摘要
微粒群算法是基于群体智能的全局优化算法,在许多领域得到广泛的应用。该算法具有简单易于实现的优点,但是容易陷入局部极值尤其是采用动态惯性因子。采用动态惯性因子有利于提高微粒群算法的收敛速度,但降低了其全局搜索能力。针对具有惯性因子微粒群算法在进化过程中微粒群多样性减弱容易陷入局部最优值的问题,以非线性动态惯性因子的微粒群算法为基础,提出1种基于部分微粒更新的微粒群算法,以提高微粒群的多样性,进而提高了算法的全局搜索能力。新算法利用Sphere、Rastrigin、Rosenbrock、Schaffer、Freudenstein-Roth、Goldstern-Price 6个经典测试函数进行测试,并与基本微粒群算法和具有线性动态惯性因子微粒群算法比较。通过模拟优化比较,新算法寻优效率高、全局性能好、优化结果稳定,新算法能有效提高微粒群的多样性,具有较好的收敛性能和全局优化能力,尤其适合多峰函数的优化。
Particle swarm optimization is a global optimization techique based on swarm intelligence, which is simple to program and is applied to many fields widely. However, PSO algorithm gets into the local optimization point easily especially for the one with dynamically changing iintertia weight factor. Application of dynmaically changing interia weight factor is propitious to convergence velocity but not to global optimization performance. The population diversity is easy to decrease in searching process for particle swarm optimization algorithm with dynamically changing inertia weight factor, which makes algorithm easy to trap into local optimum. Considering this problem, a improved particle swarm optimization algorithm based on dynmaically changing nonliear inertia weight factor and partially regenerate particle is introduced. The new algorithm is tested with Sphere, Rastrigin, Rosenbrock, Schaffer, Freudenstein-Roth and Goldstem-Price benchmark functions. The numerical simulation results show that the new algorithm have better performance of finding global optimum, hight optimization effciency and rapid convergence velocity, espacilly for multimodal function.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2010年第9期1271-1276,共6页
Computers and Applied Chemistry
基金
福州大学科技发展基金(2008-XQ-12)
关键词
微粒群
非线性
更新
惯性因子
particle swarm, nonlinear, regenerate, inertia weight factor
