An application of reinforcement learning to voltage control in power system

A novel scheme based on the principle of reinforcement learning algorithm was proposed for the voltage control in power system. The controller adopted the adaptive heuristic critic (AHC) algorithm, which consists of two networks: action selection network and action critic network. The controller parameters are adjusted by given reinforcement signal of system. The numerical simulation example of single-machine infinite power system is given, and the result shows the designed control scheme can maintain desired voltage profile and prevent/stop a developing voltage collapse in system contingencies.