Logging data noise reduction method, system and device and computer storage medium
By employing eigenmode decomposition and Hurst exponent screening, the problem of low signal-to-noise ratio in nuclear magnetic resonance logging was solved, thereby improving signal quality and enabling accurate calculation of reservoir parameters.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-30
- Publication Date
- 2026-03-24
AI Technical Summary
Current nuclear magnetic resonance logging technology suffers from a low signal-to-noise ratio, leading to a decline in signal quality and affecting the accuracy of reservoir parameter calculations. In particular, weak signals are easily filtered out as noise in low-porosity reservoirs, causing deviations in porosity calculations.
By using eigenmode decomposition, detrended fluctuation analysis, and Hurst exponent screening, the effective components and noise of the nuclear magnetic resonance echo signal are separated, a reconstructed signal is generated and corrected, and the signal-to-noise ratio is improved.
It effectively separates noise from effective signals, improves the signal-to-noise ratio of nuclear magnetic resonance logging data, and enhances the accuracy of reservoir evaluation and the precision of oil and gas well inversion.
Smart Images
Figure CN121721737A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of oil and gas exploration and development technology, and more specifically, to a method, system, equipment, and computer storage medium for noise reduction of well logging data. Background Technology
[0002] Nuclear magnetic resonance (NMR) logging, as a geophysical method in oil and gas exploration and development, can provide crucial information such as formation pore structure, fluid type, and content. However, in actual logging environments, the acquired NMR echo signals are often interfered with by various noise sources, including instrument electronic noise, downhole electromagnetic interference, and mechanical vibration noise, leading to a significant deterioration in signal quality. This directly affects the accuracy of subsequent relaxation spectrum inversion and reservoir parameter calculation. Therefore, it is necessary to perform noise reduction processing on NMR logging data, such as using moving average filtering methods.
[0003] However, while the moving average filtering method is computationally simple and easy to implement, it is essentially a low-pass filter. While suppressing high-frequency noise, it excessively smooths signal details. Nuclear magnetic resonance (NMR) echo signals exhibit typical multi-exponential decay characteristics, and early echoes contain rich pore structure information. Moving average processing blurs these key features, leading to a decrease in T2 spectral resolution. This is particularly problematic in low-porosity reservoirs, where weak, useful signals are easily filtered out as noise, causing inaccuracies in porosity calculations. Furthermore, the processing of low signal-to-noise ratio NMR logging data is often ineffective, severely impacting the reliability of reservoir evaluation.
[0004] In summary, improving the signal-to-noise ratio of nuclear magnetic resonance logging data is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0005] The purpose of this application is to provide a method for denoising well logging data, which can, to some extent, solve the technical problem of how to improve the signal-to-noise ratio of nuclear magnetic resonance logging data. This application also provides a well logging data denoising system, electronic equipment, and a computer-readable storage medium.
[0006] To achieve the above objectives, this application provides the following technical solution:
[0007] A method for denoising well logging data includes:
[0008] Acquire initial logging data obtained via nuclear magnetic resonance;
[0009] The initial logging data is subjected to eigenmode decomposition to obtain a set number of candidate modes;
[0010] Detrended fluctuation analysis is performed on the candidate modes to obtain fluctuation data;
[0011] Based on the fluctuation data, the Hurst index of the candidate mode is generated;
[0012] The candidate modes are filtered according to the Hurst index to obtain the target mode;
[0013] The target modes are superimposed to generate a reconstructed signal;
[0014] The reconstructed signal is corrected based on the initial logging data to obtain the target logging data.
[0015] Preferably, the step of performing eigenmode decomposition on the initial logging data to obtain a predetermined number of candidate modes includes:
[0016] The parameters for the eigenmode decomposition are determined, including filter length, number of filter initializations, kurtosis shift order, and maximum number of iterations.
[0017] The frequency band of the initial logging data is divided into frequency band segments equal to the number of filters, and the upper and lower cutoff frequencies of each frequency band segment are determined.
[0018] Generate FIR filters corresponding to the frequency band segments;
[0019] The initial logging data is filtered using an FIR filter to obtain the current cycle's unprocessed mode;
[0020] The parameters of the FIR filter are updated based on the mode to be processed in the current round to maximize the kurtosis value of the FIR filter output;
[0021] Number of update iterations;
[0022] If the number of iterations is less than the maximum number of iterations, then return to the step of filtering the initial logging data with an FIR filter to obtain the mode to be processed in the current round;
[0023] In response to the number of iterations being equal to the maximum number of iterations, the unprocessed modes obtained in the last round are filtered to obtain a set number of candidate modes.
