Creating a leakage detection model for water distribution networks by using heuristic and statistical methods

A significant part of the leaks that occur in drinking water distribution systems do not come to the surface due to various factors. As the time to become aware of these leaks increases, the volume of water lost also increases. For this reason, these leaks that occur in the system and do not come to the surface should be determined in a sustainable way and the system should be monitored. In this study, a unique leak detection model and methodology has been developed based on the optimization algorithm in order to detect non-surface leaks in drinking water distribution systems. In the study, a solution to the leak detection problem was proposed with an optimization algorithm using the differences between the pressure values calculated on the hydraulic model and measured in the field. The theory on which this study is based is about to determine the location of the leak by using the data of pressure differences. The resources required by the proposed methodology are a well defined water distribution network, a calibrated hydraulic model, inlet flow measurements on the water network, pressure measurements over the distribution network, hydraulic analysis software and optimization algorithm software. In this study, EPANET software was used for hydraulic analysis and MATLAB software was used for the optimization algorithm, and the EPANET-MATLAB toolkit used in the study was a useful tool that shortened the solution times and also reduced the workload. The developed model was first tested on an experimental network. Later, proposed methodology tested on Hanoi network which is widely referenced in the literature. According to the results of the analysis, the algorithm shows that in the experimental network with 25 nodes, it gives 100% correct results for 12 nodes and can predict the correct leakage point each time. The cumulative correct prediction rate for the entire network is 92%. In the Hanoi network which has 31 nodes in total; for 21 nodes (67.7%) the algorithm marked correct leakage node. The correct result for 6 of the remaining 11 nodes was determined by the algorithm with a rate of over 50%. For a total of 27 (87.1%) nodes, the rate of finding exact results in searches was over 50%. Keywords: Water network, Leakage detection, Optimization, Hanoi network, Hydraulic modeling, Calibration