Author(s): Chengchao Xu; Kevin S. Tickle
Linked Author(s): Kevin Tickle
Keywords: No Keywords
Abstract: Mathematical models of water distribution networks require many parameters before they can be used to make predictions of system behaviour under various conditions Of several types of parameters, the pipe coefficients are the most difficult to obtain since they are not directly measurable. In practice, the pipe coefficients are usually inferred from the measurement data of their dependent variables such as pipe flows and nodal heads taken at some discrete points within a network in a process of model calibration, or inverse modeling. Since, in general, the number of field measurements available is not enough to exactly identify each individual pipe coefficient within a system, it is common practice to group the pipes to reduce the number of effective parameters to be estimated. This is usually achieved by assigning the same value to the pipes in a group based on prior information on characteristics and conditions of pipes and the experience of network modeler. The grouped parameters are then estimated by applying nonlinear regression analysis. Criterion for such an analysis is usually based on the minimization of a norm of the difference between the calculated values from the model and the measurement data. Several algorithms have been developed for parameter estimation of water distribution networks, e.g, Lansey and Basnet (1991).
Year: 1997