Well-ground electromagnetic inversion method, device and storage medium

Abstract

The present application relates to the technical field of geophysical exploration, in particular to a borehole-ground electromagnetic inversion method, device and storage medium. The method comprises: setting prior information; calculating borehole-ground electromagnetic response; calculating model data fitting term; calculating model regularization term containing constraint matrix based on minimum support gradient; solving gradient and Hessian matrix of objective function; iteratively solving to obtain model increment; determining update step length to obtain updated model; using the updated model to obtain borehole-ground electromagnetic response and calculate data fitting difference; judging whether the data fitting difference meets the convergence condition and whether the iteration number k meets the convergence condition; if one of them meets the condition, the iteration is exited and the model is output. The present application enhances the ability of borehole-ground electromagnetic to depict deep formation interface, improves the precision of borehole-ground electromagnetic inversion of deep formation interface, and can more intuitively and clearly invert the interface between formations at different depths.

Description

Technical Field

[0001] This invention relates to the field of geophysical exploration technology, specifically to a well-to-surface electromagnetic inversion method, apparatus, and storage medium. Background Technology

[0002] Geophysical exploration, or geophysical prospecting for short, refers to the study and observation of changes in various geophysical fields to detect geological conditions such as lithology and geological structure. Different rock layers that make up the Earth's crust often differ in density, elasticity, electrical conductivity, magnetism, radioactivity, and thermal conductivity. These differences cause corresponding local variations in the geophysical fields. By measuring the distribution and variation characteristics of these physical fields and analyzing them in conjunction with known geological data, the geological characteristics can be inferred. Research and practical experience in the industry have revealed that traditional surface exploration methods, due to the presence of complex influences such as air waves, struggle to accurately obtain relevant electromagnetic response signals from underground strata.

[0003] Well-to-surface electromagnetic (BTS) methods have seen some development in the study of onshore resources, achieving certain results in areas such as oil and gas reservoir boundary delineation. They have also been applied in production practice in China, and some theoretical research and practical exploration have established a relevant current foundation. The well-to-surface electromagnetic method is an electromagnetic exploration method that utilizes existing mines or wells. It involves setting up a long vertical conductor in the well and transmitting a high-power alternating current, then receiving the resulting electromagnetic field response on the surface.

[0004] Well-to-surface electromagnetic (BTS) methods place the transmitter directly underground, overcoming near-surface interference and shortening the distance between the transmitter and the target, offering advantages such as a wide exploration range and great exploration depth. However, in practical applications, the spatial resolution of BTS is limited, especially in the detection of deep targets. The propagation of electromagnetic waves underground is affected by the conductivity of the medium, causing the resolution to decrease with increasing depth, resulting in poor characterization of deep strata interfaces. Summary of the Invention

[0005] To address the aforementioned technical problems in existing technologies, this invention provides a well-to-surface electromagnetic inversion method, apparatus, and storage medium. Based on a minimum support gradient model constraint strategy, and combined with an interface constraint matrix constructed from high-precision interface information extracted from seismic and well logging data, a model constraint term containing well-seismic interface information is built to constrain the well-to-surface electromagnetic inversion. This enhances the well-to-surface electromagnetic inversion's ability to characterize deep formation interfaces, improves the accuracy of deep formation interface inversion, and thus more intuitively and clearly inverts the interfaces between formations at different depths.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] On one hand, the present invention provides a well-to-ground electromagnetic inversion method, comprising:

[0008] S1. Set prior information;

[0009] S2. Calculate the borehole - surface electromagnetic response F(m0) of m0 at all measurement points according to the prior information;

[0010] S3. Calculate the model data fitting term

[0011] S4. Extract interface information according to seismic data and logging data to construct a constraint matrix g, and calculate the model regularization term containing the constraint matrix g based on the minimum support gradient

[0012] S5. Combine the model data fitting term obtained in step S3 and the model regularization term obtained in step S4 to construct an objective function, and solve the gradient g0 and Hessian matrix H0 of the objective function;

