Optimal valves regulation for calibration of pipe networks models

Abstract

The calibration of a pipe network model is faced by means of a two-steps iterative procedure. The first step of the procedure is the estimation of the parameters in the network simulation model, obtained from a first set of water head measures. The second step of the procedure is the change of the resistances provided by the valves in the network and the acquisition of a new set of measures. The inverse problem is then solved again and the procedure is repeated, until the set of computed parameters converge to stable values. The new resistances are chosen to minimize the sum of the variation coefficients, computed from the Fisher information matrix. The effect of different correlation functions between measures in the same location and different valve closures is investigated with the aid of an elementary example. An application to a larger example follows.