Client context
A power generation operator with a mixed renewable-and-thermal portfolio bidding into a regional wholesale market.
Problem size
Point forecasts for generation looked precise but were frequently wrong, leaving trading and dispatch exposed to imbalance costs across a multi-GW portfolio.
What KEPLER did
- Engineered a feature set from weather-ensemble forecasts and SCADA generation data
- Built probabilistic P10–P90 forecast models in place of single-number forecasts
- Delivered the distribution into trading and dispatch — bid and schedule against a range
- Tracked forecast accuracy and imbalance exposure in Power BI
Results
30% lower forecast error, with reduced imbalance-cost exposure and dispatch decisions made against a distribution instead of a guess.
Anonymized client engagement; identifying details withheld for confidentiality.