[0024] Preferably, the step of filtering the modes to be processed obtained in the last round to obtain a set number of candidate modes includes:
[0025] For the modes to be processed obtained in the last round, generate correlation coefficients between each pair of modes to be processed, and generate a correlation matrix based on the correlation coefficients;
[0026] In the off-diagonal region of the correlation matrix, identify the first and second modes with the largest correlation coefficients;
[0027] Generate the kurtosis values for the first and second modes respectively;
[0028] Based on the kurtosis value, discard the mode with the smaller kurtosis value between the first and second modes;
[0029] Update the correlation matrix and decrease the total number of modes by 1;
[0030] If the total number of modes is greater than the set number, then return to the off-diagonal region of the correlation matrix to determine the first and second modes with the largest values, and then proceed with the steps thereafter.
[0031] If the total number of modes equals the set number, the remaining modes are selected as candidate modes.
[0032] Preferably, the step of performing detrended fluctuation analysis on the candidate modes to obtain fluctuation data includes:
[0033] Construct the cumulative deviation sequence of the candidate modes;
[0034] The cumulative deviation sequence is segmented to obtain sequence intervals that do not overlap and have a set length;
[0035] The local trend is obtained by fitting the sequence interval using the least squares method.
[0036] The local trend is detrended and variance values are generated;
[0037] Based on the variance value, the root mean square fluctuation of the sequence interval is generated;
[0038] The root mean square fluctuation of the sequence interval is used as fluctuation data.
[0039] Preferably, generating the Hurst exponent of the candidate mode based on the fluctuation data includes:
[0040] Perform a logarithmic operation on the root mean square fluctuation to generate the first result.
[0041] Perform a logarithmic operation on the sequence interval to generate a second result;
[0042] Perform linear regression on the first calculation result and the second calculation result to obtain the linear regression result;
[0043] The Hurst index of the candidate mode is generated based on the slope of the straight line from the linear regression result.
[0044] Preferably, the step of filtering the candidate modes according to the Hurst index to obtain the target mode includes:
[0045] Obtain the set modal parameters;
[0046] If the Hurst exponent of the candidate mode is greater than the mode index, then the candidate mode is adopted as the target mode.
[0047] If the Hurst index of the candidate mode is less than or equal to the mode index, the candidate mode is eliminated.
[0048] Preferably, the step of correcting the reconstructed signal based on the initial logging data to obtain the target logging data includes:
[0049] Get the set data length value;
[0050] In the initial logging data, a first dataset corresponding to the data length value is determined;
[0051] In the reconstructed signal, a second dataset corresponding to the data length value is determined;
[0052] A correction factor is generated based on the ratio between the cumulative data values of the first dataset and the cumulative data values of the second dataset;
[0053] The reconstructed signal is processed according to the correction factor to obtain the logging target data.
[0054] A well logging data noise reduction system, comprising:
[0055] The well logging initial data acquisition module is used to acquire the well logging initial data obtained through nuclear magnetic resonance.
[0056] The feature mode decomposition module is used to perform feature mode decomposition on the initial logging data to obtain a set number of candidate modes;
[0057] The detrended fluctuation analysis module is used to perform detrended fluctuation analysis on the candidate modes to obtain fluctuation data;
[0058] An index generation module is used to generate the Hurst index of the candidate mode based on the fluctuation data.
[0059] The filtering module is used to filter the candidate modes according to the Hurst index to obtain the target mode;
[0060] The superposition module is used to superimpose the target modes to generate a reconstructed signal;
[0061] The correction module is used to correct the reconstructed signal based on the initial logging data to obtain the target logging data.
[0062] An electronic device, comprising:
[0063] Memory, used to store computer programs;
[0064] A processor is used to implement the steps of any of the above-described logging data noise reduction methods when executing the computer program.
[0065] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described well logging data noise reduction methods.
[0066] This application provides a well logging data noise reduction method, which involves acquiring initial well logging data obtained through nuclear magnetic resonance (NMR); performing eigenmode decomposition on the initial well logging data to obtain a set number of candidate modes; performing detrended fluctuation analysis on the candidate modes to obtain fluctuation data; generating Hurst exponents for the candidate modes based on the fluctuation data; filtering the candidate modes according to the Hurst exponents to obtain target modes; superimposing the target modes to generate a reconstructed signal; and correcting the reconstructed signal based on the initial well logging data to obtain the target well logging data. In this application, eigenmode decomposition (EMD) is performed on the initial logging data acquired via nuclear magnetic resonance (NMR). Since EMD can fully utilize the multi-exponential decay characteristics and non-stationary distribution features of NMR echo signals, it more thoroughly separates the effective attenuation components from background noise, reducing mode aliasing. Therefore, it can obtain candidate modes with good robustness to multi-component noise interference. Subsequently, detrended fluctuation analysis is performed on the candidate modes to generate the Hurst exponent, which removes the trend components from the modal data, revealing the inherent fluctuation characteristics of the initial logging data sequence. The candidate modes are then screened according to the Hurst exponent, objectively selecting the target modes and reducing human error. Finally, the target modes are superimposed to generate a reconstructed signal, which is then corrected based on the initial logging data. This yields logging target data that adaptively separates noise and effective signals, improving the signal-to-noise ratio of NMR logging data. The logging data denoising system, electronic device, and computer-readable storage medium provided in this application also solve the corresponding technical problems. Attached Figure Description
[0067] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0068] Figure 1 A flowchart illustrating a well logging data noise reduction method provided in this application embodiment;
[0069] Figure 2 This is a schematic diagram of a T2 spectrum containing two peaks constructed from a numerical simulation experiment;
[0070] Figure 3 This is a schematic diagram of one-dimensional nuclear magnetic resonance echo train data with an echo interval of 0.2ms, a number of echoes of 4096, and an SNR of 10.
