Definition
Whether forecast accuracy at the horizon a decision actually consumes meets the accuracy that decision requires. Catches the case where headline forecast accuracy is strong at one horizon while the decisions relying on a different horizon are running on numbers not good enough to support them.
The Measurement Gap
Forecast accuracy in renewable energy operations is not a single number. It is a curve that degrades as the time horizon extends. A day-ahead forecast might deliver 92% accuracy for aggregate production. The same model at the 4-hour horizon might deliver 85%. At the 1-hour horizon, 95%. The problem is not that these accuracies exist at different levels. The problem is that no one maps which accuracy level each downstream decision is actually consuming.
Consider the dispatch optimizer that pre-positions battery storage based on the 4-hour-ahead forecast. If that forecast is accurate to within 5%, the storage positioning works. If the 4-hour forecast is actually only accurate to within 15% during certain weather regimes, the optimizer is making expensive storage commitment decisions on information that is not reliable enough for the stakes involved. Meanwhile, the day-ahead trading desk might be using a forecast that comfortably exceeds its accuracy requirements, meaning the organization is over-invested in day-ahead accuracy and under-invested in the horizon that actually drives the higher-consequence decisions.
This misalignment is invisible without FHDA because no existing metric connects forecast accuracy at a specific horizon to the decision that consumed it. Traditional forecast accuracy reporting shows a table of numbers by horizon. FHDA shows whether each decision in the value chain is getting the forecast quality it needs.
The decision FHDA supports is whether the dispatch optimizer, trading desk, and storage management systems are each consuming forecasts at horizons where accuracy meets their specific decision requirements, or whether some decisions are being made on forecasts that are not reliable enough for the financial consequences involved.
Formula
For each decision type that consumes a forecast, identify the forecast horizon it actually used. Measure the forecast accuracy at that specific horizon. Compare against the accuracy threshold required for that decision type. The gap between delivered accuracy and required accuracy is the alignment score.
Formula (Notation):
FHDA(d) = Forecast_Accuracy(h_d) minus Accuracy_Threshold(d)
Variable Definitions:
| Variable | Definition |
|---|---|
| d | decision type (day-ahead bid, intra-day dispatch, storage pre-positioning, curtailment pre-staging) |
| h_d | the forecast horizon actually consumed by decision type d, measured in hours |
| Forecast_Accuracy(h_d) | MAPE of the forecast at horizon h_d (percentage, lower is better) |
| Accuracy_Threshold(d) | the maximum acceptable MAPE for decision type d, based on the financial consequence of forecast error at that decision point (percentage, defined per decision type) |
Edge Cases and Measurement Boundaries:
When a decision type consumes multiple forecast horizons (for example, dispatch using both 4-hour and 1-hour forecasts), compute FHDA for each horizon separately and report the worst case. Accuracy thresholds are organizationally defined based on financial exposure analysis. Provide starting recommendations but make clear these require calibration through pilot. If the forecast horizon is not currently logged in the consumption pipeline (which is common), FHDA implementation requires adding horizon metadata to the forecast-to-decision data flow before the metric can be computed.
Proposed formula, pending validation pilots.