Improvemrnt Of Tree-seed Algorithm And Its Appication In Solving Real-word Constrained Optimization Prblems

With the rapid development of society,solving large-scale complex problems(such as NP problems)has become a difficulty in the research field.As an effective means of solving complex problems,natural heuristic algorithm has attracted more and more attention.As a natural heuristic optimization method,tree species algorithm has attracted great attention of scholars since it was proposed by Kiran in 2015 because of its significant advantages such as few parameters,simple structure and good effect.However,the algorithm still has some shortcomings:(1)The existing tree seed evolution mechanism is difficult to ensure that the tree is constantly updated in the iteration,and it is easy to fall into local optimization;(2)The existing seed generation mechanism takes the tree as the center,which has the problem of "local strong and global weak",and does not achieve an effective balance between the two;(3)The algorithm is difficult to solve the optimization problem of high-dimensional search space.In view of the above shortcomings of tree species algorithm,this thesis proposes three improved algorithms based on the original algorithm,which are as follows:Improved algorithm 1、A hybrid algorithm based on tree seed and salp swarm optimization(TSA-SSA).Based on the tree species algorithm,the spatial search idea of global optimal guidance in salp swarm algorithm is introduced,which effectively enhances the global search ability of TSA-SSA algorithm and speeds up the convergence speed.On this basis,the spatial search model of SSA is used for reference,and a non evolutionary tree migration mechanism is proposed,which makes up for the difficulty of tree seed evolution in tree species algorithm and avoids falling into local optimization.Improved algorithm 2、A tree seed variant algorithm(gold＿TSA)based on the golden section rule.The golden section proportion coefficient is introduced into the algorithm,and a seed generation mechanism based on the golden section proportion is proposed to delimit the golden growth area suitable for seed growth.On this basis,the principle that non evolutionary trees should move in time is put forward,which improves the evolution mechanism of tree seed.Through comparative experiments,the proposed tree seed evolution mechanism solves the problem that the original algorithm is easy to fall into local optimization in iteration to a certain extent.Improved algorithm 3、A tree seed variant algorithm(PTSA)based on Pareto law.In order to effectively solve the non evolutionary tree problem generated in the iteration of the original algorithm,according to Pareto’s law,the group composed of trees with the top 20% fitness value in the algorithm population is defined as a high-quality tree group,and the non evolutionary tree is migrated to the high-quality tree group.At the same time,a portable seed generation mechanism is proposed to improve the lack of global search ability in the original seed generation mechanism.The experimental results show that the tree variety algorithm based on Pareto’s law can effectively make up for the defect that the evolution mechanism of tree seed is easy to stagnate in the original algorithm,so as to improve the reproduction mode of trees and seeds.Finally,this thesis uses 57 real-world constrained optimization problems,which come from the real-world single objective constrained optimization competition of the Institute of electrical and electronic engineers Evolutionary Computing Conference(IEEE CEC)in the 2020 World Conference on Computational Intelligence(WCCI),to verify the effectiveness of the proposed three improved algorithms in solving real complex optimization problems.The results show that the three improved algorithms can solve most of the actual constrained optimization problems.Compared with the top8 algorithms in the competition,they have certain competitiveness and have broad application prospects in solving complex constrained optimization problems in management applications.