Degradation-adjusted LCOH

Method, assumptions and limits

Degradation is modeled as a uniform rise in cell voltage over operating hours — the standard simplification behind every published µV/h or %/1000h figure. That voltage rise converts to specific energy through the same relationship used across all five tools: kWh/kg = 26.59 × V_cell ÷ Faradaic efficiency.

The mode that most LCOH tools get wrong

Most real projects — anything coupled to a capped grid connection, a PPA, or curtailed renewable supply — operate closer to constant power. Most published LCOH models implicitly assume constant current (or ignore degradation entirely), because it's the easier case to compute. That understates how much degradation actually costs.

The levelized cost formula

The production stream is discounted at the same rate as the cost stream — the LCOE convention. Skipping this (discounting costs but not production) understates LCOH whenever production is front-loaded relative to costs, which is exactly what happens once degradation and a mid-life stack replacement are modeled honestly: early years produce more per dollar spent than late years do.

Stack replacement

A replacement is scheduled whenever cumulative degradation pushes cell voltage past your trigger threshold (default +10%, a common warranty/EOL convention), at a cost equal to your stated share of total CAPEX (default 40%). Note this is a replacement-specific share, not the initial stack's share of a new system: a stack's share of a brand-new system's CAPEX varies widely by technology and scope (roughly 19–60% depending on what's bundled in), but replacing just the stack module — no fresh BOP, civil works, or power electronics — is consistently narrower across published PEM/AWE cost studies, in the 30–40% range. 40% sits at the top of that band on purpose (a higher assumed replacement cost is the conservative direction for LCOH). The degradation clock resets after each replacement; a long project life with a high degradation rate and low trigger threshold can schedule more than one.

The "naive" comparison

The naive figure shown alongside your result uses your BOL specific energy held constant for the entire project life, with no stack replacement scheduled — because that is what a genuinely naive LCOH calculation does, not just a simplification of one. The gap between the two numbers is not a modeling nuance; it's the cost of an assumption.

Default coefficient sources

Every default below is shown in the form and freely overridable — these are starting points, not your project's numbers:

• Fixed O&M, 2%/yr of CAPEX — published techno-economic studies for electrolyzer systems commonly report a 1–3%/yr range, with 2.5%/yr a frequently-used baseline figure; 2% sits inside that range.
• Water price, $3/m³ — this is industrial/demineralized process water delivered to electrolyzer feed spec, not raw municipal tap water. Raw water pricing alone varies enormously by region (roughly $0.05–3.50/m³ across US–EU, OECD average on the order of $1.20/m³); treatment to feed quality (deionization, degassing) adds real cost on top, which is why this default sits toward the higher end of the raw-water range rather than at the regional average.
• Water use, 12 L/kg — the stoichiometric minimum from the 2H₂O → 2H₂ + O₂ mass balance is ≈9 L/kg; real industrial PEM systems (including purification, cooling, and rejection losses) are commonly reported in the ≈9–25 L/kg range. 12 sits near the stoichiometric end.
• WACC, 8% — a commonly-used illustrative discount rate in published utility-scale renewable/electrolysis LCOH studies (typical cited range roughly 6–10%, depending on project risk and geography). This is not a substitute for your own project's actual cost of capital.

What this tool deliberately does not do

• It uses a flat electricity price, not an hourly profile — a hookup for hourly renewable/spot-price profiles is a natural next iteration but isn't in this version. A flat price is a reasonable approximation when capacity factor is set correctly, but it can't capture price-degradation correlation (e.g. curtailment-driven low prices coinciding with low-load, low-degradation operation).
• It doesn't model inflation separately — all inputs are treated as real (inflation-adjusted) terms, consistent with using a real WACC.
• It assumes uniform (not mechanism-resolved) degradation — the same simplification the underlying µV/h figure already makes.