NELGSYFeb 11, 2025

Scalable and Robust Physics-Informed Graph Neural Networks for Water Distribution Systems

arXiv:2502.12164v12 citationsh-index: 11IJCNN
Originality Highly original
AI Analysis

This work addresses the challenge of efficient planning and management of water distribution systems, which is crucial for urban infrastructure and climate change mitigation.

The authors tackled the problem of modeling water distribution systems, achieving better performance than the current state-of-the-art deep learning model and demonstrating scalability to larger systems. Their approach also improved robustness to out-of-distribution input features.

Water distribution systems (WDSs) are an important part of critical infrastructure becoming increasingly significant in the face of climate change and urban population growth. We propose a robust and scalable surrogate deep learning (DL) model to enable efficient planning, expansion, and rehabilitation of WDSs. Our approach incorporates an improved graph neural network architecture, an adapted physics-informed algorithm, an innovative training scheme, and a physics-preserving data normalization method. Evaluation results on a number of WDSs demonstrate that our model outperforms the current state-of-the-art DL model. Moreover, our method allows us to scale the model to bigger and more realistic WDSs. Furthermore, our approach makes the model more robust to out-of-distribution input features (demands, pipe diameters). Hence, our proposed method constitutes a significant step towards bridging the simulation-to-real gap in the use of artificial intelligence for WDSs.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes