A Causal Graph Attention Learning for Interpretable Cascading Failure Predictions in Power Systems With Renewable Generation via HVDC

ABSTRACT

The emergence of high‐voltage direct current (HVDC) integrated power systems introduces significant out‐of‐distribution challenges to machine learning–based analysis of cascading failures (CFs). Existing algorithms often suffer from poor interpretability and generalisation in handling diverse CF propagation modes. To address these issues, this paper proposes a novel causal graph attention (CGAT) model that captures critical information from CF events to accurately predict propagation paths and impacts. Specifically, by employing a graph attention mechanism, the model extracts both causal and trivial features as distinctive attention dual subgraphs, which aids in differentiating shortcuts in causal variables and improving prediction accuracy. These subgraphs also provide visual representations that highlight the key elements influencing CF propagation. By incorporating interventions, the CGAT model enhances prediction capabilities across a variety of scenarios that include HVDC control strategies and renewable energy sources. Extensive experiments demonstrate the effectiveness and superiority of the proposed CGAT model, which achieves prediction accuracies of 87.57% for CF chains and 92.89% for power losses on the 30‐bus system. Even on the 2737‐sop system, the accuracies remain at 84.69% and 88.14%, respectively. This model provides a highly efficient online early‐warning tool for preventing large‐scale blackouts in modern power systems.