Electrical Design and System Studies for a Hybrid Solar and Wind Generating Facility

1. Executive Summary

The Interconnection Customer held a single 300 MW interconnection service agreement at a 230 kV point of interconnection and intended to place 180 MW of Type 4 full-converter wind generation and 220 MW of photovoltaic generation behind it. The commercial logic was straightforward: two anti-correlated resources sharing one set of interconnection facilities raise the utilization of an expensive and slow-to-obtain interconnection right. The engineering consequence was not. Two dissimilar inverter-based resources, with different ramp rates, reactive envelopes, daily availability and economic value per megawatt-hour, had to share one collection system, one set of main power transformers, one plant controller, one revenue meter and one aggregate export limit, while satisfying interconnection obligations written as though the facility were one machine.

Keentel Engineering Solutions was retained as electrical design authority and study consultant for the complete facility, from the inverter and turbine terminals through the 34.5 kV collection network and collector substation to the point of interconnection. The engagement ran from Month 0 to Month 14 and covered layout and resource interaction, collection system design and optimization, medium-voltage grounding, the study set required for interconnection review, the hybrid plant controller specification, and the compliance evidence package.

Studies comprised load flow and reactive capability at the point of interconnection, short circuit and equipment duty to IEEE C37.010, protective coordination for thirteen collector feeders, arc flash to IEEE 1584-2018 and NFPA 70E, harmonics to IEEE 519 and IEEE 2800-2022, electromagnetic transient modeling in PSCAD/EMTDC for control interaction, temporary overvoltage and energization, flicker to IEEE 1453, insulation coordination and arrester duty, grounding to IEEE 80 across a footprint exceeding twelve thousand acres, and dynamic model validation planning to NERC MOD-026 and MOD-027.

The package cleared interconnection review at first submission with no deficiency letter. The architecture selected — a single common 34.5 kV collector bus with two identical main power transformers sized to the interconnection limit rather than to the resources — removed 74 MVA of installed transformation. Enabling photovoltaic inverter reactive support at zero DC power closed a 5.2 Mvar night-time lagging reactive deficit and avoided a dedicated dynamic reactive device. Economic conductor optimization cut energy-weighted collector losses from 1.18 percent to 0.79 percent while also reducing cable capital cost. Identified capital and lifetime loss savings totalled approximately 10.2 million dollars.

1.1 Study at a Glance

2. Background and Study Drivers

The Developer controlled a contiguous land position with two partially overlapping resources: a ridge-and-plateau system suited to a modern low-specific-power turbine, and broad sloping ground suited to single-axis tracker photovoltaic. Developing them separately would have required two interconnection requests and two positions in a cluster study process whose timelines and withdrawal penalties, following FERC Order No. 2023 and Order No. 2023-A, had made duplicate speculative requests financially punishing.

A single request was filed, and after two cluster study cycles an executable agreement for 300 MW of interconnection service was received. Building 400 MW of nameplate behind a 300 MW limit was deliberate. Hourly analysis showed a correlation coefficient between the resources of minus 0.21, wind concentrated in overnight and shoulder-season hours and photovoltaic in midday summer hours. Combined output exceeded 300 MW in only 6.4 percent of hours, and curtailed energy was 1.94 percent of combined gross generation. The interconnection right, worth a 27.0 percent capacity factor as a standalone solar plant or 38.0 percent as standalone wind, delivered 41.1 percent as a hybrid.

Every shared element then had to be sized, protected, grounded, controlled and metered for a facility whose two halves would rarely be at full output together and would sometimes sit in states — full wind at night with the array dark, or full photovoltaic with the turbines becalmed — that no single-technology plant experiences.

Three drivers shaped the scope. First, interconnection review risk: the Transmission Provider's facility study had used a generic aggregated model, and the design submission had to demonstrate compliance with its own studies, with a deficiency letter in this queue position carrying a documented five to seven month cost. Second, compliance exposure under the post-Order No. 901 standards. FERC Order No. 901 directed NERC to develop reliability standards for inverter-based resource performance, data sharing, model validation and planning studies; the resulting standards — principally PRC-029-1 for ride-through, PRC-028-1 for disturbance monitoring and PRC-030-1 for reporting of unexpected performance — changed what a developer must prove and when. At 400 MVA aggregate on a 230 kV connection this facility is in scope, and the Owner needed a design-stage demonstration rather than a commissioning-stage discovery. Third, capital efficiency: sizing shared equipment for 400 MW would be waste on a plant that can never export more than 300 MW, but sizing it for 300 MW required proving that no credible contingency drives a shared asset above rating.

3. Study Objectives and Scope of Work

Eleven numbered objectives defined the engagement: (1) establish the hybrid electrical architecture and its shared-versus-segregated decisions; (2) engineer the 34.5 kV collection system for thermal adequacy, loss economics and constructability, including ampacity by the Neher-McGrath method and IEEE 835; (3) select and justify the medium-voltage grounding method and derive its consequences for insulation level, arrester rating and fault detection; (4) verify reactive capability at the point of interconnection across all credible dispatch states; (5) establish short circuit duties by the ANSI/IEEE C37.010 method; (6) develop collector coordination addressing limited inverter fault contribution and high-impedance faults; (7) quantify arc flash incident energy and work practice boundaries; (8) demonstrate harmonic compliance against the correct IEEE 519 limit row and IEEE 2800-2022; (9) perform EMT studies for control interaction, temporary overvoltage and energization; (10) design grounding across the full footprint to IEEE 80 and IEEE 81; and (11) specify the hybrid plant controller and produce the compliance evidence package.

