Efficient Cascading Failure Prediction in Power Systems Based on Node-Edge Switching Graph Convolutional Network

An urgent need for real-time cascading failure assessment has emerged as the increasing penetration of renewable energy sources amplifies uncertainty in power systems. This growing uncertainty poses significant challenges to maintaining safety and reliability. However, the heavy computational burden presents a major bottleneck for the practical online application of cascading failure analysis. Data-driven methods, with their rapid computation and robust feature extraction capabilities, offer a promising solution to this challenge. In this paper, we propose an efficient cascading failure prediction method based on a Node-Edge Switching Graph Convolutional Network (NES-GConv). The proposed framework integrates the dynamic interactions between buses and transmission lines in power systems into a data-driven model, enabling effective extraction of cascading failure features. Experimental validation conducted on the IEEE standard bus system and a practical 341-bus power system demonstrates the effectiveness of the proposed method, achieving state-of-the-art performance.