CN111856559B - Multi-channel seismic spectrum inversion method and system based on sparse Bayes learning theory - Google Patents

Multi-channel seismic spectrum inversion method and system based on sparse Bayes learning theory Download PDF

Info

Publication number
CN111856559B
CN111856559B CN201910362219.9A CN201910362219A CN111856559B CN 111856559 B CN111856559 B CN 111856559B CN 201910362219 A CN201910362219 A CN 201910362219A CN 111856559 B CN111856559 B CN 111856559B
Authority
CN
China
Prior art keywords
seismic
channel
spectrum
inversion
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910362219.9A
Other languages
Chinese (zh)
Other versions
CN111856559A (en
Inventor
袁成
苏明军
李政阳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Petrochina Co Ltd
Original Assignee
Petrochina Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Petrochina Co Ltd filed Critical Petrochina Co Ltd
Priority to CN201910362219.9A priority Critical patent/CN111856559B/en
Publication of CN111856559A publication Critical patent/CN111856559A/en
Application granted granted Critical
Publication of CN111856559B publication Critical patent/CN111856559B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
    • G01V1/28—Processing seismic data, e.g. analysis, for interpretation, for correction
    • G01V1/282—Application of seismic models, synthetic seismograms
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
    • G01V1/28—Processing seismic data, e.g. analysis, for interpretation, for correction
    • G01V1/30—Analysis
    • G01V1/306—Analysis for determining physical properties of the subsurface, e.g. impedance, porosity or attenuation profiles
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
    • G01V1/28—Processing seismic data, e.g. analysis, for interpretation, for correction
    • G01V1/30—Analysis
    • G01V1/307—Analysis for determining seismic attributes, e.g. amplitude, instantaneous phase or frequency, reflection strength or polarity
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
    • G01V1/28—Processing seismic data, e.g. analysis, for interpretation, for correction
    • G01V1/30—Analysis
    • G01V1/308—Time lapse or 4D effects, e.g. production related effects to the formation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
    • G06F17/141—Discrete Fourier transforms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/15—Correlation function computation including computation of convolution operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G06F18/24155—Bayesian classification

Abstract

The invention provides a multi-channel seismic spectrum inversion method and a multi-channel seismic spectrum inversion system based on a sparse Bayesian learning theory, wherein the method comprises the following steps: acquiring preprocessed post-stack seismic data; extracting a plurality of seismic spectrum inversion parameters from the stacked seismic data; extracting seismic wavelets from the post-stack seismic data; extracting spectral information of the seismic wavelets based on Fourier transform; constructing a forward calculation sub-matrix of multi-channel seismic spectrum inversion according to the multi-channel seismic spectrum inversion parameters; extracting multi-channel average seismic channel frequency spectrum information from the post-stack seismic data based on Fourier transform; and constructing a multi-channel seismic spectrum inversion target function based on the frequency spectrum information, and solving the target function through a sparse Bayesian learning theory based on a maximum expectation algorithm to obtain a multi-channel seismic spectrum inversion result. The method can comprehensively improve the thin layer identification and reservoir fine characterization capabilities of the seismic spectrum inversion result.