[0013] S6. Iteratively solve to obtain the model increment δm0;

[0014] S7. Perform backtracking line search according to the Armijo condition to determine the update step size α0, and obtain the updated model m1;

[0015] S8. Forward - simulate using the updated model m1 to obtain the borehole - surface electromagnetic response F(m1), and calculate the data fitting difference RMS;

[0016] S9. Judge whether the data fitting difference RMS satisfies the convergence condition RMS < objRMS. If it is satisfied, go to step S12; if not, go to step S10;

[0017] S10. Judge whether the number of iterations k satisfies k > maxiter. If it is satisfied, go to step S12; if not, go to step S11, let k = k + 1, m0 = m1, update the regularization parameter, and return to step S2 until the iteration end condition is met and output the model m k .

[0018] Furthermore, in step S1, setting the prior information includes: inputting borehole - surface electromagnetic observation data d obs , the initial inversion model m0, the reference model m ref and the borehole - surface electromagnetic inversion parameters.

[0019] Moreover, the borehole - surface electromagnetic inversion parameters include, but are not limited to: inversion grid, maximum number of Gauss - Newton iterations, maximum number of linear search times, number of pre - conditioned conjugate gradient method iterations, target fitting difference, initial regularization parameter, minimum support gradient focusing parameter.

[0020] Furthermore, in step S2, the well-to-ground electromagnetic response F(m0) at all measuring points m0 is calculated based on prior information, specifically including: performing a three-dimensional forward simulation on the initial inversion model m0 in step S1.

[0021] Furthermore, a three-dimensional forward modeling simulation is performed on the initial inversion model m0, specifically as follows:

[0022] Based on the positive harmonic time e iωt The frequency domain Maxwell equations are expressed as follows:

[0023]

[0024] in, Here, is the spatial gradient operator, i represents the imaginary unit, E is the electric field strength (V / m), β is the magnetic flux density (T), ω is the angular frequency (rad / s); σ is the dielectric conductivity (S / m), μ is the magnetic permeability (H / m), and is taken as the free permeability; J0 is the current density of the external excitation source (A / m). 2 );

[0025] The pseudo-finite volume method discretization is performed using a Yee-style staggered mesh. The electric field E is defined at the center of the element edge, the magnetic induction intensity B is defined at the center of the element face, and the dielectric conductivity and permeability are defined at the element center. After discretizing equations (1) and (2), we obtain:

[0026] CURL·E=-iωB (3)

[0027] CURL T ·M fμ ·B=M eσ ·E+S (4)

[0028] Where CURL is the discrete form of the curl operator, M fμ M is the average matrix of magnetic permeability. eσ Let S be the average conductivity matrix, and S be the source term matrix.

[0029] Substituting equation (3) into equation (4) and eliminating the magnetic induction intensity B, we obtain the discrete form of the Helmholtz equation for the electric field vector:

[0030] (CURL T ·M fμ ·CURL+iω·M eσ E = -iωS (5).

[0031] Furthermore, in step S3, the model data fitting term is calculated based on prior information and the well-to-surface electromagnetic response F(m0). Specifically, this involves combining the observation data d from step S1.obs The well-to-surface electromagnetic response F(m0) obtained in step S2 is used to calculate the model data fitting term.

[0032] Furthermore, the fitting term The calculation method is as follows:

[0033]

[0034] in, This is the data covariance matrix.

[0035] Furthermore, the data covariance matrix for:

[0036]

[0037] Where δr is the relative error of the data, and the constant ∈ is the lower limit of the error of the observed data.

[0038] Furthermore, step S4 specifically includes:

[0039] Based on seismic and well logging data, interface information is extracted to construct a constraint matrix g. The constraint matrix g is then imported into the model constraint terms based on the minimum support gradient to obtain the model regularization term.

[0040] Furthermore, the model regularization term is:

[0041]

[0042] In the above formula, Let ξ be the spatial gradient operator, and ξ be the focusing parameter.

