Short-Term Generation Forecasting for Renewables Trader
A renewables trading desk was being penalised by imbalance charges due to inaccurate short-term wind generation forecasts.
Headline outcome
28% Lower forecast error
The challenge
Generic vendor forecasts didn't capture local terrain effects on wind farms, and the desk paid heavily for imbalance whenever they were wrong.
“Built a probabilistic short-term forecasting platform combining numerical weather prediction with site-level ML.”
Our approach
-
01
Ingested multiple weather providers and site SCADA data into a unified platform.
-
02
Trained site-specific ML models with quantile outputs for trading decisions.
-
03
Integrated into trading workflows with refreshes every 15 minutes.
The result
Forecast error dropped sharply, imbalance costs fell into line with best-in-class peers, and the desk now runs a standardised forecasting platform across its growing portfolio.
28%
Lower forecast error
£4.5M
Annual imbalance saved
15min
Refresh cadence
Be our next success story