Battery dispatch: margins and forecast value

A 1 MW battery is scheduled each day from a forecast made at the 12:00 D-1 gate and settled against realised DE-LU day-ahead prices. The same battery is also run on three simple rules that need no model. The difference between them is what the forecast is worth.

Result

For the -hour battery, Ridge earns EUR /MW a year more than the best simple rule, : EUR /MW on an average day, or more margin. The paired 95% interval is EUR –/MW a year. The sample covers days, to ; the annual figures are averages over it, not projections.

Most of the margin does not need a forecast. Repeating the daily shape earns the bulk of it, and the forecast adds a thin layer on days when the shape changes. The best simple rule is ranked over the whole sample, in hindsight; all three are shown.

These are simulated day-ahead arbitrage margins under the stated assumptions, not a revenue estimate for a real battery. They exclude operating, degradation and capital costs, which lower returns, and intraday and balancing revenue, which a real battery would add.

Increments in this table are totals over the whole sample, not annual figures.

The forecast's share is shrinking

The simple rule's share of the perfect-foresight margin rose from in to in . Ridge's increment in euros has been broadly stable since 2023, but relative to the simple rule it fell from to . The more regular the solar-driven shape, the more a rule that repeats it captures on its own.

Where the forecast earns

On the of days whose shape was most typical, Ridge loses to the simple rule, by EUR /MW a day. The most atypical days carry of its total gain. Negative-price days are not where it earns: EUR /MW a day there, against on other days, in line with its weak negative-price calls on the forecast page.

Surprise is judged after the fact; disagreement is known before the auction. The fifth of days on which Ridge's schedule expected most over the simple rule's carries of its gain; on the fifth where they agree most, the two earn within EUR /MW a day of each other. The forecast already behaves as the switch a desk might build: it follows the recurring shape unless it sees a reason not to. A rule that kept the simple schedule and traded on Ridge only above a disagreement threshold, refitted each day on days already settled, ended within 0.1% of Ridge's margin on every battery size (2021–2026), so it adds nothing.

How a forecast becomes a decision

On , the latest day in the sample, the whole day's schedule is set from Ridge's forecast at the gate and then settled at the prices that cleared. Ridge earned EUR , the rule EUR , and perfect foresight EUR .

What 15-minute products are worth

Since October 2025 the auction clears quarter-hours. On the same days ( to ), trading quarter-hours instead of hours raises the perfect-foresight margin by for a 1-hour battery and for a 4-hour one. Quarter-hour extremes widen the daily price range by about 16%, but a battery that moves energy over several hours captures only part of that: the shorter the battery, the more the finer products are worth. A rule that repeats the previous day's quarter-hours keeps most of the gain. The forecast on this site is still hourly, so none of this value is in its results.

Day-ahead is one market among several

Over the days from November 2020 with both markets published, aFRR capacity paid EUR /MW a day and FCR , against for day-ahead arbitrage with the Ridge forecast. A rule that picks each day's market on what is known before the balancing auctions close chose aFRR on of days and earned , close to the best market in hindsight.

Two things temper that. The balancing markets are shallow: Germany procures about 0.6 GW of FCR and about 2 GW of aFRR each way, against 21 GW of installed batteries, so they cannot absorb the fleet. And these are capacity payments, before the prequalification, energy management and bidding risk they require; aFRR is pay-as-bid, so its average accepted price is not guaranteed to any bid. Meanwhile the gap is closing: for a -hour battery, day-ahead arbitrage earned of the aFRR capacity value in 2021 and in . Intraday trading, activation energy and mFRR are not modelled.

Costs and robustness

The stresses below rerun the optimiser for Ridge without retuning it. Costs apply to grid-side charging plus discharging MWh; the rates are illustrative, not calibrated German project costs.

A second daily cycle adds margin, not forecast value. Allowing two charge–discharge episodes a day raises gross margin by , at equivalent cycles a day, while Ridge's advantage over the best simple rule moves from EUR to EUR /MW a year: the simple rule earns almost all of the extra margin too. The second cycle trades the midday solar trough, the most predictable feature of the German day.

Downside and concentration

Ridge underperforms the best simple rule on of days. Its five best days account for of all positive incremental margin, and removing them still leaves EUR /MW, so the advantage is not driven by a handful of days.

Duration

Longer duration earns more per MW but needs more energy capacity, so compare the increment per MWh as well. Ranking durations as investments would need capital costs, availability terms, lifetime degradation and other revenues, which are outside this study.

Is this margin durable as the market changes?

Solar is still deepening the trough. Solar's capture rate fell from in to in (partial year) as installed solar grew about ×, and on-peak hours now clear below off-peak on average: the signature of solar cannibalisation. Wind shows no comparable trend. With complete years, these correlations ( for solar capture, for wind) describe co-movement, not causation. The 2030 targets, GW of solar against a recorded GW AC ( GW DC) in , lie well beyond the range these data cover.

Competing storage has not yet compressed the day-ahead margin, but that may change. For the -hour battery, the perfect-foresight margin has risen with Germany's battery fleet (r=, n= years). Over so few years, with the gas shock in the sample, this cannot rule out a competition effect. The shallow balancing markets are where a larger fleet would press first, and the day-ahead spread is the only market deep enough to absorb it, which makes that spread the one to watch. Storage packs are getting cheaper: global stationary-storage pack prices fell 45% in 2025 to $70/kWh (BloombergNEF, December 2025), which lowers the barrier to new capacity. A calmer gas market or a large build of grid-scale storage would shrink the spread itself, not just the forecast's edge.