The method proposed in this paper aims to detect parameters of oscillation and locate the fault. Oscillation is often superimposed on the fundamental frequency signal for its randomness, and EMD can separate it, making Kalman state-space model more simply and reliable. When establishing the Kalman state space, the use of Z-transform makes its state updates easier. The proposed method can measure frequency and damping factor of oscillation precisely and quickly. Finally, the sample values of the starting point of the oscillation helps to determine the maximum amplitude of the oscillation. Kalman filter allows some degree of noise, which can reduce influence of noise effectively. The simulation results can be seen that the method can detect oscillation parameter efficiently and locate the oscillation range, and all of its error is acceptable. Compared with several oscillation detecting methods, the algorithm proposed in this paper was proved to be advantageous.