Theory 092/Energy & Environment /Renewable forecasting
Solar Forecast Error Follows Advection
Forecast skill is assumed to fall with cloud. It may fall with how fast the clouds move.
Solar forecasting error scales with cloud advection speed rather than with cloud cover fraction, so a cloudy but still day can be easier to forecast than a partly cloudy windy one.
01The evidence
Forecast errors vary widely at similar cloud cover, and the variation tracks wind speed at cloud level.
02The tension
Forecast models are evaluated by cloud fraction, which predicts error from how much cloud there is and misses the advection that sets how quickly it changes.
03The mechanism
A cloud field moving quickly across the array changes irradiance faster than the forecast update cycle can track, so error grows with advection speed rather than with coverage.
That makes speed the determinant, and predicts error should be largest at intermediate cover with fast advection.
04What this is not
Those models predict error from coverage. This predicts from speed, which explains hard forecasts on windy partly cloudy days.
05Predictions
- P1Error should scale with advection speed.
- P2Windy partly cloudy days should be hardest.
- P3Speed should predict better than coverage.
06Refutation
Error tracking cloud fraction at controlled advection would restore the coverage model.
07What it changes
It would change reserve scheduling.