Refinery hydrogen: separate gas, electricity and carbon costs

A transparent variable-cost model for steam reforming, with delivered-gas sensitivity and explicit emissions boundaries.

AI-generated editorial artwork: Electrical substation in front of LNG tanks and refinery pipework

AI-generated editorial illustration; it does not depict a verified project, actual prices or chart data.

1. The economic decision

For a refinery buying hydrogen or operating a reformer, the useful comparison is cost per kilogram of usable hydrogen at an agreed pressure and purity. A gas benchmark alone cannot answer that question. Feedstock gas becomes part of the product; additional fuel supplies reaction heat; electricity drives compression and auxiliaries. A purchase contract may bundle all three into an index formula. Recovering the separate exposures makes its escalator understandable and reveals which savings remain possible inside the fence.

2. What August and September actually show

The World Bank October 2 Pink Sheet reports Henry Hub at 2.77 USD/MMBtu in August and 2.95 in September, while the Netherlands TTF series moves from 21.11 to 25.42. These are different locations and contractual bases, not alternative prices for the same refinery. September TTF is 4.31 USD/MMBtu above August; applying that change mechanically to the illustrative 0.125 MMBtu/kg intensity gives 0.53875 USD/kg, only if the refinery contract genuinely tracks that index without a lag or cap.

The EIA October 6 outlook, finalized October 1, forecasts annual 2026 Henry Hub at 3.48 USD/MMBtu and US wholesale power at 52 USD/MWh. Neither is an observed October plant tariff. No verified public August/September refinery-specific hydrogen offer is used here.

3. Why the process boundary matters

In DOE's process description, reforming needs supplied heat, the shift stage produces additional hydrogen and purification separates impurities. A cost model therefore needs measured gas to the entire hydrogen unit, usable hydrogen leaving it, and electricity including the compression required by the refinery. Do not count purge gas both as an external purchased fuel and as a credit. Exported steam warrants a credit only when another unit can use it and the avoided steam cost is documented. A production emissions factor is not automatically a lifecycle factor.

4. Illustrative operating model

Assume 0.125 MMBtu HHV purchased gas, 2 kWh electricity, 0.2 USD other variable expense and 10 kg CO2 charged per kg H2. Delivered gas is hypothetically 8 USD/MMBtu, electricity 100 USD/MWh and the applicable emissions charge hypothetically 80 USD/t CO2. The gas intensity includes feed and fuel. All amounts are nominal USD at a common assumed price date, excluding tax, capital recovery, fixed labor and hydrogen distribution beyond the outlet. Carbon coverage and price are assumptions, not a universal law or market observation.

Cost itemCalculation per kg H2USD/kg
Gas0.125 × 81.0
Electricity2 / 1000 × 1000.2
Other variable expenseAssumed0.2
Covered CO210 / 1000 × 800.8
Total1.0 + 0.2 + 0.2 + 0.82.2

At 10000 kg/day this is 22000 USD/day of modeled variable expense. It is not the full cost of owning a reformer, nor the delivered price of merchant hydrogen. If emissions are not financially covered, the modeled cash cost becomes 1.4 USD/kg; that does not mean physical emissions disappear.

5. Sensitivity and the useful trigger

Gas at 4, 8 and 12 USD/MMBtu gives 1.7, 2.2 and 2.7 USD/kg with the other inputs fixed. Every additional 1 USD/MMBtu adds 0.125 USD/kg. A 20 USD/MWh electricity increase adds 0.04 USD/kg; a 20 USD/t increase in the assumed covered-carbon price adds 0.2 USD/kg. Those sensitivities explain why gas hedging, compressor improvement and emissions management affect different parts of the bill.

Hydrogen variable-cost gas sensitivity

Illustrative variable cost per kg H2 at delivered gas 4, 8 and 12 USD/MMBtu; electricity 100 USD/MWh, other variable cost 0.2 USD/kg and assumed carbon cost 0.8 USD/kg remain fixed. Not a supplier quote.

Delivered gas USD/MMBtuVariable cost USD/kg
41.7
82.2
122.7

6. What to require from the contract

Ask for the named gas index, averaging window, billing lag, HHV/LHV convention, feed/fuel coefficients and electricity meter boundary. Separately identify capacity charges, minimum take, outage hydrogen, pressure/purity guarantees and who owns steam credits. Require the treatment of allowances, free allocation and captured CO2 before accepting a carbon escalator. Compare a supplier offer against the same output specification and availability; a cheap molecule without sufficient pressure can transfer compressor cost back to the refinery.

7. Limits and frequently asked questions

Does this prove electrolysis is cheaper? No. Compare usable hydrogen, operating hours, water treatment, stack replacement and capital recovery on the same boundary; renewable certificates do not by themselves eliminate delivered power costs.

Can the 10 kg CO2 factor be reused at every reformer? No. Obtain unit mass balances and the legally covered emissions scope; capture, fuel mix and steam allocation change it.

Should a refinery cut hydrogen use when gas rises? Only within product-quality and process-safety requirements. A hydrotreating constraint can make the lost product margin larger than the saved hydrogen expense.

8. Sources and next decisions

Use the World Bank monthly benchmarks, dated EIA outlook and DOE mechanism for their distinct purposes. For purchasing boundaries continue to the gas basis guide; for compressor power use the industrial bill guide. Build the plant comparison from actual meters and quotations before using the ROI calculator.