Fluid component division and saturation calculation method based on two-dimensional nuclear magnetic logging information

Abstract

The invention discloses a fluid component division and saturation calculation method based on two-dimensional nuclear magnetic logging data. The fluid component division and saturation calculation method comprises the steps that the two-dimensional nuclear magnetic logging data are preprocessed; applying Laplacian-Gaussian transform to enhance spectral peak features, and detecting local extreme points under each scale as candidate point locations; a threshold value is set according to amplitude intensity, and weak response candidate point locations are screened out; carrying out weighted combination on the similar candidate point locations, and finally determining the number of fluid components and the central position; performing two-dimensional asymmetric Gaussian function fitting on the central neighborhood of each spectral peak, and constructing an initialization matrix of non-negative matrix factorization by using fitting parameters; performing non-negative matrix factorization by taking the initialized matrix as a starting point, and decomposing the two-dimensional nuclear magnetic spectrum into independent component spectrums; calculating the component saturation of different fluid components; and drawing a two-dimensional nuclear magnetic spectrum component decomposition result. According to the method, Laplacian-Gaussian transformation and non-negative matrix factorization are cooperatively utilized, excessive manual intervention is not needed, and fluid component decomposition and saturation calculation are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of oil exploration, and particularly relates to a fluid component division and saturation calculation method based on two-dimensional nuclear magnetic logging data. BACKGROUND

[0002] The nuclear magnetic resonance logging technology can not only realize effective evaluation of complex pore structure, but also quantitatively characterize pore fluid information in different occurrence states such as solid organic matter, clay bound fluid and movable fluid, and plays an important role in the field of oil and gas reservoir exploration and development. However, in the case of dense reservoir pore structure or coexistence of oil, gas, water and other fluids in the pore, the one-dimensional nuclear magnetic distribution spectrum has the problem of overlapping of component signal positions, and it is difficult to accurately distinguish different fluid component signals, and the application has obvious limitations. In contrast, two-dimensional nuclear magnetic distribution spectra such as T2-T1, T2-D, T2-G and T2-Pc can provide more abundant multi-dimensional information, and show greater advantages in complex fluid property identification and pore structure characterization.

[0003] In order to accurately determine the fluid properties of different component signals in the two-dimensional nuclear magnetic distribution spectrum, there are currently two methods: on the one hand, through indoor two-dimensional nuclear magnetic core experiment measurement, saturated oil, saturated water, centrifugation, drying and other methods are used to determine the distribution position and shape of component signals such as movable oil, movable water, capillary bound oil, capillary bound water, clay bound oil, clay bound water and solid organic matter in the two-dimensional nuclear magnetic spectrum, and a two-dimensional nuclear magnetic logging fluid identification chart is established accordingly. On the other hand, mathematical methods such as blind source separation can be used to process the two-dimensional nuclear magnetic spectrum, separate the overall signal into several component components, and thus determine the pore fluid properties and further calculate the proportion.

[0004] However, there are great limitations in applying the chart method and blind source separation method to fluid identification of two-dimensional nuclear magnetic spectrum: first of all, the establishment of two-dimensional nuclear magnetic fluid identification chart relies on a large number of core experiments, and the charts of different oil fields have great differences, and the universality is poor, and the chart method can only be used for qualitative judgment, and it is difficult to realize quantitative evaluation of the saturation of different fluid components. At the same time, the number of fluid components needs to be set artificially in advance before the two-dimensional nuclear magnetic spectrum is processed by blind source separation, and once the preset parameter does not match the actual number of reservoir pore fluid components, the separation result deviates from the actual situation, and it is difficult to accurately reflect the true properties and distribution state of the pore fluid.

