Residual oil micromorphology division method and device based on CT scanning and machine learning
Through CT scanning and machine learning technology, a three-dimensional digital core model is established and automatically divided, which solves the limitations of qualitative analysis and two-dimensional observation of microscopic residual oil research in the existing technology, and achieves efficient and accurate microscopic storage of residual oil Type classification guides scientific exploitation of reservoirs.
Patent Information
- Application Number
- CN202410965828.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-18
- Publication Date
- 2024-11-12
AI Technical Summary
The prior art mainly has qualitative analysis in the research of microscopic residual oil, and it is difficult to achieve accurate and efficient automatic division. The traditional method is limited to two-dimensional observation and relies on manual identification efficiency, which cannot accurately reflect the true core pore structure.
CT scan is used to establish a real three-dimensional digital core model, combine the watershed method to divide the residual oil area, calculate the micro-assembly characterization parameters, and perform cluster analysis through self-organized mapping neural network machine learning method to automatically divide the residual oil microassembly type.
Accurate and efficient automatic division of micro-export types of residual oil in the core is achieved, and classification efficiency and effect are improved. Without manual intervention, it can guide reservoir mining more scientifically.
Smart Images

Figure CN118941829A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of oil reservoir exploitation, and in particular to a method and a device for dividing residual oil microscopic morphology based on CT scanning and machine learning. Background Art
[0002] Residual oil refers to the oil that has not been produced in the developed reservoirs. Residual oil is generally divided into two categories: microscopic residual oil and macroscopic residual oil. Microscopic residual oil can be roughly divided into drop-shaped, cluster-shaped, blind-end-shaped and film-shaped residual oil according to its existence form. The distribution of residual oil is mainly affected by factors such as pore structure characteristics and particle minerals, so there are certain differences between its microscopic occurrence forms.
[0003] The study of the occurrence morphology of microscopic residual oil is of great significance to exploration, development and improving oil recovery. Therefore, studying the occurrence morphology and spatial distribution of residual oil from the perspective of microscopic pore structure is conducive to revealing the laws of residual oil utilization and conducting scientific exploitation.
[0004] At present, most of the research on microscopic residual oil at home and abroad is based on qualitative analysis. This is because in the real core pore structure, the microscopic occurrence types of residual oil are complex and diverse, so it is difficult to conduct quantitative analysis of residual oil. When studying microscopic residual oil, if a rule model is used, it cannot intuitively reflect the real core pore structure, and the rule model method also has certain limitations.
[0005] With the development of experimental technology, a variety of experimental techniques and methods for observing the microscopic occurrence morphology of residual oil have been formed, including traditional oil-bearing thin section technology, scanning electron microscopy technology, CT scanning technology, nuclear magnetic resonance imaging technology, etc. The oil-bearing thin section technology uses high-strength adhesives to make thin sections of closed coring cores to observe the pores and distribution morphology of residual oil in the thin sections. It can only observe the two-dimensional occurrence morphology of residual oil, which has great limitations and is complicated to operate. The limitation of scanning electron microscopy technology in studying microscopic residual oil is that it can only obtain high-resolution images of the two-dimensional pore structure distribution on the sample surface, and cannot study the distribution of residual oil in a larger three-dimensional space. Compared with the oil-bearing thin section technology, the biggest advantage of CT scanning technology is fast and non-destructive scanning imaging. Digital cores are constructed through software algorithms, microscopic pore characteristics are intuitively studied, and pore throats are visualized.
[0006] Among them, the CT scanning method has become a hot topic in the study of residual oil in recent years because it can non-destructively detect the distribution pattern of residual oil in pores. However, the traditional methods for distinguishing the microscopic distribution morphology of residual oil mostly rely on manual experience to identify CT images, which is a huge workload and inefficient. Therefore, it is necessary to propose a more accurate automatic machine identification method.
