Sensor Placement and Leakage Isolation with Differential Evolution

Abstract

Leakages in water distribution systems (WDS) can cause significant economic losses. Therefore finding leaks in water pipes is nowadays a crucial task for water utilities. In this paper, the effects of leakages on the hydraulics of WDS, particularly with regard to flow and pressure, are investigated. From the discrepancy of the unperturbed and the perturbed WDS due to the occurrence of leakage a methodology is developed, which enables an efficient sensor placement of flow meters and pressure sensors. This is achieved by a Fault Sensitivity Matrix (FSM). A special Genetic Algorithm (GA) called Differential Evolution (DE), which saves significantly on computation time in large WDS, is used to find the optimal position of a minimum number of sensors. DE is chosen because of its good rate of convergence.Once an optimal sensor placement is obtained, DE is also used for leakage location. The methodology was tested in two different WDS. The first WDS is the model network Poulakis (2003) used. The second is a partial network in the city of Linz, where the task is to find an ideal sensor placement to guarantee efficient leakage isolation. In this paper we show that DE performs excellent on both tasks, the sensor placement and the leakage isolation, for the two investigated systems. Additionally the implementation of demand and measurement uncertainties is outlined.