文档介绍:
改进 PSO-NN 算法在微带线 S 参数模型中的
应用
高雪莲,王光波,黄吉畴,宋宁宁*
5
10
(华北电力大学电气与电子工程学院,北京 102206)
摘要:为了克服商用软件研究微带线 S 参数时计算代价大,耗时长的缺陷,本文提出了粒子
群-神经网络算法。首先,为了检验粒子群-神经网络算法的性能,利用三种算法性能校验函
数分别对粒子群-神经网络算法、粒子群算法和 BP 神经网络算法进行测试,并将三者的结果
进行比较,比较结果表明粒子群-神经网络算法的精度最高,稳定性最好。最后,将粒子群-
神经网络算法应用于微带线 S 参数建模研究中,用 CST 软件得到的微带线 S 参数作为训练
数据和验证数据,并与粒子群和 BP 神经网络算法的结果进行比较,发现粒子群-神经网络算
法在误差和稳定性上都有明显优势,表明该算法用于微带线 S 参数建模中是可行有效的。
关键词:微带线 S 参数;测试函数;CST 软件;粒子群-神经网络算法(PSO-NN)
中图分类号:TM210
15
Application of Improved PSO-NN Algorithm in the Model
for S-Parameter of Microstrip Line
GAO Xuelian, WANG Guangbo, HUANG Jichou, SONG Ningning
(School of Electrical and Electronic Engineering, North China Electron Power University,
20
25
30
35
40
45
Beijing 102206)
Abstract: To e putational cost, time consuming defects of calculation of
S-parameters of the microstrip line with mercial software, Particle Swarm
Optimization-work (PSO-NN) algorithm is proposed. Firstly, this paper applies
PSO-NN algorithm, together with Particle Swarm Optimization (PSO) and BP work
algorithm, to three Algorithms Performance Check Functions to examine the performance of the
PSO-NN algorithm. Compared with the other two algorithms, PSO-NN algorithm is proved to
have the highest precision and the best stability. Finally, PSO-NN algorithm is applied to the
general model for S-parameter of microstrip line, using the S-parameters