Hydraulic Model Calibration For Water Distribution System

A number of mathematical models are used nowadays to describe behaviour of the real-life water distribution system (WDS). It is a well known fact that, to have any meaningful use, any WDS mathematical model must be calibrated first. Here, calibration is defined as process in which a number of WDS model parameters are adjusted until the model mimics behavior of the real WDS as closely as possible. In this thesis, WDS mathematical models that are used to model water quantity aspect only are analyzed. Steady-state flow model and quasi-steady flow model are considered. The calibration process are divided to two steps including sample design and roughness calibration. Furthermore, the fire test location is optimizated because, in the fire conditions, the better data for calibration can be collected.To conquer the difficulty in implement multi-location flow test and improve poor result of single-location flow test optimization, multi-location optimization is used in computation but the result is used to implement single-location flow test.Location optimization of fire test and measurement are both two objectives optimization problems, so it is appropriate to be resolved by the muli-objective genetic algorithm NSGA-11. The number of flow test and measurement can be determined from the obtained pareto front solution and the results that more number of testing must not lead to better nonneglect flow in WDS can be obtained. The adopted algorithm NSGA-II is very befitting for resolving such two-objective optimization problem.Through comparing the optimization result with sensitivity about nodal demand and roughness, the coincidence relation between them can be found. When the number of flow test and measurement is fixed, the method of utilizing the relation is adopted during the optimization. The ant colony algorithm have quicker convergence due to making use of the explore information that play the same important role in optimization with sensitivity, so the algorithm is chosen from the variant algorithms. Herein, location optimization model about fire test and measurement base on the Max-Min ant colony algorithm is established, which will have better convergence through utilizing the sensitivity.During the calibration, prior information play the important role in improving uniqueness and stability of the calibration problem solution. In this paper, prior information is substituted by the empirical roughness values that are estimated using all information available. Explore information in ant colony algorithms is adjusted to corresponding with empirical estimation. The Max-Min ant colony algorithms is adopted for solving the optimization problem and better solution can be obtained.Finally, all presented novel methodologies are verified successfully on two case studies that involve both artificial and real-life WDS. NSGA-II, the general genetic algorithms and the Max-Min ant colony algorithms is compared in artificial WDS. Furthermore, the sample design and calibration in real-life WDS are realized. The results show that the proposed models and methods have preferable accuracy.