Compendium

Walkthrough — Design a 100 MW Utility-Scale Solar PV Plant

54 min read

Walkthrough — Design a 100 MW Utility-Scale Solar PV Plant

A concrete, end-to-end design pass for a 100 MW_DC / 80 MW_AC photovoltaic plant in West Texas, with single-axis tracking, TOPCon bifacial modules, central inverters, 1500 V DC architecture, and a 138 kV grid interconnect. Every section ties the engineering decision back to its underlying Tier-3 reference note.

1. What we’re building

The project is a greenfield utility-scale solar PV plant sited in Reeves County, West Texas at roughly 32° N latitude, 31° 30’ N more precisely. The Permian Basin region delivers a global horizontal irradiance (GHI) of about 5.8 kWh/m²/day annually (≈ 2 100 kWh/m²/yr) and a direct normal irradiance (DNI) of roughly 7.0 kWh/m²/day, well above the U.S. national average of 4.5 kWh/m²/day. This sites the project in the top decile of U.S. solar resource quality, which is what makes a $25/MWh PPA economically tenable.

Design intent: 100 MW_DC nameplate, 80 MW_AC point-of-interconnection (POI) export, giving a DC-to-AC ratio (inverter loading ratio, ILR) of 1.25. The expected net capacity factor is 28-30%, with the upper end achievable thanks to the bifacial gain on a single-axis tracker over high-albedo sandy soil. Expected annual energy yield is 250 GWh in year one, declining at the module degradation rate of 0.45%/yr. Land take is ≈ 250 acres (≈ 101 hectares) at a packing density of 0.4 MW_DC per acre — typical for 1P single-axis tracker layouts at a ground coverage ratio (GCR) of 0.33-0.38.

The plant interconnects at 138 kV into the Oncor / ERCOT West weather zone. The PPA is a 20-year fixed-price offtake at **1-3/MWh). Commercial Operation Date (COD) targets Q2 2028, giving a 24-month development-through-energization window from notice-to-proceed.

The reference engineering note for the underlying photovoltaic conversion physics is photovoltaic-cells.

2. Spec table

ParameterValue
DC nameplate100 MW_DC
AC POI capacity80 MW_AC
DC:AC ratio (ILR)1.25
Module count175 000 × 575 W bifacial TOPCon
Module typeJinko Tiger Neo 78HL4-(V) 600 W bifacial (575 W mono-facial equivalent rating used for derate)
MountingSingle-axis horizontal tracker (1P, one-portrait module per row)
Inverters18 × central, 4.5 MVA each (Sungrow SG4400UD-MV-US class)
DC voltage1500 V DC max system voltage
MV collection36 kV underground XLPE Al cable
Substation100 MVA, 138/36 kV step-up with 138 kV breaker yard
Grid POI138 kV ERCOT (Oncor)
Site area≈ 250 acres (≈ 101 ha)
Expected annual yield≈ 250 GWh
Capacity factor28-30% (AC)
PPA$25/MWh × 20 yr
COD24 months from NTP

The module-level reference is photovoltaic-cells; the spec table consolidates the project description into the numbers every later section will trace back to.

3. Site selection

Site selection for utility solar is a five-axis optimization across irradiance, wind/snow/seismic loads, terrain, environmental constraints, and grid interconnect feasibility — and the last one is usually the binding constraint in ERCOT.

Irradiance and meteorology. We pulled the National Solar Radiation Database (NSRDB) PSM3 dataset from NREL for the candidate site coordinates, extracting 20 years of hourly GHI, DNI, diffuse horizontal irradiance (DHI), temperature, wind speed, and humidity. NSRDB confirms a GHI of 5.8 kWh/m²/day, with a P50/P90 inter-annual variability band of ±3.5%. Soiling rate is the regional median 0.6%/wk between rain events, which average every 12-15 days in summer.

Wind and seismic. ASCE 7-22 Risk Category I (low-occupancy generation facility) wind hazard for Reeves County gives a basic wind speed V = 110 mph (49 m/s) for the 700-yr MRI, IEC 61400-1 Class III (V_ref 37.5 m/s) for any small met-tower / camera-mast structures. Seismic Design Category is B with Site Class C (very dense soil / soft rock); spectral response S_s = 0.18 g, S_1 = 0.06 g — seismic is non-governing for tracker structures here, wind controls.

Terrain. A LiDAR-derived 1 m DEM constrains us to parcels with sustained slope < 5% (single-axis trackers can tolerate up to 10-12% north-south but pile install economics degrade above 5%, and the bifacial gain falls off on tilted ground due to row-to-row shading geometry). The selected parcel is a former rangeland section with < 3% maximum slope and homogeneous calcareous loam/sand soil.

Soil and foundation feasibility. Soils investigation (per ASTM D1586 SPT, ASTM D2487 USCS classification, ASTM D6951 DCP refusal) confirmed N-values 15-25 (medium-dense) at the 1.8-2.5 m embedment depth band needed for driven steel piles. No caliche refusal above 3.5 m. Driven-pile feasibility is the gating soils question; ours passes. See soil-mechanics.

Environmental. A Phase I environmental site assessment (ASTM E1527-21) screened the site for past industrial use (none — rangeland since 1920s). Wetlands delineation under Clean Water Act Section 404 found no jurisdictional waters; no USACE permit required. The Endangered Species Act (ESA) Section 7 consultation flagged the dunes sagebrush lizard (proposed-threatened) and lesser prairie-chicken; the site sits 18 mi outside designated habitat — coordination letter only, no incidental-take permit needed. Section 106 archaeological survey (NHPA) cleared 100% of disturbance footprint. NEPA was avoided because no federal nexus exists (private land, ERCOT not subject to FERC NEPA, no federal funding). Full siting reference: environmental-engineering.

Grid interconnect. ERCOT generation interconnect studies are governed by the Nodal Operating Guides and the Resource Integration Manual. We filed a generator interconnect agreement (GIA) request, joined a cluster study group (cluster studies replaced the prior serial study process in 2023), and received a system impact study (SIS) showing 78 MW of injection capability at the 138 kV Reeves Switch with $4.2M of network upgrades (a new 138 kV breaker, terminal equipment, and 3 mi of 138 kV tap line). Queue position was secured early because dev started at NTP-30 mo.

Final site shortlist. Out of nine parcels evaluated against the five-axis criteria, two passed all gates. The selected 320-acre parcel (Texas General Land Office abstract 47, block 13) carries a 30-yr ground lease at $22/acre/yr escalating 1.5%/yr, with an option to purchase at year 25. The 70 acres of buffer beyond the 250-acre fenced array provides spacing for stormwater drainage, perimeter access, and a future 20% expansion (Phase II) that would require a separate GIA filing.

Mineral rights and split estate. Reeves County is in the heart of the Permian Basin. Surface estate (which we lease for solar) and mineral estate are almost always severed; the mineral owner has dominant rights under Texas law. We negotiated a surface use accommodation agreement with the mineral lessee (a Permian E&P operator) restricting drilling to two designated perimeter pads, with horizontal-only laterals beneath the array footprint and no surface use within the fenced area. Lease term aligned with our 30-yr surface lease.

4. Module choice

The 2025-2026 utility-scale module market is consolidated around three Tier-1 manufacturers with near-identical product platforms: Jinko Tiger Neo, LONGi Hi-MO 7, and Trina Vertex N. All three are n-type TOPCon (Tunnel Oxide Passivated Contact) bifacial designs on G12 wafers (210 × 210 mm pseudo-square monocrystalline silicon), with 60- to 78-cell variants for the utility market.

Selected module: Jinko Tiger Neo N-type 78HL4-(V) 600 W bifacial. Key parameters:

ParameterValue
Cell technologyn-type TOPCon, 16-busbar (MBB)
WaferG12 (210 mm) half-cut, 156 half-cells (78 full equivalent)
Front rating (STC)600 W
Back-side rating30% bifaciality factor
Module efficiency23.23%
Voc50.5 V
Vmp42.3 V
Isc14.99 A (mono-facial)
Imp14.19 A
Temp coeff (Pmax)−0.29%/°C (TOPCon advantage vs. PERC’s −0.34%)
NOCT45 ± 2 °C
FrameAnodized 6063-T6 Al, black/silver
ConstructionGlass-glass 2.0 mm + 2.0 mm tempered (bifacial)
ConnectorStäubli MC4-Evo 2
Warranty12-yr product, 30-yr linear power (87.4% at year 30)
2025-26 ASP≈ 0.17-0.19/W landed U.S. (post-AD/CVD + Section 201)

Mono-facial 575 W equivalent is used for the structural and string-sizing calculations to be conservative on glass-glass mass (≈ 32 kg/module, ≈ 12.5 kg/m²) and on STC current. The bifaciality factor of 30% gives us a back-side gain that we’ll model separately in Section 14.

Cell-level physics, briefly. TOPCon (Tunnel Oxide Passivated Contact) is an n-type silicon architecture where a thin (1.2-1.5 nm) tunnel oxide layer plus a doped poly-Si layer sits between the n-Si base and the rear metal contact. This passivates recombination centers at the back contact, lifting open-circuit voltage by ~ 20 mV vs. PERC and boosting cell efficiency from PERC’s 22.0-22.5% to TOPCon’s 23.5-24.5% at the cell level (and 22.5-23.5% at the module level after CTM losses). The architecture also drops the temperature coefficient by 0.04-0.05%/°C vs. PERC — meaningful in West Texas where cell temperatures routinely hit 65-75 °C in summer.

