Quantitative analysis method for pore structure parameters of carbonate reservoir based on machine learning

Abstract

The invention relates to a carbonate rock reservoir pore structure parameter quantitative analysis method based on machine learning. The method comprises the following steps: acquiring a microscopic image of a casting body slice of a carbonate rock sample; analyzing the microscopic image according to the morphological characteristics of pores in the carbonate rock, respectively labeling pore and non-pore areas by adopting an interactive method, and generating a training data set; the method comprises the following steps of: extracting multi-scale Gaussian intensity features, LBP texture features and morphological gradient features of a microscopic image, and constructing a high-dimensional feature space by adopting a hierarchical fusion mode; training a random forest classifier to obtain a pore probability distribution diagram of the microscopic image; carrying out segmentation and morphological optimization on the pore probability distribution diagram of the microscopic image, and outputting a binary pore classification result; and calculating the surface porosity, the pore size, the pore size distribution and the pore connectivity index based on the binarized pore classification result. The method is high in accuracy and good in applicability; processing is very convenient, and manpower and material resource consumption is low.

Description

Technical Field

[0001] The present invention relates to the technical field of carbonate reservoir exploration and development, and in particular to a method for quantitatively analyzing pore structure parameters of carbonate reservoirs based on machine learning. Background Art

[0002] Carbonate reservoirs, due to their wide distribution, great thickness, and large scale, have become an important carrier of global oil and gas resources. As the core factor controlling the occurrence of oil and gas in carbonate reservoirs, pore structure has been extensively studied by numerous scholars at home and abroad using a variety of technical methods. For example, image observations using optical microscopes and scanning electron microscopes qualitatively characterize reservoir space type, reservoir space distribution, pore throat distribution, and other characteristics; and indirect measurements using fluid intrusion methods based on NMR nuclear magnetic resonance experiments and high-pressure mercury injection experiments quantitatively characterize pore structure parameters such as pore size, pore size distribution, and pore connectivity. However, fluid intrusion methods, such as NMR nuclear magnetic resonance experiments, are expensive and require high sample volume, making them unsuitable for large-scale sampling. This can lead to misunderstandings in reservoir characterization studies and has certain limitations.

[0003] Rock casting thin sections are low-cost, easy to obtain, and require simple sample preparation. With the development of image processing technology, thin section images can be used to extract pore structure information and thus conduct quantitative analysis.

[0004] Chinese patent publication CN108734714B discloses a Matlab-based method for analyzing carbonate rock structure. This method uses a thin-section image of a cast body to be binarized using Matlab's built-in ColorThreshold application. The bwlabel function is then used to identify pores and particles in the binary image, and pore structure parameters such as particle size and pore size are then calculated. However, this method does not effectively optimize the morphological features of the microscopic images of the cast body, resulting in the inability to identify tiny pores smaller than 2 μm. Consequently, the quantitative characterization results of the pore structure differ significantly from the actual situation. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a method for quantitative analysis of pore structure parameters of carbonate reservoirs based on machine learning.

[0006] The present invention is achieved through the following technical solutions:

[0007] The quantitative analysis method of carbonate reservoir pore structure parameters based on machine learning includes the following steps:

[0008] obtaining a carbonate rock sample, and preparing a number of standard casting thin sections based on the carbonate rock sample;

[0009] Obtain microscopic images of thin sections of the castings;

[0010] Preprocessing of microscopic images of casting thin sections;

[0011] Based on the pre-processed microscopic images, the microscopic images are analyzed according to the pore morphology characteristics in carbonate rocks, and the pore and non-porous areas are annotated using an interactive method to generate a training dataset.

