融合优质粒子分布的粒子群优化算法

Particle Swarm Optimization Algorithm Combination with the Distribution of Superior Quality Particles

针对传统的粒子群算法易于陷入局部最优和对高维空间搜索精度不高的缺点,提出融合优质粒子分布的粒子群优化算法.此算法根据分布估计算法,首先通过统计学习得到概率模型,再根据概率模型来产生优质粒子.这不仅仅能抑制传统粒子群早熟停滞的现象,还使得种群中每个粒子通过向自身历史最优值、群体最优值和优质粒子的学习而具有更佳的解决多峰值,多维搜索空间的能力.通过对几个常用测试函数的仿真实验表明,提出的新算法能够有效地跳出局部最优值,在多峰、多维空间内有更好的全局搜索能力,所以性能优于传统的粒子群优化算法.

To figure out the problems such as the conventional particle swarm algorithm is easily prone to fall into local optimum and its low accuracy of high-dimensional search, a new algorithm named particle swarm optimization algorithm combination with the distri- bution of superior quality particles is proposed. This improved algorithm is based on estimation of distribution algorithms. First of all, probabilistic model could be obtained by statistical learning, and then the improved algorithm generate the superior quality particles ac- cording to the probabilistic model. Not only the improved algorithm can overcome the shortcomings of premature convergence and stagnation of the conventional particle swarm, but also each particle in the population would have a great ability to solve the multi-peak capacity and high-dimensional search space by learning their own history optimal value, group optimal value and the superior quality particles. The results of the simulation of several commonly used test functions show that the new algorithm can effectively escape from local optimum value and have a better global search ability in the multi-peak and multi-dimensional space , so performance of the new algorithm is better than the conventional particle swarm algorithm.