Power System Low Frequency Oscillations Analysis Based on Adaptive ESPRIT method

Low frequency oscillation threatens the stability of the power system. The ESPRIT (estimation of signal parameters via rotational invariant techniques) can analyze low-frequency oscillations with high accuracy only when the model order of the analyzed signal is known. This paper proposes an adaptive method for the determination of the model order of the oscillation signal. In the proposed method, the adaptive threshold for the determination of the model order is based on the mean value of the normalization of singular values. As a result, the ESPRIT can accurately separate the signal subspace from the noise subspace. The proposed method is experimentally evaluated through simulations. Results show that the proposed method behaves much better anti-noise performance in parameter identification of low frequency signals compared with the Modified Extract Model Order (MEMO) method.