LeTID and LID, the n-type advantage. Light-induced degradation (LID) in p-type PERC is a 1-3% one-time loss in the first 24-72 hours of light exposure (boron-oxygen complex). Light- and elevated-temperature-induced degradation (LeTID) is a similar p-type-only failure mode that can compound to 3-6% over 1-2 years. n-type TOPCon has effectively zero LID and zero LeTID — a long-term yield advantage that PVsyst captures as a tighter degradation curve (0.45%/yr vs. 0.55%/yr for PERC).

Alternatives priced in parallel at procurement: LONGi Hi-MO 7 LR7-72HGD 575 W (same TOPCon platform, slight efficiency edge at 22.5%, marginally lower temp coeff) and Trina Vertex N TSM-NEG21C.20 605 W (slightly larger active area, otherwise comparable). Decision drivers: domestic-content treatment for IRA bonus credit (currently none of the three are domestic-cell — qualify the AD/CVD-exempt Southeast Asian polysilicon supply chain), Tier-1 BloombergNEF status, and dual-sourcing across at least two manufacturers to manage tariff and supply risk.

Material-science references: photovoltaic-cells (TOPCon cell architecture) and aluminum-alloys (6063-T6 anodized frames — the same alloy used for window extrusions, chosen for extrudability + corrosion resistance; T6 temper gives 240 MPa yield).

5. String and array sizing

String sizing is constrained by two voltage limits and two current limits. We work them in order.

Cold-temperature Voc limit (open-circuit, no load). NEC 690.7 (2026 edition) requires that the maximum module Voc at the lowest expected ambient temperature, when multiplied by the number of modules in series, must not exceed the maximum system voltage of any component (1500 V_DC here). The Voc temperature correction factor for n-type Si at low temp is approximately −0.25%/°C below STC’s 25 °C. West Texas ASHRAE 2% extreme-low temperature for Pecos is −10 °C (≈ 14 °F).

Voc_cold = 50.5 V × [1 + (−0.0025) × (−10 − 25)] = 50.5 × [1 + 0.0875] = 50.5 × 1.0875 = 54.92 V/module.

Max modules per string at 1500 V cold-Voc ceiling: 1500 / 54.92 = 27.31 → round down to 27. With IEC 61730 manufacturer-stated 1500 V class margin, NEC has been routinely permitting 28 modules in series in West Texas projects when designers use a more accurate site-specific extreme min (NSRDB 30-yr extreme min for the actual site coordinate gives −8 °C, yielding 54.67 V × 28 = 1531 V which exceeds 1500). We’re choosing 28 modules in series with a derated cold-Voc analysis using site-specific NSRDB data, after running PVsyst with the actual hourly temperature distribution: lifetime max instantaneous Voc never exceeds 1492 V in 50 000 hours of simulation. AHJ-acceptable in Texas; double-checked against the inverter’s 1500 V absolute max input.

MPPT window check. At maximum operating temp (cell temp ≈ 75 °C on a hot still day), Vmp drops to: Vmp_hot = 42.3 × [1 − 0.0029 × (75 − 25)] = 42.3 × 0.855 = 36.17 V/module → string Vmp = 36.17 × 28 = 1013 V. This sits inside the SG4400UD-MV MPPT range of 875-1300 V. ✓

String current and parallel count. Module Isc = 14.99 A → derated for irradiance enhancement events to 1.25 × Isc per NEC 690.8 = 18.74 A. This is the protective-device sizing current (we’ll use it for the fuse). For string topology, Isc determines whether strings can be combined in parallel within MPPT current ratings (which they always can at utility scale with central inverters), so no electrical limit on parallel count here.

Array totals. 175 000 modules ÷ 28 modules/string = 6 250 strings. Distributed across 18 inverters: 6 250 / 18 = 347 strings per inverter (we’ll use 350 nominal per inverter, with the remainder spread across the two least-loaded). 350 strings × 28 modules × 575 W mono-facial nameplate = 5 635 kW_DC per inverter at AC capacity 4.5 MVA → ILR = 5.635 / 4.5 = 1.252. ✓ This matches the 1.25 plant-level target.

Strings terminate at DC combiner boxes (24 strings each, 15 combiners per inverter) before running into the central inverter’s recombiner section.

Tracker block topology. The 6 250 strings × 28 modules layout maps onto 6 250 / (90 modules/row) ≈ 1 944 tracker rows. Each tracker row therefore hosts 28 × 3 = 84 modules of one string-block plus 6 modules of overflow that get string-combined across adjacent rows (acceptable per Nextracker installation guide as long as electrical-block boundaries align with the row-by-row backtracking control granularity). Inverter blocks are 350 strings × 28 modules / 90 modules/row = ~ 108 rows per inverter, arranged in a 6 × 18 row × column grid serving 5.6 MW_DC. Total 18 inverter blocks span a 2 700 m × 900 m site footprint.

Site DC layout summary. The full-plant DC topology in numbers: 175 000 modules → 6 250 strings × 28 modules → 270 combiners (15 per inverter × 18) × 24 strings → 18 inverters × 5.6 MW_DC → 18 step-up transformers × 4.5 MVA → 3 radial 36 kV feeders × 6 inverters → substation 100 MVA tx → 138 kV POI breaker → 138 kV tap line → Reeves Switch.

6. Inverter

We selected the Sungrow SG4400UD-MV-US, a 4.4 MVA outdoor 1500 V central inverter with an integrated 36 kV step-up MV transformer in a single ISO-container-style skid. Eighteen units give us 18 × 4.4 = 79.2 MVA nameplate AC, which we operate up to the 80 MW POI cap; an additional 1% nameplate cushion (≈ 0.8 MVA) sits in reactive-power headroom for grid-support compliance.

Key inverter specs:

ParameterValue
AC power4 400 kVA at 35 °C, 4 000 kVA at 50 °C (derated)
Max DC input voltage1500 V
MPPT voltage range875-1300 V
Number of MPPTs8 × 700 kW (segmented inputs)
Max efficiency (CEC)99.0%
EURO efficiency98.8%
Switching topology3-level NPC IGBT
CoolingForced air, smart-fan
AC output (LV side of integrated tx)645 V three-phase, integrated to 36 kV via 4.5 MVA cast-resin / oil-filled transformer
Grid supportIEEE 1547-2018 Cat. III + 1547.1 type-tested, ride-through LVRT (0% Vnom for 150 ms) + HVRT, reactive injection 0.95 lead / 0.90 lag, anti-island per UL 1741-SB
CommsSunSpec Modbus TCP, optional OPC-UA, IEC 61850 (substation side)
EnclosureNEMA 4X / IP66 power section, IP54 control
2025-26 ASP≈ 0.31M per skid

Competitive alternatives shortlisted and ultimately not selected: Sungrow SG3300UD (smaller increments, more skids needed, costlier per W), Power Electronics FS3030UX (Spain — strong EU footprint, less North American service depth), TMEIC TMA-Solar 4400 (Japan/U.S. — premium product, ~15% price premium for marginal efficiency gain).

The reason central inverters win over string inverters at this scale: string inverters at 250-350 kW each would require 230-320 units, with proportionally more DC trunk cabling, more comms drops, more O&M visits, and worse harmonic performance at the POI. The crossover where string economics overtake central is at roughly 30-50 MW_AC, depending on geography and labor cost; for an 80 MW plant in West Texas with low O&M labor cost, central is decisively cheaper at 0.09-0.10/W for string.

Reference notes: power-electronics (3-level NPC IGBT topology, MPPT control, switching losses) and electric-motor-taxonomy (the inverter-as-VSD section — same power-electronics chain used in motor drives, just running the other direction).

7. Tracker

The tracker is the second-most-impactful single product choice after the module. We selected the Nextracker NX Horizon 1P, a single-axis horizontal tracker with one-module-portrait rows (1P), independent self-powered tracker control units (TCUs) per row, and a built-in TrueCapture intelligent backtracking algorithm.

Specs:

ParameterValue
Configuration1P (one module in portrait orientation per row)
Modules per row90 modules (45 north + 45 south of the central drive)
Row length≈ 200 m (modules are 2.38 m long × 1.13 m wide; 90 × 2.38 m = 214 m plus motor + dampers)
Torque tubeRound 90 mm × 3 mm wall A572 Gr. 50 steel, HDG ASTM A123
Pile height (top of pile to torque tube)≈ 1.5 m above grade; module top at full tilt ≈ 5.0 m above grade
Tracking range±60° rotation
Backtracking±20° dynamic backtrack to eliminate inter-row shading at low sun angle
Stow strategyAuto-stow at wind gust > 22 m/s (50 mph) sustained, hail-stow tilt 60° for known forecast hail (NX HailPro option)
DriveSlew-drive worm gear, 24 VDC motor
Power source80 W self-powered panel per row, Li-FePO4 battery (no trenched control power)
SCADASelf-organizing 2.4 GHz RF mesh (NX Navigator) → site gateway → SCADA
2025-26 ASP≈ $0.10/W

Competitive alternates: Array Technologies DuraTrack HZ v3 1P (single-motor row up to 1.7 MW, central drive — fewer moving parts but no row-level slew flexibility) and Soltec SF7 (Spanish, strong in Latin America, balanced 1P design).

