A multi-fluid component quantitative evaluation method and system based on NMR T1-T2 acquisition echo data

Abstract

The application provides a multi-fluid component quantitative evaluation method and system based on collected echo data, belongs to the technical field of reservoir evaluation in oil exploration and development, and comprises the following steps: forming an original data set; obtaining compressed echo data; obtaining a spectrum of a formation; extracting signal peak values and counting to form a characteristic data set; obtaining relaxation characteristic parameters of multiple fluids; compressing to obtain; using and, performing saturation inversion under the non-negative constraint and the constraint condition that the sum is 1, and combining a weighted function of exponential decay to improve inversion accuracy. The application avoids information loss caused by inversion spectrum, and realizes quantitative and high-precision evaluation of multiple fluids.

Description

Technical Field

[0001] This invention belongs to the field of reservoir evaluation technology in petroleum exploration and development, and particularly relates to a method for determining multiple fluids in a formation. Methods and systems for characteristics and their saturation. Background Technology

[0002] Unconventional reservoirs, especially shale reservoirs, contain abundant oil and gas resources and have become important targets for oil and gas exploration and development. These reservoirs often contain multiple fluid components, and the distribution characteristics of different fluids and their corresponding saturation levels are key parameters for evaluating reservoir potential and formulating reasonable development plans. Accurately distinguishing and quantifying different fluids is not only related to the evaluation of reservoir effectiveness but also directly affects production capacity prediction and development deployment.

[0003] In the existing technology, based on The measured fluid identification and saturation calculation methods mainly rely on inversion. Spectrum. Whether establishing template intervals through experiments, applying blind source separation algorithms, or utilizing artificial intelligence methods, the essence is still the inversion obtained. The spectrum was processed. However, the inversion yielded... The spectrum is not the original collected data, but the result generated during the inversion process. Its data quality is affected by the constraints and noise during the inversion process, and often some key information is lost, resulting in incomplete fluid feature identification and large errors in saturation calculation.

[0004] Therefore, there is an urgent need to propose a method that can directly utilize the original... A new method for collecting and processing data is proposed to avoid information loss during the inversion process, thereby more accurately obtaining the characteristics and saturation of different fluids. Summary of the Invention

[0005] The purpose of this invention is to provide a method based on A method and system for quantitative evaluation of multi-fluid components based on acquired echo data directly processes the raw echo signal, avoiding the reliance on inversion in traditional methods. This method avoids information loss due to spectral distortion and more accurately preserves fluid relaxation characteristics. Through this method, the quantitative identification and saturation evaluation of different fluid components become more accurate and reliable, enabling continuous, quantitative, and high-precision evaluation of multiple fluid components, thus providing strong support for the refined characterization and efficient development of unconventional reservoirs.

[0006] This invention utilizes directly By combining the raw echo signal with data compression, data inversion, nonnegative matrix factorization, clustering, forward modeling, and constrained inversion, the characteristic parameters of multi-fluid components can be accurately extracted and their saturation can be quantitatively calculated, enabling continuous, quantitative, and high-precision evaluation of unconventional reservoir multi-fluid systems.

[0007] The present invention adopts the following technical solution: A type based on Multi-fluid component quantitative evaluation methods for acquired echo data include: Step 1. Acquire nuclear magnetic resonance echo data collected in the laboratory, well site, or downhole to form the raw dataset. ; Step 2. For the original dataset described in Step 1 Compression processing is performed to obtain compressed echo data. ; Step 3. Based on the compressed echo data described in Step 2 Inversion is performed to obtain the target stratigraphic sample. Spectral distribution ; Step 4. Regarding the steps described in step 3... Spectral distribution Perform multiple nonnegative matrix factorizations to extract signal peaks and statistically form a feature dataset. ,in, It is the longitudinal relaxation time. It is the lateral relaxation time; Step 5. Use a clustering algorithm to analyze the feature dataset described in Step 4. The process is performed to obtain the relaxation characteristic parameters of various fluids; Step 6. Construct forward echo data for different fluids based on the clustering results in Step 5. and the forward-modeled echo data Compression is performed to obtain compressed echo data. ; Step 7. Utilize the compressed echo data from Step 2 And the compressed echo data from step 6 Saturation inversion is performed under nonnegativity and summation-to-1 constraints, combined with an exponentially decaying weighted function. Improve the stability and accuracy of inversion; Step 8. Output the saturation information of different fluids in the target formation based on the inversion results of Step 7.

[0008] Furthermore, the compression method in step 2 includes: compressing the original dataset... Compression methods include window averaging, discrete cosine transform, singular value decomposition, or combinations thereof, to reduce data dimensionality and preserve key signal features.

[0009] Furthermore, the nonnegative matrix decomposition in step 4 includes: […]. Spectral distribution Multiple iterative decompositions are performed to extract the signal peaks of each fluid component, and statistical analysis is conducted to form a feature dataset for cluster analysis. .