[0007] CN105649615A discloses a method for quantitatively testing the occurrence state of tight oil in a reservoir by CT and three-dimensional visualization. By observing the CT scanning results, the different occurrence states of tight oil are manually identified, and the volume of tight oil in each occurrence state is statistically analyzed. The volume or mass of tight oil in each occurrence state is analyzed, but it is not possible to automatically divide each occurrence feature.
[0008] CN105551004A discloses a method for representing the microscopic occurrence of residual oil based on core CT image processing. The shape factor of the residual oil is calculated through CT scanning images, and the microscopic occurrence state of the residual oil is divided according to the shape factor. The parameters are relatively simple and the division standard is rough, and the accuracy of the division is not high.
[0009] CN112282748A discloses a method for distinguishing the residual oil occurrence morphology based on microfluidics and machine learning. The method for studying the distribution and occurrence state of the residual oil based on the microfluidics device has great limitations and a large gap with the real formation core, and has a narrow applicability.
[0010] Therefore, it is of great significance to develop a method that can accurately and efficiently automatically classify the microscopic occurrence forms of the remaining oil in the core. Summary of the invention
[0011] In view of the problems existing in the prior art, the present invention provides a method and device for dividing the microscopic morphology of residual oil based on CT scanning and machine learning. A real three-dimensional core model containing residual oil is established through CT scanning. Then, combined with the machine learning method, the microscopic occurrence type of the residual oil can be accurately and efficiently divided, which is conducive to revealing the utilization law of the residual oil and conducting scientific exploitation, and is of great significance to exploration and development as well as improving the recovery rate.
[0012] To achieve this object, the present invention adopts the following technical solutions:
[0013] In a first aspect, the present invention provides a method for dividing residual oil microscopic morphology based on CT scanning and machine learning, the method for dividing residual oil microscopic morphology comprising the following steps:
[0014] S1: Perform CT scanning on the core sample to establish a three-dimensional digital core model containing residual oil;
[0015] S2: using a watershed method to perform image segmentation on the three-dimensional digital core model to divide the remaining oil area;
[0016] S3: calculating the microscopic occurrence morphology characterization parameters of the residual oil according to the residual oil area, and forming a data set; the microscopic occurrence morphology characterization parameters of the residual oil include radius, surface area, shape factor and coordination number;
[0017] S4: Using machine learning methods, cluster analysis is performed on the data set to obtain the microscopic morphology division results of the remaining oil.
[0018] The method for dividing the residual oil microscopic morphology based on CT scanning and machine learning described in the present invention first uses the CT scanning method to obtain a three-dimensional digital core image, and the established three-dimensional digital core model containing residual oil is completely real and highly accurate; then the residual oil area in the core image is segmented by the watershed method, and the residual oil occurrence characteristic characterization parameters such as radius, surface area, shape factor and coordination number are calculated according to the residual oil area to form a data set, and these residual oil occurrence characteristic characterization parameters can accurately and comprehensively reflect the topological structure characteristics of the residual oil, which is convenient for the subsequent accurate division of the residual oil microscopic morphology; finally, the data set is clustered and analyzed by the machine learning method, so as to divide the microscopic occurrence types of the residual oil in the core, without manual intervention, with high classification efficiency and good classification effect. The method for dividing the residual oil microscopic morphology described in the present invention can accurately and efficiently automatically divide the microscopic occurrence morphology types of the residual oil in the core, which has guiding significance for the scientific exploitation of oil reservoirs.
[0019] Preferably, in step S1, before performing CT scanning on the core sample, an oil-water driving experiment is performed, and when no water comes out from the outlet, water is used to drive the core sample to a water content of 98% at the outlet; thereafter, the size of the core sample is processed to obtain a size suitable for CT scanning.
[0020] Preferably, after the CT scanning in step S1 obtains the three-dimensional information of the entire core sample, the three-dimensional volume data of the pore structure is obtained through image denoising, enhancement and filtering, and then a three-dimensional digital core model containing residual oil is established.
[0021] Preferably, according to the three-dimensional digital core model containing the remaining oil, it is concluded that the remaining oil includes four microscopic occurrence types: drop-shaped, cluster-shaped, blind-end-shaped and film-shaped.
