In order to improve the identification accuracy of sub-synchronous oscillation parameters in power system, aiming at the problems of strong noise interference in the sampled signals of sub-synchronous oscillation and difficulty in determining the system modal order, a sub-synchronous oscillation identification method combining fast independent component analysis (FastICA) technology and improved stochastic subspace of accurate modal order determination is proposed. Firstly, the FastICA technology is used to separate the noise from the original signal, and then the processed signal is input into the stochastic subspace algorithm as a new signal. Aiming at the problem that it is difficult to determine the order of the system modal in the stochastic subspace algorithm, the accurate modal order determination method is improved to avoid the phenomenon of false modal and modal omission, and finally the main modal is determined quickly, and the modal parameters of the signal is accurately identified. Through further analysis and verification of the ideal calculation example, and the wind farm standard model and the measured data of power grid, the results show that the method can effectively reduce noise interference, shorten the identification time, accurately identify the main modal of the signal, and improve the identification accuracy.