Description

Sparse Bayesian learning theory-based multi-channel seismic spectrum inversion method and system
Technical Field
The invention relates to the field of seismic exploration of oil and gas fields, in particular to a multi-channel seismic spectrum inversion method and system based on a sparse Bayesian learning theory.
Background
In the field of oil and gas field exploration and development, seismic data are always important guiding information for oil reservoir description and reservoir characterization due to the characteristics of wide coverage range, strong transverse continuity and the like. However, the vertical resolution of seismic data is low, so that the accuracy of the seismic data in the aspects of depicting the three-dimensional space distribution of the underground thin layer, the distribution rule of the hydrocarbon-bearing stratum and the like is usually insufficient, and the accuracy of the representation of the seismic reservoir is seriously influenced. Therefore, the improvement of the seismic vertical resolution is always an important attack and shut direction of the seismic reservoir characterization, and the seismic spectrum inversion is used as a thin layer imaging seismic processing technology, so that the vertical resolution of seismic data can be greatly improved, and the method has important significance on the identification of underground thin layers and the fine characterization of the reservoir.
Seismic spectrum inversion is a seismic data imaging processing technology which adopts a spectrum decomposition technology to improve the identification capability of a seismic lamella, and the final inversion result can be underground stratum reflection coefficient information. By performing spectrum inversion on seismic data, the range of seismic effective frequency bands can be widened, and the capability of identifying the distribution characteristics of thin layers smaller than the seismic tuning thickness is improved. In addition, the influence of colored noise with energy concentrated in a fixed frequency band range on an inversion result can be effectively avoided by seismic spectrum inversion, which is an advantage that a time domain method does not have; and the method has good suppression effect on earthquake random noise, so that the precision of thin layer identification can be integrally improved.
In the prior art, sparse seismic spectrum inversion mainly solves the sparsely constrained objective function through Basis Pursuit theory (BP) or Sparse Bayesian Learning theory (SBL). The sparse constraint is essentially one L 0 Norm optimization problem due to solving L 0 Norm belongs to NP difficult problem, BP theory simplifies the norm to L 1 And carrying out objective function solving on the norm constraint problem. But L 1 The norm naturally has the defect of insufficient sparse representation, so that the sparsity of an inversion result cannot be ensured by a BP theory under the common condition, and the accuracy of seismic thin layer identification is reduced. The SBL theory is to develop sparse parameter optimization through a certain learning criterion under a Bayes framework, and is different from the BP theory in that each basis vector is endowed with different regular parameters, so that the sparsity and the precision of an inversion result are improved.
At present, the seismic spectrum inversion method based on the SBL theory still has the following problems: 1) The sparse solution accuracy for the objective function is insufficient. The current seismic spectrum inversion method based on the SBL theory uses a Sequential Algorithm (SA) as a learning criterion to perform regular parameter training, and the method can be recorded as SBL-SA. Compared with a BP algorithm, although the sparsity of an inversion result is improved to a certain extent by the SBL-SA theory, the precision of training parameters obtained by the SBL-SA theory is usually insufficient under the noise-containing condition due to the greediness of a sequential algorithm; 2) The lateral continuity of the inversion results is insufficient: the traditional seismic spectrum inversion improves the vertical seismic resolution through a sparse theory, however, the method does not consider the transverse stability of the reflection coefficient corresponding to the transverse continuous spread of the underground stratum within a certain range, and the transverse continuity of the inversion result is poor. In addition, the low sparse solution accuracy of the traditional theory can destroy the stability of the inversion result, and further cause the reduction of the transverse continuity of the inversion result.
Disclosure of Invention
Aiming at the problems existing in the traditional seismic spectrum inversion, the invention provides a novel multichannel seismic spectrum inversion method based on a sparse Bayesian learning theory, and the precision and the transverse continuity of the seismic spectrum inversion result are improved through the following two major improvements: 1) Improving a parameter training criterion; the training criterion of the regular parameters is improved from a Sequential Algorithm (SA) in the traditional SBL-SA to an Expectation Maximization (EM) algorithm; 2) Improving input seismic trace information; the traditional single-seismic-channel spectrum input is improved to be that the average seismic-channel spectrum of a plurality of adjacent channels (Multichannel) is used as inversion input, and the method can be recorded as MSBL-EM. The invention provides a multi-channel seismic spectrum inversion method based on MSBL-EM theory, which aims to improve the accuracy of sparse solution of seismic spectrum inversion through a sparse Bayesian learning theory taking EM algorithm as a training criterion, and improve the transverse continuity of inversion results through fusing adjacent channel seismic information on the basis, thereby finally improving the accuracy of seismic thin layer identification and reservoir stratum fine characterization.
The invention provides a multi-channel seismic spectrum inversion method based on a sparse Bayesian learning theory on one hand, which comprises the following steps:
s1, acquiring preprocessed post-stack seismic data;
s2, extracting a plurality of seismic spectrum inversion parameters from the stacked seismic data;
s3, extracting seismic wavelets from the stacked seismic data;
s4, extracting frequency spectrum information of the seismic wavelets based on Fourier transform;
s5, constructing a forward calculation submatrix of multi-channel seismic spectrum inversion according to the multi-channel seismic spectrum inversion parameters;
s6, extracting multi-channel average seismic channel frequency spectrum information from the post-stack seismic data based on Fourier transform;
and S7, constructing a multi-channel seismic spectrum inversion target function based on the frequency spectrum information, and solving the target function through a sparse Bayesian learning theory based on a maximum expectation algorithm to obtain a multi-channel seismic spectrum inversion result.
In an embodiment, the preprocessing includes at least one of static correction, denoising, amplitude compensation, velocity analysis, dynamic correction, and offset.
In one embodiment, the multi-seismic-trace spectral inversion parameters include average trace number, effective spectral range, frequency sampling density, noise variance, convergence threshold, and maximum iteration number.
In one embodiment, in step S3, if the post-stack seismic data includes logging data, performing seismic wavelet extraction based on a logging curve and a well-side seismic trace in the logging data; and if the stacked seismic data does not contain logging data, performing statistical wavelet extraction on the stacked seismic data to obtain seismic wavelets.
In one embodiment, in step S5, the number of columns of the positive mathematical sub-matrix is determined according to the time-series length and the time sampling density of the seismic data, and the number of rows of the positive mathematical sub-matrix is determined according to the effective spectral range and the frequency sampling density of the seismic data.
In one embodiment, the step S6 includes:
and determining the average seismic channels corresponding to the current seismic channel and the adjacent channels thereof channel by channel according to the average channel number, and extracting the frequency spectrum information of the average seismic channels through Fourier transform to obtain the frequency spectrum information of the multi-channel average seismic channels.
In one embodiment, the step S7 includes:
calculating a reflection coefficient frequency spectrum based on a frequency domain response relation according to the frequency spectrum information of the seismic wavelets and the frequency spectrum information of the selected average seismic channel;
combining the positive calculation submatrix to construct a target function of multi-channel seismic spectrum inversion;