[0043] Furthermore, when constructing the constraint matrix, the values ​​of the elements in the matrix are all between [0,1].

[0044] Furthermore, in step S5, the gradient g0 and the Hessian matrix H0 of the objective function are solved using the Gauss-Newton method.

[0045] Furthermore, the gradient g0 and Hessian matrix H0 of the objective function in step S5 are:

[0046]

[0047] Where m are model parameters. Here is the model covariance matrix, and β is the regularization factor. Let J0 be the sensitivity matrix, and J0 represent the sensitivity matrix of the 0th iteration.

[0048] Furthermore, in step S6, the model increment is obtained by iteratively solving using the preconditional conjugate gradient method.

[0049] Furthermore, the RMS fit difference in step S8 is:

[0050]

[0051] In the above formula, N d The number of observation data. For the i-th observation data, ò i This is the lower bound of the error for the i-th observation.

[0052] Furthermore, in step S11, the regularization parameters are updated in the following manner:

[0053]

[0054] Where, γ N The attenuation factor is fixed, β0 is the initial regularization parameter, and β N Here, c is the regularization parameter after N iterations, where c is a constant and N is the number of iterations.

[0055] Furthermore, the constant c satisfies: c∈(0,1).

[0056] On the other hand, the present invention also provides a well-to-surface electromagnetic inversion device, including an operating system and a display module. The operating system is equipped with a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the above-mentioned well-to-surface electromagnetic inversion method.

[0057] On the other hand, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described well-to-ground electromagnetic inversion method.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] The well-to-surface electromagnetic inversion method provided by this invention is based on the model constraint strategy of minimum support gradient. It combines interface information mentioned in seismic and well logging data to construct an interface constraint matrix and construct model constraint terms containing well-seismic interface information to constrain the well-to-surface electromagnetic inversion. This enhances the ability of well-to-surface electromagnetic inversion to characterize deep strata interfaces, improves the accuracy of well-to-surface electromagnetic inversion of deep strata interfaces, and can more intuitively and clearly invert interfaces between strata at different depths. Attached Figure Description

[0060] Figure 1 This is a flowchart of the well-to-ground electromagnetic inversion method of the present invention.

[0061] Figure 2 This is a schematic diagram of the model used to extract interface information in the embodiment.

[0062] Figure 3 This is a schematic diagram of the constraint matrix constructed in the embodiment.

[0063] Figure 4 The image shown is a slice of the CMSG 3D model inversion in the embodiment, where y represents depth, x represents horizontal distance, and z represents vertical distance. Detailed Implementation

[0064] The technical solution of the present invention will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are not all embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0065] It should be noted that, unless otherwise specifically stated, the relative arrangement and numerical expressions of the components and steps described in these embodiments should not be construed as limiting the scope of the invention.

[0066] The following description of exemplary embodiments is merely illustrative and is not intended to limit the invention or its application or use in any way. Techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail herein, but where applicable, such techniques, methods, and apparatus should be considered part of this specification.

[0067] This invention provides a well-to-surface electromagnetic inversion method, such as... Figure 1 As shown, it includes:

[0068] S1. Set prior information; specifically including: inputting well-to-ground electromagnetic observation data d obs Initial inversion model m0, reference model m ref And well-to-surface electromagnetic inversion parameters. The well-to-surface electromagnetic inversion parameters include, but are not limited to: inversion grid, maximum number of Gauss-Newton iterations (maxiter), maximum number of linear search iterations, number of iterations of the preconditional conjugate gradient method, target fit difference, initial regularization parameter, and minimum support gradient focusing parameter.

[0069] S2. Calculate the well-to-surface electromagnetic response F(m0) at all measuring points m0 based on prior information; specifically, this includes performing a three-dimensional forward simulation of the initial inversion model m0 from step S1:

[0070] Starting from Maxwell's equations in the frequency domain, for most propagation media (air, seawater, and strata), the influence of displacement current is usually negligible within the frequency range used in geophysical controlled-source electromagnetic methods.

