Demand Response Flexibility Prediction of Residents Based on Federated Learning

Demand response (DR) flexibility prediction are of great importance to load aggregators (LG) and operators for bidding and scheduling decisions. Meanwhile, residents are increasingly concerned about data privacy protection. Therefore, a federated learning based framework is proposed for residential load demand response flexibility prediction. The dataset is constructed based on the large scale demand response project implemented in Jiangxi Province, including load curves, sociodemographic information, and historical demand response information of DR events. Compared with the centralized framework, the proposed framework improves the accuracy of the prediction model with privacy protection on the dataset.