Reinforcement learning algorithms in AC, DC, and hybrid microgrids applications: A comprehensive review

Over the past few years, the adoption of modern and intelligent energy systems, such as smart grids, microgrids, and smart buildings, has significantly increased. This surge in adoption is attributed to their advanced features, including bidirectional power flows, sophisticated metering systems, and the efficient integration of renewable energy resources. Despite the benefits, the growing adoption of these systems introduces new challenges in various aspects of power system management, particularly in operation and control. Additionally, the employment of advanced sensors and intelligent meters generates vast amounts of data, paving the way for innovative, data-driven approaches to tackle complex operational and control challenges. Among these strategies, Reinforcement Learning (RL) has emerged as a preferred technique for its applications in Energy Management System (EMS), addressing optimization challenges, controlling power flow, and beyond. This review paper provides a comprehensive analysis of RL in the context of microgrid systems. It explores RL’s fundamental principles, classifies the major algorithm types, and evaluates their applications across diverse microgrid architectures. Moreover, the paper critically examines the challenges associated with applying RL in microgrid systems and identifies promising avenues for future research, emphasizing both the limitations of current approaches and the domains that demand further investigation.