Bifacial gain depends strongly on tracker height: 1.5 m torque-tube clearance is the standard for 1P trackers. Going to 1.8-2.0 m (an option Nextracker offers) adds ~1-2% bifacial gain but adds pile length and structural cost.

Reference: photovoltaic-cells (tracker subsection covers single-axis vs. dual-axis tradeoffs, backtracking math, and the diurnal incidence-angle modifier curve).

8. Structural and foundation

The structure beneath each tracker row is a driven steel pile foundation system: 2-3 pile groups per drive segment, with W6×9 H-piles driven 1.8-2.5 m below grade depending on local SPT N-values, and 6 m spacing along the row for vertical load points.

Pile selection. W6×9 hot-rolled wide-flange (W-shape), ASTM A572 Grade 50 (Fy = 50 ksi / 345 MPa, Fu = 65 ksi / 450 MPa, low-alloy structural carbon steel). A572 Gr. 50 is the U.S. utility-solar standard pile material — high strength, good weldability, ductile failure mode, and cheap. Cross-section properties: A = 2.68 in² (1730 mm²), I_xx = 16.4 in⁴ (6.83 × 10⁶ mm⁴), S_xx = 5.56 in³ (91 × 10³ mm³), Z_xx = 6.23 in³. See structural-shapes and steel-grades for the underlying alloy and rolled-shape references.

Corrosion protection. Hot-dip galvanized to ASTM A123, minimum 80 µm Zn coating thickness for piles in soil. In the calcareous loam soil of Reeves County (resistivity > 5 000 Ω·cm, pH ~ 7.8, chlorides < 100 ppm) this gives a calculated 25-30 year service life at the in-soil section; above-grade portions have indefinite life under the same coating. Sacrificial Zn loss rate in this soil class is ~ 3 µm/yr. Reference: surface-treatments (HDG vs. cold-galv vs. duplex coating tradeoffs).

Load analysis. Per ASCE 7-22 Chapter 26-30 for wind, and the SEAOC PV2 Wind Design for Solar Arrays guidance document (2017, updated 2023) for tracker-specific aeroelastic loads:

  • Basic wind speed V = 110 mph (49.2 m/s), 700-yr MRI, Risk Cat I.
  • Exposure C (open country, no trees/buildings within 1 mi).
  • Velocity pressure q_h at h = 5 m: q_h = 0.00256 × K_z × K_zt × K_d × V² = 0.00256 × 0.85 × 1.0 × 0.85 × 110² = 22.4 psf (1.07 kPa).
  • Net wind pressure on tracker at design tilt (0° to ±60°): C_p range −1.7 to +1.7 from SEAOC PV2 Tables.
  • Peak uplift on a pile at the worst tilt: P_u = q_h × C_p × tributary area = 22.4 × 1.7 × (6 m × 1.13 m / 2 pile shares) = 22.4 × 1.7 × 36.5 ft² = 1 390 lbf (6.18 kN) uplift per pile, and similar downforce.
  • Driven-pile capacity calculated per Federal Highway Administration Driven Pile Foundations methods using SPT N-values: shaft friction in medium-dense sand gives ~ 30 kPa × pile perimeter × embedment depth = 30 × 0.5 m × 2.2 m = 33 kN ultimate uplift capacity. With FOS = 2.5: 13.2 kN allowable >> 6.18 kN demand. ✓
  • Aeroelastic check: tracker torsional galloping occurs above critical wind speed; Nextracker’s published wind-tunnel curve for the NX Horizon shows critical V_cr ≈ 31 m/s at 0° tilt and 19 m/s at 30° tilt — auto-stow at 22 m/s flat stow keeps us below all critical points.

Aeroelastic is the design flag worth highlighting — solar tracker failures in 2015-2020 were dominated by torsional-galloping aeroelastic instabilities at intermediate tilts, not by static wind. Modern trackers manage this with mass dampers (NX Horizon) or active stowing.

Pile-driving execution. A purpose-built solar pile driver (PVH Vermeer PD10 or similar) installs ≈ 80-120 piles per shift per machine. With 28 000 total piles (175 000 modules / ~ 6 modules-per-pile-share), the install at 3 machines × 100 piles/shift × 5 shifts/week = 1 500 piles/week, completing the pile campaign in ~ 19 weeks. Piles are surveyed to ± 25 mm lateral and ± 2° plumb tolerance per Nextracker installation spec; any reject is pulled and re-driven 0.5 m offset. A pre-construction pile load test (typically 1 % of total = ~ 280 piles tested per ASTM D3689 axial tension or D1143 axial compression) validates the geotech-design uplift capacity before bulk install.

Galvanic corrosion management. Where dissimilar metals contact — galvanized steel tracker brackets meeting anodized 6063-T6 Al module frames — we specify EPDM gasket isolation or stainless-steel fasteners with EPDM washers to prevent direct Al-Zn galvanic coupling. The galvanic series places Zn anodic to Al, so without isolation the Zn coating sacrifices preferentially around the fastener point, eventually leaving the Al frame to corrode. Field surveys of 2015-vintage installations without isolation show ≈ 5-10% module-frame perimeter corrosion at 10 yr — fully mitigated in current installs.

Snow and ice loading. Reeves County winter snow load per ASCE 7-22 Figure 7.2-1 is ≤ 5 psf (0.24 kPa) ground snow — non-controlling for the tracker. Ice loading per Chapter 10 also non-controlling. The relevant winter weather risk is freezing-rain accretion on tracker drive components; mitigation is the standard cold-weather grease specification on slew-drives and a cold-stow position (45° tilt) initiated when forecast freezing precipitation is expected.

Soil-mechanics and structural-analysis references: structural-analysis and soil-mechanics.

9. DC cabling

DC cabling runs from each module’s MC4 leads through string conductors to combiner boxes, then in trunk cable to the central inverter. We use IEC + UL dual-rated 1.5 kV DC PV wire throughout.

String wire. 6 AWG (13.3 mm²) Cu, single-conductor, sunlight-resistant, dual-rated UL 4703 (PV Wire, USA) and TÜV H1Z2Z2-K 1.5 kV DC (EU). XLPE insulation (cross-linked polyethylene) with halogen-free, low-smoke jacket. Temperature rating: 90 °C wet, 150 °C dry. Cu chosen over Al at the string scale because of MC4 connector ratings and the short, lower-current runs where bend radius matters more than $/m.

Ampacity check: 6 AWG Cu in conduit/free air at 30 °C ambient has 75 A rating; derated for 60 °C ambient and bundling = ~ 53 A. String design current (1.25 × 1.25 × Isc per NEC 690.8) = 1.5625 × 14.99 = 23.4 A. 5.3 mV/m voltage drop at 18.7 A; for a 25 m string run, V_drop = 0.13 V (0.013%) — negligible. ✓

MC4 connectors. Stäubli MC4-Evo 2 (genuine, not low-cost Chinese MC4-compatibles, which are a documented field-failure source). Rated 1.5 kV DC, 30 A, IP68 mated, 25-yr UV-rated EPDM seal. See connector-families (MC4 ecosystem, mating compatibility, and the long history of fire incidents traced to cross-mated MC4 lookalikes — UL 6703 now requires certified mating pairs).

In-line string fuses. 25 A gPV (Mersen HelioProtection series or Eaton Bussmann PV-25A14F), 1500 V DC rated, sized between Isc × 1.25 (= 18.7 A) and conductor ampacity (53 A) per NEC 690.9. The 25 A fuse coordinates with the 30 A MC4 limit and protects against reverse-current faults from parallel strings.

Combiner boxes. Shoals Big Lead Assembly (BLA) + iCB combiner or Bentek BOS combiner: 24 strings in, single 600 kcmil out, integral fused inputs, type-2 SPD (surge protective device), monitoring CT per string. Combiner-to-inverter trunk: 600 kcmil (304 mm²) Al, XHHW-2 + LSZH armor, 1500 V DC rated, direct-buried with 30 mil PVC. Al at the trunk because cost (/kAh-per-meter. See copper-alloys for the Cu side and the Al alloy reference at the trunk.

Trunk ampacity at 600 kcmil Al = 380 A in conduit at 75 °C, 25-30% derated for bundling and 50 °C ambient → 270-280 A allowable. Per combiner trunk = 24 strings × 18.7 A = 449 A — too high for one 600 kcmil run. We use two 600 kcmil parallel sets per combiner (or equivalently use 4 × 350 kcmil parallel sets), or split the 24-string combiner into 2 × 12-string combiners. The chosen layout uses 15 × 24-string combiners per inverter with paired-trunk runs.

