EGU General Assembly 2021
© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

Leakage detection in water pipe networks using machine learning

Yu Li1,2, Jinhui Jeanne Huang2, and Ran Yan2
Yu Li et al.
  • 1Shenzhen research institute, Nankai University, Shenzhen 518000, PR China(
  • 2Nankai University, Sino-Canada Joint R&D Centre for Water and Environmental Safety, College of Environmental Science and Engineering, Tianjin, China (

Leakage in the water supply system is a world problem that happens everywhere, not only in China but also in Japan, the US, and Europe. It not only results in the waste of water resources but also raises safety issues in drinking water. The traditional solution is the Minimum Night Flow method with manual leak detectors. This solution could only find leakage at night. The engineers have to search the leaking point randomly using leak detectors. It not only highly relies on domain knowledge and expertise but is also labor-consuming. The response time is quite long, might be a couple of days to several days. Here, time series analysis based on a dynamic time warping algorithm is used to detect anomalies in time series of pressure stations and flow stations, and the risk coefficient of each pipe network is determined by using a neural network combined with existing data. The water treatment plants don't even have to install new sensors if the budget is limited.

How to cite: Li, Y., Huang, J. J., and Yan, R.: Leakage detection in water pipe networks using machine learning, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-9345,, 2021.

Corresponding displays formerly uploaded have been withdrawn.