Research On Reliability Of Urban Water Distribution Systems

Water distribution systems (WDS) are important urban infrastructures which are designed for safely conveying potable water from treatment plant to user's tap with adequate quantity and desired quality. The reliability evaluation of WDS has positive and critical meanings for ensuring water supply security, reliability and maintaining a high level of service. A complete review of existed WDS reliability models and publications has been done before starting this study. In order to consider various factors and uncertainty sources that may affect WDS reliability, a set of methods and algorithms from subjects of system engineering, graphic theory and optimization theory has been employed in this dissertation to explore the effective WDS reliability assessment measurements and their implementation on optimal design problems. The outcome of this study could provide a theoretical basis and decision support for optimal design and operation of distribution systems.Firstly, according to WDS hydraulic theory and the relationship function between available flow and pressure, an EPANET based method is proposed to analyze the hydraulic performance of WDS under failure condition. In order to carry out WDS topological analysis when system is under failure condition, a new effective algorithm, based on Depth-first Search technique from graphic theory, has been developed to search the valves to be closed and the impacted area due to valve closing while component failure occurs in the system. The proposed algorithm provides a theoretical foundation for exploring the valve's impact on WDS reliability.Secondly, the impact of valve and adaptive pump operation has been ignored by most previous studies. In order to account all possible factors and uncertainty sources that may affect WDS reliability, a minimum cut-set based method and another Monte Carlo simulation based method have been proposed to assess reliability of distribution systems. These two methods have been implemented to evaluate reliability level of mid-sized distribution networks seperatedly. The results prove that both the number of valves and pump adaptive operation have significant impact on WDS reliability.Thirdly, flow entropy is treated as a measurement of flow uncertainty in networks to represent WDS reliability and an entropy based surrogate model has been established for reliability analysis. Also a resilience index surrogate model has been developed and in which the resilience index is a quantified measure for system's ability to overcome or reduce the impact of component failure. Both of these two surrogate models have been applied to evaluate the performance of a mid-sized network and the results show that both flow entropy and resilience index are correlated with network reliability and the high efficiency of these two surrogate models has been demonstrated as well.Finally, treating diameters and valve locations as decision variables, a multi-objective model for optimal design of distribution systems has been proposed. The two objectives of this model are to minimize the investment and operations cost and maximize WDS reliability. The shuffled frog-leaping algorithm (SFLA) has been extended to solve multi-objective problems with the help of an external Pareto archive strategy and non-dominated ranking technique from NSGA-II. The multi-objective SFLA has been used to find solutions for optimal design of a real network located in northern China; the results have demonstrated the feasibility of the optimal design model and the efficiency of the multi-objective SFLA.