Abstract: In order to solve the problem of sub-synchronous oscillation (SSO) generated by direct-drive permanent magnet synchronous wind turbines (D-PMSG), the multivariate variational mode decomposition (MVMD) method used in fault detection of mechanical fault detection is improved and combined with the second order blind identification (SOBI) to realize the subsynchronous oscillation mode identification of direct driven wind turbine. For wide area measurement system (WAMS) which has multi-channel characteristics, a method of multichannel synchronous oscillation mode identification is proposed. First, the mode number K and penalty factor α of MVMD have absolute influence on the accuracy of the algorithm, thus the MVMD algorithm is improved and a comprehensive index S_y is established to determine K and α. Secondly, On the basis of known parameters, the improved MVMD is used to decompose the SSO signal to obtain multiple intrinsic mode function components (IMFs), and the dominant IMFs are selected with the help of Fréchet distance to remove noise interference. At the same time, in order to improve the operation efficiency, the SSO signal modes are directly identified, the improved MVMD is used to decompose the SSO signal. Based on the idea of random subspace, the SOBI algorithm is improved to directly identify the frequency, damping ratio and attenuation factor of SSO signal. Finally, the proposed method is analyzed and verified by using ideal examples, simulation examples and measured data of power grid. The results show that the proposed method can be adopted to effectively and accurately identify the parameters of multi-channel sub-synchronous oscillation signals generated by direct-drive wind turbines, which lays a foundation for the research of sub-synchronous oscillation suppression.