Improved Particle Swarm Optimization Algorithm And Its Application

Optimization technique is concerned by the relevant experts and scholars, and has developed rapidly as a technique that has been widely used to solve technical optimal solution of various engineering problems. With the application areas are continuously expanding and the problems to be optimized are more complex, it becomes more difficult to seek the optimal solution, and some traditional optimization techniques can not meet the needs of the solution. The appear of intelligent optimization algorithm not only provides new ideas and means for optimize technology, and is widely used in economic, scientific and engineering problems. Particle Swarm Optimization is a population-based random optimization algorithm, comes from the study on the movement behavior of birds and fish populations. It as a new intelligent optimization algorithm, which is used to seek the optimal solution of complex problems, shows a strong advantage.There are some inherent flaws in PSO algorithm. For example, the loss of population diversity and search speed decreasing rapidly in the late running lead to stagnation and premature convergence. Therefore the adaptive inertia weight adjustment mechanism is introduced to the algorithm, and a dynamic adaptive particle swarm algorithm is presented. Using the dynamic adaptive regulation strategy inertia weight makes the inertia weight change with the group level timely based on the definition of population diversity measure, for adjusting the particle swarm search speed and direction to jump out of local optimum.In order to analyze the improved PSO algorithm performance, we carry out the experiments on the four representative testing functions, and compared with other classical algorithms. From the experimental result, we can see that the improved algorithm is effective for the premature convergence and insufficient local development.The improved algorithm proposed in this paper is used to optimize the key parameters of RBF neural network through studying the principles and steps of neural network based on PSO algorithm, and the network training algorithm is applied to time series forecasting problems. Through the prediction results we can conclude that RBF neural network trained by the improved PSO algorithm not only has fast convergence speed, but also has good generalization ability. In time series prediction problem, it has prediction accuracy, and is an effective method of time series prediction.