Deep Reinforcement Learning-Aided Pre-Positioning of Mobile Wind Turbines to Enhance Power Distribution System Resilience

Compared to stationary wind turbines, mobile wind turbines (MWTs) can move within the transportation system and supply power to electrical microgrids. With the goal to enhance the power distribution system (PDS) resilience, MWTs can help in service restoration and supply power to islands that are separated from the main grid during emergencies. However, MWTs typically start to operate only after faults driven by high-impact low-probability (HILP) events occur. To improve service restoration efficiency with MWTs, we propose a framework based on deep reinforcement learning (DRL) for MWT pre-positioning. This framework uses the DRL agent's training rewards to determine the pre-position of MWTs before HILP emergencies arise. Since the MWT movement is a discrete decision, we applied the Deep Q-learning (DQL) and Double Deep Q-learning (DDQL) algorithms as the DRL agent models. The agent provides MWTs actions on an integrated system comprising an 11-node transportation system (TS) and a 33-bus PDS.