taking the average seismic channel frequency spectrum of a plurality of adjacent channels as inversion input, solving a target function through a sparse Bayes learning theory based on a maximum expectation algorithm, and obtaining a multi-channel seismic spectrum inversion result.
On the other hand, the embodiment of the invention also provides a multi-channel seismic spectrum inversion system based on the sparse bayesian learning theory, which comprises:
the post-stack seismic data acquisition unit is used for acquiring pre-processed post-stack seismic data;
an inversion parameter obtaining unit, configured to extract multiple seismic spectrum inversion parameters from the post-stack seismic data;
the seismic wavelet extracting unit is used for extracting seismic wavelets from the stacked seismic data;
the first frequency spectrum information acquisition unit is used for extracting frequency spectrum information of the seismic wavelet based on Fourier transform;
the forward calculation sub-matrix construction unit is used for constructing a forward calculation sub-matrix of the multi-channel seismic spectrum inversion according to the multi-channel seismic spectrum inversion parameters;
the second frequency spectrum information acquisition unit is used for extracting multi-channel average seismic channel frequency spectrum information from the post-stack seismic data based on Fourier transform;
and the inversion unit is used for constructing a multi-channel seismic spectrum inversion target function based on the frequency spectrum information and solving the target function through a sparse Bayesian learning theory based on a maximum expectation algorithm to obtain a multi-channel seismic spectrum inversion result.
In an embodiment, the preprocessing includes at least one of static correction, denoising, amplitude compensation, velocity analysis, dynamic correction, and offset.
In one embodiment, the multi-seismic-trace spectral inversion parameters include average trace number, effective spectral range, frequency sampling density, noise variance, convergence threshold, and maximum iteration number.
In one embodiment, if the post-stack seismic data includes logging data, the seismic wavelet extraction unit performs seismic wavelet extraction based on a logging curve and a well-side seismic channel in the logging data; if the stacked seismic data does not contain logging data, the seismic wavelet extraction unit performs statistical wavelet extraction on the stacked seismic data to obtain seismic wavelets.
In one embodiment, the positive calculation sub-matrix construction unit determines the number of columns of the positive calculation sub-matrix according to the time series length and the time sampling density of the seismic data, and determines the number of rows of the positive calculation sub-matrix according to the effective frequency spectrum range and the frequency sampling density of the seismic data.
In an embodiment, the second spectrum information obtaining unit determines, channel by channel, an average seismic channel corresponding to the current seismic channel and an adjacent channel thereof according to the number of average channels, and extracts spectrum information of the average seismic channel through fourier transform to obtain multi-channel average seismic channel spectrum information.
In one embodiment, the inversion unit comprises:
the reflection coefficient spectrum acquisition module is used for calculating a reflection coefficient spectrum based on a frequency domain response relation according to the spectrum information of the seismic wavelets and the spectrum information of the selected average seismic channel;
the target function construction module is used for constructing a multi-channel seismic spectrum inversion target function by combining the forward calculation submatrix;
and the inversion calculation module is used for solving a target function by taking the average seismic channel frequency spectrum of a plurality of adjacent channels as inversion input through a sparse Bayes learning theory based on a maximum expectation algorithm to obtain a multi-channel seismic spectrum inversion result.
The invention has the following advantages: 1) Compared with the current SBL-SA algorithm, the MSBL-EM theory of the technical scheme optimizes the learning criterion of the regular parameters through the EM algorithm, and greatly optimizes the training criterion of the regular parameters compared with the traditional Sequential Algorithm (SA), thereby improving the accuracy of inversion results; 2) According to the technical scheme, correlation characteristics between adjacent seismic channels are considered, and the transverse continuity and stability of seismic spectrum inversion results are greatly improved by introducing the frequency spectrum information of the adjacent seismic channels; 3) Compared with the traditional SBL-SA theory, the technical scheme has higher noise resistance; 4) According to the technical scheme, the vertical sparsity of the seismic spectrum inversion result is optimized through a sparse Bayesian learning theory taking an EM algorithm as a learning criterion, and the transverse continuity of the inversion result is improved by combining spectrum information of adjacent seismic channels, so that the thin layer identification and reservoir fine characterization capability of the seismic spectrum inversion result can be comprehensively improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the prior art descriptions will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a schematic flow chart of a multi-channel seismic spectrum inversion method based on a sparse Bayesian learning theory according to an embodiment of the present invention;
FIG. 2a is a reflection coefficient of an input two-dimensional wedge formation model in an embodiment of the invention;
FIG. 2b is a synthetic seismic record of the input two-dimensional wedge formation model in an embodiment of the invention, wherein the seismic wavelets are 30Hz zero-phase Rake wavelets;
fig. 3a, 3b, 3c and 3d are inversion results and corresponding errors obtained by using the synthetic seismic record of the two-dimensional wedge-shaped stratigraphic model in fig. 2b as input in the embodiment, wherein the input is noise-free synthetic seismic data;
fig. 4a, 4b, 4c, and 4d are inversion results and corresponding errors obtained by using the two-dimensional wedge-shaped stratigraphic model synthetic seismic record in fig. 2b as input in the present embodiment, wherein the signal-to-noise ratio SNR of the seismic data =3;
fig. 5a is a Marmousi-II model according to an embodiment of the present invention, and fig. 5b is a reflection coefficient of multi-channel seismic spectrum inversion obtained based on MSBL-EM theory with fig. 5a as an input and SNR =3;
fig. 6a1 and 6b1 are the seismic spectrum inversion single-channel results obtained at 6.4km and 15km based on the SBL-SA theory when SNR =3, and fig. 6a2 and 6b2 are the seismic spectrum inversion single-channel results obtained at 6.4km and 15km based on the MSBL-EM theory when SNR =3, respectively;
FIG. 7 is the phase error robustness test result of the wavelet developed based on the single-channel data at 15km by the model Marmousi-II in FIG. 5 a;
8a, 8b, 8c, and 8d are multi-channel inversion effect analyses performed on the local data of the model Marmousi-II in fig. 5a according to the present embodiment;
FIG. 9 shows the actual three-dimensional post-stack seismic data in a work area of the east of China;
FIG. 10a, FIG. 10b, FIG. 10c and FIG. 10d are inversion results obtained by seismic spectrum inversion based on SBL-SA, MSBL-EM, SBL-SA and MSBL-EM, respectively, with the three-dimensional post-stack seismic data in FIG. 9 as input;
fig. 11 is a schematic structural diagram of a multi-channel seismic spectrum inversion system based on a sparse bayesian learning theory according to an embodiment of the present invention;
fig. 12 is a schematic diagram of the structure of the inversion unit 70 in the system shown in fig. 11.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Aiming at the problems of the traditional seismic spectrum inversion method, the embodiment of the invention provides a novel multi-channel seismic spectrum inversion method based on a sparse Bayesian learning theory. The invention is based on the following problems: (1) The traditional SBL-SA theory uses a Sequential Algorithm (SA) as a learning criterion to carry out parameter training, the training precision is usually insufficient, and the influence on the inversion result of the seismic spectrum is large; (2) The learning criterion of the regular parameters is improved through the EM algorithm based on the sparse Bayesian learning theory of the EM algorithm, compared with the optimization criterion of the Sequential Algorithm (SA) in the SBL-SA, the training precision of the regular parameters can be greatly improved, and the method has important significance for improving the precision of the inversion result of the seismic spectrum; (3) The traditional seismic spectrum inversion usually takes single seismic channel frequency spectrum information as input, and neglects the spatial correlation between adjacent seismic channels, so that the transverse continuity of an inversion result is poor; (4) By inputting the average frequency spectrum of the current trace and a plurality of adjacent seismic traces based on the sparse Bayesian learning theory of the EM algorithm, the transverse continuity of the seismic spectrum inversion result can be effectively improved, and the random noise of the earthquake can be suppressed, so that the inversion result precision is further improved, and the thin layer identification and oil reservoir fine characterization capabilities of seismic data are improved.
