Water Distribution Network Systems Optimization Based On Auto Adaptive Genetic Algorithm

The expansion or construction of water distribution network systems is an important measure to solve the problem of urban water supply.The key of this measure is to optimize the combination of pipe diameters within a network.Genetic algorithm is a commonly used optimization algorithm for solving optimization design problem in water distribution network systems.However,a common problem of the traditional genetic algorithm is that it does not combine the algorithm with the attributes of the optimization problem,which may reduce the search efficiency when the algorithm is applied to concrete problems.It is difficult to apply it directly to the large and complicated design of water supply pipe networks.Therefore,the purpose of this study is to clarify the shortcomings of optimization algorithms,and propose a new algorithm combining the genetic algorithm with the attributes of optimization problems to guide the search direction of the algorithm and improve the evolutionary efficiency.It is found that the parameters of the optimal design problem of the water supply network have corresponding physical meaning in practical problems.There is a certain correlation between the physical quantities,such as the close relationship between the flow rate and the diameter.Moreover,a fixed range of pipe diameters is adopted in the process of producing new offspring at each generation.Therefore,this study proposes auto adaptive genetic algorithm(AAGA).According to the flow rate and pipe diameter of the pipe section,the method can dynamically adjust the diameter range of the next generation individual in the algorithm,and then improve the evolutionary efficiency.In this thesis,the genetic algorithm and the auto adaptive genetic algorithm based on MATLAB software platform and EPANET hydraulic search engine are validated using small pipe network.The results show that the auto adaptive genetic algorithm and the genetic algorithm can determine the optimal solution of the small pipe network and thus prove the feasibility of the algorithm.The New York tunnel extension network is taken as a small-scale pipe network application case to evaluate the performance of the auto adaptive genetic algorithm and genetic algorithm.It is found that the auto adaptive genetic algorithm uses only 7 generations to obtain the same optimal result,while the genetic algorithm in the literature uses 10000 generations.Compared with the genetic algorithm used in this study,it is found that auto adaptive genetic algorithm can obtain better result with 7 generations.Taking a new pipe network in Yan'an as a large-scale pipe network application case,the optimal performance of the auto adaptive genetic algorithm is tested and evaluated.The results show that compared with the traditional genetic algorithm,the auto adaptive genetic algorithm uses only 2 generations to obtain better optimal result when the control parameters are selected in the low flow velocity range,and some individuals in the population of the method are involved in the adaptive process.Moreover,by optimizing the key parameters of the new algorithm,the optimal value or the optimal range of each parameter can be determined,which can further improve the optimization performance of the algorithm and provide reliable parameter suggestions for the practical application.