The Research Of Sparse Spike Deconvolution And Seismic Inversion

Post-stack deconvolution and impedance inversion are significant techniques for improving the vertical resolution of seismic record and predicting reservoir.The basic principle of deconvolution is to eliminate the effect of wavelet from seismic traces and restore the true reflection coefficient sequence as much as possible.Impedance inversion refers to the conversion of post-stack seismic profile to impedance profile,so that the interpreter can directly compare with the well logging data and intuitively understand impedance change in the underground.We usually use spike deconvolution to enhance seismic resolution.However,spike deconvolution is based on the wavelet is minimum phase and reflection coefficient is white noise.We can use autocorrelation of the seismic trace to obtain the filter operator instead of wavelet autocorrelation to compress the wavelet and improve seismic resolution.However,the actual seismic wavelet is often a mixed-phase wavelet,and the reflection coefficient does not satisfy the hypothesis of white noise.In a bid to solve above problem,scholars proposed regularized constrained sparse-spike deconvolution.Sparse-spike deconvolution is based on the assumption that reflection coefficient is sparse,then a reflection sequence composed of some large pulses can be obtained.This result is advantageous for highlighting the interfaces where the impedance changes drastically,and can effectively improve the accuracy of recursive inverse.Recursive inversion is based on deconvolution and uses the relationship between reflection coefficient and impedance to estimate impedance value.In this paper,we first use the results obtained by conventional pulse deconvolution to perform recursive inversion.It is found that the results of spike deconvolution are far away from true reflection coefficient.The method is sensitive to stochastic noise and has error accumulative effect,which lead to bad inversion result.Then,we analyse thesparse spike deconvolution from the aspect of noise level,wavelet accuracy,and thin layer thickness.The practicability of the method is tested by actual data processing.Model-based impedance inversion is the most commonly used inversion method currently.It and recursive inversion are two different development directions of impedance inversion method.Each has its own characteristics and advantages in different situations.In this paper,we first deduces the formula of the model-based impedance inversion method.In addition to the basic fitting error term in inversion objective function,the low-frequency background information obtained from logging data is added to the objective function to supplement the low frequency components of the inversion result,then we use TV constraint improves anti-noise performance of the inversion method and protect the impedance interfaces.Then the practicability of the inversion method is verified by model and real data.Whether sparse constraint deconvolution or seismic inversion,relatively accurate seismic wavelet need to be provided before inversion.At present,there are many wavelet extraction algorithms.However,most wavelet extraction methods are cumbersome and require a lot of work.In addition,many methods are based on strictly conditions.In this paper,seismic wavelet convolution matrix is expressed newly by Toeplitz-sparse matrix factorization.The decomposed wavelet convolution matrix can be used as a constraint in objective function.The modified objective function can be inversed simultaneously reflection coefficient sequence(or impedance)and seismic wavelet.Through the model tests and actual data processing,the modified sparse-spike deconvolution and impedance inversion methods can obtain stable and reliable results,which are consistent with the log curve basically.