3.1 Scope and Deliverables

4. Regulatory and Standards Basis

The facility interconnects to the bulk electric system of a Balancing Authority in the continental United States under a Large Generator Interconnection Agreement pursuant to FERC Order No. 2003 as amended, with the queue process governed by Order No. 2023 and Order No. 2023-A.

Two FERC orders set explicit performance obligations. Order No. 827 requires non-synchronous facilities to maintain 0.95 leading to 0.95 lagging power factor measured at the high side of the generator substation, which here is the point of interconnection. Order No. 842 requires equipment capable of primary frequency response with a maximum droop of 5 percent and a maximum deadband of plus or minus 0.036 Hz.

PRC-029-1 governs frequency and voltage ride-through here because both resources are inverter-based: the photovoltaic plant obviously so, and the wind plant because Type 4 machines interface entirely through a full-scale back-to-back converter. PRC-024-4 does not apply; following the introduction of PRC-029-1 it applies to synchronous generators, synchronous condensers, and Type 1 and Type 2 wind resources only, and for inverter-based resources it is superseded. This mattered commercially, because the Owner's original submission had cited PRC-024 ride-through curves and would have committed the plant to the wrong envelope.

PRC-029-1 was adopted by the NERC Board in October 2024 and filed with FERC in November 2024, becoming effective in 2026, twelve months after jurisdictional approval and effective the following calendar quarter. It applies to BES inverter-based resources and to non-BES resources 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 cover voltage ride-through (R1), reactive current and control behavior during and after disturbances (R2), frequency ride-through (R3), and documentation of hardware limitations for legacy resources (R4), addressed here by a statement of non-applicability. The expectations that drove design were riding through phase-angle jumps of at least 25 electrical degrees; not ceasing current injection anywhere inside the ride-through zone, which effectively disallows momentary cessation; returning to pre-disturbance current exchange within approximately five cycles of voltage recovery; and riding through frequency rates of change up to 5 Hz/s. PRC-028-1 requires disturbance monitoring equipment and supplies the measured event data by which operational compliance is demonstrated, and IEEE Std 2800-2022 defines the underlying performance expectations that the PRC standards make enforceable.

4.1 Standards and Regulatory Register

5. System Modeling and Data Development

5.1 Data sources and gap closure

The Transmission Provider supplied a positive-sequence transmission equivalent and fault duty at the point of interconnection for the base topology and the two most limiting outage conditions. Both converter suppliers provided control block diagrams, generic model parameters, power quality test reports and black-boxed EMT models under non-disclosure agreement. The energy consultant provided 8,760-hour time series and a wake-adjusted turbine-by-turbine production matrix.

Four material gaps were closed explicitly. Soil thermal resistivity was unmeasured; because Neher-McGrath ampacity is more sensitive to it than to any other parameter, a 34-location thermal probe survey was specified, measuring 62 to 118 degree-Celsius-centimeters per watt at design moisture and up to 210 in two sandy zones when dried out. The design adopted 90 native and 75 for engineered backfill. Aggregate harmonic emission was available only per unit, so aggregating 45 wind converters and 100 inverters required a summation law: arithmetic to the fifth harmonic, 1.4 above the fifth to the tenth, and 2.0 above. Background harmonic distortion was assumed at 1.0 percent total with a 0.7 percent fifth-harmonic component in worst-case alignment, with a 30-day pre-energization measurement later performed. Soil electrical resistivity came from a Wenner survey along eight traverses, inverted to 65 ohm-meters over 10.5 feet upon 420 ohm-meters.

5.2 Model build, equivalencing and benchmarking

Four models were kept consistent through a controlled parameter register. A PSS®E model carried the transmission equivalent with both resources fully represented, and a two-machine equivalent for contingency screening, using the second-generation renewable energy library — converter, electrical control and plant control modules for the photovoltaic plant, the same augmented by drive train, pitch and torque modules for the Type 4 wind plant, with a separate plant controller instance per resource. An ETAP model carried the unreduced plant, all 95 pad-mounted transformers and all 13 feeders segment by segment, for short circuit, coordination, arc flash and ampacity. A PSCAD/EMTDC model carried supplier black-box converters aggregated into six wind and seven photovoltaic equivalents, each representing one feeder with its true cable impedance and charging. A CDEGS model carried the substation grid, 45 turbine electrodes, the array electrode network, cable shields and counterpoise as one conductive system.

Three benchmarks preceded any reported result. The PSCAD plant reduced to a Thevenin equivalent matched PSS®E within 0.4 percent on active power, 1.1 percent on reactive power and 0.002 pu on voltage. A three-phase fault compared in both platforms over the first 500 ms showed the positive-sequence model reproducing converter current magnitude within 6 percent and recovery within one cycle, but not reproducing phase-angle-jump response at all. ETAP short circuit at 230 kV reconciled to the Transmission Provider's figure within 0.2 kA. The report stated plainly that phase-angle jump, current limiting under unbalanced faults, sub-cycle control interaction and hybrid controller mode transitions were answerable in EMT and only in EMT, while EMT could not economically cover the several hundred permutations needed for steady-state screening.

