Data-Driven Distributed Q-Learning Droop Control for Frequency Synchronization and Voltage Restoration in Isolated AC Micro-Grids

By treating each distributed generator (DG) in the isolated AC micro-grids (MG) as an intelligent agent with the adjacent information sharing mechanism, we propose a fully distributed data-driven reinforcement learning (RL) droop control method for autonomous frequency synchronization as well as voltage restoration. Since the proposed distributed control is indeed a data-driven self-learning approach, it is very suitable for plug-and-play operations of isolated AC MGs even when the operating conditions are deviated from the nominal condition under study once sufficient operational data of each DG is well-collected. To validate the performance of the proposed method, the proposed algorithm was implemented on Matlab/Simulink environment. Simulation results of modified IEEE 34-node distribution system demonstrate the effectiveness of the proposed distributed data-driven Q-learning droop control for plug-and-play operations of isolated AC MGs.