A shale multi-scale pore size distribution synthesis method based on a large-scale electron microscope

Abstract

The application provides a shale multi-scale pore size distribution synthesis method based on a large-scale electron microscope, relates to the technical fields of oil and gas development and image processing, and specifically comprises the following steps: a multi-scale shale data set is established through cluster analysis and direct interception; the pore sizes of the multi-scale shale are extracted to obtain pore size distributions under different scales; and the pore size distributions under different scales are subjected to pore area normalization to synthesize a multi-scale pore size distribution. The technical scheme of the application overcomes the problem in the prior art that the multi-scale pore size distribution of a real reservoir cannot be completely and accurately reproduced.

Description

Technical Field

[0001] This invention relates to the fields of oil and gas development and image processing, specifically to a method for synthesizing multi-scale pore size distribution of shale based on a large-size electron microscope. Background Technology

[0002] Shale reservoirs have complex mineral compositions and diverse pore types, resulting in multi-scale pore space characteristics. Existing methods for obtaining multi-scale pore size distributions are mostly based on combinations of fluid intrusion methods (such as high-pressure mercury intrusion and low-temperature nitrogen adsorption) and image scanning methods (such as CT imaging and electron microscopy). Examples include simple combination methods, fractal feature-based synthesis methods, differential distribution-based synthesis methods, and nuclear magnetic resonance combined with image scanning methods. While these methods provide a framework for synthesizing multi-scale pore size distributions, they cannot effectively handle overlapping areas between different methods, and are complex and time-consuming.

[0003] The existing methods have the following problems:

[0004] (1) Simple combination method. Based on the pore size distribution obtained by two or more fluid intrusion methods, a cutoff point is selected to synthesize the pore size distribution. Although this method yields multi-scale pore size distribution results, its reliability needs to be verified because there is no basis for combining the pore size distributions obtained by the two fluid intrusion methods.

[0005] (2) Synthesis method based on fractal features. The fractal features of the pore structure itself are used to obtain the multi-scale pore size distribution curve of the rock by combining numerical algorithms. Although the pore size curve results at different scales can be seamlessly connected, due to the multi-scale characteristics of shale reservoirs, the fractal features involved are not applicable to all shale reservoirs. At the same time, there are errors in the linear fitting involved in this method.

[0006] (3) Synthesis method based on differential distribution. This method combines high-pressure mercury intrusion and constant-rate mercury intrusion, and combines the pore size distribution obtained by the two methods by finding segments with the same or similar trends. However, since the processing of similar segments is combined with numerical algorithms, the reliability of the segment cannot be verified.

[0007] In summary, existing technologies all suffer from fitting errors or rounding errors during implementation, and cannot completely and accurately reproduce the multi-scale pore size distribution of real reservoirs.

[0008] Therefore, there is a need for a method for synthesizing the multi-scale pore size distribution of shale based on a large-size electron microscope that can completely and accurately reproduce the multi-scale pore size distribution of real reservoirs. Summary of the Invention

[0009] The main objective of this invention is to provide a method for synthesizing multi-scale pore size distribution in shale based on a large-size electron microscope, so as to solve the problem that the existing technology cannot completely and accurately reproduce the multi-scale pore size distribution of real reservoirs.

[0010] To achieve the above objectives, this invention provides a method for synthesizing multi-scale pore size distribution of shale based on a large-size electron microscope, specifically including the following steps:

[0011] S1 establishes a multi-scale shale dataset through cluster analysis and direct extraction.

[0012] S2, extract the pore size of shale at multiple scales to obtain the pore size distribution at different scales.

[0013] S3 normalizes the pore area of ​​the pore size distribution at different scales to synthesize a multi-scale pore size distribution.

[0014] Furthermore, step S1 specifically includes the following steps:

[0015] S1.1, the large-size MAPs image is traversed and segmented by a sliding window with fixed pixels to obtain the sub-image dataset for machine learning; the gray-level features of each sub-image in the dataset are extracted, and the gray-level features of the dataset are clustered based on K-means clustering, with the number of clusters being N, to obtain the clustering results of the sub-images, and then a clustered dataset is established.

[0016] S1.2, the pixel count of the MAPs is compressed to obtain compressed images of the MAPs at different nanometer resolutions, which are used to characterize the porosity information at different resolutions; the relationship between pixel count and resolution is obtained as follows:

[0017] ;

[0018] in, MAPs represent the actual physical length in nanometers. The resolution of MAPs is expressed in nanometers. This represents the number of pixels in a certain direction of the MAPs.