[0010] Furthermore, the clustering algorithm described in step 5 is... Clustering, hierarchical clustering, Gaussian mixture model ,Vague Any of the mean clustering methods can be used to distinguish the relaxation characteristic parameters of different fluids.

[0011] Furthermore, the saturation inversion in step 7 includes: solving the linear system under non-negativity constraints, and combining the exponentially decaying weighted function. This is to reduce the impact of noise on the inversion results and improve the accuracy and stability of the inversion.

[0012] Furthermore, it also includes: utilizing the saturation information described in step 8 and The spectrum is used to determine the fluid porosity at different depths of the target stratum.

[0013] The present invention also provides a method based on A multi-fluid component quantitative evaluation system for acquiring echo data includes a memory and a processor, and a computer program stored in the memory, wherein the processor executes the computer program to achieve the above-mentioned... Steps for quantitative evaluation of multifluid components by acquiring echo data.

[0014] The beneficial effects of this invention are: This invention breaks through the tradition Spectral inversion suffers from limitations such as sensitivity to noise and non-uniqueness of non-negative matrix factorization results. This invention fully utilizes the complete relaxation information contained in the original echo signal. By combining multiple non-negative matrix factorizations with cluster analysis, the stability and accuracy of fluid feature identification are improved. Window compression and multi-constraint inversion are employed to achieve direct and robust determination of the saturation of multiple fluid components. Compared with traditional methods relying on inversion spectra, this invention maintains excellent robustness and accuracy under different signal-to-noise ratios and compression levels, significantly improving the ability to identify and quantitatively evaluate multiple fluids in shale reservoirs. It provides an efficient and reliable new approach for analyzing fluid distribution characteristics and assessing reservoir parameters in complex reservoirs. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the steps of the present invention.

[0016] Figure 2 For the present invention Acquire echo data , The signal amplitude of the echo data is represented by the vertical axis. This represents the signal amplitude of the echo data, with the horizontal axis ranging from 1 to 5 × 10⁻⁶. 4 , representing echo data A total of 5×10 samples were collected. 4 One echo.

[0017] Figure 3 This invention provides a method for acquiring echo data. Compressed echo data , The signal amplitude of the echo data is represented by the vertical axis. This represents the signal amplitude of the echo data, with the horizontal axis ranging from 1 to 500, indicating the compressed echo data. It was compressed to 500 echoes.

[0018] Figure 4 For the present invention to After inversion Spectrum.

[0019] Figure 5 The five signals obtained by a single nonnegative matrix decomposition in this invention are in Characteristic distribution in spectral space. (a) is the signal 1 obtained by nonnegative matrix decomposition in... (a) Characteristic distribution in spectral space; (b) Signal 2 obtained by nonnegative matrix decomposition in The characteristic distribution in the spectral space; (c) is the signal 3 obtained by nonnegative matrix decomposition in... The characteristic distribution in the spectral space; (d) is the signal 4 obtained by nonnegative matrix decomposition in... The characteristic distribution in the spectral space; (e) is the signal 5 obtained by nonnegative matrix decomposition in... Characteristic distribution in spectral space.

[0020] Figure 6 The following diagrams are used to mark the local maxima of five signal features obtained from non-negative matrix decomposition. (a) Marking the local maxima of signal 1 obtained from non-negative matrix decomposition; (b) Marking the local maxima of signal 2 obtained from non-negative matrix decomposition; (c) Marking the local maxima of signal 3 obtained from non-negative matrix decomposition; (d) Marking the local maxima of signal 4 obtained from non-negative matrix decomposition; (e) Marking the local maxima of signal 5 obtained from non-negative matrix decomposition.

[0021] Figure 7 To convert the feature dataset Input clustering algorithm for classification. (a) is... (a) is the result of the algorithm clustering; (b) is the Gaussian mixture model. (c) shows the clustering results of the Hierarchical algorithm; (d) shows the clustering results of the Hierarchical algorithm. The results of the algorithm clustering.

[0022] Figure 8 The images show the echo data and corresponding compressed data of different fluids based on the forward modeling results of clustering. (a) is the forward echo data of fluid 1; (b) is the compressed echo data of fluid 1; (c) is the forward echo data of fluid 2; (d) is the compressed echo data of fluid 2; (e) is the forward echo data of fluid 3; (f) is the compressed echo data of fluid 3; (g) is the forward echo data of fluid 4; (h) is the compressed echo data of fluid 4; (i) is the forward echo data of fluid 5; and (j) is the compressed echo data of fluid 5. The signal amplitude of the echo data, ordinates from (a) to (j) All are signal amplitudes from echo data; the abscissas of (a), (c), (e), (g), and (i) range from 1 to 5 × 10⁻⁶. 4 This indicates that the forward echo data contains 5×10 4 The x-coordinates of (b), (d), (f), (h), and (j) range from 1 to 500, representing that the compressed echo data contains 500 echoes.