6. Study Methodology and Assumptions

6.1 Plant Configuration Summary

Both resources connect to a single 34.5 kV collector bus in two sections with a normally open tie. Section A carries wind feeders W1 to W3 and solar feeders S1 to S3, 188.4 MW of nameplate; Section B carries W4 to W6 and S4 to S7, 211.6 MW. Each section is served by one 105/140/175 MVA ONAN/ONAF/ONAF main power transformer, 230 kV grounded-wye to 34.5 kV delta, with a high-side load tap changer of plus or minus 10 percent in 33 steps and 13.0 percent impedance on the 175 MVA base.

6.2 Tool selection

Tool selection followed the deliverable: PSS®E for load flow and positive-sequence dynamics, because it is the platform in which the Transmission Provider studies and in which MOD-032 models must be submitted; ETAP for short circuit, coordination, arc flash and ampacity, because the ANSI C37.010 and IEEE 1584-2018 methods are natively implemented and auditable; PSCAD/EMTDC for transient work, because supplier models are distributed for that platform and control interaction cannot be assessed in phasor domain; CDEGS for grounding, because the footprint required a distributed, non-equipotential treatment; and Python for the 8,760-hour curtailment and loss analysis and the economic conductor optimization.

6.3 Case and Scenario Matrix

6.4 Acceptance Criteria

7. Analysis and Results

7.1 Resource layout and the competition for ground

The wind resource occupies approximately 11,400 acres at five rotor diameters crosswind by eight downwind for a 508-foot rotor at a 328-foot hub height; the array occupies approximately 1,450 acres at a ground coverage ratio of 0.33, north-south tracker rows at a 19.7-foot pitch, tracking to plus or minus 60 degrees with backtracking.

The layouts collided on 620 acres that were simultaneously the preferred position for six turbines and the flattest developable array ground, so the optimization was run jointly rather than sequentially. Moving three turbines raised array-average wake loss from 7.1 to 7.5 percent, costing about 2.4 GWh per year, and released ground for 18 MWdc of array worth about 34 GWh per year before curtailment — a trade that is obvious once posed jointly and invisible if the wind layout is frozen first. Access roads at 34 feet width and crane pads of approximately 250 by 130 feet cut the array plan into blocks, and the array was re-blocked around those corridors at a cost of six tracker rows rather than realigning 1.9 miles of road. Turbine shadows transiting the array cost 0.11 percent of photovoltaic generation annually and in two blocks disturbed the backtracking algorithm's irradiance-based mode selection, resolved by reorienting those blocks so shadow transit runs perpendicular to string wiring.

The DC:AC ratio is a genuinely hybrid result. For a standalone plant on this resource the loss-optimal ratio was 1.38; behind a shared 300 MW limit that already truncates the combined midday peak, additional DC competes not only against clipping but against wind generation for the same headroom, and the hybrid optimum falls to 1.29. The design adopted 1.32, deliberately above that, because the incremental DC earns most of its marginal energy in morning and evening shoulder hours when photovoltaic output is far below clipping and wind is statistically low, so it competes for headroom that is usually free. Modeled clipping loss was 1.9 percent of DC energy.

7.2 Collection system design and optimization

The collection question specific to a hybrid plant is whether wind and solar circuits should share a feeder. They should not. Protection settings suited to 4.6 MVA turbine transformers differ from those for 5.0 MVA inverter transformers on a very different duty cycle; a wind feeder's governing ampacity condition is a cool moist winter night while a solar feeder's is a hot dry summer midday, so a shared circuit must be designed to the worse of both; the arrays are built by different crews on different schedules; and segregated feeders yield per-resource energy attribution almost free. Segregation was adopted at feeder level and sharing at bus level: the resources share the bus, the transformers, the 230 kV yard and the point of interconnection, and nothing below the bus.

Underground ampacity was computed by the Neher-McGrath method per IEEE 835 at 42 inches to top of cable, 25 degrees Celsius ambient soil at depth, 90 degrees Celsius conductor, native thermal resistivity 90 and engineered backfill 75, at 100 percent load factor for trunk sections. Duct bank mutual heating was computed explicitly at the six locations where three or more circuits converge on the substation approach, where derating reached 21 percent and forced a conductor increase on two circuits and increased separation on a third.

Loss optimization used the economic conductor method — minimizing installed cost plus the present worth of I-squared-R losses — computed segment by segment against each segment's actual 8,760-hour loading rather than a nominal peak. The preliminary schedule, developed on a uniform sizing rule, gave 2.41 percent collector loss at coincident peak and 1.18 percent energy-weighted; the optimized schedule gave 1.62 and 0.79 percent. It did not simply upsize: it increased high-duty-cycle trunk sections and reduced 23 laterals oversized relative to their true energy throughput, so net cable capital fell by approximately 1.4 million dollars while annual losses fell by 4.3 GWh. Cable charging is a first-order quantity across 71.6 circuit-miles: 22.4 Mvar at nominal voltage, 24.7 Mvar at 1.05 pu, present whenever feeders are energized and the reason the facility can export reactive power at zero active power.