[0071] Figure 4 This is a schematic diagram of one-dimensional nuclear magnetic resonance echo train data with an echo interval of 0.2ms, a number of echoes of 4096, and an SNR of 20.
[0072] Figure 5 Decomposition of the proposed scheme Figure 3 A schematic diagram of the results for mid-echo train data;
[0073] Figure 6 Decomposition of the proposed scheme Figure 4 A schematic diagram of the results for mid-echo train data;
[0074] Figure 7 For the purposes of this application Figure 3 A schematic diagram of the denoising results for mid-echo train data;
[0075] Figure 8 For the purposes of this application Figure 4 A schematic diagram of the denoising results for mid-echo train data;
[0076] Figure 9 for Figure 3 and Figure 7 A comparison of the nuclear magnetic resonance T2 spectrum obtained by inverting the echo train data with the constructed T2 spectrum model;
[0077] Figure 10 for Figure 4 and Figure 8 A comparison of the nuclear magnetic resonance T2 spectrum obtained by inverting the echo train data with the constructed T2 spectrum model;
[0078] Figure 11 This is a schematic diagram of the structure of a well logging data noise reduction system provided in an embodiment of this application;
[0079] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application;
[0080] Figure 13 This is another structural schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation
[0081] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0082] Nuclear magnetic resonance (NMR) logging, as a geophysical method in oil and gas exploration and development, can provide crucial information such as formation pore structure, fluid type, and content. However, in actual logging environments, the acquired NMR echo signals are often interfered with by various noise sources, including instrument electronic noise, downhole electromagnetic interference, and mechanical vibration noise, resulting in a significant deterioration in signal quality. This directly affects the accuracy of subsequent relaxation spectrum inversion and reservoir parameter calculations.
[0083] Currently, methods such as moving average filtering, wavelet transform denoising, and singular value decomposition can be used to denoise NMR data. However, these methods have significant limitations: While moving average filtering is computationally simple and easy to implement, it is essentially a low-pass filter. While suppressing high-frequency noise, it excessively smooths signal details. NMR echo signals have typical multi-exponential decay characteristics, and early echoes contain rich pore structure information. Moving average processing blurs these key features, leading to a decrease in T2 spectral resolution. This is especially true in low-porosity reservoirs, where weak useful signals are easily filtered out as noise, causing errors in porosity calculations. Wavelet transform denoising achieves noise separation through multi-scale decomposition, but its performance is highly dependent on the selection of wavelet basis functions and threshold rules. The attenuation characteristics of NMR echo signals have poor matching with common wavelet basis functions (such as Daubechies and Symlets), resulting in signal energy dispersion in the transform domain, making effective noise separation difficult. Furthermore, the selection of wavelet thresholds often relies on experience and lacks adaptability, easily producing pseudo-Gibbs phenomena under complex geological conditions, introducing artificial oscillation interference. Singular value decomposition (SVD) is a method for noise reduction based on the difference in singular value distribution between signal and noise. However, it faces two main challenges in practical applications: First, the construction method of the nuclear magnetic resonance echo matrix directly affects the characteristics of the singular value distribution. An inappropriate matrix construction can lead to the overlap of singular values of signal and noise, making it difficult to separate them effectively. Second, there is a lack of objective standards for determining the singular value truncation number. Excessive truncation will result in the loss of effective signal components, while insufficient truncation will lead to noise residue.
[0084] More importantly, these traditional noise reduction methods often perform poorly on low signal-to-noise ratio NMR logging data, which seriously affects the reliability of reservoir evaluation. The logging data noise reduction scheme provided in this application can improve the signal-to-noise ratio of NMR logging data, thereby enhancing the accuracy of oil and gas well inversion and reservoir evaluation.
[0085] Please see Figure 1 , Figure 1 A flowchart of a well logging data noise reduction method provided in an embodiment of this application.