[0071] Based on the positive harmonic time e iωt The frequency domain Maxwell equations are expressed as follows:

[0072]

[0073] in, Let be the spatial gradient operator, i denote the imaginary unit, E be the electric field strength (V / m), B be the magnetic flux density (T), ω be the angular frequency (rad / s); σ be the dielectric conductivity (S / m), μ be the magnetic permeability (H / m), and ω be the free permeability; J0 be the current density of the external excitation source (A / m). 2 );

[0074] The pseudo-finite volume method discretization is performed using a Yee-style staggered mesh. The electric field intensity E is defined at the center of the element edge, the magnetic induction intensity B is defined at the center of the element face, and the dielectric conductivity and permeability are defined at the element center. After discretizing equations (1) and (2), we obtain:

[0075] CURL·E=-iωB (3)

[0076] CURL T ·M fμ ·B=M eσ ·E+S (4)

[0077] Where CURL is the discrete form of the curl operator, M fμ M is the average matrix of magnetic permeability. eσ Let S be the average conductivity matrix, and S be the source term matrix.

[0078] Substituting equation (3) into equation (4) and eliminating the magnetic induction intensity B, we obtain the discrete form of the Helmholtz equation for the electric field vector:

[0079] (CURL T ·M fμ ·CURL+iω·M eσ E = -iωS (6).

[0080] S3. Based on prior information and the well-to-surface electromagnetic response F(m0), calculate the model data fitting term.

[0081] Specifically, this involves combining the observation data d from step S1. obs The well-to-surface electromagnetic response F(m0) obtained in step S2 is used to calculate the model data fitting term. Fitting term The calculation method is as follows:

[0082]

[0083] in, This is the data covariance matrix. for:

[0084]

[0085] Where δr is the relative error of the data, and the constant ∈ is the lower limit of the error of the observed data.

[0086] S4. Based on seismic data and well logging data, such as Figure 2 A schematic diagram of the model used to extract interface information. The constraint matrix g is constructed by extracting interface information, as shown below. Figure 3 As shown, the regularization term of the model containing the constraint matrix g is calculated based on the minimum support gradient.

[0087] like Figure 4 As shown, the specific expression for the model constraint term based on the minimum support gradient (MSG) can be expressed as:

[0088]

[0089] in, ξ is the spatial gradient operator, and ξ is the focusing parameter used to control the sharpness or sparsity of the solution.

[0090] After processing and interpreting seismic and well logging data, relevant interface information is obtained. This interface information is then extracted to construct a constraint matrix g. The constraint matrix g is imported into the model constraint terms based on the minimum support gradient to obtain the model regularization term.

[0091]

[0092] When constructing the constraint matrix, all elements in the matrix take values ​​between [0,1]. Furthermore, based on the processing and interpretation of seismic and well logging data, relevant interface information is obtained. Values ​​are assigned at interface locations; matrix elements asymptotically increase to 1 near the interface and asymptotically decrease to 0 further away from the interface. Where there is no interface information, the matrix elements have a value of 0. This value assignment method ensures that g is 0 where there is no interface information, consistent with the traditional MSG model constraint inversion method, guaranteeing that constraint weights are added only at interface locations.

[0093] S5. Fitting terms for the model data obtained in step S3 and the model regularization term obtained in step S4 Construct the objective function and use the Gauss-Newton method to solve for the gradient g0 and the Hessian matrix H0 of the objective function;

[0094] The gradient g0 and Hessian matrix H0 of the objective function are:

[0095]

[0096] Where m are model parameters. is the model covariance matrix, β is the regularization factor, is the first-order derivative of the forward response of the current model with respect to the model parameters, also known as the sensitivity matrix.