[0005] Therefore, it is necessary to propose a new two-dimensional nuclear magnetic spectrum fluid component division and saturation calculation method, which can realize the autonomous detection of the number of fluid components and the center position of the component spectrum in the two-dimensional nuclear magnetic spectrum without providing too much artificial intervention, and then complete the fluid component decomposition and saturation calculation. SUMMARY

[0006] This invention aims to address the problem that existing technologies require excessive manual parameter setting during fluid component decomposition of two-dimensional nuclear magnetic resonance (NMR) spectra, leading to significant discrepancies between the decomposition results and the actual fluid distribution characteristics in reservoir pores. The core idea is to synergistically utilize multi-scale two-dimensional Laplace-Gaussian transform and non-negative matrix factorization (NMF): First, multi-scale two-dimensional Laplace-Gaussian transform is applied to enhance spectral peak characteristics, and combined with strategies such as amplitude threshold filtering and weighted merging of similar points, autonomous searching for the number of components and the center positions of spectral peaks in the two-dimensional NMR spectrum is achieved. Then, based on the search results, initialization parameters for NMF are constructed to improve the stability and accuracy of NMF, ultimately achieving effective decomposition of different fluid components in the two-dimensional NMR spectrum and accurate calculation of their saturation.

[0007] This invention is achieved through the following technical solution: the method for fluid composition classification and saturation calculation based on two-dimensional nuclear magnetic resonance logging data includes the following steps: S1. Preprocess the two-dimensional NMR spectroscopy data; S2. Apply the Laplace-Gaussian transform to perform convolution operation on the preprocessed two-dimensional NMR spectrum data from step S1, and detect the local extrema of the convolution response at each scale as candidate points. S3. Set a threshold based on the amplitude intensity of the candidate points in step S2, filter out weak response candidate points, and retain significant candidate peaks. S4. Weighted merging of similar candidate points after screening in step S3 to finally determine the number of fluid components and the center position of each spectral peak. S5. Perform two-dimensional asymmetric Gaussian function fitting on the neighborhood of each spectral peak center determined in step S4, and use the fitting parameters to construct the initialization matrix for non-negative matrix decomposition. S6. Starting from the initialization matrix in step S5, perform non-negative matrix decomposition to decompose the two-dimensional NMR spectrum into independent component spectra. S7. Calculate the saturation of different fluid components based on the component spectrum data obtained in step S6. S8. Based on the component spectrum data obtained in step S6, plot the two-dimensional NMR spectrum component decomposition results.

[0008] Optionally, step S1 includes: Interpolate the original two-dimensional NMR spectrum data to unify the number of points in both T1 and T2 dimensions to 256; then apply Gaussian filtering to the interpolated two-dimensional NMR spectrum data.

[0009] Optionally, step S2 includes: First, the Laplace-Gaussian operator is applied to perform a convolution operation with the preprocessed two-dimensional NMR spectrum data; The formula for convolving the Laplace-Gaussian operator with the preprocessed two-dimensional NMR spectral data is as follows: In the formula: Representing scale The convolution response results are as follows; This represents the two-dimensional NMR spectrum distribution function after preprocessing. For the scale The two-dimensional Laplace-Gaussian transform kernel function; The x and y coordinates of each data point in the preprocessed two-dimensional NMR spectrum; Subsequently, local extrema of the convolution response at all scales are searched, and local extrema that show significant responses at multiple scales are selected as candidate points. The local maxima condition for searching for extreme points of convolution response at different scales is: In the formula: For the scale The corresponding maximum value of the convolution is below. The coordinates of the neighborhood of the extreme point; coordinate position The convolution response value under the given conditions; coordinate position The convolution response value under the given conditions; The coordinates of the local extrema of the convolution response; It represents the coordinates of a neighborhood of a local extremum point in the convolution response.

[0010] The selection criteria for candidate sites are local extrema that show significant responses at multiple scales: In the formula: Here, m represents the minimum number of times the convolution response extrema occur at different scales, and n represents the total number of times the convolution response extrema occur at different scales.

