Chaotic Particle Swarm Optimization Algorithm Based On Adaptive Variation And Cultural Framework

Many practical problems in scientific and engineering can often be formulated to various kinds of mathematical programming problems. The traditional methods could not be applied to these problems since they require much more analytic properties of the problem. Because of simple mechanism and easy realization, inherent parallel nature, height robust and the lower requirements for analytic properties of problem, swarm intelligence algorithms have attracted widespread attention, and become one of the effective methods to solve complex mathematical programming problems. As one of the typical swarm intelligence algorithms, particle swarm optimization (PSO) algorithm not only has the advantages of swarm intelligent algorithm, but also has fast convergence speed, less parameters, easy to combine with other evolutionary algorithms, so it has a wide application. But the PSO algorithm has a low searching efficiency and easy to premature convergence in the late optimization. To improve the performance of PSO, this paper presents two kinds improved algorithms:1. To improve the global searching ability of the algorithm and increase the diversity of population, a chaotic particle swarm optimization with adaptive mutation strategy (ACPSO) is proposed by constructing a new chaotic map and a new adaptive mutation strategy. The ergodicity of chaos mapping increases the diversity of the population, and the adaptive mutation strategy improves the optimization efficiency.2. To improve the optimization performance and convergence speed further, a chaotic particle swarm optimization based on the cultural framework (CCPSO) algorithm is represented by using the complementary advantages of the thought. The new algorithm not only keeps the parallel evolution of population space and belief space of the cultural algorithm, but also keeps the fast convergence of the particle swarm algorithm. Thus the proposed method improves the global searching ability and convergence performance. In addition, the new algorithm employs the fitness variance to judge whether the population fell into local optimum in the population space, and uses Logistic chaotic map to help the algorithm jump out of local optimal area only when the algorithm fell into local optimum. So the new method avoids premature convergence and ensures the diversity of population.The simulation results of two algorithms on six benchmark functions show that the new algorithms have better searching ability and faster convergence speed.