Shale oil and gas reservoir multi-scale pore structure characterization method based on machine learning

Abstract

The invention provides a shale oil and gas reservoir multi-scale pore structure characterization method based on machine learning. The shale oil and gas reservoir multi-scale pore structure characterization method specifically comprises the following steps of obtaining a shale sample image and performing image segmentation; clustering the images by using a K-means clustering algorithm; classifying the segmented images into N classes through three features of gray vectors, variances and information entropies; and calculating a gray average value of the images in each category, and evaluating an image classification effect by using the gray average value. According to the technical scheme, the problem that in the prior art, comprehensive understanding and characterization of multi-type and multi-scale micro-pore structures of heterogeneous shale cannot be achieved at the same time is solved.

Description

Technical Field

[0001] The present invention relates to the field of shale oil and gas development, and in particular to a multi-scale pore structure characterization method for shale oil and gas reservoirs based on machine learning. Background Art

[0002] China's shale oil and gas resources are widely distributed and recoverable resources are abundant. Shale oil and gas has also become the focus of domestic unconventional oil and gas exploration and development. Compared with conventional oil and gas reservoirs, shale's unique sedimentary environment and self-generation and self-storage characteristics result in its reservoirs showing low porosity, ultra-low permeability, and multi-type and multi-scale pore structures. The fluid seepage law of shale is limited by the characteristics of the rock pore structure. Therefore, an accurate and comprehensive understanding of shale pore structure is the basis and key to accurately assessing reserves and efficiently exploiting shale oil and gas.

[0003] At present, the experimental methods for shale pore structure analysis are mainly classified into three categories: indirect fluid injection method, non-fluid injection method and intuitive image method. Image method is the most widely used technology in the field of shale pore characterization, but this technology is limited by the contradiction between the characterization unit and the resolution and the development of multi-resolution image fusion algorithm. The multi-scale understanding and characterization of shale pore structure still needs to be further improved.

[0004] Since the experimental techniques used in the past were relatively simple, the understanding and analysis of pore structure relied solely on a single experimental method, which could not accurately reveal the overall picture of shale pore structure.

[0005] Therefore, there is a need for a multi-scale pore structure characterization method for shale oil and gas reservoirs based on machine learning that can simultaneously achieve a comprehensive understanding and characterization of the multi-type and multi-scale microscopic pore structures of heterogeneous shales. Summary of the invention

[0006] The main purpose of the present invention is to provide a method for characterizing the multi-scale pore structure of shale oil and gas reservoirs based on machine learning, so as to solve the problem that the prior art cannot simultaneously achieve a comprehensive understanding and characterization of the multi-type and multi-scale microscopic pore structures of heterogeneous shales.

[0007] To achieve the above object, the present invention provides a method for characterizing the multi-scale pore structure of shale oil and gas reservoirs based on machine learning, which specifically comprises the following steps:

[0008] S1, obtain shale sample images and perform image segmentation.

[0009] S2, cluster the images using the K-means clustering algorithm.

[0010] S3, classify the segmented image into N categories through three features: grayscale vector, variance, and information entropy.

[0011] S4, calculates the grayscale mean of the images in each category, and uses the grayscale mean to evaluate the image classification effect.

[0012] Furthermore, step S1 is specifically as follows: a two-dimensional image of the shale core is obtained through a scanning experiment on the shale sample, a two-dimensional shale core image is selected, the number of horizontal and vertical blocks is set through the Matlab program, the image is segmented and divided, the number of horizontal and vertical blocks is n respectively, the core image is divided into n*n small images, and the divided n*n images are stored in a new folder.

[0013] Furthermore, step S2 specifically includes the following steps:

[0014] S2.1, initialize the cluster center, randomly select the image features stored in the new folder in step S1 as the initial cluster center, the image features include: gray vector value, information entropy and variance of the image.

[0015] S2.2, use the K-means clustering algorithm to cluster the image features into N categories, iteratively update the cluster center until the change of the cluster center is less than a certain threshold, and generate the final cluster center.

[0016] S2.3, using a clustering algorithm to determine the image category for each image; according to the category of the image, the image is saved in a corresponding folder.

