Reactive Power and Voltage Control Study for a Transmission-Connected Network

1. Executive Summary

Keentel Engineering Solutions was retained by a transmission-owning utility in the continental United States to determine why a defined sub-region of its 230 kV and 138 kV network had become progressively harder to hold at scheduled voltage, whether the condition constituted a voltage stability problem under NERC TPL-001 performance requirements, and what combination of static and dynamic reactive resources would resolve it at least cost. The engagement ran from Month 0 to Month 9 of a single study cycle. The sub-region — referred to throughout as the load pocket — is a 1,180 MW summer-peaking area fed by two 230/138 kV autotransformer banks and a 138 kV subtransmission network. Two synchronous steam units of 215 MW each, historically the reactive anchor of the pocket, had been retired. Their replacement energy came from inverter-based resources: a 250 MW solar plant, a 150 MW battery energy storage facility and a 180 MW wind plant, all located at the electrical periphery. Load had grown 14 % over the preceding planning horizon, and the motor share of that load had grown faster than the load itself. The study confirmed a genuine and progressive voltage stability deficiency, not a modeling artifact. Under the worst applicable single-contingency event, the QV reactive margin at the critical 138 kV bus was 118 MVAr against the Client's 300 MVAr criterion, and the PV real power margin was 6.4 % against a 12 % criterion. Post-contingency voltage settled at 0.87 pu, below the 0.95 pu recovery requirement, and transient simulation with a composite load model reproduced a fault-induced delayed voltage recovery plateau lasting 2.4 seconds — a duration the Client's own disturbance records independently corroborated. Root cause was not a shortage of installed MVAr. It was a shortage of the right kind of MVAr in the right place: the retired machines had supplied fast, voltage-independent, field-forced reactive power at the electrical center of the pocket, while the replacement capability sat behind long 138 kV circuits, was bounded by converter current rating, and in part fell away exactly when it was most needed. Keentel recommended a ±150 MVAr STATCOM at the critical bus, supported by two 50 MVAr mechanically switched capacitor banks under the STATCOM's coordinated control, conversion of one existing 25 MVAr bank to a C-type filter, and retuning of two plant-level voltage controllers. Reactive margin rose to 340 MVAr, post-contingency voltage to 0.952 pu, and the FIDVR plateau fell to 0.42 seconds. The package, at $62–74 million, deferred a 42-mile 230 kV line addition of $210–260 million order-of-magnitude cost by at least ten years. Figures presented are representative of the delivered study and have been generalized to protect client confidentiality.

1.1 Study at a Glance

2. Background and Study Drivers

The load pocket had been an unremarkable part of the Client's system for most of its operating life. Two synchronous steam units sat at a 138 kV station near its electrical center, carrying 430 MW of local generation and, more importantly for this study, a continuous overexcited reactive capability of approximately 266 MVAr with short-term field-forcing capability to roughly 400 MVAr combined for tens of seconds. Both were on automatic voltage regulator control against a defined schedule, and their reactive output moved instantaneously and without operator action whenever the pocket needed support. Three changes eroded that position over roughly a decade. The first was retirement. Both units were withdrawn under economic and environmental pressure. Generation planning replaced their energy contribution adequately, but the reactive and short-circuit contributions were not replaced in kind, and no process in the utility formally owned that gap. The second was displacement by inverter-based resources. The solar, storage and wind facilities that replaced the retired energy interconnect at three separate points, none at the retired station, all electrically remote from the load center. Each complies with FERC Order No. 827 and can hold power factor within the 0.95 leading to 0.95 lagging range at its point of interconnection. That is a real capability, but it is not equivalent to what was lost, for reasons the study set out to quantify. The third was load growth and composition change. Peak demand grew from approximately 1,035 MW to 1,180 MW, with a further 160 MW forecast within the planning horizon. More significantly, the fraction represented by single-phase residential air conditioning compressors rose materially. That load class is the principal driver of fault-induced delayed voltage recovery, and its growth converts a marginal reactive position into a dynamic performance failure. The immediate commercial driver was a proposed 230 kV line addition of approximately 42 miles with associated terminal work, carried in the Client's transmission plan at $210–260 million and a six-to-eight-year development and construction schedule. Internal planning had identified the line as the response to declining pocket voltages. Before committing to that expenditure and to the siting exposure of a new right of way, the Client commissioned an independent study to establish whether the underlying problem was a real power transfer limitation — which a line solves — or a reactive and dynamic support limitation — which a line solves expensively and partially. The answer determined a nine-figure capital decision. A secondary driver was compliance posture. The Client had begun to see marginal post-contingency voltage performance in its own TPL-001 assessments, and a small number of actual disturbances had produced customer reports of momentary dimming and process interruption consistent with delayed voltage recovery. The Client wanted a defensible technical basis for whatever position it took, and a methodology it could repeat without external support.

3. Study Objectives and Scope of Work

1. Establish the present and forecast reactive supply and demand balance of the load pocket, at generator terminals and at each point of interconnection, accounting for reactive losses within collector systems and generator step-up transformers.
2. Determine whether the pocket satisfies the Client's steady-state voltage schedule and ANSI C84.1 range requirements under all applicable system conditions.
3. Quantify voltage stability margin by PV and QV analysis under NERC TPL-001 planning events, identify the critical buses, and identify by modal analysis the buses and branches that drive the weak mode.
4. Determine the sensitivity of the computed margin to load modeling assumptions, including ZIP composition, motor fraction and induction motor stalling behavior.
5. Characterize fault-induced delayed voltage recovery using a composite load model, including the effect of distributed generation tripping.
6. Quantify the loss of system strength associated with synchronous retirement and evaluate the short-circuit ratio at each inverter-based resource point of interconnection.
7. Develop and compare mitigation options spanning static and dynamic reactive compensation, and recommend a preferred package with sizing, location and control philosophy.
8. Verify the recommended package by electromagnetic transient simulation for control interaction, and by harmonic and switching transient analysis for capacitor placement and energization.
9. Establish whether the proposed 230 kV line addition could be deferred or avoided, and for how long.
10. Deliver a repeatable reactive planning methodology, documented and transferable to the Client's planning staff.

