Research On Mining And Utilization Mechanisms Of Population Information For Differential Evolution

As an important member of Computational Intelligence,differential evolution has been widely concerned by researchers and has made great progress in research and application because of its simple structure,fast convergence and strong robustness.As the core operation of differential evolution,the mutation usually generates the difference vectors in a random way,which slows down the convergence of the algorithm and does not guarantee the quality of the population.Differential evolution produces a lot of population information during the process of evolution.However,the population information is not fully utilized and there is no effective learning and communication mechanism between individuals in the traditional differential evolution.The most of the neighborhood-based differential evolution uses fixed neighborhood mechanism.Although it improves the local search ability of the algorithm to a certain extent,it also limits the global search ability.It leads to the stagnation of the algorithm in the later period.There is a lack of effective individual evaluation mechanism and accurate numerical guidance in the existing algorithm.To solve these problems,there is a great research space and value to design an effective mechanism for population information mining and utilization.Based on the above analysis,a population evolution information model which includes neighborhood information,direction information and history information is constructed according to the generation method of population information.The population evolution information model is used to make full use of population information from all aspects.In this paper,four mechanisms of population information mining and utilization are proposed to enhance the ability of utilizing population information and improve the performance of differential evolution in solving complex optimization problems.The main work of this paper can be summarized as follows:(1)In order to utilize the different roles of different individuals during the search process,this paper introduces a dynamic neighborhood mechanism that depends on individual information.Individuals dynamically determine the neighborhood according to the population information and continuously adjusts it in different evolution stages.(2)To solve the problem of population quality in one-to-one individual update mechanism,this paper proposes an individual replacement mechanism.In order to improve the quality of the population,the worse individuals in the population are replaced by the eliminated offspring of better individuals.(3)In this paper,a calculation method of individual quality is proposed,which can fully reflect the quality of individuals from two aspects: individual fitness value and individual position in the population.It provide accurate reference values for constructing neighborhood and population structure optimization.(4)In order to solve the problem of slow convergence due to high randomness in the evolutionary process,cosine similarity is used as the calculation method of individual similarity,and the direction information are combined with the generation method of difference vector.The historical information is used to construct the path of historical evolution,and the experience of individual evolution is used to guide individuals to a certain direction of stable evolution.In conclusion,population information guides individuals to construct neighborhood and generate mutation,so as to make the evolution process more efficient and orderly.In this paper,a large number of experiments are used to verify the performance of these mechanisms,which provides a reference for scientific research and engineering applications.