Leak detection in Water Distribution networks with optimal linear regression models

Abstract

This paper proposes a methodology for leak detection and isolation in Water Distribution Systems (WDS). Our work is based on the construction of regression models with optimal parameters which describe the behavior of the network in a normal scenario (no leaks) and comparing these models against predicted output for actual data obtained from pressure measurements along a time horizon. The proposed detection method takes into account possible presence of noise in the demand (assumed known) and/or in the measurements obtained from sensors located in the nodes of the network. A typical water network testbed is employed to validate the proposed methodology. Epanet® software was used to perform the simulations of leak scenarios in order to validate the effectiveness of the proposed approach. Encouraging results are obtained in scenarios with noise and different demand patterns. © Springer-Verlag Berlin Heidelberg 2012.