A three-dimensional frequency domain interpretation method for ground-to-air transient electromagnetic data based on one-dimensional inversion

Through the three-dimensional frequency domain interpretation method combined with one-dimensional inversion and interleaved grid finite difference method, the problem of insufficient data acquisition of ground-space transient electromagnetic instruments is solved, and stable frequency domain conversion and three-dimensional inversion are realized, reducing the computational complexity , providing feasibility for practical applications.

CN115047530BInactive Publication Date: 2024-11-22JILIN JIANZHU UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210680143.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-06-15
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The time sampling range of ground-space transient electromagnetic instruments is limited, and the data acquisition density and acquisition time length are difficult to meet the requirements of time-frequency conversion algorithms, resulting in 'mixing effect' and 'truncation effect', which reduces the stability and calculation accuracy of the frequency response. And non-uniform time series sampling and environmental noise limit the confidence of the frequency domain electromagnetic response in the time domain signal.

Method used

The three-dimensional frequency domain interpretation method of ground-space transient electromagnetic data based on one-dimensional inversion is adopted, and one-dimensional time domain inversion is performed through finite memory quasi-Newtonian numerical optimization technology, and the frequency domain electromagnetic response is calculated using direct numerical integration and cosine transformation, and combined with The interleaved grid finite difference method performs three-dimensional frequency domain forwarding, and uses the nonlinear conjunction gradient method to perform iterative solutions to avoid explicitly calculating the sensitivity matrix.

Benefits of technology

The stable conversion from the time domain to the frequency domain is realized, which reduces the calculation scale and time, improves the stability and accuracy of three-dimensional frequency domain inversion, reduces technical difficulty, and provides a feasible solution for the three-dimensional interpretation of actual field data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115047530B_ABST
    Figure CN115047530B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field of transient electromagnetic data interpretation. It is a three-dimensional frequency domain interpretation method for ground-to-air transient electromagnetic data based on one-dimensional inversion, and includes the following steps: first, one-dimensional time domain inversion is performed on the ground-to-air transient electromagnetic data, wherein the time domain electromagnetic response of the long conductor source in the layered medium is calculated by direct numerical integration and cosine transformation; second, the resistivity model and its frequency domain electromagnetic response obtained by the one-dimensional inversion are converted into the initial model and observation data of the three-dimensional inversion; finally, the nonlinear conjugate gradient method is used to perform three-dimensional frequency domain inversion to obtain the three-dimensional resistivity distribution of the underground. The present invention adopts one-dimensional time domain inversion to realize the stable conversion of the observation data from the time domain to the frequency domain, and the conversion accuracy is not affected by the time sampling range, density and non-uniformity; the three-dimensional frequency domain forward and inversion technology is mature, which can effectively reduce the technical difficulty of direct time domain three-dimensional inversion of ground-to-air transient electromagnetic data, and reduce the calculation scale and time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of transient electromagnetic data interpretation, in particular to a three-dimensional frequency domain interpretation method of ground-to-air transient electromagnetic data based on one-dimensional inversion. Background Art

[0002] The ground-to-air transient electromagnetic method is a semi-aeronautical electromagnetic method that transmits on the ground and receives in the air. It uses a finite-length ground wire as the transmitting source, and the aircraft receives the time-domain induced electromagnetic signal in the air, which effectively solves the problem of transient electromagnetic exploration construction in areas with complex surface conditions. The ground-to-air transient electromagnetic method has the advantages of high sensitivity to underground low-resistance anomalies, short construction period, high detection efficiency and strong anti-noise ability. It has been successfully applied in geothermal surveys, volcanic structure surveys, underground tunnel surveys, groundwater salinization and groundwater monitoring, tunnel surveys, goaf detection, and ancient river channel structure detection.

[0003] In recent years, the global resources and environment have faced unprecedented pressure, and the ground-to-air transient electromagnetic method has encountered unprecedented opportunities with its unique advantages in identifying physical properties. At the same time, with the continuous deepening of natural resource exploration and development, the geological targets facing geophysical exploration are becoming more and more complex, and conventional interpretation needs to consider the multi-dimensional geophysical properties of the target. Therefore, the three-dimensional inversion of ground-to-air transient electromagnetic data has always been a hot issue in the field of geophysical exploration.

