ABSTRACT
Monitoring and locating leaks in water supply pipelines are critical to the safety of rural drinking water, which is a highlighted issue in China. To meet this need, an XGBoost-based model was developed and applied to the rural water supply network in Dingyuan, China. It could diagnose water leakage while overcoming the obstacles caused by the limited scale and incompleteness of data. In a comparative case study, the proposed model outperformed the probabilistic neural network models, which require large-scale data, in terms of both F1-score and accuracy, thus demonstrating its capability to accurately locate leakage in rural water supply pipelines.
Acknowledgements
Thanks to Mr Huang, director of a water plant in Anhui province, for providing support for the pipeline leakage simulation experiment.
Disclosure statement
No potential conflict of interest was reported by the authors.