A method, system, and storage medium for denoising time-frequency electromagnetic data

By generating a coherence matrix and utilizing the electromagnetic reciprocity principle and the least squares robust estimation method, the problem of noise interference in the time-frequency electromagnetic method is solved, achieving efficient noise reduction and improving data quality and production efficiency.

CN118091770BActive Publication Date: 2026-11-10PETROCHINA CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202211489151.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-11-25
Publication Date
2026-11-10
Estimated Expiration
2042-11-25

AI Technical Summary

Technical Problem

Existing time-frequency electromagnetic methods face severe noise interference in oil and gas exploration. Traditional filtering and noise reduction methods lead to phase accuracy and energy leakage problems, making it difficult to meet the needs of large-scale applications.

Method used

By generating a coherence matrix, performing approximate diagonalization and SVD decomposition, the reverse transmission signal is solved using the electromagnetic reciprocity principle, and the ground frequency response function is calculated using the least squares robust estimation method to remove noise.

Benefits of technology

It improves the signal-to-noise ratio of time-frequency electromagnetic data, ensures phase accuracy and amplitude authenticity, improves data quality and production efficiency, and reduces the need for manual intervention.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118091770B_ABST
    Figure CN118091770B_ABST
Patent Text Reader

Abstract

This invention provides a method, system, and storage medium for denoising time-frequency electromagnetic data, comprising: generating a coherence matrix using a time-varying spectrum table of a received time-frequency electromagnetic signal and a time-varying spectrum table of a transmitted time-frequency electromagnetic signal; performing approximate diagonalization and SVD decomposition on the coherence matrix to calculate several coherent eigenvectors and coherent eigenvalues ​​between the transmitted and received signals; using the received signal as the transmitted signal, performing forward modeling to solve the response of the reverse transmitted signal based on the current coherent eigenvalues ​​to obtain the forward modeled transmitted signal; calculating the weights of each coherent eigenvalue based on the difference between the transmitted signal and the forward modeled transmitted signal; generating a new coherence matrix based on the weights of each coherent eigenvalue and solving the linear equation using SVD decomposition to obtain the geodetic frequency response function at each frequency point in each observation point; and solving for the optimal geodetic frequency response function at each frequency point in each observation point based on the geodetic frequency response function at each frequency point in each observation point.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to a method, system, and storage medium for denoising time-frequency electromagnetic data. Background Technology

[0002] The time-frequency electromagnetic method uses a grounded long conductor to sequentially transmit square wave signals of different periods into the ground. The number of repetitions of each period of the square wave signal can be flexibly adjusted according to factors such as the intensity of environmental noise and frequency band. At the same time, the transmitter records the waveform of the transmitted signal for use in subsequent processing. The receiver uses an equatorial device and a grounded electrode parallel to the transmission direction. The receiver records the horizontal electric field component, denoted as Ex. At the same time, a magnetic rod perpendicular to the ground is used to record the vertical magnetic field component, denoted as Bz.

[0003] With the increasing number of sources of electromagnetic interference (such as high-voltage transmission lines, high-speed railways, and wireless communication base stations), especially 50Hz industrial electrical interference, the quality rate of time-frequency electromagnetic field data acquisition has decreased. In severe cases, the data rework rate remains high, making it difficult to improve data quality and production efficiency, thus restricting the economic viability of this method.

[0004] Existing technologies often improve the signal-to-noise ratio and suppress noise by increasing the current intensity and the number of repetitions of the excitation square wave signal. Although acceptable data can be obtained by increasing the current intensity and the number of repetitions of the excitation square wave signal, the excitation current intensity and the number of excitation signal repetitions cannot be increased indefinitely in field production. Therefore, the above methods have considerable limitations.

[0005] Therefore, existing technologies mostly employ filtering methods for noise reduction. Traditional filtering-based noise reduction methods utilize the design principles of finite impulse response (FIR) filters. However, truncating the original signal causes distortion of the response phase curve, compromising phase accuracy. Furthermore, the truncation frequency requires manual selection, and the frequency response curve at the truncation point cannot be perfectly polygonal, leading to energy leakage to other frequencies and amplitude distortion. Consequently, with the increasing application of time-frequency electromagnetic exploration methods in oil and gas exploration, traditional filtering-based noise reduction methods are no longer suitable for large-scale application. Summary of the Invention

[0006] To address the technical problem of difficulty in denoising time-frequency electromagnetic data obtained using the time-frequency electromagnetic method in existing technologies, embodiments of the present invention provide a method, system, and storage medium for denoising time-frequency electromagnetic data.

[0007] The embodiments of the present invention are achieved through the following technical solutions:

[0008] In a first aspect, embodiments of the present invention provide a method for denoising time-frequency electromagnetic data, comprising:

[0009] S1. Preprocess the raw time-frequency electromagnetic data to obtain the time-varying spectrum table of the time-frequency electromagnetic received signal and the time-varying spectrum table of the time-frequency electromagnetic transmitted signal.

