Baseline shift adaptive robust high resolution inversion method based on electromagnetic gradients

Abstract

The application discloses a baseline drift self-adaptive robust high-resolution inversion method based on electromagnetic gradient, belongs to the technical field of electromagnetic signal inversion, and comprises the following steps: constructing an initial inversion conductivity model; constructing an electromagnetic gradient data residual error objective function based on electromagnetic field gradient data residual error; calculating an inversion gradient data Jacobian matrix; determining parameters required by inversion; and solving an iterative updating model. The application provides a ground-to-air electromagnetic gradient signal inversion method, adopts a data residual error term of electromagnetic field gradient field action, utilizes electromagnetic gradient, avoids the risk of overall resistivity drift, maintains complete utilization of real signals, and significantly improves the resolution capacity of abnormal boundaries.

Description

Technical Field

[0001] This invention belongs to the field of electromagnetic signal inversion technology, and relates to a ground-to-space electromagnetic gradient signal inversion method, especially a baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient. Background Technology

[0002] In electromagnetic exploration, inversion technology is a core method for inferring subsurface electrical structures from measured field data. Based on electromagnetic field propagation theory and combined with numerical simulation, this method solves inverse partial differential equations to deduce the spatial variations of physical properties such as resistivity from the acquired electromagnetic induction response. The inversion process is essentially an optimization problem, requiring a trade-off between fitting the observed data and incorporating model constraints to obtain a stable solution that conforms to both the measured response and geological plausibility. The results are widely used in mineral resource exploration, oil and gas reservoir identification, groundwater detection, and geological structure analysis, serving as a crucial link in transforming geophysical observation data into reliable geological interpretations.

[0003] In electromagnetic exploration field observations, measurement data often exhibit DC bias or slowly changing baseline drift over time. This is the result of multiple factors: for example, during electric field observations, the electrochemical interaction between the grounding electrode and the soil generates polarization potential, causing a stable DC component or a slowly changing drift to be superimposed on the signal. Under complex surface conditions such as dryness, sand, or permafrost, poor electrode contact can introduce additional potential, which gradually drifts with environmental changes. Furthermore, zero-point errors in the receiver and its circuitry accumulate under conditions of temperature fluctuations, device aging, or unstable power supply, causing the baseline to gradually rise or fall. Simultaneously, in active electromagnetic methods, if the clock or phase reference of the transmitting and receiving systems is unstable, it can introduce systematic bias, causing a nearly uniform shift in the data along the entire survey line. If natural electromagnetic fields and industrial power supply interference remain approximately constant during the observation period, they can also be superimposed on the recorded signal in the form of low-frequency or quasi-DC interference, forming additional background shifts. Furthermore, if the DC component or baseline drift is not adequately removed during subsequent data processing, or if the window is improperly selected during frequency domain transformation, low-frequency energy leakage to the target frequency point may occur, resulting in a fixed bias in the extracted frequency data. Due to the combined influence of these factors, DC bias and baseline drift are very common in electromagnetic observation data: in the time domain, this manifests as an overall rise or fall in the signal baseline; in the frequency domain, it manifests as low-frequency energy anomalies or even diffusion to the target frequency point. This type of bias not only causes the apparent resistivity curve to rise or fall overall, producing a static-like effect, but also causes a systematic deviation in background resistivity during inversion, thereby obscuring the true underground electrical structure and severely affecting the reliability of imaging and interpretation.

[0004] Conventional inversion methods directly utilize observed electromagnetic field quantities, reconstructing subsurface resistivity models through weighted least squares and regularization strategies. These methods are mature and widely used. However, in actual observations, receiver zero drift, polarization effects, and system bias often introduce constant baselines. If directly used for inversion, the algorithm often absorbs these biases by making overall adjustments to the model, resulting in a systematic drift in the overall resistivity level and weakening the ability to distinguish anomaly boundaries.

[0005] To address this issue, existing methods mainly fall into several categories: One is the baseline parameter method, which simultaneously solves for the baseline and the subsurface model within the inversion framework. This method is mathematically feasible, requiring only an additional degree of freedom to describe the baseline. However, the baseline may be strongly coupled with the low-frequency / large-scale components of the model, easily leading to non-uniqueness in interpretation and affecting the stability of the results. Another approach is the projection / mean removal method, which introduces a projection matrix into the objective function to filter out constant components. While mathematically simple, the projection matrix can only strictly remove constant components and often fails to completely eliminate slowly changing low-frequency drift. In addition, a common engineering approach is preprocessing filtering to remove the baseline, such as directly shifting the data as a whole or applying a high-pass filter to make the observed curve "zero" in the reference interval. While intuitive, this method is highly dependent on experience and can easily inadvertently weaken the true low-frequency response, thus affecting the recovery of deep structures. Summary of the Invention

[0006] To overcome the problems of parameter coupling, information loss, or signal attenuation in existing methods when processing baselines, this invention provides an electromagnetic joint inversion method based on gradient data.

