Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AI-Driven Forecasting and Monitoring of Urban Water System

Qiming Guo, Bishal Khatri, Hua Zhang, Wenlu Wang

TL;DR

This paper tackles leaks and anomalies in underground urban water networks where dense sensor deployment is impractical. It introduces an AI-driven framework that combines sparse manhole-based sensing with hydraulic simulations and a spatiotemporal graph neural network, HydroNet, operating on a directed graph $G=(V,E)$ with edge-attribute embedding via $W_a$ and messages $m_{ij}=f_{message}(h_i^{(t)}, W_a a_{ij})$, using a lookback window $L=12$. The Real World Water (RWW) dataset from a campus network demonstrates that HydroNet outperforms baselines, achieving depth MAE $0.0085$ ft and flow MAE $0.0038$ cfs, validating the approach for accurate network-wide hydraulic forecasting. This work enables robust anomaly detection and scalable monitoring of underground water infrastructure, with potential applicability to diverse urban networks and extension to other pipeline systems.

Abstract

Underground water and wastewater pipelines are vital for city operations but plagued by anomalies like leaks and infiltrations, causing substantial water loss, environmental damage, and high repair costs. Conventional manual inspections lack efficiency, while dense sensor deployments are prohibitively expensive. In recent years, artificial intelligence has advanced rapidly and is increasingly applied to urban infrastructure. In this research, we propose an integrated AI and remote-sensor framework to address the challenge of leak detection in underground water pipelines, through deploying a sparse set of remote sensors to capture real-time flow and depth data, paired with HydroNet - a dedicated model utilizing pipeline attributes (e.g., material, diameter, slope) in a directed graph for higher-precision modeling. Evaluations on a real-world campus wastewater network dataset demonstrate that our system collects effective spatio-temporal hydraulic data, enabling HydroNet to outperform advanced baselines. This integration of edge-aware message passing with hydraulic simulations enables accurate network-wide predictions from limited sensor deployments. We envision that this approach can be effectively extended to a wide range of underground water pipeline networks.

AI-Driven Forecasting and Monitoring of Urban Water System

TL;DR

This paper tackles leaks and anomalies in underground urban water networks where dense sensor deployment is impractical. It introduces an AI-driven framework that combines sparse manhole-based sensing with hydraulic simulations and a spatiotemporal graph neural network, HydroNet, operating on a directed graph with edge-attribute embedding via and messages , using a lookback window . The Real World Water (RWW) dataset from a campus network demonstrates that HydroNet outperforms baselines, achieving depth MAE ft and flow MAE cfs, validating the approach for accurate network-wide hydraulic forecasting. This work enables robust anomaly detection and scalable monitoring of underground water infrastructure, with potential applicability to diverse urban networks and extension to other pipeline systems.

Abstract

Underground water and wastewater pipelines are vital for city operations but plagued by anomalies like leaks and infiltrations, causing substantial water loss, environmental damage, and high repair costs. Conventional manual inspections lack efficiency, while dense sensor deployments are prohibitively expensive. In recent years, artificial intelligence has advanced rapidly and is increasingly applied to urban infrastructure. In this research, we propose an integrated AI and remote-sensor framework to address the challenge of leak detection in underground water pipelines, through deploying a sparse set of remote sensors to capture real-time flow and depth data, paired with HydroNet - a dedicated model utilizing pipeline attributes (e.g., material, diameter, slope) in a directed graph for higher-precision modeling. Evaluations on a real-world campus wastewater network dataset demonstrate that our system collects effective spatio-temporal hydraulic data, enabling HydroNet to outperform advanced baselines. This integration of edge-aware message passing with hydraulic simulations enables accurate network-wide predictions from limited sensor deployments. We envision that this approach can be effectively extended to a wide range of underground water pipeline networks.

Paper Structure

This paper contains 7 sections, 2 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Remote Sensing System
  4. RWW Dataset: Collection and Characterization
  5. HydroNet Modeling
  6. Evaluation
  7. Conclusion

Figures (3)

  • Figure 1: Graph representation of the campus sewer network. Nodes represent manholes with flow and depth measurements, while directed edges represent pipes with physical attributes.
  • Figure 2: Schematic of the remote sensor system and its deployment configuration.
  • Figure 3: Analysis of the RWW dataset features: (a) Time-series visualization of aggregated flow rate across the network, highlighting periodic variations; (b) Average daily flow patterns by day of the week; (c) Example autocorrelation functions for depth and inflow at node 92090090, highlighted in Figure \ref{['fig:water-gnn']}(a); (d) Correlation matrix of edge features