3.1 Scope Boundaries and Deliverables

Distribution-level analysis was excluded by agreement, but the study did model the aggregate behavior of distribution-connected resources and load at the 138 kV interface, because FIDVR cannot be studied honestly without doing so. The boundary was drawn at representation rather than at design: Keentel represented what the distribution system does to the transmission system, but designed no distribution-level remedy.

4. Regulatory and Standards Basis

The Client is a Transmission Planner and Transmission Owner registered with the applicable Regional Entity, and its planning assessments are performed under NERC TPL-001. Reactive capability of interconnected resources is governed by interconnection agreements executed under FERC Order No. 2003 and, for the non-synchronous resources, by FERC Order No. 827, which requires newly interconnecting non-synchronous generators to provide reactive power within a power factor range of 0.95 leading to 0.95 lagging measured at the high side of the generator substation. The inverter-based resource landscape was material. FERC Order No. 901 directed NERC to develop reliability standards addressing IBR performance, data sharing, model validation, and planning and operating studies. The resulting NERC PRC-029-1, addressing frequency and voltage ride-through for inverter-based resources, was adopted by the NERC Board in October 2024 and filed with FERC in November 2024, becoming effective in 2026. It applies to BES IBRs and to non-BES IBRs of 20 MVA or greater aggregate capacity connected at 60 kV or above, with design compliance for non-BES resources required by January 1, 2027 or the effective date, whichever is later. Its requirements address voltage ride-through, reactive current and control behavior during and after disturbances, frequency ride-through, and documentation of hardware limitations for legacy units. Momentary cessation is effectively disallowed within the ride-through envelope, and resources are expected to return to pre-disturbance current exchange within approximately five cycles of voltage recovery. That matters directly here, because the difference between an IBR fleet that ceases current injection during a depressed-voltage plateau and one that continues to inject reactive current is, in this pocket, the difference between a 2.4 second recovery and a sub-second one. The study modeled as-installed legacy behavior and expected post-PRC-029-1 behavior separately. PRC-028-1 requires disturbance monitoring on IBRs and supplies the future evidence base for validating the study's dynamic conclusions. PRC-024-4 now applies to synchronous generators, synchronous condensers, and Type 1 and Type 2 wind resources only; PRC-029-1 supersedes it for inverter-based resources. IEEE Std 2800-2022 defines the underlying performance expectations the PRC standards make enforceable.

ANSI C84.1 Range A was applied as the normal operating band, 0.95 to 1.05 pu at 138 kV and 230 kV, and Range B as the limited-duration post-contingency band. Range B excursions were permitted only for the duration required to restore a Range A condition; any case relying on sustained Range B operation was treated as a criteria violation, not an acceptable outcome.

5. System Modeling and Data Development

5.1 Data Sources

The study began from the Client's approved planning base cases developed under its MOD-032 data collection process. Those cases represent the interconnected system at a detail appropriate for regional planning but not for the localized dynamic and harmonic questions this study needed to answer. Data development therefore proceeded in two directions at once: enriching the pocket, and reducing everything outside it.

5.2 Data Gaps and How They Were Closed

Four gaps were material. Load composition. The Client had feeder-level metering and a load research program but no mapping from that data to composite load model parameters. Keentel classified each 138/12.47 kV and 138/24.9 kV substation transformer by dominant customer class, assigned a base composition from published industry work for that class, then adjusted the aggregate until simulated summer peak reactive demand at the 138 kV interface matched metered reactive demand within 4 %. The resulting composition assigned 32 % of pocket load to single-phase air conditioner compressor motors, with the balance split among three-phase motor classes, power electronic load and static load. IBR plant model quality. Two of three plants supplied user-defined models with adequate documentation. The third supplied a generic model whose plant controller parameters did not match commissioning records; Keentel used the generic model with parameters derived from the plant's own commissioning test data and flagged the residual uncertainty. This is a common US planning condition and is what MOD-032 and MOD-033 exist to drive out over time. Distributed generation. Approximately 140 MW of distribution-connected photovoltaic capacity existed in the pocket. Interconnection records allowed classification into legacy units commissioned under IEEE Std 1547-2003 practice, which trip promptly on voltage excursion, and newer units under IEEE Std 1547-2018 with Category II or III ride-through — approximately 55 MW legacy and 85 MW modern. Both were modeled explicitly, because their behavior during a depressed-voltage plateau differs in a way that changes the answer. Reactive capability actually available at each POI, as distinct from that stated in the interconnection agreement. This was resolved by explicit modeling of the collector system, pad-mount transformers, generator step-up transformer and inverter capability curves, and is treated in Section 7.1.

5.3 Model Reduction and Equivalencing

The pocket and two electrical layers beyond it were retained in full: all 230 kV and 138 kV buses, all generation, all shunt devices, all transformers with tap controls, and explicit load representation at each step-down transformer low side. Beyond that boundary the external system was reduced to an equivalent preserving positive-sequence driving-point impedance and the negative- and zero-sequence characteristics at each boundary bus, so that fault duty and unbalanced fault behavior were preserved along with power flow behavior. Equivalencing was validated three ways: power flow comparison of reduced and full cases across all contingencies, to 0.002 pu on retained voltages and 1 % on retained flows; fault duty comparison at every retained bus to 2 %; and transient stability comparison for boundary faults, confirming that swing behavior and voltage recovery in the retained area were unaltered. Two boundary buses failed the fault duty test on the first attempt and were pushed one layer further out before the equivalent was recomputed.

