Methodology
GH2SIM runs a pure TypeScript engine on a 288-point (12×24) representative-day core: RE → electrolyzer min-load → BESS → daily-cyclic H₂ buffer → archetype grid top-up, then annualizes with month weights.
Finance is an LCOH waterfall (CRF + stack replacements + opex − SIGHT). Carbon intensity for India GHCI (≤2 kgCO₂e/kg) and EU RFNBO (≤3.38) is computed from the same dispatch trace. RFNBO post-2030 hourly correlation is approximated with 24 hourly buckets — not a bankable 8760-hour certificate.
Policy presets are expiry-aware: ISTS waiver by commissioning year, SIGHT ₹50/40/30, state CSS/AS toggles. Calibration targets the CEEW 2024 India band of roughly USD 3.5–5/kg.
See ENGINE-CONTRACT.md in the repo for the frozen I/O types.
CAPEX
CAPEX is built bottom-up and anchored to real Indian and Korean purchase orders rather than to a literature $/kW. Costs stack in this order: direct equipment ex-works → freight and customs → installation and site EPC → owner’s cost and contingency → interest during construction. The same result feeds the headline CAPEX, the LCOH waterfall and the feasibility checks, so those three cannot disagree.
| Vendor | Capacity | Contract value | Normalised |
|---|---|---|---|
AHES (KR) → KOSTEC ALK stack + BoP + install materials + commissioning | 1,000 Nm³/hr ≈ 4.8 MW | KRW 3,124,000,000 | $452/kW island · $479/kW delivered |
Nuberg → BPCL ALK equipment supply (+5% supervision, +3% spares) | 500 kW | ₹2,18,69,365 | ₹43,738.73/kW ≈ $523/kW |
Nuberg → IOCL Gujarat Downstream EPCC — purification, compression, dispensing | not stated | ₹47.95 Cr lumpsum | lumpsum benchmark — no unit rate derived |
Greenh → Medha Turnkey PEM station incl. PESO, chiller, N₂, FAT/SAT | not stated | ₹53.52 Cr net (₹63.16 Cr incl. IGST) | scope checklist — no unit rate derived |
Spirare ALK bipolar skid + DM + T&R + 2 × 50 m³ @ 20 bar | 300 Nm³/hr ≈ 1.4 MW | price not stated on document | LP buffer basis: ≈160 kg ≈ 6 h rated production |
Three things fall out of that evidence and are worth stating plainly. First, the AHES order is the only one that separates stack from balance of plant, and it puts the stack at just 45.8% of the island — BoP costs more than the stack does. Stack replacement is therefore charged against the stack alone, not the whole system. Second, fitting the two priced, capacity-stated orders (523 $/kW at 0.5 MW, 452 $/kW at 4.8 MW) gives a total-cost scaling exponent of 0.936, i.e. very weak scale economy on the equipment itself. NREL’s manufactured-cost analysis implies a much steeper curve; applied to the AHES anchor it would predict $778/kW at 0.5 MW against the $523/kW BPCL actually paid, so the vendor fit is the default and NREL is available as an alternative preset. Third, that two-point fit spans two vendors in two countries, so it carries a country effect as well as a scale effect — treat it as calibrated, not proven.
GST at 18% is reported but kept out of the capitalised total, since a registered producer recovers it as input tax credit; basic customs duty and the Social Welfare Surcharge are not creditable and do stack. Renewable assets under a PPA are shown separately and not capitalised — the tariff already carries them, and counting both would double-charge the project.
Every CAPEX line carries a confidence tag, and the inspector colour-codes them: green traces to a purchase order or a primary source, amber to cited literature, red to an engineering estimate that is not yet sourced. These remain open:
- Enacted BCD rate on imported electrolysers (currently assumed 7.5%)
- India-specific EPC / installation uplift over equipment (currently assumed 15%)
- Water treatment and RO cost per m³/h
- Grid connection / substation cost per MW
- Standalone compressor cost per kg/h
- Low-pressure (20–50 bar) hydrogen storage cost per kg
- PEM $/kW from a capacity-stated quote — the only PEM PO on file has no capacity
- Owner's cost, contingency and IDC norms from an Indian lender or DPR
Run npm run capex for the full line-by-line build, the scale curve and the back-fit against both orders.
Operating cost and the owner’s view
The same plant is also reported annually rather than levelised, because nobody runs a business in dollars per kilogram. Operating cost splits into electricity — PPA energy, grid energy, wheeling and surcharges, and compression — then consumables, then O&M, then the stack-replacement reserve, less the SIGHT incentive. Captive renewable energy carries a quarter of the residual open-access charge stack; grid import carries all of it. On a typical Indian configuration electricity lands at 70–80% of cash operating cost, which is the number that decides the project.
On top of that sits a screening-level owner’s view: revenue at the offtake price, EBITDA on cash cost net of incentive, the debt and equity split, level debt service, DSCR, undiscounted payback, and margin against breakeven. It is deliberately thin — no tax, depreciation, working capital, ramp-up, output degradation or price escalation. It is built to reject a configuration quickly, not to bank one.
One consequence is worth stating because it looks wrong at first glance: a plant can show a DSCR comfortably above 1.3× while still selling below levelised cost. Debt service is paid at the cost of debt; LCOH charges the full weighted cost of capital. The bank is fine and equity is not.