Font Size: a A A

Closed-loop Subspace Identification And Its Application

Posted on:2011-01-22Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y F LiFull Text:PDF
GTID:1118330332478568Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
Accurate system models are the premise and foundation for the application of ad-vanced control technology. As an effective multi-variable system modeling tool, sub-space identification methods have attracted a lot of attention since the early nineties of last century in many areas such as control theory, signal processing and structural en-gineering. Over the two decades, research on the subspace identification method has gradually formed the basic theoretic system, including algorithm implementation, per-formance analysis and algorithm applications. The basic subspace identification methods are limited to open-loop systems. Later subspace identification methods are proposed for the closed-loop systems and nonlinear systems. Since closed-loop system identification has great significance in practice, this work focuses on the closed-loop subspace iden-tification methods, including the improvement of the theory and the extension of the algorithms as well as the application in industrial processes.The main contents of this dissertation are outlined as follows,1. Most of subspace identification methods only consider the output noise, assuming that the input variables are not subject to noise contamination, which is obviously not the case in practice. For EIV (errors-in-variables) models, which consider both the input and output variables are subject to noise pollution, a closed-loop subspace identification method is proposed based on instrumental variables in this paper. The existing subspace methods using instrumental variables in the literature may deliver a bias in the presence of feedback control. Thus a remedy to eliminate the bias is suggested through the use of new instrument variables. Several choices of instrumental variables are presented and then a comparative study is done by numerical simulations. Compared to the existing subspace methods via both PCA (principal component analysis) and instrumental variables, the proposed method is simpler and more straightforward, and it is easy for implementation.2. Appropriate combination of the traditional closed-loop identification method and the subspace method to form a new identification method can exploit their own particular advantages for mutual benefit. In this paper, a new closed-loop subspace identification method is explored under the framework of the traditional two-stage method. In the first stage the intermediate variables are directly estimated through orthogonal decomposition of signals. The second stage comes down to an open loop identification problem where the system matrices can be estimated through a certain subspace identification method. Different subspace identification meth-ods are taken into account in the second stage and are compared by simulation examples. The proposed method is applied in a straightforward manner to iden-tify the system model without controller information. Compared with the existing subspace identification based on signal decomposition, the proposed method is of simple principle and is easy to extend. Moreover, a relatively simple approach is provided to estimate the Kalman gain and the noise variance.3. Most of the existing recursive subspace identification methods are proposed for open-loop system, and they are not suitable for online identification under closed-loop conditions. A recursive subspace identification method is developed based on predictor Markov parameters using closed-loop data. Firstly, the predictor Markov parameters are estimated through the recursive least squares method with forgetting factor. Then the subspace tracking method PAST is applied to obtain the update of the extended observability matrix. Two approaches are provided to estimate the system matrices finally. The proposed method is carried out without off-line identification of the intermediate parameters, and the algorithm is simple so as to avoid the enlargement of certain intermediate errors during the recursion process.4. The aforementioned new closed-loop subspace identification method by means of instrumental variables is applied to model one chemical process in the real world. The results illustrate the effectiveness of the method which can provide the state space model of the system for the future implementation of advanced control.Finally a summary is made, and the perspective of future studies is referred to at the end of the dissertation.
Keywords/Search Tags:system identification, subspace identification method, closed-loop identifi-cation, state space, recursive algorithm, EIV(errors-in-variables) model
PDF Full Text Request
Related items