Evolutionary Optimization For Water Losses Recognition In Water Supply Networks

A methodology to localise the losses in the water supply networks has been developed, which requires the installation of a number of flowmeters and pressure transducers on the network and the building of a numerical model. The calibration of the model to match the recorded network parameters (pressures and discharges) is done by searching an optimal set of water demands at network nodes. The comparison between the optimal set and the standard one allows the identification of the areas where the leakages are most likely to be present. The optimal set of water demands is identified by the minimisation of an objective function. In the paper, the coupling of this objective function with three evolutionary optimisation methods based on simulated annealing (SA), genetic algorithms (GA) and modified particle swarm optimization (MPSO) have been discussed and tested on a case study. The simulations show SA risks to be trapped in unfeasible zones in its search, while the methods based on GA and MPSO perform very well because in these latter methods, the individuals constituting a population work mainly in groups. Moreover, the solution obtained by GA and MPSO can be further improved by means of a simple hill climbing procedure. Considerations on the possibility of having more than one maximum of the objective function and how they can be detected are presented