[0004] At present, two methods are usually used for 3D inversion of time-domain electromagnetic method: (1) direct time-domain solution, which makes the solution process more pure, but requires iteration from time t0. The explicit iteration has strict requirements on both the time step and the space step, while the implicit iteration requires solving a large set of equations; (2) in each inversion, the electromagnetic response of dozens of frequencies is first solved, and then transferred to the time domain through frequency-time transformation. The solution accuracy and efficiency are affected by the filter coefficient. With the development of new instruments and equipment, the time range of ground-to-air transient electromagnetic data acquisition has become wider, which greatly increases the calculation scale of 3D time-domain inversion.

[0005] In order to solve the above problems, some technologies to reduce the amount of 3D inversion calculations have been proposed, such as using non-uniform time grids, approximate solutions, modified 3D inversion programs, large-scale parallel computing platforms, and pseudo-seismic imaging in 3D transient electromagnetic simulations. However, it is still a difficult task to organically combine the 3D time domain forward modeling and inversion calculation processes, which requires sophisticated design. There is still a long way to go before strict time domain 3D inversion technology can be applied to the processing of actual field data. Therefore, the current engineering applications of ground-to-air transient electromagnetic methods are still mainly for apparent resistivity calculations and one-dimensional inversion.

[0006] In recent years, the three-dimensional forward and inversion technology of frequency-domain electromagnetic data has made great progress and has moved from theory to application. Therefore, if the observed transient electromagnetic data can be converted into frequency-domain electromagnetic response, the mature three-dimensional forward and inversion technology of frequency-domain electromagnetic method can be applied to the interpretation of transient electromagnetic data. Mitsuhata et al. (2001) directly discretized the Fourier transform by piecewise integration to obtain a matrix form, and completed the conversion of time domain signal to frequency response by adding smoothness constraints to the frequency response and using the least squares inversion method. Sasaki et al. (2015) proposed the idea of ​​inverting transient electromagnetic data in the frequency domain based on the time-frequency conversion method of Mitsuhata et al. (2001), and pointed out that frequency domain inversion can effectively replace time domain inversion. Inspired by this, Weng et al. (2017) used a regularized inversion algorithm to realize the time-frequency electromagnetic signal conversion. Based on this, Li et al. (2021) completed the three-dimensional frequency domain inversion of rectangular large loop source transient electromagnetic data.

[0007] The time-frequency transform algorithm has opened up a new way for the three-dimensional inversion of transient electromagnetic data. It allows geophysicists to use mature frequency domain data processing technology to interpret time domain data, reducing the technical difficulty of time domain data inversion, reducing the scale and time of calculation, and providing a feasible solution for the three-dimensional interpretation of actual field transient electromagnetic data on personal PCs. However, due to the limited time sampling range of ground-to-air transient electromagnetic instruments, the data acquisition density and acquisition time length are difficult to meet the requirements, and "mixing effect" and "truncation effect" are prone to occur (Runshu et al., 2002; 2005), which greatly reduces the stability and calculation accuracy of the frequency response obtained by the transformation. In addition, the non-uniform time series sampling and environmental noise of ground-to-air transient electromagnetic data limit the credibility of recovering frequency domain electromagnetic responses from time domain signals. Summary of the invention

[0008] 1. Technical issues to be solved

[0009] In view of the deficiencies in the prior art, the present invention provides a three-dimensional frequency domain interpretation method for ground-to-air transient electromagnetic data based on one-dimensional inversion, which solves the problems that the time sampling range of ground-to-air transient electromagnetic instruments is limited, the data acquisition density and acquisition time length are difficult to meet the requirements of the time-to-frequency conversion algorithm, and "mixing effect" and "truncation effect" are prone to occur, thereby greatly reducing the stability and calculation accuracy of the frequency response obtained by the transformation. In addition, the non-uniform time series sampling and environmental noise of ground-to-air transient electromagnetic data limit the credibility of recovering the frequency domain electromagnetic response from the time domain signal.

