Modelling uncertainty for leak localization in Water Networks

The performance and success of model-based leak localization methods applied to water distribution networks (WDN) highly depends on the uncertainty of the system considered. This work proposes an original method of modeling the effect of uncertainties in these networks. The proposed method is based on the collection of real data in the water network in the absence of leaks. The discrepancy (residual) between the measured data and the one provided by a simulator of the network in normal operation is used to extrapolate the possible residuals in the different leak scenarios. In addition, indicators for assessing the effect of uncertainty in the performance of leak localization methods based on residual correlation analysis are provided. The error in terms of correlation intervals and leak localzation assessment between the proposed approximation and the real one is studied by means a simplified model of the WDN of Hanoi (Vietnam).Peer ReviewedPostprint (published version