Research on Optimal Control Strategy of Distributed Photovoltaic Based on Deep Reinforcement Learning

With the proliferation of large-scale distributed photovoltaic (PV) systems integrated into distribution networks, issues such as voltage fluctuations and power flow reversals have gained prominence. To ensure the safe and stable operation of the power system, it becomes imperative to investigate optimal control strategies for distributed PV systems. This paper introduces a novel optimal control approach for distributed PV systems leveraging deep reinforcement learning algorithms. It establishes a Markov decision process framework for optimizing distributed PV operation, with the objectives of minimizing network losses and maximizing PV power generation. The competitive deep Q-network algorithm is employed to train and resolve the optimal scheduling control strategy for distributed PV systems, ultimately providing guidance for photovoltaic operations.