Parameter Estimation for Water Distribution Networks

Abstract

Calibration of a water distribution network is a long, tedious task, if analyzed by an engineer, with no guarantee of determining the proper system parameters. In addition, more utilities are moving toward automated control and wish to estimate the state of the network based upon telemetry data, A rigorous, nonlinear programming algorithm, which incorporates a network simulation model, is presented to solve these problems. The model is capable of analyzing one or more independent demand patterns, or extended period simulations, or both. The model assumes the measurements are exact and has an objective of minimizing the sum of the squares or absolute values of the differences between observed and estimated values of pipe flows and nodal pressure heads. The model consistently finds optimal solutions with the objective function equal to zero with exact data. However, the estimated parameters (pipe roughness coefficients, valve settings, and nodal demands) are not always the true values, which points to a need to collect sufficient quantities of high‐quality data.