CN-121995468-A - Seismic data processing method and device based on multiple types of priori information

CN121995468ACN 121995468 ACN121995468 ACN 121995468ACN-121995468-A

Abstract

The invention provides a seismic data processing method and device based on multiple types of priori information, and belongs to the technical field of oil-gas seismic exploration. The method comprises the steps of constructing an earthquake sparse representation objective function coupled with multiple types of priori information based on original earthquake data to be processed and corresponding well logging data, wherein the multiple types of priori information comprise low-frequency impedance priori information of the well logging data, earthquake amplitude envelope priori information of the original earthquake data and supporting position priori information, solving a maximum posterior probability density solution for the earthquake sparse representation objective function by utilizing a three-dimensional multi-channel Bayesian orthogonal matching pursuit algorithm, wherein the maximum posterior probability density solution comprises a target reflection coefficient sequence of the original earthquake data, acquiring a target broadband Ricker wavelet, and reconstructing the earthquake data based on the target broadband Ricker wavelet and the target reflection coefficient sequence to obtain the target earthquake data. By adopting the method and the device, the seismic data can be reconstructed by coupling multiple types of priori information, and the resolution of the seismic data is improved.

Inventors

LI KUN
YUAN JIN
GE WEI
YIN XINGYAO
ZHENG QINGWEN

Assignees

中国石油大学(华东)

Dates

Publication Date: 20260508
Application Date: 20260407

Claims (10)

