Research On Optimal Design Of Urban Storm Drainage Network Based On Multi-objective Evolutionary Algorithm

With the advancement of urbanization and the increase of extreme rainfall events,many cities in China have suffered from waterlogging disasters to varying degrees.Urban storm drainage network is an important part of urban flood control and drainage system,and its main function is to collect,transport and treat rainwater.Reasonable design of urban storm drainage network can effectively avoid urban waterlogging and ensure the safety of urban drainage.With the development of computer technology and the enrichment of multi-objective optimization theory,more and more scholars apply multi-objective evolutionary algorithm to the optimal design of urban storm drainage network.In this paper,the optimal design of urban storm drainage network based on multi-objective evolutionary algorithm is carried out,aiming at obtaining a set of economical and effective drainage network design schemes.The main research contents are as follows:Firstly,in order to solve the problem that the selection pressure of evolutionary algorithm is reduced and the diversity is difficult to maintain due to the large number of objectives in the optimization model of urban storm drainage network,a multi-objective evolutionary algorithm(LRDEA)based on LRD dominance is proposed.In environment selection,LRD dominance is proposed as the first criterion of environment selection,which improves the selection pressure of the algorithm.Taking the angular crowding distance as the second criterion of environment selection effectively maintains the diversity of the algorithm.In the process of individual generation,using the clustering idea in brainstorming optimization algorithm for reference,new individuals are generated in two ways:within-class and between-class,which effectively balances the exploration and development in the search process.The experimental results show that LRDEA has good comprehensive performance and can output a set of high-quality non-dominated solutions.Finally,two multi-attribute decision-making methods are used to help decision makers choose the appropriate solution in the non-dominant solution set.Secondly,considering the hydraulic dynamics of urban storm drainage network,the multiobjective evolutionary algorithm is combined with SWMM simulation model in the design of urban storm drainage network,and the multi-objective evolutionary algorithm is used for optimization and the SWMM simulation model is used for individual evaluation.Aiming at the challenge of complicated constraints in the optimization model of urban storm drainage network,a tree-templatebased multi-objective evolutionary algorithm(TBEA)is proposed.In order to deal with the constraints among decision variables in the model,a tree template encoding schema and the population initialization method based on the tree’s preorder-first traversal are proposed to ensure that the solutions in the initial population meet the constraints among decision variables.The crossover and mutation operators are customized,which effectively improves the proportion of solutions that meet the constraints among decision variables in the offspring population.In order to deal with the constraints in decision space in the model,the defect that constraint dominance is easy to fall into local optimum is improved,and local constraint dominance is proposed,which effectively balances the convergence and diversity of the algorithm.The experimental results show that TBEA can effectively deal with the complex constraints in the problem and output a set of economical and effective pipe network design schemes.Finally,aiming at the shortcoming that the calculation of multi-objective evolutionary optimization of urban storm drainage network is time-consuming,a surrogate-assisted multiobjective evolutionary algorithm(S-TBEA)is proposed.In S-TBEA,RBF network is used as a surrogate model for approximately time-consuming SWMM simulation.In order to ensure the accuracy of RBF network approximation,in the initial stage,an initial population with uniform distribution is generated based on the idea of stratification;In the process of evolution,new samples are selected according to the model management strategy to update the RBF network in real time.The coding mode,individual generation and environment selection of S-TBEA are the same as those of TBEA.The experimental results show that S-TBEA can output a set of high-quality nondominated solutions in a relatively short time.Finally,an intuitive solution selection method is proposed to help decision makers choose the most suitable solution in the non-dominant solution set.