/SSCI)
Fast control interaction between your rectifiers and nearby inverter-based resource plants or HVDC
Harmonic resonance and amplification
Temporary overvoltage, insulation coordination, switching transients — capacitor bank energization, transformer inrush and sympathetic inrush, breaker restrike
Single-phase and unbalanced fault behavior, which a positive-sequence model literally cannot represent

Model quality requirements (SPP and CAISO publish nearly identical criteria; these are the de facto U.S. standard):

Note: these criteria were written for inverter-based resources. ISOs are now applying the same bar to large power-electronic loads. Budget the time and the vendor cooperation accordingly — obtaining an accurate PSCAD model of a UPS from a manufacturer who considers the control code proprietary is a commercial negotiation, not an engineering task, and it takes months.

7.6 Harmonic and power quality study (IEEE 519), flicker (IEEE 1453)

Deliverable: harmonic load flow plus a frequency scan (impedance versus frequency at the POI) across the network configurations that matter — N-0, N-1, capacitor banks in and out, minimum and maximum short-circuit levels. Resonance almost always shows up in the minimum-fault-level case, which is the one people forget to run.

Published utility requirements worth designing to:

8.1 The toolset, and who uses what

8.2 Model submission requirements by region

The one rule that saves projects: model version requirements change annually, sometimes mid-cycle. Check the specific ISO's model submission page at the start of every study cycle. A model submitted in the wrong PSS/E version is a rejected submission, and a rejected submission in a batch process can cost you six months or an entire cycle.

8.3 A note on model rejection

The most common reasons model packages get returned:

Wrong software version
Generic library blocks used where hardware-specific models are required
User-defined model submitted without source code, DLL, or documentation
PSS/E and PSCAD models that do not match each other in behavior
Missing flat-start stability — the model does not initialize cleanly
Load model that does not include protection and control behavior (no disconnection thresholds, no reconnection logic)
Default parameters instead of as-built values, with no attestation
Missing test cases or benchmark report

Every one of these is avoidable. In our experience, first-round model acceptance is the highest-leverage schedule improvement available on an interconnection project — a rejected package typically costs 30 to 90 days per round trip.

This is the checklist. Items marked ★ are the ones that most often stall projects.

9.1 Load characterization

★ MW ramp schedule by phase, expressed in MW by month, not year, for Phase 1 / 2 / 3 through ultimate build. ERCOT calls this the Load Commissioning Plan and non-performance against it can cancel your project
Peak MW, average MW, and minimum MW at each phase; annual load factor
★ IT load versus non-IT (cooling and mechanical) split, as a percentage, at multiple load levels — a NERC Level 3 Alert requirement
Power factor at the POI across the loading range, with the measurement point stated — including the light-load leading power factor case
★ Harmonic current spectrum of the IT load, magnitude and phase by order, to the 50th, plus inter-harmonic content, from your UPS and PSU vendors
Ramp rates — maximum ramp up and down in MW/min and MW/s, for startup, shutdown and normal operation, plus sub-second and sub-synchronous cyclic content
Expected sub-hourly demand profile and workload type (AI training, inference, colocation, crypto)
Forecast capability — can you provide 5-minute to 7-day-ahead forecasts? Hourly forecasts on a rolling horizon? This is becoming a tariff obligation

9.2 Electrical architecture

★ Internal single-line diagram — POI → MV switchgear → transformers → UPS → PDU → rack
Transformer sizes and impedances — MVA ratings (ONAN/ONAF), %Z on a stated base, winding configuration, tap ranges
★ UPS topology — double-conversion static, line-interactive, diesel rotary (DRUPS), or no UPS on the IT load (checkpoint-based, common on AI training). DRUPS is a flag: it produced the worst observed behavior in the 2024 event
★ Full protection and control settings of everything downstream of the POI — undervoltage and overvoltage pickup and time delay, underfrequency and overfrequency, rate-of-change-of-frequency, transfer-to-backup logic, and any disturbance-counting logic
★ Reconnection thresholds and reconnection ramp — levels, durations, and whether reconnection is automatic or manual
Ride-through capability curve as designed (ITIC/CBEMA/SEMI F47), mapped against your region's required envelope, with an honest gap analysis
Motor loads — chiller, pump and CRAH ratings, starting method (across-the-line, soft start, VFD), inrush, starting kVA, and whether they restart automatically after a dip
Backup generation — diesel or gas turbine, MW, count, N+1 or 2N, start time, and emissions permit hour limits (this constrains demand response participation and is regularly overlooked)
On-site generation and BESS — MW/MWh, inverter make and model, grid-forming versus grid-following, control mode, and whether it can export (and whether export is contractually prohibited)
Reactive devices — capacitor banks (size, kV, steps), harmonic filters, STATCOM
DC system — battery type, sizing, charger
Redundancy tier — Uptime Institute Tier III (concurrently maintainable) or Tier IV (fault tolerant); N+1, 2N, 2N+1

9.3 The interconnection ask

★ Desired POI voltage (69 / 115 / 138 / 230 / 345 / 500 kV) and preferred bus configuration
Preferred POI substation or line, with ranked alternatives
★ In-service date — and the "first power" date versus the "full load" date. These are different, both matter, and confusing them is a classic source of schedule dispute
Firm versus non-firm service, and your willingness to accept curtailment. This is the fastest schedule lever available anywhere in the country right now
Flexibility offer — MW curtailable, response time, duration, events per year, and the control path
Remote disconnect acceptance
Metering ownership preference — customer-owned versus utility-owned

9.4 Site and land

★ Site control evidence — deed, purchase agreement, or option. Some processes require site control locked for three years at application
Legal description, acreage, parcel IDs
★ POI latitude/longitude to survey accuracy, plus final yard elevation at the termination structure, conductor and shield wire attachment elevations, and structure orientation where the TO requires it
★ ALTA/NSPS survey with easements and encumbrances
★ Geotechnical report — borings, bearing capacity, groundwater, and critically soil resistivity by Wenner four-pin method, multiple traverses, for the IEEE 80 study. Ask for resistivity explicitly. Standard geotechnical scopes omit it, and discovering that later costs a full remobilization
Topographic survey, existing utilities, wetlands delineation, FEMA flood panel, Phase I environmental site assessment
Environmental and permitting status — state siting authority, threatened and endangered species, cultural resources, air permit for gensets, stormwater

9.5 Commercial readiness

★ Study deposit and milestone payment readiness. Deposits now run from $100,000 to $500,000+ per stage, plus per-MW readiness deposits and per-MW security
Evidence of financial commitment — grid operators now require quantified evidence of developer funding for load forecast inclusion
★ Disclosure of duplicative requests. Both PJM and Texas SB6 require you to disclose parallel requests for the same nameplate. Failure to disclose gets you removed from the forecast — or, in Texas, potentially denied
State and local support — some fast-track processes require documented commitment from the siting authority to expedite permitting
Executed offtake, lease, or construction commitments with financial penalties
Corporate creditworthiness documentation

PART 10 — TIMELINE AND COST

10.1 Regional process timelines

Real-world totals, from practice: roughly 12 months in an uncongested territory with existing headroom; 3 to 5 years in constrained markets; and reports of 7 years in the most congested corridors. Engineering deliverables are almost never the critical path. Transformer and breaker lead times and the grid operator's study queue are.

10.2 Engineering deliverable durations

Typical for a 2–4 position 230 kV POI station (industry practice; varies by TO and contract):

Studies run in parallel: grounding and lightning land at 60%; arc flash and coordination at 90%; EMT and dynamic model development typically start at the interconnection application and iterate for 6 to 18 months. That last item is the one developers consistently underestimate.

10.3 Cost benchmarks

Transmission and substation unit costs (from MISO's published Transmission Cost Estimation Guide, which is the best public source; includes contingency and AFUDC):

Interconnection cost per kW (from national laboratory analysis; generator-side but directionally useful): completed projects around $194/kW, active around $294/kW, withdrawn around $671/kW. Small projects (1–50 MW) around $763/kW; very large (750+ MW) around $244/kW. At $500/kW, a 100 MW project is roughly $50 million.

Order-of-magnitude data center interconnection totals (practitioner ranges, not published rates):

Core substation equipment: $3–7M
Full utility interconnection including lines and easements: $20M–$100M+
Power-constrained regions: $40–60M+
On-site backup generation: $3–10M+ per MW

Engineering fees: planning-grade estimates typically apply project management, A&G, and engineering adders totaling 8–10% of project cost. A real EPC or owner's-engineer fee on a data center POI runs higher — commonly 6–12% of installed cost for full design plus owner's engineer scope. Industry practice, not a published rate.

Study deposits: ISO study deposits currently run $50,000 to $640,000 per stage depending on size and region, plus per-MW readiness deposits ($4,000–$20,000/MW range in published processes) and per-MW financial security in some jurisdictions.

PART 13 — FREQUENTLY ASKED QUESTIONS

This is a working reference for data center developers, hyperscalers, colocation providers, industrial load owners, site selection teams, EPC contractors, and the attorneys and financiers who back them. It answers, in order:

What changed in 2025–2026 and why your 2023 interconnection playbook is obsolete
Who the U.S. grid operators are, where their rules live, and how to reach them
How a large load actually connects in each region — because the process is different in all seven ISO/RTOs and different again outside them
What an electrical power engineering firm produces for you: drawings, studies, models, reports
Exactly what data you must hand your engineer before real work can start
What it costs, how long it takes, and where projects die

The final third is a 60-question FAQ. If you are in a hurry, read Part 1, the threshold table in Part 4, the client checklist in Part 7, and the FAQ.

A word on shelf life

Large load interconnection rules in the United States are being rewritten right now, in nearly every region, simultaneously. FERC issued show-cause orders to all six RTO/ISOs on June 18, 2026; responses were due August 17, 2026. NERC must file mandatory reliability standards for computational loads by December 31, 2026. Texas paused its data center interconnection queue on August 3, 2026. Anything written in this space has a shelf life measured in months, not years. Treat every rule citation here as a pointer to a primary source — and verify against the current tariff before you commit capital. Keentel maintains updated regional briefs; contact us for the current version.

Stakeholder bodies to watch: CIFP-LLA, Load Analysis Subcommittee (LAS), Markets & Reliability Committee (MRC), Planning Committee (PC).

MISO

PART 2 — THE U.S. GRID OPERATOR MAP

2.1 The seven ISOs and RTOs

The continental United States is served by seven Independent System Operators / Regional Transmission Organizations covering roughly two-thirds of U.S. electricity demand. The remaining third — most of the Southeast, most of the non-California West, and Alaska/Hawaii — is served by vertically integrated utilities operating their own balancing authorities under FERC Open Access Transmission Tariffs (or, for municipal and federal entities, under their own board-approved rules).

2.2 Key resources by operator

PJM Interconnection

Stakeholder bodies: Large Load Working Group (LLWG) — the dedicated forum; Planning Advisory Committee (PAC); Interconnection Process Working Group (IPWG); Reliability Subcommittee (RSC). MISO also runs quarterly Large Load Additions Workshops.

SPP

Contacts: GIstudies@spp.org; SPP Request Management System (RMS). Stakeholder bodies: MOPC, Generator Interconnection Advisory Group (GIAG), Regional Tariff Working Group.

ERCOT

Note: ERCOT is not FERC-jurisdictional for transmission rates and interconnection. Its rules come from ERCOT protocols/guides approved by the PUCT. Texas legislation (SB6) sits on top.

CAISO / California

ISO New England

NYISO

PART 3 — HOW A LARGE LOAD ACTUALLY INTERCONNECTS, REGION BY REGION

Contact for modeling versions: irtt@iso-ne.com

Federal and reliability bodies

2.3 The regions with no ISO

If your site is in the Southeast (Southern Company, TVA, Duke Carolinas, Dominion South Carolina, Florida utilities), the Intermountain West (PacifiCorp, Idaho Power, NV Energy, Arizona Public Service, Public Service Co. of Colorado, NorthWestern), the Northwest (BPA, Portland General, Puget Sound Energy, Avista), or on a municipal or federal system (Salt River Project, LADWP, Imperial Irrigation District, NYPA, LIPA, Nebraska public power), there is no regional interconnection queue for load. You connect through the individual transmission provider:

Retail large load served by a vertically integrated utility — you connect under the utility's state-approved retail tariff and line-extension rules, increasingly supplemented by a purpose-built large load tariff (see the table in Part 3.8).
Wholesale arrangement — the serving entity takes Network Integration Transmission Service or Point-to-Point service under the transmission provider's OATT, and your physical connection is added as a new Point of Delivery (POD).
BPA — the Line and Load Interconnection (LLI) process governs new or modified Points of Delivery on the Federal Columbia River Transmission System. Submit a signed LLIR form to interconnection@bpa.gov with "LLIR" in the subject; requests sent anywhere else are not queued.
Non-jurisdictional entities (SRP, LADWP, IID, most munis and co-ops) — their own board-approved rules govern, not FERC and not the state commission.

Two things follow. First, requirements vary utility to utility and are evolving fast; there is no single national playbook. Second — and this matters commercially — FERC's June 2026 show-cause orders apply only to the six RTO/ISOs. Non-ISO Western and Southeastern territories are not covered, so their processes will diverge further from the emerging RTO baseline, not converge with it.

* Also note the market overlays that do not change interconnection. CAISO's Extended Day-Ahead Market (EDAM) went live May 1, 2026 with PacifiCorp; SPP's Markets+ targets an October 2027 go-live with BPA, APS, Puget Sound Energy and Salt River Project among its funding entities. Joining a day-ahead market changes how energy is dispatched and settled. It does not * change how your data center interconnects. A site in PacifiCorp territory still connects through PacifiCorp, under Oregon/Utah/Wyoming state tariffs and PacifiCorp's OATT.

The study architecture — and why it is elegant:

SPP runs the HDPS (HILL Delivery Point Study) — 90 calendar days
The host Transmission Owner runs the LCS (Load Connection Study) in parallel — equipment ratings, breaker duty, local network impacts

The sequence: complete package validated → scoping call within 10 days (which starts the 90-day clock) → 90-day base stage → supplemental stage if screening fails (outside the 90-day window) → results valid one year , within which you must elect Network Integration Transmission Service.

HDPS base stage — five mandatory analyses:

Thermal overload under contingencies
Voltage stability (steady-state power flow)
Short circuit duty per NERC TPL-001-5.1
RMS dynamic performance using the CMLD or PERC1 load model
EMT screening using short-circuit ratio (SCR), weighted SCR, composite SCR, and critical clearing time — screening thresholds are SCR/WSCR/CSCR ≥ 6.0 and CCT ≥ 0.15 s

Fai l the screen and you go to the supplemental stage: detailed PSCAD, EMT dynamic performance, sub-synchronous oscillation screening per CIGRE TB 909, converter-driven stability, emergency power control, and fault ride-through validation. SPP also runs a HILL Model Verification study comparing your RMS and EMT model responses to each other.

What SPP requires in the application package:

Delivery Point Assessment Request Form
Executed HDPS Agreement
10-year load forecast (summer, winter, and light-load seasonal values)
One-line diagram showing local facility changes
IDEV power-flow model files
Additional HILL Characteristics Form (Business Practice 7850)
Load model in CMLD and/or PERC1 format — PERC1 preferred
PSCAD EMT model — technically optional at initial submission, but submit it anyway. If you skip it and fail the EMT screen, you fall out of the 90-day window into the open-ended supplemental stage.

HILLGA — generation devoted to your load

Under Attachment BB, a generator whose POI is no more than two substations from your HILL's Local Delivery Facilities, within the same SPP Deliverability Area, can take Load Limited Resource Interconnection Service (LLRIS). A 90-day System Impact Study, 60-day restudy, agreement execution 45 business days after the study report — roughly 7 to 8 months end-to-end from validated submission. The term is 5 years from COD; to continue, the generator must enter a subsequent Consolidated Planning Process cycle seeking NRIS or ERIS.

Cost allocation: network upgrade costs are directly assigned to you during provisional service under Attachment AX, then migrate to Base Plan funding when firm service is approved.

What's changing: even SPP got a show-cause order (EL26-68). FERC wants SPP to evaluate alternative transmission technologies without being asked, memorialize ongoing operational requirements in transmission service agreements, publish searchable network upgrade cost data, and explain whether its current offerings suffice for flexible loads.

Scale: industry analysis of SPP data reports ro ughly 26.4 GW of large load interconnection requests since 2020 for facilities over 100 MW, about 9 GW of it data centers, against a system peak of roughly 54 GW. Around 7 GW converted to executed service agreements — a 27% conversion rate, which is the number every developer and every planner should have tattooed somewhere.

3.3 ERCOT — highest volume, most turbulence

ERCOT has the largest large-load queue in the world and, as of this writing, the most disrupted process in the country.

The numbers. ERCOT reported 410 GW of large loads seeking interconnection in March 2026 (about 87% data centers), 438 GW by the June 2026 Board meeting, and the Governor's August 3, 2026 letter cited 474 GW — more than five times ERCOT's record peak, roughly 90% data centers. Peak simultaneous large-load consumption actually observed in June 2026: about 3,675 MW. That gap between requests and reality is the entire story of ERCOT large load policy.

The current process: Batch Zero. PGRR145 and NPRR1325, approved by the ERCOT Board June 2, 2026 and the PUCT June 18, 2026, effective July 11, 2026, replaced the individual-study regime with a batch process.

Threshold: 75 MW or greater aggregate peak demand at a single site behind one or more common Points of Interconnection or Service Delivery Points.

Batch Zero is a one-time transitional study designed to clear the backlog that was forcing endless restudies. Batch 1 and onward run on a roughly six-month cadence with "seriousness" entry criteria; Batch 1 applications were slated to open in Summer 2027.

Three classifications:

Capacity allocation — the two numbers that define your project:

LPC (Low Power Consumption) — the maximum firm interconnection capacity you can consume with no dispatch or curtailment obligation.
MPC (Maximum Power Consumption) — your full requested peak.

Access above LPC requires electing one of two pathways:

PCLR — Provisional Controllable Load Resource (Form W, notarized, two parts). You may consume up to MPC in exchange for following ERCOT real-time dispatch, complying with ramp rates, and accepting ERCOT-imposed energy bid caps. ERCOT gives no guarantee the capacity is servable in real time.
WLPUN — Withdrawal-Limited Private Use Network (Form X, notarized, executed by both the load entity and the generator owner). This is bring-your-own-generation; your withdrawal limit is set by modeling peak demand with the on-site generation offline.

The forms you will actually fill out (all on ERCOT's Large Load Integration page):

Batch Zero Load Information Form (LIF)
Form W — Provisional Controllable Load Resource declaration
Form X — Withdrawal-Limited Private Use Network designation
Attestations under Planning Guide §9.2.1.1(1)(b)–(g), §9.2.1.2(1), §9.2.1.4(3)(d)
ERCOT Interconnecting TSP Dynamic Stability Study Form
ERCOT Affiliate Attestation Form
Energization Request for New Standalone Large Load
Notice of Proposed Net Metering Arrangement (PURA §39.169(a))
DWG Large Load Survey — required for all interconnection requests
DWG Survey for Subsynchronous Studies — required for some

The TSP is the gatekeeper, not ERCOT. You submit to your Transmission Service Provider or Distribution Service Provider — Oncor, CenterPoint Energy Houston Electric, AEP Texas, Texas-New Mexico Power, LCRA Transmission Services Corporation, or a municipal or cooperative utility. The TSP compiles and submits your eligibility package to ERCOT. A utility that misses its deadline can cause your project to be excluded, even if you did everything right. Build that risk into your schedule and your communications plan.

Texas SB6 (89th Legislature, 2025; signed June 20, 2025; most provisions effective September 1, 2025) sits on top

PUCT Project 58481 → 16 TAC §25.194 proposes a two-agreement structure for loads ≥75 MW: an Intermediate Agreement (site control, permitting progress, per-MW financial security, study fees of $100,000 for 75–249 MW or $300,000 for ≥250 MW) followed by a final Interconnection Agreement (per-MW non-refundable fee, 100% cost responsibility for interconnection infrastructure, execution within 30 days of study completion or the request is cancelled). The per-MW dollar figures are in dispute across published summaries — pull the actual rule text from PUCT Project 58481 before you model them. Verify adoption status; as of mid-2026 the rule was proposed, not adopted.

The August 2026 pause. On August 3, 2026, the Governor directed the PUCT and ERCOT to conduct a comprehensive verification and audit of all data centers advancing through the interconnection process, stating that any project failing the audit "must be denied." ERCOT suspended the August 7, 2026 Batch Zero classification-notice deadline. If your project is in the ERCOT queue, this is the single most important item on your risk register right now. Confirm current status with ERCOT and your TSP before relying on any published Batch Zero date.

3.4 PJM — no load queue, and a capacity problem instead

PJM has no formal RTO-run large load interconnection queue. Physical connection of an end-use load is executed by the Transmission Owner / Electric Distribution Company under state retail jurisdiction. PJM's roles are load forecasting (via the Load Analysis Subcommittee), RTEP reliability analysis, and — new in 2025–26 — transmission service products and capacity/curtailment rules.

PJM's large load definition: ≥50 MW at a single Point of Interconnection (smaller loads case-by-case at EDC/LSE request).

Why PJM's response has been about capacity rather than studies: the 2027/2028 Base Residual Auction cleared 5.6% short of the reliability requirement — the first shortfall in PJM history. PJM's problem is not "can we study this load"; it is "is there enough generation."

The PJM Board's five-pillar program (January 16, 2026 decision letter):

Load forecasting reform — state regulatory review of large-load submissions, removal of duplicative requests, third-party validation, standardized ramp rates and utilization factors, and a Large Load Registry. In place for the 2027 Load Forecast.
Bring Your Own New Generation (BYONG) via the Expedited Interconnection Track.
Connect and Manage, later renamed Interim Resource Adequacy Service.
Reliability Backstop Procurement.
Holistic market review.

Expedited Interconnection Track (EIT) — approved by FERC June 9, 2026, operational August 2026, sunsets end of 2027:

Curtailment is coming. Electric Distributors must implement load management for new Large Loads lacking sufficient capacity as of June 1, 2027. Reductions occur prior to Pre-Emergency Load Management — meaning new large loads get curtailed before the utility's existing demand response customers. A FERC-approved compensation rate applies to directed reductions. The Board concluded that large loads should remain in the capacity market (removing them would shift costs to existing consumers) while still being curtailed ahead of pre-emergency DR.

Co-location is now settled in outline. FERC's June 18, 2026 order approved PJM's framework: a defined Co-Located Load category, three transmission services (Network Integration Transmission Service, Firm Contract Demand, Non-Firm Contract Demand), a 50 MW threshold based on cumulative nameplate capacity, three-year grandfathering for existing network customers, non-compliance penalties including service termination, and regulation and black start charges assessed on gross demand. FERC directed further compliance on the netting threshold, extension to behind-the-meter generation customers, and treatment of qualifying cogeneration.

Reliability Backstop Procurement (filed July 31, 2026, ER26-3380): an initial target of 6,831.3 MW UCAP, terms up to 15 years, a maximum price of $555/MW-day UCAP, allocated to Electric Distributor zones by Load Adjustment / Peak Load Contribution. The target is reduced by verified new capacity serving new load via bilateral contract or self-supply — an explicit incentive to bring your own generation.

A sobering forecasting reality. A signed utility agreement does not guarantee inclusion in PJM's load forecast. PJM's January 2026 forecast cut the 2027 summer peak by roughly 4 GW and the 2028 summer peak by 4.4 GW, attributing part of the reduction to stricter data center vetting — while revising 10-year growth up to 3.6% per year, reaching roughly 222 GW by 2036. Projects have been discounted heavily or entirely for near-term in-service dates on grounds of infeasible construction and material timelines. If you want to be in the forecast, bring quantified evidence of funding, executed agreements with financial penalties, and disclosure of any duplicate requests.

3.5 MISO — mid-build

MISO has no operating large-load interconnection queue as of August 2026, and is actively building one.

Proposed large load definition: any new commercial or industrial facility or aggregation greater than 50 MW at a single site behind one or more points of interconnection — gross peak demand, new facilities only unless significantly expanded. Draft materials contemplate an effective date around September 1, 2026, with expansions of ≥25 MW to existing facilities already above 50 MW also captured.

Three-stage framework:

Reliability Requirements — consistent study expectations before energization
Speed-to-Service — "Firm Service Step Up," allowing partial service before transmission upgrades complete; also "Speed to Partial Power"
Operational — ongoing monitoring and forecasting

What MISO is proposing to require of you (from Large Load Working Group work products filed in July 2026):

Interconnection Reliability Requirements — operational monitoring, telemetry, equipment controls, remote disconnect capability, and refined definitions for "computational loads" and "large load customers"
Telemetry and PMU — real-time SCADA telemetry (MW, MVAR, voltage, breaker status) via ICCP, plus phasor measurement units, with PMU-capable equipment installed before service commencement
Operational forecasting — hourly forecasts on a 168-hour horizon, updated hourly, plus 5-minute forecasts on a 6-hour horizon, at the Commercial Pricing Node level, with accuracy monitored
Ride-through — large loads must remain connected during specified voltage and frequency disturbances unless disconnection is required for safety or equipment protection
Ramp rate limits — a 20 MW/minute figure has been discussed as an industry reference

Related mechanisms available now:

ZGIA — Zero-Injection Generator Interconnection Agreement — a co-located generation/load solution with study completion in roughly 90 days. As modified, ZGIAs associated with large loads must register as market resources, and large load customers pay transmission service on a gross basis.
ERAS — Expedited Resource Addition Study — for generation , not load, but it is how you get generation online fast to serve your load. Requires 100% site control of generator and POI, written support from the state regulatory authority for a resource adequacy need, an executed off-take agreement, an NRIS request, a $100,000 non-refundable payment, and COD within 3 years. Studies complete and GIAs issue in 3 calendar months; MISO uses reasonable efforts to complete studies within 60 days of kickoff. Caps: 15 requests per quarter, 68 total. PSCAD EMT modeling is required before GIA execution. Verify the current sunset date — published sources conflict.
EPR — Expedited Project Review — an MTEP mechanism for quickly analyzing emerging needs such as large loads.

Scale and timelines: MISO's own materials report large load additions through Expedited Project Reviews going from 0.1 GW in 2024 to a projected 13.1 GW. MISO's stated time-to-serve benchmarks: large loads 1.5–3 years; generation 4 years; transmission 7–10 years. That asymmetry — you can build in two years, the wire takes seven — is the fundamental constraint on every large load project in the country.

3.6 NYISO — a two-stage, split-jurisdiction process that surprises people

The jurisdictional threshold matrix is the single most important operational fact in New York:

Uprates count.

Expanding an existing facility can pull the entire site into NYISO review even if the original load was below threshold.

Stage 1 — NYISO Load System Impact Study.

Pre-filing coordination with the Connecting Transmission Owner to define the POI, CEII clearance for all technical staff, and a conceptual one-line. NYISO schedules a scoping call roughly two weeks after your request is deemed complete. You execute a System Impact Study Agreement within 30 days with a $150,000 deposit (trued up in both directions). The study runs against a year-five base case approved by NYISO's Operating Committee, consistent with the most recent Short-Term Assessment of Reliability. Conservative duration: about 9 months; the transmission owners quote 6–12 months.

Here is the part that surprises developers: the SIS is informational and non-binding.

NYISO assesses reliability impacts and, if adverse impacts appear, identifies potential mitigations. It does not produce good-faith cost-and-schedule estimates. Once the SIS report issues and is paid for, NYISO's role ends.

Stage 2 — CTO Facilities Study and a two-party Interconnection Agreement.

NYISO is not a party. Published transmission owner guidance quotes 9–15 months and $200,000–$500,000 for the facilities study, producing non-binding scopes with ±30% to ±15% cost estimates. Then a Cost Reimbursement Agreement, contribution-in-aid-of-construction security (roughly 10% of total project cost up front, credited against final CIAC), then 2+ years of design, procurement, construction and energization.

And a trap: the system representation in the facilities study may differ from the SIS, surfacing upgrades the SIS never identified. Budget for the possibility.

Connecting Transmission Owners: Central Hudson Gas & Electric, Consolidated Edison, LS Power Grid New York, New York State Electric & Gas, New York Transco, NextEra Energy Transmission New York, Niagara Mohawk (National Grid), Orange and Rockland, Rochester Gas and Electric. NYPA and LIPA/PSEG-Long Island are non-jurisdictional and follow their own procedures, plus NYISO SIS where thresholds are met.

Scale: the NYISO large load queue grew from 6 projects and about 1 GW in 2022 to 48 proposals and about 12 GW as of December 31, 2025, concentrated in upstate Zones D, E and H. The 2026 Gold Book cut near-term large load impact (2026 summer from 1,023 MW to 538 MW) while raising the long-term 2040 projection to 2,880 MW.

Reform in flight.

NYISO has been running its own interconnection reform track — concept discussion February–May 2026, straw proposal June–July, detailed proposal and tariff revisions August, committee approval November, Board approval and FERC filing December 2026 — on top of responding to FERC's show-cause order EL26-69. Separately, NY PSC Case 26-E-0045 ("Energize NY Development") opened February 12, 2026, with a DPS staff white paper due February 12, 2027. Both tracks will change the process described above. Verify current status.

3.7 ISO New England — the smallest pipeline and the least defined process

ISO-NE's tariff contains no defined "large load" category, no load application or study procedures, and no load-specific timelines. FERC said so explicitly in EL26-72, and even observed that ISO-NE's more limited large-load growth "may itself be attributable to the lack of clear processes."

How a large load connects in New England today:

Transmission Owner-led. You contract with the local TO — Eversource, National Grid / New England Power, Avangrid subsidiaries (United Illuminating, Central Maine Power), Versant, Unitil, Green Mountain Power, VELCO. Local service is taken under OATT Schedule 21; regional service is Regional Network Service.
ISO-NE §I.3.9 "Proposed Plan Application" review. The TO submits a PPA on your behalf. ISO-NE reviews for "significant adverse effect on reliability or operating characteristics." The Reliability Committee provides advisory review and votes.
Facilities engineering and construction by the TO.

Eversource's published sequence is representative: Project Initiation → Preliminary Transmission Study (fee-based) → Engineering & Design Agreement, during which an ISO-NE System Impact Study for I.3.9 PPA approval and regional stakeholder review is performed → Full Engineering & Construction → Operation Agreement. Eversource explicitly lists data centers among covered project types.

An important nuance.

ISO-NE Planning Procedure PP5-1 states that no notification or submittal is required for demand resources not comprised of generation. Its enumerated PPA triggers are generation ≥5 MW, reactive changes ≥5 MVAR (unit) or ≥10 MVAR (station), all transmission topology changes at ≥69 kV, RAS/ACS schemes, and interconnections at ≥69 kV with non-Market-Participants. A large load reaches the PPA process principally through the transmission facilities built to serve it — a new ≥69 kV substation or line — not through a load MW threshold. That gap is exactly what FERC identified.

PP5-3 analysis levels (useful for scoping your schedule): Level 0 = no PPA; Level I = informational; Level II = Area Transmission Steady State Assessment and/or Transfer Capability Assessment (about 1–4 months); Level III = Level II plus Area Transmission Stability Assessment and/or Dynamic Transfer Capability Assessment (about 3–12 months).

Scale: roughly 285 MW across two projects in formal study — a 200 MW data center in Northeast Massachusetts and 85 MW of general electrification in Connecticut. Contribution to peak of roughly 110 MW in the 2030s. No effect on system demand before winter 2027/2028. ISO-NE's large load forecast framework, introduced in the 2026–2035 CELT Report, uses a definition of projects >20 MW nameplate that have entered a formal study agreement with a Transmission Owner — different again from FERC's ≥50 MW above 69 kV.

ISO-NE's headline reform proposal is "Bring Your Own New Generation" excluding new large loads from the system forecast used to set the Installed Capacity Requirement, so the capacity market does not procure incremental supply on their behalf. Large loads become responsible for bringing their own energy supply or accepting curtailment. The rules are explicitly modeled on SPP's conditional-service program, with implementing rules targeted for 2027.

3.8 CAISO and California — CAISO does not interconnect load

Let's be precise about California, because this is where the most incorrect assumptions live.

CAISO does not interconnect load. In CAISO's own words: "utility tariffs and their state regulators generally govern the study, interconnection, rates, and cost recovery of new loads."

The division of labor:

Participating Transmission Owners (PG&E, SCE, SDG&E, and municipal PTOs) handle load interconnection, at distribution or transmission level, under CPUC-approved retail tariffs.
CAISO provides concurrence that a load interconnection and its network upgrades meet transmission reliability requirements, incorporates load into the annual Transmission Planning Process using CEC forecasts, runs generation and storage interconnection, and is now developing large-load technical standards.
CPUC approves the utility tariffs, rate design, and cost recovery.
CEC produces the demand forecast, mapped to substation locations, that feeds CAISO's TPP.

A correction worth internalizing: Rule 21 is not the load interconnection pathway. Electric Rule 21 governs generating facility interconnections to IOU distribution systems. It is not how your data center connects.

The actual load pathways in California:

PG&E Electric Rule 30 is the key California development. Filed November 21, 2024 (A.24-11-007), it received interim implementation from the CPUC on July 24, 2025 for applicants who agree to pay for necessary transmission infrastructure work up front. It covers transmission service facilities, interconnection upgrades, interconnection network upgrades, and network upgrades — design, construction, ownership, metering, cost allocation. Applications require a 10-year usage forecast, with applicant advances based on PG&E estimates. The final decision was still pending as of mid-2026; verify. SCE and SDG&E have no verified Rule 30 equivalent.

Flexible Service Connections are the California speed lever. Under SB 410 ("Powering Up Californians Act") and CPUC rulemaking R.24-01-018, PG&E and SCE are being required to offer Standard Offer Flexible Service Connections in tariff, letting customers access grid capacity ahead of completed upgrades by agreeing to defined load profiles — with formalized preliminary capacity assessments and no loss of queue position. Explicitly aimed at large and critical loads.

Rate design is open. CPUC R.26-04-009, opened April 2026, will consider rate design for data centers and other large load customers, coordinated with SB 57 assessment requirements on transmission-connected data center load. Roughly a 24-month proceeding.

CAISO's Large Loads initiative is the live technical track. Launched formally February 27, 2026 after a January 30 issue paper, with a Technical Requirements Straw Proposal (June 15, 2026) and a Large Loads Straw Proposal (August 11, 2026). The published schedule targets a Draft Final Proposal September 21, 2026, Board of Governors October 28, 2026, and a FERC compliance filing November 16, 2026.

What CAISO has signaled it is considering: tiered transmission service (non-firm / interim / firm contract demand) to enable faster speed-to-power during upgrade construction; technical requirements covering voltage and frequency ride-through, post-fault power recovery, ramp rate limits, management of pulsating AI compute loads, telemetry and monitoring, modeling data, commissioning tests, sub-synchronous oscillations, short-circuit levels, protection coordination and power quality; cost allocation for co-located facilities; and 24/7 curtailment capability with defined ramp rates.

Scale: CEC forecasts in CAISO materials range from +1.8 GW to +2.3 GW of data center load by 2030 and +3.3 GW to +4.9 GW by 2035–2040 — small numbers by ERCOT or PJM standards, reflecting California's cost and siting environment.

3.9 Non-ISO regions and the Western large load tariffs

WECC's Assessment of Large Load Interconnection Risks in the Western Interconnection (February 2025) is the best regional diagnostic available. Its findings:

Surveyed large load queue of 44,650 MW, roughly 80% data centers — nearly equal to current Western peak demand
Individual requests routinely exceed 500 MW; 48% of utilities reported requests over 1,000 MW
Many utilities lack formal entry criteria for load interconnection, so speculative requests cause backlogs
Data centers can be designed, permitted and built in 1–2 years versus 2–10+ years for transmission
Modeling gaps: no standardized dynamic load models in phasor or EMT domains; basic static load models miss the dynamic behavior entirely
Ride-through: data center equipment follows SEMI F47 and ITIC curves, which are building electronics curves, not grid curves — so equipment is prone to tripping on normally-cleared faults

WECC's recommendations: mandatory interconnection requirements paralleling FERC generator standards, cluster-based assessment analogous to inverter-based resource procedures, and mandatory data-sharing agreements before energization. Note that WECC does not run an interconnection process — it is the Regional Entity for compliance, standards, modeling and risk assessment.

Purpose-built Western large load tariffs (status varies; verify each):

BPA deserves separate mention because of its scale in the Northwest. The Line and Load Interconnection process governs new or modified Points of Delivery. A draft Version 4 procedure under stakeholder comment in mid-2026 proposed a 20 MW threshold for new load at existing PODs, a 180-day System Impact Study, a 180-day Feasibility Study, and a Facilities Study — with at least one participating utility characterizing the total process as approaching three years.

PART 4 — THE THRESHOLD PROBLEM: WHAT COUNTS AS A "LARGE LOAD"?

Practical rule: a ring bus is fine at 4 positions. Plan the ring so it can be cut over to breaker-and-a-half when the campus expands past 4 positions — retrofitting later is brutal. This is a 30%-stage decision with 10-year consequences.

Section and elevation drawings must show phase spacing, electrical clearances to grade and to fence, and must be designed to maximum system voltage, not nominal. A "138 kV" station is a 145 kV station for clearance, BIL and arrester duty. Governed by IEEE 1427 (clearances), IEEE 605 (rigid bus), and the NESC.

Grounding grid design and study (IEEE 80)

The deliverable contains:

Soil resistivity field data — Wenner four-pin measurements, multiple traverses and orientations, interpreted to a two-layer model
Maximum grid current derivation — worst-case single-line-to-ground fault, split factor (how much returns via shield wires and neutrals versus earth), decrement factor, future growth factor
Conductor sizing from the fusing equation — fault magnitude times clearing time; typical result 4/0 to 500 kcmil copper
Mesh spacing iteration — start conservative, relax to the economic optimum, re-verify each step
Touch and step voltage compliance against 50 kg and 70 kg body criteria
Surface layer credit — four inches of high-resistivity crushed rock materially raises tolerable touch and step voltages
Ground potential rise and transferred potential to fences, communications circuits, and to your data center building steel. This last item is frequently missed and is a genuine hazard when the building sits close to the switchyard
Verification testing per IEEE 81 (fall-of-potential)

Grid optimization routinely yields six-figure construction savings against a study costing a fraction of that. If a developer wants to skip the grounding study to save money, this is the number to show them.

Lightning protection (IEEE 998)

Electrogeometric model (rolling sphere) and/or empirical curves; compute mast heights and protective radii at a chosen exposure. Output is a shielding plan with masts and shield wires plus an exposure calculation per protected element.

Do not conflate this with NFPA 780, which governs building lightning protection. Both apply to a data center campus; they are different standards with different methods.

AC/DC schematics, wiring diagrams, panel schedules, cable and conduit

DC one-line and battery sizing calculation (IEEE 485 for VLA/VRLA, IEEE 1184 for lithium-ion). Some TOs require the DC system sized to power continuous loads for at least 8 hours. Deliverables: battery/charger sizing, DC panel schedules, DC ground detection, DC selective coordination
AC station service one-line — usually two independent sources with an automatic transfer scheme
AC elementary / schematic diagrams per device: trip circuits, close circuits, 86 lockout, 79 reclose, breaker failure initiate, permissive and blocking keying
Wiring (connection) diagrams — panel by panel, terminal-block level
Panel schedules with three-phase balance calculations
Cable schedule — every cable: tag, from/to, type, size, insulation, length, raceway assignment
Conduit, raceway and duct bank drawings with fill calculations, duct bank sections, manhole and handhole details, and Neher-McGrath ampacity derating for large duct banks — critical when routing 34.5 kV feeders across a campus

Protection and control package

This is the deliverable set with the most utility scrutiny. Expect the TO to require:

Protection one-line with all device function numbers per IEEE C37.2
Dual, fully independent primary protection systems — commonly specified as "System 1 and System 2" high-speed phase and ground fault protection on transmission elements, with separate DC supplies, separate CT cores, and separate communications paths
Breaker failure protection (50BF/62BF) on every transmission bus connection, with direct transfer trip
Line protection scheme selection. Line current differential over fiber is preferred by most TOs for new work. Where fiber is unavailable, communication-assisted distance schemes:
POTT (permissive overreaching transfer trip) — Zone 2 overreach keys permission; fails to trip on channel loss, so pair with weak-infeed echo and current-reversal logic
DCB (directional comparison blocking) — a reverse-looking element blocks; secure on channel loss but requires a coordination timer; common on power-line carrier
Relay settings report — the actual deliverable: settings files (SEL .rdb, GE .urs, etc.), time-current coordination curves, a fault study backing each setting, CT saturation checks, and a settings narrative. Coordination is verified in ASPEN OneLiner or CAPE against the TO's model
Trip scheme and DC trip logic drawings, lockout matrix, and transfer trip arrangements — a large load's POI breaker typically must be trippable by the TO
Reclosing philosophy. This is now a first-order design issue for data centers. Automatic reclosing produces exactly the repeated voltage dips that trigger disturbance-counting UPS logic. NERC's incident review explicitly recommends reconsidering automatic reclosing configurations given voltage-sensitive load proliferation. Some TOs require the customer to ride through multiple 50–70 ms reclosing shots with undervoltage pickup at or below 85%. Get this into the interconnection agreement in writing at 30% design.
PRC-005 maintenance program documentation and PRC-002 disturbance monitoring — DFRs, sequence-of-events recording, fault recording. NERC's Level 3 Alert makes fault recording devices an explicit expectation
Remote relay access so the TO can retrieve oscillography

Metering and revenue metering

Metering one-line and three-line, plus metering compartment drawings showing front, side and rear views with CT orientation on the bus work and electrical clearances
CT and PT ratios and accuracy classes. Representative utility benchmarks:
CTs: 0.3% class minimum; 0.15% class with extended accuracy range at 230 kV and above, with rating factor up to 3.0
VTs: 0.3% or better per IEEE C57.13.6, with certified ratio and phase-angle correction test certificates
Overall metering accuracy error ≤ ±1%, computed as a Gaussian combination of CT, VT, meter and data conversion error
Ratio selection so that maximum load sits at 80–95% of CT rating
The phased-load metering trap. On a three-phase campus buildout, Phase 1 load may be 20% of the ultimate CT rating, and metering accuracy collapses at low current. Either design the CT for Phase 1 accuracy or specify extended-range CTs. This gets missed constantly and shows up as a revenue dispute two years later.
Meter accuracy per ANSI C12.20; instrument transformers per IEEE C57.13
Loss compensation calculations if metering is on the low side of the transformer — requires certified transformer test data (no-load losses, load losses, %Z)

SCADA, telemetry, RTU and ICCP

This is where 2026 requirements escalated most sharply. Deliverables:

RTU/gateway points list — analog and digital, with scaling, deadbands and update rates
Communication one-line — fiber path, redundancy, protocol (DNP3 to the TO, ICCP TO-to-ISO), demarcation, cybersecurity boundary
ICCP points list if the ISO takes data directly

Typical required point content:

Analog MW and MVAr at all generation, load, and transmission line terminals
Analog kV at all buses ≥69 kV
Circuit breaker open/close status for all breakers
Device status points (auto-reclose on/off, etc.)
Bidirectional real power with at least two decimal places

Region-specific telemetry expectations worth designing to:

Design the POI with a remotely operable disconnect and a control path for it. In ERCOT this is already law (SB6 curtailment equipment and the co-location kill switch). Everywhere else it is coming.

Transmission line, structures and right-of-way

Plan-and-profile drawings (typically 1"=100' horizontal, 1"=10' vertical), PLS-CADD model, sag-tension analysis
Structure family drawings, loading trees per NESC Grade B and ASCE 74 — including extreme wind and extreme ice cases
Foundation design — drilled pier or caisson design from geotechnical borings
ROW plats, easement exhibits, crossing permits (railroad, highway, waterway, other utilities), and an encroachment matrix
Tap versus loop-in decision. Note that some utilities prohibit new taps on lines at or above 100 kV, requiring interconnection at a substation. That single constraint can move a site
Line arrester application, shield wire and OPGW design, and the OPGW splice and routing plan for differential protection and SCADA

Bill of materials, specifications, ratings sheets

Equipment specifications in CSI MasterFormat Division 26/33
Ratings sheets: transformer (MVA stages, %Z, tap ranges, LTC, sound level, BIL), breaker (continuous, interrupting, TRV, close-and-latch), switch, arrester, CT/PT, relays
Short-circuit withstand documentation and system impedance data on a 100 MVA base (Z1%, Z0%) — many TOs require this from you
Nameplate data, connection diagrams, polarity designations, and factory test results supplied in advance of installation
Submittal register keyed to specifications (a 90% deliverable)
As-builts in the TO's required CAD format

PART 7 — THE STUDIES

dominated by switching-frequency and inter-harmonic content, which is exactly why the inter-harmonic clause exists. Get this from your UPS and PSU vendors, not from a textbook table.

Flicker matters little for steady IT load and a great deal for chiller motor starting and for AI training pulse loads in the 0.1–30 Hz band, which land squarely on the sensitive region of the flicker curve.

7.7 Ride-through and the load-loss problem

This is the single most consequential technical requirement in the guide, and it is a client-side design requirement, not just a study.

ERCOT NOGRR282 / NPRR1308 — Large Electronic Load ride-through (applies where ≥50% of site demand is power-electronic computational load):

Plus, in ERCOT: continue consuming active power within 10% of pre-disturbance level during required ride-through; current must not exceed 125% of maximum normal operating current; sag-counting and instantaneous trip schemes are prohibited — filtered measurement with a window of at least one cycle is required; and settings must reflect actual equipment tolerances, not merely the minimum requirement. Initial compliance ran from November 14, 2025, with stricter 0.5–0.8 pu requirements from January 1, 2028. Non-compliance triggers a 90-day root-cause investigation, a 90-day remediation plan, and 180 days to implement. ERCOT may order immediate disconnection for imminent reliability risk.

SPP HILL fault ride-through covers similar ground with two distinctive requirements:

Constant-current control is mandatory during disturbances; constant-power control is prohibited. This is easily missed and it matters enormously — constant-power control makes a voltage sag worse by drawing more current as voltage falls
The load must tolerate six voltage fault-clearing attempts within 90 seconds and recover to ≥90% of pre-disturbance consumption within 1 second of voltage recovering to 0.9 pu
Variable-speed drives may be exempted from the constant-current requirement and some duration requirements if the study demonstrates no adverse reliability effect

Only SPP and ERCOT publish explicit numeric POI-referenced curves. Everywhere else, the ride-through envelope is negotiated in the interconnection agreement. Our recommendation: propose the ERCOT or SPP curve as the reference standard in your own agreement, so you design once rather than retrofitting when your region catches up.

The design implications for your facility are severe and specific:

Disturbance-counting logic — "transfer to backup after N dips in T seconds" — is the single biggest identified cause of the 1,500 MW Virginia event. It must be reviewed, and usually re-tuned or removed
Diesel rotary UPS (DRUPS) produced extended disconnection requiring manual reconnection in the observed event — the worst of the three behaviors documented
Static UPS, both centralized and rack-level distributed, recovered automatically — the best behavior
Your equipment is designed to ITIC/CBEMA and SEMI F47 curves. Those are building electronics immunity curves, not grid ride-through curves. They do not match transmission requirements, and the mismatch is the root of the entire problem

The study deliverable your engineer should produce: a fault-scenario load-loss quantification. For a set of representative faults and reclose sequences on your POI lines, how many MW of the campus drops, for how long, and does the grid survive it. NERC recommends exactly this analysis. It also feeds directly into the reconnection ramp rate clause of your interconnection agreement.

7.8 Reactive power, power factor, capacitor and STATCOM sizing

Utility power factor requirements at the POI vary: 0.97 lagging to 0.99 leading is a published example; others simply require operation "close to unity."
Check both extremes, not just full load. A modern data center may sit leading at light load because of power-factor-correction and filter capacitance
Deliverable: a reactive compensation study producing MVAr requirement versus load level, device selection (fixed capacitor bank, switched bank with steps, harmonic filter bank, SVC, or STATCOM), and confirmation that the device does not create resonance — tied back to the harmonic frequency scan
STATCOM is increasingly the right answer for AI campuses, because the load swings sub-second and mechanically switched capacitors cannot follow. STATCOM also provides fast voltage support that helps ride-through
Sizing inputs: the QV margin from the stability study, the flicker limit, and the MVAr band the TO requires

7.9 Arc flash, coordination, and inside-the-fence studies

These sit behind your POI but on the same critical path:

Short circuit → coordination (TCC) → arc flash, in that order, per IEEE 1584-2018 (empirical model, electrode configuration VCB/VCBB/HCB/VOA/HOA, enclosure size correction), with results labeled per NFPA 70E
Incident energy at every medium- and low-voltage work location, PPE category, arc flash boundary, printed labels
Selective coordination, including DC coordination for the UPS battery system
Load flow and voltage drop for the internal 34.5 kV → 13.8 kV → 480 V → rack path
Tools: ETAP, SKM PowerTools, EasyPower, PowerFactory

7.10 Facility study, cost estimate, network upgrades, cost allocation

Facility study deliverable: scope, single-line, general arrangement, schedule, and an AACE Class 3 or Class 2 cost estimate for the TO-side interconnection facilities and network upgrades, with a schedule showing long-lead equipment
Split into Interconnection Facilities (customer-funded, directly assigned) versus Network Upgrades (cost allocation varies by region — and FERC's 2026 show-cause orders push hard toward 100% large-load funding with no socialization). PJM's Expedited Interconnection Track already imposes 100% network upgrade cost with no cost-sharing
Include a contingent facilities list and a restudy trigger table

7.11 Curtailment, load flexibility, and loss-of-load studies

New in 2026, and increasingly the fastest path to energization:

Curtailment analysis — how many hours per year, how many MW, under what triggers. PJM's Connect and Manage framework curtails large load before emergency demand response is activated. ERCOT's Provisional Controllable Load Resource pathway trades curtailability for capacity above your firm allocation. SPP's CHILL does the same
FERC's rulemaking contemplates 60-day studies for loads willing to curtail — the single largest schedule lever available to a developer
Flexible load capability study deliverable: MW curtailable, response time, duration, notification lead time, recovery ramp, number of events per year, and — critically — the control path (SCADA setpoint, DR aggregator, on-site generation or BESS dispatch). It must be technically demonstrable. Grid operators are getting skeptical of paper flexibility
Ramp rate limits are becoming binding. Published and proposed figures cluster around 20 MW/minute. Design your workload orchestration and your on-site storage to meet whatever number your region lands on

PART 8 — SOFTWARE AND MODEL SUBMISSION

PART 9 — WHAT YOU MUST GIVE YOUR ENGINEERING FIRM

PART 11 — STANDARDS REFERENCE

11.1 IEEE — substation and physical design

11.2 IEEE — power quality and performance

11.3 IEEE — apparatus (C37.x and C57.x)

11.4 NERC Reliability Standards

Two current NERC items you must track:

The Level 3 "Essential Action" Alert on computational loads (May 2026), applicable to loads ≥20 MW connected at 60 kV containing more than 1 MW of IT load. Its seven actions — modeling data requirements, system studies, qualified-change definitions, commissioning process, protection coordination, dynamic fault recording, and operational communication — are already flowing into what your TO asks of you
NERC Project 2026-02, "Reliability Standards to Address Computational Load — Phase I." New glossary definitions for "Computational Load" and "Computational Load Entity," Rules of Procedure changes creating a registered entity category, and a consolidated standard. Board adoption targeted December 2026, aligned to FERC's December 31, 2026 filing deadline. If this lands as drafted, data centers become NERC-registered entities for the first time — with direct compliance obligations, audits and potential penalties.

11.5 Other standards

PART 12 — WHERE PROJECTS DIE

Twelve failure modes we see repeatedly. Every one is avoidable.

1. Picking the site before screening the POI. Land, water, fiber and tax abatement get evaluated; grid headroom does not. The site is bought, and then the interconnection study says the nearest available capacity is 40 miles and $80 million away. POI screening is the cheapest de-risking available and it must happen before land commitment.

2. Submitting an incomplete application. Deficiency cure periods are 10 to 15 days. In a batch process, missing one drops you to the next batch — six months to a year.

3. Wrong model version. A PSS/E case in the wrong version is a rejected submission. Check the version requirement at the start of every cycle, not from last year's notes.

4. Generic models where hardware-specific models are required. ERCOT explicitly states generic PSCAD library blocks are not acceptable for large loads. Getting a hardware-benchmarked UPS model out of a manufacturer takes months of commercial negotiation. Start that conversation at application, not at study.

5. Not disclosing duplicate requests. PJM removes undisclosed duplicates from the load forecast. Texas SB6 requires disclosure. The practice of shopping the same nameplate across five utilities to see who moves fastest is now being explicitly policed.

6. Ignoring the geotech scope. Standard geotechnical scopes do not include Wenner four-pin soil resistivity. Discovering that when the grounding study starts means a remobilization and a schedule hit for a line item that costs almost nothing if requested up front.

