Reinforcement Learning-Based Intelligent Control Strategies for Optimal Power Management in Advanced Power Distribution Systems: A Survey

Intelligent energy management in renewable-based power distribution applications, such as microgrids, smart grids, smart buildings, and EV systems, is becoming increasingly important in the context of the transition toward the decentralization, digitalization, and decarbonization of energy networks. Arguably, many challenges can be overcome, and benefits leveraged, in this transition by the adoption of intelligent autonomous computer-based decision-making through the introduction of smart technologies, specifically artificial intelligence. Unlike other numerical or soft computing optimization methods, the control based on artificial intelligence allows the decentralized power units to collaborate in making the best decision of fulfilling the administrator’s needs, rather than only a primitive decentralization based only on the division of tasks. Among the smart approaches, reinforcement learning stands as the most relevant and successful, particularly in power distribution management applications. The reason is it does not need an accurate model for attaining an optimized solution regarding the interaction with the environment. Accordingly, there is an ongoing need to accomplish a clear, up-to-date, vision of the development level, especially with the lack of recent comprehensive detailed reviews of this vitally important research field. Therefore, this paper fulfills the need and presents a comprehensive review of the state-of-the-art successful and distinguished intelligent control strategies-based RL in optimizing the management of power flow and distribution. Wherein extensive importance is given to the classification of the literature on emerging strategies, the proposals based on RL multiagent, and the multiagent primary secondary control of managing power flow in micro and smart grids, particularly the energy storage. As a result, 126 of the most relevant, recent, and non-incremental have been reviewed and put into relevant categories. Furthermore, salient features have been identified of the major positive and negative, of each selection