Water distribution system (WDS) model parameters, as all mathematical models, need calibration, before practical results can be retrieved. In this thesis a multiobjective optimization approach to calibration is described. Firstly different parameter groupings are used to improve calibration stability. A genetic algorithm based optimization algorithm is used and in depth described. Since parameter uncertainty is an important part of calibration process, a standard FOSM method and a GA population statistics based method are used in the process. In addition, the analysis tools for calibration and model fitting validation are described. Calibration algorithm shows promising results when applied to hypothetical and real life calibration scenarios. Uncertainty analysis shows sub optimal, but still very useful results.
|
