Study On Seismic Data Inversion Methods In Depth Domain

Developed after several decades,seismic exploration technology has taken great changes,from 2D to 3D,from post-stack to pre-stack,from time domain to depth domain.As is known to all,prestack depth migration is an ideal technique for seismic data imaging in the areas with complex structures and severe lateral velocity variations.It has been a conventional method for 3D seismic data processing.However,the current seismic explanation and reservoir prediction are still applied in time domain only,which means we use the “depth-time-depth” process to transfer imaging data from depth domain to time domain,then carry out inversion in time domain and finally transfer inversion results to depth domain.This processing not only waste time but also result in accuracy loss caused by resampling.The depth domain inversion hasn't been fully developed because wavelet in depth domain changes with depth so that linear convolution doesn't work out.To clarify the formation mechanism and physical meaning of reflection wavefield in depth domain,basic theory of depth migration was elaborated in this paper,take Kirchhoff migration as an example of integral type depth migration and take reverse time migration as an example of difference type depth migration,waveform and amplitude attribute of seismic events in depth domain were analyzed.Based on generalized diffraction stacking theory,depth migration imaging resolution was analyzed quantitatively.All these provide a solid foundation for the development of direct inversion techniques in depth domain.Considering the fact that wavelet changes in depth domain,we introduce the concepts of time-varying wavelet and non-stationary deconvolution of time domain into depth domain and develop the basis pursuit seismic inversion method based on “depth domain generalized wavelet”.The core of the solution is to represent the response of reflection waveform in depth domain with “depth domain generalized wavelet”.“depth domain generalized wavelet” is obtained only through seismic trace basis pursuit spectral decomposition and point spectrum simulation,it is a equivalent operator essentially which reveal local waveform attribute of seismic trace in depth domain.It can be used to achieve high precision well-seismic calibration and establish waveform dictionary.During the inversion,we use basis pursuit algorithm again based on layer sparsity constraint to realize high resolution seismic inversion in depth domain.Considering the fact that “depth-time-depth” inversion process wastes time,error accumulation and accuracy loss of resampling,we integrate the whole shifting process and derive an analytic formula for synthetic seismic record in depth domain on the basis of time-domain wavelet and depth-domain model.Then the depth-domain Jacobian matrix is derived under the framework of generalized linear inversion based on the forward modeling mentioned above.Finally we conclude a seismic data broadband constraint generalized linear inversion in depth domain.This method is derived through the “depth-time-depth” process,however,it realizes the time-depth conversion and the unified representation for velocity variables among all the inversion parameters.Meanwhile,resampling of logging and seismic data can be eliminated by inserting local time-depth relationship into continuous source function sampling in time domain.With this method,we can get higher inversion accuracy and avoid dislocation of layers which derived from error accumulation in “depth-time-depth” process.Being concerned with transmission loss,internal multiples,wave mode transformation and such factors,we apply nonlinear prestack inversion in tau-p domain based on inverse scattering theory.Proceeded from the smooth low frequency background,the inversion process solve the data function and target function by alternative iteration.Contrast parameters are obtained by adopting multiplicative regularization method in the framework of conjugate gradient.Then we acquire the whole wave field using optimized scattering Neumann series and finally high resolution underground elastic parameters are built in depth domain.Therefore,the prestack full waveform inversion in depth domain based on inverse scattering theory has been developed.In essence,the key of both basis pursuit seismic inversion method based on “depth domain generalized wavelet” and generalized linear inversion method in depth domain is to realize quantitative representation of varying wavelets in depth domain,except the former in space-wavenumber domain and the latter in space-time domain,and both have advantages and disadvantages as well as scope of application.Although prestack full waveform inversion method based on inverse scattering theory has higher resolution,but compared with the other two inversion method,it has very low computational efficiency.Model and real data tests show that the inversion methods oriented to seismic direct interpretation and reservoir prediction studied in this paper have broad prospects in application.