Hydraulic Equation-Based Method for Leak Detection in Water Supply Networks

Leakage within water supply networks (WSNs) poses significant challenges, resulting in substantial financial losses and service degradation. While conventional physical methods for leak detection prevail, contemporary research in computational intelligence has shifted toward localization techniques. Nevertheless, existing studies have yet to explore the ramifications of leaks on the governing hydraulic equations of WSNs. This paper introduces a methodology for detecting leak locations and magnitudes within WSNs. The proposed approach entails formulating and solving the system of governing hydraulic equations across varying time intervals, incorporating the concept of a pseudo-source node and leveraging pressure measurement data. We deploy a dual ensemble of deterministic algorithms, namely, the Levenberg–Marquardt and the interior-point algorithms, to unravel distinct and intricate sets of hydraulic equations. Our methodology was validated through implementation on four WSNs, each subjected to several different leak scenarios. The results highlight the notable accuracy of the proposed method in pinpointing all leak locations, with a Mean Absolute Maximum Error of less than 0.5% for leak magnitudes, confirming its effectiveness in leak detection. Moreover, the performance of the proposed method under input data uncertainty was assessed, demonstrating robust results when nodal demands were perturbed by 5%–10%. However, as WSN size increased, the accuracy decreased, underscoring the need for accurate data in practical applications.