Burst Detection for District Metering Areas Based on Soft-Voting Ensemble Model in Water Distribution System

Abstract

Water loss through leaking pipes is a major challenge for water supply enterprises, which is driving the development of efficient algorithms for detecting pipeline network leakage. Aiming at the low classification accuracy of single classifier in traditional machine learning, this paper proposes a burst detection model based on soft voting. Firstly, the variational mode decomposition (VMD) method is used to denoise the pressure data and extract the important features. Then, an Ensemble learning algorithm based on soft voting is proposed to detect the burst events, which is a machine learning method that uses a series of learners and uses some rules to integrate the learning results to achieve a better learning effect than a single learner, and improve the generalization ability of the algorithm. The method was validated on a case study involving a DMA (district metering area) with pipe-burst events which can effectively detect bursts with a low false-alarm rate and high accuracy.