一种带有二维扰动和自适应学习因子的粒子群算法  

Particle Swarm Optimization Algorithm with Two Dimensional Disturbance and Adaptive Learning Factor

作　　者：王磊 王行甫[1] 苗付友[1] WANG Lei;WANG Xing-fu;MIAO Fu-you(Department of Computer Science and Technology,University of Science and Technology of China,Hefei 230027,China)

机构地区：[1]中国科学技术大学计算机科学与技术学院,合肥230027

出　　处：《小型微型计算机系统》2018年第11期2353-2357,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61572454;61472382;61272472)资助

摘　　要：针对粒子群算法(Particle Swarm Optimization,PSO)容易陷入局部最优值、后期收敛速度慢和收敛精度低等问题,提出了一种带有二维扰动和自适应学习因子的粒子群算法(TDDALFPSO).首先,提出了自适应惯性权重和学习因子调节算法调节惯性权重、认知系数和社会系数,提高了全局搜索能力和局部搜索能力;然后提出了基于位置、速度二维扰动更新粒子位置的算法,避免了不在最优值区域的全局历史最优值对搜索的误导,提高了算法的收敛速度和精度;最后通过变异一些适应度值最差的粒子,让它们搜索空间中的其他领域,增加了种群的多样性,改善了算法容易陷入早熟的问题.仿真实验表明和基本PSO算法相比,TDDALFPSO在收敛速度、精度和稳定性上有了明显的提高;并且在大多数优化问题上和基于线性惯性权重递减的PSO算法(PSO-W)、基于综合学习的PSO算法(CLPSO)、基于适应值距离比例的PSO算法、基于三角函数动态参数选择的PSO算法(TPSO)和带正弦函数因子的粒子群优化算法(TFPSO)相比,TDDALFPSO在收敛速度、精度和稳定性上都有一定的优势.The Particle Swarm Optimization （PSO） algorithm has the shortcomings of being easy to fall into local optimal value,slow convergence and low convergence accuracy.Aiming at these problems,the Particle Swarm Optimization algorithm with Two Dimensional Disturbance and Adaptive Learning Factor （TDDALFPSO） was proposed.Firstly,the Adaptive Inertia Weight and Learning Factors Adjustment algorithm was proposed to adjust the inertia weight,cognitive coefficient and social coefficient,and the global search ability and local search ability are improved.Secondly,the Position-Speed Two Dimensional Disturbance algorithm was proposed to reduce the impact of the misleading which the historical optimal value caused,and the convergence rate and accuracy of the solution was improved.Finally some of the worst fit particles were mutated to explore other areas,the diversity of the population was increased.The simulation results show that the TDDALFPSO has a significant improvement in convergence speed,accuracy and stability compared with the basic PSO algorithm,and compared with A Modified Particle Swarm Optimizer（PSO-W）,Comprehensive Learning Particle Swarm Optimizer for Global Optimization of Multimodal Functions（CLPSO）,Fitness Distance Ratio Based Particle Swarm Optimization（FDR-PSO）,Dynamic Parameter Selection Based on Trigonometric Function in Particle Swarm Optimization（TPSO） and Particle Swarm Optimization Based on Trigonometric Sine Factor（TFPSO）,TDDALFPSO has some advantages in convergence speed,accuracy and stability.

关 键 词：粒子群优化 早熟收敛 二维扰动 自适应学习因子 惯性权重 变异 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]