7.3 Medium-voltage system grounding

The main power transformers are delta on the 34.5 kV side, so the collector has no ground source until one is given deliberately. Left ungrounded, a single line-to-ground fault raises unfaulted phases to full phase-to-phase voltage indefinitely, arcing faults can produce restrike overvoltages several times nominal, and there is no ground return current to detect.

Low-resistance grounding at 600 A was selected, implemented as one zig-zag grounding transformer per bus section with a neutral grounding resistor rated 600 A for 10 seconds. The reasoning was damage limitation and shield economics: effective grounding would have produced 14.9 kA at the bus, comparable to three-phase duty, with severe cable and termination damage and a requirement for one-third-neutral concentric wire shields throughout, while 600 A permitted a 5-mil copper tape shield and confined a termination fault to a repairable event.

The consequences were accepted explicitly and carried into every downstream calculation. Because the system is not effectively grounded the coefficient of grounding is 1.0: unfaulted phase-to-ground voltage reaches full phase-to-phase value, 1.73 pu of nominal, with an EMT transient peak of 1.82 pu. Cable insulation was specified at the 133 percent level, 420 mils on 35 kV class rather than 345, adding about 6 percent to cable cost, and arresters moved from 27 kV duty cycle with 22 kV MCOV to 36 kV with 29 kV MCOV. That reduced the insulation coordination margin at the 150 kV BIL switchgear from 68 to 27 percent, still above the 20 percent criterion but tight enough that arrester leads were limited to 18 inches. The 600 A limit also places ground fault current below the maximum load current of every trunk feeder, addressed in Section 7.6.

7.4 Load flow and reactive capability at the point of interconnection

Order No. 827 requires 0.95 leading to 0.95 lagging at the point of interconnection, which at the 300 MW limit is plus or minus 98.6 Mvar at 230 kV. The converters must produce that plus everything the network consumes, less what it generates.

The leading case is far easier because network consumption assists absorption: absorbing 98.6 Mvar at the point of interconnection requires the converters to absorb only 52.8 Mvar. That roughly three-to-one asymmetry is characteristic of a cable-rich collector and is the most useful number to establish early. In case C1, with wind at 180 MW and solar curtailed to 126 MW, available converter capability was 265.8 Mvar against 144.4 required.

The binding case is C3, full wind at night with the array dark. The requirement applies at the facility's maximum output for the condition, 180 MW, needing 59.2 Mvar at the point of interconnection. The turbines provide plus or minus 0.95 power factor at their own terminals, 1.315 Mvar each and 59.2 Mvar for all 45 — numerically identical to the requirement, and therefore short by exactly the network's consumption.

The conventional resolution is a dynamic reactive device, budgeted at approximately 4.2 million dollars for a 40 Mvar unit. The hybrid resolution is that 220 MW of photovoltaic inverters sit idle and energized all night, and an inverter with no DC power available is a fully capable static compensator limited only by its apparent power rating, unit transformer and auxiliary supply. Enabling reactive support at zero DC power provided plus or minus 60 Mvar per bus section after derating, against a 5.2 Mvar deficit, and no dynamic reactive device was installed. The same capability resolves case C5, in which the plant with no generation and all feeders energized exports 22.6 Mvar of capacitive reactive power from cable charging alone; the alternatives were nightly feeder de-energization or a shunt reactor.

7.5 Short circuit and equipment duty

Fault duty was computed by the ANSI/IEEE C37.010 method with separate first-cycle and interrupting networks, X/R-based multiplying factors and contact parting time appropriate to the breaker. Inverter-based sources were represented as current-limited contributions at 1.1 per unit of rated current rather than impedance behind voltage, and excluded from the first-cycle asymmetrical calculation because a converter cannot produce a DC offset it is not commanded to produce.

At 230 kV the three-phase duty is 16.9 kA symmetrical with X/R of 14.2 and single line-to-ground duty is 15.4 kA. With a 3-cycle breaker and 2-cycle contact parting the interrupting multiplying factor is 1.03, giving 17.4 kA required interrupting capability and a close-and-latch requirement of 27.0 kA rms asymmetrical, 43.9 kA peak; the 245 kV, 40 kA breakers carry both at 44 and 68 percent of rating.

The 34.5 kV bus produced the first significant finding. With the tie closed, both transformers in parallel present 3.71 percent impedance on a 100 MVA base which, with 1.49 percent source impedance, gives a grid contribution of 32.2 kA; converter contribution adds 4.5 kA for 36.7 kA total. That depends entirely on transformer impedance, which the original specification set at 10.5 percent, at which value the closed-tie duty is 41.8 kA — above the 40 kA switchgear rating. Of the three responses available — non-standard higher-rated switchgear, permanent open-tie operation, or higher transformer impedance — the study recommended 13.0 percent, at a cost of 7.2 Mvar of additional transformer reactive consumption that a plant with 265.8 Mvar of converter headroom absorbs for the price of a setpoint. In a plant without spare reactive capability the answer would have been different, and the report said so.