[0010] S2. Using the time-varying spectrum table of the time-frequency electromagnetic received signal and the time-varying spectrum table of the time-frequency electromagnetic transmitted signal, a coherence matrix is ​​generated by utilizing the coherence between the transmitted and received signals;

[0011] S3. Perform approximate diagonalization and SVD decomposition on the coherence matrix to calculate several coherent eigenvectors and coherent eigenvalues ​​between the transmitted and received signals;

[0012] S4. Based on the principle of electromagnetic reciprocity, the received signal is used as the transmitted signal, and the response of the reverse transmitted signal is solved by forward modeling based on the current coherent eigenvalues ​​to obtain the forward modeled transmitted signal; the initial value of the current coherent eigenvalues ​​is the coherent eigenvalue of S3.

[0013] S5. Calculate the weights of each coherent eigenvalue based on the difference between the transmitted signal and the forward-modeled transmitted signal;

[0014] S6. Generate a new coherence matrix based on the weights of each coherent eigenvalue and solve the linear equation using SVD decomposition to obtain the earth frequency response function at each frequency point in each observation point;

[0015] S7. Using the least squares robust estimation method, solve for the optimal geodetic frequency response function at each observation point based on the geodetic frequency response function at each frequency point.

[0016] Furthermore, S4. Based on the principle of electromagnetic reciprocity, the received signal is used as the transmitted signal, and the response of the reverse transmitted signal is solved by forward modeling using coherent eigenvalues ​​to obtain the forward modeled transmitted signal; including:

[0017] The coherence matrix of the response to the reverse transmission signal is formed according to formula (2), and SVD decomposition is performed to obtain the forward transmission signal X′ of the transmission signal.

[0018] X′=EY (2)

[0019] Where X′ is the forward-modeled transmitted signal; E represents the received signal, which includes environmental noise; and Y represents the ground frequency response function.

[0020] Furthermore, S5. Calculate the weights of each coherent eigenvalue based on the difference between the transmitted signal and the forward-modeled transmitted signal; including:

[0021] The weights of each coherent eigenvalue are calculated according to formula (3):

[0022] w i =1 / |x i -xi ′| (3)

[0023] Where W represents the weight, i represents the order number of the coherent feature value, and i is a positive integer greater than 0; X is the transmitted signal, which is directly recorded from the transmitter and does not include environmental noise; X′ is the forward-modeling transmitted signal.

[0024] Furthermore, S6. A new coherence matrix is ​​generated based on the weights of each coherent eigenvalue, and the linear equation is solved using SVD decomposition to obtain the ground frequency response function at each frequency point of each observation point; including:

[0025] A new coherence matrix is ​​generated based on the weights of each coherent eigenvalue and formula (4):

[0026] Y = (X T X) -1 X T WE (4)

[0027] Where Y represents the ground frequency response function, X represents the transmitted signal, E represents the received signal, W represents the weights, and T represents the matrix transpose.

[0028] Furthermore, S7. The optimal frequency response function at each observation point is obtained using the least squares robust estimation method; this also includes:

[0029] Calculate the mean square error of the optimal geodetic frequency response function at each observation point;

[0030] Determine whether the mean square error is less than the specified error. If not, return the best frequency response function of each frequency point in each current observation point as the current feature value to S4; if yes, obtain the best frequency response function of each frequency point in each observation point.

[0031] Furthermore, the method for denoising the time-frequency electromagnetic data further includes:

[0032] Output the optimal geodetic frequency response function for each observation point.

[0033] In a second aspect, embodiments of the present invention provide a denoising system for time-frequency electromagnetic data, comprising:

[0034] The preprocessing unit is used to preprocess the raw time-frequency electromagnetic data to obtain the time-series time-varying spectrum table of the time-frequency electromagnetic received signal and the time-series time-varying spectrum table of the time-frequency electromagnetic transmitted signal.

[0035] The coherence matrix forming unit is used to generate a coherence matrix by utilizing the time-varying spectrum table of the time-frequency electromagnetic received signal and the time-varying spectrum table of the time-frequency electromagnetic transmitted signal, taking advantage of the coherence between the transmitted and received signals.

[0036] The processing unit is used to perform approximate diagonalization and SVD decomposition on the coherence matrix, and calculate several coherent feature vectors and coherent feature values ​​between the transmitted signal and the received signal.

[0037] The forward modeling unit is used to solve the response of the reverse transmitted signal by taking the received signal as the transmitted signal and performing forward modeling based on the current coherent eigenvalues ​​according to the electromagnetic reciprocity principle, so as to obtain the forward modeled transmitted signal; the initial value of the current coherent eigenvalues ​​is the coherent eigenvalues ​​calculated by the processing unit;

[0038] The weight calculation unit is used to calculate the weight of each coherent feature value based on the difference between the transmitted signal and the forward-modeled transmitted signal.

[0039] The first frequency response function calculation unit is used to generate a new coherence matrix based on the weights of each coherent eigenvalue and solve the linear equation using SVD decomposition to obtain the ground frequency response function at each frequency point in each observation point; and

[0040] The second frequency response function calculation unit is used to solve for the optimal frequency response function of each observation point by using the least squares robust estimation method based on the frequency response function of each frequency point in each observation point.

