Enhancing Leak Detection Accuracy in Water Distribution Networks Using SVM-CNN-GT Algorithm

Water Distribution Networks (WDNs) experience substantial water losses due to pipeline leaks, leading to economic losses and worsening global water scarcity concerns. This research aims to improve leak detection accuracy and identify leakage locations in WDNs. The prevalence of leaks in WDNs presents a pervasive challenge affecting water utilities economically and environmentally. Traditional leak detection methods are time-consuming, resource-intensive, and prone to inaccuracies and false alarms due to sensor placement randomness. Detecting concealed background leaks presents a particular challenge due to their invisibility to the naked eye. These leaks manifest below the surface, complicating their realtime identification and repair efforts. To address these challenges, this paper proposes the SVM-CNN-GT algorithm, which is an advanced ensemble supervised Machine Learning (ML) algorithm that integrates Support Vector Machines (SVM), Convolutional Neural Network (CNN), and Graph Theory (GT). By combining multiple ML algorithms, the SVM-CNN-GT model considers various factors influencing leak detection and localization, resulting in more precise and reliable assessments of leak presence and location. The algorithm employs automatic feature extraction and heterogeneous dual classifiers, accurately evaluating leaks based on flow rate, pressure, and temperature data. Additionally, a combination probability scheme enhances leak detection efficiency by integrating diverse classifier models with distinct prediction outputs. The SVM-CNN-GT algorithm's performance is evaluated using EPANET and compared to two other algorithms, the proposed SVM-CNN-GT demonstrated remarkable proficiency with the highest average leak detection accuracy of 98%, followed by CNN at 82% and SVM at 78%.