The Improvement Of Particle Swarm Optimization And Application Research

Particle Swarm Optimization(PSO) was presented by Dr. Eberhart and others. It simulates the foraging activity of birds to obtain the optimal solution. Particle Swarm Optimization is widely used in many fields domain engineering, science, economics and management for the reasons of its fast convergence speed, the model of simple, and few parameters need to adjust. Differential evolution(DE) algorithm is presented by the Price and others. Compared with other algorithms, Differential evolution(DE) has great advantages in stability and convergence due to its advantages of convenient operation, good performance.However, the convergence precision and speed of PSO and DE have great difference for different optimization problems. For example, for some multi-modal functions, PSO has some shortcomings including not high convergence precision and trapping in local easily. DE also exists some disadvantages: the slowly convergence speed in the later period of evolution for some complex problems. Focusing on the disadvantages of the above algorithms, this paper proposes two hybrid optimization algorithms which based on two kinds of algorithms. The paper include the following contents:(1) This paper proposes a hybrid optimization based on improvement of PSO and DE.The first step evolve the population to generate the individual optimal positions by applying the PSO algorithm, then evolve the individual optimal positions by using the DE algorithm.In order to coordinate the global and local search ability, the new algorithm adopts the inertia weight and learning factor nonlinear change strategy; and in order to improve the ability to jump out of local optimum, DE algorithm adopts scaling factor nonlinear change strategy.(2) DE-PSO involves two variables: concentration and the concentration of probability,which divide the population into two parts. The first part, the particle concentration is higher,the space is more intensive and easy to fall into local optimum, this part uses DE optimization;the second part, the third part particle concentration is smaller, it uses PSO optimization.Finally, in order to test the performance of algorithm, massive experiments of standard benchmark functions in different dimension draw the following results: compared with the traditional PSO and DE, the hybrid algorithm have better convergence, higher accuracy. More importantly, it has good effect in high dimensional problems.