Deep Reinforcement Learning in Power Distribution Systems: Overview, Challenges, and Opportunities
Yuanqi Gao, Nanpeng Yu
Abstract
To facilitate the integration of distributed energy resources and improve existing operational strategies, power distribution systems have seen a rapid proliferation of deep reinforcement learning (DRL) based applications. DRL approach is well suited for dynamic, complex, and uncertain operational environments such as power distribution systems. This paper reviews the rapidly growing body of literature that develops applications of reinforcement learning in power distribution systems. These applications include active grid management, energy management system, retail electricity market, and demand response. This paper also summarizes the challenges of deploying DRL based solutions in distribution systems such as safety, robustness, interpretability, and sample efficiency. Finally, the research opportunities that can be pursued to address the challenges are provided.