A reinforcement learning approach to power system stabilizer

A reinforcement learning (RL) method is introduced into the optimization design of power system stabilizers (PSS) in this paper. Reinforcement learning (RL) theory is an important branch of the machine learning in the field of artificial intelligence, which is also the general method to deal with Markov decision process (MDP) problems. RL takes learning as trial and error process and maximizes the reward value function by choosing an action depending on the state. In recent years, RL and its application are received increasing attention of international academia. The paper presents two PSS design based on the Q-learning algorithm. One uses Q-learning to optimize the control gain of PSS. The other uses a novel Q-learning controller to replace the conventional PSS completely. The case study shows that both of them are very helpful to enhance the small-disturbance dynamics of power system.