The performance and success of model-based leak localization methods applied to water networks highly depends on the uncertainty of the system considered. One of the main sources of uncertainty in these networks are the demands in the network nodes. This work developes a methodology to evaluate the ability of an existing algorithm to locate leaks in the presence of bounded uncertainty in the node demands. Zonotopic sets are found to be specially suited to bound nodal demand uncertainty. Two consistent approaches to evaluate the effect of demand uncertainty on the decision result are shown: a Montecarlo approach and a Montecarlo combined with zonotope analysis approach. The methodology is presented using a simplified version of an existing leak location algorithm applied to an illustrative network to stress comprehension.

Peer Reviewed