闭环子空间辨识方法的分析与比较

Analysis and Comparison of Subspace Identification Methods for Closed-loop System

在实际应用中,辨识方法的辨识精度和辨识效率一直是人们关注的指标,也是人们选择辨识方法的主要依据。针对多种闭环子空间辨识方法的辨识精度和辨识效率问题的研究,首先归纳和总结了基于正交分解和基于正交投影闭环子空间辨识方法的理论和实现;然后扩展提出了基于正交分解的闭环子空间辨识方法 ORT_POMOESP、ORT_N4SID和基于正交投影的闭环子空间辨识方法 CSOPIM_W2;最后考虑系统输入输出测量噪声,针对过程噪声为白噪声和有色噪声两种情况下,通过仿真算例以数值分析的形式,对比研究了多种闭环子空间辨识方法的辨识精度和辨识效率。该研究不仅对子空间辨识方法应用于实际工业过程的建模具有实际的参考价值,而且对实际工程应用中闭环子空间辨识算法的选用具有一定的指导意义。

In practical applications, the accuracy and efficiency of the identification method are not only the important indicators, but also the main basis on choosing identification method. Focusing on Accuracy and efficiency for a variety of closed-loop subspace identification methods （CSIMs）, firstly, this paper summarizes the theory and implementation of CSIMs. Then, ORT_POMOESP, ORT_N4SID and CSOPIM_W2 methods are developed with the extension of existing CSIMs based on orthogonal decomposition and orthogonal projection respectively. Finally, considering the input and output measured noise, we compare and analyze identification accuracy and efficiency of CSIMs through simulation examples in the case that the process noise is white noise and colored noise respectively. The results of this paper have actual significance of reference and guidance for the application of CSIMs in the modeling of actual industrial process.