A Software Tool for Assessing the Performance of Water Distribution System Solution Methods based on Graph Theory

Abstract

Hydraulic simulation models have been used to simulate the steady-state of a water distribution system (WDS) for serval decades. These models have been used in WDS simulation toolkits and have played a critical role in the design, operation, and management of WDSs in industry and research. In recent years, a number of graph theory based WDS solution methods have been developed. These methods have explored the structural properties (both matrix and graph) of the problem to improve the speed and reliability of WDS simulations. One question that naturally arises is which method or combination of methods should be applied? In this thesis, a WDS simulation testbed, called WDSLib, has been developed as a tool that can be used to answer the above question. WDSLib is an extensible simulation toolkit for the steady-state analysis of a WDS. It has been created using modularised object-oriented design and implemented in C++ programming language. WDSLib can be used (1) to implement, test, and compare different solution methods, (2) to focus the research on the most time-consuming parts of a solution method, (3) to guide the choice of solution method when multiple simulation runs are used (such as occurs in a genetic algorithm run). WDSLib has been used to investigate the performance of four solution methods, namely the global gradient algorithm (GGA), the reformulated co-tree flows method, the GGA with the forest-core partitioning algorithm (FCPA), and the RCTM with the FCPA, on eight case study benchmark networks with between 934 and 19647 pipes and between 848 and 17971 nodes. The results can be used to inform the choice of the solution method for a given combination of the network features under different design settings. This work also demonstrates how to (1) use the WDSLib to implement, test, and benchmark the existing solution methods and (2) use the results to determine which method or combination of methods to used under a setting of interest. A new graph theory algorithm, called the bridge-block partitioning algorithm (BBPA), has been proposed which further partitions the WDS network in a number of bridge components and a number of block components. The BBPA is also implemented in the WDSLib in order to ensure a fair comparison with the existing methods. The BBPA is a pre-processing and post-processing method, the use of which provides significant advantages over the current methods in terms of both the computational speed and the reliability of the solution. This work also demonstrates how to (1) use the WDSLib to implement, test, and benchmark the new solution method and (2) use the WDSLib to demonstrate the efficiency of new method without having to reengineer the content of shared WDSLib functions and data representations.