[0010] (II) Technical solution

[0011] In order to achieve the above-mentioned purpose, the present invention specifically adopts the following technical solutions:

[0012] The three-dimensional frequency domain interpretation method of ground-to-air transient electromagnetic data based on one-dimensional inversion includes the following steps:

[0013] S1. First, the limited memory quasi-Newton (BFGS) numerical optimization technique is used to perform one-dimensional time domain inversion of ground-to-air transient electromagnetic data. During the inversion process, the direct numerical integration technique is used to calculate the full-space frequency domain electromagnetic response of the one-dimensional geoelectric model of the finite length wire source, and the frequency domain electromagnetic response is converted to the time domain using cosine transform. The gradient of the objective function is calculated using differential approximation. The adaptive regularization factor is used to fully fit the observed data.

[0014] S2. Secondly, the underground resistivity distribution model and its frequency domain electromagnetic response obtained by one-dimensional time domain inversion are used as the initial model and observation data for three-dimensional frequency domain inversion; the staggered grid finite difference method is used to discretize the Helmholtz equation satisfied by the quadratic field for three-dimensional frequency domain forward simulation; the inversion adopts the nonlinear conjugate gradient method (NLCG), which does not require explicit calculation and stores the sensitivity matrix; the underground three-dimensional resistivity distribution model is obtained, the time domain electromagnetic response of the three-dimensional model is compared with the measured data, and the data fitting is analyzed;

[0015] S3. Finally, interpret the results based on known geological data.

[0016] Furthermore, the one-dimensional time domain inversion process is as follows:

[0017] (1) Simplify the earth into a one-dimensional layered isotropic medium, set the thickness and resistivity of each layer of the layered model, and preset the initial model;

[0018] (2) A finite-length conductor with multiple electric dipoles is used as the source of ground-to-space transient electromagnetic. A right-hand coordinate system is established through theoretical analysis. The midpoint of the source is the origin of the coordinate system, the x-axis is along the direction of the conductor, and the positive direction of the z-axis is perpendicular to the ground and downward. The direct numerical integration technique is used to calculate the full-space frequency domain electromagnetic response of the one-dimensional layered isotropic medium model.

[0019] (3) The frequency domain and time domain electromagnetic responses can be linked by Fourier transform. When the transmitting waveform is a down-step wave, the cosine relationship of the frequency-time electromagnetic field can be derived by using the Euler formula, and the frequency domain electromagnetic response can be converted into the time domain by using the broken line approximation method.

[0020] (4) The observation data of the ground-to-air transient electromagnetic method is the rate of change of the magnetic induction intensity of each time channel over time (induced electromotive force). The objective function of the one-dimensional time domain inversion is defined as:

[0021] f(d,m)=(d-d0) T (d-d0)+α(mm r ) T(mm r ), where: d is the time domain observation data vector, d0 is the forward prediction data vector, m is the model vector, m r is the initial model vector, α is the regularization factor, and in order to fit the observed data as much as possible, the adaptive regularization factor technology is used;

[0022] (5) Use differential approximation to calculate the gradient of the objective function;

[0023] (6) Apply the limited memory quasi-Newton numerical optimization technology to obtain the resistivity distribution model of the underground layered medium. If the preset termination condition of the inversion iteration is not met, repeat (2)-(6).

[0024] Furthermore, the three-dimensional frequency domain inversion process is as follows:

[0025] (1) The staggered grid finite difference method is used to discretize the Helmholtz equation satisfied by the secondary field to perform full-space three-dimensional frequency domain electromagnetic forward simulation of a finite length wire source;

[0026] (2) Using the cubic spline interpolation data function, the prediction data of the three-dimensional frequency domain inversion is obtained;

[0027] (3) Establish the objective function of three-dimensional frequency domain inversion:

[0028] φ(m)=φ d (m)+λφ m (m), where: vector m represents the model parameters, φ m (m) represents the model constraint function, φ d (m) represents the data error function in the above norm form, λ is the regularization factor, balancing the data prediction error and the model constraint, and the nonlinear conjugate gradient method is used to solve the above objective function;

[0029] (4) The frequency domain electromagnetic response function is Taylor expanded near the reference model, and the gradient of the objective function is approximated by omitting the high-order terms:

[0030] Where: vector d obs represents the observed data, W d is the data variance matrix, C m is the model variance matrix, f(m) is the forward operator, and J is the sensitivity matrix;

[0031] The sensitivity matrix and its elements are not explicitly calculated and stored. Only the sensitivity matrix J or its transpose J needs to be calculated. T The product with a vector is sufficient;

[0032] (5) Obtain a three-dimensional underground resistivity distribution model. If the preset termination condition of the inversion iteration is not met, repeat (1)-(5).