[0022] Preferably, the method of performing image segmentation on the three-dimensional digital core model using the watershed method in step S2 comprises:
[0023] The grayscale value of the pixel in the input image is regarded as the altitude, and then the gradient amplitude is calculated to divide the remaining oil area.
[0024] Preferably, the gradient amplitude is the degree of difference between pixel gray levels.
[0025] In the present invention, a set of pixel points with greater similarity is extracted to form several sub-areas. The changes between pixel points with greater similarity are relatively small, and the gradient amplitude is relatively small. The pixel value of the gradient amplitude along the edge of the object is higher and is set as the boundary between different phases to divide the residual oil area.
[0026] Preferably, the radius in step S3 is the radius of the residual oil area, which is the radius of the inscribed sphere, and is denoted as R;
[0027] Preferably, the surface area is the surface area of a sphere inscribed within the radius of the residual oil region, denoted by S.
[0028] Preferably, the shape factor in step S3 is denoted as G, and its calculation formula is:
[0029] Where A is the cross-sectional area of the remaining oil, and P is the perimeter of the cross-sectional graph of the remaining oil.
[0030] Preferably, the coordination number is the number of connections between the residual oil and the pores or throats, denoted as C. The coordination number reflects the connectivity of the residual oil.
[0031] Preferably, the machine learning method in step S4 includes a machine learning method based on a self-organizing map neural network.
[0032] Preferably, the machine learning method based on the self-organizing map neural network has an input layer and a competition layer, the data set is imported into the input layer and the competition rate is set, and the residual oil microscopic morphology division result is obtained through the learning operation of the competition layer.
[0033] The input layer of the present invention is used to store the parameter set that needs to be clustered, and the competition layer will perform competitive learning between neurons according to the set learning rate. In this process, the competition process between neurons is to search for the neuron that is most similar to the input pattern, and the winner is called the best matching unit, which is output as the result.
[0034] As a preferred technical solution of the present invention, the residual oil microscopic morphology classification method comprises the following steps:
[0035] S1: Carry out oil-water driving experiment on core samples. When no water comes out from the outlet, water is used to drive the core outlet to 98% water content. Then, the size of the core samples is processed to obtain a size suitable for CT scanning, and then CT scanning is performed to obtain three-dimensional information of the entire core sample. The three-dimensional volume data of the pore structure is obtained through image denoising, enhancement and filtering, and then a three-dimensional digital core model containing residual oil is established. According to the three-dimensional digital core model containing residual oil, it is concluded that the residual oil includes four microscopic occurrence types: droplet-shaped, cluster-shaped, blind-end-shaped and film-shaped.
[0036] S2: using the watershed method to perform image segmentation on the three-dimensional digital core model, treating the grayscale value of the pixel point of the input image as the altitude, and then calculating the gradient amplitude to divide the remaining oil area;
[0037] The gradient amplitude is the degree of difference between pixel gray levels;
[0038] S3: calculating the microscopic occurrence morphology characterization parameters of the residual oil according to the residual oil area, and forming a data set; the microscopic occurrence morphology characterization parameters of the residual oil include radius, surface area, shape factor and coordination number;
[0039] The radius is the radius of the residual oil area, which is the radius of the inscribed sphere, recorded as R; the surface area is the surface area of the inscribed sphere, recorded as S; the shape factor is recorded as G, and its calculation formula is: Wherein A is the cross-sectional area of the residual oil, P is the perimeter of the cross-sectional graph of the residual oil; the coordination number is the number of connections between the residual oil and the pores or throats, denoted by C;
[0040] S4: A machine learning method based on a self-organizing map neural network with an input layer and a competition layer is used to import the data set into the input layer and set the competition rate. Through the learning operation of the competition layer, the microscopic morphology division result of the residual oil is obtained.
[0041] In a second aspect, the present invention further provides a residual oil microscopic morphology classification device based on CT scanning and machine learning, wherein the residual oil microscopic morphology classification device performs the residual oil microscopic morphology classification method based on CT scanning and machine learning described in the first aspect;
[0042] The residual oil microscopic morphology classification device comprises a core CT scanning module, a residual oil area classification module, a characterization parameter calculation module and a machine learning classification module which are connected in sequence.
