Research On Adaptive Differential Evolution Algorithm Based On Population Information Feedback

Differential evolution algorithm is a simple and effective population-based evolutionary algorithm,which has been widely used in various optimization problems.However,different optimization problems have different function characteristics,and the performance requirements for the solution algorithm are also different.If a fixed evolution strategy and parameters are used,it is easy to cause problems such as the algorithm getting stuck in local optimal solutions and slow convergence.Therefore,designing a suitable adaptive mutation strategy selection and parameter setting scheme has become particularly important.This paper focused on the design of corresponding adaptive methods based on population feedback from the perspectives of mutation strategy transformation and population size adjustment,in order to optimize the performance of differential evolution algorithm.The main research work of this paper is as follows:(1)Research on mutation strategy transformation mechanism based on population distribution information.Mutation strategy is an important component of differential evolution algorithm,and adopting appropriate mutation strategies at different stages of algorithm iteration can help improve the algorithm’s performance.Therefore,this paper proposed a differential evolution algorithm based on quartile strategy transformation,referred to as qSTDE.The proposed algorithm uses feedback information from the best individual and the entire population distribution to determine the location of mutation strategy transformation,thus balancing the algorithm’s exploration and exploitation capabilities.In qSTDE,first,four individuals with different performance are extracted based on quartile,and the distance ratio between these individuals and the best individual in the search space is calculated.Then,the obtained ratio is compared with a given threshold to determine the location of mutation strategy transformation,thereby dividing the evolutionary process of the algorithm into two stages.Finally,different mutation strategy pools are allocated to the algorithm based on the characteristics of different stages.To verify the effectiveness of the mutation strategy transformation mechanism based on population distribution information,the proposed qSTDE algorithm is compared with other advanced evolutionary algorithms on the CEC2005 and CEC2014 benchmark test sets.The numerical experimental results show that the qSTDE algorithm is an effective algorithm.(2)Research on population size adjustment strategy based on population evolution success rate.Population size is one of the main parameters of differential evolution algorithm and has an extremely important impact on algorithm performance.In order to adaptively adjust the population size to optimize the evolution efficiency of the algorithm,this paper proposed an adaptive population size adjustment strategy based on the success rate of population evolution(SRPS).Under the trend of linearly decreasing population size,the strategy designs an increase factor based on the success rate of algorithm iteration to adjust the population size and improve the diversity of the population.The increase factor evaluates the search prospect of the current population by calculating the success evolution rate of the current population,and adapts the population size of the algorithm according to different search prospects based on the sine function transformation.In addition,the SRPS strategy also designs an unequal probability selection method based on individual performance.When evaluating the increase of population size,the SRPS strategy selects individuals from the experimental group that failed to evolve and adds them to the current population through this selection method.This paper combined the SRPS strategy with other differential evolution algorithms and conducted numerical experiments and comparisons on the CEC2017 and CEC2011 test sets.The numerical experimental results show that the SRPS strategy is an effective population size adjustment strategy,which can improve the performance of differential evolution algorithm.