Study On Time-frequency Domain Deconvolution Method For Coalfield Seismic Data

Nowadays,it is difficult to identify the complex geological structure,small-scale geological body in the coalfield seismic exploration,which makes the problem that the resolution of coalfield seismic can not meet the actual production prominent.Because of the viscoelastic attenuation character of strata,seismic wavelet will attenuate with seismic propagation,and the seismic signal is actually non-stationary signal.The deconvolution method based on stationary seismic signal assumption is difficult to solve the problem of resolution reduction caused by the seismic wavelet attenuation.The time-frequency domain deconvolution method,which fully considers the viscoelastic attenuation character of the strata,can effectively recover the reflection and improve the seismic resolution.This paper focuses on the key contents of time-frequency domain deconvolution: time-frequency analysis methods,time-varying wavelet extraction method,and deconvolution operator design and so on.Time-frequency analysis method is the basis of the time-frequency domain deconvolution.It is important to study the time-frequency analysis method.In this paper,the time-frequency resolution,phase characteristics and signal reconstruction ability of short-time Fourier transform(STFT),Gabor transfrom,continuous wavelet transform(CWT),S-transform and generalized S-transform are compared.In terms of timefrequency resolution,the CWT,S-transform and generalized S-transform are multiresolution analysis methods,and the time-frequency resolution is higher than the STFT and Gabor transform.In terms of phase characteristics,the STFT,Gabor transform,S-transform and generalized S-transform have Fourier basis function,and there is a unified reference for phase information.Based on the complex signal analysis technology,the CWT can obtain the phase information,but there is no uniform reference for phase information.In terms of signal reconstruction,the STFT,Gabor transform,CWT,Sransform and generalized S-transform have good signal reconstruction ability.Therefore,the S-transform and generalized S-transform take into account the advantages of the STFT,Gabor transform and CWT.However,when using the S-transform and generalized Stransform to analyze low-frequency signals,there is a common problem of low time resolution.In order to solve this problem,an improved generalized S-transform(IGST)is proposed by redefining Gauss window function in this paper.The theoretical derivation shows that the improved generalized S-transform has the advantages of lossless reversibility and unified phase reference.The improved generalized S-transform enhances the time resolution of the low-frequency signals while ensuring the timefrequency resolution of the high-frequency signals,which has better performance than the S-transform and generalized S-transform.Baed on the improved generalized S-transform,the spectrum modeling method to extract time-varying wavelet in logarithm time-frequency spectrum is proposed,and then an improved time-frequency domain(ITFD)deconvolution method is established.Firstly,seismic traces are decomposed to time-frequency component by IGST,and converted it to logarithm time-frequency domain.Secondly,based on the trust region algorithm,the time-frequency spectrum of time-varying wavelet is extracted at each time of the logarithm time-frequency spectrum using the high-order Fourier series fitting,which can overcome the influence of white noise hypothesis of reflection coefficient.Thirdly,the boxcar smoothing method and the hyperbolic smoothing method are used to smooth the time-frequency amplitude spectrum of time-varying wavelet,which can eliminate the instability of the extraction results of time-frequency amplitude spectrum of time-varying wavelet.Finally,the Hilbert transform is used to estimate the time-frequency phase spectrum of time-varying wavelet,and then the time-varying wavelet is extracted completely.Compared with the time-frequency amplitude spectrum of real wavelet,timefrequency amplitude spectrum of Gabor deconvolution shows that the time-varying wavelet amplitude spectrum extraction method proposed in this paper has better perfomance.In this paper,a reasonable deconvolution operator is designed according to the time-varying characteristics of time-varying wavelet.Thus,a new time-frequency domain deconvolution method is constructed in this paper.The new time-frequency domain deconvolution method has better performance than least square deconvolution and Gabor deconvolution,and the reflection coefficient is consistent with the actual reflection coefficient.Combining the generalized standard linear solid and differential evolution algorithm,an approximate constant Q-value visco-acoustic wave equation is established.The high order staggered grid finite difference method is used to solve the approximate constant Q-value visco-acoustic wave equation.The forward modeling of elastic and viscoacoustic wave fields is carry out for the geological models of coal measure strata,faults and collapse columns.Compared with the elastic acoustic wave field,the energy of the reflected wave in the visco-acoustic wave field has the characteristics of attenuation,and the events of the reflected wave becomes wider.In order to verify the effectiveness of the time-frequency domain deconvolution method,the least-square deconvolution,Gabor deconvolution and ITFD deconvolution are used to process visco-acoustic seismic section.The results show that all three methods can effectively improve the resolution of reflection events.However,the least-square deconvolution can not compensate the attenuation of reflected wave energy caused by viscoelastic attenuation of medium,which makes it difficult to identify the deep reflection events.The Gabor deconvolution can compensate the reflection wave energy attenuation caused by viscoelastic attenuation of medium.However,the effect of time-varying wavelet extraction in Gabor deconvolution is poor.The residual wavelet reduces the signal-to-noise ratio of the seismic section.The ITFD deconvolution can not only compensate the reflection wave energy attenuation caused by viscoelastic attenuation of medium,bu also broaden the effective frequency band of seimsic,and the imaging quality is better than that of the least-square deconvolution and Gabor deconvolution.The improved generalized S-transform is used to analyze the seismic section.The results show that the ITFD deconvolution can improve the seismic resolution,broaden the seismic frequency band,and improve the recognition ability of the breakpoint and section.In order to eliminate the amplitude energy difference between seismic traces,L2 norm of standard traces is used to balance the energy and improve the continuity of reflection events.In field seismic data processing,if the time-varying wavelet extracted from standard seismic trace is used for time-frequency domain deconvolution,the continuity of seismic reflection events will be affected.The time-varying wavelet extraction method based on L2 norm can effectively balance the energy differences between traces and improve the continuity of reflection events.In addition,multi-channel weighted smoothing is used to improve the anti-noise ability of the proposed method in this paper.Thus the processing flow of time-frequency domain deconvolution for the field seismic data is established.The field seismic data verify the rationality and feasibility of the processing flow,which can effectively compensate the deep reflected wave energy and improve the vertical resolution of the coal field seismic data.