[0041] Furthermore, the second frequency response function calculation unit also includes:

[0042] The error calculation unit is used to calculate the mean square error of the optimal frequency response function at each observation point; and

[0043] The judgment unit is used to determine whether the mean square error is less than the specified error. If not, the best frequency response function of each frequency point in each current observation point is returned as the current feature value to S4; if yes, the best frequency response function of each frequency point in each observation point is obtained.

[0044] Furthermore, the weight calculation unit is also used to calculate the weights of each coherent feature value according to formula (3):

[0045] w i =1 / |x i -x i ′| (3)

[0046] Where W represents the weight, i represents the order number of the coherent feature value, and i is a positive integer greater than 0; X is the transmitted signal, which is directly recorded from the transmitter and does not include environmental noise; X′ is the forward-modeling transmitted signal.

[0047] Thirdly, embodiments of the present invention provide a storage medium storing instructions that, when executed on a computer, perform a method for denoising the time-frequency electromagnetic data.

[0048] Compared with the prior art, the embodiments of the present invention have the following advantages and beneficial effects:

[0049] This invention discloses a method, system, and storage medium for denoising time-frequency electromagnetic data. It generates a coherence matrix using a time-varying spectrum table of the received and transmitted time-varying electromagnetic signals. The coherence matrix is ​​then approximated and decomposed using SVD to calculate several coherent eigenvectors and eigenvalues ​​between the transmitted and received signals. Based on the electromagnetic reciprocity principle, the received signal is used as the transmitted signal, and the response of the reverse transmitted signal is solved using the current coherent eigenvalues ​​to obtain the forward-modeled transmitted signal. The weights of each coherent eigenvalue are calculated based on the difference between the transmitted signal and the forward-modeled transmitted signal. A new coherence matrix is ​​generated based on the weights of each coherent eigenvalue, and the linear equations are solved using SVD to obtain the ground frequency response function at each observation point. Finally, using the least squares robust estimation method, the optimal ground frequency response function at each observation point is solved based on the ground frequency response function at each observation point. This method solves the technical problem of difficulty in denoising time-frequency electromagnetic data obtained by the prior art using the time-frequency electromagnetic method. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating a method for denoising time-frequency electromagnetic data.

[0052] Figure 2 This is a schematic diagram of the system structure for time-frequency electromagnetic data.

[0053] Figure 3 This is a flowchart illustrating a method for denoising time-frequency electromagnetic data as an example.

[0054] Figure 4 The amplitude and phase diagrams of the frequency response function after denoising using the traditional Stack method and the method of this embodiment are shown for observation point A, where... Figure 4 'a' represents the amplitude diagram. Figure 4 b represents the phase map, old indicates processing using the traditional method, and new indicates processing using the method of this invention.

[0055] Figure 5 The amplitude and phase diagrams of the frequency response function after denoising using the traditional Stack method and the method of this embodiment are shown for observation point B, where... Figure 5'a' represents the amplitude diagram. Figure 5 b represents the phase map, old indicates processing using the traditional method, and new indicates processing using the method of this invention.

[0056] Figure 6 The amplitude and phase diagrams of the frequency response function after denoising using the traditional Stack method and the method of this embodiment are shown for observation point C, where... Figure 6 'a' represents the amplitude diagram. Figure 6 b represents the phase map, old indicates processing using the traditional method, and new indicates processing using the method of this invention.

[0057] Figure 7 This is a comparison chart of the amplitude and phase results after processing by the traditional Stack method and the method of this embodiment of the invention at the actual measured data observation points. Figure 7 'a' represents the amplitude diagram after processing using the traditional Stack method; Figure 7 b represents the phase comparison diagram after processing by the traditional Stack method; Figure 7 c represents the amplitude diagram after processing by the method of this embodiment of the invention; Figure 7 d represents the phase map after processing by the method in this embodiment of the invention.

[0058] In the above figures, the gray lines represent traditional methods, and the black lines represent the methods of the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0060] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other embodiments, well-known structures, circuits, materials, or methods have not been specifically described in order to avoid obscuring the invention.

[0061] Throughout this specification, references to "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "an embodiment," "an example," "an example," or "an example" appearing in various places throughout the specification do not necessarily refer to the same embodiment or example. Furthermore, specific features, structures, or characteristics can be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. Moreover, those skilled in the art will understand that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0062] In the description of this invention, the terms "front", "rear", "left", "right", "up", "down", "vertical", "horizontal", "high", "low", "inner", and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting the scope of protection of this invention.

[0063] Example

[0064] To address the technical problem of difficulty in denoising time-frequency electromagnetic data obtained using the time-frequency electromagnetic method in existing technologies, in a first aspect, embodiments of the present invention provide a method for denoising time-frequency electromagnetic data, referring to... Figure 1 As shown, it includes:

[0065] S1. Preprocess the raw time-frequency electromagnetic data to obtain the time-varying spectrum table of the time-frequency electromagnetic received signal and the time-varying spectrum table of the time-frequency electromagnetic transmitted signal.

