Font Size: a A A

Research And Application Of Constrained Differential Evolution Algorithm Based On Triangular Mutation

Posted on:2020-10-28Degree:MasterType:Thesis
Country:ChinaCandidate:C Y SuFull Text:PDF
GTID:2428330623959563Subject:Statistics
Abstract/Summary:PDF Full Text Request
Constrained optimization problem is a common problem in engineering and practice.Traditional constrained optimization methods have strict requirements on the functional properties of the problem.The differential evolution algorithm has better global optimization ability,and does not rely on the analytic properties of functions,so it's widely studied.In order to study the constrained optimization problems,this paper analyzes a variety of constraint handling mechanisms and studies the basic operators of the differential evolution algorithm.The global search ability of mutation operator in differential evolution algorithm is too strong,the parameter setting method also has shortcomings,and the diversity of population becomes worse at the end of the algorithm.Triangular mutation operator is one of differential evolution algorithm mutation operators,which has good local search ability.In view of the shortcomings of single constraint handling method and general differential evolution algorithm,this paper combines the differential mutation operator with the triangular mutation operator,and proposes several improvement schemes.The main work of this paper is as follows:1.Aiming at the deficiency of local search ability of differential evolution algorithm,this paper proposes an improved adaptive constrained differential evolution algorithm based on parameter adaptive control mechanism by introducing triangle mutation operator.Firstly,the algorithm uses the general differential evolution mutation operator for global search,and the triangular mutation operator for local search.Secondly,in order to improve the problem of slow convergence caused by triangular mutation operator,a new trial individual substitution strategy was introduced.Then in order to prevent the lack of diversity in the later stage of the algorithm,a diversity maintenance mechanism is designed.Finally,the improved adaptive differential evolution algorithm was tested on the two kinds of test sets of constraint optimization problems,and it was found that the improved differential evolution algorithm showed higher solution accuracy and faster convergence rate.2.To solve the problem of strong local search ability of triangle mutation operatorr,this paper further proposes a novel hybrid mutation differential evolution algorithm by improving the triangular mutation operator.Firstly,the hybrid mutation differential evolution algorithm improves the ability of solving complex constrained optimization problems by using parameter pool strategy.Secondly,the adaptive penalty function method is used to guide the calculation of the base vector.Then,the evolutionary strategy selection method is introduced to accelerate the convergence of the algorithm,and the algorithm implements the diversity maintenance mechanism for the worst individual.Finally,the algorithm is tested on 20 test functions,and three statistical analysis shows that the hybrid mutation differential evolution algorithm is effective in solving these problems,and the algorithm performance reaches some of current "state-of-the-art" algorithms.3.In view of the insufficiency of the interaction between mutation operators with different abilities caused by the use of multiple mutation strategies in single population,this paper proposes a differential evolution algorithm based on dual populations mutation to improve this shortcoming.Firstly,each subpopulation of the dual populations differential evolution algorithm uses the ordinary differential evolution mutation operator and triangular mutation operator respectively.Secondly,the two subpopulations use different selection methods respectively.Then,in order to exchange information between two subpopulations,this paper introduces a migration strategy of secondary individual exchange.Finally,in order to verify the algorithm effectiveness,this paper applies the improved differential evolution algorithm to particle filtering and target tracking.The simulation results of three kinds of target tracking problems show that the improved differential evolution particle filter algorithm has better resampling effect than sequential importance sampling.
Keywords/Search Tags:Differential evolution algorithm, Constrained optimization, Triangular mutation, Engineering optimization, Particle filter
PDF Full Text Request
Related items