[0097] S6. Use the preconditioned conjugate gradient method to iteratively solve for the model increment δm0;

[0098] The specific expression of the normal equation of the Gauss-Newton method can be expressed as:

[0099] H0δm0 = -g0 (12)

[0100] In the Gauss-Newton method, to obtain model updates, the normal equation needs to be solved in each iteration, and it is necessary to explicitly calculate and store the sensitivity matrix and the Hessian matrix. The solution of the sensitivity matrix requires the use of all observation data, and the explicit solution involves a large number of forward calculations. For three-dimensional CSEM inversion with multiple frequencies and multiple field sources, a large amount of computing memory and operation time are required, and existing computing resources often cannot meet the requirements. To avoid explicitly calculating and storing the sensitivity matrix, the preconditioned conjugate gradient method (PCG) or LSQR method is usually used to iteratively solve the equation.

[0101] Use the preconditioned conjugate gradient method (PCG) to iteratively solve the normal equation, which only involves solving and storing the product of the sensitivity matrix and the model vector, and the product of the transpose of the sensitivity matrix and the data vector.

[0102] S7. Perform backtracking line search according to the Armijo condition to determine the update step size α0 and obtain the updated model m1;

[0103] After obtaining the search directions for multiple iterations, line search is still required to obtain the iterative model update:

[0104] m1 = m0 + α0δm0 (13)

[0105] The step size α0 controls the magnitude of the model update. For three-dimensional inversion, exact line search is often impossible due to huge computational consumption, so a non-exact line search method is generally used to determine the update step size α k . The most common strategy is to use the Wolfe condition for non-exact line search to obtain the update step size α k . The Wolfe condition includes the sufficient decrease condition and the curvature condition:

[0106]

[0107] where the constants c1, c2 satisfy 0 < c1 < c2 < 1. In actual inversion, c1 is generally taken as 10 -4For the GN method or QN method, c2 is generally taken as 0.9. The Wolfe condition requires the use of the gradient information of the current model and the model vector of the next iteration, and the computational cost of multiple searches will be very large. Therefore, in practical applications, another commonly used method is the backtracking method, also known as the Armijo line search method. Its core idea is to start testing from the unit step size to see if it satisfies Equation (14). If not, gradually reduce the step size (such as halving the attenuation) until it satisfies the Armijo condition in Equation (14).

[0108] S8. Use the updated model m1 for forward modeling to obtain the borehole-to-surface electromagnetic response F(m1), and calculate the data fitting difference RMS; the fitting difference RMS is as follows:

[0109]

[0110] where N d is the number of observed data,[[]] is the i-th observed data, ò i is the lower error limit of the i-th observed data.[[]]

[0111] S9. Determine whether the data fitting difference RMS satisfies the convergence condition RMS < objRMS, where objRMS is the target fitting difference. If it is satisfied, go to step S12; if not, go to step S10;

[0112] S10. Determine whether the iteration number k satisfies k > maxiter, where maxiter is the maximum number of iterations of the Gauss-Newton method; if it is satisfied, go to step S12; if not, go to step S11;

[0113] S11. Let k = k + 1, m0 = m1, update the regularization parameter, and return to step S2;

[0114] Due to the underdetermination of electromagnetic inversion, in order to solve the optimization problem, it is necessary to select an appropriate regularization parameter in the inversion iteration. The regularization parameter plays a role in balancing the data fitting term and the model constraint term. The selection of the regularization parameter must take into account the stability of the inversion algorithm and the influence of the prior model on the inversion result. In this embodiment, the regularization update adopts the relaxation method to automatically select, and the expression is as follows:

[0115]

[0116] where, γ N is the fixed attenuation factor, β0 is the initial regularization parameter, β N is the regularization parameter after N iterations, c is a constant, and the constant c satisfies: c ∈ (0, 1); N is the current number of iterations of the Gauss-Newton method.[[]]

[0117] S12. Exit the iteration and output the model mk .

[0118] On the other hand, the present invention also provides a well-to-surface electromagnetic inversion device, including an operating system and a display module. The operating system is equipped with a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the above-mentioned well-to-surface electromagnetic inversion method.

