Deep reinforcement learning for energy management in a microgrid with flexible demand

Abstract In this paper, we study the performance of various deep reinforcement learning algorithms to enhance energy management system a microgrid. We propose novel microgrid model that consists wind turbine generator, an storage system, set thermostatically controlled loads, price-responsive and connection main grid. The proposed is designed coordinate among different flexible sources by defining priority resources, direct demand control signals, electricity prices. Seven were implemented are empirically compared in paper. numerical results show differ widely their ability converge optimal policies. By adding experience replay semi-deterministic training phase well-known asynchronous advantage actor–critic? algorithm, achieved highest as well convergence near-optimal