[0011] Optionally, step S3 includes: Based on the amplitude intensity of the candidate points, a threshold coefficient is set to filter out candidate points with excessively low amplitude, thereby eliminating false peaks caused by noise and retaining significant candidate peaks. The amplitude threshold filtering condition is as follows: In the formula: The amplitude intensity of the candidate site in the two-dimensional NMR spectrum; Given a relative threshold coefficient; The global maximum amplitude value of two-dimensional NMR spectroscopy data.

[0012] Optionally, step S4 includes: Candidate points falling within a specified merging radius are weighted and merged to calculate the center position and amplitude of their representative peaks, ultimately determining the precise number of fluid components in the two-dimensional NMR spectrum and the accurate center position of each peak. The formula for weighted merging and optimization of adjacent candidate points within a certain radius is as follows: In the formula: This represents the distance from each candidate point to the point of maximum amplitude; R is the specified merging radius. , , These are the horizontal and vertical coordinates and their amplitudes for each candidate point; The total number of candidate points. For the first The magnitude of each candidate point; This represents the weight assigned to each candidate point. , These are the x and y coordinates of the point with the maximum amplitude, respectively. , , This represents the weighted centroid coordinates and amplitudes of all candidate points within the merged radius.

[0013] Optionally, step S5 includes: First, local spectral regions are extracted using the center of each spectral peak determined in step S4 as the core. Second, a two-dimensional asymmetric Gaussian function model is used to perform nonlinear least squares fitting on the local region to accurately quantify the characteristic parameters of the peak shape. Finally, the fitting parameters are used to construct the initialization matrix for nonnegative matrix decomposition. The fitting formula for the two-dimensional asymmetric Gaussian function is as follows: In the formula: A represents the local two-dimensional NMR spectrum data near the center of the spectral peak; A is the amplitude of the local two-dimensional NMR spectrum. , These represent the horizontal and vertical coordinates of the center of the local two-dimensional NMR spectrum, respectively. , This represents the left and right widths of the local two-dimensional NMR spectrum along the horizontal axis. , This represents the upper and lower widths of a local two-dimensional NMR peak along the vertical axis. The initial matrix for the nonnegative matrix decomposition constructed using the fitted parameters is: In the formula: and These are the initial basis matrix and weight matrix for nonnegative matrix decomposition, respectively; , , They represent the first The horizontal and vertical coordinates and amplitudes of the center of a local two-dimensional NMR spectrum; and They represent the first The equivalent width of a local two-dimensional NMR spectrum in the horizontal and vertical axes; and They represent the first The left and right widths of a local two-dimensional NMR spectrum in the horizontal and vertical axes; and They represent the first The upper and lower widths of a local two-dimensional NMR spectrum along the vertical axis.

[0014] Optionally, step S6 includes: Starting with the initialization matrix, a non-negative matrix factorization algorithm is executed to decompose the two-dimensional NMR spectrum into the product of two non-negative matrices, thereby decomposing the independent component spectra in the two-dimensional NMR spectrum. The nonnegative matrix factorization formula and objective function are expressed as follows: In the formula: S represents the preprocessed two-dimensional NMR data matrix; W is the basis matrix, representing the distribution of different fluid components in the vertical axis; H is the weight matrix, representing the distribution of different fluid components in the horizontal axis.

[0015] Optionally, step S7 includes: First, numerical integration is performed on the component spectra of each fluid component obtained from the decomposition to obtain the relative volume of each component; then, the volume of each component is normalized to calculate the saturation of each fluid component. The formulas for calculating the saturation of different fluid components are as follows: In the formula: The first in the two-dimensional NMR spectrum The saturation of a fluid component, The first in the two-dimensional NMR spectrum Distribution functions of various fluid components; This represents the preprocessed two-dimensional NMR spectrum distribution function.

[0016] Optionally, step S8 includes: Two-dimensional NMR spectra were plotted, indicating the center positions of the automatically detected component peaks and the contour boundaries of each fluid component spectrum.

