Many-objective Brain Storm Optimization Algorithm And Its Application Research

As a classic complex optimization problem,multi-objective optimization has a wide range of applications in engineering and real life.After several decades of development,various multi-objective optimization algorithms have been proposed by scholars,and have had very good results in solving multi-objective optimization problems of 2 and 3 objectives.In recent years,scholars are increasingly interested in many-objective optimization problems with a objective number greater than 3,but they have encountered many challenges in solving such problems.As an emerging group intelligent optimization algorithm,brain storm optimization algorithm has attracted the attention of researchers.It is also one of the research hotspots to expand on it and use it to solve multi-objective optimization problems.The multi-objective brain storm optimization algorithm has achieved good results and successfully used to solve practical problems.Although it is difficult to achieve the desired effect in dealing with many-objective optimization problems,it has great potential in dealing with such problems based on its clustering strategy and individual generation mechanism.Based on the analysis of the existing multi-objective brain storm optimization algorithm,this paper starts from the problem characteristics and adopts multiple strategies to improve the classical algorithm.The main contents can be summarized as follows:Firstly,based on the model of multi-objective optimization problem and performance index introduction,the multi-objective brain storm optimization algorithm is obtained from the brainstorming method based on the multi-objective optimization significance,and two classic improved versions are introduced.A large number of simulation results show the effectiveness of the algorithm in dealing with multi-objective optimization problems and the potential and shortcomings of dealing with many-objective optimization problems.Secondly,the algorithm is improved from two aspects of convergence and diversity.For the convergence,the reason why the performance of the algorithm deteriorates when the objective increases is analyzed.From the dominance of the algorithm,the SDR dominance and decomposition strategy is adopted instead of the ordinary Pareto dominance to increase the selection pressure of algorithm.In terms of diversity,angle crowding distance,reference point selection and cluster selection strategy are used instead of the crowded distance strategy adopted by the original algorithm in the archive set update,which can more reasonably perform the population update in the high-dimensional objective space.The simulation results on the many-objective optimization problem show that the modified algorithm has greatly improved in terms of convergence and diversity.Finally,the multi-objective optimization problem is analyzed.According to the problem characteristics,the decision variables are divided into convergence-related variables and diversity-related variables,and different strategies are optimized.For the convergence optimization,the idea of "divide and conquer" is adopted,and the independent decision variables are optimized separately.The simulated binary crossover is used to generate new individuals to accelerate the convergence of the algorithm.For the optimization of diversity,an idea based on corner clustering is proposed.The brainstorming idea is used to generate new individuals by probabilistic selection of corner points,which can improve the distribution characteristics of the population and enhance the uniformity and extent of the distribution of the algorithm.Compared with other classical algorithms,the simulation results on multi-objective optimization problem and many-objective optimization problems show that the proposed algorithm can obtain a set of fast-convergent and diverse Pareto optimal solutions.Based on the effective algorithm,the human-computer interaction interface of the algorithm is designed to improve the understanding and application of the algorithm for beginners.