Leakage Detection Using Probabilistic Neural Networks and Model-Based Localization Using Quantum Genetic Algorithms in Real Water Supply Networks

作者：Yihong Guan, Wei Zhang, Mou Lv, Lingzhi Cui, Peng Qiao, Huan Zhao, Hang Li · 发表于：Journal of Pipeline Systems Engineering and Practice · 年份：2025 · DOI：10.1061/jpsea2.pseng-1944 · 研究领域：Water Systems and Optimization、Infrastructure Resilience and Vulnerability Analysis、Risk and Safety Analysis

Companies are making significant efforts to improve leakage detection efficiency. Therefore, this paper proposes the identification of leak zones using a probabilistic neural network (PNN) and model-based localization in real water supply networks. First, a large water supply network was divided into several areas using a fuzzy c-means clustering algorithm optimized with a genetic-simulated annealing algorithm (FCM-GSAA). Then the leak zone was identified using a PNN. Second, the specific leak locations were determined using a quantum genetic algorithm (QGA). The proposed method was then applied to the Qingdao water supply network in Shandong Province, China. Compared with the traditional model-based localization using QGA, which meant that the PNN were not used in advance to narrow the leakage range, the accuracy of the PNN-QGA model was improved by 12.5%, and the average calculation time was ten times faster. Therefore, this method can formulate a leak location plan reasonably and efficiently.