[0012] Based on the training data set, multi-scale Gaussian intensity features, LBP texture features and morphological gradient features of the microscopic image are extracted, and a high-dimensional feature space is constructed using a hierarchical fusion method;

[0013] Based on the training data set and the high-dimensional feature space, a random forest classifier is trained to obtain a pore probability distribution map of the microscopic image;

[0014] Segment and morphologically optimize the pore probability distribution map of the microscopic image and output the binary pore classification results;

[0015] Based on the binary pore classification results, the surface porosity, pore size, pore size distribution, and pore connectivity index are calculated.

[0016] Among them, the preprocessing of microscopic images includes grayscale normalization, noise reduction and enhancement of pore edge contrast.

[0017] Preferably, the original microscopic image is converted into an 8-bit grayscale image by grayscale normalization.

[0018] Preferably, a non-local means algorithm is used to perform noise reduction on the microscopic image.

[0019] Preferably, a limited contrast adaptive histogram equalization algorithm is used to enhance the pore edge contrast.

[0020] Among them, the multi-scale Gaussian intensity is expressed as:

[0021]

[0022] In the above formula, Expressed as I(x,y) and The result after convolution, I(x,y) represents the pixel intensity of the original image at the coordinate (x,y), Expressed as standard deviation σ i The two-dimensional Gaussian kernel function, σ represents the standard deviation of the Gaussian kernel, σ i It is represented as the standard deviation of the Gaussian kernel of the i-th scale, and (x, y) represents the position coordinates of the pixel in the kernel relative to the center.

[0023] Preferably, an adaptive threshold method is used to segment the pore probability distribution map of the microscopic image.

[0024] The pore connectivity index is calculated using the following formula:

[0025] x=N pore -N cpore +1

[0026]

[0027] In the above formula, x represents the Euler number; N pore Represents the total number of pores, which refers to the sum of the number of isolated pores and the number of connected pores; N cpore represents the number of connected pores; C represents the connectivity index. A C value close to 1 indicates better pore connectivity, and a C value close to 0 indicates worse pore connectivity.

[0028] Compared with the prior art, this application has at least the following beneficial effects:

[0029] This application labels a small number of typical pore and non-pore regions to obtain multi-scale Gaussian intensity features, LBP texture features, and morphological gradient features. This method then trains a random forest classifier to achieve high-precision classification of complex samples. Combined with an adaptive thresholding method and morphological optimization, this method improves the accuracy of the structural parameter calculation results for pore units. Compared to existing threshold segmentation methods, the proposed method significantly improves its applicability to complex pore structures. Compared to existing deep learning methods, this method requires only a small amount of annotation to train the model, making the process very convenient and requiring relatively low human and material resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0031] Figure 1 Flow chart of the method for quantitatively analyzing pore structure parameters of carbonate reservoirs in the embodiment;

[0032] Figure 2 This is a microscopic image of a cast thin section of a carbonate reservoir sample in the examples;

[0033] Figure 3 A schematic diagram of interactively marking pore and non-pore areas in an embodiment;

[0034] Figure 4 is a binary image of a carbonate reservoir sample in the embodiment;

[0035] Figure 5is a pore size distribution histogram in the embodiment. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0037] It should be noted that, in the absence of any conflict, the embodiments and features of the embodiments of the present invention may be combined with each other. It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to in conjunction with each other.

[0038] like Figure 1 As shown, this embodiment discloses a method for quantitatively analyzing pore structure parameters of carbonate reservoirs based on interactive machine learning, which specifically includes the following steps:

[0039] S1, obtaining a carbonate rock sample and preparing several standard casting thin sections based on the carbonate rock sample;

[0040] S2, taking polarized light microscopic pictures of the casting slice to obtain a microscopic image of the casting slice, such as Figure 2 As shown;

[0041] S3, grayscale normalization is performed on the casting thin section microscopic image, noise reduction is performed on the microscopic image, and pore edge contrast is enhanced, and then the pre-processed microscopic image is output.

[0042] Optionally, convert the original microscopic image to an 8-bit grayscale image. Pixel values ​​range from 0 to 255, where 0 represents black, 255 represents white, and intermediate values ​​represent varying shades of gray.

