Developing a New Calibration-Based Optimization Method for Leak Detection in Water Distribution Networks

Recently, calibration based methods are developed for leak detection in water distribution networks (WDN). These methods are collected the pressures and flows data of a real network and try to regulate the demands in hydraulic network model to match the model results with observations. Therefore, in addition of hydraulic analysis, an optimization algorithm is needed. In this study, a new population based optimization method is presented. In this method, the network first is zoned to some regions. First generation are guessed for appearance loss and leakage in each zone and answers fitness are calculated. Moreover, a new factor named minor fitness, developing in this research, is computed for each zone in all answers. A small minor fitness in a zone, predicts the correct value of leakage. With combination of the results of best answers, in each generation, a range is calculated for the next generation selection. This range become narrower in consecutive generations and the correct answer is found. The proposed method is investigated on a real network with hypothetical leakages. The results obtained in model are evaluated with the assumed value, as well as, WaterGEMS model results.  The results show the capability of the proposed optimization method in detection of the both location and the value of leakage. Comparing this method with Darwin calibrator show more accuracy, convergence and speed of the proposed method. The fitness in the present study method and WaterGEMS calibration are 0.654 and 0.817, respectively.