Font Size: a A A

Research And Application Of Decomposition Multi-objective Evolutionary Algorithm Based On Refactoring Neighborhood Strategy

Posted on:2021-02-01Degree:MasterType:Thesis
Country:ChinaCandidate:G CaoFull Text:PDF
GTID:2428330614969907Subject:Management Science and Engineering
Abstract/Summary:PDF Full Text Request
Many-objective optimization problems exist widely in scientific research and engineering practice,and have important research value and application prospects.With the increase of the objective dimension,the performance of traditional multi-objective optimization algorithms will decrease sharply.The multi-objective evolutionary algorithm MOEA/D based on decompositionreduces the pressure of non-dominant solution ranking.The algorithm has attracted widespread attention from domestic and foreign experts and scholars due to its excellent search ability and excellent convergence.However,during the algorithm update process,each sub-problem selects parent from the neighborhood for cross-mutation.The neighborhood structure remains unchanged throughout the evolution process,which limits the range of parent selection to a certain extent.In the later period,problems such as population degradation and slower convergence will occur.To solve the above problems,this paper proposes a neighborhood reconstruction strategy based on the MOEA/D algorithm to improve the quality of the parent solution set,increase the speed of population convergence,and effectively alleviating population degradation.Neighborhood reconstruction strategy changes the parent selection method in the original framework,and selects the parents from a new neighborhood set consisting of the child problem neighborhood and the elite solution.Without changing the size of the neighborhood,the quality of the parent solution set is improved,while the diversity of the understanding set is improved,and the population convergence speed is accelerated.The performance of the algorithm is tested on 2 to 10-dimensional DTLZ1-4 series test problems.The experimental results show that the algorithm effectively solves the problem of population degradation,which makes the population approach the Pareto frontier faster,effectively improving the convergence speed of the algorithm and balancing Algorithm convergence and diversity.To solve the multi-objective flexible workshop scheduling problem,a multi-objective mathematical model with minimum completion time,total machine load,workpiece delay,and total energy consumption was established,and the proposed MOEA/D-RNS algorithm was used to solve the problem.It is compared with the classic NSGA-II and other multi-objective evolutionary algorithms for comparative analysis of simulation experiments.The NSGA-II algorithm based on the dominant relationship will have problems of poor convergence and insufficient population selection pressure during the solution process,while the algorithm MOEA/D-RNS effectively improves the premature problems of algorithm,can better maintain population diversity.
Keywords/Search Tags:multi-objective optimization, refactoring neighborhood strategy, elite solution, flexible Job Shop Scheduling
PDF Full Text Request
Related items