Seismic Tomography Velocity Inversion Based On Image Domain Moveout Curvature Intelligent Recognition

With the development of Chinese economy,the dependency on foreign petroleum import increases year by year,which brings many risks and challenges to the develop-ment of national economy and security.It is of great practical significance to develop oil and gas exploration technology and improve self-supply ability of oil and gas to guar-antee economic security and social stability.As the geological structure of oil and gas exploration block becomes more and more complex,the exploration of petroleum en-counters challenging.How to improve the imaging accuracy is one of the key problems to be solved.In the process of seismic data processing,there are two ways to improve the imaging accuracy:one is to optimize imaging algorithm,the other is to improve the velocity model accuracy for imaging.the thesis is based on elastic wave equation and pattern recognition techniques from the above two aspects to improve,respectively.One is to introduce elastic inverse scattering theory into the elastic wave equation imaging al-gorithm to improve the accuracy of elastic imaging algorithm.The other is to invert the velocity model in time domain and depth domain by minimizing image domain moveout based on residual moveout curvature intelligent recognition technology,which improve the imaging accuracy.The details are as follows:Firstly,aiming at the problem that traditional elastic vector imaging condition is difficult to describe reflection coefficients,by introducing elastic inverse scattering into migration imaging,this paper proposes elastic wave equation inverse scattering migration imaging method to invert reflection coefficients.Based on the elastic Born approximation,the linear expression of reflection coefficients of seismic data is derived.On the one hand,to improve the imaging efficiency,the approximate estimation of reflection coefficients is achieved by using adjoint operator instead of large sparse matrix inversion.On the other hand,to improve the imaging accuracy,the least square method is used to realize the unbiased reflection coefficient estimation,and the uncertainty analysis of the inverse problem is further given.Numerical experiments show that reflected wave simulation effect is good and reverse time migration method have higher imaging accuracy.Secondly,to solve the problem of heavy manual picking for many times in the pro-cess of time domain migration velocity analysis,based on the idea of pattern detection and pattern recognition,this paper proposes a tomography velocity direct inversion method based on residual moveout curvature intelligent recognition.In this method,the moveout curvature normalization operator of redundant images is derived to transform the itera-tive time domain velocity modeling into the moveout recognition of redundant images.To automatically capture the moveout information,a hyperbolic moveout detection and curvature recognizer based on pattern recognition is designed to map normalized move-out in the redundant images to ratio-errors and update parameter.Finally,the intelligent inversion framework and the uncertainty analysis of inverse problem are further given.Numerical experiments show that the proposed method can realize automatical and direct velocity inversion.It effectively reduces labor cost and improves seismic time-migration imaging quality.Thirdly,focusing on the problem of the strong dependence of elastic wave equa-tion depth-imaging on velocity model for complex situation,the elastic wave equation tomography velocity inversion method is proposed based on residual moveout（depth er-ror）backpropagation.The residual moveout between the redundant images is measured byl₂norm.Combined with the inner product kernel function with respect to velocity in the elastic Born approximation,the gradient of the target functional with respect to P-and S-wave velocity is deduced by adjoint state method.The intelligent inversion frame-work and uncertainty analysis are further given.Numerical experiments show that the proposed method can effectively estimate the P-wave and S-wave velocity in the case of complex models.And the inverted velocity produces coherently superimposed images and improves quality of seismic depth-imaging.