[0086] This application provides a method for noise reduction of well logging data, which may include the following steps:
[0087] Step S101: Obtain initial logging data acquired via nuclear magnetic resonance.
[0088] In practical applications, initial logging data can be obtained first after acquiring data from oil and gas wells using nuclear magnetic resonance (NMR). It can be represented as , This refers to data collected theoretically. Let s represent the noise term, t represent the data length, t represent the time step, and T represent the transpose. To ensure that the noise statistical characteristics satisfy the Gaussian distribution assumption during subsequent signal processing, thereby improving the accuracy of decomposition and filtering, the initial logging data is preferably real part data after phase correction, rather than modulus data. The specific values of the initial logging data can be flexibly determined according to the application scenario, and are not specifically limited here.
[0089] Step S102: Perform eigenmode decomposition on the initial logging data to obtain a set number of candidate modes.
[0090] In practical applications, after obtaining the initial logging data, the initial logging data can be subjected to eigenmode decomposition to obtain a set number of candidate modes. The set number can be flexibly determined according to actual needs.
[0091] In an exemplary embodiment, during the process of performing Eigenmode Decomposition (EMD) on the initial logging data to obtain a set number of candidate modes, the parameters of the Eigenmode Decomposition can be determined. These parameters include filter length L, filter initialization number K, kurtosis shift order M, and maximum iteration count Y. Considering the characteristics of the logging data, the recommended range for the filter length is [30, 100], the number of filters must satisfy K ≥ N, the recommended range for K is [5, 10], M can be set to 1, Y can be 50, etc., and N represents the set number. The frequency bands of the initial logging data are... The frequency bands are divided into segments equal to the number of filters, and the upper and lower cutoff frequencies of each segment are determined. For the k-th segment (k=0,1,...,K−1), its upper and lower cutoff frequencies are... , , This represents the sampling frequency of the initial logging data; it generates FIR filters corresponding to the frequency band segments, for example, using MATLAB's fir1 function to generate a bandpass FIR filter for each frequency band segment, ultimately obtaining K initial FIR filters, which form a filter bank. The superscript 1 indicates the first iteration; the initial logging data is filtered using an FIR filter to obtain the current round's unprocessed mode; the parameters of the FIR filter are updated based on the current round's unprocessed mode to maximize the kurtosis value of the FIR filter output; the iteration count is updated; if the iteration count is less than the maximum iteration count, the process returns to the step of filtering the initial logging data using an FIR filter to obtain the current round's unprocessed mode; if the iteration count is equal to the maximum iteration count, the unprocessed modes obtained in the last round are filtered to obtain a set number of candidate modes.
[0092] In specific application scenarios, during the process of filtering initial logging data using an FIR filter to obtain the mode to be processed in the current round, for the k-th filter in the i-th iteration... : * indicates a convolution operation, and x represents the data to be filtered. For ease of subsequent mathematical derivation and calculation, the convolution operation is represented as matrix multiplication, with the formula: , Let X represent the k-th mode vector, with dimensions (N-L+1)×1, and let X represent the convolution matrix of the original signal, with dimensions (N-L+1)×1. , This is the k-th filter coefficient vector, with dimension L×1. Correspondingly, in updating the FIR filter parameters based on the current round's mode to maximize the kurtosis value of the FIR filter output, the current mode can be calculated first. autocorrelation function Find the first local maximum point τ1 that appears after the first zero-crossing point τ0 in the autocorrelation spectrum, and set the estimated optimal step size (in units of sampling points) as T. s =τ1, which will be used in the next step to calculate CK; then the filter will be updated to find an optimal filter f. k This makes its output signal u k =Xf k The CK value is maximized, and the objective function of CK is... H represents conjugate transpose. It is the autocorrelation matrix of signal x, with dimension L×L. Let represent the weighted correlation matrix, where Let this be a diagonal weighted matrix with diagonal elements as follows: The above maximization problem is equivalent to solving a generalized eigenvalue problem. λ is the generalized eigenvalue. New filter coefficients. It is then updated to be the same as the maximum generalized eigenvalue λ. max The corresponding generalized eigenvectors; in other words, in each iteration, the newly obtained modalities are used. And estimate the period Ts to recalculate the weighting matrix W M Then, the new eigenvectors are solved to update the filter.
[0093] In specific application scenarios, during the process of filtering the modes to be processed in the last round to obtain a set number of candidate modes, correlation coefficients (CCs) can be generated between each pair of modes to be processed from the K modes to be processed in the last round. A correlation matrix, which is a K×K matrix, is then generated based on these correlation coefficients. Within the off-diagonal region of the correlation matrix, the first mode u with the largest correlation coefficient value is determined. p Second mode u q The first and second modes are most similar; generate kurtosis values for the first and second modes respectively; discard the mode with the smaller kurtosis value from the first and second modes based on the kurtosis values; update the correlation matrix and decrement the total number of modes by 1; if the total number of modes is greater than a set number, return to the off-diagonal region of the correlation matrix to determine the first and second modes with the largest values and proceed with the subsequent steps; if the total number of modes equals the set number, use the remaining modes as candidate modes, which can be represented as... .
