Evolving Effective Hydraulic Model for Municipal Water Systems

Constructing a robust hydraulic network model is vitally important, but a time-consuming task. Over last two decades, several approaches using optimization techniques have been developed for identifying model parameters. Although most of the methods can make the model agree with field observations, few are able to achieve a good level of accuracy in terms of determining the correct model parameters for a water distribution system. The previously developed methods appear to be lacking versatility for users to specify calibration tasks given real data for a real system. This paper proposes a comprehensive framework for evolving a hydraulic network model. Calibration tasks can be specified according to data availability and model application requirements. It allows an engineer to (1) flexibly choose any combination of the model parameters such as pipe roughness, junction demand and link (pipes, valves and pumps) operational status; (2) easily aggregate model parameters to reduce the problem dimension for expeditious calculation and (3) consistently specify boundary conditions and junction demand loadings that are corresponding to field data collection. A model calibration is then defined as an implicit nonlinear optimization problem, which is solved by employing a competent evolutionary algorithm. With this methodology, a modeler can be fully assisted to carry out not only a single parameter optimization run, but also a variety of calibration tasks in a progressive manner according to practical system conditions, thus it is possible to achieve a good model calibration with high level of confidence. The method has been applied to the model of a municipal water system to demonstrate the efficacy and robustness of the evolutionary modeling practices.