A Virtual Synchronous Generator Control Strategy with Q-Learning to Damp Low Frequency Oscillation

With the global energy shortage and the deterioration of the environment, distributed generators (DGs) using renewable energy sources have been developed recently. Power electronic equipment is widely used as a common renewable energy generation equipment. However, due to the continuous increase of power electronic equipment in the power system, the power system presents an obvious trend of power electronics. This trend makes the power system face new problems and challenges, among which the most influential one is the oscillation problem in the power system. Therefore, this paper takes virtual synchronous generator (VSG) control strategy as an example, reinforcement learning (RL) algorithm is used to damp power system low frequency oscillation (LFO). Q-Learning, as a commonly used RL algorithm, is taken to replace conventional VSG adaptive controller to adjust the moment of inertia under different fault conditions. Analytical and experimental results are given to indicate that the robust droop VSG control with Q-Learning can effectively improve the system's ability of damping LFO in the power system.