A near-surface modeling method and system constrained by geological structure

Abstract

The present invention provides a near-surface modeling method and system constrained by geological structure, belonging to the field of geophysical exploration. The method first extracts structural information based on digital structural information obtained in the early stage or seismic data after migration, and then introduces constraint terms to constrain the objective function in the near-surface modeling, and obtains a high-precision near-surface velocity model by solving the constrained near-surface modeling equation. The present invention solves the problems and difficulties of low precision, low resolution, and velocity trend inconsistent with the actual situation of the velocity model obtained by the conventional near-surface modeling method. By introducing the structural constraint regularization term, a high-precision and high-resolution near-surface velocity model is obtained more stably and efficiently, providing a high-quality near-surface velocity model for subsequent migration imaging. The present invention has good application prospects in the field of seismic signal processing.

Description

Technical Field

[0001] The present invention belongs to the field of geophysical exploration, and in particular relates to a near-surface modeling method and system constrained by geological structure. Background Art

[0002] With the continuous development of oil and gas exploration in China, the focus of onshore seismic exploration has gradually shifted to the complex and undulating surface areas such as mountains, Gobi Desert, loess plateau, and desert in the west. In these undulating surface exploration areas, the rapid changes in near-surface elevation and velocity have a serious impact on the acquisition, processing, and interpretation of seismic data. The exploration areas in western my country and Sichuan have the dual complex characteristics of complex surface and complex underground structure, which is the hot spot of current seismic exploration processing research. When processing seismic data in exploration areas with complex near-surface characteristics, an accurate near-surface velocity model is the premise and foundation for achieving good results in high-precision migration imaging, and is the key to implementing the structure.

[0003] The traditional approach to modeling near-surface velocity is to first pick the first arrival, then build an initial velocity model, perform ray tracing based on the initial model, calculate travel time residuals, and then update the near-surface velocity. This process is repeated over and over again to ultimately obtain a near-surface model. For example, Chinese patent publication CN102590864A discloses a two-step tomographic inversion near-surface modeling method, the specific steps of which include: ① picking up and entering the first arrival of small refraction; ② performing tomographic inversion on the first arrival of small refraction, and outputting interpretation results that can reflect the fine changes in the velocity of the extremely shallow layer; ③ picking up the first arrival time of the cannon record; ④ using the Kriging method to interpolate the inversion results of the small refraction with the micro-logging interpretation results in the work area to establish a near-surface velocity body of the extremely shallow layer; ⑤ establishing a grid division and constrained inversion initial velocity model, replacing the shallow part of the initial model generated by the first arrival of the cannon with the extremely shallow near-surface velocity body obtained in the previous step, and supplementing the extremely shallow velocity loss caused by the cannon due to the excessive minimum offset distance; ⑥ generating a constraint weight field; ⑦ inverting the near-surface velocity model through the constrained tomography method, picking a high-speed layer top interface based on the velocity-depth model obtained by tomographic inversion, and calculating the static correction of the shot point and the detection point.

[0004] However, the tomographic equations in conventional near-surface modeling are ill-conditioned, so regularization constraints and large smoothing are required, which leads to problems such as low near-surface velocity accuracy and low resolution, and even the velocity trend is opposite to the structural direction. Summary of the invention

[0005] The purpose of the present invention is to solve the problems existing in the above-mentioned prior art and to provide a near-surface modeling method and system with geological structure constraints. When solving the tomographic equation, geological structure constraints are introduced to make the near-surface velocity more consistent with the actual structure, thereby improving the accuracy and stability of the inversion.

[0006] The present invention is achieved through the following technical solutions:

[0007] The first aspect of the present invention provides a near-surface modeling method constrained by geological structure. The method first extracts structural information based on digital structural information obtained in advance or seismic data after migration, and then introduces constraint terms into the objective function in the near-surface modeling to constrain the objective function. By solving the constrained near-surface modeling equations, a high-precision near-surface velocity model is obtained.

[0008] A further improvement of the present invention is that the constraint term is a construction constraint regularization term.

[0009] A further improvement of the method of the present invention is that the method comprises:

[0010] (1) Input digital structural information or migrated seismic data, prepare the initial near-surface velocity v0, and set the number of iterations N, i = 0;

[0011] (2) extracting structural tensors from input digital structural information or migrated seismic data;

[0012] (3) Obtain the tomographic kernel function K by performing ray tracing based on the current near-surface velocity;

[0013] (4) Calculate the difference Δt between the theoretical travel time and the actual travel time;

[0014] (5) adding a construction constraint regularization term to the tomographic kernel function K to obtain a constrained tomographic equation, then solving the constrained tomographic equation to obtain a slowness update, and obtaining a speed of a new round of iteration according to the slowness update;

[0015] (6) Determine whether i is equal to N. If not, then i=i+1 and return to step (3). If yes, then go to step (7);

[0016] (7) Output the final velocity v.