[0119] On the other hand, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described well-to-ground electromagnetic inversion method.

[0120] This invention realizes the idea and process of constructing an interface constraint matrix based on the minimum support gradient (MSG) regularization strategy combined with high-precision interface information extracted from seismic and well logging data, forming a model constraint term (CMSG) containing well-seismic interface information to constrain well-to-geomagnetic inversion, thereby improving the data inversion accuracy of the well-to-geomagnetic method.

[0121] Research shows that the well-to-surface electromagnetic inversion method based on well-seismic stratigraphic interface information constraints, when using sufficient and reliable seismic and well logging data, can extract higher-precision interface information to constrain well-to-surface electromagnetic inversion. This enhances the ability of well-to-surface electromagnetic inversion to characterize deep stratigraphic interfaces and improves the accuracy of well-to-surface electromagnetic inversion for deep stratigraphic interfaces. As a result, it can more intuitively and clearly invert the interfaces between strata at different depths, providing a more sufficient basis for future geological interpretation work.

[0122] The above specific embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to examples, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A well-to-surface electromagnetic inversion method, characterized in that, Including: S1. Set prior information; S2. Calculate the borehole - to - surface electromagnetic response F(m0) at all measurement points according to the prior information; S3. Based on prior information and the well-to-surface electromagnetic response F(m0), calculate the model data fitting term. S4. Based on seismic and well logging data, extract interface information to construct a constraint matrix g, and calculate the regularization term of the model containing the constraint matrix g based on the minimum support gradient. S5. Fitting terms for the model data obtained in step S3 and the model regularization term obtained in step S4 Construct the objective function, and solve for the gradient g0 and the Hessian matrix H0 of the objective function; S6. Iteratively solve to obtain the model increment δm0; S7. Perform a backtracking line search according to the Armijo condition to determine the update step size α0, and obtain the updated model m1; S8. Forward - simulate using the updated model m1 to obtain the borehole - to - surface electromagnetic response F(m1), and calculate the data fitting difference RMS; S9. Judge whether the data fitting difference RMS satisfies the convergence condition RMS < objRMS. If it satisfies, enter step S12; if not, enter step S10; S10. Determine if the iteration number k satisfies k > maxiter; if it does, proceed to step S12; if not, proceed to step S11, and set k = k + 1, m0 = m1, update the regularization parameters, and return to step S2, until the iteration termination condition is met and the model m is output. k .

2. The well-to-ground electromagnetic inversion method according to claim 1, characterized in that, Step S1 sets prior information, including: inputting well-to-ground electromagnetic observation data d obs Initial inversion model m0, reference model m ref And well-to-ground electromagnetic inversion parameters.

3. The well-to-surface electromagnetic inversion method according to claim 2, characterized in that, The borehole - to - surface electromagnetic inversion parameters include, but are not limited to: inversion grid, maximum number of Gauss - Newton iterations, maximum number of linear search times, number of pre - conditioned conjugate gradient method iterations, target fitting difference, initial regularization parameter, minimum support gradient focusing parameter.

4. The well-to-surface electromagnetic inversion method according to claim 2, characterized in that, In step S2, calculating the borehole - to - surface electromagnetic response F(m0) at all measurement points according to the prior information specifically includes: performing three - dimensional forward simulation on the initial inversion model m0 in step S1.

