Reinforcement Learning-Enabled Seamless Microgrids Interconnection

A reinforcement learning-enabled microgrids interconnection (RL-MIN) control strategy is presented to interconnect microgrids through automatically adjusting the control of distributed energy resources (DERs). RL-MIN is trained to gain the generic knowledge of microgrids and generate corresponding controls for DERs with an aim towards interconnecting systems in a seamless fashion. Using RL-MIN, microgrids can be smoothly connected to form a more stable system when necessary or to energize a large region after a black out occurs. Therefore, the power grid resilience can be significantly enhanced. Numerical results validate the effectiveness of RL-MIN in gaining system operational knowledge, generate corresponding controls for DERs, and interconnect microgrids seamlessly. These salient features make RL-MIN a powerful tool for operating future microgrid systems and contributing to the resilient operations of the bulk power grids.