Q-Learning-Oriented Distributed Energy Management of Grid-Connected Microgrid

In this paper11This work is supported by Niroo Research Institute (NRI)., reinforcement learning (RL) is used for energy management of agent based microgrid (MG). The Grid connected MG that contains wind turbine, fuel cell (FC), diesel generator and electric vehicle (EV) to supply its demands, is modeled as a multi-agent system (MAS). The DER and customer are considered as self-interested agents that try to maximize their profits and optimize their behavior. These agents use RL to interact with each other in distributed manner without any direct communication. The market operator of MG is responsible to gather agents' data that have been submitted and clears the market to meet the desired goals. Modeling the stochastic nature of wind power generation and demand fluctuation of customer agents, implementing demand side management program for customer agents, besides taking into account the technical constraint of diesel generator, FC and EV agent are the main strengths of this paper. The simulation results confirm the efficiency of the proposed approach.