Particle Swarm Optimization Algorithm Based On The Co-evolution Of The Parameter And The Strategy And Its Application

The problem of optimization exists in all aspects of human life, People has been looking for all kinds of suitable optimization methods to solve these problems. With the widen of the various application fields, some optimization problems has become more and more complex, and hard to be solved. Correspondingly, the optimization algorithm has been developed from some traditional algorithms to the intelligent optimization algorithm. The Particle Swarm Optimization Algorithm (PSO) is the one belongs to the intelligent optimization algorithm. As the branch of the swarm intelligence, the PSO algorithm originated in the group activities of the birds seeking food. To guaranteed find the food, in the process of the flying, the bird will collaborate with other birds and form the information sharing mechanism, the information will be exchanged between the individuals. Because of the PSO algorithm have the advantages of the simple principle, easy to be realized, few parameters and fast convergence speed, PSO has been more and more pro-gaze and has been widely applied in the fields of the scientific research and engineering application, etc.But in the meantime, PSO algorithm also exists some shortages. For example, in solving some multi-peaks problems, the PSO algorithm is easy to fall into local optimum, and easy occurs to premature convergence, especially in the search of the high dimension problems. In this paper, the basic principle of the PSO algorithm has been introduced and made some corresponding analysis on the reasons which has the influence on the performance of the PSO algorithm and made some summary on the current PSO algorithm research status. In order to improve the performance of the PSO algorithm, the following works have been done in this paper:(1) Making some analysis on the algorithm control parameter, put forward a double layer particle swarm optimization algorithm (DBPSO algorithm), the DBPSO algorithm can offers the real-time control parameters to the algorithm. The difference between this kinds of adaption evolution control parameters mechanism and the traditional adaption control parameters mechanism is this mechanism have changed the thought which make the control parameters changed according one fixed orbit. Using five benchmark test function to test the DBPSO algorithm, the experiment results shows that the DBPSO algorithm get a good performance, including the search speed and the search accuracy. (2) Concerning on the evolution strategy has the effect on the performance of the algorithm, point out that the different evolution strategy has different effect to the algorithm performance and put forward the ASPSO algorithm, forming a strategy pool by put some different evolution strategies together. In this situation, the algorithm can choose the most appropriate evolution strategy from the strategy pool according to the current status, the algorithm can determine whether to continue use current evolution strategy or choose the next evolution strategy. Using three test benchmark test function to test the performance of the ASPSO, the experiment results shows that the performance of the algorithm has been improved.(3) Making a combination of the DBPSO algorithm and the ASPSO algorithm and form a Co-evolution PSO algorithm based on the control parameters and the evolution strategy, the experiment result shows that the performance of the Co-evolution PSO algorithm have improved compared to the DBPSO algorithm and ASPSO algorithm in some aspects.The three kinds of improved PSO algorithm described above been applied in the optimization of the pressure vessel model design and get a satisfied result.