Inductive Power Grid Cascading Failure Analysis with GRU-Gated Graph Attention

2026-05-07

AI Summary

Existing cascade failure analysis methods struggle to generalize to unseen power grids due to being trained on a single grid. This work proposes a unified model based on a GRU-gated graph attention network, trained jointly across multiple grids, which enables zero-shot cross-grid transfer without fine-tuning. The model dynamically captures the mechanisms of information retention and discarding at nodes during cascading processes. It achieves, for the first time in power systems, zero-shot prediction of failures across both different grids and time periods. Evaluated on multiple previously unseen grids, the model identifies significantly more vulnerable transmission lines than structural and electrical baseline methods, thereby overcoming the generalization limitations inherent in conventional approaches.

Abstract

Identifying vulnerable transmission lines in power grids before a cascading failure occurs is challenging: existing methods can learn inter-line failure correlations from cascade data, but they are trained and evaluated on a single grid, and transferring the learned knowledge to an unseen grid remains an open problem. We address this by training a single Gated Recurrent Unit (GRU)-gated Graph Attention Network on combined cascading failure data from limited training grids and applying it directly to any unseen grid without retraining. A GRU gate controls what information each node retains or discards at each cascade iteration. Empirical evaluation shows that the model transfers zero-shot to multiple new grids spanning inter-time and inter-domain settings. Using information extracted from the trained model, we consistently identify more vulnerable lines than established structural and electrical baselines.

Problem

• cascading failure
• power grid
• transfer learning
• vulnerable line identification
• zero-shot transfer

Innovation

• GRU-Gated Graph Attention
• Zero-shot Transfer
• Cascading Failure Analysis
• Power Grid Vulnerability
• Inductive Learning