7. Treating ride-through as an IT problem. Your UPS vendor designed to ITIC and SEMI F47. Your grid operator requires a transmission ride-through envelope. These are different curves. The gap analysis has to happen at design, because retrofitting UPS control logic across a built campus is enormously expensive.

8. Leaving disturbance-counting logic in place. "Transfer to backup after three dips in 60 seconds" is what turned a normally-cleared fault into a 1,500 MW event. It is now a specific review item. Find it, and change it, before commissioning.

9. Sizing CTs for ultimate load only. Phase 1 at 20% of ultimate CT rating produces metering accuracy that will not survive a revenue audit.

10. Designing a ring bus you cannot expand. Four positions is the practical ring bus limit. If your campus roadmap goes to eight, lay out for a breaker-and-a-half conversion at 30% design.

11. Underestimating long-lead procurement. Design is not the critical path. Large HV transformers and HV breakers are. Specifications must be issued for procurement at 60% design.

12. Promising flexibility you cannot deliver. Grid operators are increasingly skeptical of paper curtailability. If you commit to 100 MW of curtailment in 10 minutes, you need a demonstrable control path — SCADA setpoint, workload orchestration, on-site generation or BESS — and you will be tested on it. Emissions permit hour limits on backup gensets frequently make the committed flexibility undeliverable in practice.

ABOUT KEENTEL ENGINEERING

Keentel Engineering is a power system and substation engineering firm serving utilities, IPPs, data center developers and industrial clients across the United States. We provide end-to-end POI interconnection support — from feasibility and site screening through dynamic modeling, protection design, and final energization.

Our POI Interconnection Services

Interconnection strategy and screening

Multi-site POI feasibility screening and ranking
Regional process and tariff analysis (PJM, MISO, SPP, ERCOT, CAISO/WECC, NYISO, ISO-NE, and non-ISO utilities)
Interconnection application preparation and data package assembly
Deficiency response and grid operator coordination

Power system studies

Steady-state load flow: N-0, N-1, N-1-1 across seasons and ramp phases
Short circuit, fault duty and breaker duty (ANSI and IEC)
Transient stability, voltage stability, PV and QV analysis
Dynamic load modeling — CMLD, PERC1 and user-defined models
EMT / PSCAD studies: sub-synchronous oscillation, control interaction, insulation coordination, switching transients
Harmonic and power quality analysis (IEEE 519), flicker (IEEE 1453)
Ride-through assessment and fault-scenario load-loss quantification
Reactive compensation and STATCOM/SVC sizing
Arc flash (IEEE 1584), coordination, and grounding (IEEE 80)

Modeling and model submission

PSS®E, PSCAD, ETAP, SKM PowerTools, DIgSILENT PowerFactory, CYME, EasyPower
Model quality testing, cross-platform benchmarking, and validation reports
MOD-032 data package preparation
First-submission acceptance is our benchmark, not a stretch goal

Substation and transmission design

Substation primary and secondary design, conceptual through IFC
One-line, three-line, general arrangement, sections and elevations
Protection and control design, relay settings, coordination and commissioning support
Grounding grid and lightning protection design
Revenue metering, SCADA, RTU and ICCP points development
Transmission line design and routing
IEC 61850 substation automation

Compliance and owner's engineering

NERC O&P 693 compliance services
NERC Level 3 Alert engineering support — modeling parameter response, study updates, commissioning procedures, fault recording, and protection coordination
PRC-019, PRC-024, PRC-025 and PRC-027 documentation
Owner's engineer services: specification, bid evaluation, design review, FAT/SAT witnessing, commissioning oversight

Adjacent capabilities

Utility-scale solar, wind and BESS interconnection
MEP engineering
Nuclear power plant services
Transmission and distribution engineering across ComEd, PJM and other territories

Regional track record

Keentel has delivered interconnection engineering across PJM (230 kV), ERCOT (345 kV), WECC (230 kV) and NYISO (138 kV) systems, including hybrid solar-plus-storage, wind, and BESS facilities. Representative outcomes include full first-submission acceptance of dynamic and short-circuit models, zero requests for information following initial package review, and sub-synchronous resonance risk assessment with EMT modeling on weak-grid interconnections.

Software depth

ETAP (15+ years) · PSS®E (14+ years) · SKM PowerTools (15+ years) · AutoCAD (27+ years) · CYME (27+ years) · SEL (27+ years) · Bentley MicroStation (14+ years) · EasyPower (10+ years) · DIgSILENT PowerFactory (8+ years) · PSCAD (5+ years)

Contact

Head Office — Tampa, FL 400 N Ashley Dr, Suite 2600, Tampa, FL 33602 · 813-389-7871

Austin, TX — 5900 Balcones Drive, Suite 100, Austin, TX 78731 · 512-591-0752 Sacramento, CA — 1401 21st St, Suite R, Sacramento, CA 95811 · 916-913-4524 Baltimore, MD — 306 W Redwood St, Suite 200, Baltimore, MD 21201 · 410-225-2181

Email: contact@keentelengineering.com Web: keentelengineering.com

Talk to us before you buy the land. POI screening across candidate sites is the cheapest risk reduction available in data center development, and the last point at which changing your answer is free. Schedule a consultation at keentelengineering.com.

SOURCES AND VERIFICATION NOTES

Primary sources

Federal and reliability

FERC Docket RM26-4, Interconnection of Large Loads to the Interstate Transmission System — ferc.gov/rm26-4
FERC Section 206 show-cause orders, June 18, 2026: PJM EL26-67, SPP EL26-68, NYISO EL26-69, MISO EL26-70, CAISO EL26-71, ISO-NE EL26-72
FERC order on PJM co-located load, EL25-49 series
FERC order directing NERC computational load standards, RD26-7, July 16, 2026
FERC Order No. 2023 / 2023-A (generator interconnection reform); Order No. 1920 / 1920-A (transmission planning and cost allocation); Order No. 845 (surplus interconnection service)
NERC, Incident Review: Simultaneous Voltage-Sensitive Load Loss , January 8, 2025
NERC Level 3 "Essential Action" Alert on computational load, May 2026
NERC Reliability Guideline, Risk Mitigation for Emerging Large Loads , 2026
NERC white paper, Characteristics and Risks of Emerging Large Loads
NERC Large Loads Action Plan and Project 2026-02 — nerc.com/initiatives/large-loads-action-plan
WECC, An Assessment of Large Load Interconnection Risks in the Western Interconnection , February 2025
WECC Data Preparation Manual and Base Case documentation — wecc.org

Grid operators

PJM Manual 14H (New Service Requests Cycle Process), Manual 14B, 14C, 14G; PJM Dynamic Model Development Guidelines; PJM EMT Model Development Guidelines; PJM Board CIFP-LLA decision letters (January and July 2026); PJM OATT — pjm.com
MISO Tariff Attachment X, Business Practice Manuals, Planning Modeling Manual, ERAS Informational Guide, Large Load Working Group materials, Transmission Cost Estimation Guide — misoenergy.org
SPP HILL Integration materials, DISIS Manual, Consolidated Planning Process Manual, HILL Fault Ride-Through Requirements v1.0, EMT Model Requirements — spp.org
ERCOT Planning Guide Sections 5, 6 and 9; NPRR1234/PGRR115; PGRR144; PGRR145/NPRR1325; NOGRR282/NPRR1308; Large Load Integration page and forms; DWG Procedure Manual — ercot.com
CAISO Large Load Considerations Issue Paper (January 2026), Large Loads Technical Requirements Straw Proposal (June 2026), Large Loads Straw Proposal (August 2026), EMT Modeling Requirements, Appendix DD (GIDAP), CAISO Tariff — caiso.com
NYISO OATT Attachment P, Attachment HH, Manual 23 (TEI Manual), Gold Book, Power Trends — nyiso.com
ISO-NE Transmission Markets and Services Tariff §I.3.9, Planning Procedures PP5-0/5-1/5-3/5-6, Interconnection Request Technical Data Submittal Guidance, CELT Report — iso-ne.com

State and utility

Texas SB6 (89th Legislature, 2025); PUCT Project 58481 / 16 TAC §25.194
CPUC A.24-11-007 (PG&E Electric Rule 30), R.24-01-018 (energization / SB 410), R.26-04-009 (data center rate design)
NY PSC Case 26-E-0045, Interconnection Reforms for Large Loads
Transmission owner FAC-001 Facility Interconnection Requirements documents (Dominion Energy Virginia, FirstEnergy, BPA, and others)
BPA Line and Load Interconnection procedure and Metering Application Guide
EEI list of large customer projects and tariffs

Cost and analysis

MISO Transmission Cost Estimation Guide (MTEP series)
Lawrence Berkeley National Laboratory interconnection cost studies

Verification notes — read before publishing or relying on specific figures

This guide was compiled on August 20, 2026 from primary regulatory documents and reputable secondary analysis. The following items were identified as uncertain, in dispute between sources, or in active flux, and should be verified against the current primary source before you rely on them commercially:

FERC show-cause responses filed August 17, 2026 in EL26-67 through EL26-72, and whether FERC granted the 90-day abeyance motions filed by several RTOs and transmission owner groups in early August 2026. Several regional processes will change materially depending on the outcome.
ERCOT Batch Zero schedule following the August 3, 2026 gubernatorial audit directive and the suspension of the August 7, 2026 classification deadline. Confirm current dates with ERCOT and your TSP.
PUCT 16 TAC §25.194 — adoption status and the per-MW security and fee figures. Published summaries conflict between $50,000/MW and $100,000/MW. Pull the rule text from PUCT Project 58481.
PJM Expedited Interconnection Track readiness deposit — published sources report $10,000, $15,000 and $20,000 per MW in different documents. Verify against the accepted tariff sheets.
MISO large load tariff filings made in late July 2026 (reliability requirements, telemetry/PMU, operational forecasting) — docket numbers and acceptance status.
MISO ERAS sunset date — sources conflict between August 31, 2027 and December 31, 2028.
MISO numerical ride-through envelopes for large loads — themes are published; specific values were not located in public documents.
SPP HILL fault ride-through numerical values — verify against SPP's HILL Fault Ride Through Requirements v1.0 directly.
ISO-NE currently in-effect PSS/E, PSCAD and ASPEN OneLiner version numbers — ISO-NE deliberately does not publish fixed versions. Contact irtt@iso-ne.com.
NYISO software version numbers and the precise tariff location of load interconnection procedures. FERC's own finding in EL26-69 is that NYISO's load procedures are not described in the tariff with sufficient clarity — much of the process lives in manuals and practice.
CAISO Large Loads Straw Proposal (August 11, 2026) contents and the resulting Draft Final Proposal — the most important open document for California.
PG&E Electric Rule 30 final decision status.
ERCOT queue totals — published figures range from 410 GW to 474 GW depending on source, date and accounting method. Cite the specific source and date each time.
Third-party queue trackers are not official grid operator data. ERCOT's own Large Load Portal was not live as of mid-2026.
Transmission owner rosters, cost benchmarks, acreage rules of thumb, stage-gate deliverable matrices, engineering fee percentages and total engineering durations are industry practice compiled from multiple sources, not published rules. They vary by transmission owner, region and contract.

Disclaimer. This guide is provided for general information and does not constitute legal, regulatory, or engineering advice for any specific project. Interconnection requirements are project-specific, region-specific, and changing rapidly. Nothing here creates an engineer-client relationship or a warranty of any kind. Verify all rules, thresholds, deadlines and figures against the current tariff, protocol, guide or rule before making commercial decisions. Keentel Engineering would be glad to help you do exactly that.

Solar + BESS Shared Bus Resonance & Harmonics Guide

Thu, 20 Aug 2026 18:07:26 GMT

Aug 20, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Shared Bus, Hidden Resonance

1. Why a shared bus goes unstable

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Co-locating a battery with a solar array on one 480 V bus creates an electrical problem that no single piece of equipment owns — and that neither manufacturer will solve for you. Here is what actually has to be calculated before the transformer is ordered.

Tesla addresses this directly in the Megapack 2 XL Design and Installation Manual at §2.9.1, and the requirement is unusually concrete. Excessive resonance is expected — their word — when a Megapack shares a common 480 V AC bus with power electronics meeting both of two conditions:

The total apparent power rating of the power electronics exceeds 20 % of the Megapack’s apparent power rating.
The equivalent switching frequency at the AC bus is less than 35 kHz.

Both conditions are met on essentially every co-located solar-plus-storage site built today. A 2-hour Megapack 2 XL has a 2,400 kVA inverter rating, so the 20 % threshold is 480 kVA — less than three utility-scale string inverters. And no utility-scale inverter on the market switches anywhere near 35 kHz; silicon-carbide designs have raised switching frequencies considerably but remain an order of magnitude below that ceiling.

Where both conditions are met, three remedies are permitted: relocate the equipment, install a line reactor of at least 0.022 mH per phase, or install an isolation transformer with impedance between 3 % and 7 %. In both cases the mitigation must carry the continuous current of the connected equipment and must sit in series between the Megapack and the common AC bus.

A location detail worth reading twice

The manual places the mitigation between the battery and the bus. Many designers place it between the solar switchgear and the bus instead, which is often the better engineering choice and interrupts the same resonant loop. It is nonetheless a documented deviation, and it costs nothing to obtain written concurrence from the manufacturer’s applications engineering before the purchase order is placed. Warranty positions have been lost over less.

Every grid-forming or grid-following inverter has an output filter, almost always an LCL arrangement, whose job is to keep switching-frequency ripple out of the network. That filter is a resonant circuit. On its own, connected to a stiff grid, its resonant peak sits at a frequency the inverter’s own control loop is designed to damp, and nothing interesting happens.

Put a second inverter system on the same bus and the picture changes. The two filters are now coupled through the bus impedance, which is small. The combined network has resonant frequencies that neither manufacturer modelled, because neither manufacturer knew what else would be connected. If one of those frequencies happens to land near a harmonic order that either system produces — the 11th, the 13th, somewhere in the low hundreds of hertz to low kilohertz — the network amplifies it rather than damping it.

Amplification of a few per cent of harmonic current into something much larger is what produces the heating, the alarms and the spurious relay operations. The magnitude depends on how sharp the resonant peak is and how much damping the network provides, which is why two apparently identical sites can behave completely differently.

The part that catches people out

Resonance is worst at light load, not full load. Absolute harmonic current from switching ripple is roughly independent of real power output, so a site at 25 % output is injecting nearly the same absolute harmonic current into a network with far less damping from load. Sites that are perfectly stable at midday can misbehave at breakfast.

2. What the battery manufacturer requires

3. Reactor or transformer

A solar array and a battery sharing a common 480 V collection bus is now the default architecture for co-located generation. It is efficient, it saves a transformer, and it works — until the two sets of power electronics start talking to each other through the bus impedance.

The failure mode does not announce itself. There is no bang, no trip on day one. What happens instead is a slow accumulation of symptoms that get attributed to everything else: nuisance inverter faults during light-load mornings, capacitor heating, harmonic alarms that the SCADA system logs and nobody reads, a protection relay that operates for no fault anyone can find. Sites run for months in this condition. Some run for years, quietly losing production and equipment life.

The cause is resonance between the output filters of two independent inverter systems that were never designed to share a bus. And the fix — an isolation transformer or a line reactor — is cheap relative to the consequences, provided it is specified correctly. That last condition is where most of the difficulty lives.

5. The grounding consequence nobody budgets for

A conclusion you can reach without measuring anything

Even in the physically unrealisable worst case, the required K-factor is 3.25. Any realistic distribution across the low orders gives between 1.0 and 1.6. K-4 is adequate under every spectral distribution consistent with a published sub-3 % distortion figure, and escalating beyond it buys physical size and cost with no technical basis. This is a bounding argument, and measurement cannot overturn it.

Assume the transformer is the chosen route. Three specification errors recur often enough to be worth naming, and all three survive vendor review because the vendor builds what the specification says.

4.1 Nobody says which winding faces the array

A specification that reads "Primary 480 V delta, secondary 480 V wye" is ambiguous in a photovoltaic application, because real power flows from the array toward the utility while transformers are conventionally described from the utility side. A vendor can reasonably build that unit with the delta facing either direction.

The direction is not a detail. Standard transformerless string inverters require a grounded wye source facing the inverters — the system’s neutral must be grounded so that line-to-ground voltage is defined and bounded. Certain variants require the exact opposite, a floating ungrounded source. Get the orientation backwards and either the inverters lose the reference they need, or a variant that must float finds itself grounded.

Write it physically, not by label

State on the purchase order and confirm on the approval drawings: wye winding facing the solar switchgear, neutral brought out and grounded at a single point; delta winding facing the medium-voltage transformer.

Do not rely on the words primary and secondary to carry that meaning, because they do not.

4.2 Sizing on real power divided by power factor

The familiar calculation is S = P ÷ PF. For a 2 MW site at 0.90 power factor that gives 2,222 kVA, rounded up to a 2,500 kVA standard rating. The arithmetic is correct and the answer often happens to be right, but the reasoning does not apply to an inverter block.

String inverters are apparent-power limited. A 200 kVA unit with adjustable power factor delivers 200 kVA total; at 0.90 PF that is 180 kW of real power, not 200 kW plus reactive on top. Ten of them total 2,000 kVA and deliver 1,800 kW at 0.90 PF. There is no operating point at which 2,222 kVA flows.

The correct basis is aggregate inverter apparent power. It matters because the same faulty reasoning applied to a 1.5 MW block gives 1,667 kVA and invites a 1,500 kVA transformer that is exactly at its rating with zero margin — for a continuous-duty source, on cold clear days, with no allowance for future inverter addition.

4.3 Buying more K-factor than physics requires

K-factor is the standard way of rating a transformer for harmonic heating, and it is routinely over-specified. The definition, per UL 1561, sums over all orders including the fundamental, with harmonic current expressed in per unit of rated RMS load current:

K = Σ ( I h, pu )² × h²

The order-squared weighting looks alarming and drives people toward K-13 or K-20. But the fundamental term dominates the sum, so a fixed quantity of distortion produces far less variation in K than the weighting suggests. Take a modern silicon-carbide inverter with published total current distortion under 3 % at rated load, and place all of that distortion at a single harmonic — the least favourable case that can physically exist:

4. Three ways the specification goes wrong

One row deserves expansion. A reactor at the manufacturer’s minimum of 0.022 mH presents 8.3 mΩ at 60 Hz, which on a 2,500 kVA, 480 V base is roughly 9 % impedance — larger than a typical transformer’s 5.75 %. The reactor is not a small impedance. But it sits in the battery branch, so it does nothing for fault duty or incident energy at the solar switchgear, which is where people are working.

The reactor is dramatically cheaper. A 0.022 mH three-phase reactor is a fraction of the cost, size and lead time of a 2,500 kVA dry-type transformer, and it satisfies the manufacturer’s stated requirement. On a project where budget is under pressure, it is the obvious answer.

It is also, frequently, the wrong one — not because it fails to solve the resonance problem, but because it solves only the resonance problem. Co-located sites usually have two or three other issues in the same part of the single-line, and the transformer disposes of all of them at once.

Install a transformer with a grounded wye winding facing the array and you have created a separately derived system under NEC Article 250.30. This is not optional and it is not free. It requires a system bonding jumper at one point only, a grounding electrode conductor sized to Table 250.66 and run to a grounding electrode at the transformer or first disconnecting means, and a supply-side bonding jumper.

None of that exists before the transformer is installed, because the array’s ground reference previously came from the medium-voltage transformer’s secondary. Adding the isolation transformer moves the reference, and every downstream assumption about ground-fault detection moves with it.

There is a second-order trap here that catches experienced installers. Standard string inverters commonly ship with the neutral terminal bonded to the ground terminal at the factory. Establish a new single-point bond at the transformer and every inverter’s internal bond becomes an additional reference tied in through the equipment grounding conductor. Manufacturers generally publish a procedure for lifting that internal bond; whether it should be lifted is a per-site determination under NEC 250.30 and 250.6, and it should be settled with the manufacturer in writing rather than decided on a ladder.

6. The question the inverter manufacturer cannot answer

A recurring exchange on these projects goes roughly as follows. The designer asks the inverter manufacturer whether it is acceptable for the neutral to be established by a separate grounding transformer rather than bonded at the inverter. The manufacturer replies that a grounded wye transformer is required. Both parties believe the question has been answered. It has not.

What the inverter requires is a grounded wye source — a defined, bounded ground reference. A zig-zag or grounded wye-delta grounding transformer provides exactly that. So the answer is yes, in principle, subject to one condition that the inverter manufacturer is in no position to evaluate: the resulting system must be effectively grounded at the inverter terminals.

Per IEEE C62.92.1 that means X0/X1 not greater than 3 and R0/X1 not greater than 1, giving a coefficient of grounding of 80 % or less of nominal line-to-line voltage — equivalently, line-to-ground voltage on the unfaulted phases held within roughly 1.38 per unit of nominal line-to-neutral during a ground fault. A grounding transformer of inadequate rating or excessive zero-sequence impedance will not meet it, and the inverter is then exposed to a temporary overvoltage it was never rated to withstand.

Whether any particular arrangement qualifies is a symmetrical-component calculation on your system. The manufacturer cannot perform it, because the answer depends on impedances they have never seen.

A related point about delta windings

A grounded-wye winding presents low zero-sequence impedance only when it is backed by a delta winding or a delta tertiary. A wye-to-grounded-wye unit without one presents high zero-sequence impedance and will generally fail the X0/X1 criterion.

This is worth checking on the existing medium-voltage transformer: a YNyn0 vector group with no tertiary means the site has no zero-sequence trap anywhere between the inverters and the utility, and triplen harmonics pass straight through. The nameplate will tell you in about ten seconds.

7. What calculation settles, and what needs a meter

Projects stall when this boundary is unclear, usually because someone has been told that nothing can proceed without a power-quality survey. That is not true, and the distinction is worth being precise about, because the equipment decision does not depend on measurement at all.

Everything governing the purchase order is in the top block. A desktop study settles the equipment decision completely. Measurement is required for the utility submission and for quantifying resonance excitation — genuinely necessary work, but not on the critical path to ordering steel.

If you are recording the data yourself

Get the measurement specification written before anyone deploys an instrument. IEEE 519 compliance is assessed statistically — twelve-cycle windows aggregated to three-second values, two hundred of those aggregated to ten-minute values, evaluated at the 95th and 99th percentile over a continuous seven-day period, orders 2 through 50, with instrumentation to IEC 61000-4-7 and IEC 61000-4-30 Class A.

A seven-day recording made with the wrong instrument class or the wrong aggregation cannot be repaired in analysis. It has to be done again.

8. Before the purchase order goes out

Confirm per-site AC capacity against the inverter schedule, not the project name — drawing revisions lag capacity changes more often than not
State winding orientation physically: which winding faces the array, and where the single grounding point is
Confirm the inverter variant, standard or floating, because they impose opposite grounding requirements
Read the existing medium-voltage transformer nameplate for the vector group and check whether a zero-sequence trap exists anywhere
Verify impedance against the battery manufacturer’s band, including the ±7.5 % manufacturing tolerance permitted by IEEE C57.12.01
Size the K-rating on a bounding calculation and resist escalation without a supporting number
Obtain written concurrence from the battery manufacturer if the mitigation sits anywhere other than where the manual puts it
Budget the separately derived system work — electrode, conductor, bonding, and revised ground-fault detection
Settle the inverter internal neutral-to-ground bond in writing with the manufacturer
Check whether the specified efficiency standard actually applies; a 480–480 V unit may qualify as special-purpose, which affects cost, size and lead time

9. Case studies

Note on these cases

The following are anonymised engineering scenarios drawn from co-located solar-plus-storage work. Client names, locations and identifying details have been removed, and figures have been rounded or adjusted where necessary to prevent identification. They are presented to illustrate the engineering, not as a client reference list.

Case 1 — The transformer that would have been built backwards

Situation.

A cooperative was developing four co-located solar-plus-storage sites and had reached the point of quoting isolation transformers to satisfy the battery manufacturer’s resonance requirement. A specification had been drafted internally and circulated to vendors: 1,500 kVA, dry type, 480 V delta primary, 480 V wye secondary, K-4, electrostatic shield, 5.75 % impedance, NEMA 3R. The intent was to install between the solar array switchgear and the existing medium-voltage transformer.

The specification was, on its face, competent. Impedance sat within the manufacturer’s band. The K-rating was reasonable. The shield was correctly called for. Nothing in it would have caused a vendor to raise a query.

What the review found.

Three issues, in ascending order of consequence.

The capacity was wrong for three of the four sites. At 480 V, 1,500 kVA is 1,804 A. A 2 MW inverter block draws 2,406 A continuously. The figure appeared to have been derived from the smallest site, where it was an exact match with zero margin, and then applied uniformly. Two of the four sites would have received a transformer overloaded by a third from the day of energisation.

The medium-voltage transformer nameplate — obtained as a photograph during the review — showed vector group YNyn0 with no delta tertiary. The site therefore had no zero-sequence trap anywhere between the inverters and the utility. This confirmed the designer’s instinct that a delta was required, but for a reason nobody had articulated, and it made the winding arrangement more consequential than anyone had realised.

Most seriously, the specification never stated which winding faced the array. Under the intended installation the inverters would draw their grounded-wye reference from the new transformer rather than from the medium-voltage unit. A vendor building delta-toward-the-array — a perfectly reasonable reading of "primary" — would have delivered four transformers that left the inverter side ungrounded. The inverters in question required a grounded wye source. The error would not have surfaced until commissioning.

Resolution.

The specification was reissued with the winding arrangement stated in physical terms rather than by primary and secondary labels, capacity revised to 2,500 kVA at the larger sites, and a note added on the separately derived system obligations the new grounded wye would create. A rating conflict between the project’s stated capacity and the inverter schedule on the issued single-line was raised as a hold point before any order was placed.

Outcome

Four transformers correctly configured at first order. The winding-orientation error alone would have been discovered at commissioning across four sites, with replacement lead times measured in months against a construction programme with no float.

Case 2 — The reactor that solved the wrong problem

Situation.

An operating site had been built with a line reactor as the resonance mitigation, selected on cost during value engineering and installed in the battery branch exactly as the manufacturer’s manual specified. The reactor was correctly sized and correctly located, and the resonance it was installed to address did not occur.

The site nonetheless generated persistent problems. Ground-fault indications appeared on the inverters during grid disturbances with no fault found on investigation. Harmonic alarms logged at the revenue meter. The utility raised a query about neutral current on the medium-voltage feeder.

What the analysis found.

The reactor had been asked to do a job it does not do. It provided series impedance and detuned the resonant interaction, which was its purpose. It provided no galvanic isolation, no zero-sequence trap and no independent ground reference — and the site needed all three for reasons unconnected to resonance.

The medium-voltage transformer was a grounded-wye to grounded-wye unit without a tertiary. Triplen harmonics and zero-sequence unbalance passed directly between the 480 V collection system and the medium-voltage network in both directions. The inverters’ ground reference came through that same transformer, so every zero-sequence disturbance on the utility system arrived at the inverter terminals essentially unattenuated. The nuisance ground-fault indications correlated with utility-side events, which is why no fault was ever found on site.

The utility’s neutral current query had the same root. With no delta anywhere in the path, triplen currents from the collection system had nowhere to circulate and appeared on the medium-voltage neutral.

Resolution.

A delta to grounded-wye isolation transformer was added in the solar branch, sized to the inverter block. The delta provided the zero-sequence trap the system had never had, the grounded wye gave the array a reference under the owner’s control rather than the utility’s, and the added series impedance reduced fault duty at the solar switchgear as a secondary benefit. The reactor was left in place; it was doing no harm and removing it would have required revisiting the manufacturer’s design review.

Outcome

Nuisance ground-fault indications ceased. The utility’s neutral current query closed. The cost of retrofitting a transformer into an energised site — outage, revised drawings, re-submission to the utility — substantially exceeded what the transformer would have cost at the original order, which is the general lesson of the case.

Case 3 — A specification that described the wrong machine

Situation.

A three-page engineering design report and technical specification was circulated for review before being issued to transformer vendors. It was well presented, carried a calculation section, cited standards, and read as competent technical work. It was titled as a specification for an isolation and harmonic mitigating transformer.

What the review found.

The threshold problem was that the document specified a 12.47 kV delta to 480Y/277 V liquid-filled pad-mount transformer with a 65 °C temperature rise and copper windings. That is a medium-voltage distribution transformer. The equipment actually under procurement was a 480 V to 480 V dry-type unit — a different class of machine on every defining parameter. A 65 °C rise is a liquid-filled parameter; dry-type units are specified at 80, 115 or 150 °C. Issued as written, vendors would have quoted a replacement for the transformer already installed on site.

More consequentially, the specified transformer would not have satisfied the battery manufacturer’s requirement under any reading. The medium-voltage transformer sits upstream of the common bus, outside the resonant loop between the two inverter systems. Placing a delta there does nothing to the impedance between them.

The calculation section contained errors that would not have survived utility review. A claimed K-factor range of 3.2 to 3.8 did not follow from the document’s own assumptions — 5 % distortion across the 5th, 7th, 11th and 13th orders yields approximately 1.1, and reaching 3.2 within those orders would require roughly 11 % at the 13th harmonic alone. The K-factor formula was written with harmonic current referenced to the fundamental rather than to rated RMS current, and omitted the fundamental term. A short-circuit figure on the primary side was labelled as available fault current when it was through-fault current referred to the primary. And a passage attributing the K-rating to prevention of "core saturation and thermal runaway" contradicted the document’s own correct statement of the eddy-current mechanism two paragraphs earlier — core saturation is a volts-per-hertz phenomenon, unrelated to harmonic load current.

Several conclusions were nonetheless correct. The 2,500 kVA rating, the K-4 selection, the delta winding for triplen trapping and the electrostatic shield were all the right answers. The derivations simply did not establish them.

Resolution.

A severity-ranked review memorandum was issued identifying three critical findings, ten significant and four minor, with the correct value stated against each. The correct conclusions were carried forward into a reissued specification. The document was withdrawn before it reached vendors.

Outcome

The reissued specification produced comparable quotations for the correct equipment. The broader lesson concerned document provenance: the report had the characteristics of generated technical content — clean structure, confident tone, specific-looking figures that did not survive arithmetic checking. Any calculation entering a procurement or a utility submission should be independently verified regardless of how authoritative the source document appears.

10. Where this usually lands

The isolation transformer question looks like a procurement item and behaves like a systems-engineering problem. The transformer itself is straightforward; what makes these projects difficult is that the decision sits at the intersection of three parties’ scopes and belongs cleanly to none of them. The battery manufacturer specifies a mitigation without knowing the array. The inverter manufacturer specifies a grounding requirement without knowing the system impedances. The utility imposes conditions at the interconnection without visibility of either.

The work that closes it is unglamorous and mostly arithmetic: aggregate the inverter ratings honestly, read the nameplate, run the symmetrical components, write the winding arrangement down in words that cannot be misread. None of it requires waiting for a measurement campaign, and all of it is cheaper than discovering the answer at commissioning.

11. Frequently asked questions

Pole-Mounted Transformer Design: IEEE & NESC Guide | Keentel

Thu, 20 Aug 2026 10:12:15 GMT

Aug 20, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Re-Engineering a Pole-Mounted Distribution Transformer to IEEE, ANSI, and NESC

1. Start with the system

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

Per ANSI/NEMA C84.1-2020 (R2025) , the standard nominal distribution system voltages are:

Insulators

The rendering shows disc (suspension) insulators on the primary. In US distribution tangent construction, the phase conductor typically sits on a pin insulator (ANSI C29.5 for low/medium voltage, C29.6 for high-voltage pin) or, increasingly, a polymer line post (C29.18 for distribution line post — note that C29.17 is the transmission line post and is a common miscitation). Suspension/deadend discs per C29.2 appear at deadends, corners, and dead-end riser poles, not at a typical tangent transformer pole.

Poles

This is the most visually obvious change. The US default is a direct-embedded wood pole to ANSI O5.1-2022, not a spun prestressed concrete pole on a base plate. Setting depth follows the standard 10%-plus-2-feet rule of thumb, checked against NESC Section 26 / Rule 261 strength requirements for Grade B or Grade C construction and the applicable Heavy/Medium/Light loading district under NESC Section 25. RUS Bulletin 1724E-150 — the federal design guide for rural distribution — is written almost entirely around unguyed wood poles, Cl asses 1–6, 35–55 ft.

Spun concrete poles do exist in the US and are covered by ASTM C1089-19 (R2025). The y are a deliberate, cost-justified exception (coastal corrosion, fire-hardening, decay zones), not the baseline. And be careful with citations: the ANSI C136 series is roadway and area lighting equipment , not utility distribution concrete poles.

Connectors

The hotline clamp translates directly, but the qualification basis changes. ANSI C119.4-2022 governs aluminum-to-aluminum and aluminum-to-copper connectors on distribution and transmission lines . Its Class A / Class B designations are electrical current-cycle test classes — Class A is 500 current cycles, Class B is 250 — plus separate mechanical strength classes 1, 2, and 3. They are not connector "types," and specifying "Class A" without also specifying the mechanical class leaves the spec incomplete.

Figure 1 , on the following page, shows the whole assembly rebuilt — elevation and single-line — alongside a summary of what the 34.5 kV version changes.

FIGURE 1 — US-practice pole-top assembly at 12.47 kV (12470GrdY/7200 V), 4-wire multigrounded neutral

The IEEE / ANSI / NESC equivalent of the IS/CBIP 12 kV — 0.4 kV pole-mounted substation. Illustrative design intent only — not for construction.

Note what changes and what does not. Primary current falls — 5.0 A at 34.5 kV versus 6.9 A at 12.47 kV for a comparable load — which is exactly why 34.5 kV distribution wins on long rural feeders. What rises sharply is everything driven by dielectric stress: BIL goes 95 → 150 kV, phase spacing widens, crossarms lengthen, and the arrester’s MCOV nearly triples.

One clearance subtlety worth flagging. NESC Rule 232 / Table 232-1 brackets open supply conductors at "over 750 V to 22 kV," and for effectively grounded systems that threshold is applied phase-to-ground. A 34.5Y/19.92 kV system sits at 19,920 V line-to-ground — still inside the 22 kV bracket. So the base vertical clearances apply, and the Rule 232C1 adder of 0.4 in (10 mm) per kV above 22 kV does not kick in. Engineers who apply the adder off the 34.5 kV line-to-line number will over-build every crossing on the line. (Verify against the codebook for your specific configuration; ungrounded and delta systems are treated differently.)

2. Voltage class: what "12 kV" actually means

The rendering’s SLD implies a straightforward radial cascade. US distribution protection is more layered, and the fuse is designed around it.

Fuse links

Expulsion links per IEEE/ANSI C37.42 come in speed characteristics defined by speed ratio — the ratio of the current that melts the link in 0.1 s to the current that melts it in 300 s (600 s for links above 100 A). Type K links are fast (speed ratio 6–8, nominally 7); Type T links are slow (10–13, nominally 12). Selecting K where a utility standardizes on T, or vice versa, breaks coordination with the upstream device.

Sizing

There is no clean percentage rule, and consultants who quote one are usually quoting a rule of thumb as if it were a standard. Real practice:

PG&E’s overhead standard states plainly that transformers are fused at approximately 2.5× nominal rating.
Classic guidance holds that minimum blowing current should exceed roughly 200% of transformer FLA, since primary fuses cannot meaningfully protect against overload — the curves don’t match — and exist for fault clearing.
S&C Data Bulletin 350-110, the industry selection reference, deliberately uses per-kVA lookup tables, not a multiplier, evaluating continuous peak load at 40 °C ambient, hot-load pickup up to 6× pre-interruption current, and cold-load pickup of 6× for 1 s, 3× for 10 s, and 2× for 15 minutes.
Major utilities (JEA among them) publish direct kVA→link tables by system voltage with no percentage rule at all.

The right framing: cold-load and hot-load pickup set the floor, not full-load amps. A transformer that has been de-energized for an hour on a summer afternoon will draw multiples of FLA on restoration, and a fuse sized off FLA will nuisance-blow every time the feeder recloses.

Coordination

The pole fuse must coordinate with the upstream feeder recloser or substation relay, and the utility’s fuse-saving versus fuse-blowing philosophy determines whether the recloser’s fast curve is enabled. That decision is made at the feeder level, not the pole level — which is precisely why a pole-top detail cannot be designed in isolation from the distribution study.

Arresters

Unde r IEEE C62.11-2020, distribution arresters are still specified as Normal Duty or Heavy Duty (with Heavy Duty Riser as an HD variant). What changed in the 2020 harmonization with IEC 60099-4 was the retirement of line-discharge-class ratings in favor of charge and energy withstand: Qrs (repetitive single-impulse charge transfer), Qth (thermal charge transfer, for distribution arresters), and Wth (switching-impulse energy). Specifications written against the older line-discharge-class language should be updated.

The rendering’s "earthing" conductor and base plate become something structurally different in the US: the pole ground is not a standalone earth electrode serving one transformer, it is one of many bonds to a continuous multigrounded neutral.

Governing rules live in NESC Section 9 :

Rule 096A — resistance to ground low enough to permit prompt operation of protective devices.
Rule 096C — multigrounded neutral systems require a minimum o f four grounds per mile.
Rule 096D — an individual made electrode on a single-grounded system shall not exceed 25 Ω.

A note on standards misuse we see regularly in submitted packages:

IEEE Std 80-2013 is a substation grounding guide — GPR, step and touch potential inside the fenced yard — and IEEE Std 142 (Green Book) covers industrial and commercial premises grounding . Neither is the governing document for a utility distribution pole ground. Citing IEEE 80 for a pole ground is a tell that the design was assembled from references rather than from practice.

8. Grounding

Transformers become single-bushing (grounded-wye primary). A 12.47 kV pole-mount unit is typically specified as 12470GrdY/7200 — one primary bushing at 7,200 V line-to-neutral with the return through the tank/neutral. The Indian design’s three fully-insulated primary bushings become one bushing per single-phase unit.
Arrester duty is set by line-to-ground voltage, not line-to-line. This is what lets a US 12.47 kV system use a 10 kV duty-cycle arrester where an ungrounded 12 kV system would need substantially more.
Single-phase service is the default, not the exception. A large fraction of US pole-top transformers are single-phase, 120/240 V, serving a handful of houses. The three-phase pole-top unit shown in the rendering is the unusual case in US practice, not the base case.
Ground-fault current is high and the return path is deliberate. Line-to-ground faults on an MGN system are true bolted faults, sized comparably to three-phase faults, which is why US fusing and coordination practice looks nothing like IEC-region practice.

And frequency matters more than people expect. A 50 Hz core operated at 60 Hz sees lower flux density and different loss and excitation behavior; impedance and regulatio n shift too. You cannot re-label a 50 Hz transformer as a 60 Hz unit — it has to be designed and tested to IEEE C57.12.00 and C57.12.20 from the start.

Two notes worth internalizing.

First, 12.47 kV, 13.2 kV, and 13.8 kV all share the 15 kV insulation class. Insulation, arresters, cutouts, and cable are ordered by class , not by nominal voltage. This is why "15 kV class" is the phrase that appears on most US distribution material specs.

Second, C84.1 Range A above 600 V is +5% / −2.5%, not the symmetric ±5% that applies at 600 V and below. On a 12,470 V nominal system, Ran ge A service voltage is 12,157–13,094 V. Designers who carry the ±5% low-voltage habit upward into MV voltage-drop studies will size regulation wrong on the low side.

Design intent: three-phase, 208Y/120 V service to a small commercial building off a 12.47 kV MGN feeder.

Three single-phase units banked is deliberate. It is the common US route to three-phase overhead service: stock is standardized, a single failed unit is a cheap replacement, and the bank tolerates open-delta operation if one unit is lost. Three-phase pole-mounted units exist — triplex designs generally 30–225 kVA, T-connected up to 500 kVA — but they are a purchased exception at most utilities, not the default.

5. Worked example B — 34.5 kV rural / collector tap

Design intent: three-phase, 480Y/277 V auxiliary service off a 34.5 kV line — a very common arrangement for site power at a solar, wind, or BESS facility, and for rural three-phase irrigation and pumping loads.

6. Where the rendering has no US equivalent: the LT MCCB box

This is the most instructive difference on the whole pole.

The rendering shows an LT MCCB box below the transformer, and the accompanying SLD runs: overhead feeder → DO fuse cutout → transformer → LT box bus → MCCB → load. That secondary molded-case breaker is standard practice in IS/CBIP distribution substation design .

On a US utility pole, it usually isn’t there. Secondary protection lives on the customer side of the service point, in customer-owned equipment governed by the NEC (NFPA 70). The utility side stops at secondary spades, triplex, and a service drop.

That demarcation is not stylistic — it is jurisdictional. NEC 90.2(D)(5) in the 2023 NEC (90.2(B)(5) in the 2020 and earlier editions) exempts installations under the exclusive control of an electric utility, on utility-owned or leased property, for the purpose of generation, transformation, transmission, or distribution. Everything on that pole is NESC (IEEE C2-2023) territory. Everything past the service point is NEC territory, and NEC Article 450 governs the customer’s transformers, not the utility’s.

Get this boundary wrong and you produce a drawing that satisfies neither code.

4. Worked example A — 12.47 kV suburban commercial tap

7. Protection philosophy: the part that doesn’t translate at all

9. If the pole serves a DER site

Increasingly, these assemblies exist to serve or interconnect distributed generation. Once generation is on the customer side of that transforme r, IEEE Std 1547-2018 (with amendment 1547a-2020, which widened Category III trip-clearing settings in Table 13) governs interconnection perform ance, ride-through, protection, and testing. That drives anti-islanding coordination, transformer winding configuration selection, and whether a direct transfer trip or a dedicated interconnection recloser is required — none of which is visible anywhere on the original rendering, and all of which changes the pole.

Standards referenc

losing

The original rendering is a good teaching drawing. It is also a reminder that a distribution assembly is not a parts list — it is the physical expression of a system philosophy. Change the neutral from three-wire to multigrounded, the frequency from 50 to 60 Hz, the governing code from IS/CBIP to NESC, and the protection basis from a secondary MCCB to a coordinated feeder recloser, and you have not adapted a design. You have replaced it.

At Keentel Engineering , distribution-class assets like these show up constantly at the edges of the work we do — auxiliary and station service at collector substations, gen-tie taps, POI interconnection packages, and utility service to solar, wind, and BESS sites. The interfaces between the utility distribution system and the project are where interconnection schedules slip, and they slip for exactly the reasons above: a clearance applied off the wrong voltage basis, a fuse sized off FLA instead of cold-load pickup, an NEC citation where a NESC rule belonged.

If you have a pole-top, service, or POI detail that needs to be right the first time, we would be glad to look at it.

KEENTEL ENGINEERING

Ta mpa, FL · Austin, TX · Sacramento, CA · Baltimore, MD

keentelengineering.com · contact@keentelengineering.com · 813-389-7871

Engineering note: standards editions, rule numbers, and clearance values cited here were verified against current published sources at time of writing. Always confirm against the governing edition adopted in your jurisdiction and the serving utility’s own construction standards, which frequently exceed code minimums.

3. Component-by-component translation

NGR Sizing Guide: Neutral Grounding Resistor Calculation

Thu, 20 Aug 2026 08:46:38 GMT

Aug 20, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Neutral Grounding Resistor (NGR) Sizing

A US Engineering Guide to IEEE, ANSI, and NEC Practice

Power Systems Design | Technical Guide, FAQ and Case Studies | August 2026

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Part 2 — Frequently Asked Questions: Large Load Interconnection

A neutral grounding resistor sits between a wye-connected source neutral — transformer secondary or generator — and ground. During a line-to-ground fault, it is the dominant impedance in the zero-sequence path, so it, not the system, decides how much ground-fault current flows.

Every benefit an NGR delivers traces back to that single fact:

Basic sizing

R = V(L-N) / Ig where V(L-N) = V(L-L) / √3

Combined: R = V(L-L) / ( √3 × Ig )

Two things to be careful about. First, use the maximum system voltage, not the nominal, if your utility or generator regulation can push the bus above nameplate — the resistor sees whatever the neutral shift actually is. Second, The ground-fault current that results is not exactly your target, because detailed short circuit analysis may be required when impedance effects become significant. At LRG levels the error is small; at HRG levels it is negligible. At low resistance values approaching solidly grounded, it stops being negligible and you need the full symmetrical-component calculation.

The charging-current constraint (HRG only)

This is the constraint most people get backwards, and it is the one that decides whether HRG is even feasible on your system. IEEE Std 142-2007 §1.4.3.1, repeated verbatim in IEEE 3003.1-2019:

"The value of the resistor is selected to limit the current, I_R, to a magnitude equal to or slightly greater than the total capacitance charging current, 3I_C0."

Ir ≥ 3 Ic0. The notation matters. Ic0 is the per-phase zero-sequence capacitive charging current. 3Ic0 is the system total. Writing "Ir ≥ Ic0" is only correct if you have quietly redefined Ic0 as the total, which contradicts IEEE usage and will confuse whoever reviews your calculation. Use 3Ic0 and state the definition.

Why the constraint exists: if the resistive component of ground-fault current does not at least equal the capacitive component, the arc at the fault can restrike and pump energy into the system L-C circuit. The resistor exists to damp that. Undersize it — meaning too high a resistance, too little current — and you have built an expensive ungrounded system.

Estimating 3Ic0. You sum contributions from every capacitive element. Typical screening values:

1. What an NGR actually does

Most of what circulates on this topic is directionally right. These specific items are not, and each of them has a cost.

1. The IEEE standard number has changed

IEEE Std 32-1972 (R1997) is superseded. The governing document is IEEE Std C57.32-2015, "IEEE Standard for Requirements, Terminology, and Test Procedures for Neutral Grounding Devices," with amendment C57.32a-2020 adding a neutral grounding resistors clause. Both are now listed as Inactive-Reserved by IEEE SA, which is a records-status flag rather than a withdrawal — they remain the referenced documents in practice, and specifications should cite them by number and year.

2. The standard time ratings are not 10 / 30 / 60 seconds

The IEEE-enumerated set is 10 seconds, 1 minute, 10 minutes, extended time, and continuous. ("Extended time" means 10 minutes or more, not exceeding 90 days per year in aggregate.) Vendors will build a 30-second unit, and there is nothing wrong with buying one, but do not present 30 s as a standard IEEE rating in a specification — it invites a bid clarification you do not need.

3. Temperature rise limits pair differently than commonly shown

Per IEEE C57.32, above 30 °C ambient: 10 s → 760 °C, 1 min → 760 °C, 10 min → 610 °C, extended → 610 °C, continuous → 385 °C.

4. "Continuous duty is not normally required" is wrong for HRG

See Example B. It is true for LRG and false for the application where the resistor matters most.

5. kW is a derived quantity, not a specification

Every manufacturer — Post Glover, I-Gard, Filnor, Powerohm — specifies by rated voltage, rated initial current, and rated time. Kilowatts appear on no data sheet. Publishing a "635 kW resistor" invites a procurement conversation that goes nowhere.

One thing the popular guides get right: the charging-current criterio n Ig ≥ 3Ic is correct and matches IEEE 142 / 3003.1 exactly. It deserves more prominence than it usually gets, not less.

And one nuance about arc flash

NGRs dramatically reduce ground-fault arc energy — that is real, and it is the single strongest safety argument for HRG. But IEEE Std 1584-2018, "IEEE Guide for Performing Arc-Flash Hazard Calculations," models three-phase arcing faults only. Single-phase and line-to-ground arcing faults are explicitly outside its scope. Its stated applicability is 208 V to 15 kV, three-phase, with bolted fault current from 700 A to 106 kA.

The practical consequence: installing an NGR does not lower the calculated incident energy on your arc-flash labels, because those labels are computed from three-phase bolted fault current, which the NGR does not touch. The NGR reduces the likelihood and severity of the ground-fault event that initiates most arcing faults in the field. That is a genuine and valuable risk reduction. It is not a label change, and anyone who promises you one is confusing two different things.

Relay sensitivity

Can your ground relay actually see Ir? On a 5 A HRG system, a 5 A ground-fault current is below the pickup of most conventional overcurrent elements and well below normal residual unbalance on a phase-CT residual connection. HRG requires either neutral-connected sensing at the resistor or core-balance CTs on feeders — never a residual connection from phase CTs.

Resistance drift with temperature

Relay coordination depends on the fault current staying near its rated value. IEEE C57.32 caps the resistance increase at 67%, but the alloy matters: purpose-made resistance alloys run around 0.00024–0.00036 Ω/°C, while general-purpose stainless (AISI 304 at 0.00092, AISI 430 at 0.00146) drifts several times more. Specify the temperature coefficient, not just "stainless steel."

Transient overvoltage

On an ungrounded system, restriking ground faults are cited in IEEE Std 242 (Buff Book) as capable of producing overvoltages of six to eight times normal phase voltage. IEEE 142-2007 §1.4.3.2 states that both high- and low-resistance grounding are designed to limit transient overvoltages to 250% of normal. Keep these separate from the steady-state effect: during a sustained single line-to-ground fault, the unfaulted phases rise to line-to-line voltage — 173% of normal line-to-ground — on any impedance-grounded or ungrounded system. That 173% is why cable and surge arrester insulation levels must be checked before an HRG conversion.

Cable and arrester insulation rating

Because the unfaulted phases sit at 173% during a sustained fault, an HRG system needs 133% or 173% insulation-level cable depending on how long faults are permitted to persist, and surge arresters rated for an ungrounded-system duty. This item alone has stopped more than one HRG retrofit at the cable-schedule review.

Neutral conductor insulation

NEC 250.187 requires the impedance grounding conductor be insulated for the maximum neutral voltage — 57.7% of phase-to-phase.

Effectively grounded verification

If a utility interconnection agreement requires an effectively grounded source, you must demonstrate X₀/X₁ ≤ 3 and R₀/X₁ < 1 (coefficient of grounding ≤ 80%). Resistance grounding by definition does not meet this. This is a frequent and expensive collision on generation interconnection projects — see Case Study 3.

Generator neutrals

IEEE C62.92.2-2017 recommends limiting generator ground-fault current to 5 A to 15 A under high-resistance grounding, because "iron burning and damage depend mainly on fault current magnitude and duration." A solidly grounded generator can see ground-fault current exceeding its three-phase fault current, since zero-sequence impedance is typically the lowest of the three sequence impedances — and the damage lands in the laminated stator core, which is the most expensive thing in the machine to repair. Where generators are paralleled, use separate NGRs per machine to avoid circulating third-harmonic current.

7. Design checks that catch real problems

NGR sizing looks like a one-line calculation. Divide line-to-neutral voltage by the ground-fault current you want, and you have your resistance. That formula is correct, and it is also about 15% of the actual engineering.

The other 85% is everything the formula does not tell you: whether your target current is defensible in the first place, whether the resistor must survive ten seconds or ten years, whether your relaying can actually see the current you selected, and whether the NEC will let you build it at all.

This guide works through the full sizing process on US voltage classes with US governing standards — IEEE C57.32, IEEE 3003.1, IEEE C62.92, and NEC Article 250 — with two complete worked examples: a 13.8 kV low-resistance grounded industrial bus and a 480 V high-resistance grounded system. It closes with a detailed FAQ and three anonymized case studies.

That last row is the one most often skipped, and it is the one that determines whether the design works in service. An NGR does not merely limit current — it manufactures a predictable current that protection engineers can build a coordination scheme around. Get the value wrong in either direction and the resistor still limits current beautifully while the relays either never see the fault or trip on load unbalance.

The grounding spectrum

US practice recognizes a continuum, not a binary. Per IEEE Std C62.92.2-2017, in order of increasing available ground-fault current:

Resonant grounded (ground-fault neutralizer / Petersen coil) — rare in US industrial work
Ungrounded — no intentional connection; capacitively coupled
High-resistance grounded (HRG) — typically ≤ 10 A
Low-resistance grounded (LRG) — 100 A to 1000 A, 400 A typical
Low-inductance grounded
Effectively grounded (solid) — X₀/X₁ ≤ 3 and R₀/X₁ < 1

The 100–1000 A LRG range and the 400 A typical value are not folklore; they are stated in IEEE Std 142-2007 §1.4.3.2 and carried forward into IEEE Std 3003.1-2019.

3. Worked Example A — 13.8 kV low-resistance grounded industrial bus

Design intent: a 13.8 kV switchgear lineup feeding MV motors and unit substations in a process plant. Ground faults must be cleared quickly and selectively; running through a ground fault is not desired or permitted by the process.

Step 1 — Line-to-neutral voltage

V(L-N) = 13,800 / √3 = 7,967 V

Step 2 — Select the target ground-fault current

400 A. This is the IEEE-stated typical value for LRG, and on a system with several thousand feet of 13.8 kV shielded cable it comfortably exceeds 3Ic0 (which will be on the order of 3–8 A here), so the resonance constraint is satisfied by an enormous margin. The binding constraint at LRG is not charging current — it is relay sensitivity and equipment damage.

