SMES Damping Controller Design and Real-Time Parameters Tuning for Low-Frequency Oscillation

A novel superconducting magnet energy storage (SMES) damping controller is designed in this paper, which adopts the method of combining integral clearing loop with PI controller, to make sure that the net energy deviation of SMES is zero and avoid deep charging-discharging of SMES during an oscillation process. Moreover, in order to appropriately set the controller parameters considering the uncertainty disturbances of power systems, the paper uses a finite Markov decision process, and adopts a deep reinforcement learning (DRL) based agent to obtain the optimal parameters. After training, the trained agent can act as a decision maker, providing the controller with real-time parameters under different operating condition. Time-domain simulation results show the usefulness of the designed controller and the advantage of the DRL approach.