论文部分内容阅读
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。
For the SISO linear differential equation model, the UD decomposition technique and the “shift property” of the data vector can be used to achieve the recursive decomposition of the information compression matrix, so that each iteration can obtain from 1 to n (n is the actual system The maximum possible order) the model of the majority of estimates and loss function values. The system order can be easily determined by using the value of each order loss function. When the system noise is colored noise, the party, government and army of the method can simultaneously determine the order of the noise model. The algorithm can significantly reduce the computational complexity of the identification process, have good numerical calculation quality and improve the identification accuracy.