| In our real life,problems in many industries can be transformed into optimization problems.With the maturity of computer technology,the use of computers to find the optimal solution to many optimization problems has become an efficient,economical and accurate way.The optimization algorithms used by computers have been gradually improved as the complexity of optimization problems has increased.Earlier mathematical optimization methods and standard evolutionary algorithms are no longer able to solve today’s more complex multimodal optimization problems with multiple global optimal solutions in the solution region.As the size and complexity of these multimodal optimization problems continue to increase,they have become a popular branch of optimization problems,attracting many scholars to participate in the research.Scholars hope to develop an efficient and accurate solution method to deal with such problems,and multimodal optimization algorithms are born.The goal of multimodal optimization algorithms is to find all optimal solutions in the solution region in a single run of the algorithm,providing more options for solving problems while improving computational efficiency.For this purpose,scholars have proposed the concept called niching techniques.By dividing the population into different subpopulations and keeping multiple subpopulations searching in parallel,the niching techniques can effectively maintain population diversity and prevent premature convergence.The niching techniques has now become the mainstream application techniques for multimodal optimization algorithms.In this paper,we propose a multimodal optimization algorithm based on group collaboration mechanism,which combines three techniques: niching techniques,differential evolutionary algorithm and mathematical optimization method.The algorithm is first divided into several research groups,and then free discussion groups are formed through independent exploration within different research groups and mutual communication between groups.Finally,the BFGS local search algorithm is applied to these outstanding individuals to achieve fast convergence to find the global optimal solution.During the operation,a lifetime mechanism is added for the population individuals to remove and re-initialize randomly the individuals that fall into local optimum for a long time in time to maintain the population diversity and improve the utilization of computational resources.When dealing with highdimensional data,principal component analysis is used to perform dimensionality reduction so that the algorithm clustering can be carried out effectively.The algorithm proposed in this paper is compared with other 11 multimodal optimization algorithms and applied in the experiments of CEC’2013 benchmark test set for multimodal optimization problems,and the experimental results show that the algorithm proposed in this paper is significantly better than the other 11 compared algorithms.Meanwhile,ablation experiments are conducted for each part of this paper’s algorithm to prove the effectiveness of each part of the algorithm.Finally,the algorithm of this paper is applied to the practical application problem of PV model parameter optimization and good results are achieved. |