Step 3 — Resistance

R = 7,967 / 400 = 19.9 Ω → specify 20 Ω. At 20 Ω the initial ground-fault current is 7,967 / 20 = 398 A, which is fine.

Step 4 — Time rating

10 seconds. Ground faults on this bus are cleared by 51G in well under 1 second, and 10 s is the standard short-time rating that gives comfortable margin over breaker failure timing.

Step 5 — Thermal check

P(initial) = I²R = 400² × 20 = 3,200,000 W = 3.2 MW

Now read that number carefully, because this is where the widely circulated version of this calculation goes off the rails. 3.2 MW is not a rating. It is the instantaneous dissipation at t = 0, and it decays as the resistor heats and its resistance rises. A 400 A / 10 s unit can be down to roughly 250 A by the end of its rated 10 seconds. You will not find "3.2 MW" on any manufacturer data sheet, and you should not put it in a specification.

What actually goes on the nameplate:

Step 6 — Protection

51G on a dedicated CT in the NGR neutral connection. Pickup at roughly 10% of 400 A, coordinated with downstream feeder ground relays.
50G / 51G on feeders using core-balance (window) CTs, which give far better sensitivity than residual connection at these current levels.
NGR continuity monitoring. An open NGR converts an LRG system into an ungrounded system silently, with no alarm an d no operational symptom until the first ground fault produces overvoltages the system was never insulated for. Continuity monitoring is inexpensive and it is the single highest-value accessory on an LRG installation.

4. Worked Example B — 480 V high-resistance grounded system

Step 3 — Select Ir

Ir must be ≥ 3Ic0 = 2.5 A. Select 5 A — roughly 2× the estimated charging current, which covers estimation error and future cable additions without pushing the fault current high enough to cause meaningful damage at the fault point.

Step 4 — Resistance

R = 277 / 5 = 55.4 Ω → specify 55 Ω (Ir = 5.0 A)

Step 5 — Time rating: CONTINUOUS

This is the critical difference from Example A, and it is where the popular version of this calculation is simply wrong. HRG exists so the system can keep running with a ground fault on it while operations locate the fault. The resistor must therefore carry its full rated current indefinitely. IEEE C57.32 continuous duty limits temperature rise to 385 °C over 30 °C ambient, versus 760 °C for short-time ratings — a different resistor, a different thermal design, a different price.

Post Glover states it plainly: continuous-duty rating is required "when the service continuity (alarm only, first fault) is prime concern." Any guidance that says continuous duty is not normally required is describing LRG and mislabeling it as universal.

Step 6 — Power dissipation

P = I²R = 5² × 55 = 1,375 W ≈ 1.4 kW, continuous

Compare: Example A produces 3.2 MW that exists for ten seconds and is not a rating. Example B produces 1.4 kW that exists forever and is a real, meaningful thermal design number. This is exactly why the industry specifies NGRs by voltage, current, and time — never by kilowatts.

Step 7 — Code compliance check, NEC 250.36

In the 2023 NEC, 250.36 is titled "Impedance Grounded Systems — 480 Volts to 1000 Volts." (It was "High-Impedance Grounded Neutral Systems" in the 2020 and earlier editions — a retitle worth noting in any spec that cites it.) Three conditions must be met:

O nly qualified persons service and maintain the installation
Ground detectors are installed
Line-to-neutral loads are not served

Condition 3 kills more HRG retrofits than any other single factor. If any 277 V lighting or single-phase load is fed line-to-neutral from that system, you cannot convert it to HRG without re-feeding those loads from a separate solidly grounded source. Confirm this before you design anything.

For systems above 1000 V, the governing section is 250.187, "Impedance Grounded Systems" — same three conditions, plus a r equirement that the impedance grounding conductor be insulated for the maximum neutral voltage (57.7% of phase-to-phase on a three-phase wye).

Step 8 — Detection and fault location

59G / device 64 neutral overvoltage or ground detection for the alarm
A pulsing contactor across part of the resistor, which modulates the ground-fault current so a clamp-on ammeter can trace the faulted feeder while the plant runs
Do not install a trip on first ground fault. If you do, you have paid for HRG and built LRG.

5. Side by side: HRG vs LRG

A cruder rule of thumb: 0.5 A per 1000 kVA at 600 V and below, 1.0 A per 1000 kVA above 1000 V. Use it for feasibility screening, not for a final specification — the cable footage drives everything, and two plants with identical transformer capacity can differ by an order of magnitude in charging current.

Measure it if you can. On a retrofit, the reliable method is a staged single-phase-to-ground connection through a known resistance with the system energized and unloaded, or a direct capacitance measurement. Estimated charging current has a wide error band, and HRG sizing has no margin to spare on the low side.

6. Five things worth correcting in the popular NGR guides

8. Anonymized case studies

Case Study 1 — Colocation data center: HRG retrofit blocked by 277 V lighting

Situation. A 30 MW Mid-Atlantic colocation facility operated its 480 V mechanical and house distribution as a solidly grounded system. Two unplanned outages in eighteen months traced to ground faults in pump-room conduit that tripped upstream and dropped mechanical cooling. The client asked for a high-resistance grounding conversion to eliminate first-fault trips.

What the analysis found

Charging current was manageable — roughly 3.5 A total on the mechanical distribution — so a 5 A HRG design at 55 Ω continuous was straightforward on paper. The problem was NEC 250.36 condition 3: the same 480Y/277 V system fed several hundred 277 V lighting fixtures line-to-neutral across the building. HRG was not permissible on that system as built.

What was done

Rather than abandon the conversion, the system was split. The mechanical and process distribution — where the ground faults were actually occurring, and where uptime mattered — was re-fed from a dedicated 480 V transformer with no line-to-neutral loads and high-resistance grounded at 5 A continuous, with pulsing fault location and 59G alarm to the BMS. The 277 V lighting load stayed on a separate, solidly grounded 480Y/277 V transformer where trips were operationally tolerable.

Outcome

Ground faults on the mechanical system now alarm instead of tripping, and are located during business hours using a clamp meter and the pulsing contactor. The additional transformer added meaningful cost relative to a straight NGR installation, which is exactly why the code check belongs at the front of the project, not at the permit review.

Engineering takeaway

Check NEC 250.36 condition 3 before you size anything. The resistor calculation is the easy part; the load inventory is what determines whether HRG is buildable.

Case Study 2 — Gulf Coast refinery: the 13.8 kV NGR that had been open for two years

Situation. A refinery experienced simultaneous cable terminator failures at two locations on a 13.8 kV bus during what should have been a routine single line-to-ground fault. The immediate damage was contained, but the failure pattern — multiple remote flashovers from one initiating fault — did not match a resistance-grounded system.

What the analysis found

The NGR on the source transformer neutral had failed open. Post-incident resistance measurement confirmed an open element. Comparison against maintenance records indicated the resistor had most likely been open since the previous turnaround, roughly two years earlier. During that period the 13.8 kV system had been operating ungrounded, with no alarm, no indication, and no operational symptom of any kind. The terminator failures were consistent with restriking-arc transient overvoltages in the range IEEE 242 describes for ungrounded systems — six to eight times normal phase voltage.

What was done

The NGR was replaced with a 20 Ω, 400 A, 10-second unit (7,967 V line-to-neutral rating) and continuity monitoring was installed on every NGR in the facility — a fleet of eleven resistors across the MV system, none of which had monitoring. Ground relay settings were reverified against the restored 400 A design value. Cable terminations at the two failure points and at three other locations showing partial-discharge indications were replaced.

Outcome

The monitoring retrofit cost a small fraction of the terminator replacements from the single event, and detected a second developing high-resistance condition on a different unit within the following year.

Engineering takeaway

An NGR is the only major protective device in a substation that can fail completely, silently, and without consequence — right up until the moment it matters. Continuity monitoring is not an accessory. Treat an unmonitored NGR the same way you would treat a relay with no self-test.

Case Study 3 — 120 MWac utility-scale solar: HRG was never an option

Situation. A 120 MWac PV project in ERCOT proposed high-resistance grounding on its 34.5 kV collector system, reasoning that a single ground fault should not curtail the whole plant. The utility's interconnection requirements simultaneously called for an effectively grounded source at the point of interconnection.

What the analysis found

Two independent problems.

First, charging current. The collector system comprised roughly 28 miles of 35 kV-class shielded cable. Scaling from published 13.8 kV values by voltage and cable capacitance put total charging current on the order of 150 A. To satisfy Ir ≥ 3Ic0, the resistor would have had to pass at least 150 A — which is squarely in the low-resistance band. High-resistance grounding was arithmetically impossible on this system, independent of anyone's preference.

Second, the interconnection requirement. Effective grounding requires X₀/X₁ ≤ 3 and R₀/X₁ < 1. Resistance grounding does not meet that criterion by definition. The two requirements in the project documents were mutually exclusive as written.

What was done

The two requirements were separated by where they apply. The GSU was configured to provide an effectively grounded zero-sequence source at the point of interconnection, satisfying the utility. The collector system was low-resistance grounded at 400 A through a 50 Ω, 10-second NGR (19,919 V line-to-neutral rating), giving fast, selective clearing of collector-circuit ground faults. Feeder ground protection used core-balance CTs at each collector circuit breaker so that a single faulted circuit trips and the remaining circuits keep producing — delivering most of the availability benefit the project had wanted from HRG, through selectivity rather than through riding out the fault.

Outcome

A ground fault on one collector circuit now curtails that circuit only, typically 8–12 MW rather than the full plant. The design cleared utility review without an exception request.

Engineering takeaway

"We want HRG" is a statement about desired behavior, not a design. On large collector systems the charging current usually decides the question before anyone gets to state a preference. And where an interconnection agreement demands effective grounding, the resolution is almost always to satisfy it at the GSU rather than to argue about the collector system.

10. Standards reference

2. The formulas — and what the notation actually means

Design intent: a 480 V distribution system serving critical process and IT load. A single ground fault must alarm, not trip. Operations locate and clear it during the next planned window.

Step 1 — Line-to-neutral voltage

V(L-N) = 480 / √3 = 277 V

Step 2 — Estimate total charging current, 3Ic0

Figure 1 , on the following page, summarizes the whole method: where the resistor sits, the grounding spectrum, the nine-step sizing sequence, both worked examples, and the IEEE C57.32 rating table.

FIGURE 1 — NGR sizing to US / IEEE practice: placement, grounding spectrum, sizing sequence, worked examples, and rating table

Illustrative design guidance — not a substitute for a project-specific grounding and coordination study.

Closing

The NGR sizing formula is trivially simple and almost never the reason a grounding design fails. Designs fail because charging current was estimated instead of measured, because a continuous-duty application was specified with a short-time resistor, because a 277 V lighting circuit made the whole scheme non-compliant, because nobody monitored a device that fails silently, or because two contract documents demanded incompatible grounding classes and nobody reconciled them until the design review.

At Keentel Engineering , system grounding studies are part of the power system studies we deliver alongside substation design, POI interconnection engineering, and owner's engineering for renewable and industrial projects. If you have an NGR specification to review, a grounding class to select, or an interconnection requirement that appears to contradict your plant design, we would be glad to look at it.

KEENTEL ENGINEERING

Tampa, FL · Austin, TX · Sacramento, CA · Baltimore, MD

keentelengineering.com · contact@keentelengineering.com · 813-389-7871

Engineering note: standards editions, section numbers, and quantitative limits cited here were verified against current published sources at time of writing. NEC section numbers are given for the 2023 edition; confirm against the edition adopted in your jurisdiction. Nothing here substitutes for a project-specific grounding and coordination study.

9. Frequently asked questions

PRC-023-6 BESS Compliance: Relay Loadability Guide | Keentel

Wed, 19 Aug 2026 13:58:18 GMT

Aug 19, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

PRC-023-6 and Utility-Scale Battery Storage

Why Your BESS Is a Generator Owner With a Transmission Relay Loadability Problem

An in-depth applicability, implementation-timing and relay-setting guide for battery energy storage Generator Owners.

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Most battery energy storage system (BESS) developers and owners build their NERC compliance program around the standards everyone talks about: PRC-024, PRC-028, PRC-029, PRC-030, MOD-026, MOD-032, FAC-008. Transmission relay loadability rarely makes the list. The assumption is understandable — PRC-023 has "Transmission" in its title, and a battery plant is a generator.

That assumption is wrong often enough to be expensive.

PRC-023-6 — Transmission Relay Loadability names the Generator Owner as an applicable functional entity in Section 4.1.2, in the same sentence and on the same terms as the Transmission Owner. It does not care whether the machine behind your relay is a steam turbine, a wind farm, or a 400 MWh lithium-iron-phosphate installation. It cares about one thing: whether you have applied load-responsive phase protection systems at the terminals of a circuit that falls within Section 4.2.1. If you have, you own Requirements R1 through R5, with a High Violation Risk Factor on R1 and a Violation Severity Level table that offers only one outcome — Severe.

Battery storage makes this harder than conventional generation for a reason that sits right in the text of the standard. The exemption that most generators rely on covers Elements "used exclusively to export energy directly from a BES generating unit or generating plant." A battery plant does not exclusively export. It imports at full nameplate, on a daily cycle, by design. Whether that single word survives contact with a bi-directional asset is the most consequential applicability question a storage Generator Owner will face, and it is not answered anywhere in the standard, the Technical Rationale, the Implementation Plan or the Implementation Guidance.

This article works through the whole problem: why a battery plant is a Generator Owner, exactly which parts of the standard reach it, how the effective dates and the 39-month implementation clock actually run, the relay-setting mathematics with fully worked 138 kV examples, the inverter-based-resource complications that make loadability and dependability collide, and what a defensible compliance program looks like in practice.

Part 2 — Frequently Asked Questions: Large Load Interconnection

PRC-023 exists because of August 14, 2003.

During the Northeast blackout, distance relays — particularly overreaching Zone 3 elements — could not distinguish heavy load from a fault. Depressed voltage and high current during the cascade produced an apparent impedance that fell inside relay tripping characteristics, and lines that were thermally fine tripped anyway. Each trip shifted load onto the remaining circuits, which pushed their apparent impedance further into their own characteristics. The protection system stopped protecting the grid and started dismantling it.

The standard's Purpose statement carries both halves of the resulting lesson:

"Protective relay settings shall not limit transmission loadability; not interfere with system operators' ability to take remedial action to protect system reliability and; be set to reliably detect all fault conditions and protect the electrical network from these faults."

That is a security requirement (do not trip when you should not) with an explicit dependability boundary (but still clear every fault). PRC-023 is fundamentally about buying the system operator time. As NERC's Statement of Purpose in the Implementation Guidance puts it, relays should "not interfere with the system operators' ability to consciously take remedial action to protect system reliability," allowing "short-term overloads with sufficient margin to allow for inaccuracies in the relays and instrument transformers."

What changed in Version 6

PRC-023-6 came out of NERC Project 2021-05, Modifications to PRC-023. There are two substantive changes.

Requirement R2 was retired and is now marked "Reserved."

The old R2 required entities to set out-of-step (power swing) blocking elements to allow tripping of phase protective relays for faults occurring during the R1 loading conditions. The Standard Drafting Team's Technical Ra tionale makes an unusually candid case for retirement:

R2's fault condition was a subset of R1's "all fault conditions," so any R2 violation was automatically an R1 violation. Retirement created no reliability gap.
R2 traced back to an error in Appendix C of the 2006/2007 reference document Determination and Application of Practical Relaying Loadability Ratings , which discussed only electromechanical power-swing-blocking schemes and concluded a timer should force a trip — which the current SDT correctly notes "is not appropriate because tripping should not occur during the identified heavy load conditions unless a fault actually occurs."
The original Attachment A language ("shall be evaluated to ensure that they do not block trip") was inverted when FERC's Order 733 ¶244 directive was implemented as a requirement ("shall set its out-of-step blocking elements to allow tripping"). That inversion shifted engineering attention from evaluating the scheme to constraining the blinder settings , and in at least one documented case an entity simply disabled two power-swing-blocking schemes rather than fight the requirement.
Ten years of evidence supported retirement: only two R2 violations ever, both found in paperwork rather than operations and both assessed as minimal risk; roughly 40,000 operations across five years of MIDAS data yielding 11 candidate events and a single relay-related one; and 18 major NERC event-analysis reports plus the February 2008 FRCC disturbance showing no case of a fault going uncleared because of power swing blocking.
Fleet modernization closed the gap independently. One entity went from 161 of 471 (34%) affected line terminals on electromechanical relays in 2011 to 19 of 699 (2.7% ) by 2022.

Attachment A, Item 2.3 was retired.

The exclusion for "protection systems intended for protection during stable power swings" referenced schemes designed to separate portions of the system during stable swings — an application the SDT concluded no longer exists in North America, with PRC-026 covering the space adequately. Items 2.3 and 2.4 now both read "Reserved."

What this means for a BESS owner:

you no longer have a standalone power-swing-blocking requirement to satisfy, but the obligation did not vanish — it was absorbed into R1's "while maintaining reliable protection of the BES for all fault conditions." If you apply PSB or out-of-step logic at your point of interconnection, you still have to demonstrate that a fault occurring during PSB assertion will be detected and cleared. You simply document it under R1 now.

The registration chain

A utility-scale BESS becomes subject to PRC-023-6 through a chain of four links, and the standard only reaches you if all four hold.

Your facility is part of the Bulk Electric System. Under the NERC BES definition, generating resources connected at 100 kV or above qualify through Inclusion I2 where individual units exceed 20 MVA gross nameplate or the plant aggregate exceeds 75 MVA, and through Inclusion I4 for dispersed power producing resources aggregating more than 75 MVA delivered through a system designed primarily for that purpose at 100 kV or above. Nearly every utility-scale battery plant — 100 MW and up, interconnected at 115 kV, 138 kV, 230 kV or 345 kV — clears these comfortably.
You register as a Generator Owner. BES generating facilities register as GO/GOP. A battery plant discharging to the grid is a generating resource; it registers exactly as a solar or wind plant does.
You apply load-responsive phase protection. Section 4.1.2 conditions Generator Owner applicability on having "load-responsive phase protection systems as described in PRC-023-6 - Attachment A, applied at the terminals of the circuits defined in 4.2.1."
The circuit is in scope under Section 4.2.1. This is where mos t of the analysis lives, and where the storage-specific difficulty appears.

An important and widely misunderstood point about thresholds

There is a persistent piece of misinformation circulating in AI-generated and secondhand compliance content claiming that PRC-023-6 applies when a plant exceeds "75 MVA at a single point of interconnection" or contains "individual units greater than 20 MVA."

PRC-023-6 contains no MVA threshold anywhere — not in the Applicability section, not in the Requirements, not in either Attachment.

Those numbers come from the BES definition and drive whether you register at all. Once registered, PRC-023-6 applicability is determined entirely by (a) the presence of load-responsive phase protection per Attachment A and (b) the voltage class and Planning Coordinator designation of the circuit. Confusing the two leads directly to two failure modes: exempting yourself on a threshold that does not exist in the standard, or performing loadability studies on circuits that were never in scope.

The Category 2 IBR wrinkle — a genuine and current exemption

FERC Order 901 drove NERC's registration of previously unregistered inverter-based resources. Category 2 IBRs — non-BES inverter-based resources at or above 20 MVA aggregate nameplate connected at 60 kV or above — became registrable with a compliance date of May 15, 2026.

Newly registered Category 2 GO/GOPs do not inherit the full standards portfolio at once. NERC published a phased list of standards with Category 2 compliance dates: BAL-001-TRE, IRO-010-5, MOD-026-2, MOD-032-2, PRC-012-2, PRC-017-1, PRC-028-1, PRC-029-1, PRC-030-1, TOP-003-6.1, VAR-001-5 and VAR-002-4.1.

PRC-023 is not on that list. NERC classified it as "Requires Modification" — meaning the standard would need revision before it could be applied to Category 2 IBRs, and no Category 2 compliance date has been established.

Part 3 — The Applicability Analysis: Does PRC-023-6 Actually Reach Your Interconnection?

Part 5 — How Attachment B Puts Your Circuit on the List

Part 1 — Regulatory Background: What PRC-023 Is Actually For

The universal evaluation condition

Every R1 criterion is evaluated at the same operating point:

0.85 per unit voltage and a power factor angle of 30 degrees.

This is an assumed stressed-system condition for evaluating loadability — not a ride-through voltage, not a fault condition, and not a measured value. It represents a depressed-voltage, poor-power-factor state where apparent impedance is at its most threatening to a distance characteristic. Confusing it with PRC-024/PRC-029 ride-through voltage-time curves is one of the more common conceptual errors in this space, and it produces studies that answer the wrong question.

Failure to evaluate at 0.85 pu and 30° is, by itself, an independent path to a Severe VSL under R1 — the VSL table lists it as a standalone "OR" condition alongside failing to use one of the criteria.

The thirteen criteria — and the one that isn't there

R1 offers criteria numbered 1 through 13, but criterion 6 is Reserved. There are twelve usable options, and you need to satisfy only one per circuit terminal. Different terminals on the same circuit may use different criteria.

For a battery plant interconnection, criteria 1, 2, 3, 5, 10 and 12 carry essentially all the weight. Criterion 5 deserves particular attention for storage sited at the end of long radial lines in weak parts of the network — a common siting pattern where land and interconnection queue position are cheapest.

The governing equations

Every distance-relay evaluation reduces to the same relationship, from the NERC Implementation Guidance:

Z relay30 = ( 0.85 × V L-L ) ⁄ ( √3 × k × I rating )

where k is the criterion's margin factor — 1.5 for criterion 1, 1.15 for criteria 2, 3, 7, 8, 9 and 13, and 1.70 for criterion 5. The result is the maximum permissible relay reach at a 30° load angle. Your actual relay reach at 30° must be at or below it.

For a mho characteristic at any maximum torque angle applied to any line angle:

Z relay = ( 1.25 × Z line ) ⁄ cos( MTA − θ line ) Z relay30 = Z relay × cos( MTA − 30° )

And where the relay actually limits circuit capability, the current the circuit can carry with the 15% margin at 0.85 pu is:

I relay30 = ( 0.341 × V relay ⁄ Z line ) × [ cos( MTA − θ line ) ⁄ cos( MTA − 30° ) ]

For three-terminal lines and lines with radial taps, every one of these expressions is rewritten in terms of apparent impedance — the in-feed-adjusted impedance seen from that terminal for a fault at the most electrically distant terminal — and loadability must be evaluated per terminal, because in-feed makes each terminal see a different apparent impedance. Storage projects that tap an existing line create exactly this configuration.

A note on the guidance document

NERC's Determination of Practical Transmission Relaying Loadability Settings V1.1 (March 2024) is published as Proposed Implementation Guidance and is marked "Not ERO Enterprise Endorsed." The equations reflect accepted industry practice and the standard expressly permits other technically comparable methodologies — but the document is not a compliance safe harbour.

If your battery plant is a BES facility with a traditional GO registration — which describes the overwhelming majority of utility-scale storage in ERCOT, CAISO, MISO, PJM and SPP — PRC-023-6 is live for you today, and the Category 2 phase-in provides no shelter.

Common assumptions. 138 kV nominal. Conductor impedance 0.12 + j0.78 Ω/mile (line angle 81.25°). Highest seasonal Facility Rating for the duration nearest four hours: 1,000 A (23 9 MVA). Zone 2 mho phase distance element set at 125% of line impedance.

Criterion 1 threshold, common to both cases:

Z relay30 (max) = ( 0.85 × 138,000 ) ⁄ ( √3 × 1.5 × 1,000 ) = 45.15 Ω

Equivalently, the relay must not operate at or below 1,500 A at 0.85 pu and 30°.

Case A — 18-mile shared gen-tie (BESS co-located with solar)

Z line = 2.16 + j14.04 Ω, |Z| = 14.21 Ω at 81.25°.

Part 7 — Worked Example: A 138 kV BESS Interconnection

Introduction

Part 2 — Why a Battery Storage Plant Is a Generator Owner

Two observations matter enormously for storage, and generic compliance summaries consistently miss both.

First, the transformer categories almost never capture a BESS GSU. Sections 4.2.1.4 through 4.2.1.6 are written in terms of the transformer's low-voltage terminal. A typical battery GSU steps 34.5 kV to 138 kV or 230 kV — its low side is 34.5 kV, so 4.2.1.4 (low side ≥200 kV) and 4.2.1.5 (low side 100–200 kV) cannot reach it. Those categories are aimed at transmission autotransformers, not generator step-up units. Only 4.2.1.6 could theoretically apply, and only if the transformer is part of the BES and your Planning Coordinator affirmatively selects it under R6 — an uncommon but not impossible outcome.

Second, the export exception is written into 4.2.1.1 and into both R6 categories (4.2.2.1 and 4.2.2.2), but not into the transformer categories. That asymmetry is deliberate, and it means the exemption analysis has to be performed per-Element against the specific subsection your circuit falls under — not as a blanket plant-level conclusion.

The exception, and why battery storage strains it

Section 4.2.1.1 excludes:

"Elements that connect the GSU transformer(s) to the Transmission system that are used exclusively to export energy directly from a BES generating unit or generating plant. Elements may also supply generating plant loads."

For a solar plant or a gas turbine this is straightforward: a radial gen-tie from the plant substation to the utility POI exports and does nothing else. It is out of scope, and PRC-025 picks up the loadability obligation instead.

For a battery plant, three separate problems appear.

Problem 1 — "exclusively to export"

A BESS gen-tie carries power in both directions as a matter of normal operation, often at full nameplate, frequently for more hours per year than it exports. Read strictly, an Element that imports several hundred megawatt-hours daily is not "used exclusively to export energy." Read purposively, the exception exists to keep radial generator leads out of a transmission standard, and a battery lead is still radial.

Neither reading is settled. The Technical Rationale for PRC-023-6 states plainly that "No changes are proposed to the Applicability of Reliability Standard PRC-023-6 from the prior version" — meaning the drafting team did not revisit this language when it revised the standard in 2021–2023, and the record contains no discussion of bi-directional resources at all. The language predates utility-scale storage as a mainstream asset class.

There is a meaningful textual signal in favor of treating bi-directionality as significant: Section 4.1.3 makes Distribution Providers applicable "provided those circuits have bi-directional flow capabilities." The drafters knew how to make bi-directional flow an applicability trigger — and they used it elsewhere in the same section.

Problem 2 — shared and hybrid configurations

The exception speaks of energy exported "directly from a BES generating unit or generating plant" — singular. Where a battery plant shares a gen-tie, a collector substation or a GSU with a co-located solar or wind facility owned by a different legal entity, or where two independently registered plants aggregate onto one lead, the "single plant, radial export" characterization becomes difficult to sustain. Hybrid and co-located configurations are now the norm rather than the exception in storage development, and they are the single most common route by which a storage owner ends up in scope.

Problem 3 — loop and three-terminal taps

Many BESS projects interconnect by looping an existing transmission line in and out of the plant substation, or by tapping it as a third terminal, because that is cheaper and faster than building a dedicated radial lead to a distant substation. The moment grid power can flow through your substation between two utility buses, the circuit is a network transmission element. It is not a generator lead under any reading, the export exception is unavailable, and if the line is 200 kV or above it is captured directly by 4.2.1.1 with no Planning Coordinator designation required at all.

The decision path

Are you a BES Generator Owner, or a Category 2 IBR? Category 2 only → PRC-023-6 is not applicable today; monitor NERC standards development. BES GO → continue.
Do you apply load-responsive phase protection per Attachment A at the terminals of any interconnection circuit? (Phase distance, phase overcurrent, out-of-step tripping, switch-onto-fault, communication-aided schemes, and phase fault detectors on current-based comm-assisted schemes that can trip on loss of communications.) No → not applicable. Yes → continue.
What is the circuit's voltage class? ≥200 kV → captured by 4.2.1.1 unless the export exception holds. 100–200 kV or sub-100 kV BES → captured only if your Planning Coordinator selects it under R6 using Attachment B.
Is the circuit radial, shared, or looped/tapped? Looped or tapped → in scope. Shared with another plant or another owner → the exception is very hard to sustain. Dedicated radial → the bi-directionality question is live and must be documented.
Have you received an R6 circuit list from your Planning Coordinator? If your circuit is named on it, the analysis is over — you are in, and your 39-month clock has started.

The disciplined answer for a genuinely radial, standalone BESS gen-tie at 200 kV or above: you probably qualify for the export exception, but you should not treat that as self-evident. Write the position down, cite the Element-level analysis, state your reading of "exclusively to export," get your Planning Coordinator and Transmission Operator to acknowledge it in writing, and retain it. An undocumented exemption is indistinguishable from an oversight when an auditor opens the file three years later.

Part 4 — Effective Dates and the Implementation Plan

This is where storage owners get caught, because the standard has two independent clocks, and the one that matters most to a new BESS is not the standard's effective date.

Clock 1 — The standard's own effective date

PRC-023-6 was developed under Project 2021-05; the Technical Rationale and Implementation Plan are dated January 2023. FERC approved PRC-023-6 by delegated letter order on January 24, 2024, in Docket No. RD23-5-000.

The Implementation Plan sets the effective date as the later of (i) the first day of the first calendar quarter after the effective date of the approving order, or (ii) the effective date of PRC-023-5. Applying that to a January 24, 2024 order places the U.S. effective date at April 1, 2024, with PRC-023-5 retired immediately prior.

Two documentation cautions worth knowing. The Version History table published inside the standard contains an apparent date typo in the Board-adoption row, and the 2024 RSAW's "Regulatory Language" section cites the March 4, 2022 order in Docket RD22-2-000 — which approved PRC-023-5, not -6. Verify enforcement dates against NERC's One-Stop-Shop rather than against the RSAW narrative.

PRC-023-6 has been enforceable for more than two years. There is no grandfathering for storage projects that reached commercial operation after that date.

Clock 2 — The 39-month rule, and why it governs new BESS projects

Section 5 of the standard and the Implementation Plan establish a separate compliance timeline for circuits that become applicable through the Planning Coordinator's R6 assessment:

Each Generator Owner, Transmission Owner, and Distribution Provider that owns circuits that become applicable to this standard pursuant to Requirement R6 shall become compliant with R1 through R5 on the later of the first day of the first calendar quarter 39 months following notification by the Planning Coordinator of a circuit's inclusion on a list of circuits per application of Attachment B, or the first day of the first calendar year in which any criterion in Attachment B applies, unless the Planning Coordinator removes the circuit from the list before the applicable effective date.

Worked timeline

Your Planning Coordinator's annual assessment identifies your 138 kV BESS interconnection under Attachment B criterion B4 and notifies you on March 12, 2027. Thirty-nine months later is June 12, 2030; the first day of the first calendar quarter following that point is July 1, 2030. If the PC's studies indicate the Attachment B criterion first applies in calendar year 2031, the later of the two dates governs and your compliance date becomes January 1, 2031.

Three things follow, and they are the practical heart of this article:

The clock starts at notification, not at commercial operation. A BESS that energizes in 2027 can be notified in 2029 and be compliant by 2032 — but only if it is watching for the notice.
The notice can arrive at any time. R6 Part 6.2 requires the PC to distribute the list within 30 calendar days of establishing it and within 30 calendar days of any change to it. Changes are driven by the PC's power flow work, not by your project schedule.
The Planning Coordinator can also remove your circuit before the effective date, which is a real and legitimate outcome to pursue where the underlying study assumptions have changed.

The other dates you need on a compliance calendar

The trap in R4 and R5: they are annual reporting obligations triggered by a setting choice . If your protection engineer selects criterion 2 or criterion 12 for a terminal and no one tells the compliance team, you acquire a recurring 15-month obligation that nobody is tracking. Both carry Lower Violation Risk Factors — but both have binary-Severe Violation Severity Levels. You either sent the report inside 15 months, or you did not.

If your interconnection is between 100 kV and 200 kV — which covers a very large share of utility-scale storage — you are in scope only through your Planning Coordinator's Attachment B assessment. Understanding that screen tells you how to anticipate the notice.

Any one of six criteria puts a circuit on the list:

B1 — The circuit is a monitored Facility of a permanent flowgate in the Eastern Interconnection, a major transfer path in the Western Interconnection as defined by the Regional Entity, or a comparable monitored Facility in the Québec Interconnection.
B2 — The circuit is selected based on Planning Assessments of the Near-Term Transmission Planning Horizon that identify instability, Cascading or uncontrolled separation adversely affecting BES reliability for planning events.
B3 — The circuit forms an agreed path supplying off-site power to a nuclear plant under NUC-001 Nuclear Plant Interface Requirements.
B4 — The circuit is identified through a de fined power flow screening sequence (detailed below).
B5 — The circuit is selected by the PC based on other technical studies or assessments, in consultation with the Facility owner.
B6 — The circuit is mutually agreed upon for inclusion by the Planning Coordinator and the Facility owner.

B4 is the one that catches storage. The sequence:

Simulate double contingency combinations selected by engineering judgment, with no manual system adjustments between the two contingencies — modeling a System Operator who has no time to react.
For 100–200 kV circuits, evaluate post-contingency loading, in consultation with the Facility owner, against a threshold based on the Facility Rating used in the PC's power flow case.
Where multiple Facility Ratings exist, use the rating for the loading duration nearest four hours.
Apply the duration-dependent threshold below.
Radially operated circuits serving only load are excluded.

Two takeaways for a storage owner.

First, B4 explicitly requires consultation with the Facility owner. You have a seat at the table, and the assumed Facility Rating used in the PC's case is worth verifying, because an understated rating in the power flow model can pull your circuit onto the list unnecessarily.

Second, a large BESS materially changes post-contingency flows on the surrounding 138 kV network in both directions. Charging at full nameplate during an off-peak double contingency is a legitimate stress case, and it is exactly the kind of scenario that pushes a neighbouring circuit past 115%. Your project can put someone else's circuit on the list — a fact worth raising early in interconnection studies rather than discovering in a PC notification letter.

A short, shared gen-tie passes criterion 1 with enormous margin. The compliance work here is documentary, not corrective : because the tie is shared between two plants it does not qualify for the export exception, so the circuit is in scope — and you need a calculation summary on file proving what is obvious to the engineer but invisible to the auditor. This is the single most common storage scenario: in scope, compliant, and undocumented.

Case B — 85-mile radial gen-tie to a remote BESS

Z line = 10.20 + j66.30 Ω, |Z| = 67.08 Ω at 81.25°.

References

The circuits in scope

Section 4.2.1 defines six circuit categories subject to R1–R5:

Part 6 — The Technical Core: Requirement R1 and the Setting Mathematics

NERC Reliability Standard PRC-023-6 — Transmission Relay Loadability (Requirements R1–R6, Attachments A and B, Violation Severity Levels, Version History).
NERC, Technical Rationale for Reliability Standard PRC-023-6, Project 2021-05, January 2023.
NERC, Implementation Plan, Project 2021-05 Modifications to PRC-023, Reliability Standard PRC-023-6, January 2023.
NERC System Protection and Control Working Group, Determination of Practical Transmission Relaying Loadability Settings V1.1 — Implementation Guidance for PRC-023-6, March 2024 (Proposed Implementation Guidance — Not ERO Enterprise Endorsed).
NERC, Reliability Standard Audit Worksheet, PRC-023-6, Version 1, March 27, 2024.
NERC Reliability Standard PRC-025-2 — Generator Relay Loadability.
NERC, ERO Enterprise CMEP Practice Guide: Application of the Registration Criteria for Category 2 Generator Owner and Generator Operator Inverter-Based Resources.
NERC, Reliability Standards Compliance Dates for Generator Owners / Generator Operators (Category 2 IBRs).
IEEE C57.109, IEEE Guide for Liquid-Immersed Transformers Through-Fault-Current Duration ; IEEE C57.91; IEEE C37.113.
FERC Order No. 733, Transmission Relay Loadability Reliability Standard ; FERC delegated letter order, Docket No. RD23-5-000 (January 24, 2024).

This article is provided for general informational purposes and reflects the referenced NERC documents as of publication. It is not legal advice and does not constitute a compliance determination for any specific facility. Applicability of PRC-023-6 depends on facility-specific configuration, registration status and Planning Coordinator determinations. Registered entities should verify current standard versions and enforcement dates against NERC's One-Stop-Shop and consult their Regional Entity and Planning Coordinator.

Part 8 — What Attachment A Actually Sweeps In at a Battery Plant

Attachment A is not a list of relays — it is a list of protective functions that could trip on load current, with or without time delay.

Included

Phase distance; out-of-step tripping; switch-onto-fault; overcurrent relays; communication-aided schemes including POTT, PUTT, DCB and DCUB; and — critically for modern storage — phase overcurrent supervisory elements (phase fault detectors) associated with current-based, communication-assisted schemes such as pilot wire, phase comparison and line current differential, where the scheme is capable of tripping for loss of communications.

That last inclusion catches a great many BESS interconnections. Line current differential is the default choice for a gen-tie, and engineers routinely assume a differential scheme is inherently load-immune. It is — until the loss-of-communications fallback logic arms a phase fault detector that is not. If your 87L scheme can trip on comms failure, the supervising phase element is in scope and must be evaluated.

Excluded

Elements enabled only when other relays or systems fail (loss-of-potential overcurrent; loss-of-communications elements except as noted above); ground fault protection; items 2.3 and 2.4 (Reserved); relay elements used only in Remedial Action Schemes approved under PRC-012 through PRC-017; protection designed only to respond in periods allowing 15 minutes or greater to respond to overload; thermal emulation relays used with dynamic Facility Ratings; and relay elements associated with dc lines and dc converter transformers.

Two exclusions are frequently over-claimed at storage sites. RAS-related elements are excluded only where the RAS is applied and approved under PRC-012 through PRC-017 — an informal owner-designed runback scheme does not qualify. And thermal emulation relays are excluded only where used in conjunction with dynamic Facility Ratings ; a thermal element on a static rating is not exempt.

Switch-onto-fault deserves its own paragraph at a BESS

Appendix D of the Implementation Guidance is explicit: SOTF protection must not operate assuming the line terminals are closed at the outset and carrying up to 1.5 times the Facility Rating. For existing SOTF schemes, the protection must not operate when a breaker closes into an unfaulted line energized from the remote terminal at a voltage exceeding 85% of nominal at the local terminal; for schemes commissioned after formal adoption of the report, the threshold tightens to 75%. Battery plants cycle their POI breaker far more often than a conventional plant ever will, which makes SOTF logic both more valuable and more exposed .

Part 9 — The Inverter-Based Resource Complications

PRC-023-6 was written for a synchronous grid, and applying it at a battery plant surfaces genuine engineering tension that the standard does not address.

Fault current is limited, and that squeezes the setting window from both ends

A synchronous machine contributes 5–6 per unit into a close-in fault. A grid-following inverter contributes roughly 1.1–1.2 per unit, because current limiting is a firmware property, not an impedance. PRC-023 pushes the relay characteristic smaller to preserve loadability, while R1's own text demands "reliable protection of the BES for all fault conditions." At a BESS, the margin between "largest load current the relay must ride through" and "smallest fault current the relay must detect" can approach unity. The classic remedy — enlarge the characteristic — is precisely what PRC-023 prohibits.

Criterion 5 is often the right tool and is often overlooked

Where the maximum line-end three-phase fault current is small relative to the conductor's thermal capability — the textbook definition of a weak source, and an accurate description of many inverter-dominated pockets — criterion 5 permits setting against 170% of the maximum end-of-line three-phase fault magnitude rather than against a thermal rating. The 1.70 factor derives from 1.15 × √2 × 1.05: margin for device error, the √2 relationship at maximum power transfer, and a 1.05 pu bus voltage assumption.

Distance element behaviour degrades

Inverter controls do not produce the negative-sequence signature that conventional directional and phase-selection logic depends on, and IBR current angle is a control decision rather than a physical impedance response. Memory polarization, phase selection and directional integrity all deserve explicit scrutiny during the same study that establishes loadability — because a reach reduction taken for PRC-023 compliance can quietly compromise dependability for a resistive line-end fault.

Quadrilateral and load-encroachment characteristics are the practical answer

Appendix C of the Implementation Guidance notes that quadrilateral and other non-mho shapes allow smaller resistive reach settings for both protection and power swing blocking without encroaching on the loadability characteristic. On modern numerical relays this is usually a better answer than shrinking a mho circle until it no longer sees faults.

Power swing blocking still has to be evaluated, R2's retirement notwithstanding

Appendix C sets out the acceptable mitigation techniques: unblocking timers, adaptive timers based on measured swing rate, negative/zero-sequence reset, quadrilateral shapes, PSB characteristics with logic requiring the locus to be inside both the protection and PSB characteristics, and continuous monitoring of swing centre voltage. Under PRC-023-6 you document these under R1 rather than R2.

Where PRC-023-6 stops and the neighbouring standards start

The elegant point, and the one to take away

PRC-025-2 Section 4.2 expressly lists "Elements that connect the GSU transformer(s) to the Transmission system" as an applicable facility. These are the same Elements that PRC-023-6 Section 4.2.1.1 carves out. The two standards were deliberately drafted to hand the gen-tie back and forth so that no Element goes unaddressed.

Escaping PRC-023 through the export exception does not reduce your obligations — it relocates them to PRC-025-2 and its Attachment 1 setting tables. Any consultant who tells you a battery plant has "no relay loadability obligation" has answered only half the question.

Part 10 — What a Battery Storage Generator Owner Actually Needs to Do

A defensible program has ten components.

Determine and document your registration track. BES Generator Owner or Category 2 IBR. The answer decides whether PRC-023-6 applies at all today. Write it down with the BES-definition analysis that supports it.
Build an Element-level inventory of the interconnection. Every terminal from the collector bus outward: MV collector feeders, LV and HV sides of the GSU, the gen-tie, the POI breaker(s), any tapped or looped transmission. For each, record voltage class, ownership boundary, radial versus networked configuration, and which section of 4.2.1 it could fall under.
Inventory protective functions against Attachment A, not against relay model numbers. Enumerate every function that can trip on load current — 21, 67, 50/51 phase, 68 out-of-step trip, SOTF, POTT/DCB/DCUB, and phase fault detectors supervising 87L schemes that can trip on loss of communications. Do thi s from settings files and logic diagrams, not from equipment lists.
Take a written position on the export exception. For each Element, state whether 4.2.1.1's exception applies, and confront the bi-directional question directly rather than eliding it. Address shared, hybrid, looped and tapped configurations explicitly. Have your Planning Coordinator and Transmission Operator acknowledge the position in writing.
Establish a Planning Coordinator interface. Confirm you are on the distribution list for the R6 circuit list. Ask when the annual assessment is performed. Participate in the Attachment B B4 consultation — the standard entitles you to it. Verify the Facility Rating your PC is using for your circuit in its power flow case, and check whether your charging profile is being modelled in the double-contingency screen at all.
Perform the loadability study at 0.85 pu and 30° for every in-scope terminal. Select the criterion per terminal. Optimize the MTA toward 90° before concluding non-compliance. Evaluate three-terminal and tapped lines per terminal using apparent impedance. Check that any reach reduction preserves fault detection at the reduced fault-current levels an inverter-based plant actually produces.
Escalate the criterion 12 decision to the business. If criterion 12 is the only viable path, the consequences — R3 tri-party Facility Rating agreement, an 87% relay component embedded in the rating, R5 annual reporting to the Regional Entity, and a public derate of the interconnection — belong in front of the asset owner before settings are issued.
Build the evidence package to the RSAW's shape, in advance. The 2024 RSAW asks for a list of relays subject to R1 and spreadsheets or calculation summaries showing each relay is set to one of the criteria; for criterion 10, coordination curves demonstrating the transformer is not exposed beyond its withstand capability; for criteria 7/8/9/12/13, the Facility Rating record plus dated correspondence evidencing PC, TOP and RC agreement; and for R4/R5, dated correspondence proving transmittal inside 15 months. Auditors sample. Every relay in the sample must resolve to a criterion and a calculation.
Put the recurring obligations on a compliance calendar with owners and 90-day reminders. R4 and R5 annual reports, three-year evidence retention, and a standing check for PC list updates. These are the requirements that fail quietly.
Re-verify after every material change. New inverter blocks, a capacity expansion, a co-located resource added to your tie, a relay replacement, a re-rated conductor, a revised Facility Rating, a network reconfiguration that turns a radial lead into a loop — any of these can move an Element into scope or invalidate a prior calculation. The Implementation Guidance is explicit that criterion 3 results "should be reverified whenever major system changes are made."

Where BESS owners get this wrong

Assuming "we're a generator, so PRC-023 doesn't apply." Section 4.1.2 names the Generator Owner directly.
Relying on an MVA threshold that does not exist in the standard.
Treating the export exception as plant-level rather than Element-level.
Overlooking phase fault detectors in line current differential schemes.
Missing the Planning Coordinator's R6 notification and losing months off a 39-month clock — or missing the compliance date entirely.
Choosing criterion 2 or 12 without registering the annual R4/R5 reporting obligation it creates.
Never optimizing the MTA, and concluding a long gen-tie cannot comply when a settings change would have resolved it.
Confusing 0.85 pu / 30° with ride-through, and producing a study that answers a PRC-024 question.
Documenting nothing, because the engineering answer was obvious. An auditor cannot sample your intuition.

Part 11 — Why Work With Keentel Engineering

PRC-023-6 compliance for a battery storage plant sits at the intersection of three disciplines that rarely live in the same team: protection engineering (relay characteristics, apparent impedance, coordination with inverter fault-current limits), transmission planning (Attachment B power flow screening, Facility Ratings, Planning Coordinator negotiation), and NERC compliance (evidence construction, RSAW alignment, recurring reporting obligations).

A study that is technically correct but undocumented fails an audit. A tidy evidence binder built on the wrong applicability determination fails worse. Keentel Engineering brings all three to the same table.

Power System Studies

Short circuit and load flow modelling, relay loadability calculations at 0.85 pu and 30° across every R1 criterion, apparent-impedance analysis for three-terminal and tapped configurations, MTA optimization, and coordination checks that confirm a reach reduction taken for loadability has not compromised fault detection at inverter-limited current levels.

POI Interconnection Engineering Support

Applicability determinations at the Element level, Planning Coordinator and Transmission Operator engagement, Attachment B criterion B4 consultation support, and Facility Rating agreements under R3 with the PC, TOP and RC.

Substation Design and Transmission Line Design

Where the answer is a design change rather than a settings change — protection scheme selection, characteristic shape, instrument transformer ratios, gen-tie configuration — the same team that identifies the problem can engineer the fix.

Utility-Scale Renewable Energy and Owner's Engineer Services

Deep familiarity with how storage and hybrid projects are actually built, contracted and commissioned, so compliance obligations are identified during design rather than discovered during an audit.

NERC O&P 693 Compliance Services

Audit-ready evidence packages structured to the RSAW, compliance calendars for the R4/R5 annual reporting obligations, three-year evidence retention programs, and support through Regional Entity engagement.

Category 2 IBR registration and the broader IBR standards portfolio

PRC-023 does not arrive alone. Keentel supports PRC-024-4 and PRC-029-1 ride-through, PRC-028-1 disturbance monitoring, PRC-030-1, PRC-025-2 generator relay loadability, MOD-026 and MOD-032 modelling, and Category 2 registration — so a single applicability review covers the whole portfolio instead of one standard at a time.

Get in touch

Phone: 813-389-7871 | Email: contact@keentelengineering.com

Offices: Tampa (HQ) · Austin · Sacramento · Baltimore | keentelengineering.com

If you own or are developing utility-scale storage and cannot answer, today, whether your interconnection Elements are in PRC-023-6 scope — or whether your Planning Coordinator has already put your circuit on a list — that is the conversation to have now, while the 39-month clock is still an asset rather than a liability.

Frequently Asked Questions

This is the most useful single result in the article. The identical relay, on the identical line, with the identical 125% reach, fails criterion 1 at an 85° maximum torque angle and passes it at 90°. Nothing about fault detection changed. The mho circle simply rotated, pulling its 30° chord away from the load region.

This is precisely why R1 criterion 12(a) instructs setting the MTA "to 90 degrees or the highest supported by the manufacturer," and why the Implementation Guidance states it is "prudent that the relays be adjusted to as close to the 90-degree MTA setting as the relay can be set to achieve the highest level of loadability without compromising the ability of the relay to reliably detect faults." On long gen-ties — a defining characteristic of remote storage siting — MTA selection is the difference between compliance and non-compliance, and it costs nothing but a settings change.

What if 90° still isn't enough?

Check the other criteria before assuming a problem. Criterion 3 for this line: X L = 66.30 Ω, so I total = 0.816 × 138,000 / 66.30 = 1,698 A, requiring the relay not to operate at or below 115% = 1,953 A — that is, Z relay30 ≤ 34.67 Ω. Criterion 3 is more restrictive here than criterion 1, which is a useful reminder that you should select the criterion that fits your circuit rather than defaulting to whichever one you used last time. You only need to satisfy one.

If no criterion can be met with an acceptable reach, criterion 12 is the designed exit — but it is an exp ensive one. It requires you to:

limit the setting to 125% of apparent impedance with MTA at 90° or the highest supported;
compute the circuit capability — here I relay30 = 1,387 A;
include a relay setting component of 87% of that current in the Facility Rating determination — 1,206 A;
under R3, adopt the calculated capability as the Facility Rating of the circuit and obtain the agreement of the Planning Coordinator, Transmission Operator and Reliability Coordinator;
under R5, report the associated circuits to your Regional Entity annually, no more than 15 months apart, so the ERO can compile a list of every circuit in North America whose capability is limited by protection.

That last point is worth internalizing. Criterion 12 does not merely change a relay setting — it publicly derates your interconnection and puts your project on an ERO-wide list of protection-limited circuits. For a merchant storage asset whose revenue depends on the ability to move energy at scale, that is a commercial decision as much as an engineering one, and it belongs in front of the asset manager before the setting is issued.

The GSU transformer under criterion 10

For a 150 MVA top-rated BESS GSU on a 138 kV system, the maximum nameplate rating corresponds to 628 A on the HV side, so 150% = 941 A. If operations has established an emergency rating of 700 A, 115% of it is 805 A. Criterion 10 takes the greater — 941 A. And under criterion 10.1, any load-responsive transformer fault protection must be shown, via coordination curves, not to expose the transformer beyond the mechanical withstand capability illustrated by the "dotted line" in IEEE C57.109, Clause 4.4, Figure 4.

Criterion 11 provides the alternative for overload protection that cannot meet criterion 10: allow at least 150% of nameplate or 115% of the emergency rating for at least 15 minutes so an operator can act, or supervise tripping with a top-oil element set no lower than 100 °C or a simulated winding hot spot element no lower than 140 °C.

PRC-029-1 Compliance Guide for Inverter-Based Resources

Tue, 18 Aug 2026 19:10:11 GMT

Aug 18, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

When the Model Becomes the Evidence: Responding to APS’s PRC-029-1 Dynamic Model Update Notice

PRC-029-1 voltage ride-through envelope for PV, BESS, and other non-wind inverter-based resources (NERC Attachment 1, Table 2).

Arizona Public Service has issued notice to interconnecting and operating inverter-based resources in its territory requiring upda ted, validated dynamic models demonstrating compliance with NERC PRC-029-1, with submittals due September 15, 2026.

Generator Owners who read that notice as a routine model refresh will under-scope the work. The notice is a combined MOD-032-1 data request and a PRC-029-1 performance demonstration, and the second half changes what counts as evidence. This article walks through what the standard actually requires, where the submittals we see most often fail, and what a defensible package looks like.

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Under PRC-024, a Generator Owner could largely discharge its obligation by demonstrating that protective relay settings were placed outside the “no trip” envelope. The evidence was a settings sheet and a coordination study. The question being asked was “are your trip thresholds in the right place?”

PRC-029-1 asks a different question: when the grid does this, what does your plant actually do?

That question cannot be answered by a settings review, because a plant can trip — or, just as consequentially, stop injecting current — for reasons that have nothing to do with a relay threshold. Inverter-level fault-handling logic, plant-controller loop gains, phase-locked loop behavior under a distorted or weak-grid condition, DC-side limits, and firmware-embedded protective functions all sit upstream of the relay and none of them appear on a settings sheet.

The consequence for the APS submittal is direct: the model is the evidence. If the model does not represent as-left field configuration, the demonstration proves nothing about the facility — and a reviewer who benchmarks positive-sequence response against EMT response will find the discrepancy.

Type 3/4 wind and wind hybrid

Part 2 — Frequently Asked Questions: Large Load Interconnection

2.1 The three operation regions

PRC-029-1 divides the voltage plane into three regions, and the obligation differs in each:

Positive-sequence tools — PSS®E, PSLF — remain the backbone of interconnection-wide study work, and APS needs current, MOD-032-1-conformant positive-sequence models in its base cases. But a phasor-domain model with a typical half-cycle time step and an assumed balanced fundamental-frequency network cannot represent several of the phenomena PRC-029-1 is specifically written around:

PLL dynamics under phase angle jump. The 25° withstand requirement is a synchronization question. A positive-sequence model with an idealized angle reference cannot demonstrate loss-of-synchronism margin.
Weak-grid and low-SCR control interaction. Sub-synchronous control interaction, controller-network resonance, and small-signal instability at low short-circuit ratio are invisible in phasor domain and are precisely the conditions under which fleets have tripped in real events.
Unbalanced faults and negative-sequence response. Single-line-to-ground faults dominate real disturbance statistics, and negative-sequence current injection behavior is an EMT question.
Inverter-internal protection and fault-handling logic. Hardware current limits, DC-link excursions, and firmware-embedded protective functions frequently live only in the manufacturer’s EMT black-box.