[0019] Furthermore, step S1.1 specifically includes:

[0020] Large-size MAPs are segmented using a 256×256 pixel sliding window to obtain a sub-image dataset for machine learning. The grayscale features of each sub-image are extracted, and K-means clustering is used to cluster the grayscale features of the dataset, with N clusters, resulting in the clustering results for the corresponding images, thus establishing a clustered dataset. After clustering, the large-size MAPs are further subdivided into N classes of smaller images, with each class representing M% of the total image size. i, i=1,2,3,4,5…, the image size is 256 pixels × 256 pixels, and the image resolution is 20 nanometers.

[0021] Furthermore, step S1.2 specifically includes:

[0022] The pixel size of MAPs is compressed to 1664 pixels × 2432 pixels, and the resolution becomes 320 nanometers; the physical size of MAPs is compressed to 832 pixels × 1216 pixels, and the resolution becomes 640 nanometers.

[0023] Furthermore, step S2 specifically includes the following steps:

[0024] S2.1 Based on the clustering dataset established in step S1.1, threshold segmentation is performed to obtain binarized images. The pore network model (PNM) of the binarized images is extracted, and the pore size distribution (PSD) of the PNM is derived. Multiple images are selected from each class to obtain the PSD, and the average PSD of the images is obtained by taking the average value. The average PSD of each class is weighted according to the clustering ratio to obtain the weighted PSD at a resolution of 20 nanometers.

[0025] S2.2, threshold segmentation is performed on the MAPs compressed images at different nanometer resolutions obtained in step S1.2 to obtain binarized images. The PNM of the binarized images is further extracted, and the PSD of the PNM is exported, thus obtaining the PSD at different nanometer resolutions.

[0026] Further, step S2.1 specifically includes:

[0027] Based on the clustering dataset established in step S1.1, threshold segmentation is performed to obtain binarized images, the PNM of the binarized images is extracted, and the pore size distribution PSD of the PNM is derived; 20 images are selected from each class to obtain PSD, and the average PSD of the class is obtained by taking the average value; the average PSD of each class is weighted according to the clustering ratio to obtain the weighted PSD at 20 nanometer resolution.

[0028] Furthermore, step S2.2 specifically includes:

[0029] Thresholding is performed on the MAPs compressed images at 320 nm and 640 nm resolutions obtained in step S1.2 to obtain binarized images. The PNM of the binarized images is further extracted, and the PSD of the PNM is exported to obtain the PSD at 320 nm and 640 nm resolutions.

[0030] Furthermore, step S3 specifically includes:

[0031] PSDs at different resolutions are normalized based on pore area. That is, the physical size of the MAPs is fixed, and the pore size is represented by the horizontal axis and the pore area by the vertical axis to describe the multi-scale PSD. The normalization method is as follows:

[0032] ;

[0033] ;

[0034] ;

[0035] in, and Re represents the width and height of the MAPs, in nanometers; Re represents the resolution, in nanometers; P represents the number of pixels, without units. and These correspond to the weights in the width and height directions, respectively, and are unitless. The pore area is in square nanometers. The pore diameter is in nanometers. The porosity PSD is a single-scale value and has no unit.

[0036] The present invention has the following beneficial effects:

[0037] This invention proposes a method for synthesizing multi-scale pore size distributions in shale based on a large-size electron microscope. The implementation process does not involve a fitting process, thus eliminating fitting errors. Furthermore, the high-precision electron microscope offers advantages such as high resolution and a large field of view, overcoming errors in connecting pore size distributions at different resolutions. By combining the physical information from shale MAPs images, this invention achieves multi-scale characterization of the heterogeneous pore structure of shale, providing technical support for the evaluation of shale reservoir space. Attached Figure Description

[0038] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0039] Figure 1 A flowchart of a method for synthesizing multi-scale pore size distribution in shale based on a large-size electron microscope, according to the present invention, is shown.

[0040] Figure 2 A schematic diagram illustrating the process of acquiring sub-image datasets from large-size electron microscope images is shown.

[0041] Figure 3The distribution of various micro-features after clustering in the original MAPs is shown.

[0042] Figure 4 Representative images of various micro-features after clustering are shown.

