基于群体适应度方差的粒子群优化算法

A PSO Algorithm Based on Colony Fitness Variance

由于粒子群算法在进化后期存在搜索速度较慢,容易陷入局部最优点以及搜索到解的时间较长且精度不高的缺点,所以对算法进行改进的研究就成为一个必要的课题。通过利用混沌的遍历性和随机性的特点,引入基于Tent映射的混沌理论机制,使算法在进化后期一旦陷入局部最优点就可以跳出局部最优点的位置,并且通过群体适应度方差的计算来判断当前群体的离散程度或聚集程度,进而判断是否需要以一定的概率选择微粒个体去进行混沌更新。几个测试函数的仿真实验结果也表明了该算法在搜索时间上、解的精度上都要远远优于标准的粒子群算法,是一种可行的优化工具,有一定的应用前景。

For the defects of particle swarm optimization algorithm such as lower search velocity, being prone to getting into local best position in later evolution phase, longer search time and lower precision, it is essential to make some researches about improvement of PSO. By making use of the ergodicity and randomicity characteristics of chaos, a chaos theory mechanism based on Tent mapping is introduced that may be helpful to guide particles to break away out of the position if it has gotten into local best position. And by calculating the colony fitness variance, it is also easy to judge its dispersion or congregation degree, furthermore, to make decisions whether to have chaos operations according to certain individual probability. The simulated experimental results show that it is prior to standard PSO in the search time and solutions＇ precision. Also it proves that it is a reasonable optimized tool and is promising.