Research On Water Distribution System Hydraulic Model Software And Sampling Design For Calibration

With the development of information construction and strict water supply requirement, many utilities realize the importance of a hydraulic model for their water distribution system (WDS). Building an accurate model is a system engineering, which costs both money and time. With the help of well-built model software and effective method for field data collecting for calibration, duplicate effort can be avoided, and the accuracy of model can be improved with low cost. Therefore, this thesis focuses on developing modeling software, and designing sampling strategy for calibration.WDS hydraulic model software is essential to build model. Much commercial model software exists, which is usually expensive, and may not satisfy the need of water utilities. Open source model software, e.g. EPANET, possessing limitation of functions, is another attractive. This thesis proposed to develop modeling software, which combines the graphic functions of GIS and the simulation functions of EPANET. ArcGIS is selected as the GIS platform. EPANET and GIS have their respective data models. The uniform platform is built based on Geodatabase. Some new functions, i.e. reading model data from Geodatabase and storing simulation results in computer memory, are added into the source codes of EPANET. The source codes are then encapsulated into a COM class. The basic functions of modeling software, e.g. displaying flow direction of pipes, creating isobar and analyzing the supply path, are implemented in .NET with ArcGIS Engine. The developed software has been successfully applied in a modeling project.Uncertainty is a main factor resulting in the inaccuracy of WDS hydraulic model. This thesis analyzes the uncertainty arising from four aspects: model structure, model parameters, model prediction and field data. In order to reduce the uncertainty and obtain an accurate model, the model must be calibrated. Calibration includes not only the process of adjusting the roughness coefficient and water demand but the activities from collecting field data to micro calibration. In essence, micro calibration is a nonlinear parameter estimation process. The confidence of unknown parameters estimation depends on the quantity and quality of the field data. Collecting of field data is an experiment. It can be designed to obtain more accurate unknown parameter estimation with low cost. Based on the theory of optimal design, the sampling design for calibration is formulated as a multi-objective optimization problem where the two objectives are cost and uncertainty. The pressure and flow monitoring point can be combined by price coefficient and accuracy weight of field measurements. The method, used to decide whether the existing monitoring points meet the need of calibration, is proposed as well. The multi-objective optimization problem is solved byε-MOEA and cross entropy method. In three case studies, the two algorithms are verified and compared with other four existing algorithms. It is shown that the two algorithms are better than the exiting algorithms. The cross entropy method converges rapidly and is more suitable for small-scale model. Theε-MOEA is more flexible and can be applied to models of different size, while its parameters must be determined based on specific case.