基于Boltzmann学习策略的粒子群算法  

Particle Swarm Optimization Based on Boltzmann Learning Strategy

作　　者：艾解清[1,2] 高济[3] 

机构地区：[1]广东电网公司信息中心,广东广州510000 [2]广东电网公司信息化评测实验室,广东广州510000 [3]浙江大学人工智能所,浙江杭州310000

出　　处：《南京理工大学学报》2012年第3期402-407,共6页Journal of Nanjing University of Science and Technology

摘　　要：针对粒子群算法过早收敛导致容易陷入局部极值的问题,提出了一种基于Boltzmann学习策略的粒子群算法(BLSPSO)。借鉴模拟退火算法的思想,在标准粒子群算法中引入Boltzmann学习策略。在BLSPSO前期粒子能够学习不同的极值点,适当保持粒子个体多样性,提高算法全局寻优能力。在BLSPSO后期粒子更倾向于学习全局最优粒子,提高收敛速度,保证算法的稳定性。仿真结果表明,所提出的算法具有寻优能力强、搜索精度高等优点,可有效避免标准PSO算法的早熟收敛。该算法在求解多极值问题上与其他PSO算法相比有较好表现。An improved particle swarm optimization （PSO）based on Boltzmann learning strategy （BLSPSO） is proposed to overcome the problem of premature convergence and easily getting into local extremum of the the standard PSO. Using the idea of the simulated annealing algorithm for reference,the Boltzmann learning strategy is introduced into the standard PSO. In the prophase of BLSPSO, the particles can study different extreme points. The diversities of the particles are preserved to improve the BLSPSO＇s global optimization ability. In the anaphase of BLSPSO, the particle tends to study the global best information. The convergence velocity is improved, and the stability of the algorithm is ensured. The simulation results show that the BLSPSO has powerful optimizing ability, higher search veracity. It can avoid premature convergence effectively and have good performance in solving multimodal problems compared with other PSO algorithms.

关 键 词：粒子群算法 Boltzmann学习策略 模拟退火 全局寻优 多极值问题 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]