Hybrid Energy Storage Control Method For DC Microgrid Based On Deep Deterministic Policy Gradient

This research offers a deep reinforcement learning-based optimal control technique for a DC microgrid hybrid energy storage system (HESS) to increase system stability in the face of uncertainty in renewable energy generation and load volatility. First, the hybrid energy storage system is modeled and analyzed, and the voltage dynamic characteristics of the DC bus and current inner-loop tracking characteristics are investigated; a framework of the input/output-based deep reinforcement learning algorithm for the hybrid energy storage system's inner-loop is provided; and the reward function and neural network are designed to achieve optimal control of the hybrid energy storage system's DC/DC converter is provided. Finally, the suggested control strategy is compared to the PI double-loop control by modeling real-world scenarios of abrupt changes in renewable energy power and load power in the MATLAB/Simulink simulation program, and the effectiveness of the proposed control approach is demonstrated.