7.6 Protective device coordination

The collector protection problem in an inverter-based plant is that the two directions of fault current differ radically. With the transformer in service the grid dominates, contributing 21.1 kA at the bus and 8.4 kA at the most remote pad, and conventional time-overcurrent works. With the transformer out, or before grid contribution arrives, the only sources are converters limited to about 1.1 per unit of their own rating — 250 to 900 A at a remote pad — which overcurrent cannot detect reliably while remaining above load current.

Three supplementary elements were specified. Negative-sequence overcurrent, 50Q and 51Q, detects unbalanced faults far below phase pickup and was set at 0.12 per unit of feeder rating with 0.35 second delay. Directional ground overcurrent, 67N, polarized from the grounding transformer neutral current transformer, provides the detection phase elements cannot, since the 600 A ground fault current sits below the 543 to 564 A maximum trunk load current; pickup was set at 40 A on an inverse curve coordinating with the neutral grounding resistor's 600 A, 10-second thermal limit. High-impedance faults on the 20.4 miles of overhead collector received specific treatment, because a broken conductor on dry soil may produce only 5 to 50 A: detection based on harmonic content, current randomness and arcing signature was specified on the six wind feeders, alarming to the control center and tripping after a confirmation interval on the four feeders with the highest public exposure. The report was explicit that these algorithms are probabilistic rather than deterministic and supplement rather than replace physical measures.

Second-harmonic restraint at 15 percent guards feeder ground elements against the unbalanced inrush of energizing up to 22 pad transformers at once; bus differential clears each bus section in 0.05 seconds; transformer differential protects each main power transformer; and the coordination interval between feeder time overcurrent and low-side backup was verified at 0.30 seconds throughout. Relay loadability was verified against PRC-025-2, and PRC-019 coordination was demonstrated by overlaying converter and plant controller limiter characteristics against the protection settings and the PRC-029-1 ride-through envelope, confirming that no protective element operates inside the no-trip zone.

7.7 Arc flash

Arc flash was assessed to IEEE 1584-2018 with NFPA 70E for work practice and PPE. A methodological point applies: the IEEE 1584-2018 empirical model is valid from 208 V to 15,000 V and does not apply to the 34.5 kV collector, so the Ralph Lee theoretical method was used for all 34.5 kV equipment at a 36-inch working distance.

The pattern is characteristic: bus incident energy is governed by differential protection and is acceptable, while incident energy at the far end of a feeder is governed by the time-overcurrent element that must be slow enough to coordinate, reaching 61.9 cal/cm² — above the 40 cal/cm² threshold at which NFPA 70E work practice effectively requires de-energization regardless of PPE. The mitigation was a maintenance setting group placing feeder phase and ground elements on instantaneous characteristics during work, selectable from the substation and the plant control system with positive indication at the work location. That reduces pad incident energy to 12.4 cal/cm² and the feeder cubicle to 17.9, both within available arc-rated PPE, at the cost of selectivity while active — an acceptable trade where any feeder trip curtails at most 35.2 MW.

7.8 Harmonic analysis

The point of common coupling is the 230 kV point of interconnection, so IEEE 519 limits are taken from the row for systems above 161 kV — the strictest row, and the one most often misapplied on renewable projects where the 69 to 161 kV row is sometimes used by mistake.

Voltage distortion at the point of interconnection was 0.62 percent total against the 1.5 percent limit, with no individual order above the 1.0 percent limit. Emission was separately checked against IEEE Std 2800-2022, which sets limits at the point of measurement as a percentage of plant rated current with a total rated-current distortion limit of 5 percent; the computed figure was 2.1 percent.

The frequency scan produced a finding. The collector's 22.4 Mvar of cable charging resonates in parallel with the source and transformer inductance. With the tie open, the short circuit level at a bus section is 1,122 MVA, placing the parallel resonance at harmonic order 7.08 — directly on the seventh — with an impedance magnitude of 42 ohms against 4.6 ohms at fundamental. With the tie closed it shifts to order 10.0. The 13.0 percent transformer impedance selected for fault duty is what moved the open-tie resonance onto the seventh; at 10.5 percent it would have sat at 7.9. This is a genuine coupling between two design decisions that would not appear if each study were run in isolation.

The resonance was accepted rather than eliminated. Aggregate seventh-harmonic emission, computed with the 1.4 summation exponent, is 0.31 percent of rated current, producing 0.44 percent voltage distortion at the 34.5 kV bus and 0.11 percent at the point of interconnection even with magnification. That margin would erode if collector cable were extended or a further resource added behind the same bus, so the design provided a spare switchgear position, duct and pad space on each section for a detuned seventh-harmonic filter — roughly 180 thousand dollars in provisions against 1.6 million for installed filters, which were not purchased.

7.9 Electromagnetic transient studies

Control interaction. Impedance-based frequency scanning identified a lightly damped 24 Hz mode appearing only when both resources were at high output and only at short circuit ratios of 3.0 or below, with 3.8 percent damping at SCR 3.0 against a 5 percent criterion. Root cause was interaction between the inverter phase-locked loop bandwidth, set at 40 Hz, and the wind converter outer voltage loop, mediated by the shared collector impedance. Reducing the PLL bandwidth to 18 Hz and slowing the wind converter outer loop time constant from 20 to 45 ms raised damping to 11.6 percent at SCR 3.0 and 19 percent at the credible worst case of SCR 5.4, at a 90 ms settling penalty well inside IEEE 2800 response requirements. Neither resource exhibited this mode alone; it existed only because they shared a bus.