Fig. 1 is a general flowchart of a multi-channel seismic spectrum inversion method based on a sparse bayesian learning theory according to an embodiment of the present invention, which specifically includes the following steps:
s1, acquiring preprocessed stacked seismic data.
And S2, extracting a plurality of seismic spectrum inversion parameters from the stacked seismic data.
And S3, extracting seismic wavelets from the post-stack seismic data.
And S4, extracting the frequency spectrum information of the seismic wavelets based on Fourier transform.
And S5, constructing a forward calculation submatrix for multi-channel seismic spectrum inversion according to the multi-channel seismic spectrum inversion parameters.
And S6, extracting multi-channel average seismic channel frequency spectrum information from the post-stack seismic data based on Fourier transform.
And S7, constructing a multi-channel seismic spectrum inversion target function based on the frequency spectrum information, and solving the target function through a sparse Bayesian learning theory based on a maximum expectation algorithm to obtain a multi-channel seismic spectrum inversion result.
In one embodiment, the preprocessed stacked seismic data obtained in step S1 refers to seismic data after a series of seismic data processing links, including but not limited to static correction, denoising, amplitude compensation, velocity analysis, dynamic correction, migration, and the like. The processing effect of each link in the early stage has great influence on the subsequent multi-channel seismic spectrum inversion, so that high-quality and high-fidelity early stage data processing is a necessary condition for ensuring the subsequent inversion quality. The finally processed post-stack seismic data is equal to the acquired data under the self-excited self-collected condition and can be recorded as s (t, x), wherein t represents the seismic wave double-travel time, namely the seismic channel time sequence, and x is the corresponding seismic channel number.
Obtaining and analyzing the pre-processed post-stack seismic data to determine a plurality of important inversion parameters, e.g., determining an effective spectral range f by frequency domain analysis span Parameters such as frequency sampling interval df and frequency domain sampling point number N; parameters such as time sampling interval dt and the number of time domain sampling points L are determined through time domain analysis, and important parameters such as convergence threshold tol, maximum iteration number MaxIter, noise variance lambda and average channel number M are finally determined.
In one embodiment, when performing seismic wavelet extraction based on the post-stack seismic data, under the condition that the post-stack seismic data obtained in the step S1 includes logging data, seismic wavelet extraction can be performed through a logging curve and a well-side seismic channel in the logging data, and the extracted seismic wavelets should meet the constraints of the logging data and the seismic data; when the post-stack seismic data does not contain logging data, statistical wavelet extraction can be carried out through the seismic data under certain assumed conditions. The extracted seismic wavelets may be denoted as w (t), where t is the wavelet time series.
When the spectrum information of the seismic wavelet is extracted based on Fourier transform in the step S4, the spectrum information can be represented by complex numbers, the mode of the complex numbers corresponds to the frequency-energy distribution rule of the seismic wavelet, and the spectrum of the seismic wavelet can be recorded as F w (f) Where f represents frequency in Hz. Earthquake deviceWave w (t) and its frequency spectrum F w (f) The response relationship of (c) may be characterized as:
F w (f)＝∫w(t)e -j2πft dt (1)
wherein, F w (f) Representing the seismic wavelet spectrum, f represents frequency and has the unit of Hz; j is an imaginary number, and j multiplied by j = -1 is satisfied.
In actual production, the seismic wavelet spectral information F w (f) It is usually obtained by discrete fourier transform relation calculation:
Figure BDA0002047175540000071
wherein L represents the seismic wavelet sequence length and k is the discrete spectrum F w (k) Corresponding subscripts of (a).
In one embodiment, when the step S5 is used to construct the forward calculation submatrix of the multi-channel seismic spectrum, the number of columns of the forward calculation submatrix depends on the time sequence length t and the time sampling density dt of the seismic data, and the number of rows depends on the effective spectral range f of the seismic data span And a frequency sampling density df. The positive operator matrix can be denoted as D (f, t), where t denotes time, where f denotes frequency in Hz.
Assuming that the number of sampling points in the frequency domain and the number of sampling points in the time domain of the post-stack seismic data are N and L, the scale of the forward sub-matrix D (f, t) is nxl. The time domain seismic convolution model may be expressed as:
Figure BDA0002047175540000081
wherein s (t), w (t) and r (t) respectively represent time domain seismic traces, seismic wavelets and reflection coefficient sequences,
Figure BDA0002047175540000085
representing a convolution operation. Corresponding to the frequency domain, seismic data spectrum F s (f) Seismic wavelet spectrum F w (f) And reflection coefficient spectrum F r (f) The response relationship of the three can be expressed as:
F s (f)＝F w (f)F r (f) (4)
on the basis, the reflection coefficient spectrum can be calculated through the spectrum information of the seismic trace and the seismic wavelet:
Figure BDA0002047175540000082
from the above formula, the reflection coefficient spectrum F r (f) The time series r (t) of the reflection coefficient can also be characterized by the following response relation:
F r (f)＝D(f,t)r(t) (6)
in the formula, D (f, t) is a positive operator matrix and satisfies the following conditions:
Figure BDA0002047175540000083
wherein, t 1 ,t 2 ,...,t L Is a corresponding time series of reflection coefficients r (t) of length L; f. of 1 ,f 2 ,...,f N As a reflection coefficient spectrum F r (f) Has a length N.
In an embodiment, when the step S6 is used to obtain the multiple channels of average seismic frequency spectrum information, firstly, the average seismic channel corresponding to the current seismic channel and its adjacent channels is determined channel by channel according to the average channel number M, and then the frequency spectrum information of the average seismic channel is extracted through fourier transform, where the frequency spectrum information can be represented by a complex number, and a modulus of the complex number corresponds to a frequency-energy distribution rule of the seismic channel. The seismic spectrum may be denoted as F s (f) Where f represents frequency in Hz.
Extracting multi-channel average seismic data s of the post-stack seismic data s (t, x) based on Fourier transform avg (t) spectrum information. Setting:
Figure BDA0002047175540000084
wherein, s (t, x) i ) Representing the xth seismic data i The method comprises the following steps of (1) obtaining channel data, wherein M is the number of selected adjacent seismic channels; s avg And (t) is average data corresponding to the M adjacent seismic traces.
s avg (t) and its frequency spectrum F s_avg (f) The response relationship of (c) can also be expressed as:
F s_avg (f)＝∫s avg (t)e -j2πft dt (9)
as shown in equation (2), s is calculated by discrete Fourier transform avg Spectrum of (t):
Figure BDA0002047175540000091
wherein L represents s avg (t) length of time series, k being discrete spectrum F s_avg (k) Corresponding subscripts of (a).
Combined formula (8) and formula (10), F s_avg (k) The non-zero portion (0. Ltoreq. K. Ltoreq.L-1) can be written as:
Figure BDA0002047175540000092
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0002047175540000094
representing seismic adjacent traces s (t, x) i ) Corresponding spectrum information. From the above formula, the average data s of multiple adjacent tracks avg (t) frequency spectrum F s_avg (k) Equal to the average of the frequency spectrum of a plurality of adjacent channels
Figure BDA0002047175540000093
The above formula provides a theoretical basis for inputting the inversion of the multi-channel seismic spectrum by using the multi-channel averaged spectral information.
