粒子群优化算法之惯性权值递减策略的改进
|
摘要
标准的粒子群算法引入惯性权重w,成为一种有效寻找函数极值的计算方法,且简单易行收敛速度快。目前普遍采用的是线性递减动态惯性权重策略,但其存在着复杂的非线性搜索过程。在线性递减策略的基础上,提出了非线性递减动态惯性权重策略,采用Griewank、Rastrigrin、Sphere、J.D.Schaffer 4个标准测试函数进行了仿真实验,与基本粒子群算法中惯性权重取固定值、线性递减LDIW和指数曲线非线性递减进行了比较。实验结果表明改进的非线性权值递减策略无论从收敛速度、收敛精度还是迭代次数都明显优于其他算法。
The standard particle group algorithm introduces inertial weight W, which has become an effective calculation method for finding the maximum value of functions, and it is simple and easy to achieve rapid convergence. Now it is generally used as a linear decreasing dynamic inertial weight strategy. Although this improvement is very successful, however, the search process is a nonlinear complex process. A nonlinear decreasing dynamic inertial weight strategy based on the linear diminishing strategy was proposed. The simulation experiments were carried out using the four standard test functions of Griewank, Rastrigrin, Sphere, and J.D. Schaffer, and the fixed weight, linear decrease LDIW and exponential cure nonlinear decline were taken with the inertial weight in the elementary particle group algorithm.The experimental results show that the improved nonlinear weight reduction strategy is superior to other algorithms in terms of convergence speed, convergence precision and iteration times.
The standard particle group algorithm introduces inertial weight W, which has become an effective calculation method for finding the maximum value of functions, and it is simple and easy to achieve rapid convergence. Now it is generally used as a linear decreasing dynamic inertial weight strategy. Although this improvement is very successful, however, the search process is a nonlinear complex process. A nonlinear decreasing dynamic inertial weight strategy based on the linear diminishing strategy was proposed. The simulation experiments were carried out using the four standard test functions of Griewank, Rastrigrin, Sphere, and J.D. Schaffer, and the fixed weight, linear decrease LDIW and exponential cure nonlinear decline were taken with the inertial weight in the elementary particle group algorithm.The experimental results show that the improved nonlinear weight reduction strategy is superior to other algorithms in terms of convergence speed, convergence precision and iteration times.
引文
[1]Kennedy J,Eberhart R C.Particle swarm optimization[C]//IEEE International Conference on Neural Networks,1995:1942-1948.
[2]吴雪娟.多目标粒子群优化算法的研究及应用[D].南京:南京邮电大学,2016.
[3]谢涛.基于(多)小波(包)、神经网络及优化的模拟电路故障诊断研究[D].长沙:湖南大学,2011.
[4]林繁涛,卢达,段永贤,等.基于收敛性线性粒子群优化算法的1-SVM电力设备状态分析方法[J].黑龙江大学工程学报,2017,8(4):75-82.
[5]Shi Y,Eberhart R C.A modified particle swarm optimizer[C]//Proceedings of the IEEE Conference on Evolutionary Computation,1998:69-73.
[6]张顶学,关治洪,刘新芝.一种动态改变惯性权重的自适应粒子群算法[J].控制与决策,2008,23(11):1253-1257.
[7]李会荣,高岳林,李济民.一种非线性递减惯性权重策略的粒子群优化算法[J].商洛学院学报,2007,21(4):16-20.
[8]陈贵敏,贾建援,韩琪.粒子群优化算法的惯性权值递减策略研究[J].西安交通大学学报,2006,40(1):53-56,61.
[9]杨巍,毛剑琳,熊晔.一种动态权值自适应粒子群优化典型函数问题[J].电子科技,2018,31(9):9-12.
[10]南杰琼,王晓东.改进惯性权值的粒子群优化算法[J].西安工程大学学报,2017,31(6):835-840.
[2]吴雪娟.多目标粒子群优化算法的研究及应用[D].南京:南京邮电大学,2016.
[3]谢涛.基于(多)小波(包)、神经网络及优化的模拟电路故障诊断研究[D].长沙:湖南大学,2011.
[4]林繁涛,卢达,段永贤,等.基于收敛性线性粒子群优化算法的1-SVM电力设备状态分析方法[J].黑龙江大学工程学报,2017,8(4):75-82.
[5]Shi Y,Eberhart R C.A modified particle swarm optimizer[C]//Proceedings of the IEEE Conference on Evolutionary Computation,1998:69-73.
[6]张顶学,关治洪,刘新芝.一种动态改变惯性权重的自适应粒子群算法[J].控制与决策,2008,23(11):1253-1257.
[7]李会荣,高岳林,李济民.一种非线性递减惯性权重策略的粒子群优化算法[J].商洛学院学报,2007,21(4):16-20.
[8]陈贵敏,贾建援,韩琪.粒子群优化算法的惯性权值递减策略研究[J].西安交通大学学报,2006,40(1):53-56,61.
[9]杨巍,毛剑琳,熊晔.一种动态权值自适应粒子群优化典型函数问题[J].电子科技,2018,31(9):9-12.
[10]南杰琼,王晓东.改进惯性权值的粒子群优化算法[J].西安工程大学学报,2017,31(6):835-840.