Deep Q-Network based Adaptive Robustness Parameters for Virtual Synchronous Generator

This paper investigates a reinforcement learning based adaptive robustness parameter tunning approach for the virtual synchronous generator (VSG). Particularly, a deep Q-network (DQN) algorithm is employed to realize the real-time parameter tuning of inertia and damping coefficient in the VSG controller. The proposed parameter tuning approach is confirmed by the simulation results and compared with the conventional VSG controller with fixed parameters.