Aiming at the problem of accurate identification of subsynchronous oscillation (SSO) signals, a multi-channel SSO identification and early warning method is proposed by combining symplectic geometry mode decomposition (SGMD) improved with dynamic time warping (DTW) and second-order blind identification (SOBI). First, SGMD is applied to the SSO signal, which is then subjected to diagonal averaging and adaptive reconstruction to obtain the initial symplectic geometric mode components (ISGMCs). The optimal distance between ISGMCs is calculated by DTW algorithm to measure the similarity of sequences, and the symplectic geometry components (SGCs) with independent modes are adaptively selected. Next, the dominant SGCs are used as observation signals and input into the SOBI algorithm. By performing joint approximate diagonalization on the observation matrix, the complete SSO source estimation signals are obtained. The least square method is introduced to improve the SOBI algorithm, enabling direct identification of the SSO oscillation frequency and attenuation factor. Finally, through comparative analysis of ideal and simulation examples, it is verified that the proposed algorithm can accurately and efficiently identify multi-channel subsynchronous oscillation signals.