Learning Assisted Resilient Load Restoration with Distributed Mobile Resources

Extreme natural disasters are characterized by high impact and low probability, which can cause significant damage to power systems. Distributed Mobile Resources (DMRs), including mobile energy storage systems (MESSs), heavy-duty electric trucks, and private electric vehicles (EVs), can enhance the resilience of power systems in the face of such disasters through their flexible charging and discharging capabilities. In this paper, we introduce a Deep Reinforcement Learning (DRL)-based method for efficiently dispatching DMRs to achieve the rapid restoration of critical loads in the event of disasters, thereby enhancing the resilience of the power grid. Firstly, this paper creates a detailed physical hurricane model and calculates equivalent wind speeds for distinct segments of distribution lines at various intervals, estimating the likelihood of line damage. DMRs, which are treated as agents in our approach, formulate real-time strategies for routing and discharging, informed by damage assessments and their current operational state. Our model-free algorithm enhances the accuracy in characterizing disaster scenarios and expedites the formulation of optimal scheduling strategies. Simulation studies demonstrate that DMRs can collaborate effectively, substantially enhancing the system's resilience in disaster situations.