A novel approach based on graph signal processing and sampling theory to set pressure sensors in water distribution networks

Optimizing the arrangement of pressure sensors in water distribution networks (WDNs) is crucial for system efficiency, enabling accurate monitoring and effective control, especially in detecting and mitigating leaks. This study addresses this issue through a joint application of graph signal processing theory and the spectral clustering algorithm. Two WDNs are modeled as graphs, and the spectral clustering algorithm is used to identify leak-sensitive regions based on the properties of the vertices. Then, three metrics from signal sampling on graphs are employed to select the vertices as possible locations for installing the sensors. A new metric, which considers the coverage rate of installed sensors, is introduced to evaluate the performance of the pressure monitoring system, in addition to helping to determine the optimal number of sensors to be deployed. The proposed approach demonstrates an increase of up to 20% in the coverage rate compared to existing methods. Detailed leak simulations show coverage rates of 84% for the Modena network and 92% for the L-Town network with optimal sensor placement. This methodology provides a more efficient and cost-effective solution for WDN managers.