German power: price shape and battery value

In German day-ahead arbitrage, the recurring solar-driven price shape carries most of the value, and a price forecast earns on the days that shape breaks. This note measures the shape in Germany–Luxembourg (DE-LU) prices, values it for a battery against simple rules and the balancing markets, and tests what a forecast made at the auction gate adds.

4-hour battery: a rule that repeats the last similar day captured → of the perfect-foresight margin ( → YTD) · of the forecast's gain comes from the most atypical fifth of days · day-ahead arbitrage earned → of what aFRR capacity paid (2021 → )

Price shape, → YTD: on-peak premium EUR → /MWh · daily price range → of the average price · solar capture rate → · negative-price hours % → %

Forecast evidence · Storage value · Methodology

1. The shape carries most of the value

A 1 MW / 4 MWh battery dispatched once a day can, with perfect foresight, earn EUR /MW a day in . A rule that repeats the last similar day's prices captured of that in and in . The forecast sits a few points above it every year. The more regular the solar-driven shape becomes, the less room is left for a forecast.

2. The forecast earns when the shape breaks

A per-hour Ridge regression, using only what is known at the 12:00 auction gate on the day before delivery, cuts the mean absolute price error by % against the best naive forecast and picks the day's cheapest and dearest hours more often. For the 4-hour battery that is worth about EUR /MW a year over the simple rule, of the battery's gross margin.

On the of days whose shape was most typical, Ridge loses to the simple rule; the most atypical fifth of days carries of its gain. Negative-price days are not where it earns. Where it will earn is visible at the auction gate: the fifth of days on which Ridge's schedule expected most over the simple rule's carries of its gain. A switch that follows the shape and trades on the forecast only on those days adds nothing, because Ridge already follows the shape when the two agree.

Forecast evidence · Storage value

3. Why the shape exists

Installed solar rose from to GW over the sample, and the price solar plants earn fell from to of the average price; wind's barely moved. The on-peak block (08:00–20:00 on weekdays) went from a EUR /MWh premium to a discount, while the gap between the day's highest and lowest hourly price rose from to of the average price. , the most expensive year, sits at : the gas crisis raised the price level, and what changed afterwards is the shape.

The day turned from a plateau into a trough between two peaks, and the trough arrives at the same hours on every sunny day. That regularity is why a simple rule captures so much: still paid the most in euros (EUR /MW a day with perfect foresight, against in ), but its spread came from gas and left with it, while today's comes from the solar profile, which is still growing.

4. Beyond the day-ahead hour

Quarter-hour products, traded since October 2025, raise the day-ahead arbitrage ceiling by for a 1-hour battery and for a 4-hour one, against the 16% by which they widen the daily price range. Balancing capacity has paid more than arbitrage. Since November 2020, aFRR up and down averaged EUR /MW a day and FCR , against for day-ahead arbitrage with the forecast. Those markets are shallow: Germany procures about 0.6 GW of FCR and 2 GW of aFRR each way, against GW of installed batteries. The gap is closing: for a 4-hour battery, arbitrage earned of the aFRR capacity value in 2021 and in .

Storage value

Risks to this view

• Gas prices. A cheaper marginal unit would lower the evening peak without lifting the midday trough.
• Grid-scale storage. Two- to four-hour batteries compete for this spread. The fleet's average duration, hours in , is the first sign of them arriving; the small balancing markets would fill first.
• Market design. Fifteen-minute products, and any change to support payments in negative-price hours, change both the trade and its measurement.
• Evidence. The results are retrospective, on history inspected during development. A prospective pilot has issued forecasts before the gate for of delivery days attempted (as of ). It will show whether the process runs reliably, not whether the value holds across seasons.

Scope

• Markets: the DE-LU day-ahead auction, with hourly prices except in the quarter-hour study, and German FCR and aFRR capacity auctions. Continuous intraday trading is not modelled, for want of a free public price source; it is the natural next step.
• Battery: 1 MW with 1, 2 or 4 MWh, 90% round-trip efficiency, one charge–discharge cycle a day, empty at the start and end of each day. Margins are simulated under these assumptions: before operating, degradation and capital costs, and not a revenue estimate for a real asset.
• Partial year: runs to mid-September, so its figures are not full-year values. In complete years since 2019, the January–September daily range differed from the full-year figure by −4% to +11% (−36% in 2021, when the gas spike came in the fourth quarter).
• Sources: Energy-Charts (Fraunhofer ISE, CC BY 4.0), which republishes ENTSO-E and SMARD data, and regelleistung.net, the German TSOs' balancing platform. Definitions and assumptions.

Author

Pedro Cabral, power market analyst covering Brazil, Chile and Argentina. This independent project applies the same work on fundamentals and price formation to the German market.

Research repository · GitHub profile

Market context

Data through .