[0007] According to one aspect of the present invention, a baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient is provided, comprising: constructing an initial inversion conductivity model; constructing a gradient data residual objective function based on electromagnetic field gradient data residuals, wherein the gradient data residuals are the residuals between actual electromagnetic gradient data and simulated gradient data; calculating the inverted electromagnetic gradient data Jacobian matrix using a forward differencing operator, wherein the forward differencing operator is a first-order difference matrix, and the Jacobian matrix is ​​a sensitivity matrix of the observed electromagnetic gradient data to model parameters; and determining the parameters required for inversion, wherein the parameters include regularization coefficients. Reference model term weight control coefficient The reference model weight matrix, gradient roughness weight control coefficient, roughness term weight matrix, and reference model are used. Based on the electromagnetic gradient data residual objective function, and according to the Jacobian matrix and the parameters, the inversion conductivity model is iteratively updated to obtain the inversion result.

[0008] Optionally, the initial inversion conductivity model is the conductivity at each measurement point. The set of elements, where each grid cell corresponds to a conductivity value, is expressed by the following formula:

[0009] ;

[0010] in, For inversion model; Indicates the total number of inversion grids; Indicates the first Conductivity at each grid point.

[0011] Optionally, after constructing the initial inversion conductivity model, the method further includes: defining the electromagnetic response obtained from the forward simulation. , It is a collection of simulated electromagnetic data from various measuring points at different frequencies, defined as follows:

[0012] ;

[0013] in, , , , These represent the simulated electromagnetic datasets for measurement points 1, 2, j, and N, respectively.

[0014] Simulated electromagnetic gradient data are obtained based on simulated electromagnetic data and the forward difference operator. :

[0015] ;

[0016] in, For forward difference operators,

[0017] ;

[0018] It is the set of real numbers; for Total number of data points; electromagnetic gradient response obtained from forward modeling. It is a collection of simulated electromagnetic gradient data at different frequencies from various measuring points, defined as follows:

[0019] ;

[0020] in, , , , These represent the simulated electromagnetic gradient datasets for measurement points 1, 2, j, and N, respectively.

[0021] Optionally, the objective function for the electromagnetic gradient data residuals is defined as:

[0022] ;

[0023] In the formula, The objective function is the residual of the electromagnetic gradient data; For the observed electromagnetic gradient data, The electromagnetic gradient response in forward modeling; This is the difference between the actual gradient data and the simulated gradient data, i.e., the gradient data residual. It is a diagonal weighted matrix that includes the uncertainty of the measured electromagnetic data, and is determined based on the reciprocal of the noise standard deviation of the electromagnetic gradient data at each measurement point.

[0024] Optionally, the Jacobian matrix of the inverted electromagnetic gradient data is calculated using the forward difference operator, including:

[0025] The formula for calculating the Jacobian matrix of the inverted electromagnetic gradient data is:

[0026] ;

[0027] In the formula, This represents the Jacobian matrix of the inverted electromagnetic gradient data; The Jacobian matrix for electromagnetic data; For forward difference operators, This is the interpolation mapping matrix between actual sampling points and grid cells, reflecting the spatial correspondence between the sampling positions of actual data and the model grid; It is a three-dimensional forward coefficient matrix. This is the three-dimensional forward response vector.

[0028] Optionally, iteratively updating the inverse conductivity model based on the Jacobian matrix and the parameters includes:

[0029] No. Step value for the next iteration This is obtained by solving the following equation:

[0030] ;

[0031] In the formula, Indicates the first The model obtained from the first step of calculation and the first step The difference between the models obtained from the first step; For the first The Jacobian matrix of electromagnetic gradient data corresponding to the next iteration model; For the first The regularization coefficient chosen in the next iteration; For the first Simulated gradient data corresponding to the next iteration model; These are the reference model item weight control coefficients; The reference model weight matrix; These are the gradient roughness weight control coefficients; This is the roughness term weight matrix; For the first Inversion model in the next iteration; This is a reference model.

