A battery energy storage system, or BESS, moves electricity between times. Its commercial result depends on when it charges, when it discharges, what services it provides and the physical limits under which it operates. Forecasting can inform those choices, but renewable generation forecasts are not price forecasts and cannot on their own establish a revenue outcome.
Power and energy define different limits
Power capacity describes how quickly the battery can charge or discharge. Energy capacity describes how much energy it can store within its operating limits. State of charge determines the headroom available at a particular moment. A dispatch plan that meets the power limit can still be infeasible because it exhausts stored energy or leaves insufficient space to charge.
Efficiency, auxiliary consumption, availability and degradation also influence the usable result. Keep the measurement boundary clear: energy at the battery, converter or grid connection is not necessarily identical. When evaluating a strategy, use the appropriate charging and discharging limits rather than assuming a lossless device with unlimited cycling.
Different services create different obligations
Storage may shift energy between periods, support a co-located renewable asset or provide system services where eligible. Each application has its own operational commitments and commercial framework. The same stored energy cannot be promised independently to incompatible services. Revenue stacking therefore requires a feasible combined schedule, not a sum of attractive standalone estimates.
Resolve the operating objective before selecting forecast inputs. A hybrid solar project may need an expected export profile and a plan for charging opportunities. A standalone market strategy may require price and dispatch information in addition to asset state. A contracted service can require reserved headroom that limits other opportunities. These distinctions shape which uncertainty matters.
What renewable forecasts contribute
Wind and solar forecasts describe expected generation and its timing. At a co-located site they can help anticipate charging energy, export-limit exposure and the need to revise a schedule. They can also support scenarios for a wider portfolio. They do not reveal all future prices, instructions or competing behaviour in a market.
Ask whether the forecast target is available production or expected delivered output. If the renewable asset can be curtailed or unavailable, the charging plan needs that context. For a hybrid site, clarify how the grid connection limit applies to renewable generation and battery discharge together. Treating each device independently can overstate deliverable power.
Uncertainty has a physical consequence
Overestimating solar output can leave less charging energy than planned. Underestimating it can leave insufficient battery headroom during a high-output period. The consequences are asymmetric and depend on the operating policy. A generic reduction in forecast MAE does not prove that a storage strategy will earn more or incur less risk.
Use coherent production scenarios across time when testing dispatch. Independent low and high interval quantiles may not describe physically plausible daily paths. Preserve the relationship between generation, state of charge and subsequent choices. Include uncertainty in availability and input freshness where it is material to the decision.
Replay decisions without hindsight
A credible backtest uses the information available at each historical decision time. Archive forecast issue times and distinguish them from later revisions. The strategy should not use realized future generation or prices as if they had been known. A perfect-foresight result can be an analytical upper bound, but it is not a production strategy forecast.
Respect charging and discharging efficiency, power limits, energy limits and the chosen initial and final state-of-charge rules. Include degradation assumptions consistently and identify which costs are excluded. Compare the strategy with a defined baseline under the same conditions. Report both commercial outcomes and physical feasibility violations, even when the latter make the result less attractive.
Questions for a forecasting evaluation
Ask which plant and battery data are needed, how revisions reach the scheduling system and what happens if telemetry is missing. Test forecast performance at the horizons when charging and dispatch choices are made. Examine variable production days and constraint events separately from quiet periods.
Before accepting a revenue study, request its market assumptions, operating policy, forecast availability, cost boundary and sensitivity analysis. No single published return can substitute for those items. The purpose of forecasting is to support a feasible decision under uncertainty, not to turn an uncertain energy strategy into a guaranteed income stream.