Frequency domain subspace identification of fractional order systems using time domain data with outliers

Abstract

AbstractThis paper focuses on the identification of the multiple‐input multiple‐output commensurate fractional order systems. Different from the assumption of known frequency domain data and prior information about noise in the general frequency domain identification algorithm, the reliable frequency domain data is measured in this paper by designing special excitation signals. Then, in order to suppress the impact of noise, the data used in algorithm are truncated from the low frequency band. In addition, this paper considers the case where the sampling data are disturbed by outliers, and a matrix decomposition method with threshold value is developed to eliminate the influence of outliers. After that, the parameters of the system to be identified are estimated by the frequency domain subspace identification method. The validity of the proposed method is demonstrated by an illustrative numerical example.