[0017] In summary, the technical effects and advantages of this invention are as follows: This invention provides a method for reservoir fluid component classification and saturation calculation based on two-dimensional nuclear magnetic resonance (NMR) logging data. It synergistically utilizes two methods: multi-scale two-dimensional Laplace-Gaussian transform and non-negative matrix factorization (NMR). First, the multi-scale two-dimensional Laplace-Gaussian transform is applied to enhance spectral peak information. Combined with strategies such as amplitude threshold filtering and weighted merging of similar points, it achieves autonomous detection of the number of components and the center position of spectral peaks in the two-dimensional NMR spectrum, avoiding interference from manually preset component numbers. Then, based on the autonomous detection results, initialization parameters for NMR are constructed, thereby improving the stability and accuracy of NMR. Ultimately, it achieves effective decomposition of different fluid components in the two-dimensional NMR spectrum and accurate calculation of their saturation. This method plays an important role in two-dimensional NMR logging data processing and has strong versatility. Attached Figure Description

[0018] 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a two-dimensional nuclear magnetic resonance distribution spectrum artificially constructed in a simulated data application example of one embodiment of the present invention; Figure 3 This is an example of the simulation data application of the present invention, in which the method is used to process two-dimensional nuclear magnetic resonance structural spectra to obtain the center position and outline of the fluid component spectrum; Figure 4 This is a measured two-dimensional nuclear magnetic resonance spectrum of a water-saturated core sample in an example of experimental data application in one embodiment of the present invention; Figure 5 This is an example of the application of experimental data in one embodiment of the present invention, in which the method was used to process the measured two-dimensional nuclear magnetic resonance spectrum of a water-saturated core sample to obtain the center position and outline of the fluid component spectrum. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Simulated data application examples like Figure 1 As shown, the method for reservoir fluid composition classification and saturation calculation based on two-dimensional nuclear magnetic resonance logging data includes the following steps: S1. Preprocessing of 2D NMR logging data; The original two-dimensional NMR spectrum data is interpolated to unify the number of points in both T1 and T2 dimensions to 256; where T1 represents the longitudinal relaxation time and T2 represents the transverse relaxation time; Gaussian filtering is applied to the interpolated two-dimensional NMR spectrum to suppress noise interference on the true two-dimensional spectrum distribution. S2. Apply the Laplace-Gaussian transform to perform convolution operation on the preprocessed two-dimensional NMR spectrum data from step S1 to enhance the spectral peak characteristics, and detect local extrema of the convolution response at each scale as candidate points. First, a multi-scale two-dimensional Laplacian-Gaussian operator is applied to convolve the preprocessed two-dimensional NMR spectral data to enhance the spectral peak features at different sizes. Then, local extrema of the convolution response at all scales are searched, and local extrema showing significant responses at multiple scales are selected as candidate locations.

[0022] The formula for convolving the Laplace-Gaussian operator with the preprocessed two-dimensional NMR spectral data is as follows: In the formula: Representing scale The convolution response under; This represents the two-dimensional NMR spectrum distribution function after preprocessing. For the scale The two-dimensional Laplace-Gaussian transform (LoG) kernel function; The x and y coordinates of each data point in the preprocessed two-dimensional NMR spectrum are shown.

[0023] The local maxima condition for searching for extreme points of convolution response at different scales is: In the formula: For the scale The corresponding maximum value of the convolution is below. The coordinates of the neighborhood of the extreme point; coordinate position The convolution response value under the given conditions; coordinate position The convolution response value under the given conditions; The coordinates of the local extrema of the convolution response; It represents the coordinates of a neighborhood of a local extremum point in the convolution response.

[0024] The selection criteria for candidate sites are local extrema that show significant responses at multiple scales: In the formula: Here, m represents the minimum number of times the convolution response extrema occur at different scales, and n represents the total number of times the convolution response extrema occur at different scales.

[0025] S3. Set a threshold based on the amplitude intensity of the candidate points in step S2, filter out weak response candidate points, and retain significant peaks; Based on the amplitude intensity of the candidate points, a threshold coefficient is set to filter out candidate points with excessively low amplitudes, thereby eliminating false peaks caused by noise and retaining significant candidate peaks.