5.4 Model Validation and Benchmarking

For six historical snapshots spanning summer peak, shoulder and light load, the steady-state model reproduced 230 kV voltages within 0.006 pu and 138 kV voltages within 0.011 pu of SCADA values with recorded tap positions and capacitor status, and matched autotransformer reactive flow within 6 MVAr. The short-circuit model agreed with the Client's fault study at fifteen retained buses within 3 % on three-phase and 4 % on single-line-to-ground duty. The dynamic model was benchmarked against three recorded disturbances. The most valuable was a normally cleared single-line-to-ground fault on a 138 kV circuit during a high-temperature afternoon, which produced a recorded voltage plateau at the critical bus of approximately 1.9 seconds. The initial model, using the Client's standard static load representation, showed no plateau at all — it predicted recovery in under 200 milliseconds. With the composite load model and the developed composition, the model reproduced a 1.8 second plateau with plateau voltage within 0.02 pu of the recorded value. That single benchmark did more to establish the credibility of the study's conclusions than any other element of the work, and it is why the central recommendation survived internal challenge.

6. Study Methodology and Assumptions

6.1 Analytical Approach

The study was structured so that each stage constrained the next. Steady-state reactive accounting established what capability existed and where. Power flow and contingency analysis established compliance with voltage criteria. PV and QV analysis quantified distance from collapse and identified critical buses. Modal analysis explained why those buses were critical and which branches drove the weak mode. Transient stability with composite load models established dynamic performance and characterized the FIDVR behavior. Short-circuit analysis quantified system strength. Only then were mitigation options developed, and only after selection in positive-sequence tools was an option verified in the electromagnetic transient domain and subjected to harmonic and switching studies. The ordering matters. It is entirely possible to size a dynamic compensator from a QV curve alone and produce a device that satisfies the steady-state criterion while failing the dynamic one, or that satisfies both while destabilizing a nearby plant controller in a low-short-circuit-ratio area. No conclusion was allowed to depend on a single tool.

6.2 Tools and Selection Rationale

Using the Client's own platform for the core work was deliberate: a study whose conclusions the Client cannot reproduce in its own tools has limited durability. Every PV and QV case, every contingency and every dynamic run was delivered as a scripted, re-runnable artifact.

6.3 Cases and Scenario Matrix

Contingencies comprised the full TPL-001 planning event set applicable to the pocket: P0 all elements in service; P1 loss of any single generator, line, transformer or shunt device; P2 non-fault single element loss including a bus section; P4 stuck breaker events at the two 230/138 kV stations; P6 overlapping loss of two single elements with system adjustment between; and P7 loss of both circuits of the double-circuit 230 kV structure entering the pocket. In total 214 contingencies were evaluated per case, with full PV and QV treatment applied to the 18 producing the lowest margins.

6.4 Acceptance Criteria

The 300 MVAr QV criterion is not a universal number and deserves explanation. The Client's long-standing practice sets required reactive margin at each critical bus equal to 1.5 times the short-term reactive output of the largest single dynamic reactive source historically relied upon in the area. With the retired units each able to supply approximately 200 MVAr under short-term field forcing, that yields 300 MVAr. Keentel reviewed the basis, found it defensible, and applied it as given rather than substituting a different convention. Where a criterion is the Client's, the study says so.

6.5 Assumptions and Limitations

The study assumed the Client's load forecast correct within its stated confidence band and tested the consequence of it being wrong by sensitivity rather than by challenging it. It assumed existing capacitor banks remain available at nameplate, and that IBR plant controllers behave as their models represent, subject to the model quality caveat above; the consequence of that assumption failing was tested explicitly. The study could not represent every distribution feeder individually. The composite load model, however carefully parameterized, is an aggregate: it reproduces observed aggregate behavior well, as benchmarking demonstrated, but cannot predict the behavior of any individual feeder. It also does not represent thermostatic recovery of air conditioning load over the minutes following a disturbance, which affects restoration but not the performance question the study was asked to answer.

7. Analysis and Results

7.1 The Physical Basis: Why Reactive Power Is a Local Resource

Every conclusion in this study rests on one relationship. Across a series impedance of resistance R and reactance X carrying real power P and reactive power Q, the approximate voltage drop between sending and receiving ends is ΔV ≈ (P·R + Q·X) / V. On a transmission network X greatly exceeds R — for the 138 kV circuits in this pocket the X/R ratio ranges from 4.6 to 9.2, and for the 230 kV circuits from 8.1 to 13.4 — so the QX term dominates. Voltage magnitude is controlled primarily by reactive flow, and the same reactance that makes reactive flow effective at moving voltage also makes reactive power expensive to transport. Carrying Q across a reactance consumes I²X in the reactance itself, and that consumption rises with the square of current. A reactive source three circuits away must produce substantially more than arrives, and the deficit grows precisely as the system becomes stressed and current rises. In the limit, supplying a reactive deficit from a distant source produces less support at the deficient bus than it produces depression along the path. That is what is meant by saying reactive power does not travel. Reactive flow over distance plainly occurs; what falls away rapidly is the marginal effectiveness of remote support, and it falls fastest when support is most needed. A MVAr inventory is therefore meaningless without a location. This pocket did not have a MVAr shortage on paper; it had one at the buses that mattered. The second distinction is between reactive sources. A synchronous machine supplies reactive power by field excitation. Its steady-state limits are set by rotor field heating when overexcited, stator end-region heating when underexcited, and armature current in between. Critically, when terminal voltage falls the excitation system increases field current, and the machine can be driven well beyond continuous rating for tens of seconds before the overexcitation limiter acts, in coordination with generator protection per the intent of NERC PRC-019. Its reactive output therefore rises as voltage falls — exactly the behavior a voltage-stressed network requires. It also supplies fault current of five to seven times rated, plus inertia, and sets a strong voltage reference for every converter connected near it. An inverter-based resource supplies reactive power by controlling converter current phase. Its response is fast — one to three cycles at current level, though plant-level controllers are much slower — and symmetrical between absorption and injection. But it is bounded by converter current rating, so available reactive power is limited by real power output unless the converter is oversized: Q ≤ √(S² − P²). A plant rated 250 MW delivering 250 MW from converters rated 250 MVA has no reactive capability at all. Furthermore, when voltage falls, an inverter at its current limit produces less reactive power, not more, because power is the product of current and voltage and current cannot rise. Where the synchronous machine leans into a voltage depression, the inverter, absent explicit oversizing and control design, leans away from it. The third distinction is the accounting between generator terminals and the point of interconnection. FERC Order No. 827 obliges non-synchronous resources to provide reactive capability within a 0.95 leading to 0.95 lagging range at the high side of the generator substation, and the reactive losses between the inverter terminals and that point are consumed out of the inverter's capability.

