Pre-Disaster Inspection Method for Distribution Networks in Typhoon Scenarios

The inspection of distribution networks in typhoon scenarios faces multiple challenges, including the uncertainty of disaster situations and the coupling of time and space. This paper constructs a pre-disaster inspection allocation model based on reinforcement learning algorithms. Additionally, state decomposition and action masking techniques are designed to optimize the performance of the reinforcement learning algorithm. A testing environment for the inspection and recovery of distribution networks is established, and simulations are conducted under various typhoon intensities and fault scenarios. Comparing the results with traditional reinforcement learning algorithms, the proposed model demonstrates superior overall performance in power restoration, providing valuable guidance for dynamic pre-disaster inspections.