Graph Attention Networks with Physical Constraints for Anomaly Detection

TL;DR

This paper introduces a hydraulic-aware graph attention network that leverages physical constraints for more accurate and robust anomaly detection in water distribution systems, effectively capturing spatio-temporal patterns.

Contribution

It presents a novel graph attention model incorporating conservation law violations and multi-scale aggregation, improving detection accuracy and robustness over existing methods.

Findings

Achieved an F1 score of 0.979 on BATADAL dataset.

Outperformed baseline models with a 3.3 percentage point gain.

Demonstrated high robustness under 15% parameter noise.

Abstract

Water distribution systems (WDSs) face increasing cyber-physical risks, which make reliable anomaly detection essential. Many data-driven models ignore network topology and are hard to interpret, while model-based ones depend strongly on parameter accuracy. This work proposes a hydraulic-aware graph attention network using normalized conservation law violations as features. It combines mass and energy balance residuals with graph attention and bidirectional LSTM to learn spatio-temporal patterns. A multi-scale module aggregates detection scores from node to network level. On the BATADAL dataset, it reaches $F1=0.979$, showing $3.3$pp gain and high robustness under $15\%$ parameter noise.