[0026] The amplitude threshold filtering condition is as follows: In the formula: The amplitude intensity of the candidate site in the two-dimensional NMR spectrum; Given a relative threshold coefficient; The global maximum amplitude value of two-dimensional NMR spectroscopy data.

[0027] S4. Weighted merging of similar candidate points after screening in step S3 to finally determine the number of fluid components and the center position of each spectral peak; To avoid duplicate counting of the same broad peak or overlapping peak, candidate points falling within a specified merging radius are weighted and merged to calculate the center position and amplitude of their representative peaks, ultimately determining the precise number and center position of the fluid components in the two-dimensional NMR spectrum.

[0028] The formula for weighted merging and optimization of adjacent candidate points within a certain radius is as follows: In the formula: This represents the distance from each candidate point to the point of maximum amplitude; R is the specified merging radius. , , These are the horizontal and vertical coordinates and their amplitudes for each candidate point; The total number of candidate points. For the first The magnitude of each candidate point; This represents the weight of each candidate point; the higher the magnitude, the greater the weight. , These are the x and y coordinates of the point with the maximum amplitude, respectively. , , This represents the weighted centroid coordinates and amplitudes of all candidate points within the merged radius.

[0029] S5. Perform two-dimensional asymmetric Gaussian function fitting on the neighborhood of each spectral peak center determined in step S4, and use the fitting parameters to construct the initialization matrix for non-negative matrix decomposition. First, using the center of each spectral peak determined in step S4 as the core, a local spectral region is extracted. Second, a two-dimensional asymmetric Gaussian function model is used to perform nonlinear least-squares fitting on this local region to accurately quantify the characteristic parameters of the peak shape. Finally, the fitting parameters are used to construct an initialization matrix for nonnegative matrix factorization.

[0030] The fitting formula for the two-dimensional asymmetric Gaussian function is as follows: In the formula: The data represents the local two-dimensional NMR distribution near the center of the spectral peak; A represents the amplitude of the local two-dimensional NMR spectral peak. , These represent the horizontal and vertical coordinates of the local two-dimensional NMR peaks, respectively. , This represents the left and right widths of a local two-dimensional NMR peak along the horizontal axis. , This represents the upper and lower widths of a local two-dimensional NMR peak along the vertical axis.

[0031] The initial matrix constructed using the fitting parameters and the nonnegative matrix decomposition is: In the formula: and These are the initial basis matrix and weight matrix for nonnegative matrix decomposition, respectively; , , They represent the first The horizontal and vertical coordinates and amplitudes of the center of a local two-dimensional NMR spectrum; and They represent the first The equivalent width of a local two-dimensional NMR spectrum in the horizontal and vertical axes; and They represent the first The left and right widths of a local two-dimensional NMR spectrum in the horizontal direction; and They represent the first The upper and lower widths of a local two-dimensional NMR spectrum along the vertical axis.

[0032] S6. Starting from the initialization matrix in step S5, perform nonnegative matrix decomposition to decompose the two-dimensional NMR spectrum into independent component spectra; Starting with the initialization matrix, a non-negative matrix factorization algorithm is executed to strictly decompose the two-dimensional NMR spectrum into the product of two non-negative matrices, thereby decomposing each independent component spectrum in the two-dimensional NMR spectrum.

[0033] The nonnegative matrix factorization formula and objective function are expressed as follows: In the formula: S represents the preprocessed two-dimensional NMR data matrix; W is the basis matrix, representing the distribution of different fluid components in the vertical axis; H is the weight matrix, representing the distribution of different fluid components in the horizontal axis.

