Power Grid Reactive Voltage Control Method Based on Improved Reinforcement Learning

This paper proposes a grid reactive voltage control method based on improved reinforcement learning. Based on the theory of reinforcement learning, the convolutional neural network (CNN) is combined with Q learning, and the gradient descent method is used to update the weights to improve the reinforcement learning algorithm. Taking the switching instructions of the discrete reactive power regulation equipment as the control variable, and the voltage deviation of each node in the network as the objective function, the grid voltage control optimization model is established. Select the appropriate action strategy based on the grid input data, use the power flow calculation to check the optimization effect, update the weights, and iteratively optimize the newly generated input state quantities to form a grid voltage deviation solution process based on improved reinforcement learning. By optimizing the programming parameters to improve the control performance and reduce the calculation error of the voltage control. The IEEE 14-bus system is used for simulation analysis on the Matpower platform, and the experimental results show the effectiveness of the proposed strategy.