1. A method of seismic data processing based on multiple types of prior information, the method comprising: Constructing an earthquake sparse representation objective function coupled with multiple types of priori information based on original earthquake data to be processed and corresponding well logging data, wherein the multiple types of priori information comprise low-frequency impedance priori information of the well logging data, earthquake amplitude envelope priori information of the original earthquake data and supporting position priori information; Solving a maximum posterior probability density solution for the seismic sparse representation objective function by using a three-dimensional multi-channel Bayesian orthogonal matching pursuit algorithm, wherein the maximum posterior probability density solution comprises a target reflection coefficient sequence of the original seismic data; Acquiring a target broadband Ricker wavelet, and reconstructing the seismic data based on the target broadband Ricker wavelet and the target reflection coefficient sequence to obtain target seismic data.
2. The method of claim 1, wherein constructing the seismic sparse representation objective function coupled to multiple types of prior knowledge based on the raw seismic data to be processed and the corresponding well log data comprises: constructing a target reflection coefficient sequence to be solved based on the original seismic data to be processed The following formula (1): (1) in the formula, Representing the raw seismic data in three dimensions, Representing a three-dimensional sequence of seismic reflection coefficients, The representation of an atomic dictionary is given, Is the L0 norm regularized weight coefficient, An L0 norm representing the three-dimensional seismic reflection coefficient sequence; Wherein, the Expressed by the following formula (2): (2) in the formula, Represents the number of atoms in the atomic dictionary, Is indicated in the position Sparse representation coefficients of the ith atom in the seismic trace, , Representing the seismic reflection coefficient of the i-th atom, Is the i-th atom in the dictionary of atoms, Is Gaussian random noise with the average value of 0; based on the above equation (1) and equation (2), a seismic data likelihood function is constructed The following formula (3): (3) in the formula, Mean value 0 and covariance 0 Is a gaussian distribution of (c); constructing low-frequency impedance priori information based on logging data corresponding to the original seismic data Likelihood function of (2) The following formula (4): (4) in the formula, In the form of an integration matrix, Covariance matrix of low-frequency impedance in logging data; constructing a three-dimensional seismic reflection coefficient sequence based on the original seismic data and the logging data Is defined as the following equation (5): (5) in the formula, Is the standard deviation of the seismic reflection coefficient of the i-th atom, Is the standard deviation of the reflection coefficient of the log data estimated based on the generalized gaussian distribution at the ith atomic position, Is the absolute value of the seismic reflection amplitude at the ith sample point; Constructing a sparse representation coefficient sequence based on the seismic amplitude envelope priori information and the support position priori information of the original seismic data Is a priori probability density distribution of (2) The following formula (6): (6) in the formula, Representing compliance parameters as Is characterized by a Bernoulli distribution, Representing the probability that the i-th atom in the atom dictionary is selected, The following formula (7): (7) in the formula, For the normalized coefficient of the probability, Representing the raw seismic data Is a hilbert transform of (c); combining the formulas (3) to (7), constructing a seismic sparse representation objective function coupled with multi-type priori knowledge, and adopting the following formula (8): (8) in the formula, And representing the target sparse representation coefficient sequence to be solved.
3. The method of claim 2, wherein said solving the maximum posterior probability density solution for the seismic sparse representation objective function using a three-dimensional multi-pass bayesian orthogonal matching pursuit algorithm comprises: assuming that the total iteration number of the three-dimensional multi-channel Bayesian orthogonal matching pursuit algorithm is M, constructing a target sparse representation coefficient of an nth iteration and an ith sampling point based on a formula (8) Index i * of the optimal atomic position for the nth iteration, and the target reflection coefficient sequence for the nth iteration ; Target sparse representation coefficient of nth iteration and ith sampling point The following formula (9): (9) in the formula, Seismic data residual after the n-1 th iteration ; Index i * of matching atomic positions for the nth iteration, formula (10) below: (10) in the formula, the target reflection coefficient sequence of the nth iteration Seismic reflection coefficient of the ith matching atom in (b) , Regularization parameters used for measuring prior information of the atomic positions; Target reflection coefficient sequence of nth iteration The following formula (11): (11) in the formula, , To match a set of atoms over n iterations, , , Regularized weight coefficients representing low frequency impedance prior information in the log data; screening out matching atoms based on the formulas (9), (10) and (11) in each iteration by utilizing a three-dimensional multi-channel Bayesian orthogonal matching pursuit algorithm, and combining and updating a target sparse representation coefficient sequence and a target reflection coefficient sequence until M iterations or seismic data residual errors are performed And (5) obtaining the maximum posterior probability density solution when the maximum posterior probability density solution is smaller than a preset iteration threshold.
4. The method of claim 1, wherein the obtaining the target wideband Ricker wavelet comprises: And optimizing the frequency integral boundary and the phase correction quantity of the basic broadband Ricker wavelet by using a particle swarm algorithm to obtain a target broadband Ricker wavelet, so that the target broadband Ricker wavelet is matched with the real reflection characteristics of the stratum and the actual seismic propagation rule.
5. The method of claim 4, wherein optimizing the frequency integration boundaries and phase corrections of the underlying wideband Ricker wavelet using a particle swarm algorithm to obtain the target wideband Ricker wavelet comprises: based on the integral of the conventional band-limited Ricker wavelet, a basic broadband Ricker wavelet is constructed, as shown in the following formulas (12) and (13): (12) (13) in the formula, And Respectively represent the integral boundaries of the center frequencies of conventional band-limited Ricker wavelets, wherein, Setting the frequency as a preset frequency; based on equation (12) and equation (13), a complex domain Ricker wavelet is constructed as follows equation (14): (14) in the formula, Is the correction phase of the complex domain Ricker wavelet, Representing the real part of the complex domain Ricker wavelet, Representing a hilbert transform operator; constructing integral boundaries for target frequencies based on correlation of parawell location seismic data and logging data And a target phase correction amount Is represented by the following formula (15): (15) in the formula, Representing the phase corrected target seismic data, , Refers to the sequence of target reflection coefficients, Is a sequence of formation reflection coefficients calculated based on the log data, Representing actual seismic data of the well bypass, Pearson correlation coefficients representing the target seismic data and a sequence of formation reflection coefficients calculated based on the log data, Pearson correlation coefficients representing the target seismic data and the actual seismic data of the well bypass, And Is a regularized weight coefficient; solving the target functional by using a particle swarm algorithm to obtain a target frequency integral boundary And a target phase correction amount ; Integrating boundaries based on the target frequency The target phase correction amount And correcting the basic broadband Ricker wavelet to obtain a corrected target broadband Ricker wavelet.
6. The method of claim 1, wherein the reconstructing the seismic data based on the target broadband Ricker wavelet and the target reflection coefficient sequence to obtain target seismic data comprises: and performing time domain convolution operation based on the target broadband Ricker wavelet and the target reflection coefficient sequence to realize high-resolution seismic data reconstruction and obtain target seismic data.
7. A seismic data processing device based on multiple types of prior information, the device comprising: The system comprises an objective function construction unit, a seismic sparse representation unit and a support position analysis unit, wherein the objective function construction unit is used for constructing an earthquake sparse representation objective function coupled with multiple types of prior information based on original earthquake data to be processed and corresponding well logging data, wherein the multiple types of prior information comprise low-frequency impedance prior information of the well logging data, earthquake amplitude envelope prior information of the original earthquake data and support position prior information; The solving unit is used for solving a maximum posterior probability density solution for the seismic sparse representation objective function by utilizing a three-dimensional multi-channel Bayesian orthogonal matching pursuit algorithm, wherein the maximum posterior probability density solution comprises a target reflection coefficient sequence of the original seismic data; The reconstruction unit is used for acquiring a target broadband Ricker wavelet, and reconstructing the seismic data based on the target broadband Ricker wavelet and the target reflection coefficient sequence to obtain target seismic data.
8. The apparatus of claim 7, wherein the reconstruction unit is configured to: And optimizing the frequency integral boundary and the phase correction quantity of the basic broadband Ricker wavelet by using a particle swarm algorithm to obtain a target broadband Ricker wavelet, so that the target broadband Ricker wavelet is matched with the real reflection characteristics of the stratum and the actual seismic propagation rule.
9. An electronic device, the electronic device comprising: processor, and A memory in which a program is stored, Wherein the program comprises instructions which, when executed by the processor, cause the processor to perform the method according to any of claims 1-6.
10. A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform the method of any one of claims 1-6.

