Box-Jenkins模型偏差补偿方法与其他辨识方法的比较

Comparisons of bias compensation methods and other identification approaches for Box-Jenkins models

对于存在相关噪声干扰的Box—Jenkins系统,本文借助于偏差补偿原理,推导了一个偏差补偿最小二乘(BCLS)辨识方法;理论分析说明BCLS方法能够给出系统模型参数的无偏估计,并将提出的方法与递推增广最小二乘算法和递推广义增广最小二乘算法进行了比较研究;用仿真试验分析了这些算法的各自特点和适用范围。

For Box-Jenkins systems with correlated noises, a bias compensation least squares （BCLS） identification method is proposed by means of the bias compensation principle. The analysis is then given to show that the BCLS algorithm can give the unbiased estimates of the system model parameters. Finally, the advantages of the proposed BCLS algorithm over the recursive extended least squares algorithm and recursive generalized extended least squares algorithm are shown by using simulation tests.