[0066] S2. Using the time-varying spectrum table of the time-frequency electromagnetic received signal and the time-varying spectrum table of the time-frequency electromagnetic transmitted signal, a coherence matrix is ​​generated by utilizing the coherence between the transmitted and received signals;

[0067] S3. Perform approximate diagonalization and SVD decomposition on the coherence matrix to calculate several coherent eigenvectors and coherent eigenvalues ​​between the transmitted and received signals;

[0068] S4. Based on the principle of electromagnetic reciprocity, the received signal is used as the transmitted signal, and the response of the reverse transmitted signal is solved by forward modeling based on the current coherent eigenvalues ​​to obtain the forward modeled transmitted signal; the initial value of the current coherent eigenvalues ​​is the coherent eigenvalue of S3.

[0069] S5. Calculate the weights of each coherent eigenvalue based on the difference between the transmitted signal and the forward-modeled transmitted signal;

[0070] S6. Generate a new coherence matrix based on the weights of each coherent eigenvalue and solve the linear equation using SVD decomposition to obtain the earth frequency response function at each frequency point in each observation point;

[0071] S7. Using the least squares robust estimation method, solve for the optimal geodetic frequency response function at each observation point based on the geodetic frequency response function at each frequency point.

[0072] Therefore, this embodiment of the invention generates a coherence matrix by using a time-varying spectrum table of the time-frequency electromagnetic received signal and a time-varying spectrum table of the time-frequency electromagnetic transmitted signal; performs approximate diagonalization and SVD decomposition on the coherence matrix to calculate several coherent eigenvectors and coherent eigenvalues ​​between the transmitted and received signals; based on the electromagnetic reciprocity principle, the received signal is used as the transmitted signal to solve the response of the reverse transmitted signal according to the current coherent eigenvalues ​​in forward modeling, thus obtaining the forward modeled transmitted signal; the weights of each coherent eigenvalue are calculated based on the difference between the transmitted signal and the forward modeled transmitted signal; a new coherence matrix is ​​generated based on the weights of each coherent eigenvalue, and the linear equation is solved using SVD decomposition to obtain the earth frequency response function at each frequency point in each observation point; using the least squares robust estimation method, the optimal earth frequency response function at each frequency point in each observation point is solved based on the earth frequency response function at each frequency point in each observation point, thus solving the technical problem of difficulty in denoising time-frequency electromagnetic data obtained by the time-frequency electromagnetic method in the prior art.

[0073] Furthermore, S4. Based on the principle of electromagnetic reciprocity, the received signal is used as the transmitted signal, and the response of the reverse transmitted signal is solved by forward modeling using coherent eigenvalues ​​to obtain the forward modeled transmitted signal; including:

[0074] The coherence matrix of the response to the reverse transmission signal is formed according to formula (2), and SVD decomposition is performed to obtain the forward transmission signal X′ of the transmission signal.

[0075] X′=EY (2)

[0076] Where X′ is the forward-modeled transmitted signal; E represents the received signal, which includes environmental noise; and Y represents the ground frequency response function.

[0077] Furthermore, S5. Calculate the weights of each coherent eigenvalue based on the difference between the transmitted signal and the forward-modeled transmitted signal; including:

[0078] The weights of each coherent eigenvalue are calculated according to formula (3):

[0079] w i =1 / |x i -x i′| (3)

[0080] Where W represents the weight, i represents the order number of the coherent feature value, and i is a positive integer greater than 0; X is the transmitted signal, which is directly recorded from the transmitter and does not include environmental noise; X′ is the forward-modeling transmitted signal.

[0081] Furthermore, S6. A new coherence matrix is ​​generated based on the weights of each coherent eigenvalue, and the linear equation is solved using SVD decomposition to obtain the ground frequency response function at each frequency point of each observation point; including:

[0082] A new coherence matrix is ​​generated based on the weights of each coherent eigenvalue and formula (4):

[0083] Y = (X T X) -1 X T WE (4)

[0084] Where Y represents the ground frequency response function, X represents the transmitted signal, E represents the received signal, W represents the weights, and T represents the matrix transpose.

[0085] Furthermore, S7. The optimal frequency response function at each observation point is obtained using the least squares robust estimation method; this also includes:

[0086] Calculate the mean square error of the optimal geodetic frequency response function at each observation point;

[0087] Determine whether the mean square error is less than the specified error. If not, return the best frequency response function of each frequency point in each current observation point as the current feature value to S4; if yes, obtain the best frequency response function of each frequency point in each observation point.

[0088] Furthermore, the method for denoising the time-frequency electromagnetic data further includes:

[0089] Output the optimal geodetic frequency response function for each observation point.

[0090] An exemplary method for denoising time-frequency electromagnetic data, refer to Figure 3 As shown, the process includes: First, preprocessing the raw data. This involves processing the received and transmitted signals to create time-varying spectrum tables for the received and transmitted signals.

[0091] The second step is to obtain the ground frequency response function Y (impedance tensor). Utilizing the coherence between the transmitter and receiver, a coherence matrix is ​​formed according to equation (1). The coherence matrix is ​​then approximately diagonalized and decomposed using SVD to obtain the coherent eigenvectors and eigenvalues ​​between the transmitted and received signals, thus obtaining the ground frequency response function Y (impedance tensor).

