Deep reinforcement learning based home energy management system with devices operational dependencies

Advanced metering infrastructure and bilateral communication technologies facilitate the development of the home energy management system in the smart home. In this paper, we propose an energy management strategy for controllable loads based on reinforcement learning (RL). First, based on the mathematical model, the Markov decision process of different types of home energy resources (HERs) is formulated. Then, two RL algorithms, i.e. deep Q-learning and deep deterministic policy gradient are utilized. Based on the living habits of the residents, the dependency modes for HERs are proposed and are integrated into the reinforcement learning algorithms. Through the case studies, it is verified that the proposed method can schedule HERs properly to satisfy the established dependency modes. The difference between the achieved result and the optimal solution is relatively small. © The Author(s), under exclusive licence to Springer-Verlag GmbH, DE part of Springer Nature 2021.