The practical rule we apply: positive-sequence demonstrates the plant meets the envelope; EMT demonstrates the model is telling the truth. Where the two disagree, the disagreement itself is the finding — and it is almost always traceable to a control path the positive-sequence model does not represent.

The following consolidates work performed for a U.S. utility-scale storage facility. Identifying details have been removed.

Facility

A standalone battery energy storage facility interconnected at transmission voltage in a non-ISO utility jurisdiction, in commercial operation, subject to BES IBR obligations.

Starting position

The Generator Owner held a positive-sequence model package and an EMT model, both delivered during the interconnection study cycle and both nominally “current.” The facility’s compliance team reasonably assumed the model refresh would be a documentation exercise.

What the model reconciliation found

Firmware drift. The inverter firmware in the field had advanced two revisions past the version the supplied EMT black-box represented. Two protective parameters had changed default values between revisions.
PPC model divergence from as-left settings. The plant controller model carried commissioning-era voltage regulation droop and reactive loop gains. The as-left settings recorded in the commissioning package differed materially, following a post-COD tuning visit that had never been reflected back into the model.
A protective element inside the mandatory region. A plant-level undervoltage element was set at 0.65 pu with a 2.0-second delay. Under the applicable table, the facility is obligated to ride through voltages below 0.70 pu for 3.00 seconds. The element sat inside the region the plant was required to survive — an exposure a settings-versus-PRC-024-envelope review had previously cleared, because the older envelope did not reach there.

What only EMT surfaced

With the corrected firmware parameters and as-left PPC gains represented, positive-sequence simulation showed the plant clearing the full ride-through set. EMT simulation of the same disturbance set at the minimum credible short-circuit ratio showed a poorly damped oscillation in the reactive control loop following fault clearance, converging slowly enough that a second disturbance inside the cumulative ten-second window would have driven the plant into current blocking within the mandatory region.

Resolution

The undervoltage element was reset outside the mandatory region with coordination verified against the plant’s remaining protection scheme. The reactive loop gains were retuned, with the retune validated in EMT and then propagated back into the positive-sequence parameter set. The updated firmware parameters were obtained from the OEM under the existing model NDA and incorporated into both model representations.

Deliverable

A benchmarked model package — positive-sequence and EMT responses overlaid across the full disturbance set, with the applicable ride-through envelope plotted on each — plus a settings gap register documenting each discrepancy found, the change made, and the verification evidence supporting it.

The generalizable finding: all three root causes were model fidelity problems, not plant capability problems. The facility was physically capable of compliant performance throughout. What it lacked was a model that demonstrated it, and a settings review conducted against the right envelope. A submittal built on the original model package would have been internally consistent, professionally presented, and wrong.

5. Requirement R4: narrower than most Generator Owners assume

6. Compliance timeline

2. What PRC-029-1 actually requires

APS’s date is positioned deliberately ahead of the October 1 BES milestone, which leaves review and remediation time inside the compliance window rather than after it. Generator Owners should treat it as firm.

7. How Keentel supports Generator Owners on this request

4. Case study: BESS ride-through failure hidden by a “current” model

R4 provides relief where physical hardware limitations prevent ride-through compliance. Three boundaries on it are frequently misread:

It is available only to IBRs in service at the enforcement date. Facilities still progressing through interconnection studies cannot plan around R4. They must be designed to comply, and a project team treating R4 as a backstop during facilities study is building on a provision that will not be there.

It covers voltage ride-through only. There is no exemption pathway for frequency ride-through or RoCoF. A facility that cannot hold through 5 Hz/s, or cannot maintain synchronism through a 25° phase jump, does not obtain relief under R4 for that limitation.

It is a documentation and communication obligation, not an application for approval. R4.1 requires documentation within twelve months identifying the limitation, the specific requirement affected, the hardware responsible, technical verification that only physical replacement resolves it, and a remediation plan — specific as to which voltage band(s) and duration(s) cannot be satisfied. R4.2 requires that documentation to go to the Planning Coordinator, Transmission Planner, Transmission Operator, Reliability Coordinator, and the Compliance Enforcement Authority, excluding proprietary manufacturer information. Follow-up information requests carry a 90-day response obligation (R4.2.1), CEA acceptance copies a 90-day distribution obligation (R4.2.2), and hardware replacement a 90-day notification obligation after which the exemption lapses (R4.3).

The documentation standard is the part that catches teams out. “The OEM says it cannot do this” is not technical verification. What is required is an engineering demonstration that the constraint is physical, and that no firmware, parameter, or control modification resolves it.

This structure is why “momentary cessation is banned” is an imprecise summary. Current blocking is permitted — but only inside the permissive region, and only with a bounded restart. Cessation anywhere in the mandatory region is a performance failure, and that is where the majority of legacy fleets have a problem.

2.2 Voltage ride-through — and the table split that gets missed

Attachment 1 contains two tables, not one. Type 3/4 wind and wind hybrid resources are held to shorter durations than everything else. PV, BESS, PV+BESS hybrids, and VSC-HVDC-connected resources fall under the second table:

PV, BESS, and other non-wind IBR

Talk to us about your submittal

Our dynamic modeling and NERC compliance practice delivers the full submittal package:

Model development and reconciliation

Positive-sequence dynamic models in PSS®E and PSLF, with MOD-032-1 data sheets refreshed to as-built and as-left configuration
EMT models in PSCAD, including plant controller representation and manufacturer black-box integration under NDA
Firmware and parameter reconciliation against field records, commissioning documentation, and OEM revision histories
Model quality review against applicable regional and interconnection-wide modeling criteria

PRC-029-1 performance demonstration

Full voltage ride-through simulation set against the applicable Attachment 1 table, including cumulative-duration sequences
Frequency ride-through and RoCoF testing with correctly windowed RoCoF computation
Phase angle jump and PLL stability assessment
Low-SCR and weak-grid control interaction screening
Reactive current priority and post-disturbance recovery verification against the R2.3 five-cycle and R2.5 one-second obligations
Positive-sequence to EMT benchmarking with documented reconciliation of any divergence

Protection and controls review

Inverter, plant controller, and plant protection settings reviewed against the PRC-029-1 envelope
Gap register with recommended settings changes and coordination verification
R4 limitation assessment and, where applicable, documentation prepared to the R4.1 evidentiary standard

Submittal and compliance support

Model verification and benchmarking report formatted to the transmission provider’s data request
Evidence retention package structured for audit
Support through reviewer comment cycles

Keentel Engineering supports Generator Owners through NERC IBR compliance — dynamic model development and validation in PSS®E, PSLF, and PSCAD, PRC-029-1 ride-through demonstration, protection and controls review, and submittal preparation for transmission provider data requests.

If your facilities received the APS notice, the constraint is calendar, not scope. Contact us to scope your package.

1. The evidentiary shift: from settings verification to simulated performance

Two implementation details carry more weight than they first appear to:

Voltage is referenced at the high side of the main power transformer, not at the inverter terminals. A plant that passes at the POI can still fail at the inverter, and vice versa, once collector impedance and MPT/GSU vector group and tap position are represented properly. Studies run at the inverter terminal answer a question nobody asked.

Durations are cumulative within a ten-second window. A single ride-through event is the easy case. Sequential faults, reclose attempts, and evolving faults accumulate against the same budget, and a plant that clears each individual event may still exceed the cumulative allowance.

2.3 Frequency ride-through and RoCoF

Frequency is measured over a 3–6 cycle window, and durations are cumulative within a ten-minute period — a different window than the voltage tables use.

RoCoF: tripping is permitted only where the absolute RoCoF magnitude exceeds 5 Hz/s, computed as an average rate of change over a window of at least 0.1 seconds. That averaging window is doing real work in the requirement. An instantaneous df/dt calculation on a fault-adjacent waveform will produce excursions far above 5 Hz/s that the standard does not treat as qualifying, and a RoCoF element configured without an equivalent averaging window will trip on events the plant was obligated to ride through.

Phase angle jump: a minimum withstand of 25 electrical degrees. R1 carves out phase angle changes above 25° that arise from non-fault switching events.

2.4 Two recovery clocks, routinely conflated

This is the most common technical error we encounter in draft submittals:

These are separate obligations with separate clocks. Compressing them into a single “restore active power within five cycles” commits the facility to a performance target roughly four times more aggressive than the standard imposes — and we have seen that error drive unnecessary controller retunes and, in one case, an unnecessary hardware conversation with an OEM.

The reciprocal error is equally costly: assuming the 1.0-second real-power clock covers the current-restart obligation, and overlooking a plant that sits blocked for 15 cycles after voltage recovery.

3. Why positive-sequence models alone will not close the APS request

8. Technical FAQ

Solar Plant Electrical Testing & Commissioning Guide

Mon, 17 Aug 2026 12:15:48 GMT

Aug 17, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Solar Plant Electrical Testing: The Fourteen Tests, and the Four That Actually Find the Fire

A Keentel Engineering Grid IQ technical guide for commissioning engineers, owner's engineers, EPC quality managers, and utility cable asset owners

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

The standard solar commissioning graphic lists fourteen electrical tests: insulation resistance, continuity, earth resistance, earth continuity, polarity, string open-circuit voltage, short-circuit current, I-V curve, transformer tests, HT cable tests, relay testing, breaker testing, CT/PT testing, and inverter testing. Underneath it says test before energization — safety, reliability and performance first.

All fourteen are real tests, and the list is a fair scope summary. But it obscures the finding that ought to reshape how the industry thinks about acceptance testing.

Aerial thermographic survey data across a cumulative fleet of roughly 373 GW-dc puts average power loss at commissioning at 4.46%, against a 2025 fleet-wide average of 5.08%. In other words, plants are being handed over carrying nearly the same anomaly burden as plants that have been operating for years. The gap between "commissioned" and "correct" is now measurable — and it is small, which means most of a plant's lifetime underperformance is baked in before commercial operation.

The fourteen tests are not catching it.

There are three reasons, and this guide is organized around them:

The tests that would find the most expensive defects are all optional. Cross-mated DC connectors — a leading cause of PV fires — pass every mandatory test in the commissioning standard. Insulation defects that only conduct when wet pass every dry insulation resistance test; a peer-reviewed study published in January 2026 found 13.5% of field-aged modules were "dry-pass / wet-fail." Sub-amp ground faults pass every fuse-based ground-fault detector. The tests that catch these three — Category 2 thermography, the wet insulation test, and documentary connector QA — are, respectively, conditional, an "additional test," and not an electrical test at all.

The acceptance criteria are mostly not in the standards. The insulation resistance table is normative.

Almost nothing else is. The 5% band on Voc, the I-V curve pass/fail, the capacity test threshold — all of it is contractual practice, and the standards say so explicitly. One of them states in as many words that determining acceptable results "is beyond the scope of this test method."

The rulebook changed underneath the industry in eighteen months, and a great many specifications have not caught up.

Introduction: Plants Are Being Handed Over 4.46% Down

Part 2 — Frequently Asked Questions: Large Load Interconnection

Between July 2024 and 2026, four foundational documents governing US solar electrical testing changed or are changing:

Part 2 — The Fourteen Tests, Rewritten

Now look at what that means on a 1500 V plant: The code-minimum acceptance value is 1 MΩ. A healthy new 1500 V array reads tens to hundreds of megohms. A string that "passes" at 1.5 MΩ has passed the standard and failed engineering — it is a defect report, not an acceptance.

This is the single most important reframing in this guide: The IEC table is a safety floor — the value below which the array is dangerous. It is not a quality criterion. A commissioning specification that adopts it as the acceptance criterion has adopted a threshold that essentially no defect can fail.

What to specify instead: a floor well above the IEC minimum (many owner's engineers use tens of megohms as an investigation trigger on 1500 V systems), plus string-to-string comparison — the outlier matters more than the absolute value — plus archival of every reading as the baseline the inverter's insulation monitoring device will be judged against for the next thirty years.

Method 1 versus Method 2: The standard offers two:

Test method 1 — test between array negative and earth, followed by a test between array positive and earth.
Test method 2 — test between earth and short-circuited array positive and negative.

The difference is not cosmetic Method 2 requires a short-circuiting device rated for the full array short-circuit current and system voltage — you are deliberately creating a bolted short on a current source that cannot be switched off. Method 1 avoids that, at the cost of two measurements and twice the connection operations on a live DC circuit.

Diagnostically, Method 1 localizes the fault to a pole; Method 2 returns only the parallel combination of the positive and negative resistances. Practical guidance: Method 2 for speed on clean commissioning batches, Method 1 the moment anything reads low.

Conduct the test in low irradiance and dry conditions, measure Voc first, and discharge array capacitance afterwards.

An instrument trap worth flagging. Most handheld commissioning test sets top out at 1000 V insulation test voltage and 1000 V Voc measurement — which is fine for a 1000 V system and insufficient for the 1500 V arrays that dominate utility-scale construction. Verify the instrument's voltage ratings against the actual array before mobilizing. This catches teams out regularly.

Voc is the workhorse field measurement, and understanding why is worth two paragraphs.

Voc depends only weakly (logarithmically) on irradiance, and strongly, linearly and predictably on temperature. Above roughly 200 W/m² you can measure it accurately without a calibrated pyranometer. It is the single best detector of module-count errors, missing or shorted modules, shorted bypass diodes, and reversed modules.

The arithmetic:

Voc(T_cell) = N_modules × Voc(STC) × [1 + β × (T_cell − 25 °C)]

where β is the temperature coefficient of Voc in %/°C, and is negative.

A note on β values, because the commonly quoted range is now out of date. The −0.27 to −0.35 %/°C band that appears in most references is correct for older p-type PERC modules. Modern n-type TOPCon is materially better — −0.24 to −0.27 %/°C is now typical, and heterojunction is better still. A current n-type utility module datasheet gives β = −0.25 %/°C, α (Isc) = +0.045 %/°C, and γ (Pmax) = −0.29 %/°C. Using a legacy β on a modern module will produce an expected Voc that is systematically wrong, and the error runs in the direction of failing good strings on hot days.

Worked example: Twenty-eight modules in series, Voc(STC) = 53.3 V, β = −0.25 %/°C:

At a cell temperature of 48 °C: 28 × 53.3 × [1 + (−0.0025)(23)] ≈ 1,406 V
At −10 °C on a winter morning: 28 × 53.3 × [1 + (−0.0025)(−35)] ≈ 1,623 V

That second number is the string-sizing calculation, and it is over 1500 V — which is exactly the point NEC 690.7 addresses and exactly the reason cold-weather string length is a design decision, not a construction one.

Acceptance. IEC 62446-1 is deliberately comparative rather than absolute here — it requires comparison against expected and against sibling strings without publishing a hard number. The ±5% band is the near-universal industry interpretation, quoted consistently across manufacturer guidance and the SolarPower Europe best-practice guidelines, which also require Voc and Isc measurement on 100% of strings and correction to actual cell temperature rather than STC.

Test 8 — I-V curve

This is where the diagnostic value lives, and where most commissioning campaigns underuse the data they already paid to collect.

The first screen is fill factor: FF = (Imp × Vmp) / (Isc × Voc). Typical crystalline silicon runs 0.75–0.85; thin film 0.55–0.75.

The five deviation classes and what each one physically is:

Part 1 — What the Sequence Is Actually For

A decision tree that works in the field: Is the shape normal? If no, look for steps and notches — that is mismatch. If yes, is the curve short (low Isc), narrow (low Voc), or both? If the dimensions are normal, compute fill factor; a low FF sends you to Vmp/Voc (low implies series resistance) and to the slope of the Isc leg (steep implies shunt resistance).

Voltage-deficit arithmetic is a genuinely useful trick: If the Voc deficit equals one module's Voc, a module is missing or shorted. If it is smaller, individual cell strings are being bypassed — a 60-cell module's bypass diode spans 20 cells, roughly one-third of module Voc. Two adjacent shaded cells in the same cell-string drop the voltage by roughly one-sixth; one shaded cell in each of two cell-strings roughly doubles that drop.

Measurement conditions are what make or break a campaign:

Minimum irradiance 400 W/m², preferably above 600–700 W/m²
Test within the four-hour window centred on solar noon, clear sky
Irradiance stability: a 1–2% change during the sweep produces questionable data; a change above 10% mimics a major array failure. Cirrus is the enemy — irregular variation plus cloud-edge magnification
Wind low — wind changes module temperature mid-sweep
Irradiance sensor: in the plane of array, a reference cell of similar technology designed for backside mounting, not a thermopile pyranometer, to minimize spectral and angle-of-incidence error
Temperature sensor: on the back centre of the module — edges run cooler, and edge-to-centre variation can reach ±15 °C

And one trap that is genuinely under-appreciated: sweep rate on high-capacitance cells High-efficiency architectures — interdigitated back contact, heterojunction — have substantially higher cell capacitance, and the sweep rate should not exceed roughly 10 V per second per cell. Sweep too fast and the capacitance produces an apparent series-resistance signature that is a measurement artefact, not a defect. Teams have condemned perfectly good modern modules this way.

Translation to STC. IEC 60891:2021 Edition 3 provides four correction procedures the classical linear superposition method (Procedure 1), a simplified one-diode model with non-linear irradiance scaling (Procedure 2), interpolation between measured curves requiring no fitting parameters (Procedure 3), and a new Procedure 4 that determines series resistance from a single I-V curve, valid between 300 and 1200 W/m² for crystalline silicon.

The acceptance fight in an EPC contract is almost never about the measurement. It is about which IEC 60891 procedure and which Rs, κ, α and β values were used to translate to STC. Specify the procedure by number in the test plan — it prevents a category of dispute that costs weeks.

Note that neither IEC 62446-1 nor IEC 61829 publishes a numeric I-V pass/fail. Any threshold in your specification is contractual.

Test 9 — Transformer tests

Verify polarity of every DC circuit — modules, source circuits, output circuits, disconnects, inverter input terminations — before closing any disconnect. A reversed string in a combiner is a classic commissioning finding, and it will destroy string fuses or blocking components on first energization.

The thing polarity testing does not catch is the defect described in Part 3: a connector mated to the correct polarity, from the wrong manufacturer.

Isc scales essentially linearly with in-plane irradiance and only weakly with temperature:

Isc(G, T) ≈ Isc(STC) × (G / 1000) × [1 + α × (T_cell − 25)]

At α ≈ +0.045 %/°C, a 40 °C temperature excursion moves Isc by under 2%. Irradiance moves it proportionally.

Which is exactly why Isc is the weaker field measurement, and why it should be treated differently from Voc:

It inherits the full uncertainty of the irradiance measurement. A 3% pyranometer error is a 3% Isc error, directly. Reference cell versus thermopile, spectral mismatch, angle-of-incidence error, soiling on the sensor itself, misalignment to the array plane — all of it lands on Isc.
Irradiance is non-stationary. A 1–2% drift during the measurement is invisible to the operator and fully present in the result.
Soiling is indistinguishable from irradiance error in an Isc reading. Both depress the curve uniformly.
Bifacial modules break the simple relationship entirely. Rear-side irradiance is spatially non-uniform and varies with albedo, row position and height, and is rarely measured well.
The measurement is itself a hazard. IEC 62446-1 permits an operational (in-service) current test as an alternative to a true short-circuit test, for exactly this reason.

Practical guidance: treat Voc as a pass/fail measurement and Isc as a comparative measurement. String-to-string consistency — within roughly 5% of the array mean under stable irradiance — is far more meaningful than absolute agreement with a datasheet.

The earth resistance test on a solar plant is the same engineering problem as on any substation, and it is subject to the same widespread misconception: IEEE Std 80's acceptance criteria are step and touch voltage limits, not a grid resistance value. The commonly specified "1 ohm" or "5 ohm" figures are rules of thumb that appear nowhere in IEEE 80 as requirements.

The correct question is whether the measured grid impedance, combined with the actual available fault current and clearing time, keeps computed step and touch potentials below tolerable limits for the surfacing installed — and whether the measured impedance matches the design model.

For the fall-of-potential measurement, the 61.8% rule derives from a uniform-soil, hemispherical-electrode idealization that a real grid violates. Require the full traverse, plotted, with a visible plateau — and expect the remote current electrode at five to ten times the maximum grid diagonal.

(One currency note for anyone writing a grounding specification: IEEE Std 80-2013 was moved to Inactive-Reserved in March 2024, with a revision project active. A great many specifications cite it as current.)

On the array side, the continuity test is checking something structurally different: that module frames, racking, tracker torque tubes and the equipment grounding conductor form a continuous bonded system across thermal expansion joints, tracker bearings and slip joints. Tracker rotation is the enemy — bonding jumpers across rotating and sliding joints are a recurring construction defect and a recurring inspection finding.

Test 10 — HT cable tests

Medium-voltage collection cable is the highest-value cable testing scope on a solar plant, and the standards moved recently:

Three points that matter on a solar collection system specifically:

Conclusion

The fourteen tests are the right scope. They are not, on their own, an acceptance program.

The acceptance criteria are mostly yours to write:

One numeric table in the commissioning standard is normative, and it is a safety floor that no realistic defect can fail. Everything else — the Voc band, the I-V threshold, the wet insulation limit, the capacity test pass mark — is contractual, and the standards say so explicitly. A specification that names standards without setting criteria has procured a set of measurements, not an acceptance.

The most expensive defects are invisible to the mandatory tests:

Cross-mated connectors, wet-conducting insulation faults and sub-amp ground faults all pass Voc, Isc, insulation resistance and continuity on day one. The three tests that find them — thermography at operating current, wet insulation testing, and documentary connector QA — are conditional, additional and non-electrical respectively. Every one of them is routinely value-engineered out.

And the commissioning record is either a baseline or it is nothing:

The insulation resistance dataset is what the plant's ground-fault protection is judged against for thirty years. The I-V dataset is what degradation is measured from. The thermographic imagery is what next year's survey is differenced against. Recorded properly they are the most valuable engineering asset the plant will ever have. Recorded as pass/fail ticks against a floor, they are a filing obligation.

The data point that should end the argument is the one that opened it:

average power loss at commissioning is 4.46%. Plants are being handed over carrying nearly the full anomaly burden of an operating fleet. That is not a testing technology problem — the instruments and the standards are adequate. It is a specification problem, and it is fixable before the first module arrives on site.

The first two matter enormously and are almost unremarked. Until 2024, utility-scale solar in the United States had no native ANSI-approved acceptance or commissioning specification — only the IEC document, written primarily around European practice and around systems much smaller than a 300 MW single-axis tracker plant. Now it has both: ATS-2025 for equipment acceptance and ECS-2024 for the commissioning process.

Every ATS Section 7 subsection follows the same three-part structure — visual and mechanical inspection, electrical tests, and test values — so §7.29 gives US practice a familiar framework for PV that maps onto how the rest of a substation is already specified. If your solar commissioning specification predates February 2025, it does not reference either document, and it should.

Test 1 — Insulation resistance

The normative table. IEC 62446-1 Table 2 gives the only hard numeric acceptance criteria in the whole commissioning standard, keyed to system voltage defined as Voc(STC) × 1.25:

Test 2 — Continuity

Tests 3 and 4 — Earth resistance and earth continuity

Test 5 — Polarity

Test 6 — String open-circuit voltage

Test 7 — Short-circuit current

The graphic lists "IR, winding resistance, TTR, vector group and more," which is right. The current governing documents are worth stating precisely, because most solar specifications cite superseded editions:

IEEE C57.12.00-2021 — general requirements. Turns ratio must be within ±0.5% of calculated, on all taps, with vector group and phase relationship verified
IEEE C57.12.90-2021 — test code
IEEE C57.152 — field diagnostic testing. (The 2013 edition, which most specifications cite, moved to Inactive-Reserved in March 2024)
IEEE C57.149-2024 — sweep frequency response analysis, superseding the 2012 edition. SFRA on arrival, before the unit is set, is the only practical field method for detecting winding displacement in transit
IEEE C57.104-2019 — dissolved gas analysis, using a percentile-based Status 1/2/3 framework stratified by transformer age and oxygen-to-nitrogen ratio, having replaced the older condition/TDCG scheme

Two solar-specific points. Collection-system transformers are cycled hard — energized at sunrise, de-energized at sunset, every day, for thirty years — so inrush duty, no-load loss and tap changer duty deserve more attention than on a conventional load-serving transformer. And the factory test report must travel with the unit, with per-bushing power factor and capacitance, winding resistance per tap at a stated temperature, and the SFRA baseline. A site result that cannot be compared to a factory value corrected to the same temperature is a number, not a result.

1.1 Which code even applies?

This is the first question on any utility-scale solar project and it is answered wrong more often than any other. NEC 90.2(B) excludes from NEC scope installations under the exclusive control of an electric utility that are on property owned or leased by the utility for the purpose of generation, transformation, transmission, distribution or metering. A solar plant owned and operated by a regulated utility on utility land is therefore outside NEC scope and is designed to the NESC (ANSI C2).NEC Article 691, "Large-Scale Photovoltaic (PV) Electric Supply Stations," exists precisely to cover the gap. Its scope covers facilities with a generating capacity of no less than 5,000 kW that are not under exclusive utility control — the IPP and merchant plant that sells into the grid but is not a utility.

What Article 691 actually does is frequently overstated. It does not blanket-exempt Article 690. It substitutes engineered, PE-stamped, independently verified design for several prescriptive requirements:

691.5 permits approval by engineering review where listed and labeled products do not exist — a genuine relief valve for utility-scale hardware
691.9 permits isolating devices located more than 6 ft from the equipment they isolate where written safety procedures exist, and relieves buildings housing only PV equipment from rapid shutdown
691.10 permits a documented fire-mitigation plan where DC arc-fault protection is not provided
691.6 and 691.7 make all of it contingent on stamped documentation and stamped independent engineering conformance reports available to the AHJ before commercial operation

It also imposes requirements: Qualified-personnel-only operation, restricted access with fencing per 110.31, interconnection through medium- or high-voltage switchgear or a substation, loads limited to generation auxiliaries, and — a sentence worth knowing — large-scale PV electric supply stations shall not be installed on buildings.

Rapid shutdown does not apply to a ground-mounted utility-scale plant: For three independent reasons: 690.12 by its own title and scope reaches only circuits on or in buildings ; 691.4 prohibits these stations from being on buildings at all; and 691.9 expressly substitutes written system shutdown procedures available at the station site.

The practical consequence for commissioning: On a single site you can have the array, DC system and collection system under NEC 690/691 up to the point of change of ownership, and the substation, GSU and interconnection under NESC. Which rule set applies where is a jurisdictional and contractual determination, not a technical one — and it needs to be settled in the specification, not argued at inspection.

One 2026 NEC change large-scale designers should note: The 100 kW threshold has been removed from 690.7(A)(3) and 690.8(A)(1)(a)(3). A licensed electrical engineer may now use engineering calculation methods for a PV system of any size, where previously the method was restricted to systems at or above 100 kW. There is also a new alternative at 690.8(A)(1)(a)(2) permitting calculation per the module manufacturer's instructions — aimed squarely at bifacial modules, where the old Isc × 1.25 rule does not describe reality.

IEC 62446-1 does not state a bare current figure. It achieves the requirement by reference: the standard requires instruments compliant with IEC 61557, and IEC 61557-4 requires a continuity test current of at least 200 mA at an open-circuit voltage of 4 V to 24 V.

A 200 mA continuity test is a continuity test, not a fault-current-withstand test. It confirms a conductive path exists. It says nothing about whether that path will survive a fault, and on a utility-scale plant with kilometres of bonded tracker rows, mechanical grounding lugs and exothermic connections, that distinction matters.

Which is why most owner's engineers additionally specify high-current bonding tests (10 A or 25 A) on a sampled basis, plus separate grounding-grid testing at the substation. Those are practice, not IEC requirements — write them in.

Test 11 — Relay testing

DC hipot is the wrong test for extruded cable. IEEE 400.1's title restricts it to laminated dielectric — PILC, pipe-type, pressurized. On XLPE and EPR, DC stress traps space charge that does not redistribute at 60 Hz and superimposes on the AC field, and on service-aged cable it can convert benign water trees into electrical trees. Use VLF or 20–300 Hz resonant AC.

The sheath integrity test belongs before backfill. It is the only after-installation test that finds pulling damage, rock in the trench bed and tool strikes — and on a solar site with tens of kilometres of directly buried collection cable installed at pace, that is the dominant damage mechanism. IEC 60229 Clause 5 specifies 4 kV DC per mm of oversheath thickness, capped at 10 kV, for one minute, with the metallic layer negative. The "10 kV" everyone quotes is the cap, not a flat value.

Convert the withstand to a monitored withstand. Published evaluation of cable diagnostics found failure-on-test rates of roughly 2.0% at 15 minutes, 2.7% at 30 minutes and 3.7% at 60 minutes — so roughly half the defects a 60-minute test finds are invisible at 15 minutes. Recording tan delta and partial discharge during the same energization costs one instrument connection and no additional outage, and turns "the circuit survived" into "the joint in run 14 has a defect."

Part 3 — The Four Tests That Actually Find the Fire

IEEE C37.233-2023 is the current guide for power system protection testing , superseding the 2009 edition that nearly every specification still cites.

The test hierarchy, and what each layer uniquely catches:

Secondary injection — per-element pickup, dropout, timing and characteristic verification. Catches settings that do not match the approved file.
Primary injection — proves the whole current path from CT primary through secondary wiring to the relay input. Nothing else catches a swapped CT lead or a shorting screw left in.
Functional trip check — every relay output actually trips the intended breaker through all interposing auxili aries, by actual breaker operation, not by jumpering.
End-to-end testing — GPS-synchronized injection at both terminals for line differential and pilot schemes.

On a solar plant, add: anti-islanding and transfer trip scheme verification, directional element polarity verification (a solar plant is a source, and directional elements set for a load-serving feeder will be wrong), and verification that the settings implement the interconnection agreement's ride-through requirements rather than the relay vendor's defaults.

Structure the commissioning record so it can serve as the first PRC-005 maintenance record if the plant is BES-jurisdictional. Building it in the wrong format means re-creating it later.

Test 12 — Breaker testing

Timing per pole and pole scatter, trip and close coil current signatures, minimum pickup voltage of trip and close coils, contact resistance, and insulation resistance pole-to-pole, pole-to-ground and across open contacts. Manufacturer data governs the limits.

The acceptance criterion worth knowing for contact resistance: compare against manufacturer's published range first, and where none exists, investigate values deviating by more than 50% of the lowest value in the comparison set — the lowest , not the average. For poles reading 45, 48 and 70 µΩ, the threshold is 45 × 1.5 = 67.5 µΩ, so the 70 µΩ pole is investigable.

For SF₆ equipment, IEEE C37.122.3-2024 covers gas handling, superseding the 2011 edition. And on any new solar substation, the SF₆ question is now a procurement question — California prohibits acquisition of new SF₆ gas-insulated equipment on a schedule already in force below 145 kV, and New York adopted a comparable schedule in December 2024.

Test 13 — CT and PT testing

IEEE C57.13-2016 governs, with C57.13.1 for field testing of relaying CTs. Field acceptance covers ratio at all taps, polarity, excitation/saturation curve, winding resistance, burden and insulation.

The item most often misunderstood: relaying accuracy class notation is a burden capability statement, not a percentage. All ANSI relaying classes require ratio error within ±10% from 1× to 20× rated secondary current. The numeric suffix is the secondary terminal voltage the CT delivers at 20× rated secondary current without exceeding that error — so a C400 CT delivers 400 V at 100 A, meaning a 4 Ω burden capability. The letter matters too: C means the error can be calculated from the excitation curve, T means it must be determined by test, K adds a guaranteed knee point at not less than 70% of the secondary terminal voltage rating.

And do not interchange knee points. The ANSI 45° knee and the IEC knee (a 50% voltage increase producing a 10% current increase) are different points, and the IEC one is higher. Substituting one for the other in a relay setting review is a real and recurring error.

Polarity errors on CT secondaries invert differential and directional protection — producing a relay that behaves perfectly under secondary injection and trips backwards on a real fault. Confirm by DC flick test: momentary DC into P1–P2, deflection direction observed at S1–S2.

Test 14 — Inverter testing

"Verifies DC input, AC output, protection, alarms and grid synchronization" is a fair summary of the functional scope, and radically understates the compliance scope.

Which framework applies depends on where you interconnect:

Distribution-connected : IEEE 1547.1-2020 specifies type, production, commissioning and periodic tests confirming conformance to IEEE 1547. Its Clause 8 defines four distinct commissioning evaluation paths depending on whether the reference point of applicability is the point of connection or the point of common coupling, and whether the DER is a unit or a composite. Which path applies is a decision to make in the specification, not on site.
Transmission-connected: IEEE 1547 does not automatically apply. The governing requirements come from the interconnection agreement, the FERC LGIA, IEEE 2800-2022, and the applicable NERC standards — principally PRC-024 and PRC-029 for ride-through and protection settings, plus MOD-026 and MOD-027 for model verification.

That last item deserves emphasis, because it is where solar projects most often get caught. The compliance risk has shifted from equipment capability to model fidelity and test evidence. An inverter platform that is fully capable of the required ride-through behaviour will still fail a compliance review if the EMT model does not reproduce that behaviour with the settings actually loaded at the site, validated in combination with the plant controller. Commissioning is where that evidence is either created or lost.

Functional scope that belongs in every inverter commissioning plan regardless: DC input verification against the string configuration; MPPT tracking verification; AC output and power quality; protection settings against the approved file; voltage and frequency ride-through verification; volt-var, volt-watt and frequency-watt response curves; anti-islanding; reconnection timing; alarm point-to-point to SCADA; and communications to the plant controller with setpoint response verified end to end.

Representative ΔT examples (from the informative annex — these are examples, not universal thresholds): a module in open circuit runs 2–7 K above its neighbours; a single-cell hot spot is Class 2 at 10–40 K and Class 3 above 40 K; a heated junction box is Class 2–3 at ≥ 3 K versus a nearby junction box; a transfer-resistance (connection) defect is Class 2–3 above 10 K.

Irradiance extrapolation matters because you rarely inspect at exactly the reference condition: ΔT₂ = (G₂/G₁)^x · ΔT₁, with exponent x = 1.5 for point abnormalities in modules, 1.6 for BOS, and 1.0 for extended-area abnormalities.

Reporting requirements include personnel names and qualifications, camera make/model/serial, environmental conditions including cloud in okta and irradiance in W/m², the emissivity and reflected temperature used, and each finding located by at least two independent methods — serial number, photograph, X-Y coordinates, permanent marking.

The standard recommends a four-year periodic interval. Utility-scale owner practice is typically annual, and increasingly continuous via docked autonomous drones — and the field data supports it: sites using autonomous docked drones show 3.0% average power loss against 5.08% for the rest of the fleet, with quarterly inspection associated with a 36% improvement and five or more inspections a year with a 119% improvement.

Here is the uncomfortable structural fact about the standard fourteen-test list: the three most expensive failure modes in operating solar plants all pass every mandatory test.

3.1 Thermographic inspection — conditional, and it shouldn't be

IEC 62446-1 makes I-V curve measurement and infrared thermographic inspection Category 2 tests — required for larger and more complex systems, but triggered by size and complexity rather than by a MW threshold, and therefore in practice invoked contractually. On utility-scale work they should be mandatory, and SolarPower Europe's best-practice guidance calls for 100% module thermography at commissioning.

IEC TS 62446-3:2017 is the governing document, and it is far more prescriptive than most people who commission aerial IR realize.

Measurement conditions:

Irradiance ≥ 600 W/m² in the plane of the module
For electrical balance-of-system components, ≥ 30% of rated system current
Wind max 4 Bft / 28 km/h
Cloud max 2 okta of cumulus
Low or no soiling; cleaning recommended
Ambient air temperature measured and documented, equipment accuracy ±2 K

Camera requirements:

NETD ≤ 0.1 K at 30 °C
Spectral band 8–14 µm for general use
Geometric resolution: max 3 cm of module edge per pixel, and minimum 5×5 pixels per 6-inch cell
Absolute error < ±2 K, traceable calibration at least every two years
Radiometric image storage required, so absolute temperatures can be recovered later

That radiometric storage requirement is worth writing into a specification explicitly. A JPEG of a hot spot is a picture. A radiometric file is evidence.

The authors conclude that dry-only testing is insufficient to guarantee safety, and propose a conservative dry screening threshold to flag modules needing wet verification.

Where this shows up in operations: intermittent inverter insulation-monitoring trips that correlate with morning dew and rain — the classic "it only faults when it's foggy" ticket — plus accelerated PID and corrosion-driven degradation.

On acceptance criteria, the US standard is refreshingly honest. ASTM E2047, the wet insulation integrity test method for PV arrays, states: "This test method does not establish pass or fail levels. The determination of acceptable or unacceptable results is beyond the scope of this test method." The user must specify the minimum acceptable value. Wet insulation acceptance is always a contractual choice — so make it deliberately rather than by omission.

A practical middle ground short of a full wet test: run the commissioning insulation test early morning while dew is still on the array, and again dry, and compare. Any string with an anomalous dry-to-wet ratio is a defect report. That costs one extra pass and finds a category of defect that a dry-only campaign cannot.

3.2 Wet insulation resistance — an "additional test," and the data says it shouldn't be

The dry insulation resistance test finds gross faults: a damaged cable jacket against a rail, a pinched conductor, a compromised junction box. It does not find micro-cracks in backsheets, marginal junction-box seals, or connector seals that conduct only when wet.

A peer-reviewed study published in January 2026 quantified the gap, testing 37 field-aged crystalline silicon modules dry (1000 V DC, two minutes) and wet (per the module qualification wet leakage method):

Part 4 — DC Arc Flash: Why the Conservative Assumption Is the Wrong One

3.3 Connector cross-mating — not an electrical test at all

This is the defect that passes every test in the standard.

Connectors from different manufacturers may physically mate while having dimensionally different contact geometries, spring-band designs and sealing profiles. The result is elevated contact resistance at the interface, which produces I²R heating, which degrades the polymer housing and seal, which admits moisture, which drives corrosion, which raises resistance further. The failure is progressive and thermally self-reinforcing, which is why it kills years after commissioning rather than at energization.

Prevalence. Roughly 3.5 billion PV connections exist globally, and connectors are described by module testing laboratories as a leading cause of fires instigated by PV systems in many markets. In a large sample of US distributed-generation system inspections, well over half of inspected systems contained safety issues requiring immediate attention, with 40% of urgent safety issues involving field-made or factory-made connectors.

The standards paradox. NEC 690.33 requires that where mating connectors are not of identical type and brand, they shall be listed and identified for intermatability per the manufacturer's instructions. UL 6703 (and its IEC counterpart IEC 62852) theoretically permits evaluated intermatability. But major connector manufacturers' published installation instructions explicitly prohibit cross-mating — and because NEC 110.3(B) mandates compliance with listing instructions, the manufacturer's prohibition supersedes any generic listing in the field. In practical terms, no truly universal PV connector exists.

What a cross-mated connector reads on day one: normal Voc, normal Isc, normal insulation resistance, normal I-V curve. Every mandatory test passes.

It is found only three ways: documentary QA of the bill of materials and connector types at design stage; visual inspection at installation; and thermographic inspection at operating current — which is precisely why the IEC 62446-3 thresholds of ≥3 K on junction boxes and >10 K on transfer-resistance defects exist, and precisely why Category 2 thermography should not be optional.

3.4 The ground-fault detection blind spot

DC ground-fault protection on PV is required for circuits above 30 V or 8 A, and its purpose is fire prevention, not personnel shock protection.

Historically it was implemented as a fuse or breaker of 0.5 to 4 A in the grounded-conductor-to-ground bond inside transformer-isolated inverters. Ground faults in an array are frequently under 1 A — below any of those thresholds. Residual-current detectors typically need on the order of 300 mA or more, and fuse-based schemes often require several amps to blow.

The failure mode. An undetected first fault on the grounded conductor sits invisibly. A second fault elsewhere creates a parallel path that bypasses the detector entirely, allowing large current to flow through racking, module frames and conductors — producing arcing, melted conductors and fire.

The canonical case. Documented in the industry literature: a 383 kW rooftop array where an initial fault of roughly 2.5 A went undetected because it sat below the detector's trip threshold. When a roof expansion joint later separated, a 311 A fault developed and routed through the pre-existing small fault, producing a rooftop fire.

Why this belongs in a commissioning article. Modern transformerless inverters use insulation monitoring devices that measure array-to-ground resistance continuously and trip on a resistance threshold rather than on fuse current. That is a much better protection scheme — and it depends entirely on knowing what "normal" looks like.

Your commissioning insulation resistance data is the baseline that device is judged against for the life of the plant. Recorded properly — per string, with irradiance, temperature and instrument identified — it is the reference dataset for every subsequent ground-fault investigation. Recorded as a tick in a box against a 1 MΩ floor, it is worthless.

Part 5 — Capacity, Performance Ratio and Availability Are Three Different Tests

DC arc flash on PV is genuinely different from AC arc flash, and the industry has been getting it wrong in a direction that surprises people.

A PV array is a current-limited source. It cannot supply unbounded fault current, and the operating point during an arc slides into the non-linear region of the I-V curve depending on irradiance. NREL's measurement work found arc voltages sitting at 10–40% of the I-V maximum power voltage — far below the normal operating point. In one test, 26.7 kW was available at maximum power but only 3.7 kW was delivered during arcing, a factor of about seven.

The consequence: the maximum-power method in NFPA 70E's informative annex overestimates PV incident energy by factors of five to ten versus measurement. Comparisons across calculation methods show 5–70% deviations depending on the method chosen; one commonly used empirical model with standard coefficients overestimated by 33–71%.

Why over-estimating is not "conservative." It drives PPE selection upward, and in a desert solar field in July, heavy arc-rated PPE introduces heat-stress and dexterity hazards that are themselves a real injury mechanism. A calculation that is wrong by 7× in the safe direction has traded one hazard for another without anyone deciding to.

NREL's recommendations are a double-iterative model combining the PV I-V characteristic with arc equations — single-diode models are adequate, within about 5% — and the observation that the industry needs a DC analogue to IEEE 1584.

And the code is moving. NFPA 70E's next edition adds a new article specific to photovoltaic systems, with the committee statement that it "addresses the unique electrical hazards associated with work practices on PV installations." The published first-revision material sets explicit thresholds: a DC arc flash hazard boundary at 150 volts and 1.2 cal/cm², with work on exposed PV terminals exceeding those thresholds prohibited, and manipulation of connectors requiring a risk assessment that determines maximum open-circuit voltage and current-interrupting capability before energized operation.

(Confirm the article number and publication status against the published edition — trade summaries and NFPA's own revision documents disagree on numbering.)

That connector clause is worth pausing on. It effectively makes "pull the connector and see" a prohibited troubleshooting method on a plant above those thresholds, which is exactly how a great deal of string troubleshooting is currently done.

A capacity test isolates what the EPC built. A PR test measures what the plant delivers — which includes weather, soiling, curtailment, grid outages and O&M quality, none of which the EPC controls after handover. Curtailment and grid-driven unavailability must be explicitly carved out of both, or you are penalizing the EPC for the offtaker's behaviour.

The ASTM pair and how they divide the work. ASTM E2939 determines the anticipated capacity at a specific location and sets the reporting conditions — irradiance, ambient temperature, wind speed. ASTM E2848 measures against them. You cannot run a defensible capacity test with one and not the other. (E2848 is currently under revision — a work item is open, which means the dominant US capacity test method is actively moving.)

The data filtering rules decide the outcome more than the measurement does. Typical practice excludes plane-of-array irradiance below 400 W/m²; excludes data where inverter output exceeds 98% of nameplate (clipping); keeps only data within roughly ±20% of the reporting-condition irradiance; constrains the distribution so no more than a 40/60 split sits above and below; excludes snow, frost, beam shading and any period where the inverter is not tracking maximum power; and requires on the order of 500 filtered data points at one-minute resolution.

IEC TS 61724-2 went to Edition 2 in September 2025, and it closes the gap that causes most modern disputes. It is retitled to include a power performance index — measured power over expected power at test conditions — it explicitly accounts for high DC:AC ratios, clipping and curtailment (the 2016 edition assumed unconstrained operation), it is adapted for bifacial systems, it shortens the test to three to five days, and it leans on PV modelling software for expected power rather than pure on-site regression.For a US audience, ASTM remains the dominant contractual method while IEC 61724-2 is more common on internationally financed projects — but the 2025 revision closes the clipping gap that ASTM has not yet closed, which is a genuine reason for US independent engineers to look at it on high DC-ratio plants.

On acceptance thresholds, be careful. No standard sets a capacity test pass threshold. ASTM E2848 provides a method, not a criterion. Published contractual examples cluster around 95–97% of modelled capacity for a pass, with liquidated damages commonly indexed from a 98% reference — but every one of those numbers is negotiated, not standardized.

And the statistical treatment is worth more than the threshold. Sophisticated contracts do not compare point estimates; they require the upper confidence bound of the measured capacity to meet or exceed the guaranteed capacity — which allocates measurement uncertainty explicitly. Whether that uncertainty falls on the EPC or the owner is a negotiated term, and on a large plant it can be worth more than the entire liquidated damages schedule.

The fourteen-test list stops at energization. The tests that determine whether the EPC gets paid come after, and conflating them is the single most common cause of solar EPC disputes.

Part 6 — Three Anonymized Case Studies

Part 7 — A Commissioning Specification Checklist

Confidentiality note. The three engagements below are presented in anonymized and generalized form. No client, developer, EPC, utility, location, vendor or date is identified, and capacities, configurations, sequences and findings have been altered or aggregated. They are included to illustrate recurring engineering and commercial patterns, not to characterize any single project.

Case Study A — The Plant That Passed Every String Test and Ran 4% Down

Situation. A large single-axis tracker plant completed commissioning with a full IEC 62446-1 Category 1 test campaign: continuity, polarity, combiner verification, Voc and Isc on 100% of strings, functional tests, and insulation resistance on every string. The test records were complete and legible, every string passed, and the plant was accepted. Within the first operating year the owner's monitoring showed production consistently below the model by a margin large enough to matter but small enough to be argued about. We were engaged to determine whether the shortfall was a build defect or a model error.

What the review found. The test campaign had been executed competently and had answered a narrower question than the owner believed.

The insulation resistance acceptance criterion was the IEC table minimum. Every string was recorded as "pass, >1 MΩ." A number of strings had in fact been recorded at values in the low tens of megohms — perfectly compliant, and dramatically lower than their neighbours. Because the criterion was a floor rather than a comparison, nobody had looked at the distribution. Re-analysis of the raw commissioning data showed a clear population of outliers clustered in specific blocks.
Voc and Isc had been recorded but not analyzed as a population. Both were checked against a ±5% band string by string. No string-to-string statistical comparison had been performed, so a systematic depression across one inverter block — well inside the 5% band on any individual string — was invisible.
Thermographic inspection had not been performed at commissioning. It was in the specification as a Category 2 item, and had been deferred to the first annual inspection on schedule grounds.
No I-V curve data existed at all. The first aerial thermographic survey, conducted after we recommended it, found a population of module-level and string-level anomalies concentrated in the blocks that the insulation resistance re-analysis had already flagged.

Outcome and lessons. The defects were remediated under warranty, but a year of production had been lost and the commercial position was weaker than it would have been at handover.

Four transferable lessons:

The IEC insulation resistance table is a safety floor, not a quality criterion. On a 1500 V array a 1 MΩ pass threshold is a test essentially no defect can fail. Specify a working floor far above it, and — more importantly — specify string-to-string comparison as the acceptance mechanism.
Commissioning data has to be analyzed as a dataset, not as a stack of pass/fail sheets. The information required to find these defects was already in the records. Nobody had plotted it.
Do not defer Category 2 thermography. Aerial IR at commissioning is cheap relative to what it finds, and it is the only test that catches thermal defects — including connector defects — which pass every electrical test.
The commissioning record is a baseline or it is nothing. Recorded per string with irradiance, temperature and instrument identity, insulation resistance data becomes the reference against which the inverters' insulation monitoring is judged for thirty years. Recorded as a tick against a floor, it has no forward value.

Case Study B — The Capacity Test That Could Not Be Settled

Situation. A plant with a high DC-to-AC ratio reached substantial completion and the contractual capacity test was run. The EPC's analysis showed a pass. The owner's independent engineer's analysis of the same data showed a fail. Both parties had used the same standard, the same measured data and the same modelling package. We were engaged to determine the source of the divergence.

What the review found. Four separate methodological choices, none of them specified in the contract, each of which moved the result by a meaningful margin — and which compounded in the same direction:

Clipping treatment. The plant's DC-to-AC ratio meant a significant fraction of clear-sky midday data was clipped. The test method the contract named was written on the assumption of unconstrained operation. The two parties had adopted different filtering rules for clipped points, and that choice alone accounted for the largest part of the divergence.
The reporting-condition irradiance band. The contract did not specify the acceptable window around the reporting-condition irradiance. One party used a narrow band, the other a wide one, and the two datasets had materially different compositions.
Which regression terms were retained. The wind term's statistical significance was marginal. Retaining or dropping it changed the fitted capacity.
Uncertainty allocation. The contract stated a guaranteed capacity and a measured capacity, and said nothing about confidence bounds. One party compared point estimates; the other compared the lower confidence bound of the measurement against the guarantee.

None of the four choices was unreasonable. All four were unspecified.

Outcome and lessons. The dispute was settled by agreeing a re-test under a jointly written test procedure that fixed every one of those choices in advance. The re-test took three weeks of clear weather to schedule and delayed final acceptance by a quarter.

Four transferable lessons:

Write the test procedure into the contract, not the standard's name. "Capacity test per ASTM E2848" is not a specification. The procedure must fix the filtering rules, the irradiance band, the regression form and the treatment of statistically insignificant terms.
Clipping is the single biggest source of capacity-test dispute on modern plants , and the older methods do not address it. IEC TS 61724-2's 2025 edition explicitly handles high DC:AC ratios, clipping and curtailment, and is worth evaluating on any high-ratio plant even where ASTM is the contractual default.
Allocate measurement uncertainty explicitly. Whether the confidence bound falls in the owner's or the EPC's favour is a negotiated term, and on a large plant it can be worth more than the liquidated damages schedule.
Carve out curtailment and grid unavailability in writing. They belong to nobody's performance and they will otherwise be argued about twice — once in the capacity test and once in every subsequent performance ratio calculation.

Case Study C — The Connector Nobody Tested For

Situation. An operating plant experienced a localized fire in the array field. There were no injuries. The owner engaged us as part of the technical investigation and to review whether the failure mode was present elsewhere on the site.

What the review found. The proximate cause was a DC connector interface that had failed thermally: elevated contact resistance had driven progressive heating, degrading the housing and seal, admitting moisture, driving corrosion, and raising resistance further until the interface arced. The connector had been in service for several years.

Two findings mattered more than the individual failure:

The bill of materials showed connectors from more than one manufacturer in the same circuits, without documentation establishing that the specific combinations were listed and identified for intermatability. The module-side and harness-side connectors physically mated, so nothing in installation flagged them.
The commissioning records were complete and showed no anomaly. Voc, Isc, insulation resistance and continuity had all been within specification for the affected circuits. Every mandatory test had passed , and every one of them would pass again on a cross-mated interface on day one, because contact resistance at commissioning current is too small to register in any of those measurements.

A subsequent thermographic survey at operating current, conducted specifically to look for the signature, identified a population of interfaces running hot relative to their neighbours.

Outcome and lessons. The affected interfaces were remediated and the connector specification was revised for the owner's remaining pipeline.

Four transferable lessons:

Connector intermatability is a documentary control, not a test. It must be caught in design review and procurement — the specification should name the connector type and require that all mating interfaces be from the same manufacturer, or that documented listed intermatability exists for the specific combination.
The manufacturer's instruction governs. Where a connector manufacturer's published instructions prohibit cross-mating, the code requirement to install per listing instructions makes that prohibition binding regardless of any generic listing. In practice this means treating universal intermatability as unavailable.
Thermographic inspection at operating current is the only field test that finds this. Not at open circuit, not at low current — at operating current, with the resolution and ΔT thresholds the thermography standard specifies. The relevant thresholds are on the order of ≥3 K for a junction box and >10 K for a transfer-resistance defect.
A clean commissioning record is not evidence that a plant is free of this failure mode. It is evidence that the tests performed do not detect it. That distinction is worth understanding before it is explained by an investigator.

