Multilayer GNN for Predictive Maintenance and Clustering in Power Grids

2025-07-09

AI Summary

Existing predictive maintenance (PdM) models for power systems neglect the coupled dependencies of faults across spatial, temporal, and causal dimensions, leading to frequent unplanned outages. To address this, this paper proposes a novel multilayer graph neural network (GNN) framework. It employs three specialized GNN variants: Graph Attention Networks (GATs) to model spatial adjacency, Graph Convolutional Networks (GCNs) to capture temporal evolution, and Graph Isomorphism Networks (GINs) to encode causal logic. These components are jointly integrated via attention-weighted embedding to achieve unified representation of the three-dimensional dependencies. Additionally, HDBSCAN is applied for risk-driven substation clustering. Experiments demonstrate that the framework achieves an F1-score of 0.8935 for 30-day fault prediction—surpassing conventional methods by over 3%. It successfully identifies eight risk clusters, with the highest-risk cluster exhibiting a mean annual fault count of 388.4; the clustering quality, measured by Silhouette Score, reaches 0.626, significantly enhancing grid resilience analysis and precision operation & maintenance capabilities.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks

Abstract

Unplanned power outages cost the US economy over $150 billion annually, partly due to predictive maintenance (PdM) models that overlook spatial, temporal, and causal dependencies in grid failures. This study introduces a multilayer Graph Neural Network (GNN) framework to enhance PdM and enable resilience-based substation clustering. Using seven years of incident data from Oklahoma Gas & Electric (292,830 records across 347 substations), the framework integrates Graph Attention Networks (spatial), Graph Convolutional Networks (temporal), and Graph Isomorphism Networks (causal), fused through attention-weighted embeddings. Our model achieves a 30-day F1-score of 0.8935 +/- 0.0258, outperforming XGBoost and Random Forest by 3.2% and 2.7%, and single-layer GNNs by 10 to 15 percent. Removing the causal layer drops performance to 0.7354 +/- 0.0418. For resilience analysis, HDBSCAN clustering on HierarchicalRiskGNN embeddings identifies eight operational risk groups. The highest-risk cluster (Cluster 5, 44 substations) shows 388.4 incidents/year and 602.6-minute recovery time, while low-risk groups report fewer than 62 incidents/year. ANOVA (p < 0.0001) confirms significant inter-cluster separation. Our clustering outperforms K-Means and Spectral Clustering with a Silhouette Score of 0.626 and Davies-Bouldin index of 0.527. This work supports proactive grid management through improved failure prediction and risk-aware substation clustering.

Problem

• Enhancing predictive maintenance in power grids using multilayer GNNs
• Improving failure prediction by modeling spatial, temporal, causal dependencies
• Enabling resilience-based substation clustering for proactive grid management

Innovation

• Multilayer GNN integrates spatial, temporal, causal dependencies
• Attention-weighted embeddings fuse Graph Neural Network layers
• HDBSCAN clustering identifies operational risk groups effectively