Evolutionary computation techniques to assess losses in water supply networks

Abstract

In this work, a methodology that combines both simulation and measurements of pressures and discharges on the water supply network is applied to a case study. The demands of the model nodes are systematically changed by means of two evolutionary algorithms and the network is simulated in order to match the readings of the instruments. By comparing the measurements with the simulated values it is possible to assess the losses and estimate their locations. Specifically, two methods have been tested and applied to a case study: the former based on Simulated Annealing (SA) and the latter on Genetic Algorithms (GAs). The simulations show that the methods based on GAs perform much better and are able to detect the different hypothesized scenarios, while the single individual used by SA risks being trapped in an unfeasible zone in its search. Moreover, the solution obtained by GAs can be further improved by means of a simple Hill Climbing procedure (HC), thus achieving satisfactory results. © 2013 WIT Press.