In one embodiment, when performing the inversion in step S7, the spectral information F of the seismic wavelet is first passed w (f) And spectral information F of the selected average seismic trace s (f) Based on frequency domain responseRelationship F s (f)＝F w (f)F r (f) Calculating the reflectance frequency spectrum F r (f) On the basis, combining with the positive calculus submatrix to construct a target function of multi-channel seismic spectrum inversion; and then, iteratively solving the target function by combining the MSBL-EM theory to obtain a multi-channel seismic spectrum inversion result, namely, solving the target function by taking the average seismic channel frequency spectrum of a plurality of adjacent channels as inversion input and based on the sparse Bayes learning theory of the maximum expectation algorithm to obtain the multi-channel seismic spectrum inversion result.
For vertical sparsity of the time domain reflection coefficient sequence r (t), the sparse Bayesian learning theory assumes that the reflection coefficient at any time position of t = t (i) satisfies the zero mean and the variance is epsilon i (i = 1...., L) gaussian distribution:
p(r(t i )|ε i )＝N(0,ε i ) (12)
when epsilon i A reflection coefficient r (t) =0 i ) =0. Thus, for sparse signals, a large fraction of ε may be trained using parameters i (i = 1.. ·, L) is equal to zero, ensuring sparsity of the reflection coefficient r (t).
On the basis of the prior information, if the inversion error of the multi-channel seismic spectrum also meets zero-mean Gaussian distribution, the likelihood function can be expressed as:
Figure BDA0002047175540000101
where λ is the noise variance and I represents the identity matrix.
From the combination of equation (12) and equation (13), the posterior probability distribution of the reflection coefficient sequence r (t) in the multi-channel seismic spectrum inversion can be written as:
Figure BDA0002047175540000102
wherein, p (F) r_avg | ε, λ) is a regularization term; mu and sigma are respectively mean and covariance of Gaussian distribution, and satisfy:
Figure BDA0002047175540000103
in the formula, Ψ = diag (ε) 1 ,ε 2 ,...,ε L )。
As can be seen from the above formula, the target parameter mu of the multi-channel seismic spectrum inversion is the seismic spectrum information F r_avg Canonical parameter ε = [ ε ] 1 ,...,ε L ]And a function of the noise variance lambda. Therefore, in the seismic spectrum information F r_avg Under the known condition, the reflection coefficient sequence r (t) of the multi-channel seismic spectrum inversion can be obtained by calculating the parameters epsilon and lambda.
Let the parameter set be γ = [ ε, λ]＝[ε 1 ,...,ε L ,λ]When upsilon is equal to the true parameter, F r_avg The probability of occurrence should be maximal, i.e. p (F) r_avg γ) max. Therefore, let p (F) be considered to be r_avg Y, the value of the maximum probability obtained by y) is the optimal parameter combination. p (F) r_avg Y) may be expressed as:
Figure BDA0002047175540000104
wherein Θ = λ I + D Ψ D T 。
Let L (γ) = -logp (F) r_avg Y), the probability maximization problem of the above formula can be converted into the minimization problem of L (y). L (γ) can be written as:
Figure BDA0002047175540000111
since the variable r (t) is not included in the formula, the optimal parameter combination γ = [ epsilon ] cannot be obtained directly by partial derivative of L (γ) 1 ,...,ε L ,λ]. Therefore, the invention introduces an EM algorithm to carry out the parameter estimation by taking r (t) as a hidden variable, and obtains the optimal parameter combination by maximizing the following objective function:
Figure BDA0002047175540000112
wherein E (.) represents desire; gamma ray (old) The combination of parameters estimated for the last iteration.
As can be seen from the above formula, the first term of Q (γ) is related to λ only and the second term is related to e only. Thus, the above formula can be abbreviated as:
Q(Υ)＝Q(λ)+Q(ε) (19)
then, by derivation of Q (epsilon) to epsilon and Q (lambda) to lambda, the optimal parameter combination γ = [ epsilon ] can be obtained 1 ,...,ε L ,λ]. Wherein:
Figure BDA0002047175540000113
in the formula, mu i The ith element representing μ; sigma ii Representing the ith diagonal element of the matrix Σ. The regularization parameter ε can be derived from the above equation i The update relationship of (1):
Figure BDA0002047175540000117
for Q (λ), we can approximate the simplification:
Figure BDA0002047175540000114
wherein λ is (old) And with
Figure BDA0002047175540000115
Respectively represent lambda and epsilon i The previous iteration result. From this, it is derived that the update relationship of λ is:
Figure BDA0002047175540000116
training the parameter combination γ = [ epsilon ] by the learning criteria of equation (21) and equation (23) 1 ,...,ε L ,λ]Then, the driving formula (15) can acquire a reflection coefficient sequence r (t) acquired by inverting the multi-channel seismic spectrum.
In an embodiment, the following working steps may be taken to implement the above technical solution: 1) Collecting the post-stack seismic data; 2) Extracting multiple channels of seismic spectrum inversion related parameters based on the stacked seismic data; 3) Performing seismic wavelet extraction based on the stacked seismic data; 4) Extracting spectral information of the seismic wavelets based on Fourier transform; 5) Constructing a positive calculation submatrix of multi-channel seismic spectrum inversion; 6) Extracting multi-channel average seismic frequency spectrum information of the post-stack seismic data based on Fourier transform; 7) Constructing a multi-channel seismic spectrum inversion target function based on the frequency spectrum information, and solving the target function through a sparse Bayesian learning theory based on a maximum expectation algorithm to obtain a multi-channel seismic spectrum inversion result; 8) Iteratively solving each channel through the step 7) until the inversion residual error is less than a given threshold value or the iteration times are more than a given maximum iteration times; 9) Repeating the steps 6) -8) for all seismic traces until the end; 10 ) finally outputting a plurality of seismic spectrum inversion results.
Fig. 2a is a reflection coefficient of a two-dimensional wedge-shaped stratigraphic model according to an embodiment of the present invention, and fig. 2b is a synthetic seismic record of the two-dimensional wedge-shaped stratigraphic model according to an embodiment of the present invention, wherein the seismic wavelets are 30Hz zero-phase rake wavelets.
Fig. 3a and 3b are seismic spectrum inversion results and errors thereof obtained based on the SBL-SA theory under the condition of no noise by using the synthetic seismic record of the wedge-shaped stratum model shown in fig. 2b as an input. Fig. 3c and 3d are seismic spectrum inversion results and errors thereof obtained based on the MSBL-EM theory under the condition of no noise by taking the synthetic seismic record of the wedge-shaped stratum model shown in fig. 2b as input. As shown in the figure, the thin layer identification capability of seismic spectrum inversion based on the MSBL-EM theory is obviously superior to that of the inversion result of the traditional SBL-SA theory, and the theoretical advantage of the invention is embodied.
Fig. 4a and 4b respectively show the seismic spectrum inversion result and its error based on the SBL-SA theory when the wedge-shaped stratigraphic model synthetic seismic record in fig. 2b is used as an input and the SNR = 3. Fig. 4c and 4d respectively show the seismic spectrum inversion result and its error obtained based on the MSBL-EM theory when the wedge-shaped stratigraphic model synthetic seismic record in 2b is used as an input and SNR = 3. As shown in the figure, the anti-noise capability of seismic spectrum inversion based on the MSBL-EM theory is obviously superior to that of the inversion result of the traditional SBL-SA theory, and the theoretical advantage of the invention is embodied.
Fig. 5a is a Marmousi-II model embodiment provided by the embodiment of the present invention, and fig. 5b is a reflection coefficient of multi-channel seismic spectrum inversion obtained based on MSBL-EM theory with fig. 5a as an input under the condition that SNR = 3. As shown by arrows in the figure, the inversion result has better thin layer depicting capability than the original seismic section, and the theoretical advantage of the invention is embodied.
Fig. 6a1 and 6b1 are the seismic spectrum inversion single-track results obtained at 6.4km and 15km based on the SBL-SA theory when SNR =3, and fig. 6a2 and 6b2 are the seismic spectrum inversion single-track results obtained at 6.4km and 15km based on the MSBL-EM theory when SNR =3, respectively. As shown by the arrows in the figure, the MSBL-EM inversion result (fig. 6a2 and fig. 6b 2) has a better matching relationship with the SBL-SA inversion result (fig. 6a1 and fig. 6b 1) and the input model in two single-pass comparisons, and the theoretical advantage of the present invention is embodied.
