The leak localisation methodologies based on data and models are affected by both uncertainties in the model and in the measurements. This uncertainty should be quantified so that its effect on the localisation methods performance can be estimated. In this paper, a model-based leak localisation methodology is applied to a real District Metered Area using synthetic data. In the generation process of the data, uncertainty in demands is taken into account. This uncertainty was estimated so that it can justify the uncertainty observed in the real measurements. The leak localisation methodology consists, first, in generating the set of possible measurements, obtained by Monte Carlo Simulation under a certain leak assumption and considering uncertainty, and second, in falsifying sets of nodes using the correlation with a leak residual model in order to signal a set of possible leaky nodes. The assessment is done by means of generating the confusion matrix with a Monte Carlo approach.

Peer Reviewed