[0033] (III) Beneficial effects

[0034] Compared with the prior art, the present invention provides a three-dimensional frequency domain interpretation method for ground-to-air transient electromagnetic data based on one-dimensional inversion, which has the following beneficial effects:

[0035] 1. The present invention obtains the observation data required for three-dimensional frequency domain inversion through one-dimensional inversion of measured data, thereby realizing the stable conversion of observation data from time domain to frequency domain, and the conversion credibility is not affected by the time sampling range, density and non-uniformity of the instrument.

[0036] 2. In the present invention, the resistivity distribution model obtained by one-dimensional inversion can be used as the initial model for three-dimensional frequency domain inversion, thereby solving the problem of three-dimensional inversion failure caused by unreasonable selection of the initial model.

[0037] 3. In the present invention, the one-dimensional inversion is equivalent to the pre-processing of filtering and denoising of the measured data, so that the three-dimensional frequency domain inversion can be iteratively run stably and the fitting error can be rapidly reduced.

[0038] 4. The present invention has mature three-dimensional frequency domain forward and inversion technology, which can effectively reduce the technical difficulty of time domain data processing, reduce the calculation scale and time, and provide a feasible solution for three-dimensional interpretation of actual field ground-to-air transient electromagnetic data on a personal PC. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the process of processing ground-to-air transient electromagnetic data in the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] Example

[0042] like Figure 1 As shown, an embodiment of the present invention proposes a three-dimensional frequency domain interpretation method for ground-to-air transient electromagnetic data based on one-dimensional inversion, comprising the following steps:

[0043] S1. First, the limited memory quasi-Newton (BFGS) numerical optimization technique is used to perform one-dimensional time domain inversion of ground-to-air transient electromagnetic data;

[0044] In the inversion process, direct numerical integration technology is used to calculate the full-space frequency domain electromagnetic response of the one-dimensional geoelectric model of a finite-length wire source, and the frequency domain electromagnetic response is converted to the time domain using cosine transform;

[0045] The gradient of the objective function is calculated using differential approximation;

[0046] Use adaptive regularization factors to fully fit the observed data;

[0047] Among them, the underground layered resistivity distribution model obtained by outputting ground-to-air transient electromagnetic data through one-dimensional time domain inversion is interpolated into a three-dimensional model, which is used as the initial model for three-dimensional frequency domain inversion. At the same time, the frequency domain electromagnetic response of the underground layered resistivity distribution model is obtained and used as the observation data for three-dimensional frequency domain inversion.

[0048] S2. Secondly, the underground resistivity distribution model and its frequency domain electromagnetic response obtained by one-dimensional time domain inversion are used as the initial model and observation data for three-dimensional frequency domain inversion;

[0049] The staggered grid finite difference method is used to discretize the Helmholtz equation satisfied by the quadratic field for three-dimensional frequency domain forward simulation;

[0050] The inversion uses the nonlinear conjugate gradient method (NLCG), which does not require explicit calculation and storage of the sensitivity matrix;

[0051] Obtain the underground three-dimensional resistivity distribution model, compare the time domain electromagnetic response of the three-dimensional model with the measured data, and analyze the data fitting;

[0052] S3. Finally, interpret the results based on known geological data.

[0053] Among them, the one-dimensional time domain inversion process is as follows:

[0054] (1) Simplify the earth into a one-dimensional layered isotropic medium, set the thickness and resistivity of each layer of the layered model, and preset the initial model;

[0055] (2) A finite-length conductor with multiple electric dipoles is used as the source of ground-to-space transient electromagnetic. A right-hand coordinate system is established through theoretical analysis. The midpoint of the source is the origin of the coordinate system, the x-axis is along the direction of the conductor, and the positive direction of the z-axis is perpendicular to the ground and downward. The direct numerical integration technique is used to calculate the full-space frequency domain electromagnetic response of the one-dimensional layered isotropic medium model.