[0017] Furthermore, step S3 specifically includes the following steps:

[0018] S3.1, obtain each image in the folder obtained in step S2, and calculate the grayscale vector and variance of the image.

[0019] S3.2, count the number of points at each image pixel level, calculate the probability of each pixel, and find the information entropy of the image pixel.

[0020] S3.3, use the grayscale vector, variance and information entropy three features to classify the segmented image into N categories. If the three features are close to a final cluster center, it will be assigned to a certain category.

[0021] Further, step S4 is specifically as follows:

[0022] Calculate the grayscale mean of the images in each category, and find the difference between the maximum mean and the minimum mean. Draw a bar graph with the category on the horizontal axis and the grayscale mean difference on the vertical axis to reflect the classification effect. The smaller the grayscale mean difference, the better the classification effect.

[0023] The present invention has the following beneficial effects:

[0024] The present invention uses the K-means algorithm in unsupervised learning. Different from the commonly used edge extraction of one image, the K-means method is applied to the classification of a large number of images; each extracted grayscale feature is regarded as a 64-dimensional vector rather than a single value; the accuracy obtained in this way is higher than that of using a single value. Combined with the characteristics of shale oil grayscale images that are only single-channel, the grayscale vector values ​​of the image are classified with better effect and faster speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the specific implementation of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the specific implementation or the prior art description. Obviously, the drawings described below are some implementations of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0026] Figure 1 A flow chart of a method for characterizing multi-scale pore structure of shale oil and gas reservoirs based on machine learning of the present invention is shown.

[0027] Figure 2 1-80 nm core images are shown.

[0028] Figure 3 Shows Figure 2 Partially segmented image.

[0029] Figure 4 A grayscale eigenvalue plot is shown. DETAILED DESCRIPTION

[0030] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] like Figure 1 A multi-scale pore structure characterization method for shale oil and gas reservoirs based on machine learning is shown, which specifically includes the following steps:

[0032] S1, obtain shale sample images and perform image segmentation.

[0033] S2, cluster the images using the K-means clustering algorithm.

[0034] S3, classify the segmented image into N categories through three features: grayscale vector, variance, and information entropy.

[0035] S4, calculates the grayscale mean of the images in each category, and uses the grayscale mean to evaluate the image classification effect.

[0036] Specifically, step S1 is as follows: a two-dimensional image of the shale core is obtained through a scanning experiment on the shale sample, a suitable two-dimensional shale core image is selected, and the number of horizontal and vertical blocks is set through the Matlab program to realize the segmentation and blocking of the image, and the number of horizontal and vertical blocks is selected to be 20 respectively, and the core image is divided into 400 small images. Each small image can also be further divided, and the 400 divided images are stored in a new folder.

[0037] Specifically, step S2 includes the following steps:

[0038] S2.1, initialize the cluster center, randomly select the image features stored in the new folder in step S1 as the initial cluster center, the image features include: gray vector value, information entropy and variance of the image.

[0039] S2.2, use K-means clustering algorithm to cluster image features into N categories, N is the number of random classifications from 4 to 7 categories, and iteratively update the cluster center until the change of the cluster center is less than a certain threshold, generating the final cluster center.

[0040] S2.3, using a clustering algorithm to determine the image category for each image; according to the category of the image, the image is saved in a corresponding folder.

[0041] Specifically, step S3 includes the following steps:

[0042] S3.1, obtain each image in the folder obtained in step S2, and calculate the grayscale vector and variance of the image.

[0043] S3.2, count the number of points at each image pixel level, calculate the probability of each pixel, and find the information entropy of the image pixel.

[0044] S3.3, use the grayscale vector, variance and information entropy three features to classify the segmented image into N categories. If the three features are close to a final cluster center, it will be assigned to a certain category.

[0045] Specifically, step S4 is as follows:

[0046] Calculate the grayscale mean of the images in each category, and find the difference between the maximum mean and the minimum mean. Draw a bar graph with the category on the horizontal axis and the grayscale mean difference on the vertical axis to reflect the classification effect. The smaller the grayscale mean difference, the better the classification effect. At the same time, the maximum mean difference in all categories is controlled to be less than 150 through looping to ensure a better classification effect.