[0032] According to another aspect of the present invention, a computer device is also provided, the computer device including a memory, a processor and a computer program stored in the memory and executable on the processor, the computer program being executed by the processor to implement the steps of the baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient as described above.

[0033] According to another aspect of the present invention, a storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient as described above.

[0034] The beneficial effects of this invention are:

[0035] This method transforms the original electromagnetic field data into electromagnetic gradient data by left-multiplying it by a forward difference operator before the observation data enters the inversion framework. Since the difference operator approximates the uniform constant term within the neighborhood to zero, interference from constant or slowly varying baselines can be naturally eliminated during the gradient data construction stage when the baseline remains essentially consistent between adjacent measurement points or changes slowly. For example, assuming the observation value at measurement point A is 100+5 (100 is the true signal, 5 is the baseline), and the observation value at measurement point B is 110+5 (110 is the true signal, 5 is the same baseline), then the gradient between A and B is (110+5)-(100+5)=110-100=10. The baseline 5 is canceled out in the gradient calculation, and the gradient result only reflects the difference in the true signal. Therefore, this method... There is no need to introduce additional baseline parameters in the inversion process, nor is it necessary to rely on projection or filtering operations. This not only avoids the risk of overall resistivity drift, but also maintains the complete utilization of the real signal. At the same time, gradient data is more sensitive to local contrast and interface abrupt changes (in regions with uniform physical properties, the gradient value is close to zero, while at the isoelectric abrupt change location at the boundary of the anomaly, such as the contact zone between the ore body and the surrounding rock, the gradient response is significantly enhanced). Therefore, inversion based on gradient data can more clearly and sharply characterize the boundary of the anomaly, improve the resolution, and achieve high-precision inversion. In contrast, traditional single-point data has a more obvious smoothing effect on the boundary response and has limited resolution.

[0036] Therefore, the gradient inversion method proposed in this invention does not require explicit modeling or filtering operations, naturally eliminates the influence of baseline drift, and still has robustness and reliability in the presence of a baseline. It also significantly improves the ability to resolve the boundaries of anomalies, providing an efficient and physically consistent new technical approach for electrical imaging under complex geological conditions. Attached Figure Description

[0037] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0038] Figure 1 This is a flowchart of the baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient according to an embodiment of the present invention;

[0039] Figure 2 This is a schematic diagram of the initial 3D model;

[0040] Figure 3 This is a schematic diagram of electromagnetic response data, in which, Figure 3 (a) represents the real part of the electromagnetic response data. Figure 3 (b) represents the imaginary part of the electromagnetic response data. Figure 3 (c) represents the real part of the electromagnetic gradient response data. Figure 3 (d) represents the imaginary part of the electromagnetic gradient response data;

[0041] Figure 4 This is a comparison and verification diagram of the performance of the high-resolution robust inversion method based on electromagnetic gradient data. Figure 4 (a) This diagram illustrates the final inversion model that this method is intended to obtain. Figure 4 (b) A schematic diagram of the final model obtained by the traditional electromagnetic data inversion method. Detailed Implementation

[0042] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, and not all of them. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present application can be combined with each other.

[0043] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0044] The terms “comprising” and “having”, and any variations thereof, in the specification and claims of this application are intended to cover non-exclusive inclusion, for example, a process, method, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or apparatus.

[0045] Reference Figure 1 , Figure 1 This is a flowchart of a baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient according to an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes:

[0046] S1, Construct the initial inversion conductivity model;

[0047] The initial inversion process is initiated by establishing an initial conductivity model as the starting point for the inversion. The inversion model is based on the conductivity at each measurement point. The set of values, where each conductivity value corresponds to a grid cell in the underground space, is expressed by the following formula:

[0048] ;

[0049] in, For inversion model; This indicates the total number of inversion grids (the size of the subdivision unit is determined according to the model size, survey area range, target scale, etc., and is usually locally densified in the source, receiver and anomaly regions, and gradually thinned out outside other areas of interest, in order to balance computational efficiency and accuracy). Indicates the first Conductivity at each grid point;

[0050] At the same time, the electromagnetic response obtained from the forward modeling is defined. This response is a collection of simulated electromagnetic data from various measuring points at different frequencies, defined as follows:

[0051] ;

[0052] in, , , , These represent the simulated electromagnetic datasets for measurement points 1, 2, j, and N, respectively.