[0056] (3) The frequency domain and time domain electromagnetic responses can be linked by Fourier transform. When the transmitting waveform is a down-step wave, the cosine relationship of the frequency-time electromagnetic field can be derived by using the Euler formula, and the frequency domain electromagnetic response can be converted into the time domain by using the broken line approximation method.

[0057] (4) The observation data of the ground-to-air transient electromagnetic method is the rate of change of the magnetic induction intensity of each time channel over time (induced electromotive force). The objective function of the one-dimensional time domain inversion is defined as:

[0058] f(d,m)=(d-d0) T (d-d0)+α(mm r ) T (mm r ), where: d is the time domain observation data vector, d0 is the forward prediction data vector, m is the model vector, m r is the initial model vector, α is the regularization factor, and in order to fit the observed data as much as possible, the adaptive regularization factor technology is used;

[0059] (5) Use differential approximation to calculate the gradient of the objective function;

[0060] (6) Apply the limited memory quasi-Newton numerical optimization technology to obtain the resistivity distribution model of the underground layered medium. If the preset termination condition of the inversion iteration is not met, repeat (2)-(6).

[0061] Among them, the three-dimensional frequency domain inversion process is as follows:

[0062] (1) The staggered grid finite difference method is used to discretize the Helmholtz equation satisfied by the secondary field to perform full-space three-dimensional frequency domain electromagnetic forward simulation of a finite length wire source;

[0063] (2) Using the cubic spline interpolation data function, the prediction data of the three-dimensional frequency domain inversion is obtained;

[0064] (3) Establish the objective function of three-dimensional frequency domain inversion:

[0065] φ(m)=φ d (m)+λφ m (m), where: vector m represents the model parameters, φ m (m) represents the model constraint function, φ d (m) represents the data error function in the above norm form, λ is the regularization factor, balancing the data prediction error and the model constraint, and the nonlinear conjugate gradient method is used to solve the above objective function;

[0066] (4) The frequency domain electromagnetic response function is Taylor expanded near the reference model, and the gradient of the objective function is approximated by omitting the high-order terms:

[0067] Where: vector d obs represents the observed data, W d is the data variance matrix, C mis the model variance matrix, f(m) is the forward operator, and J is the sensitivity matrix;

[0068] The sensitivity matrix and its elements are not explicitly calculated and stored. Only the sensitivity matrix J or its transpose J needs to be calculated. T The product with a vector is sufficient;

[0069] In this way, the overall computation time and memory requirement are much smaller than the idea of ​​directly calculating and storing the sensitivity matrix J;

[0070] This lays the foundation for three-dimensional inversion calculation on personal PC, which is also one of the reasons why this paper adopts nonlinear conjugate gradient numerical optimization technology for three-dimensional inversion.

[0071] In the calculation process: the time used for one-dimensional time domain inversion is ignored, the three-dimensional nonlinear conjugate gradient inversion is preset to be iterated 100 times, each iteration requires three forward calculations, the three-dimensional forward calculation time for each frequency is set to t, the frequencies required for frequency-time transformation are 67, and the number of frequencies commonly used in frequency domain electromagnetic inversion is 17 (2 n , refer to controlled source audio frequency magnetotelluric method);

[0072] The calculation time of three-dimensional time domain inversion based on frequency-time transformation is approximately: 100*3*67*t;

[0073] The computation time of the three-dimensional frequency domain inversion proposed in this paper is approximately: 100*3*17*t;

[0074] From the above content, it can be roughly calculated that: in terms of computing time and memory overhead, the three-dimensional frequency domain inversion method for ground-to-air transient electromagnetic data proposed in this paper reduces about 75% compared with the time domain three-dimensional inversion based on frequency-time transformation;

[0075] (5) Obtain a three-dimensional underground resistivity distribution model. If the preset termination condition of the inversion iteration is not met, repeat (1)-(5).