Description

Seismic data processing method and device based on multiple types of priori information Technical Field The invention relates to the technical field of oil-gas seismic exploration, in particular to a seismic data processing method and device based on multiple types of priori information. Background Along with the continuous improvement of the oil and gas exploration degree, the exploration target gradually changes from a conventionally constructed oil and gas reservoir to a complex and hidden lithologic oil and gas reservoir, and the requirements on the resolution of seismic data are also higher and higher. Aiming at the background of oil-gas seismic exploration of complex earth surface, complex structure and complex reservoir, the resolution of conventional band-limited seismic data is difficult to meet the urgent requirements of fine characterization of thin interbes, tiny breaks and complex lithology traps. Therefore, widening the seismic frequency band and improving the resolution of the seismic data are key to improving the accuracy of reservoir description and quantitative interpretation. In recent years, expert scholars at home and abroad research various high-resolution seismic processing methods, and main stream methods are classified into three main categories, namely deconvolution and anti-Q filtering, compressed sensing and wavelet shaping, and deep learning super-resolution processing technology. Deconvolution and anti-Q filtering mainly improve the resolution of seismic data through inversion compression wavelet and seismic wave absorption attenuation compensation. The deep learning super-resolution processing method is to directly learn the mapping relation from low-resolution data to high-resolution data through graphic image processing and artificial intelligence technology, so as to realize high-definition processing of the seismic image. The compressed sensing and wavelet shaping technology mainly realizes high-resolution seismic data reconstruction through correction of broadband Ricker wavelets or broadband trapezoidal wavelets on the premise of seismic data sparse representation. The matching pursuit algorithm based on classical Morlet wavelet and Ricker wavelet plays an important role in time-frequency decomposition of non-stationary signals, regularization of seismic data, high-resolution processing and inversion, denoising and other aspects. The matching pursuit-based seismic high-resolution processing technology is one of the most mainstream methods at present. Matching pursuit algorithms based on greedy search strategies generally face the problems of low computational efficiency, horizontal continuous difference and strong wavelet dictionary dependence, and these characteristics bring many challenges to matching pursuit high-resolution processing. Expert students have mainly solved these problems by developing multi-pass regularized matching pursuit algorithms, fast matching pursuit algorithms, and dynamic dictionary learning algorithms. However, the matching pursuit high-resolution seismic processing method at the present stage is usually mainly driven by data, ignores instantaneous amplitude information contained in seismic data, lacks consideration of high-frequency reflectivity and low-frequency priori information contained in logging data, lacks quantitative optimization mechanism in an effective frequency band range, and has yet to be enhanced in the fidelity of high-resolution seismic data processing results. Therefore, there is a need to develop high-fidelity seismic data sparse representation and high-resolution data processing techniques that couple multiple types of prior knowledge. Disclosure of Invention In view of the above, the embodiment of the invention provides a seismic data processing method and device based on multiple types of prior information, which can be coupled with the multiple types of prior information to reconstruct seismic data and improve the resolution of the seismic data. According to an aspect of the present invention, there is provided a seismic data processing method based on multiple types of prior information, the method comprising: Constructing an earthquake sparse representation objective function coupled with multiple types of priori information based on original earthquake data to be processed and corresponding well logging data, wherein the multiple types of priori information comprise low-frequency impedance priori information of the well logging data, earthquake amplitude envelope priori information of the original earthquake data and supporting position priori information; Solving a maximum posterior probability density solution for the seismic sparse representation objective function by using a three-dimensional multi-channel Bayesian orthogonal matching pursuit algorithm, wherein the maximum posterior probability density solution comprises a target reflection coefficient sequence of the original seismic data; Acquiring a target br