Subject On Modified Particle Swarm Optimization Algorithm

Particle Swarm Optimization algorithm is one kind of important intelligent optimization methods, inspired by swarm intelligence of bird flocking and fish schooling. Its characteristics are simple and only a few parameters to be adjusted, fast convergence and easy implementation. It shows great potential in the target function optimization, neural network training and engineering practice.Several kinds of improved PSO are proposed in this paper. The main works of the dissertation can be summarized as follows:First, in order to improve global convergence of PSO, two improved PSO algorithms are put forward. One is doing random variation to the current individual extreme value. The other is doing random variation to the current particle and then put the results of variation as the next generation of the particle. Two algorithms are designed to jump out of the local optimal solution. Several typical functions test results show that the new algorithms on the convergence rate and accuracy are significantly better than standard particle swarm algorithm and the literature [35] algorithm. The new algorithms are effective on avoiding the premature and convergent on global search.Second, in order to improve the PSO algorithm search performance, an advanced PSO algorithm with step factor is put forward, the original particle swarm algorithm (PSO) has been improved in each updated particle's position to add a step factor. Thus, the algorithm overcomes the difficulty of the original PSO that the particle's ability of searching is decreasing during the last time of iteration, which is caused by the step factor of every particle is fixed on one. The results show the effectiveness of the proposed method.