Reinforcement Learning Based Energy Storage Units Scheduling Considering Point of Common Coupling Constraint

In this paper, agent-aggregators cooperation has been introduced in Smart Microgrid (SMG) consisting of household users, Photo Voltaic array (PV), and Local Energy Storage System (LESS). The SMG has Energy Management System called agent deal with two aggregators to purchase the energy for the consumers. Each aggregator supplying a limited level of energy from different energy resource, where one aggregator depend on Community Energy storage System to provide power in one direction to the users and the other aggregator supplying power from the utility grid. The job of the agent is to decide the optimum amount of purchasing power from the aggregators at the lowest cost to supply the users and to charge the LESS taking into account the constraints of supplying limitations for both CESS and utility grid at Point of Common Coupling. Each aggregator supplies a limited level of energy from different energy resource, where one aggregator depend on the Community Energy storage System (CESS) to provide power in one direction to the users and the other aggregator supplying power from the utility grid. The job of the agent is to decide the optimum amount of purchasing power from the aggregators at the lowest cost to provide the users and to charge the LESS taking into account the constraints of providing limitations for both CESS and utility grid at Point of Common Coupling (PCC).A model-free Reinforcement Learning (RL) method is used as an essential strategy to achieve the proposed task using Deep Q-Network (DQN). The simulation result is done and the results are verified using Improved Particular Swarm (IPSO) as a benchmark method. The obtained results show the efficiency and effectiveness of the method used