Deep Reinforcement Learning for Control of Microgrids: A Review

A microgrid is widely accepted as a prominent solution to enhance resilience and performance in distributed power systems. Microgrids are flexible for adding distributed energy resources in the ecosystem of the electrical networks. Control techniques are used to synchronize distributed energy resources (DERs) due to their turbulent nature. DERs including alternating current, direct current and hybrid load with storage systems have been used in microgrids quite frequently due to which controlling the flow of energy in microgrids have been complex task with traditional control approaches. Distributed as well central approach to apply control algorithms is well-known methods to regulate frequency and voltage in microgrids. Recently techniques based of artificial intelligence are being applied for the problems that arise in operation and control of latest generation microgrids and smart grids. Such techniques are categorized in machine learning and deep learning in broader terms. The objective of this research is to survey the latest strategies of control in microgrids using the deep reinforcement learning approach (DRL). Other techniques of artificial intelligence had already been reviewed extensively but the use of DRL has increased in the past couple of years. To bridge the gap for the researchers, this survey paper is being presented with a focus on only Microgrids control DRL techniques for voltage control and frequency regulation with distributed, cooperative and multi agent approaches are presented in this research.