Real-Time Leak Detection in Water Distribution Systems

Abstract

Leak and burst detection in water distribution systems represents a compelling and critical issue in water conservation. Not only is the loss of treated and frequently pumped water a waste of resource, money and energy but it is also a potential health risk due to the potential of pollution ingress through cracks. New and more efficient methodologies for the real-time detection of pipe bursts are required to detect these events thus enabling company personnel to react quickly and mitigate the negative impacts of the burst. This paper presents an on-line methodology for automated detection of leaks and bursts by analysing the data collected by real-time sensors. The methodology makes use of several Artificial Intelligence techniques including Wavelets for de-noising of the recorded pressure and/or flow signals, Artificial Neural Networks for the short-term forecasting of future pressure and/or flow signal values, Statistical Process Control analysis for the analysis of discrepancies between the predicted (i.e., expected) and the actually observed signal values, and finally, the Bayesian Network based inference system for classification of discrepancies and raising of alarms. A case study from a real-life DMA in the United Kingdom with simulated (i.e., engineered) burst events is discussed in this paper. The results obtained illustrate that the methodology presented has the potential to yield substantial improvements to the state-of-the-art in real-time water distribution system management by detecting leaks and pipe bursts as they occur in a fast and reliable manner. The results obtained also show that to analyse the discrepancies by using a set of anomaly identification rules applied to consecutive time steps in the Statistical Process Control analysis improves the efficiency and reliability of the real-time leak detection system.