[0017] A further improvement of the method of the present invention is that the digital structural information and the offset seismic data in step (1) are both three-dimensional data volumes.

[0018] A further improvement of the method of the present invention is that the operation of step (3) comprises:

[0019] The path information from the shot point to the detection point is calculated by the current near-surface velocity, and then the tomographic kernel function K is obtained according to the path information.

[0020] A further improvement of the method of the present invention is that the tomographic kernel function K in step (3) is a two-dimensional matrix, in which the number of rows represents the number of rays and the number of columns represents the number of velocity grids.

[0021] A further improvement of the method of the present invention is that the theoretical travel time in step (4) is the travel time from the shot point to the detection point calculated according to the current velocity model, and the actual travel time is the travel time obtained by field observation;

[0022] Use the following formula to calculate the difference between the theoretical travel time and the actual travel time

[0023] Δt = actual travel time - theoretical travel time.

[0024] A further improvement of the method of the present invention is that the operation of step (5) comprises:

[0025] Adding a construction constraint regularization term to the tomographic kernel function K yields the constrained tomographic equation:

[0026]

[0027] Among them, L(s) is the error function in the sense of the two norm, Δs is the slowness update amount, ε is the regularization parameter, is the construction constraint regularization term; D is the preconditioner, as follows:

[0028]

[0029] Where I represents the identity matrix, is the Laplacian operator and T is the structure tensor.

[0030] In a second aspect of the present invention, a near-surface modeling system with geological structure constraints is provided, the system comprising: a memory, a processor, and a computer program stored in the memory, wherein the computer program executes the following steps when executed by the processor:

[0031] (1) Input digital structural information or migrated seismic data, prepare the initial near-surface velocity v0, and set the number of iterations N, i = 0;

[0032] (2) extracting structural tensors from input digital structural information or migrated seismic data;

[0033] (3) Obtain the tomographic kernel function K by performing ray tracing based on the current near-surface velocity;

[0034] (4) Calculate the difference Δt between the theoretical travel time and the actual travel time;

[0035] (5) adding a construction constraint regularization term to the tomographic kernel function K to obtain a constrained tomographic equation, then solving the constrained tomographic equation to obtain a slowness update, and obtaining a speed of a new round of iteration according to the slowness update;

[0036] (6) Determine whether i is equal to N. If not, then i=i+1 and return to step (3). If yes, then go to step (7);

[0037] (7) Output the final velocity v.

[0038] According to a third aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores at least one computer-executable program, and when the at least one program is executed by the computer, the computer executes the steps in the near-surface modeling method with geological structure constraints.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention solves the problems of low accuracy, low resolution, and velocity trend inconsistent with the actual situation obtained by conventional near-surface modeling methods. By introducing structural constraint regularization, a high-precision and high-resolution near-surface velocity model is obtained more stably and efficiently, providing a high-quality near-surface velocity model for subsequent migration imaging.

[0040] The present invention has good application prospects in the field of seismic signal processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flowchart of the steps of the method of the present invention.

[0042] Figure 2-1 is the true velocity model;

[0043] Figure 2-2 is the initial model;

[0044] Figure 2-3 is the near-surface velocity calculated by the conventional near-surface modeling method;

[0045] Figure 2-4 is the near-surface velocity calculated by the method of the present invention;

[0046] Figure 3 It is the trend of the average non-zero number of sparse coefficients during the iteration process. DETAILED DESCRIPTION

[0047] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0048] In order to solve the problems of low modeling accuracy, low resolution and low stability of conventional near-surface modeling methods, as well as the velocity trend that does not match the actual situation and inaccurate structural characterization, the present invention utilizes prior information such as geological outcrops and small refractions (digital structural information is obtained from geological outcrops) and converts it into digital structural information. In the process of near-surface modeling, the velocity model is constrained to obtain a high-precision velocity model that conforms to the near-surface geological structure, thereby achieving better results in subsequent mid-deep velocity modeling and seismic data migration imaging.

[0049] The method of the present invention inputs digital structural information or offset imaging profiles, calculates structural tensors for the input data to extract structural features, introduces structural constraint regularization in the process of near-surface modeling, improves the accuracy and stability of inversion, and ultimately obtains a high-precision near-surface velocity model.