[0092] The third step is to solve for the response X′ of the reverse transmitted signal. According to the principle of electromagnetic reciprocity, the received signal E is regarded as the transmitted signal. The forward model of the received signal X′, under the condition of no noise, X′ should be consistent with the transmitted signal X. The coherence matrix for solving the response X′ of the reverse transmitted signal is formed according to equation (2), and SVD decomposition is performed to solve for the response X′ of the reverse transmitted signal.

[0093] Step 4: Calculate the weights of coherent eigenvalues. Based on the difference between the transmitted signal X and the forward-modeled signal X′, calculate the weight W of each coherent eigenvalue. The calculation formula is shown in equation (3), which is the reciprocal of the absolute value of the difference between the transmitted signal X and the forward-modeled transmitted signal X′. The larger the difference, the smaller the weight.

[0094] The fifth step is to obtain the ground frequency response function Y (impedance tensor) at each frequency point of each observation point. Using the weights of each eigenvalue, a coherence matrix is ​​formed according to equation (4). The calculation method adopts SVD decomposition to solve the linear equation and obtain the ground frequency response function Y (impedance tensor) at each frequency point of each observation point.

[0095] Y = (X T X) -1 X T E (1)

[0096] X′=EY (2)

[0097] w i =1 / |x i -x i ′|, (3)

[0098] Y = (X T X) -1 X T WE (4)

[0099] Where X is the input signal, directly recorded from the transmitter, excluding environmental noise; X′ is the forward modeling input signal; E is the received (output) signal, including environmental noise; Y represents the impedance tensor, reflecting information about the underground medium; W is the weight; and T represents the matrix transpose.

[0100] The expressions for X, Y, E, X′, and W are shown below:

[0101]

[0102]

[0103] The sixth step is to use the Robust technique to solve for the optimal ground frequency response function Y (impedance tensor) at each frequency point. The least squares robust estimation method (Robust) is introduced across the entire frequency band at each measurement point to solve for the optimal ground frequency response function Y (impedance tensor) at each observation point, and the mean square error ε is calculated.

[0104] Step 7: If the mean square error ε meets the accuracy requirements, terminate the calculation.

[0105] Compare the mean square error ε with the setting error ε′ to determine whether the earth frequency response function Y obtained in step 6 is an unbiased optimal estimate. When the mean square error ε is less than the setting error ε′, the earth frequency response function Y obtained in step 6 is considered an unbiased optimal estimate, i.e., it is considered the true earth frequency response function. At this point, the calculation ends, and the earth frequency response function Y (impedance tensor reflecting the information of the underground medium) obtained in step 5 is output.

[0106] Step 8: If the mean square error ε does not meet the accuracy requirements, perform iterative calculations to continue finding the optimal solution.

[0107] If the calculated mean square error ε is greater than the set error ε′, the ground frequency response function Y obtained in step 5 is used as the initial eigenvalue for the next iteration. Returning to step 3, the forward modeling is continued to obtain X′, the weights of each eigenvalue are calculated, the ground frequency response function Y (impedance tensor) for each frequency is obtained, the optimal ground frequency response function Y (impedance tensor) is calculated using the least squares robust estimation method, the mean square error ε is calculated, and the requirement is compared. This process is repeated until the mean square error ε after the nth iteration is less than the set error ε′. At this point, the eigenvalue (Y) that has converged stably during the iteration is obtained as the optimal estimate of the coherence between the transmitted and received signals, thus obtaining the unbiased optimal estimate of noise suppression. The ground frequency response function Y (impedance tensor reflecting underground medium information) for each frequency after the nth iteration is output.

[0108] Step 8: Calculation complete, output the optimal ground frequency response function Y (impedance tensor reflecting information of the underground medium) for each frequency point.

[0109] Compared with existing technologies, this invention has the following beneficial technical effects: Traditional processing methods mainly use Stack technology (a simple statistical analysis method) to process the received signal, which is essentially an arithmetic average algorithm. When the noise signal meets the Gaussian distribution noise condition, it can effectively suppress noise. The main problem with traditional processing methods is that when there is non-Gaussian noise, the processing effect of traditional Stack technology deteriorates significantly, resulting in distortion points in the obtained frequency data, and the data cannot truly reflect the information of the underground medium. To address the above situation, processors generally take two remedial measures: one is to select time-series signals, removing signals with obvious non-random noise before time-series processing; the other is to manually edit the distortion points after obtaining the frequency data. Regardless of the remedial method, it is relatively time-consuming and labor-intensive, and at the same time reduces the reliability of the data.

[0110] This invention, based on the fundamental principles of time-frequency electromagnetic timing data processing, introduces the electromagnetic reciprocity principle. Utilizing electromagnetic reciprocity, multiple receivers are considered as array transmitters, and transmitters as receiving antennas. The statistical characteristics of the transmitted signal received are consistent with the statistical characteristics of the transmitted signal itself under noise-free conditions. The difference between the two can be used as the noise weight W in the original received signal. This noise weight W is then incorporated into the data statistics to achieve denoising. This invention satisfies denoising requirements under Gaussian noise distribution conditions while overcoming the shortcomings of traditional Stack technology in denoising non-Gaussian noise distributions. Furthermore, it eliminates the need for human-computer interaction measures to remedy the deficiencies of traditional Stack technology. This invention not only automates denoising processing, ensuring data reliability and improving data quality, but also reduces the workload of processing personnel, significantly improving work efficiency.

