速度多样性自适应选择模型的改进粒子群算法

Particle swarm optimization for adaptive selection model based on velocity diversity

针对单一进化模型无法满足粒子在不同阶段进化需求的问题,提出了一种自适应选择模型的改进粒子群算法。通过对粒子群算法的进化模型进行研究,给出了两种不同的进化模型,计算两种进化模型的速度多样性指标确定两种进化模型各自的选择概率,根据选择概率自适应选择相应的进化模式进行更新。速度多样性可以很好地反应粒子不断变化的进化情况,改进的粒子群算法可以根据进化情况自适应地调整各个进化模型的粒子比例,改善算法性能。为了进一步改进算法的整体性能,增加了一维学习机制。对比一些改进粒子群算法,进行8个测试函数的仿真实验,实验结果表明,该算法有5个测试函数的测试效果最好,Friendman检验和算法收敛对比分析结果表明,该算法具有良好的全局搜索能力和较快的收敛速度。

For single evolution model could not adaptively satisfy the requirements at different stages of the evolution, an adaptive selection model particle swarm optimization （ASPSO） was proposed. Two different evolutionary models were put forward by studying the evolutionary model of the particle swarm optimization. The particle adaptively selection choice evolution model was slected, according to probability which depended on the velocity diversity of the two kinds of evolution model. The velocity diver- sity was a reaction of the versatile evolution of the particles. The improved particle swarm optimization adjusted the ratio of each evolution model by the velocity diversity and then improved the algorithm performance. The simulation results of the problem in 8 test functions showed that, compared with other improved PSO variants, ASPSO was better in 5 test functions. What＇s more, both the Friendman test and the convergence analysis of the algorithm showed that improved ASPSO had good global search ability and faster convergence speed.