Method note · measured
Day-ahead forecast error
What it measures
How far the day-ahead wind and solar forecast misses and which way it leans, nationally and per control area (the region one grid operator balances).
Inputs
- SMARD's day-ahead forecast for wind onshore, wind offshore and solar, nationally and per control area, quarter-hourly
- SMARD's realised generation for wind onshore, wind offshore and solar, nationally and per control area, quarter-hourly
Method and weighting
The error of a quarter-hour is the forecast output minus the actual output. The miss is the sum of the absolute errors over the sum of actual output across the window's quarter-hours, an energy-weighted share of how far the forecast typically lands from what happened.
The bias, which way the forecast typically leans, is the sum of the signed errors over the same actual total, positive when the forecast ran high and negative when it ran low. Both figures are worked out by month, for each control area and nationally, and the days with the largest absolute error are listed for the national figure.
Window
The rolling 12 months to the last complete day the source has published, refreshed on the 1st and 15th of each month. The current window runs from to . Data last refreshed on .
Known limits
- The forecast is the one the grid operators published the afternoon before, so later same-day updates are left out of it.
- A control area with a small fleet gives a noisy miss and bias.
Used on
- Forecast Error in When the Grid Is Tested
Sources: SMARD. Each is described on the Data sources page.