Subsynchronous Oscillation Parameters Identification Method for DFIG Based on NGO‐Optimized MVMD and Improved TLS‐ESPRIT

To overcome the issues of mode mixing, low identification accuracy, and poor noise robustness in parameters identification of subsynchronous oscillation (SSO) in doubly fed induction generator (DFIG), an improved parameters identification method combining multivariate variational mode decomposition (MVMD) and total least squares estimation of signal parameters via rotational invariance techniques (TLS‐ESPRIT) is proposed. Initially, the grid‐connected current signal of DFIG is decomposed using MVMD. During the decomposition process, the northern goshawk optimization (NGO) algorithm is employed to select the number of modes K and penalty factor α . Correlation analysis is then performed to identify the dominant modes from the decomposed components. The dominant modes are resampled to construct the Hankel matrix, followed by singular value decomposition (SVD). The signal subspace order b is optimized based on singular value entropy (SVE), and TLS‐ESPRIT is subsequently employed to identify the parameters of the SSO components. Time‐domain simulations have been conducted using a composite signal and a 4‐machine 2‐area system incorporating a DFIG‐based wind farm, and the proposed method has also been validated using measured SSO data from the Guyuan wind farm in Hebei Province. The results show that, compared with the comparative decomposition methods, the proposed method improves data processing efficiency by more than 60%. For the composite signal, the minimum identification accuracies of amplitude, frequency, phase, and damping factor have increased by 8.50%, 9.04%, 8.56%, and 8.50%, respectively, compared with the corresponding conventional methods. The identification results obtained at different noise levels closely match the theoretical values, and the minimum identification accuracy remains above 96% even under 10 dB noise interference. These results demonstrate the accuracy and superiority of the proposed method for SSO parameters identification.