[0034] S7. Calculate the saturation of different fluid components based on the component spectrum data obtained in step S6; The relative volume of each fluid component is obtained by numerical integration of its component spectrum. Then, the volume of each component is normalized, and the saturation of each fluid component is calculated. The formulas for calculating the saturation of different fluid components are as follows: In the formula: The first in the two-dimensional NMR spectrum The saturation of a fluid component, The first in the two-dimensional NMR spectrum Distribution functions of various fluid components; This represents the preprocessed two-dimensional NMR spectrum distribution function.

[0035] S8. Based on the component spectrum data obtained in step S6, plot the two-dimensional NMR spectrum component decomposition results; Two-dimensional NMR spectra were plotted, indicating the center positions of the automatically detected component peaks and the contour boundaries of each fluid component spectrum.

[0036] Figure 2 The image shows a two-dimensional NMR spectrum artificially constructed in a simulated data application example. It contains four fluid components, with the spectral center positions of each component located at T2=0.05ms T1=0.50ms, T2=0.80ms T1=10.00ms, T2=10.00ms T1=50.00ms, T2=200ms T1=350ms, respectively; the peak amplitudes are 1.5, 0.9, 1.2, and 0.8, respectively. Based on this invention, the fluid component decomposition results obtained by processing the two-dimensional NMR constructed spectrum are as follows: Figure 3As shown in the figure, the center positions and contour boundaries of the four fluid component spectra are clearly marked. Table 1 records the parameters of the two-dimensional NMR constructed spectrum and the component spectra, including the center positions (T2 and T1 values) and peak amplitudes of the fluid component spectra. The comparison shows that the component spectra are basically consistent with the parameters of the original two-dimensional NMR spectrum constructed, verifying that the two-dimensional NMR spectrum decomposition method described in this specification has high processing accuracy and reliable processing results.

[0037] Table 1 Examples of Experimental Data Application The difference between this experimental data application example and the aforementioned simulation data application example is that the data processed is actual two-dimensional NMR experimental data from water-saturated core samples. Apart from this, the processing steps in this experimental data application example are the same as those in the simulation data application example, and therefore will not be repeated here. Please refer to the relevant descriptions in the simulation data application example for details.

[0038] Figure 4 The figure shows the measured two-dimensional NMR spectrum of an oil-saturated core sample in an example of experimental data application. Through two-dimensional NMR measurements under different experimental conditions (saturated crude oil, centrifugation, drying, etc.), the saturation information of different fluid components such as clay, residual oil, and mobile oil were obtained as 36.3%, 42.3%, and 21.4%, respectively. Based on this invention, the fluid component decomposition results obtained by processing the measured two-dimensional NMR spectrum of a water-saturated core sample are as follows: Figure 5 As shown in the figure, the center positions and outline boundaries of the three fluid component spectra are clearly marked. Table 2 records the fluid component saturation information obtained from core NMR experimental analysis and the method of this invention. The comparison shows that the saturation processing results are basically consistent, with errors all below 0.4%, verifying that the two-dimensional NMR spectroscopy data processing method described in this specification can calculate accurate and reliable fluid component saturation information.

[0039] Table 2 This invention provides a method for reservoir fluid component classification and saturation calculation based on two-dimensional nuclear magnetic resonance (NMR) logging data. It synergistically utilizes two methods: multi-scale two-dimensional Laplace-Gaussian transform and non-negative matrix factorization (NMF). First, the multi-scale two-dimensional Laplace-Gaussian transform is applied to enhance spectral peak characteristics. Combined with strategies such as amplitude threshold filtering and weighted merging of similar points, it achieves autonomous searching of the number of components and the location of spectral peak centers in the two-dimensional NMR spectrum. Then, based on this result, initialization parameters for NMF are constructed, thereby improving the stability and accuracy of NMF. Ultimately, it achieves effective decomposition of different fluid components in the two-dimensional NMR spectrum and accurate calculation of their saturation. This method plays an important role in two-dimensional NMR logging data processing and has strong versatility.

