Improved Particle Swarm Optimization Algorithm And Its Application

Particle swarm optimization(PSO),which is a swarm-based intelligent optimization algorithm,can obtain a good performance because of its simplity in principle and imple-mentation,especially in dealing with continuous optimization and discrete optimization problems.It has been widely applied in many fields such as objective function opti-mization,neural network training and fuzzy control systems,etc.However,due to the special iterative mechanism,PSO still has some shortcomings,such as premature con-vergence and easy to fall into local optimum.Therefore,in order to further improve the performance of PSO,prevent particles from falling into local optimum and enlarge the applications of PSO,this paper has the following work:(1)To enhance the performance of PSO for solving large-scale numerical optimiza-tions and engineering design problems,an adaptive disruption strategy,which originates from the disruption phenomenon of astrophysics,is proposed to balance the abilities be-tween global exploration and local exploitation better.Meanwhile,a Cauchy mutation is utilized to a certain dimension of the best particle to help particle jump out the local optima,and an adaptive disruption operator based on the iteration number is used to bal-ance the abilities between exploration and exploitation better by enhancing the diversity of the population.Nine well-known large-scale unconstrained problems,ten complicated shifted and/or rotated functions and four famous constrained engineering problems are utilized to validate the performance of the proposed DPSO algorithm compared with some of state-of-the-art algorithms.Experimental results and statistic analysis confirm the effectiveness and promising performance of the proposed algorithm.(2)A hybrid strategy-based ANN-HPSO-CA algorithm is proposed to optimize arti-ficial neural network(ANN).Firstly,to balance global exploration and local exploitation better and prevent particles from trapping in local optima,a cellular automata(CA)strat-egy is involved in the human behavior-based PSO(HPSO)algorithm,which is denoted as HPSO-CA.And the convergence performance of the proposed HPSO-CA algorithm was verified through the simulation experiment of 6 composite functions.Then,the proposed HPSO-CA algorithm is combined with ANN to prevent ANN from trapping in local min-ima.Finally,15 complex real-world datasets are used to validate the performance of ANN-HPSO-CA compared with some well-known EA-based ANN models.Experimen-tal results confirm that the proposed ANN-HPSO-CA algorithm outperforms the other predictive EA-based ANN models.