The Research On Particle Swarm Optimization Algorithm For Constrained Optimization And Dynamic Optimization

Constrained optimization problem and dynamic optimization problem are two major research fields in optimization problems. In real-world applications, they are complicated and difficult, so the subjects of solving and researching on constrained optimization problem and dynamic optimization problem are of great significance. The optimization algorithms based on swarm intelligence have shown better performance, also it have got great developments.A new optimization method named Particle Swarm Optimization (PSO) is inspired by the flocking and swarm behavior of birds, and fish schools. PSO is simple and efficient, and furthermore, its good performances and other features in dealing with single-objective optimization are useful in constrained optimization problem and dynamic optimization problem also. From some current research works,this paper works on the improved PSO algorithms based on constrained optimization and dynamic optimization respectively. The main works of this paper are:1.For constrained optimization problems, this paper describes an improved particle swarm optimization algorithm (IPSO) that incorporates a new mutation operator with dynamic mutation rate. At the early iteratives, the mutation rate is bigger and the aim is to make particles to explore smaller and isolate region and to avoid the algorithm trapping to the local optimum. While the algorithm runs, the mutation rate will be decreased to reduce the disturbing to the particles. IPSO adopts a new method that congregates some neighboring individuals to form multiple sub-populations in order to lead particles to explore new search space. Additionally, our algorithm incorporates a mechanism with a simple and easy penalty function based on the largest distance between the particle and the boundary of the feasible region to handle constraint. To compare with the state-of-art MOEAs on a well-established suite of test problems, the new approach is simple constructed, and results indicate that it works effective and has steady-state performance on constrained optimization problem.2.For the dynamic optimization problem that constructed by Sphere function, an improved PSO algorithm named w -PSO is presented. The w -PSO adopts random inertia weight and the dynamic mutation methods that are expressed in IPSO. The algorithm has better tracking capability than standard PSO. The experiment results of linear model, circular model and random model show that the w -PSO can solve this class dynamic optimization problem effectively.3.For the multi-peaks dynamic optimization problem that constructed by DF1, this paper describes a dynamic improved PSO algorithm DIPSO. DIPSO incorporates the dynamic mutation, multi-subpopulations methods of IPSO, and changes the heights and the positions of the peaks by variable parameters and chaotic model. This algorithm uses multi-subpopulations to search for the highest peak in order to track the dynamic environments. Furthermore, DIPSO apperceives the changes of the environment by tracking the fitness value of the global optimal particle. It is confirmed from the results that the proposed method is an effective algorithm for multi-peaks dynamic optimization problems.