The plant's installed inverter rating was 262 MVA, leaving 79 MVAr at full real output — short of the 117 MVAr required to meet the obligation at the POI. The gap was closed in practice by plant-level capacitor banks and by curtailing real power in the rare conditions where full reactive output was called for at full real output. That is a legitimate configuration, but it means the capability shown in the interconnection agreement is not continuously available, and a planning study that assumes otherwise overstates the pocket's reactive resources. At night or low irradiance the same plant has ample reactive capability, and collector charging contributes a further 14 MVAr of capacitive support — which mattered for the light-load overvoltage case.

7.2 Reactive Supply Inventory and Steady-State Performance

The table states the problem compactly. On a nameplate basis the pocket had 372 MVAr of controllable capability without the retired units against 266 MVAr continuous with them, and could appear better off. Under the conditions that matter — depressed voltage, low IBR output, post-contingency — the picture inverts. The capacitors deliver 81 % of nameplate at 0.90 pu; the IBR contributions are simultaneously reduced by current limits and located behind 26 to 41 miles of 138 kV circuit; and the retired units would have delivered more than nameplate, at the electrical center, without waiting for a plant controller.

Failures concentrate in the low-IBR-output cases, a direct consequence of reactive capability being tied to converter loading. The S4 case — summer peak plus the 5 % load stress the Client's criteria require for margin evaluation — did not converge under the worst P1. A non-converged power flow is not by itself proof of collapse, and Keentel treated it as an indicator requiring continuation power flow rather than as a result; Section 7.3 confirmed the non-convergence was physical. Voltage schedule and LTC coordination were examined in the same pass. The two 230/138 kV autotransformer banks tap automatically against a 138 kV-side schedule with a ±10 % range in 32 steps. Under stress their action is actively harmful: raising 138 kV voltage restores load power, which increases current and upstream reactive consumption, depressing the 230 kV side further. Under the S4 worst case the banks ran to within two steps of their boost limit while 138 kV voltage continued to fall — the classic tap-changer-driven progression toward instability. Line drop compensation on one bank carried settings inherited from a network configuration that no longer exists; recalculating for present impedance between the regulation point and the load center moved the effective regulated point by nearly 0.02 pu. Capacitor switching was on local voltage-and-time control with a 4-minute delay and no coordination across the five banks, with an observed tendency for two banks to close on one excursion when one was required.

7.3 PV Analysis and Continuation Power Flow

PV analysis was performed by continuation power flow, increasing pocket load along a defined stress direction at constant power factor and tracing the solution through the nose rather than relying on power flow divergence. Load was represented as constant power for the PV work, the conventional and conservative choice, with the ZIP sensitivity treated separately.

The Year 10 result ended the internal debate: at forecast load with low IBR output the pocket sits 12 MW below its own nose point under a single contingency, meaning it has no stable post-contingency operating point at all. Margin degradation is not linear — the curve steepens as the limit approaches, which is why voltage stability problems tend to arrive suddenly in operating experience after years of apparently gradual deterioration in studies. Curve shape was as informative as margin value. Under the worst P1 the curve at Bus A is flat near the nose over roughly 40 MW, so voltage gives almost no warning of approaching instability: an operator would see 0.93 pu, uncomfortable but survivable, with no indication that 40 MW of additional load separated the present state from collapse. This is the principal argument for planning to a margin criterion rather than a voltage criterion.

7.4 QV Analysis, Reactive Margin and Modal Identification

QV analysis was performed by installing a fictitious synchronous condenser at the study bus, varying its scheduled voltage, and recording the reactive injection required to hold each voltage. The minimum of the resulting curve is the reactive margin: the additional injection available before the system loses a solution. Margin was evaluated at 105 % of forecast peak load per the Client's criteria.

Bus A is the critical bus by a clear margin. The QV curves at Bus A, Bus C and Bus D are similar in shape and reach their minima at similar voltages — the signature of a single coherent weak area rather than three independent local problems. That finding governed the strategy: one well-placed device could address all three, whereas three local capacitor additions would have addressed none of them properly.

Modal analysis in DIgSILENT PowerFactory formed the reduced Jacobian at the stressed operating point and extracted its eigenvalues; the smallest identifies the critical mode, with zero corresponding to collapse. The separation between the smallest and second smallest under the P1 condition — 0.86 against 6.31 — shows a single dominant mode. Bus participation factors identify Bus A, Bus C and Bus D as the buses that most influence and are most influenced by it, together 0.69 of total participation. Branch participation factors, which identify the elements whose reactive losses most contribute, ranked the two 138 kV circuits between Bus B and Bus A highest at 0.31 combined, followed by the two autotransformer banks at 0.24. That ranking states directly that the weak mode is created by reactive consumption in the path from the 230 kV source into the pocket, not by a local deficiency at any one bus, and it is the analytical justification for placing compensation at Bus A rather than nearer the source. Generator participation factors quantified the earlier physical argument: the three IBR plants influence the critical mode at roughly 0.31 to 0.44 of the effectiveness of an equivalent injection at Bus A, before any consideration of current-limit behavior.

7.5 Load Modeling and Its Effect on the Computed Margin

The single largest source of variation in the computed margin was not any network parameter. It was the load model.

The constant impedance representation shows a healthy system with more than 480 MVAr of margin and no violations anywhere. It is also physically wrong for this pocket. Constant impedance load reduces consumption with the square of voltage, so as voltage falls the load relieves the very stress causing the fall. Real load does not behave that way in the time frame that matters: motor load is approximately constant power over its stable range, because a motor driving fixed mechanical torque draws more current as voltage falls; thermostatically controlled load recovers within minutes; power electronic load is constant power by design. An unrealistically stiff load model does not merely produce an optimistic number — it hides the problem entirely, because a study finding no violation generates no mitigation and no capital request. The Client's earlier internal assessments used a predominantly constant impedance representation, which is why the pocket showed as compliant in prior planning cycles while operating experience increasingly suggested otherwise. Resolving that contradiction was the most valuable single act of the study.

