Prestack Seismic Inversion Based On Sparse Constraints

Prestack seismic inversion is an important method for reservoir prediction and fluid identification.As one of the approaches,AVO technique based on seismic data is able to spo t oil and gas effectively.Meanwhile,due to increasing complicated reservoirs that have to be dealt with,there is considerable meaning of guranteeing the accuracy of physical and lithological configurations estimation,which is possible through appropriate seismic inversion.Combining inversion equation and targeted parameters in the sense of rock physics,optimized inversion under Bayesian or regularization framework provides us a proper research prospect.Seismic data are digital signals coming from accquisition.This thesis first introduces some concepts and principles of signal sparse representation,illustrates the mathematical algorithm(regularization)of solving underdetermined problems and shows the connection between Beyasian theory and regularization.Wedge-dictionary based deconvolution is tested for thin layer recognition,which involves the wedge dictionary in wavelet matrix and solves the constructed underdetermined equation by the basis pursuit(BP)algorithm.Results show that this kind of reflection coefficient(RC)inversion is good at noise immunity and resolution.To get information of practical physical and lithological parameters underground,three-term AVO inversion method constrained by exponential aprior information is proposed.This method starts from obtaining sparse RC of physical and lithological parameters,thereby trace integration can extract some seismic attributes like velocity,density and impedance exhibiting clear strata figuration.On the other hand,inversion incorporates model constraints to make sure that vertical resolution does not harm lateral continuity.The model comes from horizon interpretation of seismic profile and 2D/3D interpolation by logging data.Furthermore,this thesis proposes RC approximation equation including fluid factor to get quantitative knowledge of reservoir fluid,frequency-dependent AVO inversion using inversion spectral decomposition and the new fluid factor—frequency dependent fluid term Df.This scheme considers attenuation and dispersion effects caused by pore fluid.Several approaches of time-frequency analysis/spectral decomposition are discussed to test the resolution of inversion spectral decomposition.Beyasion a priori constraints grant reasonability and stability of inversion results,and MPI implements parallel acceleration of the trace-by-trace inversion.Model test based on realistic rock physics parameters demonstrates that the frequency-dependent AVO approximation has good accuracy if the incident angle is not large,and single trace inversion supports its ability of distinguishing gas bearing rock from water bearing rock.Real data processes show that the feasibility of the frequency-dependent AVO inversion for oil and gas identification.