Theoretical identification of leakage areas in virtual district metered areas of water distribution networks using the artificial neural network

One of the advantages of designing water distribution networks (WDNs) as a district metered areas (DMAs) is to identify the leakage in each area by controlling the input and output flow, which of course requires the separating areas and installation of flowmeters between the interconnect pipes of areas. Considering that the most existing WDNs have been expanded traditionally and not as DMA, turning them into DMAs would require huge costs and might not be even practical in some networks. In this paper, a theoretical idea of virtual DMA is presented to identify the leakage in each areas. The innovation of this paper is the ability to transform networks into DMAs using a combination of the graph theory and artificial neural network to find leaks without using a flowmeter. The proposed method, in addition to reducing costs for the flowmeters, also increases the speed of detection of leakage areas. In addition, there is no need to specify the number of leakage nodes before the leak operation begins. The proposed method has been applied for the Balerma WDN in Spain with 443 nodes and 454 pipes for two, three and four simultaneous leaks. The results of this paper show that the proposed theory in this method is able to detect leakage in each area, and this method can determine the number of optimal virtual DMA for each network. In all examples, the leakage area was correctly predicted and the maximum leakage error was about 6.5%.