Leak Modeling in Water Supply Networks Using WaterGEMS Model and Artificial Neural Network

Loss of more than 30% of the incoming water to the country's drinking water distribution networks due to leakage from the network, has caused serious concern to the officials of the country's water and sewage companies and finding the location of leakage in distribution networks is one of the important issues and concerns of users and related organizations. Reducing the amount of leakage in water supply networks is one of the main methods of managing the water distribution network in different countries. Currently, several methods have been proposed to detect leaks in water supply networks. In this paper, hydraulic modeling of a real network flow by WaterGEMS hydraulic software and inverse solution of flow equations, having measured values of pressure in a number of network nodes, location prediction and leakage rate in the network Water distribution works were carried out in Mohiabad city located in Kerman province. First, the hydraulic model of the studied network was prepared and calibrated in the hydraulic analysis software and the amount of existing leaks was collected, and then by analyzing the network for different states and number of hypothetical leaks, the pressure values in different network nodes were calculated. In the second stage, using artificial neural networks, after network training, by presenting the measured pressures in some network nodes as input data to the neural network, the position and amount of possible leaks were predicted. Investigation and comparison of the results of hydraulic analysis of the network and artificial neural network showed a very high accuracy of artificial neural networks in estimating the amount and position of leaks.