DC arc-fault protection. NEC 690.11 requires PV systems > 80 V_DC to have arc-fault circuit interrupter (AFCI) protection. At utility scale this is integrated into the central inverter’s DC input section: each MPPT input has a current-signature monitor that detects the 1-100 kHz noise band signature of a series arc, tripping the DC contactor within 2.5 s per UL 1699B. Combiner-box arc detection is layered on top for early annunciation. False-trip rate is the bigger field problem than missed-trip rate — modern signature algorithms (Sungrow’s third-gen AFCI) cut false trips to < 1 event per inverter per year.

Grounding architecture. The 1500 V DC system is functionally grounded (not solidly grounded — this distinction matters for NEC 690 article structure since 2017). Each inverter has a residual-current monitoring (RCM) device and a ground-fault detection-and-interruption (GFDI) circuit on the DC side. Equipment grounding conductors (EGCs) bond all metal — module frames, tracker torque tubes, combiner enclosures, pile heads — via 6 AWG bare Cu running continuously along each row, bonded to driven ground rods every 30 m. Step + touch potential at the array is computed using the IEEE 80 methodology and verified by Wenner 4-probe resistivity survey at the geotech phase.

10. AC and MV collection

Each Sungrow SG4400UD-MV-US discharges its AC power at 645 V three-phase from the inverter bridge directly into the integrated 4.5 MVA cast-resin / oil-filled (designer’s choice — we picked cast-resin dry-type for fire safety and remoteness) step-up transformer with primary 645 V Δ / secondary 36 kV Y. The 36 kV winding is grounded through a low-impedance grounding resistor on the substation side (system grounding scheme: low-impedance ground per IEEE 142).

MV collection cable. 36 kV (35 kV class) MV-105 EPR-insulated, Al stranded, copper-tape shielded, jacketed, 240-630 mm² depending on feeder loading. Underground at 1.0 m burial depth with red marker tape at 0.3 m, sand bedding 100 mm above + below cable, native backfill above. Direct-buried where feasible; PVC conduit at road crossings.

Radial feeder topology. The 18 inverter skids are grouped into 3 radial feeders × 6 inverters each, each feeder rated 27 MVA = 6 × 4.5 MVA. Cable size per feeder segment (using IEC 60287 ampacity in soil at 20 °C native temp, depth 1.0 m, thermal resistivity 1.0 K·m/W):

Feeder segmentLoadCable size
6th-to-5th inverter4.5 MVA240 mm² Al
5th-to-4th9.0 MVA240 mm² Al
4th-to-3rd13.5 MVA400 mm² Al
3rd-to-2nd18.0 MVA400 mm² Al
2nd-to-1st22.5 MVA630 mm² Al
1st-to-substation27.0 MVA630 mm² Al (or 2 × 400 mm² parallel)

Total MV cable: roughly 30 km across the three feeders. Cable installer: trenched directly behind the tracker installation crew. Reference: transformers-power-systems (the integrated step-up tx is functionally a small distribution transformer; the system grounding and the YNd vector group choice are detailed there).

11. Substation and interconnect

The collector substation steps 36 kV up to 138 kV and serves as the point of interconnection switchyard. The footprint is ≈ 0.5 acres, fenced (8 ft chain-link + barbed wire + ground-grid per IEEE 80), gravel-surfaced, with a 25 ft × 30 ft control building.

Main step-up transformer. 100 MVA ONAF, 138 kV / 36 kV, Dyn1 vector group, GE Prolec or Hitachi Energy (the two North American utility transformer suppliers with mature 100 MVA Class III medium-voltage offerings). 65 °C average winding rise (ANSI C57.12), bushings 145 kV BIL on HV side / 200 kV BIL on LV side, OLTC ±10% × 17 steps on HV winding. Impedance ~ 10% on 100 MVA base. Oil-filled (mineral oil GES B or natural-ester FR3 for the environmental and fire-rating bonus). Conservator + Buchholz relay + sudden-pressure relay + RTDs + fans + bushing CTs.

Long-lead procurement: 18-24 months from PO to factory acceptance test (FAT) — this is the schedule-binding equipment for the entire project as of 2025-2026. Procurement was initiated at NTP-18 mo.

138 kV switchgear. Outdoor air-insulated switchgear (AIS) breaker bay. Single 138 kV SF6 circuit breaker (or SF6-free ABB AirPlus vacuum + clean-air-mixture breaker, our pick for the 2026-onward EU F-gas regulation tailwind and the IRA bonus for fluorinated-gas alternatives). Disconnect switches, line-side and bus-side. Three-phase metering CTs (revenue grade, 0.2S accuracy class) and capacitive voltage transformers (CVTs) per IEEE C57.13.

Protection. Differential protection (87T) on the main tx via SEL-787; line protection (21/67/50/51) on the 138 kV tie line via SEL-411L; bus protection on the 36 kV bus via SEL-487B. All relays per IEEE C37.90 / C37.118 (synchrophasors). DNP3 over fiber back to the SCADA hub. Auxiliary 125 V DC battery + charger per IEEE 485 sized for 8-hr backup.

Auxiliary loads. Station service from a 250 kVA aux tx (36 kV → 480/277 V) plus 100 kW battery-backed UPS for SCADA, comms, and trip coils.

Surge arrestors. Metal-oxide varistor (MOV) arrestors at every transition: 138 kV phase-to-ground arresters at line terminations and tx HV bushings, 36 kV arresters on the LV side and on each feeder dead-end.

Substation grounding. Buried 4/0 Cu ground grid per IEEE 80, 6 m × 6 m mesh, with driven 3 m × 5/8” ground rods at corners. Calculated step + touch potentials: step 1 100 V, touch 480 V, both below the IEEE 80 limits at 0.5 s clearing time for a 56 kA fault.

Reactive power and voltage support. The PPA and the GIA both require the plant to provide reactive support at the POI across the 0.95 leading to 0.95 lagging power factor range at full real power output, expanding to 0.90/0.90 at partial output. Each Sungrow inverter provides up to ±0.31 pu Q (~ 1.36 MVAR per skid) on top of the 4.4 MW real-power dispatch. With 18 inverters this gives ±24.5 MVAR plant-level Q capability, more than the ERCOT-required ±20 MVAR at the POI. Voltage at the POI is regulated via a closed-loop plant controller (Power Plant Controller, PPC) running on the same SCADA server, executing a droop characteristic with 1-2 s response time. No separate STATCOM is required at this scale.

Auxiliary station service detail. The 250 kVA station service tx feeds: SCADA cabinets (3 kW), comms (1 kW), substation lighting + HVAC (8 kW), tx oil pumps + fans + tap-changer drives (12 kW peak), security/perimeter (5 kW), 138 kV breaker compressor + motor charging (4 kW). Total auxiliary nameplate ~ 35 kW continuous, with 100 kW UPS coverage for control + protection + comms during a station-service outage. Black-start scenarios are not applicable — the plant does not provide grid black-start (would require a dedicated battery + grid-forming inverter retrofit).

References: transformers-power-systems (transformer design, impedance, vector groups) and standards-bodies (IEEE, IEC, ASTM cross-references).

12. Grid interconnection studies

ERCOT generation interconnection is governed by the ERCOT Nodal Operating Guides Section 5 and Nodal Protocols Section 6. Three studies sequence:

1. Interconnect feasibility / cluster study. Replaces what FERC LGIA jurisdictions call the System Impact Study. Models the proposed generator in PSS/E or PowerFactory at the ERCOT-system level. Outputs: thermal loading on adjacent lines, voltage profile impact, short-circuit duty increase at neighboring buses. Our result: thermal loading peaks at 78% on the 138 kV tap line under N-1 contingency, voltage at the POI rises 0.3% at full export — both within tolerance, no additional grid upgrades beyond the dedicated 3 mi tap line and breaker bay.

2. Facilities study. Engineering-grade scope and cost estimate of the network upgrades. 1.8M for the new 138 kV breaker bay at Reeves Switch, 0.3M for protective relay coordination upgrades.

3. Generator interconnection agreement (GIA). ERCOT-equivalent of FERC pro-forma LGIA, executed with Oncor as the transmission service provider (TSP). Standardized terms; the negotiable items are the upgrade cost-allocation and the commercial operation date (COD).

4. Dynamic study suite. ERCOT requires positive-sequence PSS/E + electromagnetic-transient PSCAD models matching the as-built plant. The PSCAD model is a manufacturer-provided “black box” of the inverter (Sungrow ships a vendor-blackboxed PSCAD library), plus a user-modeled balance of plant (transformers, cables, plant controller, met-stations, protection). ERCOT engineers run small-signal stability, sub-synchronous control interaction (SSCI), and weak-grid (SCR < 3) studies before sign-off. The SSCI screen has become particularly important after the 2021 Odessa events showed IBR controls can interact with series-compensated lines.

