一种可以同时辨识模型参数和阶次的辅助变量法

An Implementation of the Instrumental Variable Method for Simultaneous Identification of Model Order and Parameters

中文摘要

      在用辅助变量法对SISO线性模型进行辨识时，如果将其数据向量和辅助向量重新排列并扩维，利用由此而出现的“移位性质”，则可构成一种新的数据乘积矩阵，其逆矩阵称作信息压缩阵。本文提出一种新的UDVT分解方法，并利用它对这个信息压缩阵进行递推分解，构成了一种能同时进行模型参数和阶次辨识的递推算法。使用这种新的辨识算法，可以减小整个辨识过程的计算量，改善数值计算品质，提高辨识精度。

英文摘要

      The paper has presented an implementation of the instrumental variable method which can be used in simultaneous identification of the model order and parameters. By extension and rearrangement of the data vectors in the IV algorithm, a covariance matrix called the Condensed Information Matrix(CIM), which is nonsymmetric, can be constructed. Applying the UDVT factorization technique to the matrix, the model order and parameters can be obtained simultaneously. This identification method is superior to the conventional IV method when the model order is high. Our algorithm has better in accuracy and requires less computational efforts.