[0094] Step S103: Perform detrended fluctuation analysis on the candidate modes to obtain fluctuation data.
[0095] Step S104: Generate the Hurst exponent for the candidate modes based on the fluctuation data.
[0096] Step S105: Filter the candidate modes according to the Hurst index to obtain the target mode.
[0097] In practical applications, the candidate modes obtained through eigenmode decomposition may also contain noise. In order to further reduce the impact of noise, considering that the Hurst exponent quantifies the long-range correlation of time series, the modes can be screened based on it. That is, the candidate modes can be detrended and fluctuation analysis can be performed to obtain fluctuation data. Based on the fluctuation data, the Hurst exponent of the candidate modes can be generated. The candidate modes can be screened according to the Hurst exponent to obtain the target mode.
[0098] In an exemplary embodiment, during the process of performing detrended fluctuation analysis on candidate modes to obtain fluctuation data, candidate modes can be constructed. Cumulative deviation sequence , Representing modes The mean of the cumulative deviation sequence Y; k The sequence is divided into non-overlapping intervals of length s. The set length can be flexibly determined according to the application scenario; for example, it can be taken from s. min to s max Multiple values to cover different time scales, s min and s max This can be determined empirically; by performing least squares fitting on the sequence interval, the local trend Y can be obtained. fit,v v represents the label of the sequence interval; detrending is performed on the local trend and variance values are generated. Based on the variance value, generate the root mean square fluctuation of the sequence interval. , N s =N / s, representing the total number of intervals of length s; the root mean square fluctuation of the sequence intervals is used as fluctuation data.
[0099] In specific application scenarios, during the process of generating the Hurst exponent for candidate modes based on fluctuation data, the fluctuation function typically satisfies a power-law relationship at different scales s. Therefore, a logarithmic operation can be performed on the root mean square fluctuation to generate the first result logF. k (s); Perform logarithmic operation on the sequence interval to generate the second operation result logs; Perform linear regression on the first and second operation results to obtain the linear regression result; Generate the Hurst index of the candidate mode based on the slope of the straight line of the linear regression result.
[0100] In specific application scenarios, during the process of filtering candidate modes according to the Hurst index to obtain the target mode, a set mode index can be obtained. The mode index can be flexibly determined according to the application scenario, such as a mode index of 0.5. If the Hurst index of a candidate mode is greater than the mode index, the candidate mode is taken as the target mode; if the Hurst index of a candidate mode is less than or equal to the mode index, the candidate mode is eliminated.
[0101] Step S106: Superimpose the target modes to generate a reconstructed signal.
[0102] Step S107: Correct the reconstructed signal based on the initial logging data to obtain the target logging data.
[0103] In practical applications, after obtaining the target modes, the target modes can be superimposed to generate a reconstructed signal. Then, the reconstructed signal is corrected based on the initial logging data to obtain the logging target data.
[0104] In an exemplary embodiment, during the process of correcting the reconstructed signal based on the initial logging data to obtain the target logging data, a set data length value E can be acquired; in the initial logging data... In the process, the first dataset corresponding to the data length value is determined; in the reconstructed signal In the first dataset, a second dataset corresponding to the data length value is determined; based on the ratio between the cumulative data values of the first dataset and the cumulative data values of the second dataset, a correction factor is generated. , The reconstructed signal is processed according to the correction factor to obtain the logging target data. , .
[0105] This application provides a well logging data noise reduction method, which involves acquiring initial well logging data obtained through nuclear magnetic resonance (NMR); performing eigenmode decomposition on the initial well logging data to obtain a set number of candidate modes; performing detrended fluctuation analysis on the candidate modes to obtain fluctuation data; generating Hurst exponents for the candidate modes based on the fluctuation data; filtering the candidate modes according to the Hurst exponents to obtain target modes; superimposing the target modes to generate a reconstructed signal; and correcting the reconstructed signal based on the initial well logging data to obtain the target well logging data. In this application, eigenmode decomposition is performed on the initial logging data acquired by nuclear magnetic resonance (NMR). Since eigenmode decomposition can fully utilize the multi-exponential decay characteristics and non-stationary distribution characteristics of NMR echo signals, it can more thoroughly separate the effective attenuation components from background noise and reduce mode aliasing. Therefore, it can obtain candidate modes with good robustness to multi-component noise interference. Then, detrending fluctuation analysis is performed on the candidate modes to generate Hurst exponents, which can remove the trend components in the modal data and reveal the inherent fluctuation characteristics of the initial logging data sequence. The candidate modes are then screened according to the Hurst exponents to objectively select the target modes and reduce human error. Finally, the target modes are superimposed to generate a reconstructed signal, and the reconstructed signal is corrected according to the initial logging data. This can obtain logging target data that adaptively separates noise and effective signals, thereby improving the signal-to-noise ratio of NMR logging data.
