Model-free sensor placement for water distribution networks using genetic algorithms and clustering

This paper presents a model-free methodology for the placement of pressure sensors in water distribution networks (WDNs) with the aim of performing leak detection/localization tasks. The approach is based on a custom genetic algorithm (GA) optimization scheme, which considers a population whose individuals are binary vectors encoding the network nodes with/without sensors. The optimization process pursues the minimization of a distance-based metric, computed considering the pipe distance from the possible sensors to the complete set of nodes of the network, hence removing the necessity of a hydraulic model of the WDN. The methodology is completed by means of an iterative clustering technique that seeks the enhancement of incoming individuals. The proposed methodology is tested over a well-known case study, L-TOWN from the BattLeDIM2020 challenge, in order to assess its performance.
Peer Reviewed
Postprint (published version)