Abstract: Stocks diversity is the precondition for ensuring the convergence of PSO algorithm. The definition of stocks diversity is clear and the operand is small, moreover which was analyzed by particle evolution degree and aggregation degree. A changed algorithm was proposed based on adjusting weight adaptively. The algorithm ensures population diversity and avoids premature convergence effectively. Simulation results indicate that this algorithm not only speeds up the population the evolution speed, but also strengthens the algorithm the overall situation astringency, and convergence of probability also increases from 15% to 100%.

Key words: Diversity ,  Particle swarm optimization algorithm , Dynamic weight , Premature 

摘要： 种群的多样性是保证粒子群优化算法收敛的前提条件, 基于此提出了一个概念清晰、运算量小的多
样性定义, 并从粒子在寻优过程中粒子聚合程度和速度进化程度出发分析粒子群的多样性。在此基础上, 提出了一
种基于动态权值的改进算法, 算法能自适应的调整惯性因子以保持种群多样性, 有效地避免了早熟收敛。仿真实验
表明该算法不仅能加快种群的进化速度, 而且还能增强算法的全局收敛性, 收敛概率也从15 %增加到100%。

关键词: 多样性 , 粒子群算法 , 动态权值 , 早熟