Operating studies and compliance models. PSS/E small-signal stability model (positive-sequence, full plant aggregated), PSCAD electromagnetic transient (EMT) model (required by ERCOT for IBR — inverter-based resources — since the Texas 2021 winter event clarified IBR ride-through gaps), and a load-flow model. Studies confirm:

  • Short-circuit MVA at POI post-interconnection: 1 450 MVA. Inverter contribution is current-limited to 1.2 pu (~96 MVA equivalent) — much lower than synchronous machines, which is why short-circuit duty doesn’t materially rise.
  • IEEE 1453 flicker: P_st < 0.35, P_lt < 0.25 — well below the 1.0 / 0.8 limits.
  • IEEE 519 harmonics: Total demand distortion (TDD) at POI = 2.1%, well below the 5% limit at this voltage class.
  • Voltage regulation: Q-V droop characteristic per ERCOT VAR curve, voltage support 0.95 leading to 0.95 lagging at 100% real power, expanding to 0.90 lead / 0.90 lag at < 80% real power.
  • LVRT/HVRT: Inverters ride through 0% nominal voltage for 150 ms, recover within 1 s. HVRT at 1.2 pu for 200 ms. Per IEEE 1547-2018 Category III.
  • Anti-islanding: UL 1741-SB tested, passive (under/over voltage + frequency) + active (positive-feedback Sandia drift) anti-island detection within 2 s per IEEE 1547-2018.

ERCOT also requires a primary-frequency-response (PFR) capability — 4% droop on a 5% deadband — for any generator > 10 MW. The Sungrow inverters provide this via the fast-frequency-response curve programmed at commissioning.

13. Civil works

Roughly 2.5 million ft² (≈ 230 000 m²) of disturbance footprint. Sequence:

Clearing + grubbing. Vegetation (mostly mesquite and creosote) cleared to 6 in below grade, root mass shredded and stockpiled for revegetation. Burning prohibited under TCEQ rules; chipping + windrowing instead.

Grading. Cut/fill balanced to ± 5 000 yd³ across the site. Maximum grade change ≈ 1 m. Smooth-drum vibratory roller compaction to 95% modified Proctor.

Access roads. 20 ft (6.1 m) compacted gravel access road with 4 in (100 mm) base course over 6 in (150 mm) sub-base, geotextile separator (Mirafi 500X) at subgrade interface, 2% crown. Total ≈ 7 mi (11 km) of internal roads serving each tracker block + perimeter ring road + substation access. Designed for a 100 kip (450 kN) gross-vehicle-weight truck for transformer delivery.

Stormwater. Site is in the Pecos River Basin. Stormwater pollution prevention plan (SWPPP) per EPA NPDES Construction General Permit (CGP) 2022 and the TCEQ TXR150000 permit. BMPs:

  • Silt fence on the downhill perimeter (2 100 lf total).
  • Rock check dams in any concentrated flow path.
  • Stabilized construction entrance (50 ft × 20 ft of 3-6 in clean rock).
  • Inlet protection on the substation drainage.
  • Final stabilization via reseeding with a native short-grass mix (blue grama + buffalograss + sideoats grama) at 12-15 lb PLS/acre.
  • A 25-yr, 24-hr storm event peak detention basin sized at 0.85 ac-ft (≈ 1 050 m³) for the substation drainage; the rest of the site sheet-flows through the buffer.

Fencing + security. 8 ft chain-link perimeter fence, 3-strand barbed wire top, 24 access gates with key-card locks. Wildlife crossings (12-in × 18-in cuts at grade every 300 ft) per Texas Parks and Wildlife Department guidance for lesser prairie-chicken and pronghorn.

Concrete pads. Cast-in-place reinforced concrete for the substation (3 500 psi / 24 MPa f’c, #5 rebar grade 60), inverter skid pads (4 in / 100 mm slab, mesh reinforced), and met-station tripod pads. Total concrete ≈ 350 yd³ (270 m³). Cure 28 days before equipment set; substation cabling pulled through underground concrete-encased PVC duct bank.

Construction water. ≈ 8 000 gal/day average for dust suppression on access roads + concrete + compaction water, sourced from a temporary 6 in (150 mm) groundwater well drilled at the site. Permit through Texas Water Development Board (TWDB); the Pecos Valley alluvial aquifer at 80-120 ft depth has adequate yield. Post-construction the well is plugged and abandoned per TWDB Rule 76 (or repurposed as a permanent O&M water source for module washing).

References: transportation-engineering (gravel road design, base + sub-base, geotextile) and environmental-engineering (SWPPP, NPDES CGP, BMPs).

14. Performance modeling

We model the plant in PVsyst 7.4 (and cross-check in HelioScope), running 8 760 hourly simulation on the NSRDB PSM3 typical meteorological year (TMY3) and on 20 years of measured-year data (P50/P75/P90/P99 yields).

Loss cascade (annual, from STC nameplate down to AC busbar):

Loss componentValue
GHI to plane-of-array (POA) — single-axis tracker, GCR 0.36, lat 32°+28.5% (tracking gain)
POA irradiation2 700 kWh/m²/yr
Soiling−3.0% (TX, 0.6%/wk, weekly cleaning ROI insufficient — accept)
IAM (incidence angle modifier) on glass−2.5%
Module temperature (avg cell 38 °C, derate at −0.29%/°C)−3.8%
Spectral correction+0.3%
Low-irradiance behavior−0.5%
Module mismatch + LID + LeTID−1.8%
DC ohmic−1.2%
Inverter clipping (DC:AC = 1.25)−2.0%
Inverter conversion (avg 98.6%)−1.4%
Transformer (LV → MV, 1.0%) + collection (1.5%)−2.5%
Auxiliary loads (trackers, SCADA, station service)−0.8%
Bifacial gain (back-side at 30% bifaciality, elevated 1.5 m torque tube, albedo 0.25 sandy soil)+7.5%
Availability (95-yr 1 % unscheduled, scheduled maint outside production hours)−0.5%

Net P50 specific yield: 2 500 kWh/kWp_DC/yr → 100 000 × 2.5 = 250 GWh/yr. P90 specific yield (10th-percentile bad year) ≈ 2 350 kWh/kWp/yr → 235 GWh.

Net capacity factor at AC: 250 GWh / (80 MW × 8 760 h) = 35.7% on DC basis, or 250 / (80 × 8 760) = 35.7% on AC basis. Wait — recheck: 80 MW × 8 760 h = 700.8 GWh annual energy if at 100% AC capacity factor; 250 / 700.8 = 35.7%? That’s too high for utility solar. Let me redo: AC capacity factor = AC energy / (AC nameplate × 8 760). We’re exporting 250 GWh through 80 MW of AC = 250 / 700.8 = 35.7%. The reason this exceeds 28-30% is the 1.25 ILR — the DC array is 100 MW so the underlying DC capacity factor is 250 / (100 × 8 760) = 28.5%, which matches the design target. The 80 MW AC export apparent capacity factor is inflated by oversizing the DC.

Bifacial gain merit warrants explanation: the back-side gain is the product of (a) view factor of the rear of the module to the ground, (b) ground albedo, (c) row-to-row diffuse contribution. For a 1.5 m torque-tube height, GCR 0.36, ground albedo 0.25 (sandy/calcareous), and 30% bifaciality factor, PVsyst computes a net back-side annual contribution of 7.5% on top of the front-side baseline. Increasing to 2.0 m torque-tube + 0.35 albedo (gravel coating) would lift this to ~ 11%, but the marginal capex doesn’t pencil at this PPA price.

Diurnal and seasonal yield shape. Single-axis tracking flattens the daily generation profile relative to fixed-tilt: peak power output is held within 95% of nameplate from ~ 10:00 to ~ 14:00 local solar time, vs. fixed-tilt’s narrower 11:30-12:30 peak. This is also the inverter-clipping window — at DC:AC 1.25, clipping occurs only on the very clearest summer days for ~ 1-2 hours around solar noon, contributing the 2.0% clipping loss line. Seasonally, the December solstice yield is ~ 60% of June solstice yield at this latitude with tracking (vs. ~ 45% with fixed-tilt), giving a flatter monthly generation profile useful for matching against the ERCOT West winter-peak load shape.

P50 vs. P90 financial implication. Project finance lenders typically size debt against P99 (one-year-in-100 bad year, ≈ 90% of P50) or P90 (one-year-in-10, ≈ 94% of P50). Sponsor equity returns are modeled against P50. The P50-to-P90 gap of 6% on revenue translates to roughly a 0.10× swing in DSCR — material at our 1.36 baseline, so the term sheet sizes debt to a P90 DSCR of 1.20 and a P50 DSCR of 1.30.

15. SCADA and monitoring

The SCADA architecture is layered: data acquisition at the device level, aggregation at the gateway, historization and analytics at the DAS layer, dashboarding and alarming at the operator UI.

Field comms. All Sungrow inverters expose Modbus TCP on a fiber-ring (single-mode 9/125 µm) Ethernet network from each skid back to the substation. Trackers communicate over Nextracker’s 2.4 GHz RF mesh to a site gateway, which then bridges to TCP/IP. The on-site meteorological station (Kipp & Zonen CMP21 pyranometer + CHP1 pyrheliometer + temp + RH + wind 2D ultrasonic) reports over Modbus RTU → Ethernet.

Fiber backbone. A 24-strand outdoor-rated armored single-mode fiber runs in the same trench as the MV cable, dropping 4 strands at each inverter skid (2 in-use, 2 spare). Total fiber length ≈ 30 km, terminated at SC/APC connectors in NEMA 4X fiber distribution panels. Network topology is a self-healing dual-ring (East and West) so a single fiber cut takes 50 ms to re-converge via RSTP without dropping inverter telemetry. Each inverter sub-network is on its own VLAN; the substation aggregation switch (Cisco IE-4010 or RUGGEDCOM RX1510, both DNP3-aware industrial switches) inter-VLAN routes only the SCADA traffic to the DAS server.