[0043] The core CT scanning module described in the present invention is used to obtain the real structural image of the core containing residual oil; the residual oil area division module is used to divide the residual oil area in the digital core model to facilitate the calculation and analysis of the next part; the characterization parameter calculation module is used to calculate the characterization parameters of the residual oil area shape; the machine learning classification module is used to divide the residual oil microscopic morphology to obtain the final division result.
[0044] Compared with the prior art, the present invention has at least the following beneficial effects:
[0045] (1) The method for dividing the residual oil microstructure based on CT scanning and machine learning provided by the present invention uses a CT scanning method to obtain a three-dimensional image of a core containing residual oil, and the established three-dimensional digital core model containing residual oil is completely real and highly accurate;
[0046] (2) The present invention provides a method for dividing the microscopic morphology of residual oil based on CT scanning and machine learning. The method is based on a machine learning method of a self-organizing map neural network to divide the microscopic morphology of residual oil. The entire process is completed automatically by a machine without the need for human intervention. The speed is much faster than manual identification, and it has the advantages of high classification efficiency and good classification effect. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 It is a flow chart of a method for dividing residual oil microscopic morphology based on CT scanning and machine learning provided in a specific embodiment of the present invention.
[0048] Figure 2 It is a three-dimensional digital core model containing residual oil in a specific embodiment of the present invention.
[0049] Figure 3 It is a schematic diagram of the remaining oil area divided by the watershed method in a specific implementation manner of the present invention.
[0050] Figure 4 It is the result of the microscopic morphology division of the residual oil based on the machine learning method in the specific implementation manner of the present invention.
[0051] Figure 5 It is a schematic diagram of a residual oil microscopic morphology classification device based on CT scanning and machine learning in a specific embodiment of the present invention. DETAILED DESCRIPTION
[0052] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific implementation methods.
[0053] The present invention is further described in detail below. However, the following examples are only simplified examples of the present invention and do not represent or limit the scope of protection of the present invention. The scope of protection of the present invention shall be subject to the claims.
[0054] As a specific embodiment of the present invention, a method for dividing the microscopic morphology of residual oil based on CT scanning and machine learning is provided, and its flow chart is as follows: Figure 1 shown.
[0055] The residual oil microscopic morphology classification method comprises the following steps:
[0056] S1: Use 5 natural cores from Gudao Oilfield to conduct oil-water drive experiments until no water comes out of the outlet, and then use water to drive the core outlet to a water content of 98%. After that, the core samples are machined and processed to the size suitable for CT scanning. The core is placed in Xradia 800Ultra CT for scanning experiments until 360-degree rotation and scanning are completed, so as to obtain the three-dimensional information of the entire sample. Through image denoising, enhancement, filtering and other processing, the three-dimensional volume data of the pore structure that meets the analysis requirements is obtained, and a three-dimensional digital core model containing residual oil is established, such as Figure 2 According to the three-dimensional digital core model containing residual oil, it is concluded that the remaining oil includes four microscopic occurrence types: drop-shaped, cluster-shaped, blind-end-shaped and film-shaped.
[0057] S2: The watershed method is used to segment the three-dimensional digital core model in S1 to divide the remaining oil area. The watershed method first regards the grayscale value of the pixel of the three-dimensional digital core image containing remaining oil in S1 as the altitude, and then calculates its gradient amplitude, which is defined as the degree of difference between the pixel grayscale values. Then, a set of pixels with large similarity is extracted to form several sub-areas. The changes between pixels with large similarity are relatively small, and the gradient amplitude is relatively small. The pixel values along the edge of the object are higher and set as the boundary between different phases. In this way, a total of 563 remaining oil areas are divided, such as Figure 3 shown.