As shown in FIG. 7, single-channel data at 15km of a Marmousi-II model is used as input, and a wavelet error robustness test of seismic spectrum inversion is carried out on the basis of the MSBL-EM theory. The result shows that when the phase error is in the range of-pi/6, the inversion result and the input model channel have better matching relation, and the theoretical advantage of the invention is embodied.
Fig. 8a, 8b, 8c, and 8d are multi-channel inversion effect analyses performed on the local data of the Marmousi-II model in fig. 5a according to this embodiment. As shown in fig. 8a to 8d, local data in a Marmousi-II model synthetic seismic record (fig. 5 a) is used as input, and multi-channel (M = 5) and single-channel seismic spectrum inversion is carried out based on the MSBL-EM theory under the condition of SNR = 3. As shown by arrows in the figure, compared with the traditional single-channel inversion, the multi-channel inversion result has better transverse continuity and stronger thin layer characterization capability, and the theoretical advantages of the invention are reflected.
FIG. 9 shows the actual three-dimensional post-stack seismic data of a work area in the east of China, and FIGS. 10a, 10b, 10c and 10d are inversion results obtained by seismic spectrum inversion based on SBL-SA, MSBL-EM, SBL-SA and MSBL-EM, respectively, with the three-dimensional post-stack seismic data of FIG. 9 as input. By taking the actual three-dimensional post-stack seismic data in fig. 9 as input, a seismic spectrum inversion result section (Inline = 100) obtained based on the SBL-SA theory is shown in fig. 10a, and a seismic spectrum inversion result section (Inline = 100) obtained based on the MSBL-EM theory is shown in fig. 10b, as shown by an oval area enclosed by a dotted line in the figure, the method better maintains the transverse continuity and stability of the inversion result compared with the traditional SBL-SA theory. Furthermore, a comparison of the stratigraphic amplitude slices of the synthetic seismic records based on the inversion results also shows that: compared with the inversion result (see fig. 10 c) obtained by the traditional SBL-SA theory, the inversion result (see fig. 10 d) obtained by the method has more fine description capability on the sedimentary features of the underground geologic body, and the practical application advantage of the method is reflected as shown by an oval area in the graph.
Based on the same inventive concept as the sparse bayesian learning theory-based multi-channel seismic spectrum inversion method shown in fig. 1, the embodiment of the present application further provides a system, as described in the following embodiments. The principle of solving the problems of the system is similar to that of the multi-channel seismic spectrum inversion method based on the sparse Bayesian learning theory in the figure 1, so the implementation of the system can refer to the implementation of the multi-channel seismic spectrum inversion method based on the sparse Bayesian learning theory in the figure 1, and repeated parts are not repeated.
In another embodiment, the present invention further provides a multi-channel seismic spectrum inversion system based on the sparse bayesian learning theory, the structure of which is shown in fig. 11, and the system includes: the system comprises a post-stack seismic data acquisition unit 10, an inversion parameter acquisition unit 20, a seismic wavelet extraction unit 30, a first spectrum information acquisition unit 40, a forward calculation submatrix construction unit 50, a second spectrum information acquisition unit 60 and an inversion unit 70.
The post-stack seismic data acquisition unit 10 is used for acquiring pre-processed post-stack seismic data; the inversion parameter obtaining unit 20 is configured to extract multiple seismic spectrum inversion parameters from the post-stack seismic data; the seismic wavelet extracting unit 30 is configured to extract seismic wavelets from the post-stack seismic data; the first spectrum information acquisition unit 40 is used for extracting spectrum information of the seismic wavelets based on Fourier transform; the forward calculation sub-matrix construction unit 50 is configured to construct a forward calculation sub-matrix for multi-channel seismic spectrum inversion according to the multi-channel seismic spectrum inversion parameters; the second spectrum information obtaining unit 60 is configured to extract multiple channels of average seismic channel spectrum information from the post-stack seismic data based on fourier transform; the inversion unit 70 is configured to construct a multi-channel seismic spectrum inversion target function based on the spectrum information, and solve the target function through a sparse bayesian learning theory based on a maximum expectation algorithm to obtain a multi-channel seismic spectrum inversion result.
In an embodiment, the preprocessing includes at least one of static correction, denoising, amplitude compensation, velocity analysis, dynamic correction, and offset.
In one embodiment, the multi-seismic channel spectral inversion parameters include an average channel number, an effective spectral range, a frequency sampling density, a noise variance, a convergence threshold, and a maximum number of iterations.
In one embodiment, if the post-stack seismic data includes well-log data, the seismic wavelet extraction unit 30 performs seismic wavelet extraction based on a well-log curve and a well-side seismic trace in the well-log data; if the stacked seismic data does not contain logging data, the seismic wavelet extraction unit 30 performs statistical wavelet extraction on the stacked seismic data to obtain seismic wavelets.
In one embodiment, the forward sub-matrix construction unit 50 determines the number of columns of the forward sub-matrix according to the time-series length and the time sampling density of the seismic data, and determines the number of rows of the forward sub-matrix according to the effective spectrum range and the frequency sampling density of the seismic data.
In an embodiment, the second spectrum information obtaining unit 60 determines, channel by channel, an average seismic channel corresponding to the current seismic channel and an adjacent channel thereof according to the number of average seismic channels, and extracts spectrum information of the average seismic channel through fourier transform to obtain spectrum information of multiple average seismic channels.
In one embodiment, the inversion unit 70 is configured as shown in FIG. 12, and mainly includes: a reflection coefficient spectrum obtaining module 71, configured to calculate a reflection coefficient spectrum based on a frequency domain response relationship according to the spectrum information of the seismic wavelet and the spectrum information of the selected average seismic trace; an objective function constructing module 72, configured to construct an objective function of multi-channel seismic spectrum inversion by combining the forward calculus submatrix; and the inversion calculation module 73 is configured to use the average seismic channel frequency spectrum of the multiple adjacent channels as an inversion input, and solve the target function through a sparse bayesian learning theory based on a maximum expectation algorithm to obtain a multi-channel seismic spectrum inversion result.
Due to the adoption of the technical scheme, the invention has the following advantages: 1) Compared with the current SBL-SA algorithm, the MSBL-EM theory of the technical scheme optimizes the learning criterion of the regular parameters through the EM algorithm, and greatly optimizes the training criterion of the regular parameters compared with the traditional Sequential Algorithm (SA), thereby improving the accuracy of the inversion result; 2) According to the technical scheme, the correlation characteristics between adjacent seismic channels are considered, and the transverse continuity and stability of the seismic spectrum inversion result are greatly improved by introducing the frequency spectrum information of the adjacent seismic channels; 3) Compared with the traditional SBL-SA theory, the technical scheme has higher anti-noise capability; 4) According to the technical scheme, the vertical sparsity of the seismic spectrum inversion result is optimized through a sparse Bayesian learning theory taking an EM algorithm as a learning criterion, and the transverse continuity of the inversion result is improved by combining the spectrum information of adjacent seismic channels, so that the thin layer identification and reservoir stratum fine characterization capability of the seismic spectrum inversion result can be comprehensively improved.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention has been described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The principle and the implementation mode of the invention are explained by applying specific embodiments in the invention, and the description of the embodiments is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.