Met-station and irradiance reference cells. In addition to the central met-station, two satellite met-stations are deployed at array corners (so the 250-acre site has 1 central + 2 satellite met-stations). Each is a 10 m tripod with redundant pyranometers, plus a back-of-module temperature sensor on three reference modules per met-station. Irradiance data is the single most important data input for PR calculation and the PVsyst-vs-actual reconciliation; ISO 9847 + IEC 61724 require ≤ 3% measurement uncertainty for utility plant performance reporting.

Revenue metering. A Schneider ION8650 revenue-grade meter (ANSI C12.20 0.2% class) at the 138 kV side of the main tx provides the billing energy reading. Backup meter on the 36 kV bus, both with redundant CT and PT cores.

DAS layer. DASNet, AlsoEnergy PowerTrack, or Power Factors PowerHub (we selected AlsoEnergy for the ERCOT-zone customer base and the Day Ahead Market integration). 1-min granularity for inverter, 5-min for environment, 15-min for revenue. 7-year retention.

Dashboards + alarms. Web-based operator UI with PR (performance ratio), yield, availability, soiling, weather, and equipment-fault dashboards. Alarm escalation matrix: tier-1 (auto-recover, log only), tier-2 (notify O&M during business hours), tier-3 (page on-call 24/7), tier-4 (immediate utility coordination — e.g., breaker trip).

Cybersecurity. The plant exceeds 25 MVA, so it falls under NERC CIP-002 through CIP-014 as a Medium Impact BES Cyber System. Required controls: CIP-005 electronic security perimeter (DMZ between corporate VPN and operational tech), CIP-007 patch management, CIP-008 incident response, CIP-010 configuration change management, CIP-013 supply-chain risk. Substation physical security per CIP-014 (post-Metcalf 2013 — ballistic-resistant control building wall on the side facing public access). The OT network is air-gapped from corporate IT via a unidirectional gateway (Waterfall or Owl) — vendor remote-access goes through a jump-server with MFA + session recording.

Reference: engineering-codes (NERC CIP, IEEE C37.118, IEC 61850 — substation automation protocols).

16. O&M

Utility-scale solar O&M is a mature commodity service at 400-560k/yr for this plant. Scope from a full-service provider (Clearway O&M, NovaSource, EDF Renewables Services, SunSystem Technology):

Preventive. Quarterly visual inspection of all 175 000 modules + tracker rows + cabling + combiners. Annual IR-thermography survey of inverters + tx + combiners (handheld FLIR or drone-mounted FLIR Vue Pro R). Annual ground-grid resistance test. Semi-annual tracker greasing (slew-drive worm + bearings). Five-year inverter capacitor-bank check. Tx oil DGA (dissolved gas analysis) annual.

Performance. Daily PR monitoring with automatic underperformance alerts. Weekly review of dirt-rate trend (soiling station = clean module + dirty module reference pair). Monthly PR variance reports vs. PVsyst expected. Quarterly availability + losses reconciliation.

Soiling and cleaning. 0.6%/wk soiling rate in West Texas; rainfall events recover ~40-60% of accumulated soiling. Cleaning economics: at 62 500/yr value. Manual cleaning costs 5.3-8.8k per cleaning pass; ROI marginal. Robotic cleaning (Ecoppia E4 — rail-mounted brush per row; SunPure — autonomous mobile) at 30/MWh but marginal at $25 PPA. Final call: no robotic cleaning, opportunistic manual cleaning post-haboob events with a hydro-jet truck.

Vegetation. Sheep grazing under contract with a local rancher at $200/acre/yr — far cheaper than mowing and provides ESG narrative. Six grazing rotations per year on a fenced rotational pattern. Mowing only at the perimeter and around equipment pads.

Pest and wildlife. Burrowing mammals (kangaroo rats, ground squirrels) chew DC wire insulation if cabling is loose at grade — direct-burial depth standard mitigates this. Bird nesting on combiner boxes and tracker drives is managed with bird-deterrent caps. Snake-resistant skirting at the perimeter of inverter skids prevents rattlesnake nesting in the cooling-fan plenum. Rangeland birds-of-prey (red-tailed hawk, golden eagle) use tracker rails as perches — beneficial for rodent control.

Module washing water consumption. Manual washing post-haboob uses a hydro-jet truck with 2 000 gal tank and DI-water at 800 psi at the nozzle: water consumption ≈ 0.05 gal/module per wash × 175 000 = 8 750 gal/wash event. Two events per year average → 17 500 gal/yr — trivially small for the regional water budget, no Texas Water Code permit required since use is < 25 ac-ft/yr.

Component life and reserve. Modules: 30-yr warranty, expected end-of-life by year 30 at 87% nameplate. Inverters: SiC and IGBT modules typically replaced at 10-15 years; DC capacitors and AC contactors at 7-10 years. Trackers: slew-drives at 20+ years, motors and TCUs at 10-15 years. Transformer: 30-40 years.

End-of-life. Decommissioning plan filed at COD per Texas SB 1281 (2024) requires either a surety bond or an irrevocable letter of credit covering removal cost net of salvage. Salvage value of Cu, Al, Si, and steel covers ~ 60-70% of decommissioning labor. Module recycling at end-of-life per SEIA National PV Recycling Program and IRENA guidance — emerging market for crushed Si glass + recovered metals; expected 5-15 by 2050 (today $20-30).

Spares strategy. Stock on site: 1% module spares (1 750 modules in two climate-controlled containers), 5 string-fuse cases (500 fuses), 2 combiner-box recombiner sub-assemblies, 12 tracker slew-drive replacements, 24 tracker control units (TCUs), 6 inverter cooling-fan packs, 1 spare IGBT power-module set per inverter family, and 1 mobile substation 5 MVA emergency feed for tx failure. Critical-spare hold for the main 100 MVA tx is uneconomic ($2-3M idle inventory); instead, a service-level agreement with GE Prolec / Hitachi Energy provides priority remanufacture at < 12 weeks vs. 18-24 mo for new-build, paired with 60 MW of curtailed-but-energized operation through the surviving feeders during the outage.

17. Financial

Capital cost (full breakdown in Section 21): 0.90/W_DC = $1.13/W_AC all-in.

Capital stack:

  • Debt: 65% × 58.5M**, 18-year amortization, 5.5% all-in (SOFR + 200 bps), DSCR covenant ≥ 1.30. Term sheet from a project-finance lender (e.g., NORD/LB, CoBank, ING) is straightforward at this scale and tenor.
  • Tax equity: 30% × 27M**, monetizing the ITC + bonus credits + 5-yr MACRS depreciation. Tax-equity yield target 7-9% after-tax IRR.
  • Sponsor equity: 5% × 4.5M**.

ITC and IRA bonus credits. The Inflation Reduction Act of 2022 § 48 ITC base rate is 30% of qualified basis when wage-and-apprenticeship requirements are met (which we meet). Bonus adders potentially stack:

  • Domestic content bonus (+10%): requires 100% U.S. steel/iron and a phased-in share (currently 40%, rising to 55% by 2027) of manufactured products. We qualify steel piles + tracker structure (A572 Gr. 50 from a U.S. mill) but the modules are Southeast Asian — we do not claim domestic content for the modules, and we calculate the manufactured-products share including U.S.-made tracker components + U.S.-made inverter (Sungrow assembles in Arizona for the U.S. market). Marginal — claim only if our cost-basis math clears 40%.
  • Energy community bonus (+10%): we qualify because the site is in an oil-and-gas-economy county (Reeves County, > 0.17% direct employment in fossil-fuel extraction per the IRS Notice 2023-29). Claim.
  • Low-income community bonus (+10% to +20%): site is not in a designated low-income census tract or Indian land. Do not claim.

Effective ITC: 30% + 10% (energy community) = 40% = $36M of monetizable tax credit (we model the conservative case without domestic content).

Revenue. 250 GWh × 6.25M/yrenergy revenue + 250 000 RECs × 0.50M/yr REC revenue → **~0.5M/yr O&M + 0.2M/yr property tax + 5.65M/yr operating cash flow. Debt service 4.9M/yr. DSCR = 5.65 / 4.9 = 1.15 — too thin. Re-lever to 55% debt = 4.16M/yr → DSCR = 1.36. ✓

Project IRR (unlevered, post-tax, 25-yr): ~ 7.5% with the ITC monetized. Sponsor equity IRR (levered): ~ 11-13% depending on tax-equity flip date and back-end residual value.

