Better understanding of water quality evolution in water distribution networks using data clustering

Abstract

The complexity of water distribution networks raises challenges in managing, monitoring and understanding their behavior. This article proposes a novel methodology applying data clustering to the results of hydraulic simulation to define quality zones, i.e. zones with the same dynamic water origin. The methodology is presented on an existing Water Distribution Network; a large dataset of conductivity measurements measured by 32 probes validates the definition of the quality zones. The results show how quality zones help better understanding the network operation and how they can be used to analyze water quality events. Moreover, a statistical comparison with 158,230 conductivity measurements validates the definition of the quality zones.

Keywords: Data clustering; Drinking water network; Epanet; K-means; Origin of water; Quality zones.