Part 8 — Frequently Asked Questions

Framework and scope

Applicable code stack decided and stated — NEC 690/691 versus NESC, by system area
ANSI/NETA ATS-2025 §7.29 and ECS-2024 PV sections referenced explicitly, alongside IEC 62446-1
IEC 62446-1 Category 2 mandated — I-V curve and thermography, not optional
Clause 8 additional tests specified: wet insulation resistance, string voltage to ground, blocking diode, shade evaluation

DC array

Insulation resistance acceptance criterion set well above the IEC table minimum, plus string-to-string comparison as the primary mechanism
Insulation test method (1 or 2) specified, with the short-circuiting device rating stated if Method 2
Instrument voltage ratings verified against the actual array — 1500 V, not 1000 V
Voc acceptance band and temperature correction method specified, with the module's actual β
Isc specified as a comparative test with a stated statistical criterion, and an operational current test permitted in lieu of short-circuit
I-V curve measurement conditions specified: minimum irradiance, solar-noon window, irradiance stability limit, sensor type and mounting, temperature sensor location, and sweep rate limits for high-capacitance modules
IEC 60891 translation procedure specified by number, with the Rs, κ, α and β values to be used

Thermography

100% module thermography at commissioning, per IEC TS 62446-3
Camera specification stated: NETD, spectral band, cm-per-pixel, pixels per cell, calibration currency
Radiometric image storage required as a deliverable
Classes of Abnormality adopted as the acceptance framework, with ΔT thresholds and the irradiance extrapolation exponents
Findings located by at least two independent methods

Balance of plant

MV cable: VLF or resonant AC (not DC on extruded), monitored withstand, sheath test before backfill
Transformer: factory test report travels with the unit; SFRA on arrival before setting
Protection: primary injection and functional trip checks by actual breaker operation; directional element polarity verified for a source, not a load
Inverter: applicable framework identified (IEEE 1547.1 path, or interconnection agreement / IEEE 2800 / NERC), with model validation evidence as a deliverable

Safety

DC arc flash study using a PV-appropriate method, not the maximum-power assumption
Connector intermatability documented at design stage and verified at installation
Commissioning record format specified — per test: instrument and calibration date, conditions, technician, criterion and its source, measured values, comparison baseline, disposition

How Keentel Engineering Can Help

Keentel Engineering provides EHV, HV and MV power system engineering to utilities, developers, EPCs, generator owners and public agencies, with offices in Tampa, Austin, Sacramento and Baltimore. Our utility-scale solar practice covers:

Owner's engineer and independent engineer services — specification development, design review, technical bid evaluation, construction-phase QA, and commissioning package audit
Commissioning scope development — IEC 62446-1 Category 2 test plans, acceptance criteria with stated basis, thermography specifications, and turnover record format
Interconnection and POI engineering — feasibility, system impact and facilities studies, POI design support, and large-load and generation interconnection
Power system studies — short circuit, coordination, arc flash including PV-appropriate DC arc flash analysis, grounding grid design and step/touch potential per IEEE 80, and EMT/PSCAD stability and control-interaction analysis
Protection and control engineering — setting calculations, settings management, commissioning test plan development, and independent review of commissioning packages
Performance testing support — capacity and performance ratio test procedure development, independent analysis, and dispute resolution
NERC compliance support — PRC-024/PRC-029 ride-through compliance, MOD-026/027 model verification, and PRC-005 program structure
Specification currency review — auditing existing solar and substation testing specifications against current standard editions

References and Further Reading

PV commissioning and testing standards

IEC 62446-1:2016+AMD1:2018, PV systems — Requirements for testing, documentation and maintenance — Part 1: Grid connected systems — https://webstore.iec.ch/en/publication/24057
IEC 62446-2:2020, Part 2: Maintenance of PV systems — https://webstore.iec.ch/en/publication/27382
IEC TS 62446-3:2017, Part 3: Photovoltaic modules and plants — Outdoor infrared thermography — https://webstore.iec.ch/en/publication/28628
IEC 61829:2015, PV array — On-site measurement of current-voltage characteristics — https://webstore.iec.ch/en/publication/23561
IEC 60891:2021 Ed. 3.0, Procedures for temperature and irradiance corrections to measured I-V characteristics — https://webstore.iec.ch/en/publication/61766
IEC 61724-1:2021, PV system performance — Monitoring — https://webstore.iec.ch/en/publication/65561
IEC TS 61724-2:2025 Ed. 2.0, Power performance index and capacity evaluation method — https://webstore.iec.ch/en/publication/66710
IEC 62548-1:2023, PV arrays — Design requirements — https://webstore.iec.ch/en/publication/64171
IEC TS 62804-1:2025 Ed. 2.0, Test methods for the detection of potential-induced degradation — https://webstore.iec.ch/en/publication/71747
IEC 63027:2023, PV power systems — DC arc detection and interruption — https://webstore.iec.ch/
ANSI/NETA ATS-2025 (§7.29 Solar Photovoltaic Systems) — https://webstore.ansi.org/standards/neta/ansinetaats2025
ANSI/NETA ECS-2024, Standard for Electrical Commissioning Specifications — https://www.netaworld.org/
ASTM E2848-13(2023), Reporting Photovoltaic Non-Concentrator System Performance — https://webstore.ansi.org/standards/astm/astme2848132023
ASTM E2939-13(2023), Determining Reporting Conditions and Expected Capacity — https://store.astm.org/e2939-13r18.html
ASTM E2047, Wet Insulation Integrity Testing of Photovoltaic Arrays — https://store.astm.org/e2047-10r19.html
UL 1699B, Photovoltaic (PV) DC Arc-Fault Circuit Protection — https://webstore.ansi.org/standards/ul/ul1699bed2018

Codes and interconnection

NFPA 70, National Electrical Code (2026), Articles 690, 691, 705 — https://www.nfpa.org/product/nfpa-70-national-electrical-code-nec/p0070code
NFPA 70E, Standard for Electrical Safety in the Workplace — https://www.nfpa.org/product/nfpa-70e-standard/p0070ecode
NFPA 70E A2026 First Revision statements (new PV article; 150 V / 1.2 cal/cm²) — https://docinfofiles.nfpa.org/files/AboutTheCodes/70E/70E_A2026_EEW_AAA_FD_FRStatements.pdf
IEEE 1547.1-2020, Conformance Test Procedures for Equipment Interconnecting DER — https://standards.ieee.org/standard/1547_1-2020.html
OSHA 29 CFR 1910.269 — https://www.osha.gov/laws-regs/regulations/standardnumber/1910/1910.269
EC&M, A New Look at PV Supply Stations (NEC Article 691) — https://www.ecmag.com/section/codes-standards/new-look-pv-supply-stations

Balance-of-plant test standards

IEEE Std 400-2023 and the 400.x series — https://standards.ieee.org/ieee/400/7618/
IEEE C57.12.00-2021 / C57.12.90-2021 / C57.152 / C57.149-2024 / C57.104-2019 — https://standards.ieee.org/
IEEE C57.13-2016, Standard Requirements for Instrument Transformers — https://standards.ieee.org/ieee/C57.13/4867/
IEEE C37.233-2023, Guide for Power System Protection Testing — https://ieeexplore.ieee.org/document/10258048
IEEE Std 80, Guide for Safety in AC Substation Grounding — https://standards.ieee.org/ieee/80/4089/
IEC 60229:2007, Tests on extruded oversheaths with a special protective function — https://webstore.iec.ch/en/publication/1066

Research and field data

NREL/CP-5K00-78331, Methods for Evaluating DC Arc Incident Energy in PV Systems (2021) — https://docs.nrel.gov/docs/fy21osti/78331.pdf
NREL/SunSpec, Best Practices for Operation and Maintenance of Photovoltaic and Energy Storage Systems , 3rd ed. — https://sunspec.org/
Sandia National Laboratories, Recommendations for RCD Ground Fault Detector Trip Thresholds (OSTI 1313068) — https://www.osti.gov/servlets/purl/1313068
Poulek, Beranek, Finsterle & Kozelka, "Dry Pass, Wet Fail: Ground Impedance Testing of Field-Aged PV Modules," Sustainability, January 2026 — https://doi.org/10.3390/su18031212
Raptor Maps, 2026 Global Solar Report — https://pages.raptormaps.com/
Hernday, Field Applications for I-V Curve Tracers (Solmetric / SolarPro) — https://www.solmetric.com/wp-content/uploads/2022/11/SolarPro-FieldApplicationsOf-I-V-Curve-Tracers-Hernday.pdf
Mayfield Renewables, Mitigating DC Connector Risks in PV Systems — https://www.mayfield.energy/technical-articles/mitigating-dc-connector-risks-in-pv-systems/
Fluke, DC ground faults in PV systems — https://www.fluke.com/en-us/learn/blog/grounding/dc-ground-faults-pv-systems
SolarPower Europe, Solar Best Practices — System Commissioning — https://solarbestpractices.com/guidelines/detail/system-commissionin

GIS Substation Engineering: Design, Safety & Testing

Sun, 16 Aug 2026 21:53:05 GMT

Aug 16, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Gas-Insulated Substations, Rewritten as Engineering: Three Claims Worth Challenging and Four Things the Overview Leaves Out

A Keentel Engineering Grid IQ technical guide for utility planners, substation engineers, owner's engineers, and industrial and data-center power teams

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

The standard one-page GIS explainer is genuinely useful. It shows the bus, the circuit breaker, the disconnector, the earthing switch, the CT and the VT, all sealed inside a metal enclosure filled with SF₆. It lists the advantages — compact, high reliability, low maintenance, enhanced safety, suitable for harsh environments — and the applications: urban substations, generation plants, refineries, offshore, rail, and increasingly data centers.

Almost all of that is correct. Three things are not, and they matter.

"Disconnect switch — provides visible isolation." This is the most consequential error in circulation. In a gas-insulated substation there is no visible break. The disconnector's contacts sit inside a sealed aluminium enclosure filled with opaque gas. You cannot see them, and IEC does not require you to. Safety derives instead from a type-tested isolating distance combined with a position indicator that is mechanically linked to the moving contact — a fundamentally different assurance model from the one most US switching and clearance rules were written around. Getting this wrong is not a semantic issue; it is a personnel-safety procedure issue.

"GIS — powering a reliable and sustainable future." SF₆ is the most potent greenhouse gas the IPCC evaluates, with a 100-year global warming potential in the range of 22,800 to 24,300 depending on which assessment report you cite, and an atmospheric lifetime commonly given as 3,200 years. In the European Union, new SF₆ switchgear is already prohibited at medium voltage, and prohibitions bite at transmission voltage in 2028 and 2032 — with a threshold set at GWP ≥ 1, which catches the fluorinated alternatives that were supposed to replace it. In California, acquisition of new SF₆ gas-insulated equipment below 145 kV has been prohibited since January 2025. GIS is an excellent technology; SF₆ is a liability with a schedule attached.

"Requires only 10–20% of the space of conventional AIS." This is a 400 kV number quoted as though it were universal. AIS clearances scale with voltage; GIS bay width barely does. The saving is dramatic at EHV and considerably more modest at 72.5 to 145 kV, and repeating the EHV figure on a 138 kV project will produce a land budget that does not survive first contact with a general arrangement drawing.

And then there is what the overview does not mention at all: very fast transient overvoltages, enclosure circulating currents, what an internal arc does to the gas and to the people nearby, and the fact that a GIS has a commissioning regime that has almost nothing in common with an AIS yard.

This guide covers all of it.

Part 2 — Frequently Asked Questions: Large Load Interconnection

A gas-insulated substation is a coaxial system. The current-carrying conductor runs down the centre of a grounded aluminium tube, supported at intervals by cast epoxy insulators (spacers), with pressurized SF₆ filling the annulus.

That geometry is why GIS is compact. In air-insulated switchgear the insulating medium is the atmosphere, and the phase-to-earth and phase-to-phase clearances required grow roughly linearly with voltage — which is why a 400 kV AIS bay is enormous and a 400 kV GIS bay is a few metres wide. SF₆ at typical operating pressure has roughly two to three times the dielectric strength of air, and about a hundred times better arc-quenching capability, which is what makes the interrupter compact as well as the bus.

But the number that actually governs the design is much lower than the raw strength figure suggests. Reliable service design stress in SF₆ is on the order of 5 kV/mm RMS at power frequency and about 15 kV/mm peak for lightning impulse, at typical operating pressures of 400–600 kPa absolute. The reason for the derating is the single most important physical property of SF₆ for a GIS engineer to understand:

SF₆ is exceptionally sensitive to local field enhancement. Its dielectric strength collapses in a non-uniform field far more sharply than air's does. A sharp protrusion, a burr left by machining, or — most commonly — a loose conductive particle rattling around inside the enclosure will reduce the withstand capability of the gap dramatically.

Everything else in GIS engineering follows from that one sentence. It is why the enclosures are assembled in clean-room-like conditions, why every compartment has a particle trap, why the on-site test begins with a conditioning voltage sequence rather than going straight to the withstand level, and why the commissioning partial discharge test is calibrated against a moving metal particle rather than a void.

2.1 Bus bar and enclosure — and the choice most specifications skip

The infographic describes the bus as the "main current-carrying conductor… enclosed in SF₆ gas for insulation." True, but incomplete in a way that has real cost consequences.

There are two architectures. In a three-phase (common) enclosure, all three conductors share one tube; the phases' magnetic fields largely cancel, so induced enclosure current is inherently small, and the assembly is cheaper and more compact. In a single-phase enclosure, each phase has its own tube; a phase-to-phase fault inside the enclosure becomes essentially impossible, which is why single-phase enclosure dominates above about 245 kV.

Three-phase enclosure is standard at 72.5–145 kV, mixed in the 145–245 kV band, and rare above that. The choice affects footprint, fault-mode assumptions, gas volume per compartment, and — as Part 3 explains — the enclosure bonding design.

2.2 Circuit breaker

The GIS circuit breaker interrupts using SF₆ in a puffer or self-blast interrupter, and the standards that govern it are the same as for any other HV breaker: IEC 62271-100 (current edition 3.0:2021, consolidated 3.1 with Amendment 1 in August 2024).

Wha t a specification should state, and frequently does not, are the class designations:

E1 / E2 — electrical endurance. E2 breakers are designed not to require maintenance of the interrupting parts during expected operating life.
M1 / M2 — mechanical endurance.
C1 / C2 — capacitive current switching, defined by restrike probability. C2 means very low probability of restrike, and it is what you want for capacitor bank, cable and unloaded line switching.
S1 / S2 — S1 for a breaker in a cable system, S2 for a line system or a cable system connected directly to an overhead line.

Writing "IEC 62271-100 compliant" procures nothing in particular. Writing "E2, M2, C2, S2" procures a specific machine.

2.3 Disconnector — where the infographic is wrong twice

Error one: there is no visible isolation.

IEC 62271-102 (current edition 2.0:2018 with Amendment 1:2022) handles disconnectors and earthing switches, and its Edition 2.0 changelog is explicit that it modified "the way to comply with the requirements of the isolating distance of disconnectors" and modified "design and construction requirements for position-indicating devices, aligning the requirements for position indication and signalling." Clause 6.104.3 carries the design and construction requirements for position indication and signalling; Clause 7.105 requires verification tests.

The concept is this: safety comes from a type-tested isolating distance — a gap proven to withstand a specified voltage across the open contacts — combined with a position indicator mechanically linked to the moving contact, not to the operating mechanism or to a limit switch that a broken linkage could defeat. The indicator must be incapable of showing "open" when the contact is not open. That is the substitute for the visible break.

An inspection window is permitted but not required. IEC 62271-203 acknowledges the case where a window exists and requires that dielectric testing be performed with "an earthed metal foil covering the accessible side" — i.e. the standard treats a viewport as a dielectric weak point to be shielded during test, not as a safety feature to be encouraged. Windows are also, as Part 3 explains, one of the apertures through which transient enclosure voltage escapes onto the outside of the enclosure.

Why this matters operationally. Most US utility switching, clearance and tagging procedures were written for AIS, where the operator physically sees an open air gap before applying protective grounds. GIS replaces that with a position indicator, an interlock scheme, and an integral earthing switch. Before a GIS is energized, three things need to be true:

The utility's switching orders, clearance procedures and tagging rules have been reviewed and updated for a no-visible-break environment.
The interlocking — mechanical and electrical — has been fully verified during commissioning, every permitted and prohibited combination, not sampled.
Operator training reflects the change.

Item 2 is one of the highest-value line items in a GIS commissioning scope and one of the first things truncated under schedule pressure.

Error two: "operated only under no-load condition."

Directionally right, dangerously incomplete. A GIS disconnector is not a load-break device, but it does switch something every time it operates: the capacitance of the bus section it is isolating. IEC 62271-102 Clause 5.108 establishes a rated bus-transfer current and rated bus-transfer voltage as declared values with an associated test duty, and Edition 2.0 added a classification of bus-charging switching capability. More importantly, that switching operation is the origin of the single most GIS-specific transient phenomenon in the substation — covered in Part 3.

2.4 Earthing switch — and the classes nobody specifies

The infographic's description ("grounds isolated equipment, protects maintenance personnel, eliminates trapped charges") is correct. What it omits is that earthing switches come in classes, and the difference is safety-critical.

Per IEC 6 2271-102, short-circuit making capacity classes:

E0 — not capable of making short-circuit current
E1 — one short-circuit making operation at rated short-circuit making current
E2 — two short-circuit making operations at rated short-circuit making current

And mechanical endurance classes M0 (100 operations), M1 (1,000 operations — a class newly added for earthing switches in Edition 2.0) and M2 (10,000 operations).

The distinction that matters: a fast-acting (high-speed) earthing switch rated E1 or E2 can be closed onto an energized circuit without destroying itself. An E0 maintenance earthing switch cannot. On a circuit that could be inadvertently energized, or where induced voltage from a parallel circuit is significant, specifying E0 to save money is a decision that will eventually be examined by an investigator.

2.5 Instrument transformers — and the quiet revolution

The infographic shows conventional CTs and VTs. That is still the mainstream, but it is changing, and the change is one of the main routes by which GIS cost is coming down.

Conventional inductive VTs and CTs inside a GIS are large, heavy, gas-filled and expensive, and a VT compartment is a significant source of failure — the independent GIS operational study cited in Part 5 found failures concentrated in circuit breakers, disconnectors and voltage transformers. Low-power instrument transformers (LPITs) — Rogowski coils for current and capacitive or optical dividers for voltage — are substantially smaller, have no ferroresonance behaviour, no saturation, and no risk of an open-circuited secondary. The published economics of GIS cost reduction attribute a meaningful share of it to replacing electromagnetic instrument transformers with optical sensors and Rogowski coils.

The trade is that LPITs deliver a low-energy analogue or digital output rather than 1 A or 5 A, so the protection and metering scheme must be designed for them from the start — typically an IEC 61850-9-2 process bus architecture. That is a system-level decision made at concept stage, not a component substitution made at procurement.

2.6 Combined disconnector/earthing switch

The infographic is right that this is a genuine space and component saving, and it is standard practice. The engineering caveat is interlocking: a three-position device (closed / open / earthed) makes some unsafe combinations mechanically impossible, which is a real safety gain, but it also means the position indication and the interlock logic carry more weight than they would with separate devices. Verify both exhaustively at commissioning.

2.7 The components not on the diagram

Every GIS compartment also contains, and every specification should address:

Spacers — cast epoxy insulators supporting the conductor and, at gas barriers, separating compartments. The gas-barrier spacer is a pressure-retaining component.
Particle traps — low-field recesses adjacent to spacers designed so a hopping particle falls in and cannot escape.
Adsorbers (molecular sieve) — dual-function: controlling moisture, and scavenging reactive decomposition products. This is why an adsorber in a heavily arced compartment must be handled as contaminated waste.
Density monitors, per compartment, with alarm and lockout stages. Density, not pressure — the whole point is temperature compensation.
Rupture discs / pressure relief devices, per compartment, aimed so the discharge plume does not impinge on operator positions or adjacent bays.
Enclosure bonding connections — see Part 3.

3.1 Very fast transient overvoltages

The mechanism. A GIS disconnector moves slowly, because it is not designed to interrupt load current. Switching a short bus section leaves trapped charge on the isolated side. As the contacts travel, the gap breaks down repeatedly — a train of pre-strikes and re-strikes. Breakdown in compressed SF₆ is extremely fast, so each strike injects a step wave into what is, electrically, a low-loss coaxial transmission line. The wave reflects off every discontinuity — open disconnectors, bushings, T-junctions, transformer terminals — and builds standing waves.

The numbers , which vary by source and should be presented as ranges:

Why it matters. Three consequences, in descending order of how expensive they are:

A nanosecond front does not distribute linearly across a transformer winding. It concentrates on the first few turns and discs of the line-end coil, producing inter-turn stresses far above what the winding's BIL rating implies. This is the reason a direct transformer-to-GIS connection needs a VFTO study . The relevant standard, IEC 62271-211, went to a new Edition 2.0 in September 2024 — worth noting, because it superseded a 2014 first edition and added type and routine test requirements plus new safety and environmental clauses.

Transient enclosure voltage (TEV) — the subject of the next section.

EMI into secondary systems. The transient field couples into control cables, instrument transformer secondaries and IEDs. This is why GIS secondary cabling discipline — routing, shielding, shield bonding at both ends, segregation from primary enclosures — is materially more demanding than in an AIS yard.

Mitigation. Damping resistors fitted in series with or parallel to the disconnector main contacts are the established solution, with 110 Ω used in Italian practice and 500 Ω in China, Japan and Korea. High-frequency resonators and nanocrystalline rings on the conductor each offer roughly 20% damping. But the cheapest mitigation is specification: choose disconnectors with inherently low trapped-charge-voltage characteristics (faster contact travel), and require the VFTO study before the arrangement is frozen rather than after.

Note one currency item: IEEE C37.122.1-2014, which explicitly listed very fast transients in its scope, was moved to Inactive-Reserved on 27 March 2025. Its content was largely absorbed into IEEE C37.122-2021 and C37.122.7-2021. A 2026 contract that specifies "per IEEE C37.122.1" is specifying a withdrawn document.

3.2 Enclosure earthing and circulating currents — the opposite of cable practice

This is where an engineer trained on HV cable systems can get GIS badly wrong.

The physics. In a single-phase-enclosed GIS the enclosure is a coaxial return conductor. The centre conductor's magnetic field induces a longitudinal EMF in the enclosure. If the enclosure is bonded at both ends, a current flows in it that nearly cancels the external field. Published figures put the enclosure return current at up to 90% of the operating current in normal service, and similarly high proportions of fault current during a fault, with measured values in one case running 50–85% of a 2,000 A primary current.

Why both-end bonding, when cable sheaths use single-point bonding. For HV cable, single-point bonding eliminates sheath circulating losses and is standard practice. For GIS it is wrong, for three reasons:

Without a return path, the induced EMF appears as a standing voltage on the enclosure — a touch hazard on a structure personnel walk past continuously, at head height, in an enclosed building.
The circulating current is exactly what cancels the external magnetic field, keeping stray field in the GIS hall low and protecting secondary equipment.
The bonded enclosure is the fault-current return path, giving low-impedance clearance and limiting ground potential rise.

So GIS uses multipoint bonding — short, low-impedance interconnections between enclosures and the ground grid at close intervals, on the order of every ten metres, over a high-density earth mat under the hall. Contrast that with AIS practice, where a perimeter loop plus equipment risers is often sufficient.

Transient enclosure voltage. When very fast transients reach an enclosure discontinuity they couple onto the outside of the enclosure and escape through apertures — SF₆-to-air bushings, gas-to-cable terminations, non-metallic viewing ports, insulated flanges. Published magnitudes vary widely; a documented field case gives roughly 45 kV, about 26% of the voltage input to the bushing, on a 525 kV installation, and the literature range commonly quoted is 10–100 kV with durations up to about 10 µs. Simulation values an order of magnitude higher circulate, and should be treated as artefacts of idealized grounding models.

The practical consequences are not electrocution — the documented outcome is "no injuries reported but possible dangerous reactions," meaning startle-reaction falls — but destruction of electronic components in secondary equipment, and sparking in air between grounded parts where multipoint bonding is inadequate.

Touch potential inside a GIS hall is a greater hazard than step potential, which inverts the usual AIS assumption, and it requires dedicated grounding analysis rather than a standard grid calculation.

One currency note worth flagging to anyone writing a grounding specification: IEEE Std 80-2013 was moved to Inactive-Reserved on 21 March 2024, with a revision project active but unpublished. A great many US GIS grounding specifications cite it as current.

3.3 Internal arc, and what the gas does to people afterwards

IEC 62271-203 requires compartments to be fitted with pressure relief devices to limit the external effects of an internal arc, with performance criteria based on arc duration and short-circuit current, and Annex B of the current Edition 3.0 (2022) containing the methods for testing under arcing due to internal fault. Filling pressure relief valves must prevent gas pressure exceeding 10% above design pressure.

The two failure modes the design must prevent are enclosure burn-through — the arc root melting the aluminium wall and ejecting molten metal and hot gas into an occupied building — and enclosure rupture from pressure rise. The rupture disc is sized to relieve before the enclosure yields, and aimed away from operator positions.

What the arc does to the gas is the part that is systematically under-communicated. Per US EPA's technical material on SF₆ byproducts, a typical post-arcing composition includes:

Look at the first and last rows. A particle lying on the enclosure floor and the same particle stuck to a spacer surface are separated by roughly a factor of nine in inception voltage. That single ratio is the argument for everything: for the particle traps, for the conditioning sequence, for the cleanliness regime during erection, and for why a GIS is assembled under conditions closer to a semiconductor fab than to a switchyard.

The mechanism. A free metallic particle resting on the earthed enclosure acquires charge by contact in the applied field. Above a threshold, electrostatic force exceeds gravity and the particle lifts off, beginning a "firefly" hopping motion and recharging on each bounce. Its motion is mechanically rather than electrically synchronized, which is exactly why its PRPD pattern covers all 360° — the diagnostic giveaway that distinguishes a particle from every other defect.

The danger is not the particle in mid-gas. It is migration to and adhesion on a spacer. A particle standing on the epoxy creates a triple junction — metal, solid dielectric, gas — with severe field enhancement, and the solid surface provides a low-strength path. Surface discharge then propagates along the spacer and develops into flashover. Laboratory work tracking this evolution has observed it progressing over hours through distinct stages before transitioning, in a final couple of seconds, to leader-like discharge and flashover.

Practical ranking for a field engineer. In GIS, the two defects that actually cause forced outages are protrusions and free particles. A protrusion is the more immediately dangerous — low inception voltage relative to operating stress, a fixed and permanent field enhancement, and capable of escalating to breakdown under a lightning or switching impulse with little warning. A free particle is the more common, and it is what the CIGRE 5 pC sensitivity criterion is deliberately calibrated against. Voids in spacers are largely a manufacturing-quality issue caught by the factory PD routine test, not a field finding. Floating electrodes erode hardware and need correction at the next outage but rarely fail suddenly.

Acoustic PD detection is complementary rather than redundant here: a hopping particle produces mechanical impacts on the enclosure that an acoustic sensor hears directly and a UHF sensor infers.

Part 1 — What GIS Actually Is

The structural insight: the GIS equipment premium is roughly +20%, and it is more than offset by land, civil, assembly, maintenance and outage savings. The folk claim that "GIS costs two to three times AIS" compares equipment prices, not installed cost.

Voltage matters to the crossover. Published analysis puts GIS at a clear cost advantage at 420 kV and above, with no clear advantage either way across roughly 72.5–245 kV once the complete substation is costed. An important secular trend also sits underneath this: AIS costs fell substantially over three decades while GIS costs were static, with GIS cost reduction now coming from function integration and from replacing electromagnetic instrument transformers with optical sensors and Rogowski coils.

We could not locate any authoritative published US GIS-versus-AIS cost ratio, and the available structured data is European and Indian. For a US audience the defensible framing is: expect an equipment premium of roughly 20% or more; the decision is driven by land cost, permitting timeline and outage exposure; and total installed cost favours GIS wherever land is expensive or scarce — which is precisely the urban and data-center case.

5.3 Schedule — a counterintuitive point

G IS is not automatically faster. It requires a building — or at minimum a foundation and weather protection — that must be complete and clean before erection starts, and GIS erection is a cleanliness-critical, sequential activity that resists parallelization and is highly sensitive to site dust. An AIS yard can be built across multiple parallel work fronts.

The GIS schedule advantage is real, but it lives in permitting and land acquisition, not necessarily in construction duration. Where schedule is the binding constraint on a brownfield bay addition, hybrid switchgear (see 5.5) is frequently the better answer than full GIS.

5.4 Reliability — fewer failures, longer repairs

CIGRE's fourth international reliability survey on switching equipment, covering 2014–2017, reports that "the reliability of the GIS for all voltage ratings is continuously improving through the surveys," with GIS-type designs showing lower failure rates than AIS designs — and, strikingly, live-tank breaker reliability degrading over the same period while dead-tank and GIS-type improved.

Clean benchmark numbers from that survey, per bay-year: 220 kV = 0.0014 failures/yr, 500 kV = 0.00172 failures/yr.

An independent operational study of 110–500 kV GIS matched the 220 kV benchmark almost exactly (0.0015 failures/yr across 2,582 cell-years) while finding the 500 kV rate an order of magnitude worse (0.015 across 454 cell-years — a small sample). Dominant causes: gas gap bridging for major failures; actuator malfunction and gas leakage for minor failures. Failures concentrated in circuit breakers, disconnectors and voltage transformers.

And here is the counterpoint that belongs in every GIS business case. GIS trades failure frequency for failure duration . One utility comparison gives GIS outage time of 56 hours versus AIS at 25 hours, while separately putting GIS breaker and disconnector failures at about a quarter, and busbar failures at about a tenth, of AIS rates.

That is the real reliability story: far fewer failures, but each one takes roughly twice as long to fix. The reason is structural — you must recover gas, open a sealed compartment, work under a clean-condition regime, re-evacuate, refill, re-verify gas quality, and frequently re-test dielectrically before returning to service.

For a data-center or industrial client with a tight recovery-time objective, that MTTR asymmetry may matter more than the MTBF advantage. It is a direct argument for finer gas compartmentalization and for a spares strategy agreed at contract, not after the first failure.

5.5 Maintenance

The most defensible published figure comes from a current manufacturer datasheet: first major inspection after more than 25 years, with expected lifetime more than 50 years. Comparative studies put GIS maintenance cost at roughly 50% of AIS, though one study claims a far more aggressive 80% saving — treat the wide spread as evidence that the answer is fleet- and philosophy-dependent.

5.6 Where hybrid switchgear fits

The decision axis is really how much of the bay do you gas-insulate:

AIS (none) → dead-tank breaker with integral CTs (breaker and CTs) → hybrid / mixed technology switchgear (whole bay in gas, air-insulated busbar) → GIS (everything, including the busbar).

Hybrid products — marketed under names like PASS and HIS, typically available across roughly 72.5 to 420 kV — enclose all high-voltage bay functions in a gas-insulated housing while using conventional air-insulated busbars for connections. They are fully assembled and high-voltage tested in the factory, which is the real proposition: site work and site dielectric testing shrink dramatically.

Hybrids are the right answer when the driver is schedule and civil works rather than absolute footprint — brownfield bay additions into a live AIS yard, generation and renewables interconnections, and data-center feeds where the land exists but the schedule does not. One vendor positions it as "close to GIS compactness at AIS value," which is fair as positioning; note that no manufacturer publishes a quantified footprint reduction or time saving for hybrids, so do not put one in a business case without deriving it.

Part 6 — The SF₆ Problem and the Regulatory Clock

Part 3 — The Four Things the Overview Leaves Out

Four defect types dominate partial discharge findings in GIS: free moving particles, protrusions on the HV conductor or enclosure, floating electrodes, and voids in spacers. Their PD inception voltages differ enormously — and in a way that explains the entire design philosophy.

Measured inception voltages from a controlled GIS study (specific to that test geometry, so treat as illustrative of the ordering rather than as universal values):

That last row deserves emphasis, because it is the industry's best argument: US electric power SF₆ emissions have fallen by about four-fifths since 1990, largely through voluntary leak management and gas handling discipline. The regulatory pressure now arriving is not because the industry failed; it is because the gas is extraordinarily persistent and the remaining emissions still matter.

Over a GIS's life, use-phase leakage dominates the lifetime climate impact, with filling and recovery losses comparatively minor at typical leak rates of 0.05–0.5% per year. Which is exactly why the tightening of the type-test leakage requirement from 0.5% to 0.1% per year per compartment matters more than it sounds.

6.2 United States: no federal ban, two significant state bans

T here is no federal rule restricting SF₆ use in switchgear. Two clarifications worth making explicitly, because both are commonly misstated:

The AIM Act does not cover SF₆. It regulates a statutory list of 18 hydrofluorocarbons. SF₆ is not an HFC and is not on the list, and the technology-transition provisions apply only to HFCs and their substitutes in listed subsectors — none of which is electrical switchgear.
The EPA SF₆ Emission Reduction Partnership has ended. EPA describes it in the past tense and has archived its resources.

Federal reach is limited to reporting, under 40 CFR Part 98 Subpart DD (Electrical Transmission and Distribution Equipment Use). Two changes there matter:

The trigger moved from a nameplate-capacity threshold to an emissions-based 25,000 metric tons CO₂e per year.
Scope expanded beyond SF₆ to all insulating gases with GWP above 1 — which means the fluorinated alternatives are captured too. Only gases with weighted-average GWP at or below 1 (clean air, CO₂/O₂) are excluded.

California is the binding constraint. CARB's regulation at 17 CCR §95350 et seq., amended in 2020 and effective 1 January 2022, prohibits acquisition of new SF₆ gas-insulated equipment on this schedule:

6.1 The numbers

7.1 The three chemistries

Vacuum interruption with clean air insulation. Clean air — technical air, roughly 80/20 nitrogen/oxygen — has GWP zero and no ozone depletion potential, no liquefaction limit, and no regulatory tail of any kind. Interruption is by vacuum, which is mature technology at distribution voltage and has been pushed upward.

Fluoronitrile mixtures (C4-FN with CO₂ and O₂) — marketed as g3 and used within the EconiQ portfolio. Typical HV mixtures run around 3.5–5% C4-FN with 10–13% O₂ in CO₂.

Fluoroketone mixtures (C5-FK) — used primarily at medium voltage and in some 72.5–145 kV applications; C5-FK's higher boiling point limits partial pressure and therefore HV applicability.

7.2 What is actually available at what voltage — the key question

Clean air plus vacuum is commercially mature to 145 kV and demonstration-stage above it. Current portfolios offer 72.5 kV and 145 kV GIS and 145 kV live-tank and dead-tank breakers. 245 kV and 420 kV are EU LIFE-funded demonstration projects: a single 420 kV bay at a Belgian TSO substation running from October 2024 to September 2028 with one year of monitoring and five further years of tracking, and a 245 kV live-tank breaker project with a Spanish TSO kicking off in April 2026. An independent assessment in December 2025 put it plainly: the technology "is not yet fully ready for 420 kV."

Fluorinated mixtures reach the top of the range. Fluoronitrile GIS and breakers are available at 72.5, 145, 245 and 420 kV, with the world's first 420 kV g3 circuit breaker announced in 2022. Fluoronitrile-based EconiQ products extend to 550 kV: the world's first SF₆-free 550 kV GIS was announced for a Chinese utility in May 2025, and the first project described as having the entire 550 kV GIS SF₆-free was announced for a Japanese utility in March 2026. A 420/550 kV gas-insulated busbar using a high-pressure CO₂/O₂ mixture became orderable in December 2025.

So the practical answer for 2026 is uncomfortable but clear: anyone specifying SF₆-free above 145 kV is choosing a fluorinated mixture or a CO₂/O₂ design, not clean air.

7.3 The engineering penalty

Introduction: A Good Diagram With Three Problems

Part 2 — The Components, Corrected

Two consequences follow. Pressure goes up — roughly 9 bar for fluoronitrile mixtures against 6–7 bar for SF₆ — which affects enclosure design, seals, and gas handling equipment. And the minimum operating temperature becomes a design input rather than an afterthought: the liquefaction limit is set by the C4-FN fraction, so a colder site requires a leaner mixture and therefore a larger enclosure or a higher pressure. On a northern US or Canadian site this is a real constraint that must be stated in the specification, not discovered at the factory acceptance test.

7.4 The PFAS problem — the risk nobody prices

Both C4-FN and C5-FK contain fully fluorinated methyl groups and are therefore PFAS under the OECD structural definition. The EU's universal PFAS restriction proposal has been under development since January 2023 and was reportedly narrowed in October 2025, with targeted derogations under discussion — but whether energy transmission equipment or insulating gases receive one, and for how long, was not something we could establish.

An independent research institute's December 2025 assessment states the risk clearly: broad PFAS restrictions under development "could ban both C4F7N and Teflon" — the latter used in arc-quenching nozzles — with current proposals suggesting a complete ban in new products roughly 6.5 years after a restriction enters force, and a potential 20-year postponement for spare parts.

Layer that on top of the EU's GWP ≥ 1 threshold and the position is stark: fluoronitrile mixtures sit above GWP 1, so they face EU prohibition at HV in 2028 and 2032 regardless of the PFAS question. The chemistry that first made SF₆-free transmission switchgear possible is scheduled for prohibition in the market that drove its development.

And there is an unresolved supply question that deserves a direct answer from vendors. The principal manufacturer of the fluorinated fluid used in these mixtures announced in December 2022 that it would exit all PFAS manufacturing by the end of 2025. We could find no public statement from that manufacturer or from the switchgear OEMs confirming whether production ceased, whether the product was carved out, or who the replacement supplier is. If you are specifying fluoronitrile equipment with a 40-year design life, the feedstock supply question belongs in your technical clarifications, alongside the gas top-up and end-of-life recovery commitments.

7.5 What this means for a specification written today

In the United States, below 145 kV: clean air plus vacuum is a real, mature, zero-GWP choice, and in California and New York it is increasingly the only compliant one. Specify it.

In the United States, above 145 kV: SF₆ remains legal and available, and California's ban does not reach above 245 kV until 2033. Fluorinated alternatives are available and are climate-compliant everywhere in the US today. But they are not durable in Europe, and they carry an unquantified PFAS regulatory and supply risk.

Everywhere: the design decision that ages best is the one that reduces gas quantity and leakage regardless of chemistry — tighter type-tested leakage rates, finer compartmentalization to reduce the gas released per intervention, permanent density monitoring with trending rather than alarm-only, and contractual commitments on gas recovery at end of life.

Standards to cite. IEC 62271-4:2022 Edition 2.0 is the current handling standard, and its title changed to cover all insulating and switching gases, with gas-specific annexes — a 2026 specification should cite it alongside IEC 60376 and IEC 60480 rather than relying on any one. CIGRE TB 802 (WG B3.45, 2020) is the reference brochure on non-SF₆ gases and mixtures. On the IEEE side, PC37.122.10, a guide for handling non-SF₆ gas mixtures, is under development with completion expected in the late-2026 to 2028 window.

Plus SF₄, H₂S, and solid metal fluorides as a white, grey or tan powder.

S₂F₁₀ is the one to know about. EPA describes it as "more than 43 times more toxic to cell cultures than the other SF₆ byproducts tested" — and, critically, "generally odorless in pure form at typical environmental temperature." You cannot smell your way to safety. (SOF₂ and SF₄ produce a rotten-egg odour at low concentration, which is a useful but unreliable warning.)

Exposure limits give the scale of the hazard:

Documented human effects from real incidents include burning and watering eyes, nosebleeds, throat irritation, chest tightness and wheezing, coughing (in one case producing blood), nausea, fatigue and headaches — plus pulmonary edema, skin and eye burns, bronchitis and lung haemorrhage in severe cases. One severe case resulted in 45% reduced lung capacity. Symptom resolution ranged from immediate to weeks or months.

The procedure after an internal arc, per IEEE C37.122.3 for heavily arced gas (SO₂ + SOF₂ above 1%):

Connect an additional pre-filter at the inlet of the gas reclaimer.
After evacuation, wait at least one hour before opening, to let switching dust settle.
Wash and neutralize all parts with 10% soda solution or equivalent.
Document the fault fully, including photographs.
Measure oxygen before entering any compartment or low-lying space — SF₆ is roughly five times denser than air and pools in pits, trenches and cable basements.

Required PPE: single-use protective clothing, shoe covers and hair cap; acid-proof gloves; a full-face mask (preferred) or at minimum a breathing protective mask; and goggles. The byproducts are acid-forming on contact with mucous membranes.

Two specification consequences. First, a GIS building needs oxygen depletion monitoring and forced ventilation at low level, with the ventilation and detection design coordinated to the largest single compartment gas mass. Second, the site emergency response plan must be written for this specific hazard before energization, and the people who would respond must be trained on it. This is not a generic arc-flash plan.

3.4 The commissioning regime — which shares almost nothing with AIS

An AIS bay is commissioned by testing its components. A GIS is commissioned by testing an assembled, sealed, gas-filled system that cannot be inspected. The regime is correspondingly different, and it is where NETA-based specifications fall short.

On-site dielectric test. The IEEE approach in C37.122 / C37.122.7-2021 is a field withstand at 80% of the rated low-frequency withstand voltage performed in the factory, applied for one minute, preceded by a conditioning voltage application at levels and durations specified by the manufacturer.

The IEC approach in IEC 62271-203 defines three procedures: a power-frequency test for one minute at the tabulated value for equipment up to 170 kV; the same plus PD measurement at 245 kV and above; or the same plus lightning impulse tests, three of each polarity, as an alternative at 245 kV and above. The standard's own note states the tabulated on-site values are approximately 80% of the type-test values. In the current Edition 3.0 (2022) this content sits in Annex C with the voltages in Table 7 — get the numbers from the standard rather than from a reproduction, because the clause structure changed between editions.

Practically, for a 123 kV GIS the on-site withstand is around 200 kV held for 60 seconds, applied via a series-resonant set operating in the range of roughly 10–300 Hz, with the voltage raised in steps rather than applied directly.

The conditioning sequence is not a formality. Its stated purpose is to "drive any small particles, if they exist, to low electric field intensity locations such as particle traps." It is the mechanism by which manufacturing and erection debris is neutralized before the substation carries load. Skipping or shortening it because the outage is tight defeats the main purpose of the test.

Standard setup items: all enclosures solidly earthed; instrument transformer secondaries shorted and earthed; surge arresters, CVTs and electronic sensors disconnected; and any inspection windows covered with earthed metal foil.

Partial discharge by UHF, not by IEC 60270. On site, the conventional apparent-charge method at frequencies below about 1 MHz is swamped by ambient noise — corona from adjacent energized equipment, converter noise, and the resonant test set itself. A partial discharge in SF₆ has a sub-nanosecond current rise and radiates strongly into the GHz region, and the GIS enclosure acts as a coaxial waveguide that propagates those UHF modes while attenuating low-frequency ambient. Practical instrumentation covers roughly 100 MHz to 2 GHz, with the diagnostically useful band commonly 300 MHz to 1.5 GHz.

The sensitivity verification procedure is what makes a UHF result mean anything, and it is the item most often missing from a commissioning specification. The CIGRE method — set out in TB 654 (WG D1.25, 2016) and extended in TB 933 (WG D1.66, 2024) — is two-step:

In the laboratory, an artificial PD source (a hopping metal particle, typically 3–5 mm long and 1 mm diameter) is placed in a reference GIS section and adjusted to produce 5 pC apparent charge measured per IEC 60270. Simultaneously an electrical pulse of known amplitude is injected at the same location, and the amplitude producing the same UHF response is recorded.
On site, that same calibrated pulse is injected into the installed GIS, and the installed UHF system must detect it at every sensor position of interest.

Detecting the pulse demonstrates the system can see a 5 pC defect. Without that verification, "no PD detected" means only that nothing exceeded an unknown threshold.

Gas quality acceptance. Three measurements are mandatory before energization: moisture content, purity (SF₆ percentage) and density against the manufacturer's nominal filling requirement. Reference values:

A practical detail worth writing into the specification: measure moisture at installation and again at least five days after final filling. Water is released slowly from the epoxy spacers themselves, and the equilibrium value is what matters. In-service moisture can rise for years after commissioning with zero external ingress.

Leakage. The long-standing requirement in IEC 62271-203 is that leakage from any single compartment to atmosphere and between compartments shall not exceed 0.5% per year for the service life of the equipment. Edition 3.0 (2022) tightened the type-test requirement for gases with GWP above 1,000 from 0.5% to 0.1% per year per gas compartment — and the market has moved with it: current manufacturer datasheets for 145–170 kV GIS now quote type-tested leakage below 0.1% per year per compartment, where older brochures for the same product family said 0.5%.

The rest of the scope: contact resistance on all main current-carrying circuits including bus joints, breakers, disconnectors, earthing switches, bushings and cable connections; timing tests; mechanical operation tests; and — the item worth doubling the allocated time for — exhaustive interlock verification.

A note on NETA. ANSI/NETA ATS-2025 has no GIS-specific section. GIS falls across the switchgear, busway, SF₆ switch and SF₆ circuit breaker sections. The practical consequence for an owner's engineer: a specification clause reading "test in accordance with ANSI/NETA ATS" does not procure a GIS commissioning scope. You must explicitly add the on-site dielectric test per IEEE C37.122.7 or IEC 62271-203 Annex C, the UHF PD measurement with CIGRE sensitivity verification, and the gas quality regime — or you will not get them.

Part 4 — Free Particles: The Defect That Drives the Whole Design

Part 5 — GIS versus AIS, Honestly

5.1 Footprint — the figure to stop repeating

The published claims for GIS footprint relative to AIS range from 10–20% to "usually less than half," and almost none are broken out by voltage class. One government-adjacent study applies a flat 80% reduction across 132 kV, 220 kV and 400 kV, which is exactly the oversimplification to avoid.

The physics is not complicated. AIS clearances scale roughly with voltage — phase-to-earth clearance grows from around a metre at 145 kV to several metres at 420 kV, and bay width, height and length grow with it. GIS bay width grows far more slowly: a current 145 kV GIS product family offers bay widths of 650, 800 and 1,200 mm, and even a 550 kV bay is measured in a small number of metres. So the ratio improves sharply with voltage.

The honest framing is that the saving widens with voltage — modest at 72.5–145 kV, substantial at 245 kV, dramatic at 420 kV and above. The 10–20% figure is a 400 kV-plus number. If footprint is the deciding factor on your project, derive it from the actual GIS bay dimensions in the vendor's general arrangement and your own AIS layout at the same voltage and configuration. It takes an afternoon and it is the only number that will survive scrutiny.

5.2 Cost — the equipment premium is not the story

The most structured published comparison is a utility ten-year life-cycle cost study of an H-configuration with three circuit breakers, with AIS normalized to 100%:

(Read §95352 Tables 1 and 2 directly before relying on a boundary — the published renderings differ slightly on inclusive/exclusive treatment at 38, 145 and 245 kV, and that distinction determines whether a specific purchase is caught.)

CARB also sets annual emission limits with an annual emission factor of 1.0% for systems at or above 10,000 MTCO₂e capacity and 2.0% below, dropping slightly from 2035, with a floor of 50 MTCO₂e. Exemptions are available — and the most commonly used ground is that non-SF₆ equipment of the required type and rating is unavailable from at least two suppliers, which, as Part 7 shows, is a live argument above 145 kV.

New York adopted 6 NYCRR Part 495 in December 2024, with a phase-out schedule running from 2027 to 2030 by voltage and interrupting rating, a 1% systemwide emission limit on a rolling three-year average from 1 January 2030 for owners emitting above 7,500 MTCO₂e, first reporting year 2027, and an exemption window with requests due by 1 September 2026.

Massachusetts regulates leak rate rather than acquisition: 310 CMR 7.72 caps the maximum annual SF₆ emission rate at 1.0% from 2020 onward, and requires that newly manufactured GIS placed under an owner's control on or after 1 January 2015 be represented by the manufacturer as having a 1.0% maximum annual leak rate. There is no Massachusetts prohibition on SF₆ equipment.

(We found no SF₆ switchgear rule in Washington State, and no federal procurement preference for SF₆-free equipment. If a specification claims either, ask for the citation.)

6.3 Europe: the schedule that reshapes the product market

Regulation (EU) 2024/573 replaced the 2014 F-gas Regulation, published in the Official Journal on 20 February 2024. Its switchgear prohibitions, corroborated across government and DSO sources though the Annex text should be read directly before relying on it:

Three points that matter enormously.

The HV threshold is GWP ≥ 1, not GWP ≥ 1,000. That means the fluoronitrile and fluoroketone mixtures that first replaced SF₆ at transmission voltage are themselves prohibited at HV from 2028 and 2032. Only chemistries with GWP below 1 — clean air, CO₂/O₂ — survive. The technology that solved the problem is on the same clock as the problem.

The trigger is "placing on the market." At least one DSO defines it as handover to the operator for use in its final location rather than manufacture or energization, with equipment ordered before the regulation's entry into force exempt on documentary evidence. Verify against the transitional provisions before relying on it for a specific order.

From 1 January 2035, only reclaimed or recycled SF₆ may be used for maintenance or servicing of electrical switchgear in the EU. For anyone operating an SF₆ fleet in Europe, that is a substantial O&M planning item that arrives well inside the life of equipment being installed today.

Elsewhere: the UK's post-Brexit position does not clearly mirror the EU schedule and should not be assumed to. Japan and South Korea rely on voluntary commitments and emissions trading respectively rather than prohibitions. China has no national switchgear rule — and is simultaneously the largest SF₆ emitter and the buyer of the world's first SF₆-free 550 kV GIS, which is a contradiction worth noticing.

Part 7 — SF₆-Free Alternatives: Where the Technology Actually Is

Part 8 — Extending a GIS: The Decision You Make Once

This is the most commercially consequential GIS topic and it is entirely absent from every overview.

There is a dedicated standard: IEEE C37.122.6-2013, Recommended Practice for the Interface of New Gas-Insulated Equipment in Existing Gas-Insulated Substations Rated above 52 kV , currently under revision with a scope extended to cover GIL-to-GIS interfaces. Its existence tells you the problem is real.

Four difficulties compound:

Vendor lock-in. GIS enclosure geometry, flange patterns, conductor diameters, spacer designs, gas pressures and interlock philosophies are proprietary. There is no interchangeability standard equivalent to AIS bus dimensions. In practice, extension means returning to the original OEM or engineering a bespoke transition module.

Obsolescence, on a 50-year asset. GIS service life exceeds 50 years and manufacturer datasheets say so. Over that span the original bay design is very likely out of production, the OEM may have been acquired — the industry has seen exactly that at scale — and spares and tooling may be gone. A 25-year-old GIS being extended in 2026 is a normal case, not an exceptional one.

Dielectric testing the extension without de-energizing the adjacent bays. The new section needs its on-site withstand test at approximately 80% while the adjacent bays are live. That requires a gas-tight, dielectrically rated isolating or buffer compartment, temporary earthing and short-circuiting arrangements, and a manufacturer risk assessment. If the original GIS was not designed with buffer compartments or spare disconnector positions, this can be impossible without a full busbar outage — which is the true cost of a cheap original design, paid twenty years later.

Gas handling on a live installation — evacuating and refilling adjacent compartments and re-verifying gas quality across the interface.

The specification tool that solves half of this is CIGRE TB 870, Service Continuity Guide for HV GIS above 52 kV (WG B3.51, 2022), which introduces the MRE code — a standardized two-numeral notation for service continuity during Maintenance, Repair and Extension, where the first numeral is the number of busbar sections out of service and the second is the number of feeders. "MRE12" means one busbar section and two feeders out for the activity.

This is genuinely useful. It lets an owner state a service-continuity requirement in the tender rather than discovering the outage consequences after commissioning. TB 870's design levers for improving it: additional busbar separations, additional disconnectors and earthing switches, buffer compartments, and intermediate support insulators — with the honest caveat that higher service continuity correlates with higher complexity and cost. It also gives a layout recommendation worth adopting: distribute critical feeders on opposite sides of a busbar sectionalizer rather than clustering them.

The design recommendation is simple: buy extensibility at the outset. Specify spare bay positions with capped busbar extensions and buffer compartments, state the required service continuity using the MRE code, and secure long-term spares and interface-drawing commitments contractually. Retrofitting extensibility is dramatically more expensive than designing it in — and sometimes it is not possible at any price.

Part 9 — Three Anonymized Case Studies

Confidentiality note. The three engagements below are presented in anonymized and generalized form. No client, utility, location, vendor, manufacturer or date is identified, and voltages, configurations, sequences and findings have been altered or aggregated. They are included to illustrate recurring engineering and commercial patterns, not to characterize any single project.

Case Study A — The Bay That Could Not Be Added

Situation. An owner needed to add two feeder bays to a high-voltage GIS installed roughly two decades earlier, to serve a new large load. The substation had physical room in the building. The budget assumed a straightforward extension. We were engaged to develop the scope and the outage plan.