Claims (12)

1. A multi-channel seismic spectrum inversion method based on a sparse Bayesian learning theory is characterized by comprising the following steps of:
s1, acquiring preprocessed stacked seismic data;
s2, extracting a plurality of seismic spectrum inversion parameters from the post-stack seismic data;
s3, extracting seismic wavelets from the stacked seismic data;
s4, extracting frequency spectrum information of the seismic wavelets based on Fourier transform;
s5, constructing a forward calculation submatrix of multi-channel seismic spectrum inversion according to the multi-channel seismic spectrum inversion parameters;
s6, extracting multi-channel average seismic channel frequency spectrum information from the post-stack seismic data based on Fourier transform;
s7, constructing a multi-channel seismic spectrum inversion target function based on the frequency spectrum information of the seismic wavelets and the frequency spectrum information of the multi-channel average seismic channel, and solving the target function through a sparse Bayes learning theory based on a maximum expectation algorithm to obtain a multi-channel seismic spectrum inversion result;
wherein, the step S7 specifically includes:
calculating a reflection coefficient frequency spectrum based on a frequency domain response relation according to the frequency spectrum information of the seismic wavelets and the frequency spectrum information of the selected average seismic channel;
combining the positive calculation submatrix to construct a target function of multi-channel seismic spectrum inversion;
taking the average seismic channel frequency spectrum of a plurality of adjacent channels as inversion input, solving a target function through a sparse Bayes learning theory based on a maximum expectation algorithm, and obtaining a multi-channel seismic spectrum inversion result.
2. The sparse bayesian learning theory based multi-channel seismic spectrum inversion method of claim 1, wherein the preprocessing comprises at least one of static correction, denoising, amplitude compensation, velocity analysis, dynamic correction, migration.
3. The sparse bayesian learning theory-based multi-channel seismic spectrum inversion method according to claim 1, wherein the multi-channel seismic spectrum inversion parameters include an average channel number, an effective spectrum range, a frequency sampling density, a noise variance, a convergence threshold, and a maximum iteration number.
4. The multi-channel seismic spectrum inversion method based on the sparse Bayesian learning theory as recited in claim 1, wherein in step S3, if the post-stack seismic data comprises logging data, seismic wavelet extraction is performed based on a logging curve and a well-side seismic channel in the logging data; and if the stacked seismic data does not contain logging data, performing statistical wavelet extraction on the stacked seismic data to obtain seismic wavelets.
5. The sparse Bayesian learning theory-based multi-channel seismic spectrum inversion method according to claim 1, wherein in step S5, the number of columns of the positive operation submatrix is determined according to the time series length and the time sampling density of the seismic data, and the number of rows of the positive operation submatrix is determined according to the effective frequency spectrum range and the frequency sampling density of the seismic data.
6. The sparse Bayesian learning theory-based multi-channel seismic spectrum inversion method according to claim 3, wherein the step S6 comprises:
and determining the average seismic channel corresponding to the current seismic channel and the adjacent channel thereof channel by channel according to the average channel number, and extracting the frequency spectrum information of the average seismic channel through Fourier transform to obtain multi-channel average seismic channel frequency spectrum information.
7. A sparse Bayesian learning theory-based multi-channel seismic spectrum inversion system, the system comprising:
the post-stack seismic data acquisition unit is used for acquiring the pre-processed post-stack seismic data;
the inversion parameter acquisition unit is used for extracting a plurality of seismic spectrum inversion parameters from the post-stack seismic data;
the seismic wavelet extracting unit is used for extracting seismic wavelets from the stacked seismic data;
the first frequency spectrum information acquisition unit is used for extracting frequency spectrum information of the seismic wavelet based on Fourier transform;
the forward calculation sub-matrix construction unit is used for constructing a forward calculation sub-matrix of the multi-channel seismic spectrum inversion according to the multi-channel seismic spectrum inversion parameters;
the second frequency spectrum information acquisition unit is used for extracting multi-channel average seismic channel frequency spectrum information from the post-stack seismic data based on Fourier transform;
the inversion unit is used for constructing a multi-channel seismic spectrum inversion target function based on the frequency spectrum information of the seismic wavelets and the frequency spectrum information of the multi-channel average seismic channel and solving the target function through a sparse Bayesian learning theory based on a maximum expectation algorithm to obtain a multi-channel seismic spectrum inversion result;
wherein the inversion unit comprises:
the reflection coefficient spectrum acquisition module is used for calculating a reflection coefficient spectrum based on a frequency domain response relation according to the spectrum information of the seismic wavelets and the spectrum information of the selected average seismic channel;
the target function construction module is used for constructing a multi-channel seismic spectrum inversion target function by combining the forward calculation submatrix;
and the inversion calculation module is used for solving a target function by taking the average seismic channel frequency spectrum of a plurality of adjacent channels as inversion input through a sparse Bayes learning theory based on a maximum expectation algorithm to obtain a multi-channel seismic spectrum inversion result.
8. The sparse bayesian learning theory based multi-channel seismic spectrum inversion system according to claim 7, wherein the preprocessing comprises at least one of static correction, denoising, amplitude compensation, velocity analysis, dynamic correction, migration.
9. The sparse bayesian learning theory-based multi-channel seismic spectrum inversion system of claim 7, wherein the multi-channel seismic spectrum inversion parameters include an average number of channels, an effective spectral range, a frequency sampling density, a noise variance, a convergence threshold, and a maximum number of iterations.
10. The multi-channel seismic spectrum inversion system based on the sparse bayesian learning theory as claimed in claim 7, wherein if the stacked seismic data includes logging data, the seismic wavelet extraction unit performs seismic wavelet extraction based on a logging curve and a well-side seismic channel in the logging data; if the stacked seismic data does not contain logging data, the seismic wavelet extraction unit performs statistical wavelet extraction on the stacked seismic data to obtain seismic wavelets.
11. The sparse bayesian learning theory based multi-channel seismic spectrum inversion system according to claim 7, wherein the forward sub-matrix construction unit determines the number of columns of the forward sub-matrix according to the time sequence length and the time sampling density of the seismic data, and determines the number of rows of the forward sub-matrix according to the effective spectrum range and the frequency sampling density of the seismic data.
12. The sparse bayesian learning theory-based multi-channel seismic spectrum inversion system according to claim 9, wherein the second spectrum information obtaining unit determines, channel by channel, average seismic channels corresponding to the current seismic channel and its neighboring channels according to the average channel number, and extracts spectrum information of the average seismic channels through fourier transform to obtain multi-channel average seismic channel spectrum information.
CN201910362219.9A 2019-04-30 2019-04-30 Multi-channel seismic spectrum inversion method and system based on sparse Bayes learning theory Active CN111856559B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910362219.9A CN111856559B (en) 2019-04-30 2019-04-30 Multi-channel seismic spectrum inversion method and system based on sparse Bayes learning theory

