Evidence
How H2Hub is tested
H2Hub is a screening tool. Here is how we test it, what it gets close to, and where it differs from published studies.
Results as of 25 October 2026 · engine version 2026-10-29-incentive-assessment-v3 · 437 automated checks
1 · The cost formula
An independent calculator written outside H2Hub (year-by-year discounted cash flow) reproduces H2Hub's plant-gate LCOH to within 0.0000001% on 200 random plants. Separate checks confirm that electricity cost equals price × energy per kg, that fixed costs per kg double when utilisation halves, that plant size alone does not change $/kg, and that stack replacements only fall inside the plant life.
These now run as automatic tests inside H2Hub on every change (7 golden cases + 6 invariants), so a future edit cannot silently move a result.
2 · Published cost studies (US DOE H2A)
H2Hub's plant-gate cost was compared with seven published US DOE cases. Four land within 5%; three are lower by 7–13%.
| Case | Published $/kg | H2Hub method $/kg | Difference | Source |
|---|---|---|---|---|
| DOE Record 19009, current PEM, 3¢/kWh | 2.31 | 2.24 | −2.9% | DOE Hydrogen Program Record 19009 (2020) |
| DOE Record 19009, future PEM, 3¢/kWh | 1.86 | 1.83 | −1.4% | Record 19009 |
| DOE Record 19009, current PEM, 7.35¢/kWh | 4.83 | 4.66 | −3.6% | Record 19009 |
| DOE AMR 2022, current alkaline, 3¢/kWh (preliminary) | 2.41 | 2.33 | −3.5% | DOE AMR 2022, P204 |
| DOE Record 20006, SOEC, 3¢/kWh | 2.36 | 2.19 | −7.3% | Record 20006 (2020) |
| DOE Record 14004, current PEM | 5.12 | 4.43 | −13.4% | Record 14004 (2014) |
| DOE Record 14004, future PEM | 4.20 | 3.70 | −12.0% | Record 14004 (fixed O&M not published per kg; entered as 0) |
Why the misses
Every miss is lower, and every miss is in the capital charge. DOE's H2A model uses after-tax financing (8% after-tax real return, 21% tax, depreciation) and adds indirect capital costs. H2Hub uses one pre-tax real discount rate (default 8%) with no tax or construction period. To match DOE's capital charge, H2Hub would need roughly 9–13%. That is why every H2Hub result now also shows the cost at a 10% discount rate.
Electricity and fixed O&M terms match DOE in every case.
3 · Real plants
H2Hub was tested against the reported first-year output of Sinopec's Kuqa plant in Xinjiang (300 MW solar, 260 MW electrolysis).
| Measure | H2Hub (ERA5 2024) | Reported |
|---|---|---|
| Hydrogen, one year | 8,229 t | “more than 7,000 t” in the first 12 months (Hydrogen Insight, 2 Sep 2024) |
| First ~6 months | ≈4,100 t | ≈2,010 t to 21 Dec 2023 (Sinopec, via Jiemian / SCMP) |
| Electrolyser load | 18.7% | ≈20% in 2023 (Wood Mackenzie); “less than one-third” (BloombergNEF) |
| Design | — | 20,000 t/yr |
H2Hub's solar-only estimate sits above the reported first-year output, as expected for a plant still ramping up. The first six months produced about half the modelled amount, consistent with the technical and grid problems Sinopec reported, which no weather model captures. 300 MW of solar supports about 8,000 t a year in H2Hub, far below the 20,000 t design, so the design must rely on power that has not been published.
Inputs: 260 MW alkaline (52 units), 300 MW solar, no wind, battery or grid (none published), site at a town-level estimate (41.72, 82.94). Tested 25 Oct 2026.