7.6 Fault-Induced Delayed Voltage Recovery

FIDVR arises from the behavior of single-phase induction motors, principally residential and small commercial air conditioning compressors. These motors have low inertia and drive a compressor load that does not unload when speed falls. When a transmission fault depresses voltage below roughly 0.6 pu for more than a few cycles, electromagnetic torque — which varies with the square of voltage — falls below load torque and the motor decelerates. Below a critical slip it stalls. A stalled single-phase motor draws roughly five to six times rated current at very poor power factor: it becomes a large uncontrolled reactive load switched on at the worst possible moment. The consequence is that even after the fault is normally cleared and the network is intact, the aggregate stalled population holds local voltage on a plateau, typically between 0.6 and 0.8 pu. Recovery waits on thermal protection: each stalled compressor's overload device eventually disconnects it, and voltage recovers as the stalled population is shed. Because those devices have a wide statistical spread in operating time, recovery is gradual. Plateau duration is therefore set by protection statistics rather than network dynamics, which is why it cannot be shortened by faster fault clearing once the stall has occurred. The composite load model represents this explicitly, with an aggregated single-phase air conditioner component having defined stall voltage, stall time, thermal protection characteristic and restart behavior, alongside three-phase motor components, an electronic load component and a static component, all behind a distribution equivalent impedance and an aggregated distribution transformer with tap.

Two rows deserve comment. The legacy-trip versus full ride-through comparison shows a 0.6 second difference in plateau duration and a material difference in plateau voltage: the 55 MW of legacy distribution photovoltaic capacity trips within the first few hundred milliseconds of the plateau, removing real power injection and, in effect, increasing the load the transmission system must support during it. This is precisely the behavior IEEE Std 1547-2018 ride-through categories and, at transmission scale, NERC PRC-029-1 are designed to eliminate. The study reported the dependence explicitly and noted that the improvement is not available to the planner as a mitigation, because it depends on equipment turnover outside the Client's control. The comparison between the STATCOM and an equal MVAr rating of switched capacitors is the technical heart of the recommendation. The capacitor bank contributes 150 MVAr at 1.0 pu but only 93 MVAr at the 0.79 pu plateau voltage, arriving after a mechanical switching delay of three to six cycles plus control decision time, at a moment when the network needs reactive current immediately and at depressed voltage. The STATCOM produces rated reactive current within about two cycles and holds that current as voltage falls, so its reactive power falls only linearly with voltage — 136 MVAr at 0.91 pu — and its response arrives while motors are decelerating but before most have stalled. Preventing the stall is dramatically more effective than remedying it, because once a substantial population has stalled the reactive requirement roughly triples and no economically sized device can hold voltage up.

7.7 System Strength and Short-Circuit Ratio

Retiring synchronous machines removes short-circuit current as well as reactive capability, and the two losses compound. Weak systems exhibit greater voltage change for a given reactive disturbance — the same statement as saying the QX term is larger — and they destabilize converter control loops tuned for stiffer conditions.

Fault duty at Bus A fell by 40 % on retirement. The short-circuit ratio at the solar plant POI fell to 2.3 and the weighted short-circuit ratio across the pocket's 640 MVA of aggregate IBR capacity to 1.9. Values below 3 are conventionally treated as weak and below 2 as very weak, requiring EMT-domain verification of converter control stability rather than reliance on positive-sequence models. The post-mitigation column carries a caveat the study made prominently. A STATCOM contributes fault current of approximately its rated current — 150 to 165 MVA equivalent here — with a controlled, current-limited characteristic rather than the sustained, angle-referenced contribution of a synchronous machine. Counting it as strength is defensible for converter control stability, a small-signal impedance question, and much less defensible for protection coordination, where the question is whether a relay sees enough current for long enough. Keentel reported the STATCOM contribution separately in every table and recommended that protection studies use the synchronous-only figure. The improvement from 1.9 to 2.4 weighted SCR is real but modest and does not restore the pre-retirement condition; a synchronous condenser would have, and that trade-off is treated in Section 10.

7.8 Harmonic and Resonance Assessment for Capacitor Placement

A shunt capacitor bank forms a parallel resonance with source inductance at approximately h = √(S\_sc / Q\_c). Reducing short-circuit level moves that resonance downward toward the low-order characteristic harmonics that IBR plants and distribution load actually produce. Synchronous retirement therefore creates a harmonic problem out of a capacitor installation that was previously benign — a coupling frequently missed because harmonic assessment and retirement studies live in different groups.

The pre-existing three-bank configuration at Bus C had shifted to a parallel resonance at order 4.97 — effectively coincident with the fifth harmonic, the dominant characteristic harmonic on the system. Frequency scans confirmed fifth-harmonic impedance magnification of 11.4 times, and measured 138 kV voltage total harmonic distortion at Bus C of 3.9 % with all three banks in service against the IEEE Std 519-2014 limit of 2.5 % for systems above 69 kV through 161 kV, with individual harmonic voltage distortion of 2.9 % against the 1.5 % limit. This was an existing, unrecognized non-compliance rather than a consequence of the proposed work, and the Client was notified as a discrete finding on identification rather than at report issue. Converting one bank to a 25 MVAr C-type filter tuned near order 4.5, with damping sized for a quality factor of 2.5 at the tuning point, moved the parallel resonance to 3.61 and damped it, reducing fifth-harmonic magnification to 2.3 times and computed voltage THD to 1.8 % with individual distortion of 1.1 %. The C-type topology was selected over a simple detuned bank because it provides damping with low fundamental-frequency loss in the damping resistor, which matters for a continuously connected transmission-class installation. At Bus A the two proposed 50 MVAr banks resonate at order 7.82 individually and 5.53 together; the latter sits between the fifth and seventh harmonics with adequate separation, and the frequency scan confirmed no magnification above 3.1 times at any characteristic harmonic. Because the banks switch under the STATCOM's coordinated control, the study specified that the control logic must never permit a bank combination producing a resonance within 0.3 harmonic orders of the fifth, seventh or eleventh — trivially satisfied here, but binding if a future bank is added at the same bus.

