Day-ahead vs intraday battery optimization: why dispatch should keep evolving

Day-ahead and intraday solve different decisions

In Single Day-Ahead Coupling (SDAC), bids across participating European bidding zones are coupled with cross-zonal capacity and cleared for the next delivery day. The result creates prices, matched trades and scheduled exchanges. For a battery, the day-ahead decision establishes a baseline position before the operator knows next day’s exact PV output, demand, outages, reserve activation or intraday prices.

Single Intraday Coupling (SIDC) and local intraday products let participants trade after day-ahead closure and closer to delivery. Intraday is not merely a second arbitrage market. It is the principal recourse layer for responding to forecast revisions, plant availability and changing opportunity while respecting what has already been bought or sold.

A battery schedule is a rolling timeline

At every update, the optimizer should carry forward the actual commercial state: cleared day-ahead quantities, completed intraday trades, reserve obligations and any locked nominations. Past decisions are fixed. Future decisions can be revised only within available market windows.

Europe’s move to 15-minute SDAC market time units for delivery from 1 October 2025 increased the resolution of this process. More intervals can represent solar ramps and battery behavior more accurately, but they also multiply forecast, bid, schedule and telemetry records.

SOC is inventory—and optionality

A full battery cannot absorb a negative-price or downward-activation opportunity. An empty battery cannot respond to a high price or upward activation. Mid-range SOC can therefore have option value even when an immediate charge or discharge appears profitable.

SOC reservation is not a fixed percentage. It depends on price uncertainty, later opportunities, required reserve headroom, charging access, terminal obligations and the cost of failing to deliver. A risk-averse operator may keep more headroom; a merchant operator may accept more exposure when the current spread is strong.

Forecast revisions drive intraday decisions

Relevant updates include price distributions, PV and wind forecasts, load, cross-border capacity, outages and site availability. The optimizer should know when each forecast was created. Evaluating a D-1 decision with a forecast updated on the delivery morning introduces look-ahead bias.

Intraday liquidity also matters. A modeled price is not necessarily a fillable trade. Bid-ask spread, depth, partial fills, order type, fees and gate closure can determine whether a theoretical improvement is executable. A useful system separates forecast value from execution assumptions.

Trading and physical dispatch must reconcile

An intraday trade changes the position; it does not move the battery. The updated net schedule must pass through asset and site constraints, operational validation and the EMS/PCS control chain. If the BMS derates power or communications fail, the expected position and physical response diverge. That difference can create imbalance exposure.

A closed loop records order, trade, schedule version, approved command, acknowledgement, revenue-meter power and settlement interval. Battery trading software should therefore connect to optimization and operations rather than remain an isolated order screen.

Illustrative rolling decision

Illustrative example—not market or asset performance. At D-1, a price forecast shows a €60/MWh spread between midday and evening. After assumed round-trip losses, fees and degradation, the optimizer schedules a partial charge and discharge but leaves 25% of usable energy uncommitted. On delivery morning, a revised PV forecast lowers expected midday prices and an evening plant derate reduces available discharge power. The intraday run increases charging where liquidity allows, reduces the infeasible evening sale and moves some discharge to a later interval.

The improved decision is not simply “trade more intraday.” It uses intraday to reconcile new information with the remaining physical envelope. A benchmark should compare this rolling policy with day-ahead-only operation using the same original forecast and with a separately labeled perfect-foresight upper bound.

Bids, trades, schedules and setpoints are not interchangeable

A bid is an offer to transact. A trade is an accepted commercial position. A schedule or nomination describes intended energy delivery under the relevant process. A setpoint requests instantaneous or interval power from the plant. Conflating them creates subtle errors: an optimizer may dispatch an unfilled order, a trader may change a position without updating the EMS, or a controller may continue an expired schedule.

Use stable identifiers across the chain. A decision record should identify forecast vintage, optimization run and proposed order. A fill record should update the position. A dispatch schedule should cite the position version it serves. Each command should have a validity window and acknowledgement. Metered energy should reconcile to the interval and commercial boundary.

Transaction costs change switching behavior

Without bid-ask spread, fees and degradation, a model can alternate charge and discharge on every small price movement. Add a minimum economic threshold, realistic fill model and any operational dwell/ramp behavior. Robust policies often make fewer decisions than an oracle because preserving flexibility and avoiding weak trades has value.

What to do when the forecast worsens

A less certain intraday outlook does not always imply closing the position. The battery may have physical flexibility to absorb the error, or closing liquidity may be poor. Compare three actions—trade out, cover physically, or accept residual imbalance—using the same probabilistic outcomes and constraints. The responsible choice depends on who carries imbalance and how delivered energy is settled.