Research Of Multi-objective Optimization Based On Combinational Algorithm

Many multi-objective optimization problems is used in the field of scientific researchand engineering practice. For example the transportation of cities,the layouts of buildings,thedistribution of energy resources,the operation of financial capital and so on. Thus,the researchof multi-objective optimization algorithms is very important. Because of the conflict betweendifferent objectives, the solution of multi-objective optimization problems is not the only.Asolution that makes all the objectives the best can’t be found.So solving multi-objectiveoptimization problems is not a easy thing. Although scholars have already made a lot ofmulti-objective optimization algorithms, a single algorithms can not meet needs ofmulti-objective optimization research. Estimation of distribution algorithm、particle swarmoptimization algorithm and differential evolution algorithm is mixed in the proposedalgorithm,and two combinational multi-objective optimization algorithms are raised.To deal with the phenomenon of particle swarm optimization algorithm being oftentrapped in local optima for multi-objective optimization problems,a multi-objectiveoptimization algorithm composed of particle swarm optimization and differential evolution isproposed. Both particle swarm optimization algorithm and differential evolution algorithm areused to create new particles.A controlling factor is used to control the proportion of the use oftwo algorithms.The velocity updating formula of particle swarm optimization algorithm ischanged to improve the search efficiency.Six test functions are used to evaluate theperformance of the proposed algorithm,and the proposed algorithm is compared withNSGA-II,σ-MOPSO,NSPSO and MOPSO.The experimental results show that the Pareto setsobtained by the proposed algorithm have good convergence and diversity performance, andthe proposed algorithm is stable and fast.By combining estimation of distribution algorithm and differential evolution algorithm,amulti-objective optimization algorithm composed of estimation of distribution and differentialevolution is proposed. Particles of population are manufactured by estimation of distributionalgorithm and differential evolution algorithm.A selective factor is used to select thegeneration method of each particle. The proportion of the use of two algorithms is changing,ithas a relationship with the search period.Early in the search,estimation of distributionalgorithm is used to quickly locate.Using differential evolution algorithm to precisely search.The variation factor of differential evolution algorithm has been changed.The newvariation factor is changing.The range of variation of differential evolution algorithm indifferent search periods is under control of the new variation factor.Six test functions are usedto test the proposed algorithm.The results of the test of the proposed algorithm is comparedwith NSGA-II and RM-MEDA.The results of the comparisons show that the Pareto optimalsolution set of the proposed algorithm has good convergence and diversity performance,andthe proposed algorithm is effective and stable.