Water distribution network leak localization with histogram-based gradient boosting

Abstract Accurate and rapid leak localization in water distribution networks is extremely important as it prevents further loss of reduces scarcity. A framework for identifying relevant event parameters such the location, leakage area, start time presented this paper. Firstly, proposed data-driven methodology consists acquiring pressure data at nodes network through hydraulic simulations by randomly changing initial conditions (leak time). Pressure uncertainties are added to sensor measurements order make problem more realistic. Secondly, acquired then used train, test, validate a machine learning model predict parameters. The random forest histogram-based gradient boosting algorithms investigated compared detection problem. approach with algorithm shows high accuracy predicting true location.