WIT Press


Demand Uncertainty In Modelling WDS: Scaling Laws And Scenario Generation

Price

Free (open access)

Paper DOI

10.2495/SDP160621

Volume

210

Pages

12

Page Range

735 - 746

Published

2017

Size

808 kb

Author(s)

R. Magini, F. Capannolo, E. Ridolfi, R. Guercio

Abstract

Water distribution systems (WDS) are critical infrastructures that should be designed to work properly in different conditions. The design and management of WDS should take into account the uncertain nature of some system parameters affecting the overall reliability of these infrastructures. In this context, water demand represents the major source of uncertainty. Thus, uncertain demand should be either modelled as a stochastic process or characterized using statistical tools. In this paper, we extend to the 3rd and 4th order moments the analytical equations (namely scaling laws) expressing the dependency of the statistical moments of demand signals on the sampling time resolution and on the number of served users. Also, we describe how the probability density function (pdf) of the demand signal changes with both the increase of the user’s number and the sampling rate variation. With this aim, synthetic data and real indoor water demand data are used. The scaling laws of the water demand statistics are a powerful tool which allows us to incorporate the demand uncertainty in the optimization models for a sustainable management of WDS. Specifically, in the stochastic/robust optimization, solutions close to the optimum in different working conditions should be considered. Obviously, the results of these optimization models are strongly dependent on the conditions that are taken into consideration (i.e. the scenarios). Among the approaches for the definition of demand scenarios and their probability-weight of occurrence, the moment-matching method is based on matching a set of statistical properties, e.g. moments from the 1st (mean) to the 4th (kurtosis) order.

Keywords

water distribution systems, water demand, scaling laws, scenario generation, stochastic/robust optimization