Graph Neural Networks for Pressure Estimation in Water Distribution Systems
Графовые нейронные сети для оценки давления в системах водоснабжения
2024-07-01
SCID: 54.1/pzt9u2yg
Discuss with AI
GAT with Residual Connectionsgraph neural networkspressure estimationrandom sensor placementwater distribution networks
Figures from the paper
Abstract (AI)
Abstract Pressure and flow estimation in water distribution networks (WDNs) allows water management companies to optimize their control operations. For many years, mathematical simulation tools have been the most common approach to reconstructing an estimate of the WDNs hydraulics. However, pure physics‐based simulations involve several challenges, for example, partially observable data, high uncertainty, and extensive manual calibration. Thus, data‐driven approaches have gained traction to overcome such limitations. In this work, we combine physics‐based modeling and graph neural networks (GNN), a data‐driven approach, to address the pressure estimation problem. Our work has two main contributions. First, a training strategy that relies on random sensor placement making our GNN‐based estimation model robust to unexpected sensor location changes. Second, a realistic evaluation protocol that considers real temporal patterns and noise injection to mimic the uncertainties intrinsic to real‐world scenarios. As a result, a new state‐of‐the‐art model, GAT with Res idual Connections, for pressure estimation is available. Our model surpasses the performance of previous studies on several WDNs benchmarks, showing a reduction of absolute error of ≈40% on average.
Key Findings
1
Across several water distribution network benchmarks, the model reduces absolute error by approximately 40% on average compared with previous studies.
2
The evaluation protocol incorporates realistic temporal patterns and injected noise to emulate real-world hydraulic data uncertainties.
3
The proposed Graph Attention Network with residual connections establishes a new state-of-the-art model for pressure estimation.
4
The study combines physics-based water distribution modeling with graph neural networks to estimate pressures under partial observability and uncertainty.
5
Training with randomly placed sensors improves robustness to unexpected changes in sensor locations.
Research Object
Pressure and hydraulic state of water distribution networks (WDNs)
Research Subject
Accurate and robust pressure estimation under partial observability, sensor-location changes, measurement noise, and real-world temporal patterns using physics-informed graph neural networks
Publication Details
Publication Date
2024-07-01
Journal
Publisher
ISSN
Cited by
69
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest
References available in scid.ai12
Deep Residual Learning for Image Recognition2016
Adam: A Method for Stochastic Optimization2014
Semi-Supervised Classification with Graph Convolutional Networks2016
PyTorch: An Imperative Style, High-Performance Deep Learning Library2019
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale2020
A Comprehensive Survey on Graph Neural Networks2020
Proceedings of the 24th international conference on Machine learning2007
Neural Message Passing for Quantum Chemistry2017
Randaugment: Practical automated data augmentation with a reduced search space2020
Graph Convolutional Neural Networks for Web-Scale Recommender Systems2018
Graph convolutional networks: a comprehensive review2019
Knowledge Graph Completion: A Review2020