UAV-Assisted Emergency Restoration Strategy for Distribution Systems With Deep Learning

After natural disasters, distribution system (DS) restoration is often challenged by the failure of some recovery functions due to the damage of communication facilities. This paper proposes an unmanned aerial vehicle (UAV)-assisted emergency restoration strategy. Specifically, the system considers the wireless communication characteristics between the UAV and the feeder terminal units (FTUs), and constructs probabilistic Line-of-Sight (LoS) and Non-Line-of-Sight (NLoS) link propagation constraints. Based on this, the UAV flight trajectory and DS reconfiguration plan are jointly optimized to maximize load restoration. This joint optimization problem is a complex non-convex optimization problem. In this paper, a deep neural network (DNN) approach is introduced to transform the problem into a mixed-integer linear programming (MILP). The results show that the joint restoration strategy can fully exploit the UAV communication recovery potential and enhance DS resilience.