Fault diagnosis of water distribution networks based on state-estimation and hypothesis testing

Abstract

A novel approach for the diagnosis of water distribution networks is proposed. The method combines information about the model of the network and the prior knowledge about the daily patterns of water consumption. In order to infer the presence of faults, typical field measurements characterised by nodal pressures and the flow at the in-let of the district metered areas are used. Then, an implicit state-estimation technique is applied for the estimation of the nodal demands from the available measurements. Fault diagnosis is then achieved based on a signature matrix produced by means of hypothesis testing, which compares the statistics of the estimate with the prior of the demand. Simulation results in realistic conditions show the ability of the method to correctly detect and isolate leaks as well as parametric faults. © 2012 IEEE.