Leakage Diagnosis Framework for Water Distribution Networks using ABC

In this work, we present a novel approach for leakage localization and identification in water distribution networks using Approximate Bayesian Computation(ABC). A reduced pressure graph theory-based model is derived for the water networks. A leakage is considered an unknown parameter of the model which is to be estimated. Consumer demands are considered stochastic in nature and owing to that the model output, which is pressure, is also stochastic. The first moment of the model and the pressure measurements are compared to estimate probable leakages. Further, the probability distribution for the estimated leakages is computed using ABC. Results from an experimental test on a small-scale water network are also presented to demonstrate the approach.