7.9 Switching Transient Analysis

Capacitor energization was studied in PSCAD/EMTDC with a 5 microsecond time step, explicit representation of the bank, switching device, bus geometry and local cable and bus inductance, and statistical switching over 200 shots per configuration.

Back-to-back energization — closing the second 50 MVAr bank while the first is energized — is the governing duty, because the inrush loop is formed by the two banks and the small inductance between them rather than by the source. The computed 18.7 kA at 5.1 kHz exceeds the definite-purpose capacitor switching capability of a standard 138 kV breaker under IEEE C37.06, which is bounded in both peak current and frequency; exceeding the frequency limit is as significant as exceeding the peak, because interrupter duty relates to the rate of rise of current at the contacts. Adding 200 microhenry current-limiting reactors in each bank connection brought the duty to 8.9 kA at 1.9 kHz, within capability with margin. Transient overvoltage at 138 kV reached 1.62 pu on the worst statistical shot, within arrester capability but an unnecessary stress on a device operating several times daily; controlled point-on-wave closing at voltage zero with ±1 millisecond scatter reduced it to 1.13 pu. The larger benefit is at distribution level: energizing a transmission capacitor can excite a series resonance formed by the step-down transformer and a distribution capacitor bank. The computed 2.41 pu at a 12.47 kV bus with 4.8 MVAr of distribution capacitors would have caused nuisance tripping and arrester duty; controlled closing reduced it to 1.48 pu. Restrike was addressed by specification rather than mitigation. A restrike on de-energization can produce overvoltage above 2.7 pu by trapping charge and re-establishing the arc at peak recovery voltage, and no economically reasonable arrester scheme makes a restriking device acceptable for daily switching. The study specified a definite-purpose, restrike-free-rated device with controlled opening and periodic verification of contact condition, since restrike probability rises with wear.

7.10 Electromagnetic Transient Verification of the Dynamic Device

The final step was verification in PSCAD/EMTDC that the recommended STATCOM would remain stable at a short-circuit ratio of 2.3 to 2.5 in the presence of three nearby plant-level controllers, with no adverse control interaction. This is not optional in a weak system: positive-sequence tools represent converter controls only within the electromechanical bandwidth, and control interaction occurs above it. The scan swept STATCOM voltage droop from 1 % to 6 %, plant controller gains across their commissioned ranges, and network strength from SCR 1.5 to 4.0, running frequency-domain impedance scans and time-domain disturbance tests at each combination. A growing oscillation near 14 Hz appeared whenever STATCOM droop was below 2 % and local SCR below 2.0, with the mode involving both the STATCOM voltage regulator and the solar plant's plant-level controller; damping ratio at the worst combination was −0.02, that is, marginally unstable. Increasing droop to 4 % and slowing the plant controller integral time from 0.8 to 2.0 seconds moved damping to +0.11 across the entire swept envelope, verified at the extremes rather than only at the expected operating point. A second finding concerned coordination rather than stability. The proposed logic for switching the two 50 MVAr banks used a simple threshold on sustained STATCOM output. Under a fault-and-recovery sequence it switched a bank in during the recovery transient and out again 40 seconds later, producing an avoidable second voltage step. A 15 second inhibit following any excursion below 0.9 pu, plus a switch-out deadband wider than the bank step size, eliminated the hunting across all tested sequences.

8. Sensitivity and Scenario Analysis

Four results were decision-relevant. The load model dominates everything. Spread attributable to load representation exceeds spread attributable to all network and dispatch variables combined. That is why the study invested disproportionate effort in load composition development and disturbance benchmarking, and why the delivered methodology places load model development ahead of case building rather than treating it as a parameter choice at the end. Low IBR output governs, not high. Planners accustomed to thermal analysis instinctively stress the system with high generation output; for reactive adequacy in an IBR-served pocket the binding condition is the opposite, because reactive capability is tied to converter availability and the plants are not obligated to be online. A summer peak evening after solar output has collapsed but before load has, with wind low, is the governing condition, and it occurs routinely. The two queued IBR additions changed margin by 8 MVAr. That mattered commercially, because one internal position had been that the queue would resolve the reactive problem without capital investment. It does not, for the same locational reason the existing IBRs do not. The STATCOM rating sensitivity set the sizing. A ±100 MVAr device reaches 291 MVAr, just below criterion, with no allowance for forecast error. A ±200 MVAr device reaches 384 MVAr at roughly 27 % higher installed cost with no additional benefit against any criterion. The ±150 MVAr device reaches 340 MVAr, clears the criterion by 13 %, and retains headroom for about 90 MW of growth beyond the Year 10 forecast. Sizing was selected on the margin curve, not on the reactive deficit: the deficit at Bus A under the worst P1 is approximately 182 MVAr, but a device at Bus A also reduces reactive losses along the supply path, so its effect on margin exceeds its rating. Two sensitivities were tested and found not to matter, which is worth recording. Ambient variation within the summer design range altered margin by less than 5 MVAr. Re-deriving the external equivalent from a different seasonal base case altered margin by 7 MVAr, confirming that the pocket's problem is internal and not an artifact of the boundary.