Temporary overvoltage. Opening the 230 kV breaker at 300 MW export gave 1.28 pu peak at the point of interconnection and 1.34 pu at the 34.5 kV bus, decaying within 180 ms, with arrester energy duty of 2.1 kJ per kV of MCOV against 7.8 station-class capability. As anticipated from the grounding decision, a collector single line-to-ground fault gave 1.73 pu sustained and 1.82 pu transient peak on unfaulted phases, confirming the 36 kV duty-cycle, 29 kV MCOV arrester selection and the 133 percent insulation level, with arrester energy at 34 percent of capability for maximum 10-second clearing.

Transformer energization. Uncontrolled energization of a 175 MVA transformer from the 230 kV side gave 6.1 per unit first-peak inrush with a 2.4-second decay, a 4.2 percent voltage dip at the point of interconnection, sympathetic inrush in the second unit, and — through inrush harmonics interacting with cable capacitance — 1.31 pu at the 34.5 kV bus for approximately 900 ms, long enough to matter for arrester duty and converter protection. Controlled point-on-wave closing with independent pole operation and residual-flux measurement reduced first-peak inrush to 0.9 per unit, the dip to 0.6 percent and the collector overvoltage to 1.08 pu, and was specified on both 230 kV transformer breakers. Collector energization was limited by procedure to 22 pad transformers per operation.

7.10 Flicker

Flicker was assessed to IEEE Std 1453 for the wind resource only, using flicker coefficients from the turbine power quality test report at the applicable network impedance angle, summed over 45 turbines by the root-sum-square law. Short-term flicker severity at the point of interconnection was 0.015 against a planning level of 0.35; switching-operation flicker from turbine start-up gave 0.005 with a maximum relative voltage change of 0.068 percent against a 3 percent step limit. These are so far below the limits that the honest conclusion is that flicker is not an engineering constraint at this system strength. It was assessed because the interconnection agreement required it, and the sensitivity case at SCR 5.4 still gave only 0.045.

7.11 Insulation coordination

The 230 kV arresters were selected at 180 kV duty-cycle rating with 144 kV MCOV, appropriate to an effectively grounded transmission system; the 34.5 kV margins are governed entirely by the grounding decision. The 27 riser poles at overhead-to-underground transitions received particular attention, because a traveling wave entering a cable reflects at the open remote end and can approach twice the incident magnitude. Arresters were applied at both ends of every transition, with lead length, shield bonding and effective protective level calculated individually at each location rather than assumed.

7.12 Grounding across the plant footprint

The substation grid covers 4.8 acres and was designed to IEEE 80 with 4/0 AWG bare copper on a 20-foot grid and 92 rods of 5/8 inch by 30 foot in the measured two-layer soil, giving 0.42 ohms standalone. With 15.4 kA of single line-to-ground fault current, a split factor of 0.38 for shield wire and neutral return paths and a decrement factor of 1.05 at 0.30 seconds, grid current is 6.14 kA and ground potential rise 2,580 V.

The large footprint changes the problem in two ways. The adverse effect is transferred potential: every cable shield and counterpoise conductor bonds the substation grid to 45 turbine electrodes and 50 photovoltaic block electrodes across twelve thousand acres. CDEGS modeling showed 810 V at the nearest turbine pad 0.7 miles away, falling below 150 V beyond 2.4 miles. Mitigation at every pad comprised graded ring electrodes at 3 and 6 feet, crushed rock surfacing at 3,000 ohm-meters within the working area, and equipotential bonding of all exposed metalwork. All plant communications are fiber optic end to end with no metallic telecommunications entering the site, which removed the IEEE 367 and IEEE 487 question entirely — a choice made specifically because of the transferred potential result.

The favorable effect is that the array is an enormous electrode. Approximately 47,000 driven tracker piles, bonded through the array grounding conductor network and a bare copper counterpoise, present a computed array resistance of 0.09 ohms. Bonding that counterpoise to the substation grid through four dedicated 4/0 AWG connections reduced composite resistance from 0.42 to 0.31 ohms and ground potential rise from 2,580 V to 1,905 V, bringing touch potential within limits without additional grid conductor.

7.13 The hybrid power plant controller

The controller is what makes the facility one plant rather than two. The architecture is a single master hybrid plant controller measuring at the point of interconnection, commanding two subordinate resource controllers each retaining local authority over its own units.

The arbitration problem is that the resources are not interchangeable. Inverters follow an active power command in 100 to 200 ms and ramp at effectively any rate; turbines respond in 0.5 to 2 seconds, are limited by pitch and drive-train constraints to roughly 10 percent of rated power per second, and accumulate fatigue duty under repeated fast down-regulation. Photovoltaic reactive capability is limited by inverter apparent power rating at high output and is essentially unlimited at night, while wind reactive capability tracks instantaneous output. The specification therefore allocates fast active power regulation preferentially to the photovoltaic resource, sustained curtailment by economic merit, and reactive regulation to whichever resource has headroom.