[0023] Figure 9 The following are the results of multi-fluid saturation inversion using the method of this invention. (a) shows the saturation data of fluid 1, (b) shows the saturation data of fluid 2, (c) shows the saturation data of fluid 3, (d) shows the saturation data of fluid 4, and (e) shows the saturation data of fluid 5. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention are described clearly and completely below. Obviously, the described embodiments are only some embodiments of this invention, not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0025] like Figure 1 As shown, one of the present invention is based on A method for quantitative evaluation of multi-fluid components from acquired echo data includes the following steps: Step 1. Echo data acquisition: First collect Echo data, denoted as the original dataset. The original dataset It can be derived from laboratory cores Experiments, well sites Experiment or downhole Well logging. Among them, The size is , This refers to the number of stratigraphic points or the number of samples. This represents the number of echoes.

[0026] Different pulse sequences have different NMR response equations and construct different kernel matrices, achieving saturation recovery. Taking pulse sequences as an example, echo data and The relationships between the spectra are as follows: ; in, It is echo data, that is Data from one sample, such as Figure 2 As shown, For the waiting time, yes Spectrum It's a noise signal. For time.

[0027] The matrix form of the formula is: ; in, It is echo data, and its size is , It is the number of echoes. yes The spectrum, its size is nuclear matrix It can be done The calculated size is , yes Spectrum length, The number of distribution points is , The number of distribution points is , .

[0028] Step 2. Echo data compression: For the original dataset Compression processing is performed to obtain compressed echo data. The compression methods include window averaging compression, discrete cosine transform, singular value decomposition, or combinations thereof, to reduce data dimensionality and retain key signal features. Figure 3 Demonstrates the use of the window averaging method for Figure 2 The result after compression processing of mid-echo data.

[0029] Specifically, the compression method applies to the kernel matrix. and the original dataset Each line of echo data All are compressed, resulting in a compressed kernel matrix. and compressed echo data The sizes are respectively and .in, This represents the number of echoes after compression. yes Spectrum length, The number of distribution points is , The number of distribution points is , Compressed echo data is obtained by compressing the data at each depth point. Its size is . This refers to the number of stratigraphic points or the number of samples. This represents the number of echoes after compression.

[0030] ; ; in, This is the compressed echo data, and its size is , It is the number of compressed echoes. It is echo data, and its size is , It is the number of echoes. This is the compressed kernel matrix, and its size is , It is the first The boundary index of an average window, , This represents the number of echoes within the window. It is the window index. It is the index of the echo within the corresponding window.

[0031] Step 3. Based on compressed echo data, perform... Spectral inversion: Using compressed echo data Nuclear magnetic resonance inversion method was used to obtain the target stratum sample. Spectral distribution, denoted as Nuclear magnetic resonance inversion methods include , and Regularization, etc., serves to invert time-domain echo data into... Spectrum. Figure 4 Demonstrates the use of Methods Figure 3 The result after data inversion processing.

[0032] The input data for the inversion method is the compressed kernel matrix. and compressed echo data The compressed echo data It is compressed echo data Data from one sample. Compressed echo data from all formations. Inversion is performed to obtain the target stratigraphic sample. Spectral distribution Its size is . This refers to the number of stratigraphic points or the number of samples. yes Spectrum length, The number of distribution points is , The number of distribution points is , .

[0033] Step 4. Nonnegative matrix decomposition and feature extraction: Because the results of nonnegative matrix factorization are random, for the target formation sample... Spectral distribution Perform nonnegative matrix factorization, extract the signal peak coordinates in each factorization, and statistically analyze them to form a feature dataset. .

[0034] Specifically, in each decomposition process, the nonnegative matrix decomposition yields the product of two nonnegative matrices, which represent the weight matrix and the feature matrix, respectively, as shown in the following formula: ; in, It is a weight matrix. For the characteristic matrix, Indicates the various fluid components in Feature distribution in spectral space This represents the number of fluid components in the formation. For each decomposition, the characteristic matrix is... arranging each row according to direction and The directional dimension unfolds into a two-dimensional matrix. This two-dimensional matrix corresponds to a certain fluid component in The distribution characteristics on the spectrum, where the five signals obtained from a first nonnegative matrix decomposition are... Feature distribution in spectral space, such as Figure 5 As shown. Subsequently, in each two-dimensional matrix The search term searches for the location of a local maximum; the location of this maximum value is the location of the fluid component in the search. Peak coordinates in the spectrum Figure 6 Showing the Figure 5 The results of marking local maxima points in the feature distribution. The peak coordinates extracted during all decomposition processes are statistically analyzed and summarized to form the final feature dataset. This feature dataset Capable of stably characterizing different fluid components in the target formation The distribution characteristics in the spectral space provide a basis for subsequent fluid identification and quantitative evaluation.