[0053] Forward difference operator It has the following forms of expression:

[0054] ;

[0055] in, It is the set of real numbers; for Total number of data items;

[0056] Based on simulated electromagnetic data and forward difference operator The simulated electromagnetic gradient data was obtained:

[0057] Among them, the electromagnetic gradient response obtained from forward modeling It is a collection of simulated electromagnetic gradient data at different frequencies from various measuring points, defined as follows:

[0058] ;

[0059] in, , , , These represent the simulated electromagnetic gradient datasets for measurement points 1, 2, j, and N, respectively.

[0060] S2, constructing the objective function of gradient data residuals based on electromagnetic field gradient data residuals;

[0061] The gradient data residual objective function measures the mismatch between the observed data (i.e., the actual data) and the forward simulation data. It is the function that needs to be minimized during the inversion process and is defined as follows:

[0062] ;

[0063] In the formula, The objective function is the residual of the electromagnetic gradient data; For the observed electromagnetic gradient data, The electromagnetic gradient response in forward modeling; This is the difference between the actual gradient data and the simulated gradient data, i.e., the gradient data residual. This is a diagonal data weighting matrix used to weight the residuals of the observation data point by point according to the uncertainty. Each diagonal element is equivalent to the inverse of the noise standard deviation of the electromagnetic gradient data at each measurement point, and includes the uncertainty of the observed electromagnetic field gradient data, which is used to adjust the reliability of the gradient data.

[0064] S3, calculates the Jacobian matrix of the inverted electromagnetic gradient data using the forward difference operator;

[0065] The Jacobian matrix (also known as the sensitivity matrix) is the core of the inversion process, representing the degree of influence of small changes in model parameters (conductivity) on the forward modeling data. In this invention, the Jacobian matrix for inverting electromagnetic gradient data is calculated using the following formula:

[0066] ;

[0067] In the formula, The Jacobian matrix representing the inverted electromagnetic gradient data is a matrix composed of the sensitivity of the observed electromagnetic gradient data to the model parameters, expressed as: ;

[0068] The electromagnetic data Jacobian matrix refers to the matrix composed of the sensitivity of observed electromagnetic data to model parameters, and its expression is: ,in Refers to simulated electromagnetic data;

[0069] This is a forward difference operator, whose main function is to act as a transformation matrix between electromagnetic data and electromagnetic gradient data. The specific transformation relationship is as follows: ;

[0070] This is an interpolation mapping matrix between actual data sampling points and grid cells, reflecting the spatial correspondence between the actual data sampling locations and the model grid. This is primarily because the actual electromagnetic data and electromagnetic gradient data acquisition locations are determined based on the actual conditions of the survey area and the detection requirements, while the grid subdivision in the inversion is determined based on the survey area, the target body, and the overall model size. Therefore, their positions may not correspond perfectly. If the acquisition location and the grid location do not correspond perfectly, a matrix is ​​needed to address this. Establish a mapping relationship between the actual acquisition location and the inversion grid location, thereby establishing a spatial mapping relationship between the acquired data and the model forward response;

[0071] The coefficient matrix is ​​a three-dimensional forward modeling matrix, typically referring to the linear operators obtained by discretizing Maxwell's equations onto a mesh / element that satisfy the following linear equation system. ,in, The source term is the vector projected onto the external electrical or magnetic source according to the discrete basis functions. and All parameters, including model settings, mesh parameters, and physical parameters, are determined before the forward simulation solution. For the inversion method, however, and All of these are known matrices that can be directly used in the inversion stage of the forward modeling process; The coefficient matrix expression pair The partial derivatives, due to The expression is about Therefore, this matrix can be determined during the formation of the forward coefficient matrix. For the inversion method, it is also a known matrix that can be directly used when the forward process parameters are input to the inversion stage. For the above system of linear equations The unknown electromagnetic field to be solved, i.e., the three-dimensional forward response vector, is relevant to the inversion method. It also belongs to the known vectors that can be directly used when the solution of the forward modeling process is input into the inversion stage.

[0072] S4, determine the parameters required for the inversion;

[0073] Set the control parameters needed during the inversion iteration process to ensure the stability and convergence of the inversion. These parameters include:

[0074] Regularization coefficient : Used to balance the contribution of data residuals and model errors;

[0075] Reference model term weight control coefficient Used to control the contribution of the reference model;

[0076] Reference model weight matrix : Control the model to not deviate from the reference model;

[0077] Gradient roughness weight control coefficient : Controls the intensity of model smoothing;

[0078] Roughness term weight matrix : Control the model to be reasonably smooth in space;

[0079] Reference Model As a reference for model constraints, the reference model is constructed based on known geological data and combined with available geophysical prior information, inversion results and other prior constraints. When there is no known information, it can usually be set as a uniform background model.