[0040] Finally, it should be noted that the preferred embodiments of the present invention have been described above, but the present invention is not limited to the specific embodiments described above. The devices and structures not described in detail should be understood to be implemented in the ordinary way in the art. Any simple modifications, equivalent changes and modifications made by any person skilled in the art to the above embodiments based on the technical essence of the present invention without departing from the scope of the technical solution of the present invention shall still fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for fluid composition classification and saturation calculation based on two-dimensional nuclear magnetic resonance logging data, characterized in that, Includes the following steps: S1. Preprocess the two-dimensional NMR spectroscopy data; S2. Apply the Laplace-Gaussian transform to perform convolution operation on the preprocessed two-dimensional NMR spectrum data from step S1, and detect the local extrema of the convolution response at each scale as candidate points. S3. Set a threshold based on the amplitude intensity of the candidate points in step S2, filter out weak response candidate points, and retain significant candidate peaks. S4. Weighted merging of similar candidate points after screening in step S3 to finally determine the number of fluid components and the center position of each spectral peak. S5. Perform two-dimensional asymmetric Gaussian function fitting on the neighborhood of each spectral peak center determined in step S4, and use the fitting parameters to construct the initialization matrix for non-negative matrix decomposition. S6. Starting from the initialization matrix in step S5, perform non-negative matrix decomposition to decompose the two-dimensional NMR spectrum into independent component spectra. S7. Calculate the saturation of different fluid components based on the component spectrum data obtained in step S6. S8. Based on the component spectrum data obtained in step S6, plot the two-dimensional NMR spectrum component decomposition results.

2. The method for fluid composition classification and saturation calculation based on two-dimensional nuclear magnetic resonance logging data according to claim 1, characterized in that, Step S1 includes: Interpolate the original two-dimensional NMR spectrum data to unify the number of points in both T1 and T2 dimensions to 256; then apply Gaussian filtering to the interpolated two-dimensional NMR spectrum data.

3. The method for fluid composition classification and saturation calculation based on two-dimensional nuclear magnetic resonance logging data according to claim 1, characterized in that, Step S2 includes: First, the Laplace-Gaussian operator is applied to perform a convolution operation with the preprocessed two-dimensional NMR spectrum data; The formula for convolving the Laplace-Gaussian operator with the preprocessed two-dimensional NMR spectral data is as follows: In the formula: Representing scale The convolution response results are as follows; This represents the two-dimensional NMR spectrum distribution function after preprocessing. For the scale The two-dimensional Laplace-Gaussian transform kernel function; The x and y coordinates of each data point in the preprocessed two-dimensional NMR spectrum; Subsequently, local extrema of the convolution response at all scales are searched, and local extrema that show significant responses at multiple scales are selected as candidate points. The local maxima condition for searching for extreme points of convolution response at different scales is: In the formula: For the scale The corresponding maximum value of the convolution is below. The coordinates of the neighborhood of the extreme point; coordinate position The convolution response value under the given conditions; coordinate position The convolution response value under the given conditions; The coordinates of the local extrema of the convolution response; The coordinates of a neighborhood of a local extremum point in the convolution response; The selection criteria for candidate sites are local extrema that show significant responses at multiple scales: In the formula: Here, m represents the minimum number of times the convolution response extrema occur at different scales, and n represents the total number of times the convolution response extrema occur at different scales.

4. The method for fluid composition classification and saturation calculation based on two-dimensional nuclear magnetic resonance logging data according to claim 3, characterized in that, Step S3 includes: Based on the amplitude intensity of the candidate points, a threshold coefficient is set to filter out candidate points with excessively low amplitude, thereby eliminating false peaks caused by noise and retaining significant candidate peaks. The amplitude threshold filtering condition is as follows: In the formula: The amplitude intensity of the candidate site in the two-dimensional NMR spectrum; Given a relative threshold coefficient; The global maximum amplitude value of two-dimensional NMR spectroscopy data.