[0111] This invention preprocesses the raw time-frequency electromagnetic data to eliminate noise or noise that does not conform to a Gaussian distribution. Utilizing electromagnetic reciprocity, multiple receivers are considered as array transmitters, and transmitters as receiving antennas. Under noise-free conditions, the statistical characteristics of the transmitted signal received should be consistent with those of the transmitted signal itself. The difference between these two characteristics can be seen as a measure of noise in the received raw signal. Based on this, a coherence matrix is ​​formed, approximately diagonalized, and decomposed using SVD to obtain the coherent eigenvectors and eigenvalues ​​between the transmitted and received signals, forming an array dataset. An iterative sequence is constructed, weights are introduced, and an unbiased optimal estimate of the suppressed noise is obtained. This unbiased optimal estimate satisfies the high-order statistical analysis requirements of robust processing, yielding a time-frequency electromagnetic method geodetic frequency response function that approximates the actual underground conditions. Employing the electromagnetic reciprocity algorithm avoids the problem of getting trapped in local optima when non-random noise is present, improving the speed and accuracy of the solution, making time-frequency electromagnetic noise suppression more efficient, and providing a valuable tool for geophysical exploration personnel. This invention achieves better noise suppression, with an accuracy of over 90%. It can serve as an effective tool for time-frequency electromagnetic processing personnel.

[0112] Sample point test

[0113] This invention employs actual production data from a specific work area for method testing and research. Based on on-site reconnaissance, data from measurement points near a strong interference source were selected for comparative analysis of traditional Stack and electromagnetic reciprocity-based processing techniques. Three measurement points, A, B, and C, were selected within a certain range, with reference to... Figure 4-6 As shown, by comparing the underground medium information obtained under conditions with and without interference sources, the application effect of the method in this embodiment of the invention is illustrated.

[0114] When the measuring point is far from the interference source, it is recorded as observation point A, and the processing result is as follows: Figure 4 As shown. Reference Figure 4 The results of the two methods, a and 4b, are basically consistent, with the amplitude and phase essentially overlapping and the data showing regular changes. Both methods effectively captured the real signals generated by the underground medium, indicating that the method of the present invention is feasible and effective.

[0115] When the distance between the measuring point and the interference source is moderate, it is recorded as observation point B, and the processing result is as follows: Figure 5 As shown. Reference Figure 5 For examples a and 5b, traditional methods distort the mid-to-high frequency signals, causing amplitude and phase jumps at local frequency points. While the data is generally usable, further human-computer interaction is needed to carefully identify the true mid-to-high frequency data generated by the underground medium. The electromagnetic reciprocity-based method significantly reduces noise from strong interference, showing regular amplitude and phase changes, thus better restoring the mid-to-high frequency data generated by the underground medium.

[0116] When the observation point is close to the interference source, denoted as observation point C, the processing result is as follows: Figure 6 As shown. Reference Figure 6 The results of methods a and 6b are basically consistent in the low-frequency range. The traditional method shows complete distortion in the mid-to-high frequency range, with abrupt jumps in amplitude and phase, making it impossible to identify the mid-to-high frequency data generated by the underground medium, rendering the data unusable. The method based on electromagnetic reciprocity still shows good noise reduction for strong interference, with the obtained amplitude and phase exhibiting regular changes, essentially restoring the mid-to-high frequency data signal generated by the underground medium.

[0117] The comparative test results at the three measurement points above show that traditional Stack technology acquires distorted data in non-Gaussian noise (strong interference source) areas and cannot effectively acquire signals generated by the underground medium. The electromagnetic reciprocity principle-based processing technology overcomes the shortcomings of traditional Stack technology in non-Gaussian noise (strong interference source) areas, effectively eliminating or suppressing interference, achieving good noise reduction, and effectively acquiring underground medium signals. Compared with traditional methods, it features a fully automated, high-fidelity, and efficient process, demonstrating outstanding practicality and economy.

[0118] Therefore, this invention utilizes the principle of electromagnetic reciprocity and the coherence between transmitted and received signals to optimize the denoising algorithm for acquired data containing non-Gaussian noise. This not only yields a more realistic and reliable time-frequency electromagnetic method geodetic frequency response function that approximates the actual underground conditions, but also improves the speed and accuracy of the solution, making time-frequency electromagnetic processing more efficient and providing geological interpreters with more reliable underground medium information parameters. This invention employs an iterative algorithm that supports electromagnetic reciprocity, thereby improving the accuracy of denoising and obtaining effective and reliable original information about the underground medium, exhibiting superior generalization ability compared to other methods.