[0080] Initial model: The initial model refers to the starting model of the inversion iteration. It is set manually and can usually be set as a uniform earth model.

[0081] Reference Model and The combined effect facilitates the incorporation of prior geological information, integrates a reliable background into the reference model, and allows for the transfer of information between the reference model and the reference model. Weighting is used to guide the model inversion towards the reference model.

[0082] S5, Solve the iterative update model;

[0083] Based on the objective function of the electromagnetic gradient data residual, the inversion conductivity model is iteratively updated to minimize the objective function, thus obtaining the inversion result, i.e., the accurate conductivity model.

[0084] No. Step value for the next iteration This is obtained by solving the following equation:

[0085] ;

[0086] In the formula, Indicates the first The model obtained from the first step of calculation and the first step The difference between the models obtained from the first step; For the first The Jacobian matrix of electromagnetic gradient data corresponding to the next iteration model; For the first The regularization coefficient chosen in the next iteration; These are the reference model item weight control coefficients; The reference model weight matrix; These are the gradient roughness weight control coefficients; This is the roughness term weight matrix; For the first Inversion model in the next iteration; This is a reference model.

[0087] By continuously iterating and updating the model ( For the first The model is inverted in each iteration until the objective function no longer decreases significantly, the model change is less than a preset threshold, or the number of iterations reaches a preset number (usually set to 100). Then the inversion result is output to obtain an accurate conductivity model.

[0088] This invention first establishes an initial inversion model to determine the inversion starting point; then it defines the objective function to clarify the inversion objective function; then it defines the gradient Jacobian matrix to prepare for subsequent calculations; finally, it provides an iterative calculation formula, and through iteratively solving the model, it achieves interference-resistant high-resolution inversion of electromagnetic gradient field data.

[0089] Reference Figure 2 , Figure 2 As a realistic 3D model, inversion is performed using the model's forward response. The main purpose is to compare the inversion results of the electromagnetic gradient data from this model with those of conventional electromagnetic field inversion, providing forward response data support. A 3D coordinate system is established with the center of the conductor source as the origin, the direction of the survey line distribution as the x-axis, the vertical distance of the survey line from the center of the conductor source as the y-axis, and the vertical upward direction of the survey area as the z-axis. The conductivity in space is... The electrical conductivity of the earth is There are two conductors with a conductivity of 1. The anomaly is shown in the figure. The distribution of the selected survey lines is shown in the figure.

[0090] Reference Figure 3 , Figure 3 This is a schematic diagram comparing electromagnetic response data and electromagnetic gradient response data. Figure 3 The black dashed line represents clean data, and the red dashed line represents data containing baseline drift. Figure 3 (a) represents the real part of the electromagnetic response data. Figure 3 (b) represents the imaginary part of the electromagnetic response data. Figure 3 (c) represents the real part of the electromagnetic gradient response data. Figure 3 (d) represents the imaginary part of the electromagnetic gradient response data. It can be seen that baseline drift has a significant impact on the electromagnetic response data; using gradient data can effectively avoid the influence of baseline drift.

[0091] Reference Figure 4 , Figure 4 Verification diagram showing the improvement in performance of the high-resolution robust inversion method based on electromagnetic gradient data. Figure 4 (a) This diagram illustrates the final inversion model that this method is intended to obtain. Figure 4 (b) A schematic diagram of the final model obtained by the traditional electromagnetic data inversion method.

[0092] The proposed method achieves a higher degree of consistency between the final model and the real model. In contrast, due to baseline drift, the model obtained through traditional electromagnetic data inversion exhibits significant errors compared to the real model, particularly in the accuracy of background resistivity values ​​and the longitudinal consistency of anomaly boundaries. This demonstrates that the proposed method effectively avoids the influence of baseline drift and improves inversion robustness. Furthermore, the proposed method yields a higher resolution model for anomaly boundaries, with a particularly significant improvement in the accuracy of deep-to-deep anomaly boundary localization, showcasing the high-resolution target localization capability of this method.

[0093] This invention also provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps in any of the above method embodiments.

[0094] This invention also provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when it is run.

[0095] Specific examples in this embodiment can be found in the examples described in the above embodiments and optional implementations, and those already described will not be repeated here.