[0106] To facilitate understanding of the noise reduction effect of the well logging data in this application, a numerical simulation experiment was conducted to verify the effectiveness of the application, taking one-dimensional nuclear magnetic resonance well logging data noise reduction as an example. The numerical simulation experiment first constructed a T2 spectrum with two peaks, with T2 values of 8ms and 180ms corresponding to the two peaks, respectively. Then, a certain amount of Gaussian white noise was added to the forward modeling results to simulate echo train data with different signal-to-noise ratios (SNR).
[0107] Figure 2 It is a T2 spectrum with two peaks constructed from a numerical simulation experiment. Figure 3 and Figure 4It is forward-modeled one-dimensional nuclear magnetic resonance echo train data. Figure 3 The mid-echo interval is 0.2ms, the number of echoes is 4096, and the SNR is 10. Figure 4 The mid-echo interval is 0.2ms, the number of echoes is 4096, and the SNR is 20. Figure 5 and Figure 6 This is the result of decomposing one-dimensional nuclear magnetic resonance echo train data according to the scheme in this application. Figure 5 Decomposition of the proposed scheme Figure 3 Results of mid-echo train data, Figure 6 Decomposition of the proposed scheme Figure 4 Results of the mid-echo train data. Figure 7 and Figure 8 This is the result of denoising one-dimensional nuclear magnetic resonance echo train data according to the scheme proposed in this application. Figure 7 For the purposes of this application Figure 3 The result of denoising the mid-echo train data is that the signal-to-noise ratio of the denoised data is 29. Figure 8 For the purposes of this application Figure 4 The result of denoising the mid-echo train data is that the signal-to-noise ratio of the denoised data is 55. Figure 7 and Figure 8 It can be seen that the signal-to-noise ratio of the echo data is significantly improved after denoising.
[0108] Will Figure 3 , Figure 4 , Figure 7 and Figure 8 The number of echo data points was compressed to 30, and the Levenberg-Marquardt (LM) method was used to invert the compressed echo data to obtain the NMR T2 spectra of data with different signal-to-noise ratios before and after denoising, such as... Figure 9 and Figure 10 As shown, Figure 9 yes Figure 3 and Figure 7 A comparison of the nuclear magnetic resonance T2 spectrum obtained by inverting the echo train data with the constructed T2 spectrum model; Figure 10 yes Figure 4 and Figure 8 The figure shows a comparison between the nuclear magnetic resonance T2 spectrum obtained from the echo train data inversion and the constructed T2 spectrum model. Figure 9 and Figure 10 It can be seen that the T2 spectrum retrieved after denoising is closer to the constructed model, indicating that the denoising method proposed in this application can effectively improve the signal-to-noise ratio of nuclear magnetic resonance logging data, thereby improving the retrieval accuracy.
[0109] Please see Figure 11 , Figure 11 This is a schematic diagram of a well logging data noise reduction system provided in an embodiment of this application.
[0110] This application provides a logging data noise reduction system, which may include:
[0111] The well logging initial data acquisition module 101 is used to acquire well logging initial data acquired by nuclear magnetic resonance.
[0112] The eigenmode decomposition module 102 is used to perform eigenmode decomposition on the initial logging data to obtain a set number of candidate modes;
[0113] The detrended fluctuation analysis module 103 is used to perform detrended fluctuation analysis on candidate modes to obtain fluctuation data.
[0114] The index generation module 104 is used to generate the Hurst index of candidate modes based on fluctuation data;
[0115] The filtering module 105 is used to filter candidate modes according to the Hurst index to obtain the target mode;
[0116] The superposition module 106 is used to superimpose the target modes to generate a reconstructed signal;
[0117] The correction module 107 is used to correct the reconstructed signal based on the initial logging data to obtain the logging target data.
[0118] This application provides a logging data noise reduction system. The Eigenmode Decomposition (EMD) module can be specifically used for: determining Eigenmode Decomposition parameters, including filter length, number of filter initializations, kurtosis shift order, and maximum iteration count; dividing the frequency band of the initial logging data into frequency band segments equal to the number of filters, and determining the upper and lower cutoff frequencies of each frequency band segment; generating FIR filters corresponding to the frequency band segments; filtering the initial logging data using the FIR filters to obtain the current round's unprocessed modes; updating the parameters of the FIR filters based on the current round's unprocessed modes to maximize the kurtosis value output by the FIR filters; updating the iteration count; responding to an iteration count less than the maximum iteration count, returning to the step of filtering the initial logging data using the FIR filters to obtain the current round's unprocessed modes; and responding to an iteration count equal to the maximum iteration count, filtering the unprocessed modes obtained in the last round to obtain a set number of candidate modes.
