Use case · Grid-scale battery · Germany · 2024
Two hours or four?
A 20 MW German battery, backtested across all 366 days of 2024 on realised EPEX day-ahead and intraday prices, with cell wear priced into every trade. Doubling the energy bought 30 % more income — not 100 %.
| Same 20 MW connection | 2h | 4h |
|---|---|---|
| Total income, net of wear | €3,086,271 | €4,012,697 |
| Per MW, net of wear | 154.3 k€/MW/yr | 200.6 k€/MW/yr |
| Per MW, before wear | 183.0 k€/MW/yr | 235.0 k€/MW/yr |
| Per kWh installed | €77.2/kWh/yr | €50.2/kWh/yr |
| Equivalent full cycles | 955.63 | 573.25 |
Duration is concave
Doubling the energy returned 30% more income. Throughput rose only 20%: the longer asset does not cycle more, it cycles better, picking wider spreads. The extra 40 MWh earned €23,161 per MWh per year against €77.2/kWh for the first 40 — the marginal kWh was worth about 30% of the average one, and that is the figure a capex case has to clear.
The intraday desk added about 40%
On top of the day-ahead schedule, under forward-only conditions rather than with hindsight: 189 re-optimisations per trading day, each traded only if it improved income by at least €1.00 and actually changed the schedule. Two-thirds of the 69,174 candidates were computed and rejected. This is not a perfect-foresight number.
Cycling decides whether the 2h case is bankable
Nothing capped the 2h asset but the wear price, and it ran 955.63 equivalent full cycles — past the 730 per year most warranties and published benchmarks assume. Respecting that cap costs roughly 9%. The 4h asset needs no such caveat, at 573.25.
Day-ahead income holds a broad June–September plateau. The intraday uplift does not — April, May, September and December carry 77% of the year’s entire intraday result.
Cross-checked against a published benchmark
On the comparable basis — before wear, since the benchmark charges no degradation — our 2h asset lands +17 % to +32 % above the realised German battery index published by KYOS (138.7 k€/MW/yr equivalent). Different years, so read it as an order-of-magnitude check, not a validation.
The range is the answer, not a hedge. Our run was not cycle-limited and took 955.63 equivalent full cycles, where the benchmark battery is capped at 730. We did not re-run the year under that cap, so we bound it instead of guessing. Giving up whole days until the budget is met yields 162.2 k€/MW/yr (+17 %) — a floor, because a genuinely cycle-constrained optimizer would surrender the cheapest cycles within each day rather than discarding entire days. The uncapped result, 183.0 k€/MW/yr (+32 %), is the ceiling, since adding a constraint cannot raise income. Days are optimized independently here, so both bounds hold exactly and the true capped figure sits between them.
Figures from KYOS Energy Storage report Nr 10, November 2025. KYOS has not reviewed, endorsed or validated this analysis; the comparison and every normalisation in it are ours. Their index window is November 2024 – October 2025, a period they describe as weaker than the equivalent 2024 months. Their reference battery assumes 90 % round-trip efficiency against our 95 %, which favours us; their index trades intraday with perfect price foresight, which favours them. Against their forward-looking 2026 assessment (151.6 k€/MW) the same range is +7 % to +21 %. Their published intraday figure is, on our reading of their methodology, a two-market result — we have asked them to confirm; on the alternative reading the matched pairing is day-ahead only, where we reach 110.4 k€/MW against their 106.4, +3.8 %.
Assumptions and scope
- Wholesale gross margin only — before grid fees, taxes, levies, trading costs, capex and opex. Not project returns.
- Degradation priced at €15/MWh discharged, inside the optimizer’s objective. “Before wear” figures add it back for comparability with benchmarks that assume none.
- Energy arbitrage only — day-ahead plus intraday continuous. No FCR, aFRR or imbalance revenue.
- One market, one year, one price path. Realised German 2024 prices — a comparatively strong year. A backtest, not a forecast. Excluding the four largest days, the 2h year falls to 124.2 k€/MW/yr.
- Deterministic mixed-integer optimization (SCIP) on a versioned, artefact-tracked pipeline. No LLM produces any figure here.
Run your asset through the same pipeline
Your power and energy, your grid limits, your wear price, the reference years you care about — and the generated report with the artefact-level trail behind every number. Send us the spec sheet and the question you need answered.