SISO线性模型的阶次辨识和参数估算

对于ＳＩＳＯ线性差分方程模型,利用ＵＤ分解技术和数据向量的“移位性质”,可实现信息压缩阵的递推分解,从而在每步递推中能同时获得从１到ｎ（ｎ为实际系统的最大可能阶次）各阶的模型多数估值和损失函数值。利用所得的各阶损失函数值,可方便地确定系统的阶次、当系统噪声为有色噪声时,本方法党政军可同时确定噪声模型的阶次。本算法可显著减少辨识过程的计算量,具有良好的数值计算品质,提高了辨识精度。

A Recursive Algorithm for Parameter Estimation and Order identification of SISO Model

In the identification of SISO linear difference equation model, implementing recursive UD factorization technique and "the Shift Structure" of data vector, the so-called Condensed information Matrix (CIM) can be recursively factorized, thus all the model parameters and loss functions from order 1 to n (n is the maximum possible order of the real process) can be obtained simultaneously in every recursion. According to the loss functions, the model order can be easily determined and the order of the noise model can also be determined simultaneously if the model noise is colored. The proposed algorithm can considerably reduce the computational burden of identification and also have excellent numerical property.