5. The well-to-ground electromagnetic inversion method according to claim 4, characterized in that, Performing three - dimensional forward simulation on the initial inversion model m0 specifically is: Based on the positive harmonic time e iωt The frequency domain Maxwell equations are expressed as follows: in, Here, σ is the spatial gradient operator, i represents the imaginary unit, E is the electric field strength, B is the magnetic induction intensity, ω is the angular frequency, σ is the dielectric conductivity, μ is the magnetic permeability, and is taken as the free permeability; J0 is the current density of the external excitation source. Using the Yee - type staggered grid method for mimetic finite - volume discretization. The electric field Ε is defined at the center of the unit edge, the magnetic induction intensity Β is defined at the center of the unit face, and the dielectric conductivity and magnetic permeability are defined at the center of the unit. After discretizing equations (1) and (2), we get: CURL·E=-iωB (3) CURL T ·M fμ ·B=M eσ ·E+S (4) In the above formula, CURL is the discrete form of the curl operator, M fμ M is the average matrix of magnetic permeability. eσ Let S be the average conductivity matrix, and S be the source term matrix. Substitute equation (3) into equation (4) and eliminate the magnetic induction intensity Β to obtain the discrete form of the Helmholtz equation for the electric - field vector: (CURL T ·M fμ ·CURL+iω·M eσ )E=-iωS (5).

6. The well-to-ground electromagnetic inversion method according to claim 2, characterized in that, In step S3, the model data fitting term is calculated based on prior information and the well-to-surface electromagnetic response F(m0). Specifically, this involves combining the observation data d from step S1. obs The well-to-surface electromagnetic response F(m0) obtained in step S2 is used to calculate the model data fitting term.

7. The well-to-surface electromagnetic inversion method according to claim 6, characterized in that, Fitting term The calculation method is as follows: in, This is the data covariance matrix.

8. The well-to-surface electromagnetic inversion method according to claim 7, characterized in that, Data covariance matrix for: where δr is the relative data error, and the constant ò is the lower limit of the error of the observed data.

9. The well-to-ground electromagnetic inversion method according to claim 1, characterized in that, Step S4 specifically includes: Extract interface information according to seismic data and logging data to construct a constraint matrix g, and import the constraint matrix g into the model constraint term based on the minimum support gradient to obtain the model regularization term.

10. The well-to-surface electromagnetic inversion method according to claim 9, characterized in that, The model regularization term is: In the above formula, Let ξ be the spatial gradient operator, and ξ be the focusing parameter.

11. The well-to-surface electromagnetic inversion method according to claim 9, characterized in that, When constructing the constraint matrix, the values of the elements in the matrix are all between [0, 1].

12. The well-to-surface electromagnetic inversion method according to claim 1, characterized in that, In step S5, use the Gauss - Newton method to solve the gradient g0 and Hessian matrix H0 of the objective function.

13. The well-to-surface electromagnetic inversion method according to claim 12, characterized in that, The gradient g0 and Hessian matrix H0 of the objective function in step S5 are: Where m are model parameters. Let be the model covariance matrix, and β be the regularization factor; d represents the sensitivity matrix; J0 represents the sensitivity matrix of the 0th iteration; d obs For observation data, This is the data covariance matrix.

14. The well-to-surface electromagnetic inversion method according to claim 1, characterized in that, In step S6, adopt the pre - conditioned conjugate gradient method to iteratively solve to obtain the model increment.

15. The well-to-surface electromagnetic inversion method according to claim 1, characterized in that, The fitting difference RMS in step S8 is: In the above formula, N d The number of observation data. For the i-th observation data, ò i Let F(m) be the lower limit of error for the i-th observation data, and F(m) be the well-to-ground electromagnetic response.

16. The well-to-surface electromagnetic inversion method according to claim 1, characterized in that, In step S11, the regularization parameter is updated in the following way: Where, γ N The attenuation factor is fixed, β0 is the initial regularization parameter, and β N Here, c is the regularization parameter after N iterations, where c is a constant and N is the number of iterations.

17. A well-to-surface electromagnetic inversion device, comprising an operating system and a display module, wherein the operating system is internally equipped with a processor and a memory, and the memory stores a computer program, characterized in that, When the computer program is executed by a processor, it implements the borehole - to - surface electromagnetic inversion method according to any one of claims 1 - 16.

18. A computer-readable storage medium, characterized in that, A computer program is stored on the storage medium, and when the computer program is executed by a processor, it implements the borehole - to - surface electromagnetic inversion method according to any one of claims 1 - 16.

Images