[0050] The embodiments of the method of the present invention are as follows:

[0051] [Example 1]

[0052] like Figure 1 As shown, the method of the present invention comprises:

[0053] (1) Input digital structural information or migrated seismic data Y, prepare the initial near-surface velocity v0, and set the number of iterations N, i = 0;

[0054] Either digital structural information or migrated seismic data Y can be used, wherein digital structural information is structural information artificially depicted by observing geological outcrops, and migrated seismic data is structural information obtained based on data drive. Both digital structural information and seismic data are three-dimensional data bodies.

[0055] Generally, the initial velocity uses a gradient model with a velocity field of mxnxo, where m is the depth z direction, n is the horizontal y direction, and o is the horizontal x direction. At z=0, the velocity is v1, the velocity gradient is a, and the velocity function is v(x,y,z)=v1+z*a.

[0056] (2) extracting structural tensors from input digital structural information or migrated seismic data;

[0057] The method for extracting the structure tensor adopts the existing method, which is briefly introduced as follows:

[0058] For a three-dimensional data volume, the structure tensor T(x) of a certain point can be represented by a 3x3 symmetric positive definite matrix:

[0059]

[0060] Where dx, dy, and dz are the gradients of the data along the x, y, and z directions respectively. Through the eigenvalue decomposition of the matrix, the above formula can be expressed as

[0061] T(x)=λ u uu T +λ v vv T +λ w ww T (2)

[0062] where λ u , v and λ w is a positive real number distributed between 0 and 1, representing the eigenvalue of the matrix T(x), and u, v, and w are the corresponding eigenvectors. Usually, λ is defined u ≥λ v ≥λ w ≥0, in this case, the eigenvector u represents the direction with the largest gradient, that is, the direction perpendicular to the structural dip. The eigenvectors v and w represent the directions parallel to the structure, respectively. The larger the λ, the larger the smoothing scale along the corresponding characteristic direction.

[0063] (3) Obtain the tomographic kernel function K by performing ray tracing based on the current near-surface velocity;

[0064] The existing method is used for implementation, which is briefly introduced as follows: velocity modeling is a process of continuous updating and iteration. The current near-surface velocity is vi. Ray tracing is to calculate the path from the shot point to the detection point through the current near-surface velocity, and then obtain the tomographic kernel function K based on the path information.

[0065] The tomographic kernel function K is the ray path obtained by ray tracing. K is a two-dimensional matrix, in which the number of rows represents the number of rays and the number of columns represents the number of velocity grids. Its physical meaning is the length of the ray passing through each velocity grid, so K can be calculated based on the ray path.

[0066] (4) Calculate the difference Δt between the theoretical travel time and the actual travel time;

[0067] The theoretical travel time is the travel time from the shot point to the detection point calculated based on the current velocity model. The real travel time refers to the travel time obtained by field observation. Δt = real travel time - theoretical travel time, that is, the difference between the picked travel time and the calculated travel time.

[0068] (5) A constrained regularization term is added to the tomographic kernel function K to obtain a constrained tomographic equation, and then the constrained tomographic equation is solved to obtain a slowness update amount, and the speed of a new round of iteration is obtained according to the slowness update amount.

[0069] The details are as follows:

[0070] The analytical solution of the anisotropic diffusion equation is introduced, and its implicit expression is the following partial differential equation:

[0071]

[0072] Where f(x) is the input seismic data, g(x) is the output seismic data, α is a positive real number used to control the smoothing strength; T(x) is the structural tensor, which contains the gradient information of the strike, dip and normal of the structure.

[0073] Introducing the finite difference approximation, the preconditioner D can be written as:

[0074]

[0075] Where I represents the identity matrix, is the Laplace operator, is a symmetric matrix, and overall represents a smoothing of the solution along the constructed gradient.

[0076] The conventional near-surface modeling formula is as follows:

[0077]

[0078] Where K is the tomographic kernel function, Δt is the difference between the picked travel time and the calculated travel time. The present invention introduces a construction constraint regularization term to obtain the constrained tomographic equation, as shown in formula (6):

[0079]

[0080] Among them, L(s) is the error function in the sense of the two norm, Δs is the slowness update amount to be calculated, ε is the regularization parameter, is the construction constraint regularization term.

[0081] Conventional near-surface modeling is formula (5). In the case of some high-steep structures, the velocity model calculated by the conventional method of formula (5) cannot well characterize the velocity model. In the process of near-surface modeling, the present invention introduces a structural constraint regularization term. After introducing structural information, a near-surface velocity model that is closer to the actual velocity can be obtained.

