Performance of partitioned water distribution networks under spatial-temporal variability of water demand

Water network partitioning (WNP) into District Meter Areas allows improving pressure management, water budget, leaks detection, compared to classical redundantly looped networks. Anyway, WNP can significantly worsen the hydraulic performance, because often many gate valves are required, and the reduced set of possible paths for water flow makes the network vulnerable, especially to concentrated peaks of water request. Hence, reliably modeling the spatial and temporal distribution of water demand would allow better understanding the effects of WNP, improving both network design and management. Aiming at taking into account the behavior of a partitioned network under spatial and temporal variability of water demand, a design procedure is proposed, coupling a stochastic model of water demand with novel partitioning techniques, based on spectral clustering. The application of the procedure to a real medium-sized network shows that demand variability may significantly affect the hydraulic performance of the partitioned network, evaluated by means of topological and hydraulic indices.