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910362219.9A CN111856559B (en) 2019-04-30 2019-04-30 Multi-channel seismic spectrum inversion method and system based on sparse Bayes learning theory

Publications (2)

Publication Number Publication Date
CN111856559A CN111856559A (en) 2020-10-30
CN111856559B true CN111856559B (en) 2023-02-28

Family

ID=72965769

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910362219.9A Active CN111856559B (en) 2019-04-30 2019-04-30 Multi-channel seismic spectrum inversion method and system based on sparse Bayes learning theory

Country Status (1)

Country Link
CN (1) CN111856559B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113777650B (en) * 2021-08-12 2022-10-25 西安交通大学 Sparse time-frequency spectrum decomposition method, device and equipment based on mixed norm and wavelet transform and storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2113792A1 (en) * 2008-04-29 2009-11-04 ExxonMobil Upstream Research Company Spectral shaping inversion and migration of seismic data
CN103293552A (en) * 2013-05-24 2013-09-11 中国石油天然气集团公司 Pre-stack seismic data retrieval method and system
EP2713185A1 (en) * 2012-09-28 2014-04-02 Dal Moro Giancarlo Ditta Individuale Method and apparatus to detect and analyze seismic signals
CN103792571A (en) * 2012-10-26 2014-05-14 中国石油化工股份有限公司 Point constraint Bayes sparse pulse inversion method
CN105842732A (en) * 2016-03-16 2016-08-10 中国石油大学(北京) Inversion method of multichannel sparse reflection coefficient and system thereof

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
ATE483175T1 (en) * 2006-06-06 2010-10-15 Total Sa METHOD AND PROGRAM FOR CHARACTERIZING THE TEMPORARY DEVELOPMENT OF A PETROLEUM REQUIREMENT
US9651697B2 (en) * 2014-03-28 2017-05-16 Cgg Services Sas Noise attentuation using a dipole sparse Tau-p inversion
CN106324675B (en) * 2016-10-09 2018-09-07 中国石油大学(华东) A kind of broad-band teleseismic wave impedance low-frequency information prediction technique and system

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2113792A1 (en) * 2008-04-29 2009-11-04 ExxonMobil Upstream Research Company Spectral shaping inversion and migration of seismic data
EP2713185A1 (en) * 2012-09-28 2014-04-02 Dal Moro Giancarlo Ditta Individuale Method and apparatus to detect and analyze seismic signals
CN103792571A (en) * 2012-10-26 2014-05-14 中国石油化工股份有限公司 Point constraint Bayes sparse pulse inversion method
CN103293552A (en) * 2013-05-24 2013-09-11 中国石油天然气集团公司 Pre-stack seismic data retrieval method and system
CN105842732A (en) * 2016-03-16 2016-08-10 中国石油大学(北京) Inversion method of multichannel sparse reflection coefficient and system thereof

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Discrete Tomography by Bayesian Labeling with its Efficient Algorithm;Yuan Cheng et al.;《Proceedings of SPIE》;20001231;第4121卷;第30-38页 *
时频联合域贝叶斯地震反演方法;印兴耀等;《石油物探》;20170331;第56卷(第2期);第250-260页 *

Also Published As

Publication number Publication date
CN111856559A (en) 2020-10-30

Similar Documents

Publication Publication Date Title
Xue et al. Amplitude-preserving iterative deblending of simultaneous source seismic data using high-order radon transform
CN105842732B (en) The inversion method and system of the sparse reflectance factor of multiple tracks
US20070274155A1 (en) Coding and Decoding: Seismic Data Modeling, Acquisition and Processing
CN111723329A (en) Seismic phase feature recognition waveform inversion method based on full convolution neural network
CN111368247B (en) Sparse representation regularization prestack AVO inversion method based on fast orthogonal dictionary
CN112946749B (en) Method for suppressing seismic multiples based on data augmentation training deep neural network
CN110895348B (en) Method, system and storage medium for extracting low-frequency information of seismic elastic impedance
CN113962244A (en) Rayleigh wave seismic data noise removal method, storage medium and electronic device
CN103364826A (en) An earthquake blind source deconvolution method based on independent component analysis
CN105093315B (en) A method of removal coal seam strong reflectance signal
CN115905805A (en) DAS data multi-scale noise reduction method based on global information judgment GAN
CN111856568B (en) MWV model-based frequency domain multi-channel reflection coefficient joint inversion method and system
CN111856559B (en) Multi-channel seismic spectrum inversion method and system based on sparse Bayes learning theory
CN112526599A (en) Wavelet phase estimation method and system based on weighted L1 norm sparsity criterion
US11372122B2 (en) High-resolution processing method for seismic data based on inverse multi-resolution singular value decomposition
CN109655916B (en) Method and system for separating effective waves and multiples in seismic data
CN116068644A (en) Method for improving resolution and noise reduction of seismic data by using generation countermeasure network
CN111239805B (en) Block constraint time-lapse seismic difference inversion method and system based on reflectivity method
CN109991657B (en) Seismic data high-resolution processing method based on inverse binary recursive singular value decomposition
CN112578439B (en) Seismic inversion method based on space constraint
Wu et al. Iterative deblending based on the modified singular spectrum analysis
CN112363217A (en) Random noise suppression method and system for seismic data
CN113009564A (en) Seismic data processing method and device
CN110687605A (en) Improved K-SVD algorithm-based algorithm analysis application in seismic signal processing
CN115327624B (en) Inversion method and inversion system for seismic wavelets and reflection coefficients

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant