Data-based optimal microgrid energy management with a discrete iterative reinforcement learning

This article developed a reinforcement learning (RL) scheme to obtain the management strategy for microgrids with a novel discrete adaptive law which is facilitated to learn the critic network weight. The primary objective is to extend the battery's lifespan while reducing the microgrid cost. First, the system model and cost function are defined with the battery and load power. Then an iterative RL algorithm incorporating the new adaptive law is proposed to learn optimal microgrid control. To support this, both action and critic networks are implemented using neural network (NN) approximators. This paper also thoroughly examines the convergence behaviour of the NN and the stability of the overall control system. Numerical simulations are conducted to demonstrate the effectiveness of the proposed method.