[0119] This application provides a well logging data noise reduction system. The characteristic mode decomposition module can be specifically used to: generate correlation coefficients between pairs of modes to be processed for the modes obtained in the last round, and generate a correlation matrix based on the correlation coefficients; determine the first mode and the second mode with the largest correlation coefficient values in the off-diagonal region of the correlation matrix; generate kurtosis values for the first mode and the second mode respectively; discard the mode with the smaller kurtosis value in the first mode and the second mode according to the kurtosis value; update the correlation matrix and decrement the total number of modes by 1; in response to the total number of modes being greater than a set number, return to the steps of determining the first mode and the second mode with the largest values in the off-diagonal region of the correlation matrix and the subsequent steps; in response to the total number of modes being equal to the set number, use the remaining modes as candidate modes.
[0120] This application provides a logging data noise reduction system. The detrending fluctuation analysis module can be specifically used for: constructing a cumulative deviation sequence of candidate modes; segmenting the cumulative deviation sequence to obtain sequence intervals that do not overlap and have a set length; performing least squares fitting on the sequence intervals to obtain local trends; detrending the local trends and generating variance values; generating root mean square fluctuations of the sequence intervals based on the variance values; and using the root mean square fluctuations of the sequence intervals as fluctuation data.
[0121] This application provides a well logging data noise reduction system, in which the index generation module can be specifically used to: perform logarithmic operation on the root mean square fluctuation to generate a first operation result; perform logarithmic operation on the sequence interval to generate a second operation result; perform linear regression on the first operation result and the second operation result to obtain a linear regression result; and generate the Hurst index of the candidate mode based on the slope of the straight line of the linear regression result.
[0122] This application provides a logging data noise reduction system, in which the screening module can be specifically used to: obtain a set modal index; if the Hurst exponent of a candidate mode is greater than the modal index, then the candidate mode is taken as the target mode; if the Hurst exponent of a candidate mode is less than or equal to the modal index, then the candidate mode is removed.
[0123] This application provides a logging data noise reduction system. The correction module can be specifically used to obtain a set data length value; determine a first dataset corresponding to the data length value in the initial logging data; determine a second dataset corresponding to the data length value in the reconstructed signal; generate a correction factor based on the ratio between the cumulative data value of the first dataset and the cumulative data value of the second dataset; and process the reconstructed signal according to the correction factor to obtain the logging target data.
[0124] This application also provides an electronic device and a computer-readable storage medium, both of which have the corresponding effects of the well logging data noise reduction method provided in the embodiments of this application. Please refer to... Figure 12 , Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.
[0125] An electronic device provided in this application includes a memory 201 and a processor 202. The memory 201 stores a computer program, and when the processor 202 executes the computer program, it implements the steps of the well logging data noise reduction method described in any of the above embodiments.
[0126] Please see Figure 13 Another electronic device provided in this application embodiment may further include: an input port 203 connected to the processor 202 for transmitting commands input from the outside to the processor 202; a display unit 204 connected to the processor 202 for displaying the processing results of the processor 202 to the outside; and a communication module 205 connected to the processor 202 for enabling communication between the electronic device and the outside. The display unit 204 may be a display panel, a laser scanning display, etc.; the communication method adopted by the communication module 205 includes, but is not limited to, Mobile High-Definition Link (MHL), Universal Serial Bus (USB), High-Definition Multimedia Interface (HDMI), wireless connection: Wireless Fidelity (WiFi), Bluetooth communication technology, Bluetooth Low Energy communication technology, and communication technology based on IEEE 802.11s.
[0127] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the well logging data noise reduction method described in any of the above embodiments.
[0128] The computer-readable storage media involved in this application include random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs (compact disc read-only memory), or any other form of storage media known in the art.
[0129] This application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the well logging data noise reduction method described in any of the above embodiments.
[0130] For descriptions of relevant parts of the well logging data noise reduction system, electronic device, and computer-readable storage medium provided in this application's embodiments, please refer to the detailed description of the corresponding parts in the well logging data noise reduction method provided in this application's embodiments, which will not be repeated here. Furthermore, parts of the technical solutions provided in this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.
