Volume 66

An Intelligent Diagnosis Method for Power Distribution Network Energy Security Vulnerability Based on Random Walk and Embedding Learning Rongnan Deng, Jiani Zhou, Meng Song, Li Sun

Abstract

Ensuring energy security under extreme weather conditions is a major challenge faced by modern urban power grids. Flood disasters cause large-scale power outages by submerging distribution equipment, seriously endangering the reliability and resilience of power supply systems. To improve the disaster prevention capability of distribution networks and fortify the defense line for energy security, this paper proposes an intelligent diagnosis framework integrating physical mechanisms and graph learning. First, this study established a high-precision flood fault-power flow reconfiguration model to quantify the failure risk of key equipment such as ring main units (RMUs) under different flood depths. Furthermore, a two-layer random walk network was innovatively constructed: by simulating the adaptive walk of virtual particles between layers, the topological connectivity and electrical vulnerability of nodes under disaster conditions are captured simultaneously. Finally, the Skip-gram model was adopted to map distribution network nodes into low-dimensional embedding vectors, thereby realizing unsupervised and efficient intelligent diagnosis of node vulnerability. Case studies show that this method can accurately identify the key vulnerable nodes that are most likely to trigger systemic collapse during floods. The research results can provide precise decision support for preventive reinforcement, emergency control, and rapid recovery of distribution systems under extreme weather, and serve as a key intelligent technology for building a more resilient and safer future energy system.

Keywords Energy Security, Two-Layer Walk Network, Distribution Network Resilience, Graph Theory Algorithms