[0096] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0097] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0098] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0099] The steps in the methods of the above embodiments of the present invention can be adjusted, combined, or deleted according to actual needs. The technical features can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the embodiments are described. However, as long as the combinations of these technical features do not contradict each other, they should all be considered within the scope of the present invention.

[0100] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient, characterized in that, include: Construct an initial inversion conductivity model; A gradient data residual objective function is constructed based on the electromagnetic field gradient data residual, wherein the gradient data residual is the residual between the actual electromagnetic gradient data and the simulated electromagnetic gradient data. The Jacobian matrix of the inverted electromagnetic gradient data is calculated by a forward difference operator, wherein the forward difference operator is a first-order difference matrix, and the Jacobian matrix is ​​the sensitivity matrix of the observed electromagnetic gradient data to the model parameters. Determine the parameters required for inversion, including regularization coefficient, reference model term weight control coefficient, reference model weight matrix, gradient roughness weight control coefficient, roughness term weight matrix, and reference model; Based on the electromagnetic gradient data residual objective function, and by iteratively updating the inversion conductivity model according to the Jacobian matrix and the parameters, the inversion result is obtained.

2. The baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient according to claim 1, characterized in that, The initial inversion conductivity model is the conductivity at each measurement point. The set of elements, where each grid cell corresponds to a conductivity value, is expressed by the following formula: ； in, For inversion model; Indicates the total number of inversion grids; Indicates the first Conductivity at each grid point.

3. The baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient according to claim 2, characterized in that, After constructing the initial inversion conductivity model, the method further includes: Define the electromagnetic response obtained from forward modeling. , It is a collection of simulated electromagnetic data from various measuring points at different frequencies, defined as follows: ； in, , , , These represent the simulated electromagnetic datasets for measurement points 1, 2, j, and N, respectively. Simulated electromagnetic gradient data are obtained based on simulated electromagnetic data and the forward difference operator. : ； in, Forward difference operator, ； It is the set of real numbers; for Total number of data points; electromagnetic gradient response obtained from forward modeling. It is a collection of simulated electromagnetic gradient data at different frequencies from various measuring points, defined as follows: ； in, , , , These represent the simulated electromagnetic gradient datasets for measurement points 1, 2, j, and N, respectively.

4. The baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient according to claim 3, characterized in that, The objective function for the gradient data residual is defined as follows: ； In the formula, The objective function is the residual of the electromagnetic gradient data; For the observed electromagnetic gradient data, The electromagnetic gradient response in forward modeling; This is the difference between the actual electromagnetic gradient data and the simulated electromagnetic gradient data, i.e., the gradient data residual. It is a diagonal weighted matrix that includes the uncertainty of the measured electromagnetic data, and is determined based on the reciprocal of the noise standard deviation of the electromagnetic gradient data at each measurement point.

5. The baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient according to claim 4, characterized in that, The Jacobian matrix of the inverted electromagnetic gradient data, calculated using the forward difference operator, includes: The formula for calculating the Jacobian matrix of the inverted electromagnetic gradient data is: ； In the formula, This represents the Jacobian matrix of the inverted electromagnetic gradient data; The Jacobian matrix for electromagnetic data; Forward difference operator, This is the interpolation mapping matrix between actual sampling points and grid cells, reflecting the spatial correspondence between the sampling positions of actual data and the model grid; It is a three-dimensional forward coefficient matrix. This is the three-dimensional forward response vector.

6. The baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient according to claim 5, characterized in that, The inverse conductivity model is iteratively updated based on the Jacobian matrix and the parameters, including: No. Step value for the next iteration This is obtained by solving the following equation: ； In the formula, Indicates the first The model obtained from the first step of calculation and the first step The difference between the models obtained from the first step; For the first The Jacobian matrix of electromagnetic gradient data corresponding to the next iteration model; For the first The regularization coefficient chosen in the next iteration; For the first Simulated electromagnetic gradient data corresponding to the next iteration model; These are the reference model item weight control coefficients; The reference model weight matrix; These are the gradient roughness weight control coefficients; This is the roughness term weight matrix; For the first Inversion model in the next iteration; This is a reference model.

7. A computer device, characterized in that, The computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient as described in any one of claims 1 to 6.

8. A storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, implements the steps of the baseline drift adaptive robust high-resolution inversion method based on electromagnetic gradient as described in any one of claims 1 to 6.

Images