[0082] By solving formula (6), a high-precision near-surface velocity model can be obtained: Formula (6) is solved by existing methods, such as LSCG, and the least squares conjugate gradient method is used to solve formula (6). L(s) is the error function in the sense of the two norm. By taking the derivative of L(s), setting the derivative = 0, the slowness update is calculated. The new slowness is obtained by the current slowness (the inverse of the current speed) and the slowness update, and the new speed is obtained by the new slowness.

[0083] Each time formula (6) is solved, the slowness (i.e., the inverse of the speed) update amount is obtained, and the speed of a new round of iterations can be obtained based on the slowness update amount.

[0084] (6) Determine whether i is equal to N. If not, then i=i+1 and return to step (3). If yes, then go to step (7);

[0085] (7) Output the final velocity v.

[0086] The present invention also provides a near-surface modeling system constrained by geological structure, and the embodiments of the system are as follows:

[0087] [Example 2]

[0088] The system comprises: a memory, a processor, and a computer program stored in the memory, wherein the computer program executes the following steps when executed by the processor:

[0089] (1) Input digital structural information or migrated seismic data, prepare the initial near-surface velocity v0, and set the number of iterations N, i = 0;

[0090] (2) extracting structural tensors from input digital structural information or migrated seismic data;

[0091] (3) Obtain the tomographic kernel function K by performing ray tracing based on the current near-surface velocity;

[0092] (4) Calculate the difference Δt between the theoretical travel time and the actual travel time;

[0093] (5) adding a construction constraint regularization term to the tomographic kernel function K to obtain a constrained tomographic equation, then solving the constrained tomographic equation to obtain a slowness update, and obtaining a speed of a new round of iteration according to the slowness update;

[0094] (6) Determine whether i is equal to N. If not, then i=i+1 and return to step (3). If yes, then go to step (7);

[0095] (7) Output the final velocity v.

[0096] The digital construction information and the offset seismic data in step (1) are both three-dimensional data volumes.

[0097] The operation of step (3) includes: calculating the path information from the shot point to the detection point through the current near-surface velocity, and then obtaining the tomographic kernel function K according to the path information.

[0098] The tomographic kernel function K in step (3) is a two-dimensional matrix, in which the number of rows represents the number of rays and the number of columns represents the number of velocity grids.

[0099] The theoretical travel time in step (4) is the travel time from the shot point to the detection point calculated according to the current velocity model, and the real travel time is the travel time obtained by field observation;

[0100] Use the following formula to calculate the difference between the theoretical travel time and the actual travel time

[0101] Δt = actual travel time - theoretical travel time.

[0102] The operation of step (5) includes:

[0103] Adding a construction constraint regularization term to the tomographic kernel function K yields the constrained tomographic equation:

[0104]

[0105] Among them, L(s) is the error function in the sense of the two norm, Δs is the slowness update amount, ε is the regularization parameter, is the construction constraint regularization term; D is the preconditioner, as follows:

[0106]

[0107] Where I represents the identity matrix, is the Laplacian operator and T is the structure tensor.

[0108] The present invention also provides a computer-readable storage medium, and an embodiment of the computer-readable storage medium is as follows:

[0109] [Example 3]

[0110] The computer-readable storage medium stores at least one computer-executable program, and when the at least one program is executed by the computer, the computer executes the steps in the near-surface modeling method with geological structure constraints.

[0111] The embodiments of the present invention are as follows:

[0112] [Example 4]

[0113] The results of damped dictionary learning and conventional K-SVD dictionary learning are compared. Figure 2-1 is the dictionary learned by the damped dictionary, Figure 2-2 is the dictionary obtained by the conventional K-SVD algorithm, Figure 3 is the average non-zero number of sparse coefficients in the iteration process, where the solid line is the method of the present invention and the dotted line is the traditional K-SVD algorithm. Figure 2-1 and Figure 2-2 It can be seen from the comparison that the dictionary obtained by the improved dictionary learning algorithm of the present invention contains fewer noise atoms. On the other hand, Figure 3 The comparison of the average non-zero number also reflects that the improved algorithm can represent seismic data more sparsely, indicating that the improved algorithm can better adapt to seismic data.

[0114] right Figures 2-1 to 2-4 By comparison, it can be found that the inversion results obtained by near-surface modeling based on structural constraints are generally better than those of conventional near-surface modeling; especially in the velocity inversion in the valley, the advantages of structural constraints are clearly reflected. Near the right near-surface, near-surface modeling based on structural constraints can well describe the near-surface velocity changes with higher resolution. Figure 3 The comparison of the curves extracted in the depth direction shows the ability of the near-surface modeling based on structural constraints to effectively restore the true velocity on the near-surface, which illustrates the advantages of the near-surface modeling algorithm based on structural constraints.

