The Research Of Improved Particle Swarm Optimization Algorithm For Function Optimization Problems

Particle Swarm Optimization algorithm is a representative of the intelligent optimization algorithms. The algorithm is characterized by simple, few parameters, fast convergence and easy in implementation. So it attracted wide attention from experts and scholars, and at the same time has been a rapid development since it had been introduced. But now, the research on the Particle Swarm optimization algorithm is not perfect. At the same time, Particle Swarm optimization has some shortcomings, for example, low convergence accuracy, slow convergence speed, and easy falling into local optimum. Therefore, the study on the particle swarm optimization algorithm is very meaningful.The paper focuses on the applications in continuous function optimization problem. Based on the theoretical and experimental verification, this paper proposes two improved particle optimization algorithms. The main work and innovation are as follows: (1) A new Particle Swarm Optimization (PSO) algorithm is presented based on three methods of improvement in standard PSO. First, the iteration formula of PSO is changed and simplified by removal of velocity parameter that is unnecessary during the course of evolution. Second, the personal best value of each particle is replaced by the mean value of the personal best value of all particles. Third, acceleration coefficients are adaptively adjusted to improve the search performance of algorithm. The experimental results show that the proposed algorithm can effectively avoid premature convergence problem. At the same time, the convergence speed is faster than standard PSO and some other modified PSO algorithms.(2) A new particle swarm optimization (PSO) algorithm is presented to overcome disadvantages that standard PSO has shown in solving complex functions, including slow convergence rates, low precisions and premature convergence, etc. The proposed algorithm improves the performances of standard PSO by following methods:a) applying chaotic initialization for swarm, b) using adaptive inertia weight to enhance the balance of global and local search of algorithm, and c) introducing disturbance factors to avoid being trapped in local optimum. The experimental results show that the new algorithm has great advantages of convergence property over the standard PSO and some other modified PSO algorithms, and the global convergence performance improved significantly.Finally, base on summarization of the work, we put forward some further research directions.