[0076] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A three-dimensional frequency domain interpretation method for ground-to-air transient electromagnetic data based on one-dimensional inversion. It is characterized in that The steps include: S1. First, a limited memory quasi-Newton numerical optimization technique is used to perform one-dimensional time domain inversion of ground-to-air transient electromagnetic data; In the inversion process, direct numerical integration technology is used to calculate the full-space frequency domain electromagnetic response of the one-dimensional geoelectric model of a finite-length wire source, and the frequency domain electromagnetic response is converted to the time domain using cosine transform; The gradient of the objective function is calculated using differential approximation; Use adaptive regularization factors to ensure adequate fit to observed data; S2. Secondly, the underground resistivity distribution model and its frequency domain electromagnetic response obtained by one-dimensional time domain inversion are used as the initial model and observation data for three-dimensional frequency domain inversion; The staggered grid finite difference method is used to discretize the Helmholtz equation satisfied by the quadratic field for three-dimensional frequency domain forward simulation; The inversion uses the nonlinear conjugate gradient method, which does not require explicit calculations and does not require the storage of sensitivity matrices; Obtain the underground three-dimensional resistivity distribution model, compare the time domain electromagnetic response of the three-dimensional model with the measured data, and analyze the data fitting; S3. Finally, interpret the results based on known geological data; The one-dimensional time domain inversion process is as follows: (1) Simplify the earth into a one-dimensional layered isotropic medium, set the thickness and resistivity of each layer of the layered model, and preset the initial model; (2) A finite-length conductor with multiple electric dipoles is used as the source of ground-to-space transient electromagnetic. A right-hand coordinate system is established through theoretical analysis. The midpoint of the source is the origin of the coordinate system, the x-axis is along the direction of the conductor, and the positive direction of the z-axis is perpendicular to the ground and downward. The direct numerical integration technique is used to calculate the full-space frequency domain electromagnetic response of the one-dimensional layered isotropic medium model. (3) The frequency domain and time domain electromagnetic responses can be linked by Fourier transform. When the transmitting waveform is a down-step wave, the cosine relationship of the frequency-time electromagnetic field can be derived by using the Euler formula, and the frequency domain electromagnetic response can be converted into the time domain by using the broken line approximation method. (4) The observation data of the ground-to-air transient electromagnetic method is the rate of change of the magnetic induction intensity of each time channel over time. The objective function of the one-dimensional time domain inversion is defined as: f(d,m)=(dd 0 ) T (dd 0 )+α(mm r ) T (mm r ), where: d is the time domain observation data vector, d 0 is the predicted data vector obtained by forward modeling, m is the model vector, and m r is the initial model vector, α is the regularization factor; (5) Use differential approximation to calculate the gradient of the objective function; (6) Apply the limited memory quasi-Newton numerical optimization technology to obtain the resistivity distribution model of the underground layered medium. If the preset termination condition of the inversion iteration is not met, repeat (2)-(6).

2. The three-dimensional frequency domain interpretation method of ground-to-air transient electromagnetic data based on one-dimensional inversion according to claim 1, Features: The three-dimensional frequency domain inversion process is as follows: (1) The staggered grid finite difference method is used to discretize the Helmholtz equation satisfied by the secondary field to perform full-space three-dimensional frequency domain electromagnetic forward simulation of a finite length wire source; (2) Using the cubic spline interpolation data function, the prediction data of the three-dimensional frequency domain inversion is obtained; (3) Establish the objective function of three-dimensional frequency domain inversion: φ(m)=φ d (m)+λφ m (m), where: vector m represents the model parameters, φ m (m) represents the model constraint function, φ d (m) represents the data error function in norm form, λ is the regularization factor, balancing the data prediction error and model constraints, and the nonlinear conjugate gradient method is used to solve the above objective function; (4) The frequency domain electromagnetic response function is Taylor expanded near the reference model, and the gradient of the objective function is approximated by omitting the high-order terms: Where: vector d obs represents the observed data, W d is the data variance matrix, C m is the model variance matrix, f(m) is the forward operator, and J is the sensitivity matrix; The sensitivity matrix and its elements are not explicitly calculated or stored. Only the sensitivity matrix J or its transpose J needs to be calculated. T The product with a vector is sufficient; (5) Obtain a three-dimensional underground resistivity distribution model. If the preset termination condition of the inversion iteration is not met, repeat (1)-(5) of the three-dimensional frequency domain inversion process.