Abstract: An approach to identifying low frequency oscillation modes based on the stochastic response signals of the system is presented. First, the free oscillation signals are extracted from the stochastic response signals of the system by the natural excitation technique（ NEx T）. Then the mode parameters of the obtained free oscillation signals are identified by the total least squares-estimation of signal parameters via rotational invariance techniques（ TLS-ESPRIT） algorithm. Finally, the frequencies and damping ratios of low frequency oscillations are obtained. The effectiveness of the proposed method has been verified by the simulation data from a 16-machine system and the measurement data in Sichuan power grid. Compared with the random decrement technique（ RDT） and Prony method, the proposed method is more accurate in identifying the damping ratios and has a better noise immunity performance.