Aggregator-Grid Interactive Building Dual-Layer Price-Responsive Demand Response Scheduling Based on Federated Deep Reinforcement Learning
Wei Zhang, Yiyang Li
Abstract
The energy sector transition to more decentralized and renewable structures requires greater participation by demand-side resources, which can be achieved by establishing a dual-layer model of demand-side resource aggregators based on grid-interactive intelligent buildings. To maximize the use of local flexibility resources connected to the city distribution network, these grid-interactive intelligent buildings typically involve resources such as AC, EV, ESS, etc. Based on price-based demand response, this work proposes a novel solution model based on federated reinforcement learning for this dual-layer structure, aiming to maximize the efficiency of solving the aggregator pricing problem and building energy management problem under the premise of considering the privacy of all parties, and to meet the needs and interests of all parties. Finally, the effectiveness of the proposed method is proven through case study.