[0043] Optionally, the non-local means algorithm is used to reduce noise on the microscopic image, with a search window of 7×7, a domain window of 4×4, and a noise reduction intensity of 10;

[0044] Optionally, a contrast limited adaptive histogram equalization algorithm is used to enhance the pore edge contrast.

[0045] S4, based on the pre-processed microscopic images, the microscopic images of the casting thin sections were analyzed according to the pore morphology characteristics in carbonate rocks, and the pore and non-porous areas were annotated separately using an interactive method (e.g. Figure 3and generate a training dataset.

[0046] It is worth noting that when labeling, attention should be paid to the number of samples of each category, and the number of labels of each category should be manually controlled to balance each category to avoid the model being biased towards the majority category.

[0047] S5, based on the training data set obtained in S4, extracts the multi-scale Gaussian intensity features of the casting thin section micrograph to obtain the intensity distribution difference between the pore area and the non-pore area; the multi-scale Gaussian intensity is expressed as:

[0048]

[0049] In the above formula, σ represents the standard deviation of the Gaussian kernel, which controls the smoothness of the filter, and σ i It is represented as the standard deviation of the Gaussian kernel at the i-th scale; (x, y) represents the position coordinates of the pixel in the kernel relative to the center, and the range is usually x, y∈[-kσ, kσ];

[0050] By extracting the LBP (Local Binary Patterns) texture features of the casting thin section micrograph, the internal microstructural distribution differences between the porous area and the non-porous area are obtained.

[0051] By extracting the morphological gradient features of the casting thin section micrograph, the edge information of the pore area and the non-pore area is highlighted.

[0052] Multi-scale Gaussian intensity features, LBP texture features and morphological gradient features are combined to construct a high-dimensional feature space using a hierarchical fusion approach to improve the recognition accuracy of complex targets in thin-section microscopic images.

[0053] S6, based on the training dataset from S4 and the high-dimensional feature space from S5, trains a random forest classifier and generates a pore probability map for the microscopic images. During training, the input data is the high-dimensional feature values ​​of the original images after hierarchical fusion, and the output is the pore distribution probability.

[0054] S7, the adaptive threshold method is used to segment the pore probability map of the microscopic image to ensure that different areas of the same image match the appropriate threshold; then morphological optimization is performed to fill the interior of the pore area through closing operations, and to filter out noise points with an area smaller than the threshold by deleting small areas; finally, the binary pore classification results are output, such as Figure 4 shown.

[0055] S8, based on the binary pore classification results of S7, combined with Image J image analysis software, used the AnalyzeParticles command to statistically and quantitatively calculate the pore structure parameters, including face ratio, pore size, pore size distribution, and pore connectivity.

[0056] The face rate is determined by the ratio of the number of pixels in the identified pore area to the total number of pixels in the image, which can be expressed as:

[0057]

[0058] In the above formula, φ represents the face rate, Q pore represents the number of pixels in the pore area, Q total Indicates the total number of pixels in the image;

[0059] The aperture size is calculated based on the area equivalent circle radius and can be expressed as:

[0060]

[0061] In the above formula, R eq represents the radius of the area equivalent circle, and S represents the pore area.

[0062] After obtaining the pore size of a single pore, the proportion of pores in different pore size intervals is counted to obtain a pore size distribution histogram, such as Figure 5 As shown;

[0063] The pore connectivity is quantitatively characterized based on the connectivity index. By obtaining the Euler number, the connectivity index is calculated. The connectivity index can be expressed as:

[0064]

[0065] Where x represents the Euler number, N pore Represents the total number of pores, which refers to the sum of the number of isolated pores and the number of connected pores; N cpore It indicates the number of connected pores. It should be noted that when pores are interconnected to form a pore network, they should be regarded as one connected pore. C represents the connectivity index. A C value close to 1 indicates better pore connectivity, and a C value close to 0 indicates worse pore connectivity.

