Study Of Particle Swarm Optimization Based On Trust Region Method

Because the particle swarm optimization algorithm is easy to understand,fast to assemble and has less adjustable parameters,low requirements for optimization function,it has been widely used in scientific research,engineering practice and other fields.However,like other stochastic optimization algorithms,due to the blindness in the process of particles flight,the slow search speed and the high probability to lose population diversity,the particle swarm is easy to fall into the local optimal solution.To improve the optimization performance of particle swarm optimization algorithm,it is appropriate to introduce deterministic search in random search to improve the search efficiency of the algorithm.The trust region method has fast local convergence and ideal convergence under certain conditions.In addition to this,the method has stable numeric performance.By introducing the trust region method in the particle swarm optimization algorithm to guide the search of particles in the better direction,it can not only guarantee local convergence,speed up the convergence rate,but also has high certainty.At the same time,in order to maintain the diversity of the population,learn from the advantages of the trust region and mutation operator,the Cauchy mutation based on the trust region technology is implemented when the particles are caught in the local optima,then the particles are prevented from escaping the local optimum to improve the algorithm global optimization performance.In this paper,the trust region method is introduced into the particle swarm optimization to enhance the search ability of population.And two improved hybrid particle swarm optimization are proposed to advance the search performance and precision in the case of a valid search space.The main work of this thesis is as follows:(1)A novel hybrid adaptive particle swarm optimization embedded trust region method is proposed.Based on the diversity of ARPSO,the algorithm uses the trust region method to search locally,and uses the obtained potential optimal solution to adjust the search direction and avoid the blind duplicate invalid search.Compared with the standard particle swarm algorithm and several other improved algorithms,the experimental results show that the algorithm achieves better results in convergence accuracy and stability,and requires a much less iteration.Besides,it is analyzed that the algorithm can converge to the global optimal solution with higher probability theoretically.(2)Based on the idea of social class,a random recombination hierarchical particle swarm optimization with mutation based on trust region technique is proposed on the basis of standard particle swarm optimization.The proposed algorithm divides the population into three different levels according to the idea of the social class.The higher level is mainly responsible for the global exploration,and it is expected to find the local region of the optimal solution.And the Cauchy variation based on the trust region technique is introduced in the middle class to ensure the diversity of the population,avoiding falling into local extremes and the loss of the ability to continue searching.The role of the bottom of the particles group is to execute local search in detail,speed up the convergence rate and enhance the accuracy of convergence.Compared with the previously proposed algorithm,this algorithm does not need to obtain a search direction,reduce the computational complexity,and has no analytic requirement for the objective function.The experimental results show that the improved particle swarm optimization algorithm is superior to the standard particle swarm optimization and other related improvements.In this paper,based on the in-depth discussion and analysis of the principle of particle swarm optimization,the deterministic trust region method is introduced to find a potential descent direction to guide the particles,avoiding blind duplicate search,and drawing on the idea of social division of labor,combined with the mutation operation,the global search performance and convergence are guaranteed.The work of this paper provides a new idea for improving the performance of particle swarm optimization algorithm based on hybrid search.