Stochasticity of Demand and Probabilistic Performance of Water Networks

Accurate prediction and modeling of water demand is crucial to understanding the long-term performance of systems, as it is to mounting an effective design and operational planning effort. The paper investigates the influence of cross correlation and autocorrelation in demand on the probabilistic, hydraulic performance of water networks, as measured with the mean and variance of nodal pressures. A stochastic demand model that accounts for lag-1 autocorrelation and lag-0 cross correlation between demands is applied to generate synthetic series of correlated demands. A Monte Carlo Simulation is coupled with EPANET2 to generate time series of pressures and update the mean and variance of nodal pressures. Preliminary results indicate that enforcing a strong lag-0 cross correlation in demand decreases the mean of pressures and increases the variance of pressures. This indicates that the frequency of low-pressure, hydraulic failures is contingent on the level of correlation measured or assumed in a reliability study. Enforcing a strong lag-1 autocorrelation memory at system nodes produces little or no changes in the mean and variance of nodal pressures, but it is found to govern the period of time a pressure signal can persist below a minimum-pressure constraints and remain in a hydraulic failure state.