：针对粒子群算法进行多极点函数优化时

存在的局部极小点和搜寻效率低的问题，引入了小

生境的思想到粒子群算法中，以粒子的最好位置为

中心，粒子的最好的个体解对应的适应值为半径建

立圆形小生境。stretching

技术，其次对子群体采用解散策略，即当在子群体中找到一个极值点后把子群体解散回归主群体，最

后设置子群体创建时的半径阈值，避免子群体半径过大。该算法解决了标准的NichePS0算法在处理

多峰函数时，极值点的个数依赖于子群体个数及极值点容易出现重复、遗漏等问题。对3个常用的基

本测试函数的实验表明，新算法(SNPsO)在多峰函数寻优中解的稳定性、收敛性和覆盖率均优于标准

NichePS0。-Niching PSO algorithm to deal with complex multimodal function optimization problems exist some defects, asked the the an improved the niche sNPsO algorithm. sNPSO algorithm introduced in which the order of the niche ideas stretching technology is the first application in the main group, the dissolution of the strategy followed by the sub-groups, find an extreme point in the sub-groups after the handle the groups dissolution regression main groups, the final set radius threshold value, the creation of the subset populations avoid subpopulations radius is too large. The algorithm to solve the standard NichePS0 algorithm multimodal function, the number of extreme points depends on the number of sub-groups and the extreme points prone repeat omissions. 3 basic test function experiments show that the new algorithm (SNPsO) in multimodal function optimization solution stability, convergence and coverage are better than standard NichePS0.