A Data-Driven Closed-Loop System Identification Approach Using Canonical Correlation Analysis

Abstract

In this paper, a data-driven closed-loop system identification approach using canonical correlation analysis (CCA) is proposed. Based on the subspace identification methods (SIM), an auxiliary matrix is constructed by utilizing known controller information in the structure of the closed-loop system. With the help of the projection, the equivalent relationship between past input and future output is found, and the optimal identification of system parameters is implemented by using the CCA method. The proposed method solves the biased identification problem caused by the correlation between closed-loop input and noise. The main results are tested and verified via the numerical example, and the superiority of the identification estimation results of the proposed method is compared through different methods.