What the review found. The extension was feasible. It was also going to cost several times the assumed figure and require an outage the owner had not contemplated, for four compounding reasons:

The original OEM no longer produced that bay design, and corporate consolidation meant the successor organization's engineering records for the platform were incomplete. A bespoke transition module had to be engineered and type-tested.
The original design had no buffer compartments and no spare disconnector positions. The busbar was continuous through the region where the new bays would connect. There was no way to isolate a section for the new work while keeping the adjacent bays energized.
Consequently the on-site dielectric test of the extension — approximately 80% of the type-test value, applied to a section physically continuous with live equipment — could not be performed without de-energizing the busbar. The outage required to commission two feeder bays was a full busbar outage on a substation that had no redundancy at that voltage.
The interlocking philosophy had changed between the original design and current practice, so the new bays' interlock scheme had to be reconciled with the existing one rather than simply extended.

Outcome and lessons. The project proceeded with a staged outage negotiated over two seasons, at materially higher cost and schedule than budgeted. The transferable lessons:

Extensibility is a design decision made at the original build, and it is nearly free then. Spare bay positions with capped busbar extensions and buffer compartments add a small percentage to the original cost and remove an entire category of future risk. Retrofitting them is often impossible.
Specify service continuity in the tender, not in hindsight. The CIGRE MRE code exists precisely so an owner can state "maintenance, repair and extension shall be achievable with no more than one busbar section and one feeder out of service" as a requirement the bidder must design to.
Secure long-term commitments contractually — spares availability, interface drawing release, and a defined obsolescence notification period. Fifty years is longer than most corporate memories.
A GIS building with physical space in it is not the same thing as an extensible GIS. The constraint is electrical and procedural, not spatial, and it is invisible on a general arrangement drawing.

Case Study B — The Commissioning Scope That Said "Per NETA"

Situation. A high-voltage GIS forming part of an industrial facility's incoming supply was approaching energization. The commissioning specification, drafted by a competent team more familiar with air-insulated and medium-voltage work, required testing "in accordance with ANSI/NETA ATS." The contractor had priced accordingly. We were asked to review the commissioning package before energization.

What the review found. The contractor had done exactly what was asked and it was not close to sufficient — because the referenced standard has no GIS-specific section. Four items were missing entirely:

No on-site dielectric withstand test. The GIS had been assembled, gas-filled and checked for leaks, and the assumption was that factory type testing covered dielectric performance. It does not. The on-site test exists specifically to prove that a system assembled on site — with field joints, field gas filling, and whatever debris the erection process introduced — is sound. Its conditioning phase is the mechanism by which erection debris is driven into particle traps before the substation carries load.
No partial discharge measurement. And no UHF sensors specified, which meant retrofitting external sensors at accessible flanges rather than using integral internal ones — reducing sensitivity and coverage.
No sensitivity verification. Even once PD measurement was added, the initial proposal was to record UHF activity and report "no PD detected." Without the CIGRE calibrated-pulse verification, that statement means only that nothing exceeded an unknown threshold at an unknown detection capability.
Interlock verification was sampled, not exhaustive. In a substation with no visible break, where the operator's assurance that a circuit is isolated comes from a position indicator and an interlock scheme, sampling is not a defensible basis.

Gas quality measurement was present but scheduled once, immediately after final filling — before moisture released from the epoxy spacers had reached equilibrium.

Outcome and lessons. The scope was extended, the schedule slipped modestly, and the tests were performed. PD was detected during the withstand test and localized to a compartment; investigation found foreign material introduced during erection. It would not have been found by any test in the original scope.

Three transferable lessons:

"In accordance with ANSI/NETA ATS" does not procure a GIS commissioning scope. The on-site dielectric test, UHF PD with sensitivity verification, and the gas quality regime must be specified explicitly, by reference to IEEE C37.122.7 or IEC 62271-203 Annex C and the CIGRE brochures.
Specify integral UHF sensors at design stage. Retrofitting external sensors after the fact costs sensitivity and coverage, and forecloses permanent monitoring later.
A PD result without a documented sensitivity verification is not a result. Require the verification record as a deliverable, per sensor position.

Case Study C — The Compliance Date Nobody Had Mapped

Situation. An owner with a multi-site portfolio was standardizing on a GIS product family for a programme of substation builds and replacements running several years into the future. Procurement was well advanced. Our scope was a technical review of the standardization decision.

What the review found. The technical selection was sound. The programme timeline was not compatible with the regulatory timeline in two of the jurisdictions involved, and nobody had mapped one against the other.

Two sites fell in a jurisdiction with an acquisition prohibition on new SF₆ gas-insulated equipment at the relevant voltage class, with a date inside the programme window. The equipment could be ordered before the date under the applicable transitional provisions, but the analysis of whether each specific site's schedule would clear the trigger had not been done — and the trigger is defined by acquisition, not by energization.
The interrupting-current rating mattered. The phase-out tables in the governing regulation are keyed to voltage and short-circuit rating, and a modest increase in the specified interrupting capability — which the fault study was likely to require at one site — moved that site from one phase-out date to another several years apart.
An exemption pathway existed, based on the unavailability of non-SF₆ equipment of the required type and rating from at least two suppliers — which at the voltage in question was a genuinely arguable position. But exemption requests take time, have their own deadlines, and permit acquisition only within a defined window after approval. Nobody had started.
The alternative-technology assessment had not distinguished between chemistries. The standardization study treated "SF₆-free" as one option. It is not: clean air plus vacuum is commercially mature only to a certain voltage, while above it the available alternatives are fluorinated mixtures that carry their own separate regulatory exposure in other jurisdictions and an unresolved feedstock supply question.

Outcome and lessons. The programme was resequenced so that the affected sites were procured on the compliant technology, with an exemption request prepared for the one site where no two-supplier alternative existed at the required rating.

Four transferable lessons:

Map the regulatory calendar against the procurement calendar at programme inception, per site, per voltage class, per interrupting rating. The phase-out tables are keyed to more than voltage, and a fault-study result can move a site across a boundary.
Understand what triggers the prohibition — acquisition, placing on the market, or energization. The definitions differ between jurisdictions and they determine which side of a date a given purchase order falls on.
"SF₆-free" is not a single option. Vacuum plus clean air and fluorinated mixtures have different voltage ceilings, different physical properties including minimum operating temperature, and radically different regulatory outlooks. Assess them separately.
If your equipment has a 40-year design life, ask the feedstock question. For any fluorinated insulating medium, the technical clarifications should cover long-term gas supply, top-up availability, and end-of-life recovery commitments — in the contract, not in the brochure.

Part 10 — A GIS Specification Checklist

Items that are routinely omitted and expensive to add later:

Architecture and future-proofing

Single-phase versus three-phase enclosure, stated with the rationale
Service continuity requirement stated using the CIGRE MRE code
Spare bay positions with capped busbar extensions and buffer compartments
Gas compartmentalization philosophy, driven by target MTTR rather than first cost
Contractual spares, interface-drawing release, and obsolescence notification commitments

Equipment

Circuit breaker class designations spelled out — E, M, C and S
Earthing switch making-capacity class (E0/E1/E2) per location, with rationale
Conventional instrument transformers versus LPITs, decided at concept stage with the protection architecture
Type-tested leakage rate, stated as a requirement — 0.1% per year per compartment, not 0.5%
Minimum operating temperature stated explicitly, especially for any alternative-gas design

Transients and grounding

VFTO study required, particularly for any direct transformer-to-GIS connection
Disconnector trapped-charge-voltag e characteristics specified, or damping resistors required
Enclosure multipoint bonding design, with touch potential analysis inside the building
Secondary cabling routing, shielding and segregation requirements

Commissioning

On-site dielectric withstand per IEEE C37.122.7 or IEC 62271-203 Annex C, with the conditioning sequence explicitly required
UHF PD measurement with integral sensors, and CIGRE sensitivity verification records as a deliverable, per sensor position
Gas quality measurement at filling and again at least five days after final filling
Exhaustive interlock verification, every combination, recorded
Contact resistance on all main circuits, against the engineered value

Safety and operations

Oxygen depletion monitoring and low-level forced ventilation, sized to the largest single compartment charge
Internal-arc emergency response plan written and trained before energization
Switching, clearance and tagging procedures reviewed for the absence of a visible break
SF₆ (or alternative gas) inventory tracking aligned to the applicable reporting rule

Conclusion

Gas-insulated switchgear is a genuinely excellent technology, and the case for it is strong on exactly the grounds the overview claims: it is compact, it is reliable, it needs little maintenance, and it performs where air-insulated equipment cannot. On a constrained urban site, a refinery, an offshore platform or a data-center campus, it is frequently the only workable answer.

But the one-page version leaves out the four things that determine whether a GIS project goes well.

The safety model is different. There is no visible break. Assurance comes from a type-tested isolating distance, a mechanically linked position indicator and an interlock scheme — which means procedures, training and commissioning rigour carry weight they do not carry in an AIS yard.

The failure physics is different. SF₆'s sensitivity to field enhancement makes a millimetre of loose metal a system-level threat, which is why the commissioning regime is built around conditioning and partial discharge rather than around component testing — and why "in accordance with ANSI/NETA ATS" does not procure it.

The transients are different. Disconnector operation generates nanosecond-front overvoltages that stress directly connected transformer windings, couple onto the enclosure, and disrupt secondary systems. None of that has an AIS analogue.

And the clock is different. SF₆ is the most potent greenhouse gas assessed, already prohibited in new European medium-voltage switchgear and prohibited at transmission voltage from 2028 and 2032 on a GWP ≥ 1 threshold that catches the alternatives too. California and New York have their own schedules. Below 145 kV a zero-GWP answer exists today. Above it, every available choice carries a regulatory question that has not yet been answered.

For an owner, the practical conclusion is that the GIS decision is not a technology preference. It is a commitment to a specific safety procedure regime, a specific commissioning scope, a specific spares and extension strategy, and a specific position on a moving regulatory landscape — all of which are cheapest to get right before the specification is issued.

How Keentel Engineering Can Help

GIS and AIS substation design — technology selection studies, physical and electrical design, general arrangement development, and derived footprint and cost comparison
Owner's engineer services — specification development, technical bid evaluation, factory acceptance test witness, construction-phase QA and turnover package audit
Power system studies — short circuit, coordination, insulation coordination, VFTO analysis for transformer-to-GIS connections, arc flash, and grounding grid design including touch and step potential inside GIS buildings
Commissioning oversight — GIS commissioning scope development, on-site dielectric and UHF PD test witness, sensitivity verification review, and independent review of commissioning packages
Interconnection and POI engineering — including large-load and data-center interconnection
NERC compliance support — O&P 693 compliance and protection system maintenance program structure
Specification currency review — auditing existing substation and testing specifications against current standard editions

The highest-value engagement point is before the technology selection is frozen, while the architecture, service-continuity requirement, commissioning scope and gas strategy are all still open. Contact Keentel Engineering to discuss your project.

References and Further Reading

IEEE

IEEE C37.122-2021, Standard for High-Voltage Gas-Insulated Substations Rated Above 52 kV — https://standards.ieee.org/ieee/C37.122/7281/
IEEE C37.122.7-2021, Guide for Field Testing of Gas-Insulated Substations Rated Above 52 kV — https://standards.ieee.org/ieee/C37.122.7/7399/
IEEE C37.122.6-2013, Interface of New Gas-Insulated Equipment in Existing Gas-Insulated Substations — https://standards.ieee.org/ieee/C37.122.6/11330/
IEEE C37.122.3-2024, Guide for SF₆ Gas Handling for High-Voltage Equipment — https://standards.ieee.org/ieee/C37.122.3/4428/
IEEE C37.122.5-2013, Moisture Measurement and Control in SF₆ Gas-Insulated Equipment — https://standards.ieee.org/ieee/C37.122.5/11329/
IEEE Std 80, Guide for Safety in AC Substation Grounding (Inactive-Reserved; P80 revision active) — https://standards.ieee.org/ieee/80/4089/
IEEE Std 693-2018, Recommended Practice for Seismic Design of Substations — https://standards.ieee.org/ieee/693/4996/
IEEE PES, SF₆ Gas Alternatives (WG K19 / PC37.122.10 status) — https://ieee-pes.org/trending-tech/sf6-gas-alternatives/

IEC

IEC 62271-203:2022 Ed. 3.0, AC gas-insulated metal-enclosed switchgear above 52 kV — https://webstore.iec.ch/en/publication/65853
IEC 62271-1:2017+AMD1:2021, Common specifications — https://webstore.iec.ch/
IEC 62271-4:2022 Ed. 2.0, Handling procedures for gases for insulation and/or switching — https://webstore.iec.ch/en/publication/64701
IEC 62271-100:2021 (+AMD1:2024), AC circuit-breakers — https://webstore.iec.ch/en/publication/62785
IEC 62271-102:2018 (+AMD1:2022), AC disconnectors and earthing switches — https://webstore.iec.ch/
IEC 62271-200:2021, AC metal-enclosed switchgear above 1 kV to 52 kV — https://webstore.iec.ch/en/publication/63466
IEC 62271-204:2022, Rigid HV gas-insulated transmission lines — https://webstore.iec.ch/en/publication/65588
IEC 62271-207:2023, Seismic qualification for gas-insulated switchgear assemblies — https://webstore.iec.ch/en/publication/68303
IEC 60376:2018, Technical grade SF₆ — https://webstore.iec.ch/en/publication/33028
IEC 60480:2019, Re-use of SF₆ and its mixtures — https://webstore.iec.ch/
IEC 61936-1:2021, Power installations exceeding 1 kV AC — https://webstore.iec.ch/en/publication/64490

CIGRE

CIGRE TB 654 (WG D1.25, 2016), UHF partial discharge detection system for GIS: Application guide for sensitivity verification — https://www.e-cigre.org/publications/detail/654-uhf-partial-discharge-detection-system-for-gis-application-guide-for-sensitivity-verification.html
CIGRE TB 933 (WG D1.66, 2024), Requirements and Application of UHF PD Monitoring Systems for Gas Insulated Systems — https://electra.cigre.org/334-june-2024/technical-brochures/requirements-and-application-of-uhf-pd-monitoring-systems-for-gas-insulated-systems.html
CIGRE TB 870 (WG B3.51, 2022), Service Continuity Guide for HV GIS above 52 kV — https://electra.cigre.org/322-june-2022/technical-brochures/870-service-continuity-guide-for-hv-gis-above-52-kv.html
CIGRE TB 802 (WG B3.45, 2020), Application of non-SF₆ gases or gas-mixtures in medium and high voltage gas-insulated switchgear — https://www.e-cigre.org/publications/detail/802-application-of-non-sf6-gases-or-gas-mixtures-in-medium-and-high-voltage-gas-insulated-switchgear.html
CIGRE, Fourth international reliability survey on switching equipment — https://cigre.org.uk/web-cont1001/uploads/CIGRE-fourth-reliability-survey-on-switching-equipment.pdf
CIGRE Science & Engineering N°40, The Impacts of the F-gas Regulation (EU) 2024/573 on High Voltage Gas Insulated Switchgear Operators — https://cse.cigre.org/cse-n040/

Regulation

US EPA, 40 CFR Part 98 Subpart DD, Electrical Transmission and Distribution Equipment Use — https://www.ecfr.gov/current/title-40/chapter-I/subchapter-C/part-98/subpart-DD
US EPA, Byproducts of Sulfur Hexafluoride (SF₆) Use in the Electric Power Industry — https://www.epa.gov/system/files/documents/2022-05/sf6_byproducts.pdf
US EPA, SF₆ Emission Reduction Partnership (archived) and state regulation roundup — https://www.epa.gov/eps-partnership
CARB, Regulation for Reducing Sulfur Hexafluoride Emissions from Gas Insulated Switchgear , 17 CCR §95350 et seq. — https://ww2.arb.ca.gov/rulemaking/2020/sf6
NYSDEC, 6 NYCRR Part 495, Sulfur Hexafluoride Standards and Reporting — https://dec.ny.gov/sites/default/files/2024-12/part495apc.pdf
MassDEP, 310 CMR 7.72 — https://www.mass.gov/info-details/reducing-sulfur-hexafluoride-sf6-emissions-from-gas-insulated-switchgear-310-cmr-772
Regulation (EU) 2024/573 on fluorinated greenhouse gases — https://eur-lex.europa.eu/

Technical literature

Riechert, Very Fast Transients in GIS (IEEE Switchgear Committee) — https://ewh.ieee.org/soc/pes/switchgear/presentations/tp_files/2012-2_Thu_e_Riechert.pdf
Yao et al., Insulation Defect Partial Discharge Characteristics in GIS , Energies 11(4):971 (2018) — https://www.mdpi.com/1996-1073/11/4/971
Billen, Pernigotto et al., Life-cycle assessment of SF₆ alternatives in HV switchgear , Energies 13(7):1807 (2020) — https://www.mdpi.com/1996-1073/13/7/1807
An et al., Global SF₆ emissions , Atmospheric Chemistry and Physics 24, 12465 (2024) — https://acp.copernicus.org/articles/24/12465/2024/
SINTEF, What is the status of phasing out SF₆ gas in switchgear and circuit breakers? (December 2025) — https://blog.sintef.com/energy/what-is-the-status-of-phasing-out-sf6-gas-in-switchgear-and-circuit-breakers/

Part 11 — Frequently Asked Questions

Cable Testing Before Energization

Sun, 16 Aug 2026 16:03:47 GMT

Aug 16, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Cable Testing Before Energization: The Twelve-Step Sequence, Rewritten as Engineering

A Keentel Engineering Grid IQ technical guide for commissioning engineers, owner's engineers, EPC quality managers, and utility cable asset owners

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

There is a widely circulated infographic — you have almost certainly seen a version of it — laying out the cable testing sequence for HV, MV and AC cables before energization. Twelve numbered steps: visual inspection, continuity, insulation resistance, AC withstand, DC withstand where applicable, sheath integrity, phase identification, polarity, earth continuity, megger trending, final documentation, system readiness. The banner underneath reads: "Don't assume it's OK — test it before you energize."

That banner is correct and the twelve items are all real tests. As a reminder of scope, the list is useful.

As a sequence , it has three problems that cost real money:

The sheath test is in the wrong place. It sits at step 6, after the withstand tests. In practice the sheath test is the only after-installation test that finds pulling damage, rock in the trench bed, and tool strikes during joint-bay work — and it needs to happen before backfill, not after. Finding a jacket puncture at step 6 on a buried circuit means re-excavating.

Steps 4 and 5 are alternatives, not a sequence. You perform an AC withstand or a DC withstand, chosen by insulation type and cable age. Running both on the same extruded circuit is not thoroughness — a DC hipot on service-aged XLPE can create the defect it was meant to find.

The step that would actually catch the dominant defect is missing. Roughly three-quarters of failures on modern HV extruded cable systems occur in the accessories — the joints and terminations that are hand-built on site — and a simple pass/fail withstand test gives you no information about them unless they break during the test. Partial discharge measurement, or a monitored withstand, is what turns "this circuit survived" into "joint bay 4 has a defect."

This guide takes the twelve steps and rewrites them as engineering: the governing standard for each, the actual acceptance criterion, what the published failure data says about which tests are worth the money, and a corrected sequence.

Introduction: The Sequence Is Right. The Order and the Criteria Are Where Projects Fail.

Part 2 — Frequently Asked Questions: Large Load Interconnection

1.1 Withstand versus diagnostic — the distinction that determines everything

IEEE Std 400-2023, Guide for Field Testing and Evaluation of the Insulation of Shielded Power Cable Systems Rated 5 kV and Above, is the umbrella document for this whole subject. It catalogues the available field test methods, gives advantages and disadvantages for each, and — importantly — deliberately does not carry the detailed acceptance criteria, which live in the "point" documents: 400.1 (HVDC), 400.2 (VLF), 400.3 (partial discharge), 400.4 (damped AC).

The taxonomy it uses is the one that should shape your test specification:

Withstand test — a go/no-go proof. The insulation either survives the applied voltage for the applied duration, or it does not. A pass tells you the circuit had no defect severe enough to break down under that specific stress. It tells you nothing about condition.
Simple withstand — voltage applied, pass or fail, nothing recorded.
Monitored withstand — voltage applied for the withstand duration while a diagnostic quantity (tan delta, partial discharge) is recorded. Same energization, same outage, pass/fail plus condition data.
Diagnostic test — measures a property of the insulation (tan delta, PD, dielectric response) to estimate condition. No inherent pass/fail; interpretation is against criteria, baselines and trends.

If your commissioning specification calls for a simple withstand and nothing else, you have specified a test that answers one question — will it break today? — and declines to answer the more valuable one: where is the weak accessory?

The single most useful upgrade available to most cable commissioning specifications is to convert the withstand test to a monitored withstand. It costs one additional instrument connection and no additional outage time.

1.2 Where cables actually fail

Test effort should follow the failure data, and the failure data is unusually clear.

HV and EHV extruded systems — the accessories dominate. The EPRI/NEETRAC/DOE Cable Diagnostic Focused Initiative (CDFI), examining North American extruded cable systems installed since 2000, found accessories accounting for roughly 68–75% of failures, with the report's own summary noting "the propensity of the failure to occur in the accessories (3 to 4 times more frequently than the cable)." It also found that for those installations "the failure rate in the first three years of life is slightly higher than the failure rate after three years" — classic infant mortality, which is precisely the signature of installation defects rather than aging.

CIGRE's international survey confirms the direction. Technical Brochure 815 (WG B1.57, 2020), covering 2006–2015 across roughly 29,350 circuit-km of AC land cable, recorded 744 faults on land cable systems, split cable 56%, joints 16%, terminations 18%, other 9% — so accessories account for 34% of all faults, internal and external combined. Since external damage overwhelmingly strikes the cable body, the accessory share of internal failures is considerably higher. The internal-cause rates make this explicit for 60–500 kV XLPE AC land cable:

Part 2 — The Twelve Steps, Rewritten

(Capacitance values are typical; state the assumed value in any project calculation.)

A 72 MVA, 276 kV power-frequency test transformer is not a transportable object. That is why nobody 60-Hz-hipots EHV cable, and why a conventional 50/60 Hz set can test roughly 50 feet of 15 kV cable where a VLF set handles a couple of miles.

Series resonant test systems solve this. A variable-frequency converter feeds an exciter transformer and a fixed HV reactor in series with the cable capacitance. The converter sweeps frequency until X_L = X_C, at f_res = 1/(2π√(LC)). At resonance the reactive power circulates between reactor and cable, and the supply only makes up the losses — so the required source rating is roughly Q_test divided by the circuit quality factor, typically 50–100 for modern reactors. The 72 MVA EHV case becomes a feed of roughly 0.7–1.4 MW, and a transportable one.

The accepted band is 20–300 Hz, which appears explicitly in IEC 60502-2's after-installation AC option and in ANSI/NETA ATS-2025 Table 100.6.5, "Shielded Power Cables, AC (20 Hz–300 Hz) Test Voltages" — near-power-frequency resonant AC is now a first-class tabulated acceptance method in a US standard, not an exotic alternative.

After-installation AC test values. The IEC voltage hierarchy is:

The equivalent "rest of world" table for the same insulation sets Action Required at tip-up above 1.0 and mean above 2.0 — figures roughly fifty times tig hter. If a client specification or a vendor report shows numbers around 1–2 ×10⁻³ as the boundary for "good XLPE," it is on the other table. Applying it to US service-aged cable will condemn a great deal of healthy plant.

Two further cautions:

Do not apply XLPE criteria to EPR. New EPR typically sits in the 8–10 ×10⁻³ range at U₀, against under 1 ×10⁻³ for new XLPE. The XLPE "Action Required" threshold of 50 would be comfortably passed by healthy EPR, but the "No Action Required" threshold of 4 would condemn every EPR cable ever made. ANSI/NETA ATS-2025 Table 100.6.7.2 is now the accessible published source for EPR evaluation criteria in US practice, alongside Table 100.6.7.1 for PE-based insulations.

Do not apply aged-cable criteria to a new cable. IEEE 400.2 its elf carries the caveat that for newly installed cable "insufficient data have been collected to make precise estimates of criteria, consequently the criteria are likely to contain considerable errors." Tan δ criteria are calibrated on service-aged populations. On a commissioning test, the more defensible use of tan δ is as a baseline and as a comparison between phases of the same circuit — three phases installed together by the same crew from the same reel should read alike, and the one that does not is the interesting one.

Tan delta is the most useful single diagnostic in MV cable commissioning and the easiest to apply wrongly. Three parameters, all expressed in units of 10⁻³:

Mean VLF-TD — mean tan δ at U₀; the absolute loss level
VLF-TDTS — temporal stability; the standard deviation of readings over time at fixed voltage (the 2024 edition recommends computing at both 1.0 U₀ and 0.5 U₀)
VLF-DTD — differential tan delta, or tip-up: TD at 1.5 U₀ minus TD at 0.5 U₀

The trap: IEEE 400.2 does not publish one table. It publishes a family, split by insulation type and by geography — and the numbers differ by more than an order of magnitude between them. US service-aged XLPE has historically run far higher tan δ than European and Asian dry-cured XLPE, so the criteria diverge.

For PE-based insulations (PE, XLPE, TR-XLPE) under the criteria applicable to US practice:

Plus Appendix B, "Guidance for Circuit Reliability Considerations for Medium and High Voltage Cable Methods of Test."

Four things that structure tells you abo ut where the industry has moved:

* Cable acceptance is now organized by insulation type and test method *, rather than one DC table for everything.
Damped AC and 20–300 Hz resonant AC are now first-class tabulated methods in a US acceptance standard.
NETA now publishes its own tan delta evaluation criteria, so a US commissioning specification no longer has to reach into IEEE 400.2 for thresholds — and, usefully, NETA is the accessible source for the EPR criteria.
Appendix B is a philosophical shift. NETA is telling the industry to select the test method by circuit reliability requirement — how critical is this feeder, what does an outage cost, what is the consequence of a breakdown on test — rather than by habit. That is exactly the conversation an owner's engineer should be having at specification stage.

(The numeric values in all of these tables are copyrighted and changed in 2025. Do not reuse values from an older NETA reproduction; obtain the current edition.)

IEEE 400.2-2024 also moved, with reported changes including extended voltage coverage beyond the 2013 edition's 69 kV ceiling, restriction of monitored-withstand parameters to tan delta and partial discharge (dropping leakage current), tan delta stability recommended at both 1.0 U₀ and 0.5 U₀, a 60-minute minimum acceptance duration for cables at 66 kV and above, and a requirement to ramp voltage down slowly rather than switching off. Those changes are described in vendor summaries rather than confirmed against the standard text; read them from the document before writing them into a specification.

For reference, the widely reproduced VLF withstand table from the 2013 edition — sinusoidal waveform, phase to ground, kV rms / kV peak — remains the most commonly cited version in the field:

Maintenance runs at roughly 75% of acceptance. Verify against the 2024 edition before use — the tables were reorganized by cable age and insulation type.

Part 1 — What the Sequence Is Actually For

Cable testing checklists circulate globally, and many originate on projects governed by an owner standard rather than by IEC or IEEE directly. Understanding that hierarchy matters, because it changes what "compliance" means.

Take the Saudi Aramco stack as a representative example, since a great deal of Gulf-origin cable testing material derives from it. Aramco maintains SAES-P-xxx engineering standards (mandatory design and installation requirements), 15-SAMSS-xxx materials specifications, SAIC-P-xxxx inspection checklists that an inspector physically signs, and SATIP typical inspection plans with SATR test records. The power cable document is SAES-P-104, Wiring Methods and Materials.

Its cable testing content is instructive because it is far more prescriptive than the IEC and IEEE documents it references. Where IEC 60502-2 offers four alternative after-installation tests and lets the parties choose, the owner standard names one and gives the number. Where IEEE 400 catalogues methods and declines to set acceptance criteria, the inspection checklist gives the inspector a signable pass/fail. Notably, it adopts the IEC 60229 after-installation rule verbatim for direct-buried cable at 5 kV and above — DC hipot between insulation shield and ground at 4 kV per mm for one minute, not exceeding 10 kV.

Its treatment of DC hipot is also more nuanced than it is often represented. DC withstand values are specified for new cable only; routine DC hipot is described as not recommended; cables in service more than five years are exempt; and VLF is the sanctioned method for condition assessment on aged cable. That is entirely consistent with the space-charge science in Step 5 — and citing it as "the owner requires DC hipot" without the qualifiers would misrepresent it.

Two cautions. The publicly circulating copy of that standard is a 2010 revision; owner standards are controlled documents on roughly five-year cycles, and current revisions are not public. Obtain the current revision through the owner or the EPC rather than relying on a circulating PDF. And the same pattern holds worldwide — Gulf national oil companies, European TSOs, and US utility construction "blue books" all do the same thing. The IEC and IEEE documents remain the technical basis. The owner standard removes the optionality. Know which one your contract incorporates, and check that the two do not conflict before the first test.

Part 8 — A Corrected SequenceNew Paragraph

And at 1.7 U₀, approximatel y 92% of PD sites were detected within 60 minutes — which is why CDFI recommends at least 30 minutes at 1.7 U₀, with 60 minutes preferred to capture the remaining 8%.

Read those two findings together and the point is stark: at operating voltage, PD is detectable in fewer than 2% of cases. At 1.7 U₀, it is detectable in more than half. A soak test at U₀ — the fallback option in several specifications — will find almost nothing.

3.4 The defect mechanisms PD actually finds

All of them are workmanship, and all of them are invisible once the joint is closed: semiconductive layer not fully removed or over-cut into the insulation; contaminated or incorrectly lubricated interfaces; incorrect stress cone or deflector positioning; insufficient interface pressure; voids in field-poured resin; moisture in a joint shell; wrong assembly dimensions; missing or misapplied field-distribution elements.

The consequence for test design: the method must be able to localize, not merely detect. A single-ended time-domain measurement with reflection mapping, or a double-ended or distributed measurement, is what turns "this circuit has PD" into "excavate joint bay 4." IEEE 400.3-2022 gives guidance on when each is appropriate.

This is the part most cable-testing content avoids, and it is the part an owner should care most about. The Cable Diagnostic Focused Initiative — an EPRI, NEETRAC and US DOE program under DOE award DE-FC02-04CH11237, final Phase 1 report December 2010, with Phase II chapters following — remains the largest independent evaluation of cable diagnostics ever published, covering roughly 10,824 conductor miles across all techniques.

4.1 The headline conclusions, in CDFI's own words

"Diagnostic tests can work. They often show many useful things about the condition of a cable circuit, but not everything desired."

"Most diagnostic technologies examined do a good job of accurately establishing that a cable circuit is 'good'. They are not as good at establishing which circuits are 'bad'."

"No one diagnostic is likely to provide sufficient information to accurately establish the condition of a cable circuit."

"Diagnostics are generally unable to determine definitively the longevity of the circuit under test."

That second quote is the one to internalize. A cable diagnostic is a good negative test and a mediocre positive one. Passing tells you a great deal. Failing tells you less than the marketing suggests. This is why CDFI advocates combining features rather than relying on a single method, and why a monitored withstand — which gives you a withstand resu lt and diagnostic data from one energization — is such good value.

4.2 Simple withstand: how well does it actually perform?

CDFI Chapter 9 evaluated 7,875 miles tested between 2001 and 2008 across XLPE, paper and EPR:

Accuracy: raw median 93.0%, weighted median 87% of tested cable did not fail within two years of the test.
Failure-on-test rates at IEEE 400.2 recommended voltages: roughly 2.0% at 15 minutes, 2.7% at 30 minutes, 3.7% at 60 minutes.

T hat second row is the argument for duration, and CDFI states the consequence directly: "Ten failures representing more than 230 conductor miles would have gone undetected if the test had been terminated at 15 min." Nearly half of the defects a 60-minute test finds are invisible at 15 minutes. If your specification says 15 minutes because that is what fits the outage window, you are buying roughly half the test.

There is also a finding that reframes what "failing on test" means. CDFI found the likelihood of subsequent failure notably lower for sections that failed in a controlled manner during the test and were then repaired than for sections that passed. That is not a paradox — it is selection. A test failure identifies and removes a defect; a test pas s leaves whatever was below threshold in the ground. (CDFI's own chapter contains internally inconsistent figures on the magnitude of this effect, drawn from different datasets with small samples, so treat the direction as robust and the magnitude as unsettled.)

4.3 Does a diagnostic program pay?

CDFI's benefits chapter gives two field cases:

Case A — utility VLF withstand program, four years : 9 outages with diagnostics versus 19 without, a 53% reduction. Note the nuance, because it is the honest part: total failures were higher with diagnostics (32 versus 19), because defects were being caught on test rather than in service. The program did not reduce the number of defects; it changed where and when they surfaced — from an unplanned outage to a controlled t est.
Case B — service-provider PD program, ten years : 13 versus 19 outages, a 32% reduction.
Economic outcome for Case A: roughly $350,000 benefit, four to five times the cost of the program. For HV/EHV commissioning testing, CDFI puts the benefit between $1.3 and $2.6 million over a fifteen-year horizon.

And a conclusion that belongs in every business case: "Intangible costs are decisive. Without them Diagnostic Programs are not beneficial." If your economic model counts only crew time and replacement cable, and not the cost of an unplanned outage to the customer, the program will not justify itself on paper — regardless of whether it is worth doing.

4.4 How often does a new circuit fail its commissioning test?

Published figures exist but need careful handling. CIGRE's Science & Engineering journal reports that "12% of all on-site acceptance tests on newly installed circuits resulted in breakdowns, mostly in cable joints and terminations" — attributed within that paper to a further reference, which is worth chasing before you rely on it. The same source reports that 40% of surveyed practitioners had observed an insulation breakdown during testing, and that in more than 70% of those cases PD was detected before breakdown — which is, again, the argument for monitoring during the withstand rather than simply applying it.

CIGRE TB 841 compares non-pass rates for terminations and joints between near-power-frequency and damped AC commissioning tests, drawing on a test-experience database spanning 1997 to end-2016. The actual percentages are in the brochure. If commissioning failure rates matter to your business case, that is the document to buy.

A directional figure worth stating carefully: a large commerc ial PD-testing database covering more than 250,000 field tests across 17 countries reports roughly 37% of systems at 5–500 kV having at least one substandard component, rising to about 45% for HV and EHV, with terminations defective in about 14% of HV/EHV cases. Those are vendor-published figures from a firm selling the test, and the headline before/after comparisons in that dataset compare non-randomized populations of very different sizes — but the component-level defect percentages are consistent with CDFI and CIGRE, and are worth knowing.

The most consequential US change is the restructuring of NETA's cable acceptance tables. ATS-2025 expanded Table 100.6 into a family of eight sub-tables:

Cable testing standards moved more in the last three years than in the decade before, and a great many specifications in circulation have not caught up.

If the failure data says accessories cause three-quarters of HV failures, and a simple withstand test cannot see an accessory defect unless it breaks, then the sequence needs a test that can. That test is partial discharge measurement — and the data on it is unusually good.

3.1 What PD is measured against

IEEE Std 400.3-2022 governs field PD diagnostic testing of installed shielded cable systems — cable, joints and terminations, online and offline. Note the retitle from the 2006 edition ("…in a Field Environment" → "Field Diagnostic Testing"): it signals the shift from PD as a pass/fail gate to PD as a diagnostic and localization tool. IEC 60270 defines apparent charge in picocoulombs and the calibration procedure that makes a pC number mean anything. IEC 60885-3 covers the factory method on extruded cable lengths.

3.2 Why there is rarely a universal on-site pC limit

This is worth understanding, because specifications that demand one are unenforceable:

Calibration breaks down on an installed system. IEC 60270 calibration injects a known charge at the terminals. On an installed circuit the PD source may be a kilometre away, and attenuation and dispersion of the PD pulse along the cable mean the charge measured at the terminal is not the charge at the defect — with a length- and frequency-dependent relationship between them. A 50 pC reading from a termination and a 50 pC reading from a joint 2 km out are completely different physical defects.

Online measurement cannot legitimately report pC at all, because the coupling path is uncalibrated. Online PD is inherently comparative and trended, not absolute.

Site noise floor is variable and often dominant — achievable sensitivity is set by the substation environment, not by the instrument.

And the meaningful quantity is inception voltage relative to service stress, not magnitude. A discharge that initiates at 1.7–2.0 U₀ and extinguishes at 1.5 U₀ is a different risk from a small discharge that never extinguishes above 1.0 U₀ — the second one is active every day in service.

The factory number people mistakenly carry into the field is IEC 60502-2's routine test: raise to 2 U₀, hold 10 seconds, reduce to 1.73 U₀, with no detectable discharge above a declared sensitivity of 10 pC for routine tests or 5 pC for type tests. That is a number obtained on a drum, in a screened room, on a calibrated circuit. It does not transfer.

A commissioning specification that is actually enforceable requires: (a) a documented sensitivity and noise-floor check at the start of the test; (b) no detectable PD above the demonstrated sensitivity at and below a stated multiple of U₀; (c) PD inception voltage at or above a stated multiple of U₀; (d) localization of an y detected source to a specific accessory; and (e) raw data archived as the baseline for future trending. "≤5 pC" is not enforceable. That list is.

3.3 The data on when PD appears and how long to hold

CDFI's HV and EHV work measured the distribution of PD inception voltages and onset times directly, and the numbers make the case for the test duration on their own:

Part 9 — Three Anonymized Case Studies

Confidentiality note

The three engagements below are presented in anonymized and generalized form. No client, utility, location, vendor, contractor or date is identified, and voltages, lengths, sequences and findings have been altered or aggregated. They are included to illustrate recurring en gineering and process patterns, not to characterize any single project.

Case Study A — The Circuit That Passed, Twice

Situation

A newly installed medium-voltage feeder of moderate length, extruded insulation, with several joints and two sets of terminations, was commissioned with a VLF withstand test. The test was passed. The circuit was energized. It failed in service in the first year, at a joint. The joint was repaired, the circuit was re-tested with the same VLF withstand protocol, passed again, and was returned to service. It failed a second time, at a different joint, within the following year.

The owner engaged us to review the commissioning approach before repeating the cycle.

What the review found

Nothing in the test record was wrong. The voltage was correct for the cable class, the equipment was calibrated, the technician was qualified, and the results were properly documented. The problem was that the test performed had never been capable of answering the question the owner cared about.

Three specific findings:

The test was a simple withstand. No tan delta, no partial discharge, no monitoring of any kind. It could only report survival, and it had reported survival accurately both times.
The duration was set to fit the outage window, at the shorter end of the standard's range. The published evidence is that shortening a withstand from 60 minutes to 15 substantially reduces the defects it detects — roughly half of the defects found at 60 minutes are invisible at 15.
The circuit had no baseline of any kind. No tan delta, no TDR trace, no PD data. After two failures the owner still had nothing against which to compare a third test, and no way to distinguish "this joint is marginal" from "this joint is fine."

Outcome and lessons

The circuit was re-tested as a monitored withstand — the same energization, the same outage duration extended, with tan delta and partial discharge recorded throughout. PD was detected and localized to a specific accessory that had passed both previous simple withstand tests. That accessory was rebuilt and the circuit retested clean.

Three transferable lessons:

A withstand pass is a weak positive result. The published evaluation of these methods is explicit that diagnostics are good at confirming a circuit is sound and considerably less good at identifying which are not. Passing a withstand means no defect was severe enough to break down under that particular stress for that particular duration. It does not mean the accessories are good.
Test duration is not an outage-planning variable. It is a detection-probability variable. If the outage window will not accommodate the specified duration, the correct response is to change the outage window, not the test.
The marginal cost of converting a simple withstand to a monitored withstand is close to zero — one additional instrument connection, no addi tional outage. The marginal information is the difference between "it survived" and "rebuild the joint in bay 4."

Case Study B — The Bonding Scheme That Was Never Energized as Designed

Situation

A high-voltage single-core circuit was installed with a special bonding arrangement, with link boxes at the section transitions. The circuit was commissioned, passed its insulation tests, and entered service. Over the following operating seasons the owner observed conductor temperatures consistently higher than the ampacity study predicted at equivalent loading, with no obvious explanation. The circuit was derated administratively while the cause was investigated. We were engaged to review the installation and commissioning records against the design.

What the review found

The circuit was carrying circulating sheath current that the design had been specifically arranged to eliminate. Two independent causes, either of which alone would have produced the effect:

A jacket defect on one minor section. The oversheath had been damaged — most probably during backfill, given its position relative to a duct crossing — creating a sheath-to-earth path. On a specially bonded section, that quietly converts the section to solidly bonded. Nothing trips, nothing alarms, and the only symptom is lost ampacity and elevated temperature. The sheath test had been performed after backfill, and had been performed at a flat voltage taken from a generic specification rather than the thickness-proportional rule — under-testing the jacket relative to the applicable criterion.
A link box configured contrary to the bonding drawing. At one transition, the link arrangement did not match the design. The as-built documentation recorded the box as complete; nobody had audited the internal configuration against the drawing, because the commissioning checklist item read "link boxes installed and secured."

Additionally, the SVLs had not been disconnected during the sheath test — so even a correctly executed sheath test at the correct voltage would have been partially masked by SVL conduction.

Outcome and lessons

Sheath fault pre-location by bridge, pin-pointing by step voltage method, a single excavation, jacket repair, link box reconfiguration, and re-test. The circuit's ampacity was restored to design.

Four transferable lessons:

The sheath test is the only commissioning test that proves the bonding scheme is electrically what the drawings say. No conductor-side test will ever reveal a sheath-to-earth fault on a specially bonded circuit.
Test the sheath before backfill. The defect was almost certainly created by backfill; testing before it would have found nothing, but testing before and after brackets the damage and localizes the cause.
The sheath test voltage is thickness-proportional, not a flat number. Four kV per millimetre capped at 10 kV — a generic "5 kV" or "10 kV" lifted from another project can under-test or over-test depending on the jacket.
"Link boxes installed" is not a commissioning check. The inte rnal link configuration must be audited against the bonding drawing, link by link, and photographed before closing. It is nearly impossible to verify later without dismantling.

Case Study C — The Specification That Cited a Reserved Standard

Situation

An owner's cable testing specification, several years old and used across a portfolio of projects, was applied to a new installation. During execution a dispute arose between the contractor and the owner's site representative about which test was required, at what voltage, for how long. Both parties were reading the same specification and reaching different conclusions. We were asked to review the specification and resolve the ambiguity.

What the review found

The specification was not badly written. It was out of date, in four separate and compounding ways:

It cited superseded editions throughout. Several of the IEEE cable testing documents it named by year had been revised, and one had been moved to Inactive-Reserved status entirely — meaning the issuing body no longer maintains it. The contractor's proposed method complied with the current edition; the site representative's expectation came from the superseded one. Both were defensible readings of an ambiguous instruction.
It cited a DC withstand as the default method for extruded cable, without the age qualification that the underlying technical basis requires. The contractor, correctly, objected to DC-hipotting a circuit that included a section spliced to existing service-aged cable.
It specified acceptance criteria for tan delta without stating the insulation type they applied to. The values were the PE-based figures. Part of the installation was EPR, where those criteria would have condemned sound cable.
It specified a withstand duration at the short end of the range, apparently inherited from an earlier outage-constrained project and never revisited.

None of these were errors of engineering judgment when the specification was written. All four became errors through the passage of time.

Outcome and lessons

The specification was reissued with current editions, a method selection table keyed to insulation type and cable age, insulation-specific tan delta criteria, and a duration set by detection probability rather than by outage convenience. The immediate dispute resolved itself once the instruction was unambiguous.

Three transferable lessons:

A standards currency review is one of the cheapest high-value scopes available on any cable specification older than about three years. It takes days, not weeks, and removes an entire class of contractual ambiguity. In this field the cable testing standards have moved unusually fast: multiple IEEE 400-series documents revised since 2018, two moved to Inactive-Reserved in successive years, and the principal US acceptance standard restructured in 2025.
If you retain a superseded edition deliberately, say so. "IEEE 400.2-2013 (superseded; retained for consistency with the existing fleet baseline)" is a defensible position that a contractor can price. Citing it silently because nobody checked is not.
Acceptance criteria without a stated scope are worse than no criteria, because they will be applied where they do not belong. Every tan delta threshold needs i ts insulation type attached; every withstand voltage needs its cable class and age condition attached; every duration needs its rationale.

Part 10 — Frequently Asked Questions

Conclusion

The banner on that infographic is right: don't assume it's OK, test it before you energize. The twelve steps behind it are all real tests and all belong in a commissioning scope.

But the value is not in the list. It is in three decisions the list does not make for you.

The sequencing decision

The sheath test belongs before backfill, because it is the only test that finds installation damage and the only one that proves a special bonding scheme is electrically what the drawings say. Everything else is recoverable. That one is not, once the trench is closed.

The method decision

Steps 4 and 5 are alternatives selected by insulation type and cable age, not a sequence to be worked through. DC belongs on laminated dielectric and on jackets. AC and VLF belong on extruded cable. And the choice between a simple withstand and a monitored one determines whether you learn anything at all about the accessories that will cause most of your failures.

The criteria decision

Every acceptance value in this field carries a scope — an insulation type, a cable age, a geography, a duration — and detaching a number from its scope is how healthy cable gets condemned and defective cable gets energized. A test record that states its criterion and the source of that criterion is worth ten that say "pass."

The failure data is consistent about where the risk lives. It lives in the joints and terminations built by hand on site, in the first three years of service, in the jacket damage nobody looked for before the backfill went in, and in the bonding configuration nobody audited against the drawing. None of it is exotic. All of it is findable before energization, by a sequence designed around the engineering rather than around the checklist.

How Keentel Engineering Can Help

Cable system design — sizing and ampacity studies, thermal and duct bank modelling, route and installation design, and sheath bonding design with induced voltage calculation
Test specification development — method selection by insulation type, cable age and circuit criticality; acceptance criteria with stated scope; and hold and witness point definition
Commissioning oversight — witness testing, independent review of commissioning packages, and turnover record definition and audit
Standards currency review — auditing existing cable and testing specifications against current editions, including the recent IEEE 400-series revisions and the ANSI/NETA ATS-2025 restructuring
Failure investigation — root cause analysis of cable and accessory failures, and diagnostic program design
Owner's engineer services — acr oss substation, interconnection and transmission scopes

The highest-value engagement point is before the specification is issued, while the method selection, acceptance criteria, record format and test sequencing are all still open. Contact Keentel Engineering to discuss your project.

References and Further Reading

IEEE cable field testing standards

IEEE Std 400-2023, Guide for Field Testing and Evaluation of the Insulation of Shielded Power Cable Systems Rated 5 kV and Above — https://standards.ieee.org/ieee/400/7618/
IEEE Std 400.1-2018, Field Testing of Laminated Dielectric, Shielded AC Power Cable Systems Using HVDC — https://webstore.ansi.org/standards/ieee/ieee4002018
IEEE Std 400.2-2024, Field Testing of Shielded Power Cable Systems Using Very Low Frequency — https://standards.ieee.org/ieee/400.2/11049/
IEEE Std 400.3-2022, Partial Discharge Field Diagnostic Testing of Shielded Power Cable Systems — https://standards.ieee.org/ieee/400.3/5316/
IEEE Std 400.4 (Inactive-Reserved) and the P400.4 revision — https://standards.ieee.org/ieee/400.4/4847/
IEEE Std 575 (Inactive-Reserved) and the P575 revision — https://standards.ieee.org/ieee/575/4346/

IEC and NETA

IEC 60229:2007, Tests on extruded oversheaths with a special protective function — https://webstore.iec.ch/en/publication/1066
IEC 60502-2:2014 + AMD1:2024, Power cables 6 kV to 30 kV — https://webstore.iec.ch/en/publication/71892
IEC 60840 Ed. 5.1:2023, Power cables above 30 kV to 150 kV — https://webstore.iec.ch/en/publication/63025
IEC 62067 Ed. 3.0:2022, Power cables above 150 kV to 500 kV — https://webstore.iec.ch/en/publication/66951
IEC 60228:2023, Conductors of insulated cables — https://webstore.ansi.org/standards/iec/iec60228ed2023
IEC 60270:2000+AMD1:2015, Partial discharge measurements — https://webstore.iec.ch/en/publication/23842
IEC 60885-3 Ed. 2.0:2015, PD measurements on lengths of extruded power cables — https://webstore.ansi.org/standards/iec/iec60885ed2015
ANSI/NETA ATS-2025, Standard for Acceptance Testing Specifications for Electrical Power Equipment and Systems — https://webstore.ansi.org/standards/neta/ansinetaats2025

CIGRE

CIGRE TB 815 (WG B1.57, 2020), Update of Service Experience of HV Underground and Submarine Cable Systems — https://electra.cigre.org/312-october-2020/technical-brochures/update-of-service-experience-of-hv-underground-and-submarine-cable-systems.html
CIGRE TB 841 (WG B1.38, 2021), After Laying Tests on AC and DC Cable Systems with New Technologies — https://electra.cigre.org/318-october-2021/technical-brochures/after-laying-tests-on-ac-and-dc-cable-systems-with-new-technologies.html
CIGRE TB 797 (WG B1.50, 2020), Sheath Bonding Systems of AC Transmission Cables — https://electra.cigre.org/309-april-2020/technical-brochures/sheath-bonding-systems-of-ac-transmission-cables-design-testing-and-maintenance.html
CIGRE TB 283 (WG B1.18, 2005), Special Bonding of High Voltage Power Cables — https://www.e-cigre.org/publications/detail/283-special-bonding-of-high-voltage-power-cables.html
Smeets & Verhoeven, Reliability of key T&D equipment , ELECTRA No. 323 (2022) — https://electra.cigre.org/323-august-2022/technology-e2e/reliability-of-key-td-equipment-test-laboratory-and-field-experience.html
Gulski et al., Testing and Diagnosis of Power Cables using Damped AC Voltages , CIGRE Science & Engineering N°29 (2023) — https://cse.cigre.org/cse-n029/testing-and-diagnosis-of-power-cables-using-damped-ac-voltages.html

Diagnostics evidence — the CDFI program

Hartlein & Hampton et al., Diagnostic Testing of Underground Cable Systems (Cable Diagnostic Focused Initiative) , DOE Award DE-FC02-04CH11237, Phase 1 Final Report, December 2010 — https://neetrac.gatech.edu/files/2023/09/CDFI_Phase_1_Final-Report.pdf
CDFI Phase II, Chapter 2, Medium Voltage Cable System Issues — https://neetrac.gatech.edu/files/2023/09/2-MV-Issues_25_with-Copyright.pdf
CDFI Phase II, Chapter 3, HV & EHV Cable System Aging and Testing Issues — https://neetrac.gatech.edu/files/2023/09/3-HV-Issues-7_with-Copyright.pdf
CDFI Phase II, Chapter 9, Simple Withstand — https://neetrac.gatech.edu/files/2023/09/9-Simple-Withstand-12_with-Copyright.pdf
CDFI Phase II, Chapter 13, Benefits — https://neetrac.gatech.edu/files/2023/09/13-Benfits-25_with-Copyright.pdf

DC testing of extruded cable

US NRC, DC HiPot Testing of Aged XLPE-Insulated Cables , ML102100460 — https://www.nrc.gov/docs/ML1021/ML102100460.pdf
Megger, Cable insulation test voltages: how high is too high? — https://www.megger.com/en/et-online/june-2019/cable-insulation-test-voltages-how-high-is-too-hig
EPRI 1000273, Estimation of Remaining Life of XLPE-Insulated Cables — https://restservice.epri.com/publicdownload/000000000001000273/0/Product

Field practice

Prysmian, On-Site Testing Guidelines for MV Cables — https://uk.prysmian.com/sites/uk.prysmian.com/files/media/documents/On%20Site%20Testing%20Guidelines%20(2019)%20(3).pdf
OMICRON, resonant test systems for on-site cable testing — https://www.omicronenergy.com/
BAUR, cable sheath fault location — https://www.baur.eu/en/products/cable-fault-location/cable-fault-location-process/cable-sheath-faults
Doble/Techimp, Partial Discharge Testing of Medium Voltage Cables: Online or Offline — https://www.doble.com/wp-content/uploads/Partial-Discharge-Testing-of-Medium-Voltage-Cables_Online-or-Offline-article.pdf
High Voltage Inc, VLF test voltages per IEEE 400.2 — https://hvinc.com/wp-content/uploads/2024/02/IEEE-test-voltages.pdf

Keentel Engineering — Tampa · Austin · Sacramento · Baltimore. EHV, HV and MV power system engineering for utilities, developers and EPCs.

Per unit, a termination in that dataset is roughly 25 times more likely to fail internally than a joint. Overall cause split: internal 64%, external 29%, unknown 7%.

Medium voltage is a different picture. CDFI's North American MV surveys put the median utility failure rate at 3.5 failures per 100 miles per year (lower quartile 1.6, upper quartile 8, with a long right tail — use the median, not the mean, for any "typical utility" statement). The component split across all MV insulation types: cable 56.2%, splices 37.1%, terminations 5.6% — accessories at roughly 43%, with the balance between splices and terminations inverted relative to HV, which makes sense given the population of direct-buried repair splices.

The engineering conclusion. On a new HV or EHV circuit, the cable body arrives as an extruded, continuously monitored, factory-tested and factory-PD-tested product. The joints and terminations are built by hand, on site, often in a manhole or a trench, sometimes at night. Your commissioning test is, in practical terms, a test of the jointer's work. Design it accordingly.