[0043] Figure 5 PSDs with different types of microscopic features are shown.

[0044] Figure 6 Weighted composite PSDs of different types of micro-features are shown.

[0045] Figure 7 A multi-scale aperture distribution map based on area normalization is shown. Detailed Implementation

[0046] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0047] like Figure 1 The method for synthesizing multi-scale pore size distribution of shale based on a large-size electron microscope, as shown, specifically includes the following steps:

[0048] S1, Multi-scale data establishment: A multi-scale shale dataset is established through cluster analysis and direct extraction.

[0049] S2, Single-scale pore size distribution acquisition: Extract the pore size of multi-scale shale to obtain the pore size distribution at different scales.

[0050] S3, Multi-scale pore size distribution synthesis: Normalize the pore area of ​​the pore size distribution at different scales to synthesize a multi-scale pore size distribution.

[0051] Specifically, step S1 includes the following steps:

[0052] S1.1, the large-size MAPs image is traversed and segmented by a sliding window with fixed pixels to obtain the sub-image dataset (i.e., the image patches segmented by the sliding window) for machine learning; the gray-level features of each sub-image in the dataset are extracted, and the gray-level features of the dataset are clustered based on K-means clustering, with the number of clusters being N, thereby obtaining the clustering results of the sub-images and establishing the clustered dataset. Figure 3 This represents the distribution of the clustered sub-images within the original MAPs.

[0053] After clustering, large-size MAPs are subdivided into N classes of smaller-size images, with each class comprising M images. i(i=1,2,3,4,5), the image size is 256 pixels × 256 pixels, and the images in each class have the same structural features. Since the multi-scale dataset is built directly based on MAPs, the images are not compressed and the resolution is maintained at 20 nanometers.

[0054] Large-size map images are segmented using a 256×256 pixel sliding window to obtain a dataset of sub-images (i.e., image patches segmented by the sliding window) for machine learning. The grayscale features of each sub-image are extracted, and the grayscale features are clustered using a machine learning method (K-means clustering) with N clusters, thus obtaining the clustering results for the corresponding images. After clustering, the large-size map images are subdivided into N classes of smaller images, with each class representing M% of the total image size. i (i=1,2,3,4,5…), the image size is 256 pixels × 256 pixels, and the images in each class have the same structural features. Since the multi-scale dataset is built directly based on MAPs, there is no data loss due to image compression, and the resolution is maintained at 20 nanometers.

[0055] Select a MAPs image and perform cluster analysis, such as Figure 2 As shown. The MAPs have a pixel size of 26624 pixels × 38912 pixels. The resolution is 20 nanometers, and the physical size is 0.53 mm × 0.78 mm. The MAPs are traversed using a 256 pixel × 256 pixel sliding window, dividing them into a series of smaller images of 256 pixels × 256 pixels, resulting in a dataset at 20 nanometer resolution. The gray-level co-occurrence matrix (GMM) of the images in the dataset is extracted, and the GMM is clustered using the K-means clustering algorithm, with a cluster size of 5. Finally, the MAPs are subdivided into 5 feature image classes and the proportion of each class, which are used for subsequent PSD extraction. Figure 4 The image shown is the feature image of each cluster after clustering, and the corresponding clustering results are shown in Table 1.

[0056] Table 1 Clustering Results

[0057]

[0058] S1.2, the pixel count of the MAPs is compressed to obtain compressed images of the MAPs at different nanometer resolutions, which are used to characterize the porosity information at different resolutions; the relationship between pixel count and resolution is obtained as follows:

[0059] ;

[0060] in, MAPs represent the actual physical length in nanometers, regardless of image compression. Remain unchanged; The resolution of MAPs, in nanometers. This represents the number of pixels in a certain direction of the MAPs. Follow The value changes with the resolution, and based on this relationship, compressed MAPs at a specific resolution can be obtained. The essence of compressed MAPs is to compress the number of pixels.

[0061] Direct cropping: The pixel size of the MAPs is compressed to 1664 pixels × 2432 pixels. Since the physical size remains unchanged, its resolution becomes 320 nanometers. The physical size of the MAPs is compressed to 832 pixels × 1216 pixels, and the resolution becomes 640 nanometers. The MAPs images at 320 nanometer and 640 nanometer resolutions are used for subsequent PSD extraction.

