基于广义泊松矩泛函的闭环子空间辨识  

Closed-loop subspace identification via generalized Possion moment functionals

作　　者：于淼 王友谊 魏永涛 YU Miao;WANG Youyi;WEI Yongtao(School of Control Engineering,Northeastern University at Qinhuangdao,Qinhuangdao 066004,China)

机构地区：[1]东北大学秦皇岛分校控制工程学院,河北秦皇岛066004

出　　处：《系统工程与电子技术》2024年第6期2092-2098,共7页Systems Engineering and Electronics

基　　金：国家自然科学基金(62003082);河北省自然科学基金(F2021501018);河北省教育厅科学技术研究项目(ZD2022148)资助课题。

摘　　要：针对连续系统在闭环条件下进行辨识,由于将来输入和噪声的相关性,会导致辨识方法产生有偏估计结果的问题,提出一种基于广义泊松矩泛函的闭环子空间辨识算法。首先,采用广义泊松矩泛函变换得到输入输出信号的滤波器模型,从而得到连续系统的输入输出矩阵方程。然后,利用故障诊断领域中的等价空间代替子空间辨识过程中的观测空间进行辨识。最后,为了解决闭环辨识中由反馈控制器产生的估计偏差辨识问题,采用主成分分析法和辅助变量法,达到辨识系统一致性估计的目的。仿真结果验证了所提方法的有效性和精确性。To solve the problem of biased estimation due to the correlation between future input and noise in the identification process of continuous system,a closed-loop subspace identification algorithm utilizing generalized Possion moment functionals is presented.Firstly,the filter model of the input-output signals is obtained by generalized Possion moment functionals transformation,and then the input-output matrix equation of the continuous systems is obtained.Secondly,instead of using the observable subspaces in the process of subspace identification,this paper focus on the parity space which is commonly employed in fault detection.Finally,the system is estimated consistently by the principal component analysis and instrumental variable method,which solves the identification problems of biased results for the system operates in closed-loop identification due to the feedback controller.The effectivity and accuracy of the proposed method are verified by the simulation results.

关 键 词：子空间辨识 闭环辨识 广义泊松矩泛函 等价空间 

分 类 号：N945.14[自然科学总论—系统科学]