[0119] Implementation Cases

[0120] To verify the effectiveness of the algorithm, data from an actual survey line in a certain work area were processed using both traditional methods and the time-frequency electromagnetic data denoising method of this invention. The processing results are as follows: Figure 7 As shown.

[0121] from Figure 7 As can be seen from a and 7b, the data processed using traditional Stack technology is significantly affected by interference from multiple frequency bands, manifested as multiple irregular jumps in amplitude and phase at high frequencies, which deviate from the true response of the underground medium information. Figure 7 c and 7d are the results processed using an electromagnetic reciprocity-based method. Compared with the results processed by traditional Stack technology, which showed multiple jump points (noise) in the mid-to-high frequency band, the new method can effectively suppress or remove mid-to-high frequency noise. The obtained amplitude and phase curves show regular changes and no obvious jump points are observed, which is a true response of the underground medium information.

[0122] Secondly, embodiments of the present invention provide a denoising system for time-frequency electromagnetic data, with reference to... Figure 2 As shown, it includes:

[0123] The preprocessing unit is used to preprocess the raw time-frequency electromagnetic data to obtain the time-series time-varying spectrum table of the time-frequency electromagnetic received signal and the time-series time-varying spectrum table of the time-frequency electromagnetic transmitted signal.

[0124] The coherence matrix forming unit is used to generate a coherence matrix by utilizing the time-varying spectrum table of the time-frequency electromagnetic received signal and the time-varying spectrum table of the time-frequency electromagnetic transmitted signal, taking advantage of the coherence between the transmitted and received signals.

[0125] The processing unit is used to perform approximate diagonalization and SVD decomposition on the coherence matrix, and calculate several coherent feature vectors and coherent feature values ​​between the transmitted signal and the received signal.

[0126] The forward modeling unit is used to solve the response of the reverse transmitted signal by taking the received signal as the transmitted signal and performing forward modeling based on the current coherent eigenvalues ​​according to the electromagnetic reciprocity principle, so as to obtain the forward modeled transmitted signal; the initial value of the current coherent eigenvalues ​​is the coherent eigenvalues ​​calculated by the processing unit;

[0127] The weight calculation unit is used to calculate the weight of each coherent feature value based on the difference between the transmitted signal and the forward-modeled transmitted signal.

[0128] The first frequency response function calculation unit is used to generate a new coherence matrix based on the weights of each coherent eigenvalue and solve the linear equation using SVD decomposition to obtain the ground frequency response function at each frequency point in each observation point; and

[0129] The second frequency response function calculation unit is used to solve for the optimal frequency response function of each observation point by using the least squares robust estimation method based on the frequency response function of each frequency point in each observation point.

[0130] Furthermore, the second frequency response function calculation unit also includes:

[0131] The error calculation unit is used to calculate the mean square error of the optimal frequency response function at each observation point; and

[0132] The judgment unit is used to determine whether the mean square error is less than the specified error. If not, the best frequency response function of each frequency point in each current observation point is returned as the current feature value to S4; if yes, the best frequency response function of each frequency point in each observation point is obtained.

[0133] Furthermore, the weight calculation unit is also used to calculate the weights of each coherent feature value according to formula (3):

[0134] w i =1 / |x i -x i ′| (3)

[0135] Where W represents the weight, i represents the order number of the coherent feature value, and i is a positive integer greater than 0; X is the transmitted signal, which is directly recorded from the transmitter and does not include environmental noise; X′ is the forward-modeling transmitted signal.

[0136] Thirdly, embodiments of the present invention provide a storage medium storing instructions that, when executed on a computer, perform a method for denoising the time-frequency electromagnetic data.

[0137] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for denoising time-frequency electromagnetic data, characterized in that, include: S1. Preprocess the raw time-frequency electromagnetic data to obtain the time-varying spectrum table of the time-frequency electromagnetic received signal and the time-varying spectrum table of the time-frequency electromagnetic transmitted signal. S2. Using the time-varying spectrum table of the time-frequency electromagnetic received signal and the time-varying spectrum table of the time-frequency electromagnetic transmitted signal, a coherence matrix is ​​generated by utilizing the coherence between the transmitted and received signals; S3. Perform approximate diagonalization and SVD decomposition on the coherence matrix to calculate several coherent eigenvectors and coherent eigenvalues ​​between the transmitted and received signals; S4. Based on the principle of electromagnetic reciprocity, the received signal is used as the transmitted signal, and the response of the reverse transmitted signal is solved by forward modeling based on the current coherent eigenvalues ​​to obtain the forward modeled transmitted signal; the initial value of the current coherent eigenvalues ​​is the coherent eigenvalue of S3. S5. Calculate the weights of each coherent eigenvalue based on the difference between the transmitted signal and the forward-modeled transmitted signal; S6. Generate a new coherence matrix based on the weights of each coherent eigenvalue and solve the linear equation using SVD decomposition to obtain the earth frequency response function at each frequency point in each observation point; S7. Using the least squares robust estimation method, solve for the optimal geodetic frequency response function at each observation point based on the geodetic frequency response function at each frequency point.

