Uncertainty reduction in water distribution network modelling using system inflow data

Abstract

In water distribution network (WDN) modelling, nodal demand is the sum of flows taken by users associated with a computational node. User demands are not fixed in time; rather they are stochastic. Hence, nodal demand is a model parameter with high uncertainty, which is propagated throughout the WDN model, thus also rendering the output values (node pressures and pipe discharges) uncertain. Total water inflow into the network can be accurately measured using flow meters. This paper investigates how knowledge of system inflow can be used as a constraint in WDN modelling, taking into consideration the uncertain nodal demands, and consequently reducing the uncertainty of the model output. Fuzzy sets were used to represent the uncertain demands and modified genetic algorithms were used to find the optimal solutions. As a test case, a set of data from a real WDN was used. The uncertainty of the WDN model output was computed for two cases: first, with the total network inflow taken into consideration; and second, with the inflow used as a constraint. Although the methodology that handles the constraints needs significantly greater computational effort, its results provide a more realistic insight into model uncertainty. The proposed methodology was verified using Monte Carlo simulation.