In order to take into account the scarcity of the water resource and the increasing of the population, the management of drinking water networks has to be improved with the use of new tools and actions that allows fighting against wasting water. The monitoring of drinking water networks is based on the use of sensors to locate malfunctions (leaks, quality/contamination events, etc.). Practical implementation has to be carried out by optimizing the placement of the number of sensors and improving the detection and localization of malfunctions. Techniques for the detection and localization of leaks have been proposed in the last years based on the evaluation of residuals obtained by means of the comparison between the measurements obtained by the sensors and the values obtained by simulating the water network in a leak free scenario. In this paper, a data-driven approach based on the use of statistical classifiers working in the residual space is proposed for leak localization. The classifiers are trained using leak data scenarios in all the nodes of the network considering uncertainty in demand distribution, additive noise in sensors and leak magnitude. Finally, the proposed approach is tested using the well-known Hanoi network benchmark.

In order to take into account the scarcity of the water resource and the increasing of the population, the management of drinking water networks has to be improved with the use of new tools and actions that allows fighting against wasting water. The monitoring of drinking water networks is based on the use of sensors to locate malfunctions (leaks, quality/contamination events, etc.). Practical implementation has to be carried out by optimizing the placement of the number of sensors and improving the detection and localization of malfunctions. Techniques for the detection and localization of leaks have been proposed in the last years based on the evaluation of residuals obtained by means of the comparison between the measurements obtained by the sensors and the values obtained by simulating the water network in a leak free scenario. In this paper, a data-driven approach based on the use of statistical classifiers working in the residual space is proposed for leak localization. The classifiers are trained using leak data scenarios in all the nodes of the network considering uncertainty in demand distribution, additive noise in sensors and leak magnitude. Finally, the proposed approach is tested using the well-known Hanoi network benchmark.