Research And Application Of Adaptive Multipopulation Differential Evolution Algorithm

Differential evolution(DE)algorithm has become one of the most popular heuristic algorithms and has been widely used because of its advantages of high efficiency,simple structure,and fewer control parameters.The performance of differential evolution algorithm mainly depends on the design of population structure,parameters(NP,F and Cr)and the use of mutation operator.Reasonable population structure,parameters and mutation operator can improve the performance of DE algorithm.This paper studies the population structure,parameter control,and selection of mutation operators of the differential evolution algorithm,and proposes two adaptive differential evolution algorithms based on multiple populations,and applies them to the power network problem of smart home.The main content and innovations of the article work are as follows:First,a globally optimized multi-population differential evolution algorithm(AMMADE)with adaptive mutation and local search is proposed in order to better coordinate the cooperation among populations and rationally utilize resources.In this algorithm,during the evolution of each generation,the population is divided using the Euclidean distance sorting method,and the sub-population with the best performance can obtain more computing resources in the next generation to coordinate the relationship among the sub-populations.cooperate.In addition,an adaptive local search strategy is adopted for the subpopulation with the best performance to achieve a balanced search.The effectiveness of the strategy proposed by the algorithm is verified by solving the optimization problem in the CEC2014 benchmark problem.Experimental results show that the algorithm can reach or outperform the rela ted algorithms.Second,a multipopulation differential evolution algorithm(EMPADE)with adaptive parameters is proposed.In this algorithm,an adaptive parameter control method based on diversity,fitness value and iterative feedback is designed to control the parameters F and CR of each subpopulation.Aiming at the problem that the diversity of the top-ranked subpopulations given by the Euclidean distance grouping algorithm is small,an improved mutation strategy DE/current-pbad/1 with archive is designed.In addition,an improved population size reduction strategy is designed to adaptively eliminate underperforming individuals to speed up population convergence.The effectiveness of the proposed algorithm strategy is verified by solving the optimization problem in the CEC2014 benchmark problem.Experimental results show that the proposed algorithm can achieve or outperform the related algorithms.Finally,this paper applies the proposed multi-population adaptive differential evolution algorithm to solve the power network problem of smart home.Compared with the classical heuristic algorithm,the experimental results show that the proposed algorithm can reduce the cost of smart grid.