[0115] Finally, it should be explained that the above technical solution is only one implementation method of the present invention. For those skilled in the art, it is easy to make various types of improvements or modifications based on the application methods and principles disclosed in the present invention, and it is not limited to the method described in the above specific implementation method of the present invention. Therefore, the method described above is only preferred and does not have a restrictive meaning.

Claims

1. A near-surface modeling method constrained by geological structure, characterized by: The method first extracts structural information based on digital structural information obtained in the early stage or seismic data after migration, and then introduces constraint terms into the objective function in near-surface modeling to constrain it, and obtains a high-precision near-surface velocity model by solving the constrained near-surface modeling equations; The method comprises: (1) Input digital structural information or migrated seismic data, prepare the initial near-surface velocity v0, and set the number of iterations N, i = 0; (2) extracting structural tensors from input digital structural information or migrated seismic data; (3) Obtain the tomographic kernel function K by performing ray tracing based on the current near-surface velocity; (4) Calculate the difference Δt between the theoretical travel time and the actual travel time; (5) adding a construction constraint regularization term to the tomographic kernel function K to obtain a constrained tomographic equation, then solving the constrained tomographic equation to obtain a slowness update, and obtaining a speed of a new round of iteration according to the slowness update; (6) Determine whether i is equal to N. If not, then i=i+1 and return to step (3). If yes, then go to step (7); (7) Output final velocity v; The operation of step (5) includes: Adding a construction constraint regularization term to the tomographic kernel function K yields the constrained tomographic equation: Among them, L(s) is the error function in the sense of the two norm, Δs is the slowness update amount, ε is the regularization parameter, is the construction constraint regularization term; D is the preconditioner, as follows: Where I represents the identity matrix, is the Laplacian operator and T is the structure tensor.

2. The near-surface modeling method constrained by geological structure according to claim 1, characterized in that: The constraint term is a construction constraint regularization term.

3. The near-surface modeling method constrained by geological structure according to claim 1, characterized in that: The digital construction information and the offset seismic data in step (1) are both three-dimensional data volumes.

4. The near-surface modeling method constrained by geological structure according to claim 1, characterized in that: The operation of step (3) includes: The path information from the shot point to the detection point is calculated by the current near-surface velocity, and then the tomographic kernel function K is obtained according to the path information.

5. The near-surface modeling method constrained by geological structure according to claim 1, characterized in that: The tomographic kernel function K in step (3) is a two-dimensional matrix, in which the number of rows represents the number of rays and the number of columns represents the number of velocity grids.

6. The near-surface modeling method constrained by geological structure according to claim 1, characterized in that: The theoretical travel time in step (4) is the travel time from the shot point to the detection point calculated according to the current velocity model, and the real travel time is the travel time obtained by field observation; Use the following formula to calculate the difference between the theoretical travel time and the actual travel time: Δt = actual travel time - theoretical travel time.

7. A near-surface modeling system constrained by geological structures, characterized in that: The system comprises: a memory, a processor, and a computer program stored in the memory, wherein the computer program executes the following steps when executed by the processor: (1) Input digital structural information or migrated seismic data, prepare the initial near-surface velocity v0, and set the number of iterations N, i = 0; (2) extracting structural tensors from input digital structural information or migrated seismic data; (3) Obtain the tomographic kernel function K by performing ray tracing based on the current near-surface velocity; (4) Calculate the difference Δt between the theoretical travel time and the actual travel time; (5) adding a construction constraint regularization term to the tomographic kernel function K to obtain a constrained tomographic equation, then solving the constrained tomographic equation to obtain a slowness update, and obtaining a speed of a new round of iteration according to the slowness update; (6) Determine whether i is equal to N. If not, then i=i+1 and return to step (3). If yes, then go to step (7); (7) Output final velocity v; The operation of step (5) includes: Adding a construction constraint regularization term to the tomographic kernel function K yields the constrained tomographic equation: Among them, L(s) is the error function in the sense of the two norm, Δs is the slowness update amount, ε is the regularization parameter, is the construction constraint regularization term; D is the preconditioner, as follows: Where I represents the identity matrix, is the Laplacian operator and T is the structure tensor.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one computer-executable program, and when the at least one program is executed by the computer, the computer executes the steps in the near-surface modeling method with geological structure constraints as described in any one of claims 1-6.

Images