Three-dimensional longitudinal wave impedance full-waveform inversion method and apparatus

Abstract

The invention discloses a three-dimensional longitudinal wave impedance full-waveform inversion method and an apparatus. The method comprises the following steps of determining an adjoining equation, longitudinal wave speed gradient, density gradient and longitudinal wave impedance gradient calculating formulas corresponding to a time-domain three-dimensional first-order speed-stress variable density sound wave equation; determining a forward-propagating wave field and a residual reverse back-propagating wave field by adopting a high-order staggered-mesh finite difference method according to the time-domain three-dimensional first-order speed-stress variable density sound wave equation and the corresponding adjoining equation; determining the longitudinal wave speed gradient, the density gradient and the longitudinal wave impedance gradient according to the forward-propagating wave field, the residual reverse back-propagating wave field, and the longitudinal wave speed gradient, density gradient and longitudinal wave impedance gradient calculating formulas; and updating a parameter model by adopting a local optimization algorithm according to the longitudinal wave speed gradient, the density gradient and the longitudinal wave impedance gradient to output a three-dimensional longitudinal wave impedance full-waveform inversion result satisfying a convergence condition. The feasibility of an inversion process can be improved. Great numerical dispersion suppressing capabilities can be further achieved.

Description

Technical field [0001] The invention relates to the technical field of seismic elastic parameter inversion for petroleum exploration, in particular to a three-dimensional longitudinal wave impedance full waveform inversion method and device. Background technique [0002] Traditional elastic parameter inversion methods, such as the elastic impedance inversion method based on the Zeoppritz equation and the AVO (Amplitude Versus Offset, amplitude change with offset) type inversion method, mainly have the following problems: First, it is assumed that the plane wave is incident The limitation is that the assumption of the reflection amplitude of pre-stack seismic data based on the convolution model is not suitable for the situation of large angle incidence; second, it is difficult to obtain a reliable low-frequency model when the well is sparse. These problems make elastic impedance inversion methods and AVO-type inversion methods unsuitable for detailed description of complex litholo...

AI Technical Summary

Problems solved by technology

[0002] Traditional elastic parameter inversion methods, such as the elastic impedance inversion method based on the Zeoppritz equation and the AVO (Amplitude Versus Offset, amplitude variation with offset) type inversion methods, mainly have the following problems: First, they are limited by the plane wave incident assumption. limitations, the assumption of the reflection amplitude of pre-stack seismic data based on the convolution model is not suitable for the case of large-angle incidence; secondly, it is difficult to obtain a reliable low-frequency model in the case of sparse wells

[0005] It can be seen that since P-wave velocity and density are coupled with each other, simultaneous P-wave velocity and density inversion is a more feasible inversion strategy. However, Jeong et al., Prieux et al. and Yang Jizhong et al. used the second-order scalar frequency domain variable density acoustic wave equation, However, the memory requirement for forward modeling in the frequency domain is huge, especially for the LU matrix (in linear algebra, it can be decomposed into a matrix product of a lower triangular matrix and an upper triangular matrix, where L and U are lower triangular and upper triangular matrices, respectively). The memory requirement is extremely high, and it is even difficult to realize; at the same time, the second-order scalar acoustic wave equation is usually solved by the high-order central grid finite difference method. To reduce the influence of numerical dispersion, it is necessary to increase the finite difference order, which will increase the amount of calculation.

Images