Total curtailed energy is identical under every strategy, because the export limit is a hard constraint that does not care which resource is reduced; what differs is which megawatt-hours are lost and what the network loses delivering the rest. The Owner's revenue structure valued wind energy higher at the margin, its incentive being production-based while the photovoltaic asset's was capacity-based, so the selected strategy curtails photovoltaic output first, breaks ties on collector loss, and applies a fairness floor preventing either resource from absorbing more than 55 percent of the aggregate reduction in a rolling month so that neither asset's availability warranty calculations are distorted. Hard constraints override the economic logic in all cases: transformer loading against the ONAF2 rating, feeder ampacity, bus section balance with the tie open, pad transformer voltage, and the aggregate export limit.

Primary frequency response under Order No. 842 was specified at 5 percent droop with a plus or minus 0.036 Hz deadband applied at plant level rather than unit level, so the response seen at the point of interconnection has the correct droop regardless of how many units are running. Down-regulation is always available and delivered by the photovoltaic resource within 400 ms; up-regulation requires headroom, which exists only when the plant is curtailed, and available up-response is reported continuously to the host Balancing Authority.

Failure modes were specified explicitly because the aggregate limit is a compliance obligation. Loss of point-of-interconnection measurement forces fallback to an estimate from the transformer low-side meters with a conservative 280 MW limit; loss of communication to one resource controller reduces the other's limit assuming the unreachable resource is at maximum; and loss of both drops each resource controller to a local static 150 MW limit, so the aggregate cannot exceed 300 MW under any communication failure.

7.14 Metering and telemetry for two resources on one interconnection

One interconnection means one revenue meter, and one revenue meter cannot by itself say how much energy came from wind and how much from solar. The architecture comprises a primary revenue meter and check meter at the point of interconnection, both 0.2 accuracy class with 0.3 class instrument transformers; sub-metering at each transformer low side; and feeder-level metering on all 13 collector feeders, which because feeders are resource-segregated gives direct per-resource attribution.

Attributing losses is not trivial, because losses are quadratic and shared. A loss-compensation matrix expressing each resource's marginal loss factor as a function of both resources' output was derived from 240 load flow solutions across the dispatch space, applied in settlement to reconcile the sub-metered quantities to the single revenue figure and supporting MOD-025 reporting for each resource separately. Disturbance monitoring to PRC-028-1 comprises point-on-wave recording at 256 samples per cycle at the point of interconnection and both bus sections, synchrophasors at 60 frames per second, sequence-of-events recording at 1 ms, and continuous trending, with telemetry by redundant DNP3 paths to the substation RTU and ICCP onward at a 2-second scan.

7.15 Dynamic models and validation planning

Positive-sequence models were delivered per MOD-032 using the second-generation renewable energy library, with separate model instances and plant controllers per resource. The study documented an honest limitation: the standard library contains no representation of a hybrid supervisory controller arbitrating between two resource controllers. The submitted model uses coordinated but independent plant controllers whose parameters reproduce aggregate behavior for continuous small-signal conditions, and does not reproduce discrete mode transitions such as allocation switching, constraint activation and communication-failure fallback; those were demonstrated in EMT, with a recommendation that the Transmission Provider treat the positive-sequence model as valid for planning studies and the EMT model as the reference wherever mode transitions are material.

The MOD-026 and MOD-027 validation plan was written at design stage rather than left to commissioning, specifying staged tests at energization — voltage reference steps of plus and minus 2 percent with both resources and with each alone, reactive setpoint steps, frequency reference bias steps of 0.5 Hz to exercise the droop path, and active power reference steps to characterize ramp limiting — each with acceptance bands defined in advance.

8. Sensitivity and Scenario Analysis

Three sensitivities changed the design. Soil thermal resistivity in the two dried-out sandy zones drove trunk ampacity to 447 A, below the 543 A design load, forcing 2.1 miles of engineered thermal backfill at approximately 210 thousand dollars — an expenditure that would otherwise have surfaced as a cable failure years into operation. The transformer impedance sensitivity revealed the coupling between fault duty and harmonic resonance described in Section 7.8. The harmonic summation exponent produced a range wide enough that the compliance conclusion depends on the assumption, stated openly alongside the recommendation for the 30-day measurement that later confirmed the middle of the assumed range. Flicker is the clearest example of an element that did not matter: severity never exceeded 13 percent of the planning level anywhere in the range.

9. Findings and Root Cause Assessment

Findings F2, F3 and F8 exist only because this is a hybrid plant: one resource's reactive capability disappears when its primary energy source does while the shared network's reactive appetite does not; the shared collector concentrates the cable charging of two plants against one transformer impedance; and two converter control systems from different suppliers, each stable alone, shared a bus. None would have been found by studying either plant in isolation.

10. Mitigation Options and Recommendations

The recommended package was to raise transformer impedance to 13.0 percent and accept the reactive cost; enable inverter reactive support at zero DC power and delete the dynamic reactive device; implement the maintenance setting group with positive local indication; retune the converter control parameters as validated in EMT; specify point-on-wave closing on both 230 kV transformer breakers; specify high-impedance fault detection on the overhead collector; provide filter provisions without filters; and specify the controller fallback logic.

Grid-forming control was evaluated seriously and not recommended: it would have addressed F8 more robustly than retuning and improved behavior at low system strength generally, but commercial availability at this power level, with warranty support and an operating track record, was insufficient to justify the schedule risk. The report recommended instead that the converters be required to be firmware-upgradeable to grid-forming operation and that the contract preserve that option, which it did.