9. Findings and Root Cause Assessment

The proximate cause of F1 through F4 is the removal of 266 MVAr of continuous and approximately 400 MVAr of short-term dynamic reactive capability from the electrical center of the pocket without locational replacement. The replacement capability is nominally comparable but sits at the periphery, is bounded by converter current rating rather than field current, falls with voltage rather than rising, responds on plant-controller timescales of seconds rather than excitation timescales of cycles, and is unavailable when the resource is not generating. Load growth and the growth of stalling-prone motor load converted a reduced margin into a criteria failure, and would eventually have converted a criteria failure into an operating event. The underlying cause is a process gap, and the study said so directly. The Client's generation retirement review examined energy adequacy, capacity adequacy and thermal transfer capability. It did not examine reactive adequacy, dynamic reactive adequacy or system strength, because no procedure assigned those questions to any group. The interconnection studies for the replacing IBRs each examined the local impact of that project and each correctly found no adverse impact, because no single project caused the deficiency. The deficiency is the cumulative consequence of a sequence of individually acceptable decisions, and nothing in the process was watching the cumulative position. F7 compounded it: the load model in use made the accumulating deficiency invisible in the very assessments that should have detected it. F5 is independent of the main narrative but was elevated because it represents present non-compliance with IEEE Std 519-2014 at an existing installation, arising from the same retirement through a mechanism nobody had connected.

10. Mitigation Options and Recommendations

Six technology families were evaluated, each sized to deliver comparable steady-state reactive support at Bus A so the comparison isolated the characteristics that differ.

The comparison turns on four properties. Speed. Mechanically switched devices act in three to six cycles once commanded, and the command follows a control decision deliberately delayed to avoid hunting. Motors stall within roughly 100 to 200 milliseconds of fault initiation, so a mechanically switched device is remedying a condition rather than preventing one, and the required MVAr after stalling is roughly three times that required before it. Voltage dependence. A capacitor's output falls with the square of voltage, delivering 81 % of nameplate at 0.90 pu and 64 % at 0.80 pu — least effective exactly when most needed. An SVC in its capacitive range shares this, because its thyristor-switched branches are capacitors. A STATCOM regulates current, so its reactive power falls linearly with voltage, delivering 90 % at 0.90 pu and 80 % at 0.80 pu, and most designs support a short-term overload. A synchronous condenser can be driven above rating by field forcing for tens of seconds and can therefore maintain or exceed rated output into a depressed voltage. System strength. Only the synchronous condenser materially restores what retirement removed: sustained fault current of roughly five times rating, inertia, and an angle-referenced voltage source that stabilizes nearby converters without depending on their own control loops. The STATCOM contributes approximately its rated current, useful for converter stability but not for protection sensitivity. Grid-forming control on the existing storage plant provides a voltage reference behind an impedance and improves small-signal stability in a weak network, but its current contribution is bounded by converter rating and its availability by commercial dispatch. Cost, footprint and maintenance. The synchronous condenser costs roughly 1.6 to 1.7 times the STATCOM, needs a substantially larger footprint with auxiliary, lubrication, cooling and starting systems, incurs continuous no-load losses near 1 % of rating, and carries annual maintenance of $1.2–1.8 million against roughly $0.3–0.5 million for a static device. It is nonetheless the correct answer where system strength, not reactive margin, is the binding constraint.

10.1 Recommendation

Total installed cost order of magnitude for the package is $62–74 million, against $210–260 million for the 230 kV line addition it defers. The STATCOM was preferred over the synchronous condenser because the binding constraints were reactive margin and dynamic voltage recovery, both of which it satisfies at roughly 60 % of the condenser's installed cost and without rotating plant. The system strength deficiency, though real, did not itself cause a criteria violation once converter controls were retuned and verified in the EMT domain. Had EMT work found instability that retuning could not resolve, or had protection sensitivity failed at reduced fault duty, the recommendation would have changed, and the report said so explicitly with the switching point identified: a weighted SCR below approximately 1.5, or a further synchronous retirement in the adjacent area, reverses the conclusion. The STATCOM was preferred over the SVC on worst-case rather than typical performance. At operating points where both satisfy criteria the SVC is $7–10 million cheaper. At the depressed voltages that characterize the events the device exists to manage, the SVC's capacitive output falls with V² while the STATCOM's falls with V, and the margin between meeting and failing the recovery criterion was narrow enough that the difference mattered. The two 50 MVAr banks are not redundant with the STATCOM; they are what makes the ±150 MVAr rating sufficient. Steady-state support at summer peak is supplied by the capacitors, preserving the STATCOM's full dynamic range as reserve, which is also how the Client's criterion of 50 % unused dynamic capability at peak is met. Placing the banks under the STATCOM's coordinated control rather than on independent local relays makes that reserve management automatic. Grid-forming control on the existing storage plant was recommended as a complementary measure to be pursued commercially, not as a substitute. Its small-signal benefit in a weak network is real and its cost low, and it reduces sensitivity to further synchronous retirement, but it depends on a third-party asset whose availability the Client cannot guarantee at the moment of a contingency.

11. Implementation Support and Field Validation

Keentel supported procurement and commissioning under a follow-on scope, producing the technical input to the functional specification: the reactive capability curve including short-term overload, required response time to a step voltage change, droop and deadband settings, capacitor coordination logic including the resonance constraint and post-excursion inhibit, harmonic performance requirements at the point of connection, and a required EMT model deliverable with a defined validation test set. That last item deserves emphasis. The specification required a validated, non-black-box-limited PSCAD model of the supplied controls together with a factory test record demonstrating agreement between model and physical controller for a defined disturbance set. Without it, the EMT verification of Section 7.10 would have had to be repeated on a vendor-generic model of unknown fidelity, and any commissioning surprise would have been unresolvable. Validation proceeded in three stages: factory acceptance against a real-time simulation derived from the study's EMT case, including the low-SCR condition and the 14 Hz interaction scan; staged commissioning tests covering STATCOM step response, bank switching with transient recording at 138 kV and at a nearby 12.47 kV bus, and coordinated response to a schedule change; and finally a recorded system disturbance approximately five months after energization.

The recorded disturbance was a normally cleared single-line-to-ground fault on a 138 kV circuit at high ambient temperature — the same class of event as the pre-study benchmark that produced a 1.9 second plateau. The post-mitigation event produced a 0.38 second plateau. The comparison is not perfectly controlled, since the events differ in fault location, pre-fault load and IBR output, and the study reported it as corroboration rather than proof. Simulating the specific recorded event in the validated model reproduced the measured plateau within 0.05 seconds and plateau voltage within 0.015 pu, which is the more rigorous statement. Steady-state validation used SCADA and PMU data over the first summer peak season. QV behavior cannot be observed directly on an operating system, so validation of the margin claim was indirect: the model was re-benchmarked against 22 operating snapshots, matching 138 kV voltages within 0.009 pu, and margin was recomputed from the re-benchmarked model. It returned 334 MVAr against the predicted 340 MVAr.

