Graph Attention Networks with Physical Constraints for Anomaly Detection

Mohammadhossein Homaei, Iman Khazrak, Ruben Molano, Andres Caro, Mar Avila
1/18/2026
cs.LGcs.CR

No official code found. Best implementation: Homaei/Physics-GAT (1 star, MIT, last commit 2025-11-06).

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.

View on arXivView PDF

Code Implementations(1)

Homaei/Physics-GATCommunity68%
MIT

Graph Attention Networks with Physical Constraints for Anomaly Detection

10Jupyter NotebookNov 6, 202511 months agoMIT

Similar Papers

Graph Neural Network-Based Anomaly Detection for River Network Systems

85% match1 repo362023

Unsupervised Online Detection of Pipe Blockages and Leakages in Water Distribution Networks

83% match2025

Algorithm-Informed Graph Neural Networks for Leakage Detection and Localization in Water Distribution Networks

82% match2024

Graph Neural Networks for Pressure Estimation in Water Distribution Systems

81% match1 repo272023

Spatio-Temporal Attention Graph Neural Network: Explaining Causalities With Attention

81% match2026

Explainable Fuzzy GNNs for Leak Detection in Water Distribution Networks

80% match2026

Cite this paper

@article{homaei2026graph,
  title  = {Graph Attention Networks with Physical Constraints for Anomaly Detection},
  author = {Mohammadhossein Homaei and Iman Khazrak and Ruben Molano and Andres Caro and Mar Avila},
  year   = {2026},
  eprint = {2601.12426},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2601.12426v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2601.12426.svg)](https://codeofpaper.com/paper/2601.12426)

Discussion