Abstract: With the increasing proportion of renewable generations in the power system, the problem of sub-synchronous oscillation (SSO) has become prominent. Accurate identification of SSO signals is the premise of identifying sources and mitigating SSO, which is crucial to the stable operation of power systems. In many SSO events, the oscillation frequency and amplitude are time-varying, making it difficult for existing methods to accurately identify SSO parameters. In this paper, an SSO detection method based on adaptive-reset extended Kalman filter (EKF) is proposed. Firstly, a four-state SSO signal model is constructed to enable the EKF algorithm to detect signal amplitude, frequency and attenuation coefficient, and an algorithm for identifying multi-mode SSO signal is designed. Secondly, an adaptive- reset method of EKF parameters based on the innovation criterion is proposed, and the accurate identification of time- varying SSO signals could be realized by adaptively resetting the covariance matrix. Finally, the proposed algorithm is verified by simulation and hardware experiment. The results showed that method proposed in this paper can accurately identify time-varying SSO modal parameters, and the algorithm has strong real-time performance and has high practical value.