Step 1 — Visual inspection

Verify routing, glands, terminations, supports and identification tags; confirm no physical damage or improper installation.

The item worth adding is one that cannot be verified after the fact. Pulling tension, sidewall bearing pressure and bend radius are only verifiable while the pull is happening. Conductor stress limits of roughly 0.008 lb/cmil for soft copper and hard-drawn aluminium, sidewall pressure limits commonly 500 lb/ft for 600 V–15 kV non-shielded and 300 lb/ft for 25–35 kV cable, and minimum bending radii of 12× the shielded-conductor diameter or 7× the overall diameter — none of these leave a visible mark on an installed cable, and all of them produce a defect that surfaces years later.

A dynamometer chart from the pull, retained in the record, is evidence. A tick in a box after the cable is in the duct is not. If the specification does not require pull-tension recording, "visual inspection" cannot verify installation stress at all — and that is worth raising before the cable goes in the ground.

The second addition: photograph every accessory build. Semicon cut position, stress cone seating, interface preparation and shell closure are permanently invisible once the joint is closed. A photographic record of the build is the only evidence that will exist if the joint fails in year three.

Step 2 — Continuity test

Confirm conductor continuity end to end; detect open circuits or incorrect connections.

Continuity is a floor, not a test. Three upgrades convert this step from a formality into useful data:

Conductor DC resistance against a standard. IEC 60228:2023 (Edition 4.0) gives maximum DC resistance at 20 °C by nominal cross-section and conductor class. Measure with a micro-ohmmeter — 100 A or more on HV conductors — correct to 20 °C, and compare. A significant deviation from the IEC 60228 value points at a bad compression joint or the wrong conductor size, and it is the only test in the sequence that would catch either.

TDR as a commissioning fingerprint. Fire a fast low-voltage pulse into the conductor-to-screen transmission line and record the reflectogram. Every impedance discontinuity — each joint, each termination, the far-end open — produces a signature. Distance is d = v·t/2 , where the velocity of propagation for XLPE is typically around 0.5–0.58 c (take it from the cable datasheet; VOP is the largest single error source in TDR distance).

Why it belongs in a commissioning scope: it is a free, non-destructive as-built record . It proves circuit length, proves the number and position of joints matches the as-built drawing, and becomes the reference trace against which any future fault-location TDR is differenced. A fault trace read against a commissioning baseline is dramatically easier to interpret than one read cold, at 3 a.m., with the circuit out. Very few US commissioning specifications require it. Adding it costs an hour.

Cross-check against the as-built joint schedule. If the TDR shows four reflections and the drawing shows three joints, somebody added a joint. That is worth knowing before energization, not after.

Step 3 — Insulation resistance test

Measure insulation resistance between conductors and earth using a megger; ensure insulation integrity before energization.

This step is genuinely necessary and almost universally over-interpreted.

What IR does well

It detects gross problems — a conductor shorted to screen, a flooded or heavily contaminated termination, an omitted insulation step, moisture in an accessory, a conductive path across a stress cone surface. It is fast and cheap. And it performs one function nobody puts in the specification but everyone relies o n: it protects the expensive test set from being connected to a dead short.

What IR does not do — and this is the part that gets misrepresented:

It cannot detect water trees, the dominant aging mechanism in extruded MV cable. Water trees are AC-field-driven, partially conductive structures. At 5 kV DC they present essentially infinite resistance.
It cannot detect voids, delaminations or PD sites. A void is a capacitive defect. It passes no DC.
It cannot detect a defective accessory interface that will discharge at 1.3 U₀.
It says very little about remaining life. A cable that will fail next month typically meggers as infinity.

Are PI and DAR meaningful on cable? Largely no

Polarization Index (10 minutes ÷ 1 minute) and Dielectric Absorption Ratio (60 s ÷ 30 s) measure dielectric absorption — the slow polarization current of a multi-layer, hygroscopic insulation system. They were developed for, and are defined in, IEEE Std 43 for rotating machine windings, where mica-and-resin tape systems produce a large absorption current. A homogeneous extruded polyethylene dielectric has very little absorption current: the charging transient is dominated by geometric capacitance and settles quickly, so PI on healthy XLPE tends toward 1 and carries essentially no diagnostic content.

On PILC — a genuine laminated, oil-paper, moisture-sensitive system — absorption-based ratios retain real meaning, and IEEE 400.1's DC test evaluation is built on exactly that current-versus-time physics.

Working rule: PI and DAR are meaningful on machines and on PILC, and close to meaningless on extruded cable. Specifying a PI acceptance criterion for an XLPE feeder is a requirement nobody can fail and nobody learns anything from.

Step 4 — AC withstand (hi-pot) test

Apply specified AC test voltage to verify dielectric strength; detect hidden insulation defects before service.

AC is technically the right stress — it excites the same failure mechanisms the cable will see in service, unlike DC. The problem is arithmetic.

Cable is a distributed capacitor. Charging current and the reactive burden on the source are:

I_c = 2πfCV and Q = 2πfCV²

Work that through:

IEC 60502-2 offers four alternative after-installation tests for 6–30 kV cable:

DC at 4 U₀ for 15 minutes — new cables only
AC at rated phase-to-phase voltage U for 15 minutes, at 20–300 Hz
AC at U₀ for 24 hours at power frequency — the "soak test"
VLF at 3 U₀ for 15 minutes at 0.1 Hz

For HV extruded cable, CIGRE Technical Brochure 841 (WG B1.38, 2021) recommends a commissioning test of 1.7 U₀ for 60 minutes at near power frequency for AC XLPE/EPR cables in the 220–400 kV range, with a 1.0 U₀ soak for 24 hours as the recognized alternative where a resonant set cannot be mobilized. The per-U_m values live in IEC 60 840 and IEC 62067 Clause 16.3, Table 12 — get them from the standard rather than from a vendor summary, because the published summaries conflict on whether the multiplier is flat at 1.7 U₀ or descends with voltage class.

Two practical notes

First, a soak test at U₀ for 24 hours proves far less than an over-voltage test — it applies no more stress than normal service. It is a legitimate fallback, not an equivalent. Se cond, an AC withstand with PD monitoring is worth several times an AC withstand alone, for reasons Part 3 makes quantitative.

Step 5 — DC withstand (where applicable)

Verify insulation performance under DC test voltage according to project specifications and cable type.

The parenthetical "where applicable" is doing an enormous amount of work, and it deserves unpacking, because this is the most misunderstood test in the sequence.

Where DC is correct:

Laminated dielectric cable — PILC, pipe-type, pressurized. This is the entire premise of IEEE Std 400. 1-2018, whose titl e states the restriction explicitly: Guide for Field Testing of Laminated Dielectric, Shielded AC Power Cable Systems Rated 5 kV to 500 kV Using High Voltage Direct Current. Oil-impregnated paper does not trap space charge the way crosslinked polyethylene does, and the field distributi on in a laminated dielectric under DC is resistive and stable. DC is the appropriate proof medium here, and the current-versus-time behaviour is genuinely diagnostic.

The jacket or oversheath integrity test — see Step 6. The jacket is a thin unshielded polymer barrier with no water-tree population and no service AC stress, so the space-charge objection does not apply.

Low-voltage insulation resistance at 2.5–5 kV DC, which is safe on all cable types.

Where DC is wrong, and why:

On extruded insulation — XLPE, TR-XLPE, EPR — DC withstand is contraindicated for two documented mechanisms. The US NRC's technical review DC HiPot Testing of Aged XLPE-Insulated Cables is the most useful primary-source summary of the underlying literature:

Space charge. DC field application injects and traps charge carriers in the polymer. Work by Hozumi and colleagues showed DC-treated polyethylene retaining more trapped charge than untreated specimens after equal aging, with charge levels increasing further during subsequent AC aging; Takeda and colleagues identified the mechanism in XLPE as dissociation of impurities — antioxidants and acetophenone crosslinking byproducts — producing hetero space charge attracted to the electrodes. When the cable is re-energized on AC, the trapped charge does not redistribute at 60 Hz. It superimposes on the AC field and produces local stress enhancement above design stress. Dissipation can take up to 24 hours.

Water tree to electrical tree conversion. The EPRI-sponsored Detroit Edison study is the canonical experiment. Aged cable segments that received a DC hipot consistently failed before non-DC-te sted aged segments when returned to accelerated water-treeing conditions — the result summarized as "the dc tested aged cable segment always failed before the non-dc tested previously aged cable segment," with 100% consistency and, in the investigator's words, "no statistical analysis required."

The accurate framing, which matters for writing a defensible specification: DC hipot is not banned. It is contraindicated for service-aged extruded insulation, where it can create the defect it was meant to find. And it is uninformative everywhere on extruded cable, because DC does not excite the AC failure mechanisms. Notably, the same EPRI work found that DC testing prior to aging did not appear to influence cable life — which is why several owner specifications still permit DC hipot on new extruded cable at commissioning while explicitly prohibiting it for maintenance testing of aged cable.

CDFI's own verdict on DC withstand for extruded circuits is blunt: "no evidence that it provides significant benefits for extruded cable circuits", and it "can cause premature failures in aged, XLPE insulated cables."

Steps 4 and 5 are therefore a decision, not a sequence. The decision tree:

Step 6 — Sheath integrity test

Check the cable outer sheath for damage or moisture ingress; confirm proper sheath continuity.

This is the most undervalued test in the sequence and the one most often performed too late.

The number. IEC 60229:2007 (Edition 3.0), Electric cables — Tests on extruded oversheaths with a special protective function , Clause 5, specifies the after-installation test as:

4 kV DC per mm of specified nominal oversheath thickness, capped at 10 kV, applied for 1 minute, with no breakdown.

The metallic layer is driven negative with respect to earth — deliberately, so the test is cathodic rather than corrosive, matching the polarity of the factory routine test.

This resolves a long-running confusion. The "10 kV DC for one minute" that everyone quotes is the ceiling of a thickness-proportional rule, not a flat value. A 2.5 mm HDPE oversheath hits the cap exactly (4 × 2.5 = 10 kV). A thinner MV jacket gets propo rtionally less, and applying 10 kV to it is an over-test. The competing "5 kV" figure that also circulates comes from a different lineage — German VDE practice gives 3 kV for PVC sheath and 5 kV for PE sheath; UK manufacturer practice per the BS 6622 / BS 7835 families gives 8 kV and 5 kV respectively for one minute.

For context, the same standard's factory routine test is 8 kV/mm capped at 25 kV for 1 minute, and the post-abrasion type test is a flat 20 kV — so the after-installation value is deliberately the gentlest of the three.

Leakage current is the real acceptance criterion in practice. The pass/fail is "no breakdown," but a rising or step-changing leakage current is the actionable signal — values moving from the microampere range into milliamperes indicate jacket damage even without a flashover. Test-set manufacturers publish permitted leakage bands that scale with route length, and a 10-minute duration is common practice for QA-grade sheath testing rather than the standard's 1 minute.

Why it matters — three reasons, in order of severity:

1. It is the water ingress path. A punctured oversheath admits moisture to the metallic screen, and from there longitudinally under the screen to the insulation screen and the insulation itself. On extruded cable that is the front end of the water-treeing mechanism: a jacket defect at commissioning becomes an insulation failure in year eight. On lead-sheathed cable it is also a corrosion initiation site, since the jacket is the only barrier between the sheath and the soil.

2. It is the only after-installation test that finds installation damage. Sidewall pressure damage, sharp bends, rock in the bedding, tool strikes during joint-bay backfill — none of these appear on the conductor insulation test, and all of them appear on the sheath test. Which is why it belongs before backfill, not at step 6. Testing before backfilling is the difference between a repair and an excavation.

3. On HV systems it is the only test that proves the bonding scheme is what the drawings say. This one deserves its own section.

Locating a sheath fault. Two techniques, used together:

Pre-location by Murray loop bridge — the faulted screen and a healthy screen or phase form a resistance ratio arm, shorted together at the far end and nulled from the near end. Works for low- and high-resistance faults over kilometre-scale routes. Its accuracy degrades if the two screens differ in specific resistance.
Pin-pointing by step voltage (A-frame) — pulsed DC of defined polarity is fed into the faulty screen; the return current through the soil creates a surface voltage gradient. The operator walks the route with two earth spikes roughly 0.8 m apart. The deflection direction points toward the fault, the reading peaks directly over it and reverses polarity past it. Accuracy is quoted in centimetres, and the polarity reversal is what distinguishes the real fault from a secondary defect.

Bridge to get within tens of metres, A-frame to get to the trowel.

Step 6b — The section the checklist omits: sheath bonding and link boxes

On any single-core HV circuit, current in the core induces a longitudinal EMF in the parallel metallic sheath. You can either let that EMF drive a circulating current — and lose ampacity to sheath I²R loss — or block the current and accept a standing voltage on the sheath. The three schemes:

Both-end (solid) bonding. Sheath earthed at both ends. Standing sheath voltage near zero, no sheath voltage limiters needed, safest to touch — but a closed loop exists, circulating sheath currents flow, and the resulting losses can derate the circuit substantially. Used on short circuits where the ampacity penalty is affordable.
Single-point bonding. Sheath earthed at one end only; the other end open through a link box. No circulating current, no sheath loss, maximum ampacity. The penalty is a standing induced voltage at the open end, proportional to circuit length and load current — which is what limits minor-section length, since the standing voltage must stay under the project's touch-voltage limit while the sheath insulation and the SVLs survive far larger transients during through-faults and surges. A parallel earth continuity conductor is normally laid alongside.
Cross-bonding. The route is split into groups of three equal minor sections, with the sheaths transposed at each of the two intermediate joint bays so that each sheath sees, in turn, the induced EMF of each phase. Over a full major section the three EMFs sum to approximately zero — no circulating current and no large standing voltage. The cost is complexity: sectionalized joints, a cross-bonding link box at each transposition, and degraded cancellation if section lengths are unbalanced.

Here is why the sheath test matters so much on these systems. In a single-point-bonded or cross-bonded circuit, the sheath is deliberately held off earth for the length of a minor section. A sheath-to-earth fault silently converts a single-point-bonded section into a solidly bonded one. Nothing trips. The circuit simply begins circulating sheath current, loses ampacity, and runs hotter than design for the rest of its life. On a cross-bonded system, one jacket fault also unbalances the transposition, so the vector cancellation stops working.

The sheath test is the only commissioning test that detects this. Nothing in the conductor test sequence will.

Sheath voltage limiters are metal-oxide surge arresters rated for the sheath's much lower continuous voltage — typically 0.8 to 4.8 kV MCOV for link-box mounted units, 4 to 14 kV outdoors, with selection driven by a project-specific insulation coordination study. Their real-world failure modes are worth knowing: thermal runaway from undersizing, moisture ingress into a flooded buried link box (the dominant killer), failure to earth (which again silently converts single-point to solid bonding), and loose or corroded link-box connections.

Commissioning scope for link boxes, per current CIGRE guidance, typically includes: verifying enclosure and gland sealing against a submerged duty if the chamber floods; auditing every link position against the bonding drawing (this is where cross-bonding transpositions get built wrong, and it is nearly impossible to detect later without dismantling); torque checks on all bolted connections; insulation resistance on each SVL, disconnected, as a baseline; continuity and resistance of bonding leads and the earth continuity conductor; and phase identification of the coaxial bonding leads, which are a classic mislabelling point. Photograph the interior before closing.

One procedural note that catches people: the sheath test must be performed with the SVLs temporarily disconnected, or they will conduct at 10 kV and mask the jacket entirely.

Two currency flags for anyone writing a specification. IEEE Std 575-2014, the guide for bonding shields and sheaths of single-conductor cables, moved to Inactive-Reserved on 27 March 2025, with a revision project active and retitled to emphasize induced voltage and current calculation. IEEE Std 400.4-2015 (damped AC) moved to Inactive-Reserved on 26 March 2026, also with an active revision. Neither method is withdrawn — the documents are in reserve — but a specification citing them by year is now citing reserved standards. CIGRE TB 797 (2020) is the current reference for bonding system design, testing and maintenance at 66 kV an d above.

Step 7 — Phase identification

Verify correct phase sequence and cable identification; prevent incorrect equipment connections.

Three distinct checks get collapsed into this one line, and they happen at different times with different equipment:

Phase identification (de-energized). Confirms which conductor at end A is which at end B. The standard method: ground one phase at the remote end, leave the others floating, and measure from the near end with the insulation tester at 1 k V DC. The grounded phase reads near zero; the others read in the gigaohm range. Repeat per phase. The trap: all remote conductors must be lifted from ground first. If they are not, every phase reads co ntinuous and the test is worthless. Station a safety watch at the remote end.

Phasing check (energized). Confirms two sources or sections are in phase before paralleling — a live phasing set or voltmeter across the open point reads near zero between corresponding phases and full phase-to-phase voltage between non-corresponding ones.

Phase sequence / rotation (energized). Confirms A-B-C versus A-C-B at the switchgear with a rotation meter, on the primary or via VT secondaries.

Why it matters: closing onto an out-of-phase source produces a fault of up to twice normal magnitude through a transformer or generator, with mechanical damage to windings and possible destruction of the machine. Downstream, motors run backwards and every directional and differential protection element is mis-polarized. The acceptance criterion is unambiguous and worth writing down: each phase identified consistently at both ends, sequence matching the approved single-line, no ambiguity anywhere in the record.

Step 8 — Polarity test

Confirm correct polarity, especially for DC systems and control cables.

Verify positive and negative continuity end to end on station battery and DC distribution feeders, and on all control, CT and VT secondary, and trip circuits, before energization.

Two specific hazards justify the step . Reversed polarity into a charger, inverter or polarized load is immediately destructive. And polarity errors on CT secondaries invert differential and directional protection — producing a relay that be haves perfectly under secondary injection and trips backwards on a real fault. Confirm CT polarity by DC flick test: momentary DC into P1–P2, deflection direction observed at S1–S2. Nothing else proves it as installed.

Step 9 — Earth continuity test

Verify the continuity of protective earthing conductors; ensure a low-resistance grounding path.

For the cable system specifically, this means proving the metallic screen or concentric neutral is continuous end to end and correctly bonded at the intended point or points. Measure screen resistance with a micro-ohmmeter and compare against the calculated value for the screen cross-section and route length.

On single-point-bonded and cross-bonded HV systems this is emphatically not a simple continuity check. The screen is intentionally discontinuous at sectionalized joints. The test must follow the bonding diagram section by section, with link boxes configured, and the acceptance criterion is that the measured configuration matches the design — not that everything reads continuous. A test technician who "fixes" a discontinuity they were not expecting has just destroyed the bonding scheme.

Step 10 — Megger trending

Compare insulation resistance values with previous records; identify insulation deterioration over time.

The intent is right — baselines and trends beat single reading s — but insulation resistance is close to the worst quantity on a cable to trend, for four reasons:

1. The measurement is dominated by the terminations, not the cable. Surface leakage across two stress cones in a humid switchroom swamps the volumetric leakage of a kilometre of XLPE. Two readings a year apart largely record the weather and how recently somebody wiped the terminations.

2. The reading sits near the top of the instrument's useful range. Expected values for XLPE are quoted at above 10,000 MΩ·km at 20 °C. You are trending noise against the instrument ceiling.

3. Temperature sensitivity is brutal. Insulation resistance is roughly exponential in temperature — the working rule is a halving of IR per 10 °C rise. The same XLPE quoted at above 10,000 MΩ·km at 20 °C falls to roughly 10 MΩ·km at 90 °C: a thousand-to-one range. Without accurate insulation temperature — not ambient; cable core temperature, which lags by hours — and correction to a common base, a year-over-year trend is meaningless.

4. Length must be normalized. In M Ω·km or MΩ per 1,000 ft, or a 500 m circuit and a 3 km circuit are incomparable.

What to trend instead. If the objective is genuinely to track condition over the asset's life, the quantities worth baselining at commissioning and repeating are:

Tan delta at U₀ and its tip-up (VLF-TD and VLF-DTD), on MV extruded cable
Partial discharge inception and extinction voltages relative to U₀, and the location of any detected source
The TDR reflectogram, as a physical fingerprint of the circuit
Conductor and screen resistance, temperature-corrected
Sheath leakage current at the standard test voltage

Every one of those has more diagnostic content than an IR reading, and all of them are obtainable during commissioning at marginal additional cost.

Step 11 — Final documentation

Record all inspection and test results; ensure compliance with project specifications before energization.

The commissioning record for a cable circuit is not a completion artifact. It is the baseline for the asset's entire diagnostic life, and it has to be built that way from the start.

A defensible cable test record contains, per test:

The instrument, its serial number, and its calibration certificate date — dated before the test
Test date, ambient conditions, and where obtainable, cable temperature
Technician name and certification level
The acceptance criterion applied and its source standard and edition
Measured values, in normalized units where applicable
The comparison baseline, where the criterion is comparative
Disposition

Plus, for the circuit as a whole: the TDR baseline trace, the as-built joint and termination schedule with positions, the bondi ng diagram with as-built link config urations, the accessory build photographs, the pull-tension record, and the sheath test result with the pre-backfill date on it.

A record that says "pass" is not re-evaluable. A record that says "measured 0.0032 tan δ at U₀; criterion per IEEE 400.2-2024 PE-based insulation, No Action Required below 4×10⁻³; instrument SN 12345, calibrated 14 months prior" can be defended, audited and trended for thirty y ears.

Step 12 — System readiness

Confirm all test results are acceptable; remove temporary test connections and issue clearance for energization.

Two items here are worth elevating from housekeeping to formal hold points:

Removal of temporary connections and grounds. Temporary safety grounds left on, shorting screws left in CT circuits, test leads left connected to a link box — each of these produces an energization event rather than an energization. Make it a signed checklist against a numbered register of every temporary connection installed, not a walk-around.

Restoration of the bonding configuration. Every link box opened for testing must be restored to the design configuration and re-verified against the bonding drawing before energization. SVLs disconnected for the sheath test must be reconnected. This is the single most likely way for a correctly designed and correctly tested bonding scheme to be energized wrong.

Part 3 — The Step That's Missing: Partial Discharge

Part 4 — What the Evidence Says About Whether Any of This Works

Part 5 — Tan Delta Criteria: The Trap Nobody Warns You About

Part 6 — What Changed Between 2023 and 2026

Part 7 — Owner Standards and Why They Override Everything

Putting the engineering back in order, for a new MV or HV extruded circuit:

Before and during installation

Reel tests on receipt — insulation resistance and, where specified, sheath test on the drum, before anything is pulled. A cable damaged in transit should not be installed.
Pull tension and sidewall pressure recording during the pull, retained in the record.
Accessory build photography — semicon cut, stress cone seating, interface preparation, shell closure. Before closing.
Sheath integrity test before backfill. 4 kV/mm capped at 10 kV, one minute, metallic layer negative, leakage current recorded. This is the single most important resequencing in this article.

After installation, before energization

Insulation resistance — as a go/no-go gate and to protect the test set, not as a diagnostic.
Conductor and screen DC resistance, temperature corrected, against IEC 60228 and against the calculated screen value.
TDR baseline trace, archived, cross-checked against the as-built joint schedule.
Phase identification, de-energized, both ends, with all remote conductors lifted from ground.
Polarity verification on DC, control and instrument transformer circuits, including CT flick tests.
Sheath integrity test repeated after backfill and jointing complete, with SVLs disconnected.
Link box audit against the bonding drawing, with torque records and SVL insulation resistance baselines.
Monitored withstand — VLF with tan delta and PD for MV, resonant AC at near power frequency with PD for HV. At 1.7 U₀ for 60 minutes where the standard and the equipment allow, not 15 minutes.
PD localization of any detected sour ce to a specific accessory, with repair and retest.

At energization

Temporary connection register closed out, signed, item by item.
Bonding configuration restored and re-verified, SVLs reconnected.
Phasing check against the adjacent system before paralleling.
Commissioning record issued as a baseline package, not as a completion certificate.

Substation Drawing Sets & Engineering Design

Sun, 16 Aug 2026 10:15:26 GMT

Aug 16, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

The Substation Drawing Set Is the Project

IN BRIEF

Substation projects rarely fail because someone got the fault current wrong. They fail because a drawing went to the field one revision behind the calculation that governs it, because a 60% set was issued before vendor data arrived, or because nobody could say with certainty which PDF was the current one.

This article breaks down the deliverable architecture of a substation electrical package: a 122-sheet E-series drawing set, the four submittal gates that control it, the twenty engineering studies that feed it, and the folder structure that keeps all of it traceable from kickoff to as-built.

It is the same structure Keentel Engineering uses on every substation package we deliver.

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

Ask a substation project manager why the last job slipped and you will hear about long-lead transformers, utility outage windows, or a geotechnical surprise. You will almost never hear the real answer, because it is unglamorous: the engineering deliverables were not under control.

Consider a failure sequence we have walked into more than once. A 13.8 kV feeder relay panel is fabricated from a 60% wiring diagram because the vendor needed to start early. The arc flash study finishes two weeks later and drives an instantaneous setting change, which changes the CT ratio, which changes the panel wiring. The revised sheet is issued — but the fabricator is working from a PDF that was emailed, not transmitted, and the folder it lives in has no revision suffix. The panel arrives at site wired to a superseded drawing. Nobody catches it until functional testing, eleven days before the outage window.

Nothing in that sequence is an engineering error. Every calculation was correct. The failure was entirely one of deliverable architecture: which document is current, who has it, what it depends on, and what had to be finished before it was allowed to be issued.

This is why, at Keentel Engineering, the folder structure and the drawing register are set up before the first sheet is drafted. They are not administrative overhead. They are the control system for the entire design effort, and they are the first thing we hand a client at kickoff.

1. The failure mode nobody puts in the lessons-learned report

3. The four gates: 30% / 60% / 90% / IFC

Part 2 — Frequently Asked Questions: Large Load Interconnection

A substation electrical package is not one document. It is roughly 122 drawing sheets, fourteen calculation packages, twenty study reports, three specification divisions, ten vendor submittal streams, a bill of materials, and a cost estimate — each issued up to four times, each with its own revision history, and most of them dependent on at least one of the others.

Multiply that out and a single substation generates several thousand controlled documents over its design life. Without an imposed structure, that volume degrades in a predictable way: files accumulate in email threads, revisions get distinguished by filename suffixes like "final_v2_REVISED_use-this-one," and the question "is this the current sheet?" becomes unanswerable without a phone call.

The structure that prevents this has three properties. It is decided once, at project setup, before anyone is under schedule pressure. It separates received information from produced information, so client data and vendor prints can never be silently edited. And it makes the milestone the top-level organizing principle inside the drawing folder, so that the 60% set physically cannot be confused with the 90% set.

The percentage milestones are not measures of how much drafting is done. They are decision gates, each with a specific purpose, a specific deliverable list, and — most importantly — specific exit criteria. A milestone that is issued without meeting its exit criteria has not de-risked anything; it has only moved the risk downstream, where it costs more.

3.1 30% — Concept Design

The 30% submittal exists to answer one question: is this the right substation? Not is it detailed correctly, but is the arrangement, the equipment count, the voltage configuration and the footprint the right answer to the client's need. Getting a "yes" here is what makes the remaining seventy percent of the effort worth spending.

Typically 26 of the 122 electrical sheets are issued at 30%: the general sheets, the overall site plan and general arrangement, the governing one-line diagrams, a preliminary ground grid plan, the duct bank concept, the control house layout, and the major equipment schedule. Alongside them go the Basis of Design report, preliminary load flow and short circuit results, transformer sizing, and an AACE Class 4 cost estimate.

Exit criteria: client has approved the arrangement and one-line configuration in writing; utility fault duty and interconnection requirements are received; soil resistivity testing is complete or scheduled; the Basis of Design is signed.

The most expensive mistake in substation engineering is skipping or rushing 30%. Every arrangement decision deferred past this gate gets re-litigated later, when foundations are dimensioned and steel is ordered.

3.2 60% — Developed Design

At 60% every sheet in the set exists and is developed. This is the coordination milestone — the point at which electrical, civil, structural and the client's operations group are working from the same geometry. It is also the gate that releases long-lead procurement, because equipment datasheets are mature enough to issue for quotation.

About 111 of 122 sheets are issued: all one-lines and three-lines, all AC and DC schematics, grounding, raceway, control house, SCADA and equipment details. Draft Division 26 and Division 33 specifications, a preliminary bill of materials and an AACE Class 3 estimate accompany them.

Exit criteria: interdisciplinary clash review complete; grounding, cable sizing, DC battery and station service calculations at least preliminary and consistent with the drawings; long-lead datasheets released for quotation; all 30% comments formally dispositioned.

3.3 90% — Pre-Final Check Set

Ninety percent is the check set. Every sheet is complete, every calculation is signed, and the package has been through internal QA/QC with a documented back-check. Nothing at this gate should be a technical surprise; if it is, the 60% gate was not held properly.

The full 122-sheet set is issued, together with final specifications, a priced bill of materials, the complete signed calculation package and an AACE Class 2 estimate. Relay setting sheets are drafted here even though they are usually issued separately later.

Exit criteria: internal back-check complete and signed; every study report final; all vendor certified drawings received and reconciled against the design; every open client comment closed. No comment may remain open at IFC.

3.4 IFC — Issued for Construction

IFC is not a completion percentage. It is a legal and professional act. The Engineer of Record applies a seal, the CAD files are frozen, the index becomes a controlled document, and the transmittal is signed and logged. From that moment, every change is a numbered revision or a bulletin — there is no such thing as quietly updating an IFC drawing.

This is the discipline that most distinguishes a mature design shop from an immature one. It is also the discipline that protects the client, because after IFC the drawing set becomes the definitive record of what was engineered, what was built against it, and where the responsibility for a deviation sits.

THE KEENTEL STANDARD

We publish the exit criteria for each gate in the Project Execution Plan at kickoff, and we do not issue a milestone that has not met them. If the utility fault duty has not arrived, the 30% goes out flagged as conditional rather than going out silently assuming a number.

Every client comment is logged, dispositioned and closed in a tracked comment register. We close 100% of comments before IFC — not as a target, as a precondition.

Our IFC release is a controlled event: seal applied, CAD frozen, index issued, transmittal signed and logged. Post-IFC changes are issued as numbered revisions with a bulletin describing what changed and why.

A well-structured electrical set is organized by series, and the series numbering carries meaning. A field engineer who knows the convention can find the DC trip circuit for a 138 kV breaker without opening the index, because it is in the E-600s. This is not cosmetic; on a project with 122 sheets, navigability is a safety and schedule issue.

The breakdown below is for a typical greenfield 138 kV / 13.8 kV distribution substation with one main power transformer, an air-insulated switchyard and a prefabricated control house. Scope drives the count — a gas-insulated station, a transmission switching station or a brownfield expansion will shift sheets between series — but the architecture holds.

5. The studies that drive the drawings

Two observations are worth drawing out of that table.

First, the E-600 protection and control series is the largest single block at 23 sheets — roughly one sheet in five. It is also the series most sensitive to late changes, because a relay setting change propagates into schematics, panel wiring, terminal blocks and the SCADA point list simultaneously. Sequencing the coordination and arc flash studies early is the single highest-leverage schedule decision on the electrical package.

Second, only 26 sheets are issued at 30%, but those 26 govern everything else. The one-line diagram alone constrains equipment ratings, protection philosophy, bus arrangement, grounding and the entire E-600 series. An hour of scrutiny on sheet E-201 is worth a week of drafting downstream.

Note what sits in the 30% column. Five studies must be underway before the concept submittal, and three of them depend on data the engineer does not own — utility fault duty, soil resistivity, and the client's load forecast. Chasing that data is a project-management task that starts on day one, not an engineering task that starts when the drafter needs a number.

6. The folder structure that holds it together

2. A folder structure is an engineering deliverable

Below is the top level of the structure we instantiate on every substation project. The principle is that a document's location tells you its status — received or produced, current or superseded, which milestone it belongs to — without opening it.

7. Six ways substation packages go wrong

4. Anatomy of a substation electrical drawing set

Drawings are the output. Studies are the reason the drawings say what they say. A substation electrical package normally rests on around twenty engineering studies, and the order in which they are executed determines the schedule far more than drafting capacity does.

The critical chain runs roughly like this. Utility source impedance and fault duty arrive from the interconnecting utility. That feeds the short circuit study, which sets breaker interrupting and momentary duties and therefore the equipment specification. Breaker and relay selection feeds the coordination study, which produces the settings. The settings feed the arc flash study, which produces incident energy, boundaries and the label schedule. Each link is a hard dependency: you cannot coordinate devices you have not selected, and you cannot compute incident energy from settings you do not have.

A parallel chain runs through grounding. Soil resistivity testing feeds the two-layer soil model, which feeds the IEEE Std 80 grid study, which sizes the conductor and sets grid geometry, step and touch potential and ground potential rise. If the soil testing has not been performed, the ground grid plan cannot leave preliminary status — and because the grid is installed before anything else in the yard, that becomes a construction-sequence problem quickly.

WHAT THIS ARTICLE COVERS

Section 2 — why a document structure is an engineering deliverable, not filing.

Section 3 — the four submittal gates (30 / 60 / 90 / IFC) and the exit criteria for each.

Section 4 — the anatomy of a 122-sheet substation electrical drawing set, series by series.

Section 5 — the twenty studies that drive those sheets, and the dependency chain that sets your schedule.

Section 6 — the project folder structure that makes all of it traceable.

Section 7 — six failure modes we see repeatedly, and the control that prevents each one.

THE KEENTEL STANDARD

Every Keentel substation project is instantiated from the same folder template on day one — 122 folders across four levels, from 00_Project_Administration through 11_Project_Closeout_and_Archive.

Received data lives under 01_Reference_and_Input_Data and is read-only. Client standards, as-builts, survey, soil resistivity and utility fault duty all land there unaltered, with the date and source they arrived under.

Drawings live under 04_Drawings, subdivided first by milestone — 30%, 60%, 90%, IFC — and only then by file type. A superseded issue is moved to 05_Superseded_and_Void and stamped, never deleted and never left in place.

We hand this structure, and the live drawing register that indexes it, to the client at kickoff. You always know what exists, what revision it is at, and where it sits.

WHAT KEENTEL DELIVERS

Full power system study capability: load flow, short circuit, protective device coordination, arc flash and incident energy, motor starting, harmonics and power quality, and reliability analysis — modelled in ETAP, SKM PowerTools and EasyPower.

Substation grounding to IEEE Std 80 including two-layer soil modelling and step/touch potential analysis, lightning shielding to IEEE Std 998, insulation coordination, bus and cable ampacity, and DC battery sizing to IEEE Std 485.

Stamped study reports and relay setting packages, issued as standalone deliverables that a client's protection group can audit line by line — not as an appendix nobody reads.

Arc flash label schedules delivered ready for field application, with the incident energy basis traceable to the coordination study that produced it.

8. Working with Keentel Engineering

Each of the following is a failure we have been brought in to fix on someone else's project. In every case the underlying cause was structural, and in every case a specific control would have prevented it.

7.1 The 30% gate gets compressed

Under schedule pressure, the concept submittal is issued without a signed Basis of Design and without written arrangement approval. Every arrangement question then reopens at 60%, when foundations are dimensioned and steel is being ordered. Rework at that point costs five to ten times what it would have cost at concept.

Control: written arrangement and one-line approval is a hard exit criterion for 30%. No approval, no 60% start.

7.2 Drawings run ahead of the calculations that govern them

A schematic gets drafted with an assumed CT ratio because the coordination study is not finished. The assumption is never revisited. It survives into IFC and is discovered during functional testing.

Control: the drawing register carries an explicit dependency for each sheet, and any value taken from an unissued calculation is flagged as a hold on the sheet — visibly, in a hold list, not in a drafter's memory.

7.3 Vendor data arrives after the drawings that depend on it

Certified transformer or switchgear drawings arrive at 85% design, and the bushing arrangement, control cabinet terminal numbering or dimensions differ from the design assumptions. The E-600 and E-800 series need rework at exactly the wrong moment.

Control: long-lead datasheets are released for quotation at 60%, and the design freezes only against certified vendor drawings — with the reconciliation itself a documented 90% exit criterion.

7.4 Comments are answered but never closed

A client marks up the 60% set. Some comments are addressed, some are addressed differently than intended, and a handful are never dispositioned. Six months later the same comment appears as a construction RFI.

Control: a tracked comment register with a disposition and a responder for every single comment, reviewed with the client before the next gate opens.

7.5 Revision control breaks at IFC

Post-IFC changes get issued as "revised" PDFs by email rather than as numbered revisions with transmittals. Within weeks, three parties hold three different versions of the same sheet and none of them is identifiable as current.

Control: IFC freezes the CAD and the index. Every subsequent change is a numbered revision with a bulletin, a signed transmittal and a superseded copy moved and stamped.

7.6 The as-built never happens

Construction finishes, the team demobilizes, and the contractor redlines sit in a box. The owner operates for a decade against an IFC set that no longer matches the station — and the next expansion project starts by paying an engineer to walk down and re-document what is there.

Control: as-built incorporation is a contracted deliverable with its own folder, its own budget line and its own seal — not a courtesy at the end of the job.

WHAT YOU GET ON DAY ONE

A Project Execution Plan with published exit criteria for every gate — so "60% complete" means the same thing to you as it does to us.

The full project folder structure, instantiated and shared, so your document control team is never guessing where something lives.

A live drawing register listing every sheet, its file names, its status at each milestone and its current revision — updated continuously, not reconstructed before each submittal.

A named Engineer of Record who holds the seal from kickoff to as-built.

Keentel Engineering provides electrical power engineering design and studies services for utilities, industrial owners, developers and EPC partners. We take substation and switchyard packages from concept through IFC, construction support and as-built — and we do it inside the deliverable structure described in this article, on every project, without exception.

Design services

Substation and switchyard design , 4.16 kV through 345 kV, AIS and GIS, greenfield and brownfield expansion
Complete E-series drawing packages: arrangement, bus design, one-line and three-line development, protection and control schematics, grounding and lightning protection, raceway and cable systems, control house electrical, SCADA and communications equipment detailing and schedules
Industrial and facility power distribution, medium-voltage switchgear and MCC design, standby and emergency power systems
Technical specifications, Division 26 and Division 33, and equipment datasheets issued for quotation

Studies and analysis

Load flow, short circuit, protective device coordination, arc flash and incident energy, motor starting, harmonics and power quality, and reliability analysis in ETAP, SKM PowerTools and EasyPower
Substation grounding to IEEE Std 80 with two-layer soil modelling; lightning shielding to IEEE Std 998; insulation coordination; bus and cable ampacity; DC battery sizing to IEEE Std 485
Arc flash label schedules and NFPA 70E compliance packages, delivered ready for field application
Relay setting calculations, setting files and functional test procedures

Project delivery

Owner's engineer and design review services, including independent review of a third party's 30/60/90 submittals
Construction support: RFI response, field change notices, site observation and submittal review
Commissioning support, FAT and SAT witnessing, energization planning and as-built record drawings

Have a substation package that needs to be under control?

Whether you are scoping a new station, holding a 60% set that will not stop moving, or preparing to issue for construction and wanting a second set of eyes before the seal goes on — talk to us. Initial scoping conversations are complimentary, and we will tell you plainly whether your package is ready for the next gate.

Grid-Forming vs Grid-Following BESS Inverters | Keentel

Sat, 15 Aug 2026 12:29:56 GMT

Aug 15, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

Grid-Forming vs. Grid-Following BESS Inverters: What Actually Changes Inside the Control Loop — and Inside Your Interconnection Study

A Keentel Engineering Grid IQ technical guide for utility planners, interconnection engineers, and storage developers

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

If you have sat in an interconnection study review in the last three years, you have seen some version of the comparison graphic that prompted this article. On the left, a grid-following (GFL) inverter: a phase-locked loop (PLL) reading a transmission-grid voltage waveform, an injected current that dutifully follows it, and the caption "needs an existing grid to lock onto." On the right, a grid-forming (GFM) inverter: a self-generated voltage reference V, θ, a current that follows its own waveform, and the caption "can start and run a grid on its own."

The diagram is correct. It is also, on its own, dangerously incomplete — because it implies the difference is a binary switch, and it implies the GFM column is simply better.

Neither is true. The real difference is a change in what the converter regulates over a specific time frame, with consequences that propagate all the way out to your short-circuit model, your relay settings, your PSCAD study scope, and your interconnection agreement date. And the GFM column carries a set of hard engineering constraints — current limiting, transient stability under saturation, headroom, and its own weak-and-strong-grid instability modes — that rarely make it into the marketing.

This guide unpacks what the two panels of that graphic actually mean at the control-loop level, what the field evidence says, where the standards and grid codes now sit as of mid-2026, and how the choice shows up in real project work.

Introduction: A Two-Panel Diagram That Hides a Decade of Engineering

2. Grid-Following: How It Works, and Exactly Where It Breaks

Part 2 — Frequently Asked Questions: Large Load Interconnection

The most common definition — "grid-following controls current, grid-forming controls voltage" — is a useful mnemonic and a poor engineering specification. Both devices ultimately switch semiconductors to produce a voltage; both end up injecting a current.

NERC, the UNIFI Consortium, and ESIG have converged on a time-domain definition instead, and it is the one that belongs in your interconnection documents:

Grid-forming: maintains an internal voltage phasor — AC voltage magnitude and angle — that is constant or nearly constant in the sub-transient to transient time frame.

Grid-following: maintains its output current phasor magnitude and angle during the sub-transient period, then transitions to power control in the transient time frame.

The sub-transient window is roughly 0–5 cycles; the transient window is tens of cycles. UNIFI is careful to note these are general descriptions rather than hard limits.

Why the framing matters: it makes the distinction testable. UNIFI Version 3 (January 2026) converts "voltage source behind an impedance" into a measurable criterion — high-frequency voltage-phasor stiffness demonstrated by a passive response rising ten-fold per decade of frequency, or upper gain bounds of 0.005–0.02 p.u./p.u., or an estimated series inductance below a maximum within a ±5 Hz band. You can put that in a functional specification. You cannot put "controls voltage instead of current" in one.

It also disposes of a persistent misconception: grid-forming is not the same thing as islanding capability, and it is not the same thing as black start. UNIFI V3 organizes GFM into four capability tiers — Tier 1 (voltage magnitude support only, e.g. a GFM STATCOM), Tier 2 (voltage angle and frequency support with minimal islanding), Tier 3 (full support with extended islanding — where grid-scale BESS sits), and Tier 4 (Tier 3 plus black start). A device can be legitimately grid-forming and still be unable to energize a dead bus.

2.1 The architecture

A GFL plant is, electrically, a Norton equivalent: a controlled current injection behind a large shunt impedance. Its control chain is familiar:

Measure PCC voltage.
The PLL estimates the voltage phase angle and establishes a rotating dq reference frame.
Outer loops (active power / DC-link voltage, reactive power / AC voltage) generate d- and q-axis current references.
Fast inner current loops drive the modulator to track those references.

The critical property is that the plant has no angle of its own. Its angle is a measurement of the grid's angle. That single fact explains every GFL limitation that follows.

Remove the external voltage reference and the PLL has nothing to track — which is why ESIG's capability comparison lists black start for GFL as "not usually possible." It is not a tuning problem. It is architectural.

2.2 The weak-grid instability mechanism

As the Thévenin impedance seen at the POI rises — as short-circuit ratio falls — the inverter's own current injection begins to materially move the PCC voltage angle. The PLL measures that movement. The controller responds. The response changes the injection. The loop gain of this positive-feedback path scales with grid impedance.

The mechanism has been characterized in impedance terms: the PLL introduces a negative resistance in the Z_qq impedance component in the low-frequency range, and the second-order low-pass filter used in more sophisticated PLL structures can further degrade damping in weak grids. In one national-laboratory impedance study, instability was triggered as PCC short-circuit ratio fell from 2 to 1.15, producing a 3 Hz oscillation on a line trip and an undamped 10 Hz oscillation after curtailment.

The design trade-off is unforgiving and well quantified. CIGRE small-signal analysis found that a PLL bandwidth of 11 Hz produced a 44 rad/s mode with poor damping, while reducing bandwidth to 4 Hz eliminated the mode and achieved 86% damping at SCR = 2.3 — at the cost of response speed to genuine grid changes. You buy weak-grid stability with slower dynamics.

2.3 On the SCR threshold — resist the urge to quote a single number

Screening heuristics in wide circulation place "strong" above SCR 3.0, "weak" between 2.0 and 3.0, and "very weak" below 2.0. These are useful for deciding when a detailed EMT study is required. They are not a physical instability boundary, and presenting them as one will get you challenged in a study review.

NERC's own position is explicit: "There is no specific short circuit threshold one can consider as a 'weak grid.'" Weak-grid problems are system-specific. And CIGRE's comparative analysis found that a well-tuned GFL resource held 50% damping at post-contingency SCR of 0.9, with limitations appearing only in exceptional post-contingency scenarios near SCR 1.0 — while a poorly tuned GFM exhibited instability comparable to GFL.

The defensible framing: SCR 2–3 is a study-trigger band; the true limit is tuning-dependent and, for modern well-tuned equipment, sits considerably lower.

What grid codes do pin down are test floors. UNIFI V3 requires testing at SCR of 10, 5, 3, 1.5 and 1.2 (X/R = 6) with no tripping. AEMO's voluntary specification steps SCR from 20 down through 10, 3, 2, 1.5 to 1.25, with stability at 1.25 as the pass criterion. MISO's draft framework uses a ramp-down to SCR 1.25.

2.4 The field evidence: Odessa I and Odessa II

The strongest real-world case that PLL-based synchronization is a bulk-system reliability concern comes from ERCOT.

Odessa, 9 May 2021. A single-line-to-ground fault on a GSU transformer near Odessa produced a total 1,340 MW reduction, of whic h 1,112 MW was solar PV. The single largest identified failure mode was PLL loss of synchronism at 389 MW — inverters tripping because the voltage phase angle shifted roughly 10 degrees during the fault. Two facilities alone accounted for 239 MW and 150 MW. A further 153 MW was legacy momentar y cessation below 0.9 p.u. voltage.

Odessa, 4 June 2022. A B-phase-to-ground fault on the 345 kV system caused 2,555 MW of loss, includ ing 1,711 MW of solar PV across fourteen facilities. The breakdown is instructive: inverter AC overcurrent 459 MW; passive anti-islanding (phase jump) 385 MW; AC overvoltage 295 MW; DC voltage unbalanc e 211 MW.

The critical detail is that second line item. Following the 2021 event, PLL loss-of-synchronism protection had been disabled at many facilities. Inverters from the same manufacturer then tripped on passive anti-islanding, which misinterpreted the phase-angle shift on fault recovery as an islanding signature — triggering when the angle between voltage and current phasors exceeded 15 degrees within 500 ms.

That is the whole argument in one paragraph. Patching a PLL symptom moved the failure to an adjacent phase-angle-sensitive function, because a device that infers grid state from measured angle will trip on angle transients somewhere. The issue is architectural, not a settings error.

3. Grid-Forming: Voltage Source Behind an Impedance

6. Protection: The Consequence Nobody Budgets For

This is the section that separates a serious GFM discussion from a brochure. It is also where most project risk actually lives.

5.1 The fundamental constraint

A synchronous machine's fault contribution is set by physics — it will deliver several times rated current, roughly 3–5× per ESIG's comparison, with higher sub-transient values, because its thermal mass and magnetic circuit permit it. An inverter has essentially no thermal mass. Its current limit is instantaneous and strict, set by the semiconductor's safe operating area. Exceeding it, even momentarily, is not permitted.

Reported GFM fault-current figures diverge widely, and the divergence is a reporting artifact worth understanding:

ESIG: IBRs provide 1–1.5× rated current.
Sandia expert interviews: most GFM inverters provide at most 1.1–1.2 p.u.; Sandia's own measurements of single-phase GFM units ranged 1.3–2.25 p.u. depending on manufacturer.
NREL's GFM research roadmap cites sub-transient values of 4–6 p.u. for under 10 cycles, with steady-state below 2 p.u.
UNIFI's illustrative short-term rated current: 1.5× full-rated for 2 seconds.

These are not contradictory. The 4–6 p.u. figures describe sub-cycle behavior limited by internal impedance before control acts; the 1.1–1.5 p.u. figures describe sustained contribution after limiting engages. Conflating them is the single most common error in GFM writing, and it will produce a short-circuit model that does not match reality.

5.2 Why limiting is hard to do well

Every current-limiting strategy trades something away:

Current-reference saturation. Fast and accurate — but the device becomes a current source. Voltage-source behavior, the thing you bought, is lost precisely when you need it.
Switch-level limiting. Faster than anything else, acting within a few switching cycles — but it bypasses the entire GFM control structure, risking controller instability and severe waveform clipping.
Virtual impedance. Subtracts a synthetic voltage drop from the reference, requiring |Z_vi| ≥ E₀/I_max. Retains voltage-source behavior, at the cost of slower action and reduced small-signal margin.
Power set-point modulation. Simple, but the governing equations contain no dependence on I_max, so current limiting within the inverter's boundaries cannot be guaranteed.
Voltage-based limiting. Documented drawback: the inverter becomes a controlled current source during limiting, similar to a GFL inverter, which can have a destabilizing effect in weak grids — the exact failure mode GFM was procured to avoid, reappearing at the worst moment.
Hybrid schemes. Current-reference limiting for immediate action, virtual impedance taking over — best trade-off, most complex implemen tation.

5.3 The transient stability consequence

Under normal operation the power-angle relationship is the familiar P = (E·V_g / X_g)·sin δ. Once current limiting engages, it becomes P = V_g·I_max·cos(δ − φ) — sinusoidal to cosinusoidal. Output power drops immediately, the internal reference angle accelerates under droop dynamics, and loss of synchronism follows if the critical clearing angle is exceeded.

Two mitigations matter in review:

Increasing the current-angle priority φ raises the critical clearing angle — but φ cannot be pushed too far without confining the unit to permanent limiting.
Virtual impedance X/R is decisive: a purely resistive virtual impedance can quickly cause instability through a small critical clearing angle, while an inductive one substantially increases it. A ratio of X_vi/R_vi ≈ 5 has been proposed as a workable compromise.

Integrator windup in the cascaded loops is a named additional hazard requiring explicit anti-windup logic.

5.4 Headroom — and a genuine disagreement worth knowing

A GFM inverter's ability to hold its internal voltage phasor against a disturbance is bounded by available current. Dispatched at 1.0 p.u., there is nothing left with which to form voltage.

Here the specifications diverge, and the divergence is substantive:

UNIFI and ESIG hold that a guaranteed power margin must be reserved to enable a firm inertia response independent of base loading, and that GFM performance may be reduced at or near power, energy or current limits.
MISO deliberately targets capabilities that do not require holding capacity or energy in reserve, explicitly avoiding requirements that would force hardware oversizing or materially affect state-of-charge management.
ERCOT's Advanced Grid Support takes the same position: AGS does not require additional short-circuit current capability, does not require reserving capacity or state of charge for a specified inertia contribution, and is framed as achievable without additional hardware or maintained energy reserves.

These positions are reconcilable once you scope them. Core voltage-source behavior is essentially free in firmware. Firm, contracted inertia and fast frequency response services are not — those require reserved power and energy, and that is a commercial decision, not a control-design one.

One genuine gap: there is essentially no authoritative quantitative literature on GFM-specific thermal or battery-degradation implications from continuous small active-power injections for damping and inertia. Treat vendor claims in this area as unverified.

GFM provides inertia equivalent to a synchronous machine

It does not, and the field data is unambiguous. During a 12 June 2023 NEM frequency event with RoCoF near −0.17 Hz/s, an Australian grid-forming battery's inertial response peaked roughly 480 ms later than a comparable synchronous generator. At the 200 ms mark, the synchronous machine had released 13.7% of its two-second response; the battery had released 2.0%. The Australian program's own conclusion is that GFM inertia differs qualitatively and quanti tatively from synchronous inertia and is not a direct replacement for it. It is genuinely valuable and genuinely different.

The Iberian blackout proves we need more inertia

It does not. The ENTSO-E expert panel's final report, published 20 March 2026, concluded that even with higher system inertia the loss of synchronism would not have been avoided — simply adding more rotating mass would not have been enough. The event was driven by voltage and reactive-power control gaps and oscillations, including a 0.63 Hz converter-d riven forced oscillation and a 0.2 Hz inter-area mode in the preceding half hour, with renewables in fixed power factor mode and conventional units meeting required reactive output less than 75% of the time. GFM is relevant to that event — but as a source of voltage-source behavior and dynamic voltage support, not inertia. Making the inertia argument from Iberia will be corrected by anyone who has read the report.

More grid-forming is always better

No. ESIG has documented a tested GFM inverter that becomes unstable when short-circuit ratio rises above 4 — the mirror image of the GFL weak-grid problem, and a predictable one: voltage sources paralleled through low impedance exchange large circulating currents for small angle errors. Analytical work reaches the same conclusion, that too many GFM converters lead to instability and the allocation between GFM and GFL must be balanced. Meanwhile national-laboratory guidance suggests roughly 25–30% GFM penetration as a target and notes stability issues rising above 60–70% instantaneous GFL penetration, while ESIG places the threshold where GFM becomes necessary at around 75–80% IBR penetration. There is an optimum, not a maximum.

