A Classification Model for Real-time Identification of Solar Curtailment in the California Grid

Abstract

Higher penetration of renewable generation has led to a large increase in 'curtailment', i.e. periods in which generation from renewable resources is lost as a result of insufficient demand or lacking grid transmission capacity. Electricity consumers could help avoid curtailment - and reduce emissions - by shifting their consumption to time periods with curtailment. However, their ability to do so is severely limited by a lack of real-time curtailment information. To address this issue, we present a classification model based on gradient-boosted learning which identifies solar curtailment in real time for the California grid. The model relies only on publicly-available, real-time grid information, and is tuned specifically to capture time periods with high curtailment. Our analysis shows that the proposed classifier can precisely and reliably identify solar curtailments.