Modification And Application Of Particle Swarm Optimization

In 1995,J.Kennedy and R.C.Eberhart proposed the Particle Swarm Optimization Algorithm(Particle swarm optimization,PSO).Particle swarm optimization algorithm is a new intelligent optimization algorithm based on the study of foraging behavior of birds in the biological world.Because the initial point of PSO algorithm is given randomly,the algorithm has low precision and slow convergence speed when solving the complex multi-peek function optimization problem.In this thesis,two classes of improved particle swarm optimization algorithms and a class of improved simplified particle swarm algorithm are proposed for the shortcomings of existing particle swarm optimization algorithms,such as local optimum,low precision of solution and slow convergence rate.The main work of this thesis is as follows:1.Briefly introduce the background of PSO algorithm.The existing results related to the inertia weight in the PSO algorithm are compared under the same parameter setting.We also analyzed the pros and cons.2.When the inertia weight of particle swarm optimization algorithm takes the linear decreasing function,particle swarm optimization algorithm is prone to show the shortcomings of premature convergence and low precision.In order to overcome them,a class of nonlinear decreasing inertia weight is proposed,which is that we take inertia weight as a nonlinear decreasing cosine function in the early iteration,and the inertia weight is taken as a smaller random value at the end of iteration.And the effectiveness of the improved algorithm is verified by numerical experiments.3.When the learning factor of PSO algorithm is a fixed value,the algorithm has the disadvantages of low precision and slow convergence speed.To overcome them,a class of learning factors based on exponential function change instead of fixed value is proposed in this paper.The learning factor1 c is taken as a nonlinear exponential function and2 c is a small random value in the early iteration.However,at the end of iteration,the learning factor2 c is taken as a nonlinear exponential function and the learning factor1 c takes a smaller random value.The traveling salesman problem verifies the effectiveness of the improved algorithm.4.When the inertia weight of the particle swarm optimization algorithm is fixed,the algorithm is easy to fall into local optimum.To overcome this shortcoming,an improved simplified particle swarm optimization algorithm is proposed in this thesis.In the thesis,we replaced the fixed value of inertia weight with a nonlinear decreasing exponential function.PID parameter tuning is solved by the improved simplified PSO algorithm,which shows the effectiveness of the improved simplified particle swarm optimization algorithm is verified.5.Finally,we proved that simplified particle swarm optimization algorithm for single particle and population particle to satisfy Markov process.