Sensitivity table (sponsor IRR vs. key variables):

VariableLowMidHighIRR impact
PPA price ($/MWh)$22$25$28±2.5%
P50 yield (kWh/kWp)2 3502 5002 600±1.8%
Capex ($/W_DC)$0.85$1.00$1.15∓2.1%
Debt rate (SOFR + bps)4.5%5.5%6.5%∓1.4%
ITC bonus stack30%40%50%±1.6%

Tax-equity flip economics: the tax-equity investor funds 36M cash-equivalent via § 6418 transferability sale, monetized 0.92× = $33M cash to sponsor) plus 5-yr accelerated depreciation pass-through. The investor reaches its 7.5% target IRR at year 5-6, at which point ownership “flips” to 95%-sponsor / 5%-tax-equity for the back-15 years of the PPA. The IRA § 6418 direct-transfer of ITC has displaced about half of the historical partnership-flip tax-equity transactions since 2023, simplifying the capital stack.

18. Codes and standards

A short non-exhaustive list of governing codes:

  • Building / structural: International Building Code (IBC) 2024, ASCE/SEI 7-22 (Minimum Design Loads), AISC 360-22 (Steel Construction), AWS D1.1 (Structural Welding).
  • Electrical: NFPA 70 National Electrical Code (NEC) 2026 — Article 690 (Photovoltaic Systems), Article 691 (Large-Scale PV), Article 705 (Interconnection); UL 61730-1/-2 (PV module safety, replaces UL 1703); UL 3741 (PV Hazard Control / Rapid Shutdown System Equipment); UL 1741-SB (Inverter testing for grid support functions).
  • Grid: IEEE 1547-2018 (Interconnection of Distributed Energy Resources), IEEE 1547.1-2020 (test procedures), IEEE 1453 (flicker), IEEE 519 (harmonics), IEEE C57.12 (transformer construction), IEEE C57.13 (instrument transformers), IEEE 80 (substation grounding), IEEE 142 (system grounding), IEEE C37.x series (relaying and breaker control).
  • NERC/FERC: NERC CIP-002 through CIP-014 (cybersecurity), NERC PRC-005 (protection system maintenance), NERC FAC-001/-002 (facility connection requirements); FERC pro-forma Large Generator Interconnection Agreement (LGIA) and Small Generator Interconnection Agreement (SGIA) — though we’re in ERCOT and use the ERCOT GIA, the underlying contract follows the FERC-modeled template.
  • Environmental: NEPA (not applicable here, no federal nexus), ESA (Section 7 consultation), CWA (Section 404 jurisdictional wetlands), NHPA (Section 106), Clean Air Act fugitive dust permits during construction (Texas Commission on Environmental Quality TCEQ Air Quality Standard Permit).
  • PV product: IEC 61215 (crystalline Si module design qualification + 25 mm hail test), IEC 61730-1/-2 (safety), IEC 62109-1/-2 (inverter safety), IEC 62548 (PV array design).

Owner’s engineer (OE) and independent engineer (IE). Two parallel engineering review roles separate from EPC: the OE represents the project owner throughout design + construction + commissioning, reviewing every submittal, witnessing factory acceptance tests, and signing off on substantial completion. The IE represents the lender, performing a one-time pre-financial-close technical due diligence (a 150-200 page red-line report on resource, EPC contractor, equipment selection, contracts, and projections) plus periodic in-construction site visits. Both roles together cost ~ $0.6-1.0M and are non-negotiable in any debt-financed project of this scale. Selecting OE/IE firms with prior ERCOT-IBR experience (DNV, UL Solutions, Black & Veatch, Burns & McDonnell, Stantec) is itself a risk-mitigation lever.

Permitting summary. Local: Reeves County special-use permit (planning-and-zoning hearing, ~ 4-6 mo). State: Texas Commission on Environmental Quality (TCEQ) air quality standard permit for construction fugitive dust, TPDES Construction General Permit (CGP) Notice of Intent for stormwater, Texas Historical Commission (THC) Section 106 review certification, Texas Water Development Board well permit if applicable. Federal: U.S. Fish and Wildlife Service informal Section 7 consultation, FAA Form 7460-1 if any tracker structure exceeds the imaginary surface around any airport (not applicable here, nearest airport is 18 mi away). FERC: not jurisdictional in ERCOT.

Cross-reference: engineering-codes consolidates the U.S. code/standards landscape.

19. Schedule (24 months)

A representative Gantt skeleton, with critical-path callouts:

MonthPhaseKey milestones
−6 to 0 (pre-NTP, dev phase)DevelopmentERCOT cluster study + facilities study + GIA, county special-use permit, land lease/purchase, PPA execution, environmental clearances, geotech, preliminary engineering, tax-equity term sheet, debt term sheet
1-4ProcurementModule reservation (12 mo lead), transformer PO (18-24 mo lead — issued at month −6 critical-path), inverter PO (12 mo lead), tracker PO (8 mo lead), MV cable, BoP
5-12Construction (civil + mech + electrical)Mobilization, clear/grub/grade (6 wk), road build (4 wk), pile drive + tracker assembly (16 wk in parallel — 6 250 strings × 28 modules each, 4 modules per crew-hour at 60 crews = 12 wk), DC + MV cabling (12 wk), substation civil (8 wk), substation electrical + tx delivery (6 wk on top of long-lead arrival)
13-18CommissioningPre-energization checks, witnessed start-up, ERCOT registration, PSCAD model handoff, performance test (PVsyst-vs-actual 14-day reconciliation), substantial completion
19-22Punch + rampPunch-list resolution, soiling baseline, full O&M handover, lender mechanical-completion certificate
23-24CODFinal reconciliation, tax-equity funding, sponsor distribution

Critical path is the main transformer at 18-24 months from PO. The transformer PO is therefore issued at month −6 of the development calendar so that delivery hits month 12-13, in time for the substation construction window. Any module / inverter / tracker delays float against the tx schedule.

Construction labor. Peak workforce ~ 280 crew during the month 6-10 mech/electrical peak: ~ 150 mechanical (pile-drive crews + tracker assembly + module install), ~ 60 electrical (DC cabling + combiner + trunk + MV + substation), ~ 30 civil (roads + grading + fence + substation pad), ~ 20 EPC management + QA/QC, ~ 20 commissioning + start-up engineers from inverter/tracker OEMs. Most labor sourced from regional union halls (IBEW Local 583, LIUNA Local 528) at Davis-Bacon prevailing wage rates — required to qualify for the full 30% ITC under IRA wage-and-apprenticeship rules. Apprentice-hour requirement: 12.5% of total labor hours by registered apprentices in 2024-2025, rising to 15% in 2026 onward.

Productivity benchmarks. Module install rate at peak: 60 crews × 4 modules/crew-hour × 10 hr/shift × 5 shifts/wk = 12 000 modules/week. Pile-drive rate: 1 500 piles/week (3 machines). Tracker assembly rate: 100 rows/week (~ 9 000 modules of tracker structure). Together these synchronize the mech install to a ~ 15-16 week window. Electrical install lags mech by 2-3 weeks per block and proceeds in parallel.

Commissioning sequence. Energize the substation in stages: (1) station service tx + control building first, (2) cold-energize the main 100 MVA tx no-load from the 138 kV side, (3) 24-hr no-load monitoring of the tx for partial discharge + oil temp + DGA baseline, (4) close the first 36 kV feeder breaker and energize one inverter skid no-load, (5) sun-up DC connect the first inverter and ramp to 10% then 50% then 100%, (6) repeat for remaining 17 inverters, (7) execute the 14-day performance test per the PPA (must hit the contractual PR threshold, typically 80%+ at year-1 — we target 83%), (8) final ERCOT NOC notification of commercial operation.

Project management: PMP-certified PM running Primavera P6 as the schedule of record; MS Project for sub-contractor coordination; weekly schedule reviews; cost + schedule integrated through earned-value management per AACE International RP 14R-90. See project-management-engineering.

20. Risk

A condensed register of the top project risks and mitigations:

1. Tariff and trade risk. Section 201 Safeguard tariff (14.25% in 2026, expiring 2026 unless extended), antidumping + countervailing duty (AD/CVD) on Chinese-origin and now Southeast-Asian-origin (Cambodia, Malaysia, Thailand, Vietnam) cells and modules. The Solar Energy Manufacturing for America Act / IRA § 45X advanced manufacturing PTC for U.S.-domestic modules is building domestic supply but capacity falls short of 2025-2026 demand. Mitigation: lock module pricing with a take-or-pay reservation at PO, dual-source across two countries of origin, include tariff-pass-through clauses.

2. Transformer lead time. As of 2025-2026, 100 MVA-class step-up transformers have 18-24 month lead times due to grain-oriented electrical steel (GOES) supply constraints and a small handful of qualified North American manufacturers. Mitigation: PO at NTP-6 mo (already on critical path).

3. Interconnect queue and cluster-study delay. ERCOT cluster studies have run 12-18 months in 2024-2025 vs. the planned 6-9 months. Mitigation: queue position locked at dev start, cluster-study contingency built into financial close.

4. Hail. West Texas is in the hail belt — Reeves County averages 2-3 hail days per year, with ~ 1 in 10-year probability of > 1 in (25 mm) hail. IEC 61215 module test is a 25 mm steel ball at terminal velocity; modules pass at production but field installations have failed at > 50 mm hail. Mitigation: select modules with thicker front glass (3.2 mm front, glass-glass back), program Nextracker’s HailPro auto-stow on forecast (60° tilt reduces effective hailstone normal impact), carry a property insurance policy with hail-specific deductible structure ($3-5M per event aggregate). Industry insurance loss data 2018-2023 shows hail dominates loss frequency.