11. Implementation Support and Field Validation

Keentel supported construction and energization from Month 10, providing settings files, the collector energization sequence, factory acceptance test witnessing for the transformers and switchgear, and on-site support for staged testing.

Grid resistance was measured by fall-of-potential using the slope method with a 2.9-mile current lead, which on a composite electrode of this extent is the only defensible technique; a conventional 62 percent measurement here would have produced a meaningless number. Staged tests for MOD-026 and MOD-027 were executed to the pre-written plan with both resources online, with wind alone and with photovoltaic alone, and all responses fell within the pre-declared acceptance bands. The single retune required was to the plant controller reactive integral gain, set for symmetric two-resource dispatch and hunting mildly at 0.3 Hz when the photovoltaic resource operated alone at low output; the gain was scheduled against the number of participating resources.

The PRC-029-1 design compliance demonstration package was assembled from converter type test reports, EMT-simulated response at the ride-through envelope boundaries, phase-angle jump tests at 25 and 35 electrical degrees, rate-of-change-of-frequency tests to 5 Hz/s, and evidence that no plant or unit protective element and no converter limiter operates inside the no-trip zone. The PRC-028-1 disturbance monitoring architecture was commissioned and verified against a staged event, so the measured data needed for ongoing operational compliance is available from first energization.

12. Results and Value Delivered

The 10.2 million dollars comprises approximately 2.9 million in avoided transformation with its bay and foundation cost, 4.2 million in the avoided dynamic reactive device, 1.4 million in reduced cable capital, and 1.7 million in the present worth of avoided collector losses, less the incremental cost of thermal backfill and 133 percent insulation. The less quantifiable outcome was schedule: the package cleared review at first submission.

13. Lessons Learned and Engineering Insights

A hybrid plant is a shared-network problem, not two generation problems. Every finding unique to this project — the night-time reactive deficit, the seventh-harmonic resonance, the 24 Hz control interaction — arose at the interface, not inside either resource. Studying the two plants separately and superposing the results would have found none of them.

Size shared equipment to the interconnection limit, not to the sum of nameplates. The instinct to give each resource its own transformer is usually wrong when the plant is export limited. A common bus with two transformers sized to the export limit removed 74 MVA of installed capacity and gave the controller a single place to enforce the aggregate constraint. The cost is that a bus fault takes the whole facility, which is why the 0.05 second bus differential clearing time was not negotiable.

Idle inverters are the cheapest reactive plant on the project. Two hundred and twenty megawatts of inverters spend every night energized and doing nothing. Configuring them for reactive support at zero DC power cost a firmware option and a controller mode, and displaced a 4.2 million dollar device. This is the most transferable finding from the engagement and applies to any hybrid or storage-augmented plant.

Design decisions in different disciplines are coupled, and the coupling is invisible in siloed review. Raising transformer impedance to solve a fault duty problem moved a harmonic resonance onto the seventh. Choosing resistance grounding to limit damage raised temporary overvoltage, which forced heavier arresters, which cut the insulation coordination margin from 68 to 27 percent and turned arrester lead length into a compliance-critical dimension.

Measure the soil, and state what the model cannot do. Neher-McGrath ampacity is more sensitive to thermal resistivity than to anything else in the calculation, and a survey costing a small fraction of one percent of the cable budget found two zones where design load exceeded true ampacity by 21 percent. Equally, positive-sequence models cannot represent phase-angle jump response or hybrid controller arbitration; saying so in the report, and pointing to the EMT work that answers those questions, is the difference between a reviewer trusting the document and auditing it.

14. Keentel Capability Summary

• Hybrid generating facility electrical architecture: shared versus segregated transformation, bus and feeder decisions, and the point of measurement for aggregate limits
• 34.5 kV collection system design, overhead and underground: conductor and cable selection, Neher-McGrath and IEEE 835 ampacity, duct bank mutual heating, economic conductor optimization
• Medium-voltage grounding studies: grounding transformer and neutral grounding resistor sizing, ground fault overvoltage, insulation level and arrester consequences
• Load flow, reactive capability and voltage control studies to FERC Order No. 827 at the point of interconnection
• Short circuit and equipment duty to ANSI/IEEE C37.010, including current-limited inverter source representation
• Protective coordination for collector feeders: limited fault contribution, directional ground protection, high-impedance fault detection
• Arc flash to IEEE 1584-2018 and the Lee method above 15 kV, with NFPA 70E work practice and maintenance mode design
• Harmonic analysis and frequency scanning to IEEE 519 and IEEE 2800-2022, including aggregation and resonance assessment
• Electromagnetic transient studies in PSCAD/EMTDC: control interaction, temporary overvoltage, energization and point-on-wave closing
• Flicker assessment to IEEE 1453; insulation coordination to IEEE 1313.2 and IEEE C62.22
• Grounding design and field validation to IEEE 80 and IEEE 81 across large distributed footprints, including transferred potential
• Hybrid power plant controller specification, curtailment allocation logic, failure-mode design, metering and telemetry architecture
• Dynamic model development, MOD-026 and MOD-027 validation planning, and PRC-029-1 and PRC-028-1 compliance evidence packages

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.