Study Of Differential Evolution Algorithm Based On Target Optimization

Differential evolution algorithm is a kind of intelligent optimization algorithm,intelligent optimization algorithm is inspired by the natural life of a natural phenomenon or process and the establishment of artificial intelligence algorithm.They can implement highly parallel,self-organizing,self-learning and self-adaptive,and the differential evolution algorithm because of its less controlled parameters,easy to learn,easy to realize,has attracted more and more attention of scholars,differential evolution algorithm has become one of the important topics in the field of optimization algorithm,this paper study and analysis the traditional differential evolution algorithm and the other improved differential evolution algorithm,put forward to combine dynamic adaptive parameter setting and randomly choose mutation strategy are studied.Firstly,this paper analysis the performance of differential evolution algorithm and the factors that affect the performance of differential evolution algorithm.According to the existing algorithm in the characteristics of the objective function of optimization is not universal,so the two algorithms in this paper are used to dynamically adjust the parameters and self-adaptive strategies,thus the convergence speed of the algorithm is guaranteed,and also guarantee the quality of finding solutions of optimal algorithm.Secondly,this type of objective function to be optimized is uncertain,and the parameters need to be optimized different values and mutation strategy types of demand is not the same,so before each iteration,using the normal distribution function produce proportional factor F value and Cr value of the crossover rate,at the same time through the rand(0,1)were randomly selected to participate in the mutation strategy type of mutation,and a combination of two selected mode of operation,proposed a single objective optimization differential evolution algorithm,which ensures fast convergence,but also conducive to quickly find the global optimal value.Thirdly,based on the analysis and research of multiobjective optimization problems,the convergence and diversity of Pareto non-dominated solutions are not good.Therefore,normal distribution function is used to generate F and Cr values,and by roulette probability select mutation strategy type,combined with the definition of non-dominatedsolutions in the NSGA-ii algorithm Pareto,and improved congestion,on-demand operation for cutting,proposed the multiobjective optimization differential evolution algorithm,thereby improving the convergence and the diversity of non-dominated solution.Finally,using the simulation tool of MATLAB and Microsoft Visual Studio2010 doing a simulation experiments.Respectively on simulation single objective optimization differential evolution algorithm and multiobjective optimization differential evolution algorithm,in order to verify the validity and superiority of the proposed algorithm comparing the proposed algorithm with the other existing algorithm experiment from the point of convergence and the quality of multiple performance indicators.