12. Results and Value Delivered

The Year 10 row carries a deliberate qualification. With the recommended package, Year 10 forecast peak under the worst P1 yields 11.4 % margin against a 12 % criterion — a marginal shortfall at the far end of the horizon. The study neither concealed it nor sized the device upward to eliminate it, because doing so would spend capital now against a forecast whose uncertainty exceeds the shortfall. Instead the report defined a monitoring trigger: when measured pocket peak exceeds a stated threshold, or a further synchronous retirement is proposed in the adjacent area, the reactive assessment is re-run, with a staged second capacitor bank or a STATCOM overload uprate pre-identified as the incremental response. Value delivered falls into four categories. First, a defensible reactive plan supported by an auditable chain from load research through EMT verification to field measurement. Second, deferral of a $210–260 million transmission line for at least ten years at a package cost of $62–74 million, with the siting and permitting exposure of a 42-mile right of way deferred with it. Third, demonstrated compliance with the Client's planning criteria and with NERC TPL-001 performance requirements across the applicable planning event categories, plus correction of a previously unrecognized IEEE Std 519 non-compliance. Fourth, a documented, repeatable reactive planning methodology delivered with scripted cases, which the Client's own staff executed independently in the following planning cycle.

13. Lessons Learned and Engineering Insights

The load model is the study. More analytical error is available through load representation than through any other single choice in voltage stability work. A constant impedance representation of this pocket showed 486 MVAr of margin and no violations; the benchmarked composite representation showed 118 MVAr and four criteria failures. That is not a refinement, it is the difference between a compliant system and one with no stable Year 10 operating point. The only reliable defense is benchmarking against a recorded disturbance, and the corollary is that a planner without disturbance records should acquire them before acquiring more sophisticated analysis. Reactive nameplate is not reactive capability. Every reactive resource in this pocket carried a rating that overstated what it delivers under the conditions that matter. Capacitors deliver 81 % at 0.90 pu. IBR plants deliver their Order No. 827 obligation only after GSU and collector losses are covered, and only when generating. A reactive inventory that sums nameplate across a region without locational weighting, voltage dependence and availability conditioning is not an engineering document. Retirement studies must include reactive and strength assessment, and few do. The deficiency here was created by a retirement review examining energy, capacity and thermal transfer, and by interconnection studies each correctly finding no adverse project-specific impact. Nothing was done incorrectly by any individual process. The gap was that no process owned the cumulative reactive and system strength position of a defined area over time. Assigning that ownership costs nothing and would have surfaced this problem years earlier. Preventing a motor stall is an order of magnitude cheaper than remedying one. Once a substantial single-phase motor population stalls, reactive demand roughly triples and the recovery timescale is set by thermal protection statistics that no network device can accelerate. The entire economic case for a fast device over a mechanically switched one rests on acting inside the 100 to 200 millisecond window before the stall, not on steady-state MVAr. Weak systems must be verified in the EMT domain, and the model must be contractually secured. The 14 Hz interaction mode found in PSCAD was invisible in every positive-sequence case, would have appeared at commissioning as an unexplained oscillation, and would have been resolved under schedule pressure by whatever setting change stopped it. Securing a validated, unrestricted EMT model of the supplied controls before award is the difference between engineering the resolution and improvising it. Non-convergence is a symptom, not a result. The S4 case that failed to converge was genuinely unstable, and continuation power flow proved it. It could equally have been a numerical artifact. Reporting a non-converged case as a violation without continuation analysis is a common and avoidable error that undermines a study's credibility precisely when the conclusion is expensive.

14. Keentel Capability Summary

- Reactive power planning: supply and demand accounting at generator terminals and points of interconnection, including collector and GSU reactive loss accounting under FERC Order No. 827. - Voltage stability assessment: PV and QV analysis by continuation power flow, reactive margin computation, critical bus identification and margin criteria development. - Modal and small-signal analysis: reduced Jacobian eigenvalue analysis with bus, branch and generator participation factors in DIgSILENT PowerFactory. - Dynamic performance studies: transient stability with composite load models, FIDVR characterization, DER trip and ride-through representation. - Load model development: load research mapping, composite load model parameterization, and benchmarking against recorded disturbance data. - System strength assessment: short-circuit duty, SCR and weighted SCR evaluation, and synchronous retirement impact analysis. - Electromagnetic transient studies in PSCAD/EMTDC: converter control interaction scans, impedance-based stability assessment, and compensator control verification in low-SCR networks. - Power quality and switching studies: harmonic frequency scans, C-type and detuned filter design, IEEE Std 519 assessment, capacitor energization, back-to-back inrush, restrike and controlled switching specification. - Compensation technology evaluation: MSC, MSCDN, SVC, STATCOM, synchronous condenser and grid-forming inverter-based support, with sizing, siting and control philosophy. - Model development and validation: network equivalencing, steady-state and dynamic benchmarking, MOD-032 and MOD-033 data practice. - Compliance support: NERC TPL-001 planning assessment, FAC-011/014 SOL methodology input, and PRC-029-1 and IEEE Std 2800-2022 interpretation. - Owner's engineer services: functional specification input, real-time simulation and factory acceptance witnessing, commissioning test design, field validation, and methodology transfer to client planning staff.

15. Frequently Asked Questions

16. Glossary of Terms and Abbreviations

17. Confidentiality and Use Statement

This case study has been prepared for informational purposes. All client identities, project locations, contract details, and proprietary data have been withheld or generalized. Technical parameters, study results, and figures presented are representative of work performed by Keentel Engineering Solutions and have been adapted so that no individual project, owner, or facility can be identified. Nothing in this document constitutes a design recommendation for any specific installation. Any reuse of the methodologies described requires project-specific engineering analysis by a qualified professional engineer.