Five plants in total, all run live in H2Hub on 25 Oct 2026 with ERA5 2024 weather, default costs, output only:
| Plant | Power supply | H2Hub | Reported | Outcome |
|---|---|---|---|---|
| Sinopec Kuqa, China | 300 MW solar | 8,229 t/yr | more than 7,000 t in first 12 months (Hydrogen Insight, Sep 2024) | Close — model above, as expected while ramping up |
| Energiepark Mainz, Germany | 8 MW wind (plus a grid connection) | 164 t/yr from wind alone | about 180 t/yr (ZfK, Feb 2020) | Close — 9% below; the plant also used grid power |
| Datang Duolun, China | 150 MW wind + solar (split not published) | 4,154–5,018 t/yr (two splits) | about 2,400–2,700 t/yr (22.6 million Nm³ in ~10 months, Science and Technology Daily, Nov 2025; 3,993 t to June 2026, Xinhua) | Miss — model 1.5–2.1× higher |
| SPIC Da'an, China | 700 MW wind + 100 MW solar + 80 MWh battery | about 7,960 t in the first 141 days | about 3,550 t hydrogen (from more than 20,000 t of ammonia in ~5 months, Sina Finance, Dec 2025) | Miss — model about 2.2× higher |
| HyBalance, Denmark | grid only, 1.2 MW | 20 t/yr (grid as backup) · 180 t/yr (grid as primary supply) | about 54 t/yr (120 t from 2018 to Sep 2020, Air Liquide) | Miss — the two grid rules sit either side of the reported ~54 t; the real plant ran to its customers' demand, not to capacity |
Why the misses
Duolun sends surplus power to the grid and was in its first year; H2Hub sends all usable power to the electrolyser. Da'an's figure is ammonia output during start-up, limited by the ammonia plant rather than by power. HyBalance ran to its customers' hydrogen demand: H2Hub's backup-only grid rule gives 20 t and the grid-as-primary rule gives 180 t, either side of the reported ~54 t. For demand-led plants, enter a hydrogen demand profile. Exact wind/solar splits, storage sizes and site coordinates are not published for several plants, so inputs are best public estimates.
Inputs used
- Sinopec Kuqa: 41.72, 82.94 · 260 MW alkaline (52 units) · 300 MW solar · no wind, battery or grid.
- Datang Duolun: 42.2, 116.5 · 70 MW alkaline · 100/50 and 50/100 MW wind/solar (two splits) · no battery or grid.
- SPIC Da'an: 45.5, 124.3 · 240 MW alkaline · 700 MW wind · 100 MW solar · 80 MWh battery · no grid.
- Energiepark Mainz: 49.96, 8.27 · 6 MW PEM · 8 MW wind · no grid.
- HyBalance: 56.64, 9.79 · 1.2 MW PEM · 1.2 MW grid import · run with both grid rules.
4 · Weather data
Every 8,760-hour run uses hourly ERA5 data (ECMWF / Copernicus) for the project's coordinates. In our NEOM case study, the modelled solar capacity factor is 23.2% compared with 22.4% from the Global Solar Atlas. Wind is a harder case: at NEOM's wind site ERA5 gives 7.67 m/s against 10.68 m/s from the Global Wind Atlas 4 at 120 m. H2Hub shows this cross-check next to the ERA5 mean and lets you apply the Atlas value as a correction; it never applies it silently. Runs that fall back to modelled climatology are flagged in amber.
Data obtained from the Global Wind Atlas version 4.0, a free, web-based application developed, owned and operated by the Technical University of Denmark (DTU). The Global Wind Atlas version 4.0 is released in partnership with the World Bank Group, utilizing data provided by Vortex, using funding provided by the Energy Sector Management Assistance Program (ESMAP). For additional information: https://globalwindatlas.info
ReneEnergy created grid-aligned regional subsets with native compressed tiles copied unchanged: no resampling or alteration of wind values. ReneEnergy sampled one native approximately 250 m raster cell and sheared its modelled 2008–2017 mean from 100 m to 120 m (α 0.14). This is not a site measurement or the 10% windiest-area statistic. The original publishers have not reviewed or endorsed this analysis.
5 · What we have not tested
- Delivered cost and the survivability rating use unsourced delivery and stress assumptions. They are tagged Indicative wherever they appear; see the indicative assumptions.
- Grid-only plants: H2Hub now offers both grid rules (backup only, or primary supply). Our one test (HyBalance) is bracketed by the two; demand-led plants need a hydrogen demand profile to match real output.
- Cost results have not been compared with any real plant's audited costs, because none publishes them.
- Ammonia, storage and shipping steps are outside the plant-gate cost.
Run it yourself
Put your own site, power price and electrolyser through the same tests.
Disclaimer
Disclaimer: The outputs, estimates, projections, and educational materials available through H2Hub are provided for informational and educational purposes only. They do not constitute financial, investment, tax, legal, or professional advice. References to engineering analysis or engineering insights describe analytical content and do not constitute licensed professional engineering services. Schematics and models are conceptual and feasibility-stage, and are not for construction, permitting or investment reliance without independent professional review. Past performance, model outputs, and market data are not guarantees of future results. Please consult a qualified professional before making any business, investment, or project decisions.