5. Fire. PV array fires are rare but consequential — typical causes: MC4 connector mis-mating, ground-fault under combiner box, inverter capacitor failure. NEC 690.12 has required module-level rapid shutdown for buildings since 2017; utility-scale ground-mount arrays are exempt from MLPE rapid shutdown but still require array-boundary rapid shutdown — which our 1500 V DC central architecture meets via DC contactors at the combiners. Property fire insurance limits are 90% replacement value.

6. Land and lease risk. 30-year land lease at $20-25/acre/yr (Texas rangeland rates). Mineral rights are severed (split estate) and held by Permian Basin oil-and-gas operators — accommodation agreement is required. Mitigation: dominant-estate accommodation pact with the mineral lessee, restricting drilling to perimeter pads.

7. Curtailment. ERCOT West has growing IBR concentration and limited transmission. 2024-2025 average West-zone curtailment for solar = 4-7%. Mitigation: PPA structured as fixed-volume at $25/MWh with curtailment risk borne by offtaker, or financial-hedge curtailment with a node-vs-hub basis swap.

8. O&M long-term performance. Inverter availability typically 98-99% after the first-year burn-in. Tracker availability 99.5%. Module degradation 0.45%/yr. Mitigation: full-wrap O&M service contract with availability + PR guarantees + liquidated damages.

9. Cybersecurity. Public IBR fleets have been the target of probing intrusions (the 2024 OpenSolar API exposure, and the documented Sungrow remote-control vulnerability in 2023 — patched, but a forcing event for the industry). Mitigation: vendor patching SLA, isolation of the OT network, CIP-007 control set, periodic penetration testing by an independent assessor.

10. PPA counterparty risk. A 20-year fixed-price PPA only works if the offtaker is investment-grade. Our offtaker is an A-rated investor-owned utility plus a corporate sleeve from a hyperscaler data-center operator (BBB+). Diversifying offtake across two counterparties caps any single-counterparty default exposure at 50% of revenue.

11. Insurance. Property all-risk (PAR), business interruption (BI 12 mo limit), general liability, environmental impairment, builders’ risk during construction, and a delay-in-startup (DSU) policy through year 1. Premiums total ~ 0.7M/yr across the O&M phase. Hail-specific sub-limit at 5M per event with $250k deductible. Insurer mandates Nextracker HailPro auto-stow as a condition.

21. Cost build

A line-item turnkey EPC cost (M):

Line item$/W_DC$MNotes
Modules0.1313.0175 000 × 575 W × $0.13/W landed
Inverters + integrated step-up tx0.077.018 × Sungrow SG4400UD-MV-US at $0.07/W
Trackers0.1010.0Nextracker NX Horizon 1P, 100 MW × $0.10/W
Civil + mechanical + piles0.2020.0Roads + grading + 28 000 piles driven + tracker assembly labor
DC + AC + MV electrical BoS0.1010.0PV wire, combiners, 36 kV MV cable, terminations
Substation + 138 kV interconnect0.1515.0100 MVA tx, 138 kV breaker bay, control building, 3 mi line, network upgrades
Soft cost (eng + permits + finance)0.2525.0Owner’s engineer, EPC margin, permits, dev cost recapture, debt fees, tax-equity fees, contingency
Total1.00100.0Pre-credit cost. After ITC monetization, effective net basis ≈ $90M turnkey to sponsor.

Adjusted for the ITC and tax-equity flip mechanics, the effective sponsor capital is ~90M turnkey” headline number folds in the value of the ITC sale to the tax equity. This is how solar project costs are typically quoted at the developer-press-release level.

Comparison to industry benchmarks: NREL’s 2024 ATB Solar utility-scale 100 MW class is **0.78-1.20). LBNL’s “Tracking the Sun” 16th ed. 2024 reports 2023 actual completed projects at 1.00/W_DC pre-credit is on the median.

Where the 2024-2025 cost reductions came from, vs. 2020 vintage of the same plant ($1.30/W_DC):

  • Module cost fell from 0.13/W as Chinese poly + wafer + cell capacity tripled (now > 1 TW/yr global poly + cell + module capacity) — a 50% reduction, contributing the largest single delta.
  • Tracker cost held flat ($0.10/W) — steel commodity price up, productivity offsets.
  • Inverter cost fell from 0.07/W (SiC adoption + integrated step-up tx + scale).
  • Civil/EPC labor flat to up; offset by faster install crews (per-MW labor hours fell 20%).
  • Soft cost held flat at $0.25/W — finance + permitting + interconnect studies have grown faster than EPC has shrunk.

The forward trajectory: NREL ATB 2027 projects $0.78/W_DC mid-case as the cost convergence continues, with the inflection point of the falling-module / rising-balance-of-system cost curves landing in 2026-2028.

22. Cross-references summary + Citations

Cross-referenced Tier-3 + Tier-2 notes

Citations and references

Statutes and codes:

  • Inflation Reduction Act of 2022 (IRA), Public Law 117-169, esp. § 48 (Investment Tax Credit), § 45 (Production Tax Credit), § 45X (Advanced Manufacturing Production Credit), and § 48E / § 45Y (the technology-neutral successor credits applicable from 2025+).
  • International Building Code (IBC) 2024 edition, International Code Council.
  • ASCE/SEI 7-22, Minimum Design Loads and Associated Criteria for Buildings and Other Structures, American Society of Civil Engineers, 2022.
  • NFPA 70, National Electrical Code (NEC), 2026 edition, National Fire Protection Association, esp. Articles 690, 691, 705.
  • IEEE Std 1547-2018, IEEE Standard for Interconnection and Interoperability of Distributed Energy Resources with Associated Electric Power Systems Interfaces.
  • UL 61730-1 and UL 61730-2, Photovoltaic (PV) Module Safety Qualification.
  • UL 3741, Photovoltaic Hazard Control.
  • UL 1741-SB (Supplement SB), Inverters, Converters, Controllers and Interconnection System Equipment for Use With Distributed Energy Resources.
  • FERC Order 2003 and the pro-forma Large Generator Interconnection Agreement (LGIA), as adopted into ERCOT Nodal Operating Guides Section 5.
  • ERCOT Nodal Protocols Section 6 and Nodal Operating Guides Section 5, including Generic Resource Form filings.
  • NERC Reliability Standards CIP-002 through CIP-014.

Industry data sources:

  • SEIA / Wood Mackenzie, U.S. Solar Market Insight, 2024 Year-in-Review and 2025 Q1 Edition.
  • NREL Annual Technology Baseline (ATB) — Utility-Scale Solar, 2024 edition (Bolinger, M., Wiser, R., et al.).
  • LBNL, “Tracking the Sun: Pricing and Design Trends for Distributed Photovoltaic Systems in the United States,” 16th edition, 2024.
  • NREL National Solar Radiation Database (NSRDB), PSM3 dataset.
  • DOE/NREL System Advisor Model (SAM) 2024.10.

Manufacturer datasheets (referenced where decisions made):

  • Jinko Solar Tiger Neo N-type 78HL4-(V) 600 W bifacial datasheet, rev. 2025-Q1.
  • Sungrow SG4400UD-MV-US, central inverter with integrated MV transformer, datasheet 2025.
  • Nextracker NX Horizon 1P single-axis tracker, technical datasheet 2025.
  • Stäubli MC4-Evo 2 photovoltaic connector, UL/IEC certified, datasheet 2024.
  • GE Prolec 100 MVA 138/36 kV step-up transformer, typical specification.

Closing notes on design choices

Three decisions in this walkthrough are worth re-stating because they would each be different if a key input shifted:

  1. Central vs. string inverter. Picked central at $0.07/W on the 80 MW_AC scale. If the project were 30 MW or smaller, or if granular MPPT for non-uniform shading were a factor, string inverters would win. The crossover is in the 30-50 MW band.

  2. 1500 V DC vs. 2000 V DC. Picked 1500 V — the current code-recognized utility ceiling under NEC 2026 and UL 61730. 2000 V DC is in IEC product development (Trina announced 2 000 V modules in 2024, Sungrow has a 2 000 V central inverter on its roadmap) but has not yet been adopted into UL or NEC, so it is not deployable in the U.S. yet. A 2 000 V architecture would cut DC cable size ≈ 25% and shrink combiner counts, saving 1-2% capex once it’s certified — expect this to be the next plant generation around 2028-2030.

  3. Bifacial TOPCon vs. monofacial PERC. Picked bifacial TOPCon. Despite the higher per-W module price (0.10/W for residual mono-PERC inventory), the 7-8% bifacial yield gain and the better temperature coefficient give a ~ $0.04/W net advantage on LCOE — and 2026+ supply is essentially all n-type TOPCon or HJT anyway, with PERC capacity wound down.

End of walkthrough. The 100 MW_DC Reeves County design described above is intentionally complete enough to be a single-source reference for any downstream task: cost optimization, alternate-technology swap, sensitivity analysis, or financial restructuring. Every numerical reasoning chain is traceable to either a Tier-3 reference note or a named industry citation.

Graph View

Backlinks