[0035] Step 5. Cluster analysis and multi-fluid feature recognition: feature dataset The input is used for classification by a clustering algorithm, and the result is as follows: Figure 7 As shown. The clustering algorithms used include Clustering, Gaussian mixture model Hierarchical clustering and fuzzy clustering Mean value clustering, etc. By statistically analyzing the clustering results, the values ​​of various fluid components can be obtained. , Analyze the characteristic parameters in the direction and calculate their ratios. This enables the effective identification and differentiation of different fluid components.

[0036] Step 6. Constructing forward echo data: Based on the fluid characteristic parameters obtained from cluster analysis, forward echo data for different fluids are constructed. The result is as follows Figure 8 As shown in (a), (c), (e), (g), and (i), compressed echo data is obtained using the same compression method as in step 2. The result is as follows Figure 8 As shown in (b), (d), (f), (h), and (j). The size is , The size is . This represents the number of fluid components in the formation. The number of echoes This represents the number of echoes after compression.

[0037] Step 7. Multifluid saturation inversion: Compress echo data using step 6 And the compressed echo data from step 2 The saturation inversion of multifluid components was performed, and the calculation results are as follows: Figure 9 As shown. The inversion satisfies the following constraints: Non-negative constraints, saturation of each fluid ≥ 0; As a constraint, the sum of the saturations of all fluids is 1.

[0038] Furthermore, considering that the echo signal follows a multi-exponential decay characteristic, a weighting function of exponential decay is introduced. This data is compressed and added to the inversion model to enhance the stability and accuracy of the inversion results. The inversion model can be represented as: ; in, It is compressed echo data. It is an exponentially decaying weighted function. It refers to the saturation of multiple fluids. It is the first The saturation of a fluid, It is the number of fluid components in the formation. It is an index of fluids. To obtain the matrix in it Norm; ; in, It is a coefficient used to weigh the weights of the preceding and following echoes. For the weights of different echoes, It is the weight of the first echo. It is the first The weight of each echo, It is the first The time corresponding to each echo.

[0039] Step 8. Output the results: Ultimately, the saturation information of multi-fluid components is obtained, enabling continuous, quantitative, and accurate evaluation of different fluids. Furthermore, the saturation information of multi-fluid components and... The spectrum is used to determine the fluid porosity at different depths of the target stratum.

[0040] The present invention also provides a method based on A multi-fluid component quantitative evaluation system for acquiring echo data includes a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to achieve the above-mentioned... Steps for quantitative evaluation of multifluid components by acquiring echo data.

[0041] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method based on A method for quantitative evaluation of multi-fluid components based on acquired echo data, characterized in that: include: Step 1. Acquire the nuclear magnetic resonance echo data collected in the laboratory, at the well site, or downhole to form a raw data set ; Step 2. Compressing the original data set described in Step 1 to obtain compressed echo data ; Step 3. Compressing the echo data based on the description in Step 2 Performing inversion to obtain the target formation sample spectrum distribution ; Step 4. To the solution described in step 3 spectral distribution perform multiple non-negative matrix factorization, extract signal peaks and count to form a feature dataset , is the longitudinal relaxation time, is the transverse relaxation time; Step 5. Processing the feature data set described in Step 4 using a clustering algorithm to obtain relaxation characteristic parameters for a variety of fluids are processed to obtain relaxation characteristic parameters for a variety of fluids; Step 6. Construct forward echo data for different fluids based on the clustering results in Step 5. and the forward-modeled echo data Compression is performed to obtain compressed echo data. ; Step 7. Utilize the compressed echo data from Step 2 and step 6 compressing echo data Saturation inversion is performed under nonnegativity and sum-to-one constraints, combined with an exponentially decaying weighted function. Improve the stability and accuracy of inversion; Step 8. Output the saturation information of different fluids in the target formation based on the inversion results of Step 7.

2. The method of claim 1, wherein, The compression methods in step 2 include: window average compression, discrete cosine transform, singular value decomposition, or a combination thereof.

3. The method of claim 1, wherein, The clustering algorithm described in step 5 is Clustering, hierarchical clustering, Gaussian mixture model and blur Any type of mean clustering.

4. The method of claim 1, wherein, Step 7, saturation inversion, involves solving the linear system under non-negative constraints.

5. The method of claim 1, wherein, Also included is the use of the saturation information described in step 8 in conjunction with spectrum to determine fluid porosity at different depth points of the target formation.

6. A method based on A multi-fluid component quantitative evaluation system for acquiring echo data, comprising a memory and a processor, and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the method of any one of claims 1-5 based on Steps of a multi-fluid component quantitative evaluation method of acquired echo data.

Images