[0066] This application labels a small number of typical pore and non-pore regions to obtain multi-scale Gaussian intensity features, LBP texture features, and morphological gradient features. This method then trains a random forest classifier to achieve high-precision classification of complex samples. Combined with an adaptive thresholding method and morphological optimization, this method improves the accuracy of the structural parameter calculation results for pore units. Compared to existing threshold segmentation methods, the proposed method significantly improves its applicability to complex pore structures. Compared to existing deep learning methods, this method requires only a small amount of annotation to train the model, making the process very convenient and requiring relatively low human and material resources.

[0067] This application is applicable to pore identification, segmentation and parameter calculation of casting thin section microscopic images, and can provide data support for reservoir evaluation.

[0068] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A quantitative analysis method for pore structure parameters of carbonate reservoirs based on machine learning, characterized in that: The following steps are involved: obtaining a carbonate rock sample, and preparing a number of standard casting thin sections based on the carbonate rock sample; Obtain microscopic images of thin sections of the castings; Preprocessing of microscopic images of casting thin sections; Based on the pre-processed microscopic images, the microscopic images are analyzed according to the pore morphology characteristics in carbonate rocks, and the pore and non-porous areas are annotated using an interactive method to generate a training dataset. Based on the training data set, multi-scale Gaussian intensity features, LBP texture features and morphological gradient features of the microscopic image are extracted, and a high-dimensional feature space is constructed using a hierarchical fusion method; Based on the training data set and the high-dimensional feature space, a random forest classifier is trained to obtain a pore probability distribution map of the microscopic image; Segment and morphologically optimize the pore probability distribution map of the microscopic image and output the binary pore classification results; Based on the binary pore classification results, the surface porosity, pore size, pore size distribution, and pore connectivity index are calculated.

2. The method for quantitative analysis of pore structure parameters of carbonate reservoirs based on machine learning according to claim 1, characterized in that: The preprocessing of microscopic images includes grayscale normalization, noise reduction and enhancement of pore edge contrast.

3. The method for quantitative analysis of pore structure parameters of carbonate reservoirs based on machine learning according to claim 2, characterized in that: The original microscopic image is converted into an 8-bit grayscale image through grayscale normalization.

4. The method for quantitative analysis of pore structure parameters of carbonate reservoirs based on machine learning according to claim 2, characterized in that: The non-local means algorithm is used to reduce noise in microscopic images.

5. The method for quantitative analysis of carbonate reservoir pore structure parameters based on machine learning according to claim 2, characterized in that: The contrast of pore edges is enhanced using the contrast-constrained adaptive histogram equalization algorithm.

6. The method for quantitative analysis of pore structure parameters of carbonate reservoirs based on machine learning according to claim 1, characterized in that: The multi-scale Gaussian intensity is expressed as: In the above formula, Expressed as I(x,y) and The result after convolution, I(x,y) represents the pixel intensity of the original image at the coordinate (x,y), Expressed as standard deviation σ i The two-dimensional Gaussian kernel function, σ represents the standard deviation of the Gaussian kernel, σ i It is represented as the standard deviation of the Gaussian kernel of the i-th scale, and (x, y) represents the position coordinates of the pixel in the kernel relative to the center.

7. The method for quantitative analysis of pore structure parameters of carbonate reservoirs based on machine learning according to claim 1, characterized in that: The adaptive threshold method is used to segment the pore probability distribution map of the microscopic image.

8. The method for quantitative analysis of pore structure parameters of carbonate reservoirs based on machine learning according to claim 1, characterized in that: The pore connectivity index was calculated using the following formula: x=N pore -N cpore +1 In the above formula, x represents the Euler number; N pore Represents the total number of pores, which refers to the sum of the number of isolated pores and the number of connected pores; N cpore represents the number of connected pores; C represents the connectivity index. A C value close to 1 indicates better pore connectivity, and a C value close to 0 indicates worse pore connectivity.

Images