Improvement And Application Of Brainstorming Optimization Algorithm In Several Categories Optimization Problems

Brain Storm Optimization(BSO),a new swarm intelligence algorithm is generated by simulating the process of human beings.Although the thought of the BSO algorithm is new,the research of it is still in an early stage and many aspects need to be improved and modified,especially its application fields need to be further expanded.Therefore,this paper tries to apply the BSO algorithm into multimodal optimization,multi-objective optimization and constrained optimization problems.The pap er mainly studies the following three aspects of research work:Firstly,an improved algorithm named Self-adaptive Multi-objective Brain Storm Optimization(SMOBSO)is proposed.On one hand,using a simple clustering operation replaces choosing points randomly to represent the exact k-means clustering in basic algorithm.On the other hand,the paper introduces a probability of openness.By choosing an adaptive mutation method to avoid the algorithmtrapping into local optimum and improve uneven distribution of defects effectively.What's more,a loop crowded distance is used to maintain the external collection archive.Lots of simulations of multiply benchmark functions have showed that the improved algorithm not only increases the diversity of population,but also improves the abi lity of convergence.Secondly,for multimodal optimization,a simple clustering operator is added to the basic algorithm.Cluster the individuals firstly and then update the individuals by mutation.Beyond that,a retention mechanism of adaptive parameter method is introduced to select individuals.As a result,the algorithm not only can find global optimization in the feasible region,but also can offer a number of meaningful local optimal solutions to solve multimodal optimization problem better.At last,by simulating the benchmark functions of the international standard testing library and comparing with other common algorithms,the effectiveness and applicability of the proposed algorithm have been verified.Finally,for constrained optimization problem,this paper combines the modified brain storm optimization and information of constrained optimization and makes full use of its convergence operator.Divide the population into several subpopulations according to the certain rules to make the individuals close to the feasible region from different directions to improve the search ability of algorithm.Then according to the selection mechanism of BSO,choose individuals of each subpopulation unevenly to maintain and increase the diversity of population.Besides,the parallel search is constantly used to local region that can effectively avoid the premature in the algorithm.