Reserch Of Many-objective Brain Storm Optimization Algorithm Based On Decomposition

In the real world,multi-objective optimization problems are ubiquitous.We usually call it a multi-objective optimization problem when the number of objectives is two or three.We call it a many-objective optimization problem when the number of objectives is greater than three.After years of continuous exploration,the multi-objective optimization algorithm has made good progress in both the theoretical level of the algorithm and the practical application.However,it is difficult to obtain a better Pareto frontier for the many-objective optimization problem.Therefore,how to ensure the convergence and diversity of the population solution set becomes the key to improving the performance of the algorithm.The brain storm optimization algorithm is an intelligent optimization algorithm based on group iterative search.It has the advantages of simple model,fast convergence speed,and few parameters.Researchers have applied it to engineering optimization,pattern recognition,multi-objective optimization and other fields,and have achieved good results.Based on this,the paper combines the brain storm optimization algorithm with multiple strategies to solve many-objective optimization problems.The main study work of the thesis includes the following aspects:(1)Firstly,in order to solve the problem of selection pressure reduction in the population evolution process of the existing multi-objective brain storm optimization algorithm based on Pareto dominance when solving the many-objective optimization problem,the MOBSO-SDR algorithm based on SDR dominance and angular crowding distance is proposed.The main idea is to use SDR dominance and angular crowding distance to update the archive set to preserve the convergence and distribution of the solution set during the iteration.The experimental results show that the convergence of the algorithm is significantly improved when solving problems with a large number of targets,and the diversity effect is improved compared with the cyclic crowding distance,but there is still room for improvement.(2)In view of the poor distribution of MOBSO-SDR,the decomposition-based strategy and the dominance-based method are combined,and a many-objective brain storm optimization algorithm MOBSO-SD which integrates dominance and decomposition strategies is proposed.The main idea is applying the strategy and archive set based on the neighborhood in the decomposition idea to the selection of high-quality parents in the brain storm optimization algorithm to increase the diversity of the algorithm,generating new individuals with simulated binary crossover to make the algorithm faster convergence,using SDR dominance and angular crowding distance to update the archive set to retain non-dominated solutions in the iterative process to improve the convergence and diversity of the algorithm,and using penalty boundary intersection method in the population update process.The simulation results show that the proposed algorithm can maintain good convergence and distribution in solving many-objective problems.(3)Finally,a many-objective brain storm optimization algorithm ADMOBSO based on reference points and decomposition is proposed in order to further improve the convergence and diversity of the irregular frontier when algorithm solves many-objective problems.The main idea is to adopt a strategy of adaptive clustering based on reference point poles,introduce archive sets,neighborhood ideas,and elite classes,and select high-quality parents in different types with probability according to brain storm ideas.And generating new individuals by the way of parental similarity to make the algorithm have better distribution.Using SDR dominance and weighted boundary intersection method to update the archive set to retain the non-dominated solution in the iterative process to improve the convergence and diversity of the algorithm.Similarly,the population individuals are updated by the penalty boundary method,and the reference point position is continuously updated according to the population target value during the algorithm iteration process to ensure the distribution of the population solution set.Finally,the results of a large number of simulation experiments show that the many-objective brain storm optimization algorithm proposed based on reference points and decomposition in this paper has better performance in solving linear,concave,and multi-peak problems.MOBSO-SD and ADMOBSO algorithms are applied to solve the storm drainage problem,and the experimental results have obtained a set of Pareto optimal solutions with uniform distribution.The performance of the algorithm proposed in this paper in solving many-objective optimization problems has been significantly improved,and the ability of the algorithm to solve practical problems has been enhanced,and it has important application prospects.