Real-Time Leak Localization Using Hydraulic Modeling

Abstract

Leaks are present in all Water Distribution Networks (WDN), and they degrade the water efficiency of the utility, leading to wasted economical and environmental resources. Lack of drinking water affects almost half of the planet’s population and enhancing the water efficiency of the utilities by finding the leaks in the network faster is a way to mitigate this issue. Advances in measuring technology such as the development of Supervisory Control and Data Acquisition (SCADA) and smart sensor systems has increased the amount of data in the field, which has led to growth of available modern leak detection and localization systems for the utilities and enabled real-time implementations. The enhanced data availability has also revealed issues in the underlying systems. Malfunctions in measurement devices such as lagging or drifting has been identified as a major issue and hinders the implementation of robust and reliable data driven solutions.

This study aimed to evaluate a near real-time hydraulic modelling-based method for localizing leaks in WDN to assess its accuracy, validate approaches and identify its requirements and limitations. A complete, large-scale hydraulic network model and SCADA data were utilized to find previously verified leaks and evaluating the localization results. The results were verified by conducting refence samples with identical conditions without actual leak event present and comparing the result to the actual localization result. The influence of enhanced demand allocation via utilization of smart sensor consumption data in hydraulic simulation results was also investigated.

The results consisted of 28 one-hour leak localization runs in 6 District Metered Areas (DMA) in the network, and two different simulations with enhanced and general demand allocations. The hydraulic distance between the identified leak locations and the leak locations had the median of 970 meters, but the 10th percentile of 213 meters and 90th percentile of 3231 meters indicated high variability in the performance. The parameter correlation analysis revealed that while the results are still hydraulically intuitive, no single parameter was found that had strong correlation with performance, but r-value of -0.19 of pressure drop indicated potential to affect the results. This highlights that the performance is influenced by the combined hydraulic characteristics of individual DMAs. Analysis reference samples revealed that the system exhibits weak robustness to unfavourable conditions.

The analysis of simulations with enhanced and general demand allocations consisted of two DMAs where the average pressure was decreased by 0.3 meters and one DMA where average pressure increased by 0.1 meters, whereas other DMAs did not exhibit distinguishable differences. Two individual notable effects were pinpointed at specific locations. In the first localized effect enhanced allocation revealed the consumption to be near-zero whereas the same junction consumed 1.2 L/s with general allocation, leading to 0.3-meter increase in pressure. Second localized effect presented a 26-kilometer trunk main with 1.0-meter pressure decrease due to increased unit head loss with increased flow.

Conclusively, the near-real time hydraulic modelling-based leak localization method used in this study was identified as a potential tool for enhancing WDN water efficiency, but it requires improvements for enhancing the robustness to unfavourable hydraulic conditions and dealing with inaccurate data or model discrepancies. Smart meter data proved modest potential for enhancing hydraulic simulation results via more accurate demand allocation. The usage of measured consumption data as a tool for flagging false positive leak detections is suggested as a future improvement.