Research And Application On Memetic Algorithm For Multi-objective Optimization Problems

Optimization problems widely exist in various social and economic departments,and the research on this kind of problem has broad application prospects and great scientific research value.How to solve optimization problems better has been paid more and more attention in the field of scientific research.At the same time,a large number of related work has emerged in practice.Evolutionary algorithms(EAs)have good scalability,robustness,and global searchability,which can make up for the shortcomings of traditional mathematical methods.It has been widely used to solve various optimization problems.However,due to the lack of local search,evolutionary algorithms will encounter convergence challenges when solving some more complex problems.Memetic algorithm is a heuristic algorithm proposed by Pablo Moscato,which analogizes meme information evolution to data evolution theory,based on the conventional global search algorithm,it introduces the local search operator,and optimizes the individuals obtained from the global search to enhance the performance of the algorithm.This strategy also makes MA different from other heuristic algorithms.MA is more like a hybrid algorithm framework,which has attracted a lot of researchers' interest after it was proposed.This paper explores the application of memetic algorithm in the field of multi-objective optimization(1)Firstly,in this thesis,we analyze the widely-used hybrid algorithm framework,points out their shortcomings,and proposes a local search system based on self-guidance.The system uses the individual's genes to guide the generation of neighbors,to reduce meaningless neighbors.Then,the simulated annealing operator using the Gaussian probability density function is proposed as a local search operator of the algorithm,and an acceptance strategy based on crowding distance variance is designed.Based on the above two points,in Chapter 3,a self-guided multi-objective cultural gene algorithm with elite development strategy(S-SANSGA-II)is proposed,and its performance is verified by comparative experiments.(2)Aiming at the problem that the local search operator consumes too much computing resources,a new local search operator is proposed.The operator uses the hyperbolic tangent function as the turbulence generator and affects the fluctuation intensity by adjusting the value range.The operator has the advantages of simple and fast,which improves the convergence of the algorithm and ensures the simplicity of the process.At the same time,a resource allocation strategy based on crowding distance Roulette is designed to maintain population diversity.In addition,an intensity intervention mechanism of individual development is designed,which allows the decision-maker to adjust the bias factor to intervene in the intensity of individual development,and emphasizes the influence of the decision maker's preference on the evolution process.Based on the above three points,a fast self-guided multi-objective memetic algorithm with elite exploitation and resource allocation strategy(NSGA-II-Bn F)is proposed in Chapter 4.Finally,the algorithm and nine competitive algorithms are used to solve 36 multi-objective problems with different complexity,and the results show the superiority of our proposed algorithm.