5. The method for fluid composition classification and saturation calculation based on two-dimensional nuclear magnetic resonance logging data according to claim 4, characterized in that, Step S4 includes: Candidate points falling within a specified merging radius are weighted and merged to calculate the center position and amplitude of their representative peaks, ultimately determining the precise number of fluid components in the two-dimensional NMR spectrum and the accurate center position of each peak. The formula for weighted merging and optimization of adjacent candidate points within a certain radius is as follows: In the formula: This represents the distance from each candidate point to the point of maximum amplitude; R is the specified merging radius. , , These are the horizontal and vertical coordinates and their amplitudes for each candidate point; The total number of candidate points. For the first The magnitude of each candidate point; This represents the weight assigned to each candidate point. , These are the x and y coordinates of the point with the maximum amplitude, respectively. , , This represents the weighted centroid coordinates and amplitudes of all candidate points within the merged radius.

6. The method for fluid composition classification and saturation calculation based on two-dimensional nuclear magnetic resonance logging data according to claim 5, characterized in that, Step S5 includes: First, local spectral regions are extracted using the center of each spectral peak determined in step S4 as the core. Second, a two-dimensional asymmetric Gaussian function model is used to perform nonlinear least squares fitting on the local region to accurately quantify the characteristic parameters of the peak shape. Finally, the fitting parameters are used to construct the initialization matrix for nonnegative matrix decomposition. The fitting formula for the two-dimensional asymmetric Gaussian function is as follows: In the formula: A represents the local two-dimensional NMR spectrum data near the center of the spectral peak; A is the amplitude of the local two-dimensional NMR spectrum. , These represent the horizontal and vertical coordinates of the center of the local two-dimensional NMR spectrum, respectively. , This represents the left and right widths of the local two-dimensional NMR spectrum along the horizontal axis. , This represents the upper and lower widths of a local two-dimensional NMR peak along the vertical axis. The initial matrix for the nonnegative matrix decomposition constructed using the fitted parameters is: In the formula: and These are the initial basis matrix and weight matrix for nonnegative matrix decomposition, respectively; , , They represent the first The horizontal and vertical coordinates and amplitudes of the center of a local two-dimensional NMR spectrum; and They represent the first The equivalent width of a local two-dimensional NMR spectrum in the horizontal and vertical axes; and They represent the first The left and right widths of a local two-dimensional NMR spectrum in the horizontal and vertical axes; and They represent the first The upper and lower widths of a local two-dimensional NMR spectrum along the vertical axis.

7. The method for fluid composition classification and saturation calculation based on two-dimensional nuclear magnetic resonance logging data according to claim 6, characterized in that, Step S6 includes: Starting with the initialization matrix, a non-negative matrix factorization algorithm is executed to decompose the two-dimensional NMR spectrum into the product of two non-negative matrices, thereby decomposing the independent component spectra in the two-dimensional NMR spectrum. The nonnegative matrix factorization formula and objective function are expressed as follows: In the formula: S represents the preprocessed two-dimensional NMR data matrix; W is the basis matrix, representing the distribution of different fluid components in the vertical axis; H is the weight matrix, representing the distribution of different fluid components in the horizontal axis.

8. The method for fluid composition classification and saturation calculation based on two-dimensional nuclear magnetic resonance logging data according to claim 7, characterized in that, Step S7 includes: First, numerical integration is performed on the component spectra of each fluid component obtained from the decomposition to obtain the relative volume of each component; then, the volume of each component is normalized to calculate the saturation of each fluid component. The formulas for calculating the saturation of different fluid components are as follows: In the formula: The first in the two-dimensional NMR spectrum The saturation of a fluid component, The first in the two-dimensional NMR spectrum Distribution functions of various fluid components; This represents the preprocessed two-dimensional NMR spectrum distribution function.

9. The method for fluid composition classification and saturation calculation based on two-dimensional nuclear magnetic resonance logging data according to claim 1, characterized in that, Step S8 includes: Plot the original two-dimensional NMR spectrum, indicating the center positions of the automatically detected component peaks and the contour boundaries of each fluid component spectrum.

Images