Parameters are often identified by using Prony algorithm,but this method is sensitive to the noise of signals and has a high demand to the input signal.Therefore,the paper proposes a algorithm for identifying power system low frequency oscillation,which combines FastICA(Fast Independent Component Analysis) and Prony algorithm.First of all, wide area measurement signal is used as the input signal.And the preprocessing signal is denoised by ICA algorithm.In the end,the parameters of low frequency oscillation in the power system are obtained by using Prony algorithm,which analysis the denoised signal.Through analyzing the case studies on the ideal signal and four-machine system,it was found that the method improve the ability of accuracy,rapidity and anti-noise after the signal is denoised to identify the parameter of low frequency oscillation.