2. The method for denoising time-frequency electromagnetic data as described in claim 1, characterized in that, S4. Based on the principle of electromagnetic reciprocity, the received signal is used as the transmitted signal, and the response of the reverse transmitted signal is solved by forward modeling using coherent eigenvalues ​​to obtain the forward model of the transmitted signal; including: The coherence matrix of the response to the reverse transmission signal is formed according to formula (2), and SVD decomposition is performed to obtain the forward transmission signal X′ of the transmission signal. X′=EY (2) Where X′ is the forward-modeled transmitted signal; E represents the received signal, which includes environmental noise; and Y represents the ground frequency response function.

3. The method for denoising time-frequency electromagnetic data as described in claim 2, characterized in that, S5. Calculate the weights of each coherent eigenvalue based on the difference between the transmitted signal and the forward-modeled transmitted signal; including: The weights of each coherent eigenvalue are calculated according to formula (3): w i =1 / |x i -x i ′| (3) Where W represents the weight, i represents the order number of the coherent feature value, and i is a positive integer greater than 0; X is the transmitted signal, which is directly recorded from the transmitter and does not include environmental noise; X′ is the forward-modeling transmitted signal.

4. The method for denoising time-frequency electromagnetic data as described in claim 3, characterized in that, S6. Generate a new coherence matrix based on the weights of each coherent eigenvalue and solve the linear equation using SVD decomposition to obtain the ground frequency response function at each observation point; including: A new coherence matrix is ​​generated based on the weights of each coherent eigenvalue and formula (4): Y=(X T X) -1 X T WE (4) Where Y represents the ground frequency response function, X represents the transmitted signal, E represents the received signal, W represents the weights, and T represents the matrix transpose.

5. The method for denoising time-frequency electromagnetic data as described in claim 1, characterized in that, S7. Solving for the optimal geodetic frequency response function at each observation point using the least squares robust estimation method; also includes: Calculate the mean square error of the optimal geodetic frequency response function at each observation point; Determine whether the mean square error is less than the specified error. If not, return the best frequency response function of each frequency point in each current observation point as the current feature value to S4; if yes, obtain the best frequency response function of each frequency point in each observation point.

6. The method for denoising time-frequency electromagnetic data as described in claim 1, characterized in that, Also includes: Output the optimal geodetic frequency response function for each observation point.

7. A noise reduction system for time-frequency electromagnetic data, characterized in that, include: The preprocessing unit is used to preprocess the raw time-frequency electromagnetic data to obtain the time-series time-varying spectrum table of the time-frequency electromagnetic received signal and the time-series time-varying spectrum table of the time-frequency electromagnetic transmitted signal. The coherence matrix forming unit is used to generate a coherence matrix by utilizing the time-varying spectrum table of the time-frequency electromagnetic received signal and the time-varying spectrum table of the time-frequency electromagnetic transmitted signal, taking advantage of the coherence between the transmitted and received signals. The processing unit is used to perform approximate diagonalization and SVD decomposition on the coherence matrix, and calculate several coherent feature vectors and coherent feature values ​​between the transmitted signal and the received signal. The forward modeling unit is used to solve the response of the reverse transmitted signal by taking the received signal as the transmitted signal and performing forward modeling based on the current coherent eigenvalues ​​according to the electromagnetic reciprocity principle, so as to obtain the forward modeled transmitted signal; the initial value of the current coherent eigenvalues ​​is the coherent eigenvalues ​​calculated by the processing unit; The weight calculation unit is used to calculate the weight of each coherent feature value based on the difference between the transmitted signal and the forward-modeled transmitted signal. The first frequency response function calculation unit is used to generate a new coherence matrix based on the weights of each coherence eigenvalue and solve the linear equation using SVD decomposition to obtain the earth frequency response function of each frequency point in each observation point. as well as The second frequency response function calculation unit is used to solve for the optimal frequency response function of each observation point by using the least squares robust estimation method based on the frequency response function of each frequency point in each observation point.

8. The noise reduction system for time-frequency electromagnetic data as described in claim 7, characterized in that, The second frequency response function calculation unit also includes: The error calculation unit is used to calculate the mean square error of the optimal frequency response function at each observation point; and The judgment unit is used to determine whether the mean square error is less than the specified error. If not, the best frequency response function of each frequency point in each current observation point is returned as the current feature value to S4; if yes, the best frequency response function of each frequency point in each observation point is obtained.

9. The denoising system for time-frequency electromagnetic data as described in claim 7, characterized in that, The weight calculation unit is also used to calculate the weights of each coherent feature value according to formula (3): w i =1 / |x i -x i ′| (3) Where W represents the weight, i represents the order number of the coherent feature value, and i is a positive integer greater than 0; X is the transmitted signal, which is directly recorded from the transmitter and does not include environmental noise; X′ is the forward-modeling transmitted signal.

10. A storage medium, characterized in that, The storage medium stores instructions that, when executed on a computer, perform the noise reduction method for time-frequency electromagnetic data as described in any one of claims 1-6.

Citation Information

Patent Citations

  • Magnetotelluric impedance measuring method

    CN101071179A

  • Remote reference magnetotelluric impedance calculation method based on correlation detection

    CN105866852A