GFM is just a firmware setting

Partly. Baseline GFM control — voltage-source behavior, phase-jump response, droop, damping — is predominantly a software matter, and several major Australian projects were commissioned as conventional plants and later retrofitted to GFM through firmware and a connection-alteration process without hardware modification. What costs hardware is overcurrent headroom beyond the existing rating (oversized semiconductors and cooling), black-start capability, and energy-buffer sizing. S cope those three separately in any budget.

8. The Compliance Landscape as of August 2026

1. The Definition That Matters (and the One to Stop Using)

Standards

IEEE 2800-2022 remains the anchor for transmission-connected IBRs, and it is deliberately technology-neutral — it neither requires nor defines grid-forming controls . That changed structurally in December 2025, when the IEEE Standards Association approved three related projects:

Note P2800a's framing carefully: it is about removing barriers to GFM, not mandating it. Mandates are coming from grid codes and ISOs, not from IEEE.

IEEE 2800.2-2026, the conformity-assessment recommended practice for test and verification of IBRs interconnecting with bulk power systems, was approved by the IEEE SA Standards Board in February 2026 and published on 30 June 2026. It is not GFM-specific; GFM test and verification is being routed to P2800.1.

At distribution level, the IEEE 1547 series has no grid-forming standard. IEEE 1547-2018 and 1547a-2020 are built around grid-following smart-inverter functions — volt-var, volt-watt, frequency-watt, ride-through. Intentional islanding is handled by IEEE 1547.4-2011, currently under revision. There is no 1547-series project comparable to P2800.1.

NERC's September 2023 white paper, Grid Forming Functional Specifications for BPS-Connected Battery Energy Storage Systems , remains the document most US ISOs are building on. Its three headline functional specifications are phase-jump performance, system-strength support in the sub-transient time scale, and stable operation through and following loss of the last synchronous machine.

UNIFI Version 3 (January 2026) is currently the most quantitatively detailed public specification, and the trajectory from V1's deliberately non-prescriptive "performance principles" to V3's testable numeric thresholds is the story of the last three years in one document series.

On the modeling side, WECC-approved generic GFM models REGFM _A1 and REGFM_B1 are in the standard library, with REGFM_C1 (hybrid control) and REPCGFM_C1 (plant control) advanced through WECC in September 2025. This matters more than it sounds: study-grade generic models are what let a GFM claim survive an interconnection queue.

Grid codes and mandates

There is no US federal grid-forming mandate. FERC Order No. 901 drove the IBR reliability standards program and the registration workplan (completed May 2026, with 531 IBR facilities identified and registration effective 15 May 2026). Order No. 909 approved PRC-024-4 and PRC-029-1 — ride-through and protection settings, not GFM. Notably, FERC explicitly declined a recommendation to require grid-forming capability, saying instead that it encourages NERC to remain proactive in using its standards development authority.

ERCOT is the first US mandate, and it is already in force. NOGRR272 and PGRR121, establishing Advanced Grid Support (AGS) requirements for inverter-based Energy Storage Resources, were approved by the ERCOT Board in September 2025 and by the PUCT in November 2025, effective 1 December 2025. Applicability: ESRs with a Standard Generation Interconnection Agreement dated 1 April 2026 or later. ERCOT separately proposed an AGS Incentive Program in May 2026 (NPRR1333) for resources not otherwise covered — $1,500/MW one-time, availability-adjusted, under a $25 million cap — which remains pending as of August 2026.

If you are developing storage in ERCOT, that SGIA date is now a design gate, not a compliance footnote.

Elsewhere: Great Britain's GC0137 added a GBGF technical specification to the Grid Code as an optional capability, with the current Guidance Note (Issue 4, December 2025) prescribing ≤5 ms response, ≥60° phase-jump withstand, 2 Hz/s RoCoF withstand over a rolling 500 ms window, damping factors of 0.2–5.0, and control bandwidth below 5 Hz. GC0163 (July 2024) removed the restriction to physical impedance, permitting virtual (software) impedance — a meaningful shift toward performance-based rather than method-based compliance. Australia's AEMO specification remains voluntary, with a Technical Requirements Review running to a draft report in Q3 2026, final in Q4 2026, and a rule change to the AEMC targeted for Q1 2027. The EU's ENTSO-E published its Phase II technical report on grid-forming requirements in November 2025, intended for integration into NC RfG 2.0 — but RfG 2.0 has not been adopted, and ENTSO-E's own December 2025 position paper notes the process has been deprioritised with no communicated timeline. Elsewhere, VDE FNN in Germany, RTE in France, Fingrid in Finland and CEN in Chile have mandatory requirements in place.

Markets: capability is running ahead of procurement

The most instructive fact about GFM economics in 2026 is that three sophisticated markets reached three different answers within five months.

Germany launched market-based procurement of instantaneous reserve ( Momentanreserve ) on 22 January 2026, with premium products priced at roughly €805–€888.50 per MW·s per year and basic products at €76–€109.50, against 90% and 30% availability requirements respectively. This is a real, bankable revenue stream for grid-forming storage.
Australia's AEMC decided against creating an operational inertia market in October 2025, finding the benefits outweighed by the cost of designing and running it.
Great Britain's NESO awarded its Mid-Term (Y-1) Stability Market Round 2 in February 2026 — 7.3 GVA·s of inertia for £10.3 million — and awarded nothing to battery storage, with all grid-forming BESS submissions failing at the technical assessment stage. This despite Stability Pathfinder Phase 2 ha ving contracted five grid-forming batteries in Scotland, and a portfolio of GFM projects at Blackhillock, Kilmarnock South and Eccles collectively contracted for several GVA·s of inertia.

A vendor's summary to ERCOT captures the developer's position exactly: the cost categories are CAPEX for controls and stability buffers, balance of plant, OPEX for state-of-charge recalibration and troubleshooting complexity, time to revenue through extended testing, and revenue risk — "no upside without market."

On cost premium specifical ly: there is no credible published GFM-versus-GFL cost premium figure from any national laboratory, ISO, or industry consortium. ESIG's position is that enabling GFM controls in a BESS carries low incremental cost when incorporated early in the design process and is materially cheaper than synchronous condensers, STATCOMs or tra nsmission reinforcement. Any specific percentage you encounter online traces to vendor marketing or paywalled market research, not engineering study. Treat it accordingly.

4. Side-by-Side: The Engineering Comparison

5. The Hard Part: Current Limiting and Transient Stability

7. Four Claims to Push Back On

Two things stand out. First, everyone agrees on the physics — voltage-source behavior in the sub-transient frame, inertia without frequency measurement, weak-grid stability, positive damping. Second, nobody agrees on the numbers. A vendor cannot build one product that "complies" globally; compliance is per-market.

Traditional overcurrent protection assumes fault current substantially exceeds load current. In a system where the dominant sources contribute 1.1–1.5 p.u., distinguishing a fault from a temporary overload by magnitude alone becomes unreliable. ESIG's recommendation is to move toward differential protection and communications-assisted transfer-trip or blocking schemes rather than magnitude-based discrimination.

Specific findings that should shape your protection scope:

Negative sequence. GFM IBRs — like GFL IBRs and synchronous machines — should provide negative-sequence current during faults, because a great deal of conventional protection depends on negative-sequence quantities. IEEE 2800-2022 already requires incremental negative-sequence reactive current dependent on negative-sequence terminal voltage.
Distance protection. GFM units with a hard current limiter can cause distance-rela y under- or overreach. GFM units using virtual impedance can potentially enable correct operation — a direct, practical engineering argument for VI-based limiting that goes beyond stability .
Power swing protection. No published literature specifically examines GFM impact, though practitioners consistently identify it as a major concern. This is an open risk, not a solved problem.
Waveform quality. Hardware limiting and fast switch action produce distorted waveforms in the first few cycles before software control restores a sinusoidal current — relevant to any relay algorithm relying on early-cycle phasor estimation.
Expert consensus is that GFM units should remain in GFM mode during faults, supporting voltage while limiting current, rather than switching to a following mode.

Sandia has published seven named knowledge gaps in this area, including measured GFM fault currents, ride-through strategies aligned with relay operation, GFM–GFL interaction during and after faults, and adequate GFM-plus-relay models for protection studies. If a project's protection philosophy is being built on GFM assumptions, that scope needs EMT-based validation, not a spreadsheet.

Australia has made this the explicit gating question. AEMO has stated that GFM BESS has demonstrated voltage-waveform stability but is not yet confirmed to deliver protection-quality fault current — sufficient in magnitude, duration and sequence composition for relays — and has launched a dedicated Grid-Forming Inverter Protection-Quality Fault Current Trial to find out, with contracts running from July 2026. This is despite an Australian grid-scale battery fleet above 9,000 MW and a NEM pipeline of 33.2 GW in which roughly 74% is grid-forming.

3.1 The architecture

A GFM converter behaves as a Thévenin equivalent — a voltage source behind an impedance. It establishes its own internal voltage magnitude and angle, and active and reactive power flow arise from the angle and magnitude difference between that internal phasor and the grid, exactly as they do for a synchronous machine.

The consequences follow directly:

Synchronization is inherent, not measured. There is no PLL in the primary control path, so there is no PLL to destabilize.
Inertial response is inherent. A change in system frequency changes the angle difference, which changes power, with no frequency measurement and no measurement window.
Phase-jump response is inherent. A step in grid angle immediately changes the angle difference and therefore the power — opposing the disturbance.
The device can energize a dead network, subject to sufficient DC energy, overload capability for transformer inrush, and a ground reference.

3.2 The five control families

"Grid-forming" is a behavior class, not an algorithm. Five families dominate, and the distinctions matter when you are reviewing a vendor's model.

Droop control

Linear P–ω and Q–V characteristics. Enables system-wide synchronization with all units converging on a common frequency, and supports parallel operation without communications. The most common approach in microgrids, and deployed at scale in grid-connected BESS. Its known pitfall is the low-pass filter on the power measurement, which is the knob balancing harmonic rejection against response speed; it also lacks explicit inertia unless added.

Virtual synchronous machine (VSM) / synchronverter

Emulates synchronous machine dynamics digitally, with virtual inertia, damping and flux linkage as tunable parameters. ESIG identifies VSM as the most common approach in GFM pilots, largely because its behavior is familiar to system planners. The trade-off: you inherit the machine's pathologies along with its virtues, including its oscillatory modes, and the parameter tuning burden is real.

Virtual oscillator control (VOC) / dispatchable VOC (dVOC)

Emulates a nonlinear oscillator whose natural frequency is nominal grid frequency. Droop behavior emerges from the dynamics rather than being imposed. Offers near-instantaneous synchronization and load sharing using local measurements only. Still largely in pilot and development phase.

Matching control

Emulates synchronous behavior from the DC side — the DC-link voltage plays the role of rotor speed. Conceptually distinct because it structurally couples DC-bus energy to frequency rather than treating DC dynamics as an afterthought.

Power synchronization control (PSC)

Notable for preserving voltage-mode operation during faults rather than reverting to current-source behavior — directly relevant to the current-limiting problem discussed in Section 5.

3.3 Quantified performance expectations

Specifications have moved from principles to numbers. Representative requirements:

9. Three Anonymized Case Studies

10. A Practical Selection Framework

Confidentiality note

The three engagements below are presented in anonymized and generalized form. Ratings, locations, utility identities, dates and specific study results have been altered or aggregated to protect client confidentiality. They are included to illustrate the engineering decision patterns that recur in this work, not to characterize any single project.

Case Study A — Standalone BESS in a Low-SCR Transmission Pocket

Situation. A developer held a queue position for a ~200 MW / 800 MWh standalone battery at a rural 138 kV substation in a region already hosting several hundred MW of utility-scale solar. The transmission provider's screening study returned a composite short-circuit ratio in the low 2s under system-intact conditions and below 1.5 under a credible N-1 outage of the strongest source. The queue cluster restudy identified control-interaction risk and required full EMT analysis.

What the study found

With a conventional grid-following control model, the plant remained stable at system-intact SCR but exhibited poorly damped oscillation in the 5–10 Hz range following the N-1 contingency, with damping falling below acceptable margin as the neighboring solar plants were dispatched to high output. Sensitivity runs showed the mechanism was PLL-driven: reducing PLL bandwidth restored damping but degraded the plant's reactive response speed enough to threaten a separate voltage-recovery requirement. The two requirements could not be satisfied simultaneously with GFL controls at that location.

Re-running the same contingency set with the vendor's grid-forming control model — a droop-based implementation with inductive virtual impedance for current limiting — eliminated the oscillatory mode across the full dispatch sweep and remained stable through the SCR ramp-down to 1.25.

Outcome and lessons

The developer elected GFM. Two points proved decisive in negotiation with the transmission provider, and both generalize:

The alternative was network reinforcement. The comparison was never GFM cost versus GFL cost; it was GFM cost versus a synchronous condenser or a line upgrade with a multi-year schedule.
The current-limiting method was a study input, not a detail. An ear ly vendor model using current-reference saturation showed loss of synchronism for a close-in three-phase fault at the reduced-SCR condition. The virtual-impedance variant did not. Had the model been accepted without interrogating the limiter, the study would have produced a stable-looking result for a plant that was not.

The project also absorbed an unbudgeted cost: the plant's short-circuit contribution changed enough between GFL and GFM assumptions to require the interconnecting utility's protection group to revisit relay coordination on two adjacent lines.

Case Study B — Industrial Campus Microgrid with Islanding and Black Start

Situation. A manufacturing client with a process load that cannot tolerate a momentary interruption sought resilience against a distribution feeder with a poor historical outage record. The scope was a ~15 MW / 60 MWh BESS alongside existing rooftop and carport PV and two legacy standby diesel generators, with a requirement for seamless transition to island and the ability to restore the campus from a fully de-energized state.

Engineering decisions

Grid-forming was not optional here — islanding and black start are Tier 3 and Tier 4 capabilities, and no grid-following architecture provides them. The consequential decisions were elsewhere:

No mode switching. The initial vendor architecture switched between GFL control when grid-connected and GFM control when islanded, gated by a PCC status signal. Comparative studies of transition schemes show this approach settling in roughly 2 seconds on islanding, with voltage dipping to about 0.92 p.u. and a frequency overshoot near 61.2 Hz — against 2 cycles and no voltage transient for an architecture that stays in GFM mode continuously and maintains the same operating point through the transition. For a process load that trips on a 100 ms sag, that difference is the entire project. The design was changed to continuous GFM operation.
Sizing was driven by inrush, not energy. Black start required overload headroom for transformer energization. A controlled voltage ramp on the order of 0.1 p.u./s was adopted for soft-start, and interconnection transformers were configured delta–wye-grounded to provide a ground reference for the island without the substation source.
Load shedding was designed in, not bolted on. Shedding roughly one-fifth of non-critical load during restoration materially reduced the required GFM capacity and therefore the capital cost of the battery.
Protection was redesigned, not re-set. In islanded operation the available fault current fell to roughly 1.2 p.u. of the BESS rating. Overcurrent coordinatio n built for a utility source was unusable. The scheme moved to differential protection on the campus loop with a communications-assisted blocking scheme, plus a separate settings group for islanded operation.

Outcome and lessons

The islanded transition met the process requirement. The generalizable lesson is tha t the protection redesign, not the battery, was the schedule-critical path — and it was discovered late, because the initial scope treated the BESS as an equipment purchase rather than a change to the c ampus's fault-current source.

Case Study C — ERCOT Storage Portfolio and the Advanced Grid Support Deadline

Situation. A portfolio owner held several inverter-based Energy Storage Resource positions in ERCOT at different stages: two with executed interconnection agreements predating the AGS applicability date, and three whose agreements would be executed after it. The owner's initial assumption was that AGS was a future concern and that a firmware update near commissioning would satisfy it.

What the review found

Three issues surfaced, each of which recurs across the ERCOT queue:

The applicability boundary split the portfolio. The projects with earlier agreements were outside the mandate; the later ones were inside it. Two units of the same model, from the same supplier, on the same balance sheet, now faced different requirements. The commercially rational move — standardizing on AGS-capable configuration across the portfolio — was only obvious once the split was quantified.
Model quality, not inverter capability, was the binding constraint. The requirement is not merely that the equipment be capable; it is that the plant demonstrate compliance through model quality and unit testing. The supplier's available EMT model for one platform did not reproduce the required phase-jump behavior with the settings proposed for the site, and the plant-level controller model had not been validated in combination with the unit model. Resolving that took longer than any hardware question.
The scope of AGS was narrower than assumed — in the owner's favor. ERCOT's requirement does not oblige additional short-circuit current cap ability and does not require reserving capacity or state of charge for a specified inertia contribution. Once that was established, the projected comm ercial impact on merchant dispatch was far smaller than the owner's initial planning assumption. The cost of AGS was a testing and modeling cost, not an energy-arbitrage cost.

Outcome and lessons

All five projects proceeded to an AGS-capable configuration. The transferable lesson is that in a mandate regime , the compliance risk sits in model fidelity and test evidence rather than in the inverter itself — a nd that engineering review should start at the interconnection-agreement date, well before equipment selection.

11. Frequently Asked Questions

Grid-forming is not a default and not a differentiator on its own. Work through these in order:

What is the SCR at the POI — system-intact and post-contingency, composite rather than plain? If post-contingency composite SCR approaches or falls below the low 2s, GFM belongs in the study scope.
Is there a mandate? In ERCOT, check the SGIA date first. In Europe, check the national TSO requirement, not the pending EU code.
Is islanding, seamless transition, or black start in scope? If yes, GFM is required — and specify continuous GFM operation, not mode switching.
What is the alternative being displaced? GFM's business case is almost never against GFL. It is against synchronous condensers, STATCOMs, curtailment, or network reinforcement.
Is there a revenue mechanism? Germany pays for inertia. ERCOT has proposed an incentive. Great Britain's enduring stability market has not yet awarded a battery contract. This determines whether GFM is a cost or an asset.
Interrogate the current limiter. Ask which method the vendor uses, and require the EMT model to demonstrate transient stability for close-in faults at post-contingency SCR. Virtual impedance with an inductive X/R is materially better behaved than current-reference saturation.
Separate firmware scope from hardware scope. Baseline GFM behavior is controls. Overcurrent margin, black start and energy buffer are hardware and belong in a separate line item.
Budget the protection study. Fault-current contribution, distance-relay reach, negative-sequence availability and power-swing behavior all change. This is the most commonly under-scoped item in GFM projects.
Check model availability early. WECC-approved generic models exist. Vendor-specific EMT models and validated plant controllers are the long pole.
Confirm the mix, not just the technology. Both too little and too much GFM cause problems. The allocation between GFM and GFL at a common electrical area is a study question.

12. Conclusion

The two-panel graphic is a good starting point and a poor stopping point. The engineering reality is that grid-following and grid-forming are two answers to one question — where does this converter get its phase reference? — and that answer propagates into weak-grid stability, fault behavior, protection coordination , black-start capability, model requirements, compliance obligations, and revenue.

Three things are true simultaneously in mid-2026, and holding all three is what separates a sound project decision from a fashionable one:

Grid-following architecture has demonstrated bulk-system reliability limits. The Odessa events showed that a device inferring grid state from measured phase angle will trip on angle transients somewhere, and that disabling one such function simply relocates the failure.
Grid-forming solves real problems and introduces new ones. Current limiting, transient stability under saturation, protection-quality fault current, and its own high-SCR instability mode are open engineering questions, not settled ones.
The regulatory and market landscape is fragmenting, not converging. ERCOT mandates. FERC declined to. Germany pays. Great Britain has not. Europe's code is stalled. Australia is still testing the fault-current qu estion that would let GFM count toward system strength.

For a developer, that means the grid-forming decision is no longer a technology preference — it is an interconnection strategy, a protection scope, a modeling program, and, depending on jurisdiction, a compliance deadline with a date attached.

How Keentel Engineering Can Help

Keentel Engineering supports utilities, developers, EPCs and generator owners across EHV, HV and MV power system work, with offices in Tampa, Austin, Sacramento and Baltimore. Our storage and IBR practice covers:

Interconnection and POI engineering — feasibility, system impact and facilities studies, and POI design support
Power system studies — load flow, short circuit, EMT/PSCAD stability and control-interaction analysis, system strength and SCR assessment
Model review and validation — vendor EMT and positive-sequence model quality, plant controller validation, and functional-specification test evidence
Protection engineering — coordination in low-fault-current systems, distance-relay performance with IBRs, and islanded-operation settings groups
Compliance support — IEEE 2800, NERC PRC standards, and ERCOT NOGRR/PGRR requirements including Advanced Grid Support
Owner's engineer services — technology selection, specification development, factory and site test witness

If you are evaluating grid-forming for a project, the highest-value conversation usually happens before equipment selection — at the point where the SCR assessment, the mandate applicability date, and the protection scope are still open. Contact Keentel Engineering to discuss your project.

References and Further Reading

Specifications and functional requirements

UNIFI Consortium, Specifications for Grid-Forming Inverter-Based Resources, Version 3, January 2026 — https://docs.nlr.gov/docs/fy26osti/98381.pdf
NERC, White Paper: Grid Forming Functional Specifications for BPS-Connected Battery Energy Storage Systems, September 2023 — https://www.nerc.com/globalassets/our-work/white-papers/white_paper_gfm_functional_specification.pdf
NERC, Grid Forming Technology: Bulk Power System Reliability Considerations, December 2021 — https://www.nerc.com/globalassets/our-work/white-papers/white_paper_grid_forming_technology.pdf
AEMO, Voluntary Specification for Grid-forming Inverters: Core Requirements Test Framework, January 2024 — https://www.aemo.com.au/-/media/files/initiatives/engineering-framework/2023/grid-forming-inverters-jan-2024.pdf
NESO, Grid Forming Guidance Note, Issue 4, December 2025 — https://www.neso.energy/document/289921/download

Standards

IEEE SA, P2800a — Amendment: Reduce Barriers for IBRs with Grid-Forming Equipment — https://standards.ieee.org/ieee/2800a/12386/
IEEE SA, IEEE 2800.2-2026 — https://standards.ieee.org/ieee/2800.2/10616/
IEEE SA Standards Board minutes, December 2025 — https://standards.ieee.org/wp-content/uploads/2025/12/1225sasb-meeting-minutes.pdf

Technical foundations

NREL, Research Roadmap on Grid-Forming Inverters, NREL/TP-5D00-73476 — https://docs.nlr.gov/docs/fy21osti/73476.pdf
NREL, Overcurrent Limiting in Grid-Forming Inverters, NREL/CP-5D00-88404 — https://docs.nlr.gov/docs/fy24osti/88404.pdf
NREL, Study of Seamless Microgrid Transition Operation, NREL/TP-5D00-86102 — https://docs.nlr.gov/docs/fy24osti/86102.pdf
NREL, Parallel Grid-Forming Inverter-Driven Black Start, NREL/CP-5D00-87257 — https://docs.nlr.gov/docs/fy24osti/87257.pdf
ESIG, Grid-Forming Technology in Energy Systems Integration, 2022 — https://www.esig.energy/wp-content/uploads/2022/03/ESIG-GFM-report-2022.pdf
ESIG, Grid-Forming BESS Brief, March 2025 — https://www.esig.energy/wp-content/uploads/2025/03/ESIG-GFM-BESS-brief-2025.pdf
CIGRE Science & Engineering N°37, Grid-Forming and Grid-Following Inverters: A Dynamic Performance Evaluation — https://cse.cigre.org/cse-n037/grid-forming-and-grid-following-inverters-a-dynamic-performance-evaluation-using-rms-emt-and-small-signal-analysis.html
Sandia National Laboratories, Protection of 100% Inverter-Dominated Power Systems, SAND2024-04848 — https://www.osti.gov/servlets/purl/2429968/
NERC, Integrating Inverter-Based Resources into Low Short Circuit Strength Systems, 2017 — https://www.nerc.com/globalassets/who-we-are/standing-committees/rstc/irpwg/item_4a._integrating-_inverter-based_resources_into_low_short_circuit_strength_systems_-_2017-11-08-final.pdf

Events and field performance

NERC, Odessa Disturbance Report, May 2021 — https://www.nerc.com/globalassets/our-work/reports/event-reports/odessa_disturbance_report.pdf
NERC, 2022 Odessa Disturbance Report — https://www.nerc.com/comm/RSTC_Reliability_Guidelines/NERC_2022_Odessa_Disturbance_Report%20(1).pdf
ENTSO-E, Expert Panel Final Report on the 28 April 2025 Iberian Blackout, March 2026 — https://www.entsoe.eu/news/2026/03/20/entso-e-publishes-expert-panel-final-report-on-28-april-2025-blackout-in-spain-and-portugal/
ENTSO-E, Grid Forming Capability of Power Park Modules — Report on Technical Requirements, November 2025 — https://eepublicdownloads.entsoe.eu/clean-documents/Publications/SOC/20251104_GRID_FORMING_CAPABILITY_OF_POWER_PARK_MODULES.pdf
ARENA / Ekistica, Grid-Forming Battery Portfolio Series Summary Report, 2025 — https://arena.gov.au/assets/2025/07/Ekistica-ARENA-Grid-Forming-Battery-Portfolio-Series-Summary-Report.pdf
ARENA, Hornsdale Power Reserve Virtual Machine Mode Testing Summary Report, 2022 — https://arena.gov.au/assets/2022/03/hornsdale-power-reserve-virtual-machine-mode-testing-summary-report.pdf
AEMO, Black System South Australia 28 September 2016 — Integrated Final Report — https://www.aemo.com.au/-/media/files/electricity/nem/market_notices_and_events/power_system_incident_reports/2017/integrated-final-report-sa-black-system-28-september-2016.pdf
NESO / Ofgem, Technical Report on the events of 9 August 2019 — https://ofgem.gov.uk/system/files/docs/2019/09/eso_technical_report_-_final.pdf

Grid codes, mandates and markets

ERCOT, NOGRR272 — Advanced Grid Support Requirements for Inverter-Based ESRs — https://www.ercot.com/mktrules/issues/NOGRR272
ERCOT, NPRR1333 — Establish an Incentive Program for Advanced Grid Support — https://www.ercot.com/mktrules/issues/NPRR1333
ERCOT, Advanced Grid Support BESS Functional Specification and Test Framework, July 2024 — https://www.ercot.com/files/docs/2024/07/10/2024_07_ERCOT_IBRWG_Advanced%20Grid%20Support%20Inverter-Based%20ESR%20Functional%20Specification%20and%20Test%20Framework_v1.pdf
Netztransparenz, Marktgestützte Beschaffung von Momentanreserve — https://www.netztransparenz.de/de-de/Systemdienstleistungen/Frequenzhaltung/Marktgest%C3%BCtzte-Beschaffung-von-Momentanreserve
NESO, New contracts awarded under Mid-Term (Y-1) Stability Market Round 2, February 2026 — https://www.neso.energy/news/new-contracts-awarded-under-mid-term-y-1-stability-market-round-2
AEMO, Grid-Forming Inverter Protection-Quality Fault Current Trial — https://www.aemo.com.au/energy-systems/electricity/national-electricity-market-nem/nem-forecasting-and-planning/transition-planning/transitional-services---type-2-services/grid-forming-inverter-protection-quality-fault-current-trial
AEMC, Efficient Provision of Inertia — Final Determination, October 2025 — https://www.aemc.gov.au/sites/default/files/2025-10/ERC0339%20Final%20determination%209%20October.pdf

Keentel Engineering — Tampa · Austin · Sacramento · Baltimore. EHV, HV and MV power system engineering for utilities, developers and EPCs.

MOD-025 & MOD-026 After COD: Verification & Model Validation Guide

Sat, 15 Aug 2026 11:50:50 GMT

Aug 15, 2026 | Blog

A Coordinated Electric System Interconnection Review—the utility’s deep-dive on technical and cost impacts of your project.

contact@keentelengineering.com

813-389-7871

Schedule a Consultation

MOD-025 and MOD-026 After COD: The Complete Guide to Post-Commissioning Verification Testing and Model Validation

Keentel Engineering | Generator verification, dynamic modeling and NERC compliance engineering

ERCOT enforces all of the above through simulation, which means your model is your compliance case. The bar is now high:

Whole-facility scope. The model must represent everything the IT load, the UPS and power conversion, the cooling plant, the protection and control systems in formats compatible with ERCOT's study platforms (PSS/E, PSCAD, TSAT).
Real control loops, not approximations. Generic textbook representations are unacceptable. The model must capture the actual inner control behavior of your power electronics.
Hardware-validated converter models. For electronic loads, the PSCAD model must be benchmarked against actual hardware testing including voltage ride-through and subsynchronous response. A model assembled from standard PSCAD library blocks fails by definition, because a generic block has never been tested against your vendor's hardware. The good news: validation is a hardware-type test, so results for a given converter product are reusable across every facility that uses it.
Format migration. Facilities that previously submitted the older composite load model (CMLD) format must transition to EPRI's PERC1 format.
Three checkpoints. Models are reviewed before the stability study begins (no model, no study), before each quarterly stability assessment, and for electronic loads one final time before energization, when you must submit as-built models with a documented comparison against the previously studied data and a sworn attestation that the model matches actual field settings. ERCOT's review takes 10 business days, extendable by 20 put it on your critical path.
A living obligation. Change your technology, controls, or relay settings in a way that affects ride-through including converting a crypto mining site to an AI data center — and you've triggered a new interconnection study, even if your megawatts don't change.

A generating facility spends years in the interconnection queue. Studies are run, models are submitted, equipment is procured, the plant is built, and one day it produces its first megawatt-hour. Commercial operation date. The project team celebrates and moves on to the next asset.

And then, quietly, two compliance clocks start running.

The first is MOD-025 — verification of the facility's real and reactive power capability, due within twelve months of the commercial operation date. The second is MOD-026 — verification and validation of the dynamic models the facility relies on, due within roughly a year of commissioning for a new resource. Neither of these is satisfied by the interconnection study package. Neither is satisfied by the OEM's generic model. Both require the registered entity to demonstrate, with evidence, that the as-built, as-commissioned, as-operating facility behaves the way the grid planner's models say it does.

This is where a surprising number of otherwise well-run projects get into trouble. The interconnection model was a design-stage artifact. The plant that actually got built has different firmware, different plant controller settings, different protection thresholds, and often a different reactive capability than the one that was studied. Post-COD verification is the process of closing that gap — and documenting that you closed it.

This guide walks through what MOD-025 and MOD-026 actually require after COD, how the testing works on site and off site, what "model verification" and "model validation" really mean in PSS®E and PSCAD, the findings that come up again and again, and how to build an evidence package that survives an audit.

The 12 months after COD are the ones that catch people out

Part 2 What MOD-026 actually requires

Part 2 — Frequently Asked Questions: Large Load Interconnection

The short version

MOD-025-2, Verification and Data Reporting of Generator Real and Reactive Power Capability and Synchronous Condenser Reactive Power Capability , asks a deceptively simple question: can your plant actually produce the real and reactive power that the planning models and registration data say it can?

The Generator Owner must verify both, document the verification on a standardized form, and submit it to the Transmission Owner, Transmission Planner and Planning Coordinator.

Who it applies to

The standard applies to Generator Owners with:

individual generating units greater than 20 MVA, or
generating plants with an aggregate nameplate rating greater than 75 MVA,

that are directly connected to the Bulk Electric System. Transmission Owners carry a parallel obligation for synchronous condensers greater than 20 MVA.

For a utility-scale solar or storage facility, the aggregate threshold is the one that matters. Practically every utility-scale inverter-based resource in North America is in scope.

The requirements

R1 — Real Power. Verify the real power capability of the generating units in accordance with Attachment 1, and submit the completed Attachment 2 form within 90 calendar days of the date the test data was recorded or the historical data was selected.
R2 — Reactive Power. Verify reactive power capability, over-excited and under-excited, on the same basis and with the same 90-day submittal clock.
R3 — Transmission Owners carry the equivalent obligation for synchronous condensers.

The periodicity and the post-COD trigger

Three timing rules govern MOD-025:

New facilities: within 12 calendar months of the commercial operation date. This is the post-COD trigger and it is the one most frequently missed.
Recurring: at least once every five years, with no more than 66 calendar months between verifications.
On change: within 12 months of discovery of a capability change exceeding ten percent that persists for more than six months.

That third trigger deserves attention at storage and hybrid facilities, where augmentation, cell replacement, inverter firmware changes and repowering can move capability materially.

The 90-day trap

Entities often plan carefully around the 12-month COD deadline and then lose the submittal. The 90-day clock runs from the data recording or selection date , not from the deadline. A test run in month three of the compliance year must be submitted by month six — waiting until month eleven is a violation even though the annual deadline has not passed.

Attachments 1 and 2

Attachment 1 contains the substance: the periodicity schedule, the verification specifications, the data recording requirements, and the specification for the one-line diagram that must accompany the submittal. It tells you what has to be measured, where, and to what standard.

Attachment 2 is the reporting form — the one-line diagram, the data table organized by measurement point, the verification data summary, and the ambient conditions at the time of verification.

The point of measurement is a recurring source of confusion. Capability is not simply the sum of inverter nameplate ratings. It is what the facility can deliver at the agreed measurement point, after collector system losses, transformer losses, auxiliary and station service load, and any reactive limits imposed by the plant controller. Getting the measurement point and the correction methodology right is most of the engineering.

What is different about an inverter-based resource

MOD-025 was written in a world of synchronous machines, and the language still reflects that. Applying it to a solar plant or a battery introduces practical problems that have nothing to do with the standard's text:

Real power capability depends on resource availability. A solar plant cannot demonstrate full real power output at 4 p.m. in December. Verification must be scheduled against irradiance, or qualifying historical operating data must be screened out of the historian.
Curtailment interferes. ERCOT curtailment, congestion, or a basepoint below capability can invalidate a test window.
Reactive capability is voltage-dependent and control-mode-dependent. The reactive capability an IBR can actually deliver at the point of measurement varies with POI voltage, inverter AC voltage, active power output, and whether the plant controller is in voltage, power factor or reactive power control mode.
Reactive capability is often limited by settings, not hardware. This is the single most common MOD-025 finding at IBR plants — the equipment can do more than the plant controller is configured to allow.
Storage adds a second quadrant. A BESS must be considered charging as well as discharging, and reactive capability may differ between the two.

None of this changes the requirement. It changes the planning that has to go into it.

Verification versus validation

MOD-026 is the standard that trips people up, largely because two similar words mean two very different things.

Model verification asks: does the model represent the equipment that is actually installed? It is a documentation and parameter exercise. You compare the model's configurable, site-specific parameters against the as-commissioned inverter settings, plant controller configuration, protection thresholds, transformer data and firmware versions, and you demonstrate traceability from the field to the model.

Model validation asks: does the model reproduce what the plant actually did? It is a simulation exercise. You take a real event — a staged test or a recorded system disturbance — replay it against the model, and compare simulated response to measured response.

MOD-026 requires both. Passing one does not satisfy the other. A model can be perfectly parameterized and still fail to reproduce a recorded fault response; a model can happen to match one event and still contain parameters that bear no relationship to the installed equipment.

Applicability under MOD-026-2

The revised standard developed under Project 2020-06 substantially broadens the resource types in scope:

That last row matters. A Transmission Planner who observes a mismatch between your plant's recorded behavior and its modeled behavior can require verification on technical justification alone, outside the normal periodicity.

Version check. MOD-026-1 has been the enforceable version for years, and MOD-026-2 was developed and balloted under Project 2020-06 with a phased implementation. Before you build a compliance plan around specific dates, confirm the enforceable version, the approved implementation plan, and the phase applicable to your facility with your Regional Entity and Planning Coordinator. The engineering below applies either way; the deadlines do not.

The requirements, in plain language

R1 — the Transmission Planner and Planning Coordinator jointly define the rules. Model specifications consistent with MOD-032, acceptance criteria, submittal process, and — critically for inverter-based resources — EMT model requirements and formats. You cannot scope a MOD-026 submittal without first obtaining your TP/PC's R1 criteria. They differ between regions, and ERCOT's are specific.
R2 — submit a verified positive-sequence dynamic model. The obligation is to demonstrate that configurable, site-specific parameters represent the parameters of the in-service equipment, with supporting documentation of model structure and parameter accuracy.
R3 — submit a verified EMT model for inverter-based resources, FACTS, HVDC and synchronous condensers. This includes large-signal disturbance test comparisons, verification of site-specific parameters, representation of enabled protections and limiting functions, and validation against recorded active and reactive power events.
R4 — change management. When hardware, software or control settings change in a way that alters dynamic response, submit an updated model or a verification plan within the timeframes in Attachment 2 — up to 180 days after return to service.
R5 — the Transmission Planner reviews within 120 calendar days and either accepts or denies with explanation.
R6 — you have 120 calendar days to respond to a denial or a model review request with updated models, a verification plan, or additional justification.

Periodicity

What can be done remotely, and what cannot

A well-run post-COD verification program divides cleanly into work that must happen at the plant and work that is better done off site.

Requires site presence or site coordination:

Staged reactive capability testing that sweeps the plant across its operating envelope
Verification of instrumentation, CT/PT circuits and metering accuracy at the point of measurement
Confirmation of time synchronization at the device level — GPS, IRIG-B, PTP or NTP distribution
Retrieval of settings files and event records from devices without remote access
Witnessing of inverter or plant controller functional behavior where remote visibility is limited

Better done off site:

Test plan development and TSP/ISO coordination
Screening of historian and disturbance records for qualifying events
Parameter comparison between field settings and model files
All PSS®E and PSCAD simulation work
Model tuning, validation plotting, and quantitative comparison
Report preparation, Attachment 2 completion, and evidence assembly

The practical model that works for most owners is remote engineering support during a site-executed test, with an option to attend on site for the first campaign, for a complex hybrid facility, or where the O&M provider has not run this type of test before. Keentel Engineering supports both — on-site witnessing and fully off-site engineering — and most programs end up as a blend.

Planning a MOD-025 verification at an IBR plant

A workable sequence:

Confirm the point of measurement and the correction methodology before anything else. Every number in the submittal depends on it.
Decide test versus historical data. If the historian holds qualifying operating points — full real power at a representative ambient, plus over- and under-excited reactive excursions — historical data avoids a staged test entirely. This is often the fastest and cheapest route, and it is fully compliant.
Coordinate with ERCOT and the TSP. Reactive testing moves POI voltage. It requires notice, a voltage schedule accommodation, and sometimes an outage coordination entry.
Set the data acquisition up properly. Real power, reactive power, voltage and current at the measurement point, plant controller mode and setpoints, individual inverter status, ambient temperature and irradiance, all time-synchronized, at a resolution and duration that satisfies Attachment 1.
Run the test against the plant's actual limits. The objective is to find the boundary, not to confirm a number someone typed into the RARF three years ago.
Reconcile with registration data. If verified capability differs from the RARF, the RARF gets updated. Silent divergence is a finding waiting to happen.
Complete Attachment 2 and submit inside 90 days.

Planning a MOD-026 validation campaign

MOD-026 validation is less about scheduling a test and more about being able to capture an event when one occurs. That means:

Disturbance monitoring that records voltage, current, active power, reactive power and frequency at adequate resolution
Records long enough to capture pre-event, event and recovery
Plant controller and inverter-level data, not just substation relay data
Time synchronization good enough to align substation records with inverter records
A retention policy that keeps the records long enough to be useful

Facilities that have this in place can usually satisfy MOD-026 validation from recorded disturbances without a staged test. Facilities that do not are forced into staged testing, OEM type-test records, or a documented explanation of why validation data could not be obtained — and that last option invites scrutiny.

This is also the point where MOD-026 and PRC-028 overlap in a useful way. The disturbance monitoring capability that PRC-028 requires is precisely the capability that makes MOD-026 validation straightforward. Owners who treat them as one program rather than two save real money.

Part 5 PSCAD and EMT modeling: why it is now unavoidable

A positive-sequence model verification is not "we sent the OEM's .dyr file." A defensible verification covers:

Structure and configuration

Correct model modules selected for the inverter, electrical controls and plant controller
Correct interconnection of those modules
Plant equivalent impedance, collector system representation and transformer data consistent with the as-built facility
Consistency with the MOD-032 data submitted to the Planning Coordinator

Parameter traceability

Every configurable, site-specific parameter mapped to a field source — an inverter settings file, a plant controller configuration export, a relay setting, a nameplate
Reactive control mode, droop, deadband, time constants and limits matching the commissioned plant controller
Active power controls, ramp rates and frequency droop matching the commissioned configuration
Voltage and frequency trip parameters matching the protection actually in service

Behavior

Clean initialization and a flat run with no drift
Numerically stable response across the study cases
Sensible response to voltage steps, faults and frequency events
No premature blocking or tripping in a region where the resource is required to ride through

That last point is the bridge between MOD-026 and PRC-024 / PRC-029. A model whose trip parameters do not match the field is not just a modeling problem — it can be masking a ride-through non-conformance, or manufacturing a phantom one that causes the Planning Coordinator to derate your plant in studies.

The findings that come up over and over

Plant controller reactive limits in the model that do not match the commissioned configuration
Voltage and frequency trip settings modeled from a design specification rather than the relay settings actually loaded
Momentary cessation represented in the model but disabled in firmware — or the reverse
Firmware version in service newer than the version the model package was issued for
Frequency droop and deadband defaults never updated to the ERCOT-required values
Ramp rate limits absent from the model entirely
Auxiliary load and station service unrepresented, shifting net capability
Collector equivalent impedance carried over from the interconnection study rather than as-built
Supplemental reactive devices — capacitor banks, STATCOMs — omitted from the plant model
Positive-sequence and EMT models that disagree with each other about the same plant

Compliance is not the analysis. Compliance is the evidence of the analysis, retained and retrievable.

A complete post-COD package generally contains:

Applicability determination and the basis for it
The TP/PC verification criteria obtained under MOD-026-2 R1
Test plans, with dates and revision history
Raw recorded data with time stamps and provenance
Settings files as extracted, with extraction dates
Firmware version evidence
The model files as submitted, versioned
Parameter comparison matrices with a source reference for every parameter
Simulation output and comparison plots
Attachment 2 for MOD-025, complete with the one-line diagram
Transmittal evidence — dated proof of submittal inside the 90-day and 365-day windows
Correspondence with the TP/PC, including review responses
A register of open items, with owners and target dates

The last item is the one that distinguishes a mature program. Auditors are generally far more comfortable with an entity that has identified a gap, documented it, and is tracking it to closure than with an entity whose package is silent on a question it should have asked.

Case studies

Part 1 — What MOD-025 actually requires

The following are anonymized composites drawn from representative engagements. Facility names, owners, capacities and dates have been changed or generalized, and no confidential client information is disclosed.

Case study 1 The reactive capability that existed on paper only

Facility: approximately 200 MWac utility-scale solar PV, ERCOT, roughly nine months past COD.

Trigger: MOD-025 verification approaching the 12-month post-COD deadline.

What we found: the registration data and interconnection agreement reflected a reactive capability of roughly 0.95 leading and lagging at the point of interconnection. Verification testing established that the facility could reach the lagging requirement, but under-excited capability fell approximately 12 percent short of the registered figure at full real power output.

The hardware was not the constraint. Three separate limits were stacking: a conservative reactive limit configured in the plant controller during commissioning and never revisited; auxiliary and station service load that had not been accounted for in the original capability calculation; and an inverter-level reactive priority setting that reduced available reactive output at high active power.

What we did: documented the verified capability, prepared the Attachment 2 submittal against measured performance rather than the registered value, and issued a separate engineering recommendation identifying the plant controller limit as recoverable capability. The owner elected to revise the plant controller configuration, after which a follow-on verification demonstrated capability consistent with the interconnection agreement, and the RARF was updated.

Takeaway: at IBR facilities, reactive capability shortfalls are usually configuration problems, not equipment problems. They are also usually invisible until someone tests for them — and they can quietly affect both compliance and the plant's ability to meet its voltage schedule.

Case study 2 — Two models, one plant, two different answers

Facility: approximately 150 MW / 600 MWh battery energy storage system, ERCOT, first MOD-026 submittal after commissioning.

Trigger: post-commissioning model verification and validation.

What we found: the PSCAD model supplied with the project represented momentary cessation during low-voltage events. The firmware actually commissioned at site used continuous current injection through the ride-through region — a different, and considerably better, behavior. The positive-sequence model was a third variant, with voltage trip thresholds that matched neither the EMT model nor the protection settings loaded in the relays.

The mismatch had gone unnoticed because each model had been reviewed in isolation, by different parties, at different stages of the project.

What we did: established the as-commissioned configuration from inverter and PCS settings files and firmware version evidence, rebuilt the parameter set for both models from that single source, and validated against a recorded transmission fault captured by the facility's disturbance monitoring. The corrected EMT model reproduced the measured active and reactive power recovery closely; the original model had predicted a cessation and recovery ramp that never occurred.

Takeaway: the value of MOD-026 is not the paperwork. A model that predicts momentary cessation where the plant actually injects current will cause the Planning Coordinator to see a weaker, less helpful resource than you own — and can influence study outcomes and curtailment decisions for years.

Case study 3 — A protection setting that only the model knew about

Facility: approximately 300 MWdc solar PV, ERCOT, roughly four months past COD.

Trigger: combined model-to-settings verification performed alongside a ride-through review.

What we found: the positive-sequence model carried an undervoltage trip threshold consistent with the original design specification. The relays in service had been commissioned with a more sensitive setting, inside the region where the facility was required to ride through. Neither the EMT model nor the model quality test results had flagged it, because the EMT model represented the design value as well.

Because the plant had not yet experienced a qualifying disturbance, the discrepancy would not have surfaced through validation. It surfaced through parameter-by-parameter comparison between the relay settings files and the model.

What we did: documented the discrepancy, quantified the exposure against the applicable ride-through envelope, and issued a recommendation to bring the protection setting into conformance. The setting was revised, the models were updated to match, and the model verification submittal reflected the corrected configuration.

Takeaway: model validation against recorded events cannot find a defect the plant has not yet been exposed to. Parameter verification can. This is exactly why MOD-026 requires both, and why treating verification as a formality is a false economy.

Frequently Asked Questions — Earthing Types and Keentel Engineering Practice

The questions below are the ones we hear most often from clients, EPC design teams, and reviewing engineers when selecting and specifying earthing systems. The answers reflect Keentel Engineering's practice across utility, industrial, and renewable-generation grounding studies.

Part 3 Post-COD testing: on site and off site

Part 4 PSS®E modeling: what verification really involves

Part 6 Building an audit-ready evidence package

Acceptable validation methods

Model validation must be based on one of two sources:

Staged tests — factory type tests, hardware-in-the-loop assessments, or manufacturer tests, and
Measured system disturbances — the facility's recorded response to actual transmission system events.

For volt/var and reactive validation, the selected event must involve a genuine dynamic reactive excursion; a slow, quiescent voltage drift proves nothing. For active power and frequency validation, the event must meet the applicable perceived frequency deviation threshold — 0.10 Hz in ERCOT and the Western Interconnection , 0.05 Hz in the Eastern Interconnection, and 0.15 Hz in Québec.

Where qualifying results cannot be obtained, the standard requires that the reason be documented. "We could not find an event" is an acceptable position only if you can show you looked, that your monitoring was capable of capturing one, and that you have said so in writing.

What Attachment 1, Table 1.2 asks of an inverter-based resource

For IBRs, FACTS and HVDC, the model content requirements group into four areas:

Facility identification — manufacturer, model numbers, and software/firmware versions actually in service.
Volt/VAR control models — inverter electronic controls, the plant controller, and any supplemental reactive resources such as STATCOMs, capacitor banks or reactors.
Frequency and active power control models — active power controls, ramp rates, and frequency response, with model validation.
Protective and limiting functions — enabled limiters (power, current, ramp rate) and protections (voltage, frequency) that directly trip or block units or the facility.

Item 4 is where model quality most often falls down, and it is the item with the most direct reliability consequence.

For inverter-based resources the EMT model is no longer an optional deliverable for unusual projects. It is a requirement under MOD-026-2 R3, and it exists because positive-sequence simulation cannot represent the behavior that actually determines whether an IBR rides through a nearby fault.

Phenomena that only an EMT model captures:

Fast inner-loop current controller behavior during and immediately after a fault
Current limiting and priority logic — active versus reactive priority under saturation
Phase-locked loop behavior under weak grid conditions and low short-circuit ratio
Unbalanced fault response and negative-sequence current injection
Momentary cessation entry and exit, and the recovery ramp
Sub-synchronous control interaction and control instability
Harmonic and switching behavior where relevant
Protection and limiter interaction on the timescale where it actually happens

What EMT verification involves

Confirming the model corresponds to the firmware and control configuration actually in service, not a generic release
Verifying site-specific parameters against the same settings sources used for the positive-sequence model
Confirming that enabled protections and limiters are represented and active in the model — a very common gap is a model shipped with protections disabled for numeric al convenience
Large-signal disturbance tests: balanced and unbalanced faults at the point of interconnection , at a range of fault impedances and durations
Voltage and frequency step response, and ride-through profile testing
Initialization, time-step, and numerical stability checks
Cross-checking EMT against positive-sequence behavior for the same disturbance

That final cross-check is worth its own line item. When the two models disagree, one of them is wrong, and finding out which during a Planning Coordinator review is considerably more expensive than finding out during your own verification.

The traceability chain

The output of a good verification program is a chain that an auditor can walk in either direction:

Field and OEM settings → PSCAD/EMT model → PSS®E positive-sequence model → simulated plant performance → recorded plant performance

Every link needs a document behind it. Where a link cannot be closed — usually because an OEM will not release a proprietary control parameter — the correct response is to document the limitation explicitly, not to quietly assume the model is right.

keentelengineering

Test Energy Before COD in WECC: CAISO BESS Guide

Part 2 — Frequently Asked Questions: Large Load Interconnection

1. Executive Summary

5.1 Separating Wholesale Charging From Retail Station Load

5.2 Telemetry Content and Cadence

Test Energy Before COD in the WECC Footprint

3. What Pre-COD Test Energy Actually Is

2.1 The Commercial Clock

2.2 The Reliability Clock

4.1 The Gating Structure

4.2 Trial Operations and Live Data Validation

7. The Model Stack for a BESS

Notice and Disclaimer

2. Two Clocks, One Project

5. Metering and Telemetry for a Storage Resource

6. The WECC Modeling Chain

6.1 The Modeling Rules That Trip Projects

10. Ride-Through and Disturbance Monitoring

9. The Registration Clock

10.1 The Three-Part Structure

10.2 The COD Interaction That Matters Right Now

4. The CAISO New Resource Implementation Path

11. Three Thresholds That Are Not the Same

8. EMT Is a Separate and Now Mandatory Obligation

12. As-Built Verification and Material Modification

12.1 Two Different Sixty-Day Windows

12.2 The Material Modification Question

12.3 Commissioning Drift

13. The Pre-COD Test Program

14. Counting Backward From COD

15. What Actually Goes Wrong

16. Keentel Interconnection and Compliance Services

16.1 Interconnection Engineering and Studies

16.2 Modeling and Model Validation

16.3 NERC Compliance Support

16.4 Owner’s Engineer and Commissioning Support

17. Frequently Asked Questions

L1-L5 Data Center Commissioning | Testing Guide

L1 Through L5 Commissioning in a Data Center

Part 2 — Frequently Asked Questions: Large Load Interconnection

The framing that matters

1. Executive Summary

6.1 Protective Device Testing and Settings

6.2 Equipment Start-Up

﻿

4. L1 — Factory and Component Verification

2. Three Numbering Systems That Are Not the Same Thing

3.1 What Level 0 Actually Reviews

4.1 Electrical Equipment

4.2 Mechanical Equipment

4.3 What L1 Produces

5.1 Electrical

5.2 Mechanical and Structural

8. L5 — Integrated Systems Testing

6. L3 — Pre-Functional and Start-Up Testing

7. L4 — Functional Performance Testing

7.1 Electrical Systems

7.2 Mechanical Systems

7.3 Redundancy Demonstration

8.1 The Load Question

8.2 The Scenario Matrix

8.3 Instrumentation and Acceptance

9. Where the L4 / L5 Boundary Actually Falls

11. What Commissioning Actually Finds

10. Designing for Testability

13. Reading the L1–L5 Chart Correctly

12. Turnover, Training, and Ongoing Commissioning

Notice and Disclaimer

5. L2 — Installation Verification

Clarification 1 — The levels are a convention, not a standard

Clarification 2 — Level 0 belongs on the chart

Clarification 3 — The levels are per system, not per facility

Clarification 4 — The redundancy tests at L4 and L5 are different tests

Clarification 5 — There is a level after L5

﻿

14. Keentel Data Center Engineering Services

14.1 Power Delivery and Interconnection

14.2 Facility Electrical Design

14.3 Power System Studies