In recent years, the safety of nuclear power plants in seismic events has received more and more attention, and automatic seismic shutdown has become one of the functions of protective shutdown in active nuclear power plants. However, how to accurately recognize the valid seismic data and reduce the false action rate of seismic shutdown has been a research difficulty that has not yet been broken through. In seismic data processing, how to effectively retain the weak effective signal while denoising and improve the data quality is the basis of seismic data analysis. According to the characteristics of random noise in seismic data, the shortcomings of using improved singular spectrum analysis (SSA) and double-tree complex wavelet block thresholding (BT-DTCWT) alone are analyzed, and combined with the advantages of the two, the joint denoising of improved SSA and BT-DTCWT is adopted. In the simulation analysis, facing the data with different signal-to-noise ratios, the proposed denoising algorithm is more stable and has a better denoising effect, and the overall signal-to-noise ratio is higher than that of the traditional wavelet thresholding method by more than 4 dB. For the actual seismic data, the denoising algorithm in this paper is compared and analyzed with the Butterworth filter, wavelet threshold denoising and other methods, and the results show that the filtering effect of the algorithm in this paper is more obvious and has better performance.