[0131] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0132] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for noise reduction of well logging data, characterized in that, include: Acquire initial logging data obtained via nuclear magnetic resonance; The initial logging data is subjected to eigenmode decomposition to obtain a set number of candidate modes; Detrended fluctuation analysis is performed on the candidate modes to obtain fluctuation data; Based on the fluctuation data, the Hurst index of the candidate mode is generated; The candidate modes are filtered according to the Hurst index to obtain the target mode; The target modes are superimposed to generate a reconstructed signal; The reconstructed signal is corrected based on the initial logging data to obtain the target logging data.
2. The method according to claim 1, characterized in that, The step of performing eigenmode decomposition on the initial logging data to obtain a set number of candidate modes includes: The parameters for the eigenmode decomposition are determined, including filter length, number of filter initializations, kurtosis shift order, and maximum number of iterations. The frequency band of the initial logging data is divided into frequency band segments equal to the number of filters, and the upper and lower cutoff frequencies of each frequency band segment are determined. Generate FIR filters corresponding to the frequency band segments; The initial logging data is filtered using an FIR filter to obtain the current cycle's unprocessed mode; The parameters of the FIR filter are updated based on the mode to be processed in the current round to maximize the kurtosis value of the FIR filter output; Number of update iterations; If the number of iterations is less than the maximum number of iterations, then return to the step of filtering the initial logging data with an FIR filter to obtain the mode to be processed in the current round; In response to the number of iterations being equal to the maximum number of iterations, the unprocessed modes obtained in the last round are filtered to obtain a set number of candidate modes.
3. The method according to claim 2, characterized in that, The process of filtering the unprocessed modes obtained in the last round to obtain a set number of candidate modes includes: For the modes to be processed obtained in the last round, generate correlation coefficients between each pair of modes to be processed, and generate a correlation matrix based on the correlation coefficients; In the off-diagonal region of the correlation matrix, identify the first and second modes with the largest correlation coefficients; Generate the kurtosis values for the first and second modes respectively; Based on the kurtosis value, discard the mode with the smaller kurtosis value between the first and second modes; Update the correlation matrix and decrease the total number of modes by 1; If the total number of modes is greater than the set number, then return to the off-diagonal region of the correlation matrix to determine the first and second modes with the largest values, and then proceed with the steps thereafter. If the total number of modes equals the set number, the remaining modes are selected as candidate modes.
4. The method according to claim 1, characterized in that, The step of performing detrended fluctuation analysis on the candidate modes to obtain fluctuation data includes: Construct the cumulative deviation sequence of the candidate modes; The cumulative deviation sequence is segmented to obtain sequence intervals that do not overlap and have a set length; The local trend is obtained by fitting the sequence interval using the least squares method. The local trend is detrended and variance values are generated; Based on the variance value, the root mean square fluctuation of the sequence interval is generated; The root mean square fluctuation of the sequence interval is used as fluctuation data.
5. The method according to claim 4, characterized in that, The step of generating the Hurst index of the candidate mode based on the fluctuation data includes: Perform a logarithmic operation on the root mean square fluctuation to generate the first result. Perform a logarithmic operation on the sequence interval to generate a second result; Perform linear regression on the first calculation result and the second calculation result to obtain the linear regression result; The Hurst index of the candidate mode is generated based on the slope of the straight line from the linear regression result.
6. The method according to claim 1, characterized in that, The step of filtering the candidate modes according to the Hurst index to obtain the target mode includes: Obtain the set modal parameters; If the Hurst exponent of the candidate mode is greater than the mode index, then the candidate mode is adopted as the target mode. If the Hurst index of the candidate mode is less than or equal to the mode index, the candidate mode is eliminated.
7. The method according to claim 1, characterized in that, The step of correcting the reconstructed signal based on the initial logging data to obtain the target logging data includes: Get the set data length value; In the initial logging data, a first dataset corresponding to the data length value is determined; In the reconstructed signal, a second dataset corresponding to the data length value is determined; A correction factor is generated based on the ratio between the cumulative data values of the first dataset and the cumulative data values of the second dataset; The reconstructed signal is processed according to the correction factor to obtain the logging target data.
8. A well logging data noise reduction system, characterized in that, include: The well logging initial data acquisition module is used to acquire the well logging initial data obtained through nuclear magnetic resonance. The feature mode decomposition module is used to perform feature mode decomposition on the initial logging data to obtain a set number of candidate modes; The detrended fluctuation analysis module is used to perform detrended fluctuation analysis on the candidate modes to obtain fluctuation data; An index generation module is used to generate the Hurst index of the candidate mode based on the fluctuation data. The filtering module is used to filter the candidate modes according to the Hurst index to obtain the target mode; The superposition module is used to superimpose the target modes to generate a reconstructed signal; The correction module is used to correct the reconstructed signal based on the initial logging data to obtain the target logging data.
9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the logging data noise reduction method as described in any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the well logging data noise reduction method as described in any one of claims 1 to 7.