[0062] Specifically, step S2 includes the following steps:

[0063] S2.1 Based on the clustering dataset established in step S1.1, threshold segmentation is performed to obtain binarized images. The pore network model (PNM) of the binarized images is extracted, and the pore size distribution (PSD) of the PNM is derived. Multiple images are selected from each class to obtain the PSD, and the average PSD of the images is obtained by taking the average value. The average PSD of each class is weighted according to the clustering ratio to obtain the weighted PSD at a resolution of 20 nanometers.

[0064] Based on the clustering dataset established in step S1.1, threshold segmentation is performed to obtain binarized images, the PNM of the binarized images is extracted, and the pore size distribution PSD of the PNM is derived; 20 images are selected from each class to obtain PSD, and the average PSD of the class is obtained by taking the average value; the average PSD of each class is weighted according to the clustering ratio to obtain the weighted PSD at 20 nanometer resolution.

[0065] Taking the first category of images as an example, 20 images with typical (organic matter) features are selected from all images in the first category. The PNM (Network Matrix) of each binarized image is extracted, and the PSD (Planetary Detail Size) of the PNM is derived. The average PSD of the 20 PSDs is then calculated to obtain the average PSD of the first category of images. Similarly, the average PSDs of the second, third, fourth, and fifth categories are obtained. The average PSDs of the five categories are weighted according to their cluster proportions to obtain the PSD at 20 nanometer resolution. Figure 5 As shown.

[0066] Based on the cluster ratio, different types of PSDs are weighted and synthesized to obtain the overall PSD at 20nm resolution, such as... Figure 6 As shown.

[0067] S2.2, threshold segmentation is performed on the MAPs compressed images at different nanometer resolutions obtained in step S1.2 to obtain binarized images. The PNM of the binarized images is further extracted, and the PSD of the PNM is exported, thus obtaining the PSD at different nanometer resolutions.

[0068] Thresholding is performed on the MAPs compressed images at 320 nm and 640 nm resolutions obtained in step S1.2 to obtain binarized images. The PNM of the binarized images is further extracted, and the PSD of the PNM is exported to obtain the PSD at 320 nm and 640 nm resolutions.

[0069] Specifically, since the probability distribution is dimensionless, PSDs at different resolutions cannot be directly synthesized. The pore area is used to normalize the PSDs at different resolutions, and then they are further combined to obtain multi-scale PSDs of shale.

[0070] Step S3 is as follows:

[0071] In PSD, the horizontal axis represents pore size, and the vertical axis represents probability distribution. Since probability distribution is dimensionless and cannot describe the quantitative relationships between different sizes, PSDs at different resolutions are normalized based on pore area. That is, the physical dimensions of the MAPs are fixed, and the horizontal axis represents pore size, while the vertical axis represents pore area to describe multi-scale PSDs. The normalization method is as follows:

[0072] ;

[0073] ;

[0074] ;

[0075] in, and Re represents the width and height of the MAPs, in nanometers; Re represents the resolution, in nanometers; P represents the number of pixels, without units. and These correspond to the weights in the width and height directions, respectively, and are unitless. The pore area is in square nanometers. The pore diameter is in nanometers. The porosity PSD is a single-scale value and has no unit.

[0076] By normalizing the pore area, PSDs from multiple resolutions are synthesized onto the same horizontal axis, forming a multi-scale PSD, such as... Figure 7As shown, the pore area normalization process avoids the drawback of directly combining probability distributions of PSDs at different resolutions. Its rationale lies in controlling the physical size of MAPs, meaning the pore area of ​​the MAPs at different resolutions is objectively quantified. At 20 nm resolution, it represents pores below 320 nm. At 320 nm resolution, it represents pores from 320 nm to 640 nm. At 640 nm resolution, it represents pores above 640 nm. Using the physically meaningful pore area as the ordinate constrains the probability distribution at different resolutions.

