The conventional seismic deconvolution methods almost assume that the seismogram is stationary, and the seismic wavelet is of zero-phase or minimum-phase. However, the actual seismic data can't meet the above assumptions. Thus, this paper proposes a double deconvolution method in time-frequency domain to improve the resolution of the non-stationary seismogram. Firstly, quadratic spectrum modeling method combined with higher order bispectrum method is used for the non-stationary seismogram to extract a single mixed-phase wavelet, and the first deconvolution is implemented by applying the spectrum division method for the entire of seismogram; then quadratic spectrum modeling method combined with higher order bispectrum in time-frequency domain is used to extract time-variant mixed-phase wavelet, and the second deconvolution is implemented by applying spectrum division method on every point spectrum in the time-frequency domain. Simulation experiments prove that the proposed method effectively overcomes the interference of adjacent strata, and greatly improve the resolution of the non-stationary seismogram.

Traditional wavelet extraction methods are generally based on the hypothesis of time-invariant seismic wavelet with zero or minimum phase. Additionally, the evaluation of wavelet precision is hard to be performed directly. This paper presents a time-varying mixed-phase wavelet estimation method based on adaptive segmentation, in which quadratic spectrum modeling and higher-order cumulants double spectroscopy method are implemented to extract the amplitude and phase of segmented wavelets. Combined with the evaluation of deconvolution results using the estimated wavelets, the method can give a secondary evaluation of wavelet precision. Compared with the mixed-phase wavelet extraction method based on high-frequency attenuation compensation and zero-phase wavelet extraction method based on adaptive segmentation, simulation experiment results proved the correctness and superiority of the proposed method.

No abstract

Abstract-To adjust to the development of dynamic prediction of oil and gas reservoir, exploration of subtle reservoir, exploration and development of thin interbeds, processed seismic profiles are required to be of high signal to noise ratio, high resolution and high fidelity. While the time-varying wavelet in seismic data is an important factor affecting the quality of seismic profiles, for the purpose of improving the effects of seismic data processing, the influence law of wavelet time-varying nature is necessarily to be analyzed and researched, so that the timevarying wavelet can be suppressed based on more targeted methods. In the paper, the formation mechanism of time-varying nature is studied, and the influence law of time-varying wavelet is analyzed in detail, which is different in various links of seismic data processing, including amplitude compensation, pre-stack deconvolution, post-stack deconvolution and migration. Meanwhile the methods and application effects for suppressing wavelet time-varying nature in every link are summarized, and the advantages and disadvantages of these approaches are concluded through analysis and comparison, so that the development direction of researching more effective methods for suppressing or eliminating wavelet time-varying nature in the future is pointed out.