Optimal Economic Energy Management of Microgrid Using Deep Deterministic Policy Gradient (DDPG)

This article presents an optimal economic energy management method of microgrid based on deep reinforcement learning (RL). Traditional energy management often relies on experience-based rules. However, it is difficult to perform optimal dispatch policies when facing complex conditions such as multiple energy sources and load variations. To address this challenge, a method based on deep deterministic policy gradient (DDPG), which is advantageous for continuous control and autonomous learning, is developed to achieve optimal control models that take time-of-use (TOU) pricing, photovoltaic (PV) generation, battery energy storage system (BESS), and load demand into account for optimal management in microgrid. Moreover, to investigate the effectiveness of the proposed method, a telecom data center microgrid system in northern Taiwan was used for validation. The results show the proposed economic energy management method improves the utilization of PV and BESS and achieves the lowest cost.