[0077] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A synthetic method for shale multi-scale pore size distribution based on large-scale electron microscopy, characterized by, Specifically comprising the following steps: S1, a multi-scale shale data set is established by cluster analysis and direct interception; S2, the pore size of the multi-scale shale is extracted to obtain the pore size distribution under different scales; S3, the pore size distribution under different scales is normalized by pore area to synthesize a multi-scale pore size distribution; Step S1 specifically comprises the following steps: S1.1, traverse the large-size MAPs image with a fixed pixel sliding window to obtain a machine learning sub-image data set; extract the gray scale features of each sub-image in the data set, cluster the gray scale features of the data set based on K-means clustering, the number of clusters is N, obtain the clustering result of the sub-image, and then establish a cluster data set; S1.2, compress the pixel number of MAPs to obtain MAPs compressed images under different nanometer resolutions to represent pore information under different resolutions; the relationship between the pixel number and the resolution is obtained as follows: ； wherein, is the true physical length of the MAPs, in nanometers; is the resolution of the MAPs, in nanometers; is the number of pixels in a direction of the MAPs.

2. The synthetic method of shale multi-scale pore size distribution based on large-scale electron microscopy according to claim 1, characterized in that, Step S1.1 specifically comprises: The large-size MAPs image is traversed and cut by a sliding window of 256 pixels x 256 pixels to obtain a machine learning sub-image data set; the gray scale features of each sub-image in the data set are extracted, the gray scale features of the data set are clustered based on K-means clustering, the number of clusters is N, the clustering result corresponding to the image is obtained, and a clustered data set is established; after clustering, the large-size MAPs are subdivided into N small-size images, and the proportion of each image is M i , i = 1, 2, 3, 4, 5…, the image size is 256 pixels x 256 pixels, and the image resolution is 20 nanometers.

3. The synthetic method of shale multi-scale pore size distribution based on large-scale electron microscopy according to claim 1, characterized in that, Step S1.2 specifically comprises: The pixel size of MAPs is compressed to 1664 pixels x 2432 pixels, and the resolution is changed to 320 nanometers; the physical size of MAPs is compressed to 832 pixels x 1216 pixels, and the resolution is changed to 640 nanometers.

4. The synthetic method of shale multi-scale pore size distribution based on large-scale electron microscopy according to claim 1, characterized in that, Step S2 specifically comprises the following steps: S2.1, based on the cluster data set established in step S1.1, threshold segmentation is performed to obtain a binary image, the pore network model PNM of the binary image is extracted, and the pore size distribution PSD of the PNM is derived; select multiple images in each class of images to obtain PSD, take the average to obtain the average PSD of the image; the average PSD of each class is weighted according to the cluster proportion to obtain the weighted PSD under a resolution of 20 nanometers; S2.2, respectively, threshold segmentation is performed on the MAPs compressed images under different nanometer resolutions obtained in step S1.2 to obtain binary images, and further extract the PNM of the binary images to derive the PSD of the PNM, i.e. to obtain the PSD under different nanometer resolutions.

5. The synthetic method of shale multi-scale pore size distribution based on large-scale electron microscopy according to claim 4, characterized in that, Step S2.1 specifically comprises: Based on the cluster data set established in step S1.1, threshold segmentation is performed to obtain a binary image, the pore network model PNM of the binary image is extracted, and the pore size distribution PSD of the PNM is derived; select 20 images in each class of images to obtain PSD, take the average to obtain the average PSD of the image; the average PSD of each class is weighted according to the cluster proportion to obtain the weighted PSD under a resolution of 20 nanometers.

6. The synthetic method of shale multi-scale pore size distribution based on large-scale electron microscopy according to claim 4, characterized in that, Step S2.2 specifically comprises: Respectively, threshold segmentation is performed on the MAPs compressed images under 320 nanometer and 640 nanometer resolutions obtained in step S1.2 to obtain binary images, and further extract the PNM of the binary images to derive the PSD of the PNM, i.e. to obtain the PSD under 320 nanometer and 640 nanometer resolutions.

7. The synthetic method of shale multi-scale pore size distribution based on large-scale electron microscopy according to claim 1, characterized in that, Step S3 specifically comprises: The PSDs under different resolutions are normalized based on the pore area, i.e. the physical size of MAPs is fixed, the horizontal axis is the pore size, and the vertical axis is the pore area to describe the multi-scale PSD, and the normalization method is as follows: ； ； ； wherein, and W and H are the width and height of the MAPs, respectively, in nanometers; Re is the resolution, in nanometers; and P is the number of pixels, unitless; and W and H are the width and height of the MAPs, respectively, in nanometers; Re is the resolution, in nanometers; and P is the number of pixels, unitless; is the pore area, in square nanometers; is the pore diameter, in nanometers; is the pore PSD at a single scale, unitless.

Images