Subspace-based methods for the identification of multivariable dynamic errors-in-variables models

Subspace-Based Optimal Iv Method for Closed-Loop System Identification

“…The use of the MOESP class of methods in a closedloop framework has received a special care since the beginning of the 90's (Verhaegen, 1993a;Chou and Verhaegen, 1997;Chou and Verhaegen, 1999;Zhao and Westwick, 2003;Oku and Fujii, 2004). The study realised in (Chou and Verhaegen, 1997) has more precisely proved that the PO MOESP algorithm could be directly applied from data collected in closed-loop if and only if u is a white noise, property which can't be checked in practice.…”

“…The study realised in (Chou and Verhaegen, 1997) has more precisely proved that the PO MOESP algorithm could be directly applied from data collected in closed-loop if and only if u is a white noise, property which can't be checked in practice. In this paper, it is proposed to modify the original PO MOESP method by introducing reconstructed past input and outputũ andỹ as instrumental variable.…”

“…It is interesting to notice that, according to the fact that the input u is not white, the most natural approach would consist in considering the following modified version of the QR factorisation developed in paragraph 4.2 of (Chou and Verhaegen, 1997) (see e.g. (Zhao and Westwick, 2003) for a second contribution)…”

“…One of the reasons for the success of SIM lies in the direct correspondence between geometric operations on matrices constructed from input/output data and their implementation in terms of well known, stable and reliable algorithms from the field of numerical linear algebra. Lot of them have been extended to the closed-loop case (Verhaegen, 1993a;V a n Overschee and De Moor, 1997;Chou and Verhaegen, 1997;Katayama et al, 2005) or directly developed in a closed-loop context (Qin and Ljung, 2003;Oku and Fujii, 2004;Huang et al, 2005). When looking at methods that can consistently identify plant models of systems operating in closed-loop while relying on simple linear (regression) algorithms, instrumental variable (IV) techniques seem to be rather attractive, but at the same time also not very often applied.…”

Research on Thermal Multivariable System Identification Based on Living Operating Data

“…SMI has a better numerical reliability and a modest computational complexity compared with the prediction error method (PEM), particularly when the number of outputs and states is large [2][3][4][5].SMI based multivariable output error state space (MOESP) has been proposed [6]. Ljung and McKelvey extended the subspace identification based on autoregressive exogenous (ARX) model into LS problems [7]; Van Overschee and De Moor provided a generic method for closed-loop subspace identifications [8].For error-in-variable (EIV) model structure, Huang et al proposed a subspace identification method based on orthogonal projection and instrumental variables [9].…”