[0047] The method provided by the present invention is described in detail below in conjunction with experiments:

[0048] (1) Image segmentation: Through scanning experiments on shale samples, such as Figure 2 As shown in the figure, a two-dimensional shale core image with a resolution of 1-80nm is selected, and the number of horizontal and vertical blocks is set by the program to achieve image segmentation and block division. Each block image can also be further divided. The number of horizontal and vertical blocks is selected to be 20 respectively, and the core image is divided into 400 small images, as shown in the figure. Figure 3 As shown. And save the 400 divided images into a new folder.

[0049] (2) K-means clustering: Initialize the cluster center and use the K-means clustering algorithm to cluster the grayscale features of the image into 6 categories. Iterate and update the cluster center until the change in the cluster center is less than a certain threshold, and then generate the final cluster center. Use the clustering algorithm to determine the category of each image. According to the category of the image, save them into the corresponding folder.

[0050] (3) Three-feature classification: Obtain the images in each new file, calculate the grayscale vectors of the images and their variances; then count the number of points at each pixel level, calculate the probability of each pixel point, and calculate their information entropy; and use the three feature results for joint classification.

[0051] (4) Classify and store images and evaluate the classification effect: To evaluate the classification effect, use the tictoc timing code to calculate the running speed, observe the similarity of images in various folders, and finally obtain the average t = 2.9381s. By calculating the grayscale mean of the images in each category and finding the difference between the maximum mean and the minimum mean, as shown in Figure 4 As shown, a bar graph is drawn with the horizontal axis as the category and the vertical axis as the grayscale mean difference, which is used to reflect the classification effect. The smaller the grayscale mean difference, the better the classification effect. At the same time, by cyclically controlling the maximum mean difference in all categories to be less than 150, a better classification effect is guaranteed.

[0052] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A multi-scale pore structure characterization method for shale oil and gas reservoirs based on machine learning, characterized in that: The specific steps include: S1, acquiring shale sample images and performing image segmentation; S2, clustering images using K-means clustering algorithm; S3, classify the segmented image into N categories through three features: grayscale vector, variance, and information entropy; S4, calculates the grayscale mean of the images in each category, and uses the grayscale mean to evaluate the image classification effect.

2. The method for characterizing the multi-scale pore structure of shale oil and gas reservoirs based on machine learning according to claim 1, characterized in that: Step S1 is specifically as follows: a two-dimensional image of the shale core is obtained through a scanning experiment on the shale sample, a two-dimensional shale core image is selected, the number of horizontal and vertical blocks is set through the Matlab program, the image is divided into blocks, the number of horizontal and vertical blocks is n respectively, the core image is divided into n*n small images, and the divided n*n images are stored in a new folder.

3. The method for characterizing the multi-scale pore structure of shale oil and gas reservoirs based on machine learning according to claim 1, characterized in that: Step S2 specifically includes the following steps: S2.1, initialize the cluster center, randomly select the image features stored in the new folder in step S1 as the initial cluster center, the image features include: gray vector value, information entropy and variance of the image; S2.2, use K-means clustering algorithm to cluster image features into N categories, iteratively update cluster centers until the change of cluster centers is less than a certain threshold, and generate the final cluster centers; S2.3, using a clustering algorithm to determine the image category for each image; according to the category of the image, the image is saved in a corresponding folder.

4. The method for characterizing the multi-scale pore structure of shale oil and gas reservoirs based on machine learning according to claim 1, characterized in that: Step S3 specifically includes the following steps: S3.1, obtain each image in the folder obtained in step S2, and calculate the grayscale vector and variance of the image; S3.2, count the number of points at each image pixel level, calculate the probability of each pixel, and find the information entropy of the image pixel; S3.3, use the grayscale vector, variance and information entropy three features to classify the segmented image into N categories. If the three features are close to a final cluster center, it will be assigned to a certain category.

5. The method for characterizing the multi-scale pore structure of shale oil and gas reservoirs based on machine learning according to claim 1, characterized in that: Step S4 is specifically as follows: Calculate the grayscale mean of the images in each category, and find the difference between the maximum mean and the minimum mean. Draw a bar graph with the category on the horizontal axis and the grayscale mean difference on the vertical axis to reflect the classification effect. The smaller the grayscale mean difference, the better the classification effect.

Images