Prony算法辨识传递函数的模型阶数选取研究  

Research on Model Order Selection of Transfer Function Identification Using Prony Method

作　　者：郭成[1] 李群湛[1] 

机构地区：[1]西南交通大学电气工程学院,成都610031

出　　处：《系统仿真学报》2009年第22期7042-7044,7047,共4页Journal of System Simulation

基　　金：国家科技支撑计划(2007BAA12B5)

摘　　要：针对Prony算法辨识传递函数的模型阶数选取问题,提出了一种基于SNR值及留数值的新方法。SNR值体现的是拟合性能,留数值体现的是模式参与程度。首先选取一个阶数初始值,然后在模型阶数取初始值条件下对输出信号进行Prony分析,最终依据SNR值及留数模值,得到适合的模型阶数。对典型传递函数的仿真分析验证了所提方法的有效性。To solve the model order selection problem of transfer function identification using prony method, an improved method was proposed based on value of SNR and residues. SNR value expressed the fitting property; absolute value of residues expressed the participation degree of mode. In this method, an initial value of order was selected firstly, and then output signal was estimated by prony analysis according to the initial value of order. Finally, the model order was obtained by SNR value and absolute value of residues. The simulation for the typical transfer function validates the effectivity of the presented method.

关 键 词：PRONY方法 传递函数 模型阶数 SNR 留数 

分 类 号：TM712[电气工程—电力系统及自动化]