A Research Of Multi-Objective Differential Evolution Algorithm

Most of optimal decision issues in engineering design and scientific research can be summarized as multi-objective optimization problems.These multiple objectives are often interrelated and mutually conditioned,which means any improvement in one objective lead to degradation of at least one objectives.So it is difficult to achieve an optimum for all objectives simultaneously.Therefore,research on the multi-objective optimization problem becomes a challenging and significant topic in evolutionary computation community.As one of the most effective evolution algorithms in recent years,differential evolution is very promising to solve the multi-objective optimization problems.At present,the research of multi-objective differential evolutionary(MODE)algorithm is in the development stage.Therefore,this dissertation mainly aims to study and solve multi-objective differential evolutionary algorithm.Main contributions and results of this dissertation are as follows:The research background and significance of multi-objective optimization problem are introduced firstly.Then,the mathematical model and some relative core concepts of multi-objective optimization are presented.Furthermore,the research status and development of multi-objective differential evolutionary algorithm are reviewed and summarized.Finally,the inadequacies of multi-objective differential evolutionary algorithm are discussed and the difficult problems are also provided.In the view of slow convergence and poor uniformity of differential algorithm in solving multi-objective optimization problems,multi-objective differential evolution algorithm with multi-strategy and ranking-based mutation is proposed.Through analyzing the different differential evolution strategies,a mutation operator of multi-strategy differential evolution based on ranking-based mutation,which can make full use of all the advantages of each mutation strategy and improve the exploitation and exploration of the algorithm.In order to preserve the diversity of the Pareto optimality,a crowing entropy-based divert measure is introduced.The experimental results prove that,compared with some recently proposed multi-objective evolutionary algorithm,the proposed algorithm has better convergence and diversity.In order to further improve the convergence and uniformity of Pareto optimal set,an improved multi-objective differential evolution algorithm with archive and spherical pruning is proposed,which is based on algorithm MODE-MSRM.To improve the convergence and robustness of the algorithm,a control parameter self-adjusted strategy is introduced.Moreover,an external archive is introduced to store all the non-dominated solution in the evolutionary process.Additionally,the spherical pruning mechanism is used to replace the crowing entropy-based divert measure and enhance the distribution and diversity of the solutions in the external archive.Experimental results show that the proposed algorithm outperforms some recently proposed improved multi-objective evolutionary algorithms.In the process of solving MOPs,a new approach which integrates preferences into evolutionary optimization is proposed to find the Pareto optimal solutions in regions interested by Decision Makers,named preference Multi-objective differential evolution algorithm based on global physical programing.On the basis of the previous algorithm MODE-ASP,global physical programing scheme is used to expresses the decision maker's preference in a more concise and effective language,which leads the population to evolve in t regions interested by Decision Makers and obtain a set of more satisfactory solutions.The experimental results indicate the effectiveness of the proposed algorithm.