[0058] S3: Based on the image segmentation results in S2, the radius, surface area, shape factor, coordination number and other parameters characterizing the occurrence characteristics of the remaining oil in the core of the isolated oil field are calculated, and a data set is formed. Among them, the radius of the remaining oil area is the radius of the inscribed sphere, denoted as R, and the surface area is S. The shape factor is a parameter that describes the relative regularity of an object, denoted as G. Its calculation formula is: Where A is the cross-sectional area of the remaining oil, and P is the perimeter of the cross-sectional graph of the remaining oil. The coordination number is the number of connections between the remaining oil and the pores or throats, denoted by C, which reflects the connectivity of the remaining oil. The resulting data set includes 563 sets of residual oil radius R, surface area S, shape factor G, and coordination number C, as shown in Table 1.
[0059] Table 1
[0060] Remaining oil number R / μm <![CDATA[S / μm 2 ]]> G C 1# 5.4 91.6 0.02 3 2# 8.2 211.2 0.045 5 3# 3.2 32.17 0.037 2 … … … … … 563# 4.5 63.52 0.022 2
[0061] S4: Use the machine learning method based on the self-organizing map neural network to perform unsupervised clustering on the residual oil microscopic occurrence characteristic characterization data set established in S3. The input layer of the neural network is used to store the parameter set that needs to be clustered. The input layer is the residual oil occurrence characteristic characterization data set generated in S3, and the competition layer performs competitive learning between neurons according to the set learning rate. In this specific implementation method, the competition rate is set to 0.5. In this process, the competition process between neurons is to search for the neuron that is most similar to the input pattern; the winner is called the best matching unit and is output as the result. The division results of the residual oil microscopic morphology are: drop-shaped, cluster-shaped, blind-end-shaped and membrane-shaped.
[0062] The results of the residual oil micro-type classification based on machine learning in this specific implementation are shown in Tables 2 and Figure 4 shown.
[0063] Table 2
[0064]
[0065] As can be seen from Table 2, the drop-shaped remaining oil is mostly located at the throat with a coordination number of 35.5%, accounting for 5.8%; the cluster-shaped remaining oil is mainly distributed in large pores surrounded by small pore throats and larger pore throats; the blind-end remaining oil mostly exists in isolated large pores, connected to other pores through throats, accounting for 38.7%; the film-shaped remaining oil is mostly attached to the surface of rock particles, accounting for 20%.
[0066] As a specific embodiment of the present invention, a residual oil microscopic morphology classification device based on CT scanning and machine learning is also provided, and its schematic diagram is shown as follows: Figure 5 shown.
[0067] The residual oil microscopic morphology classification device comprises a core CT scanning module, a residual oil area classification module, a characterization parameter calculation module and a machine learning classification module which are connected in sequence.
[0068] The residual oil microscopic morphology classification device also includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the residual oil microscopic morphology classification based on machine learning is realized.
[0069] The residual oil micro-morphology division device performs the above-mentioned residual oil micro-morphology division method based on CT scanning and machine learning; the core CT scanning module 1 is implemented according to the method described in the above step S1, and is configured to perform oil-water displacement experiments on the core to be analyzed, and the size specifications of the core sample are mechanically processed and processed, and cut to a size suitable for CT scanning. The core containing residual oil is scanned by CT, and a three-dimensional digital core model is established through image denoising, enhancement, filtering and other processing; the residual oil area division module 2 is implemented according to the above step S2, and is configured to use the watershed method to segment the core image obtained in module 1, regard the pixel gray value of the input image as the altitude, and calculate the gradient amplitude to divide the residual oil area. The characterization parameter calculation module 3 is implemented according to the above step S3, and is configured to calculate the residual oil occurrence characteristic characterization parameters such as the radius, surface area, shape factor, coordination number, etc. of each residual oil area in the core of the Shengli Island Oilfield for the image segmentation result in module 2, and form a data set. The machine learning classification module 4 is configured to use a machine learning method based on a self-organizing map neural network to import the generated residual oil microscopic morphology characterization data set into the input layer and set the competition rate. Through the learning operation of the competition layer, the residual oil occurrence characteristic characterization data set is clustered and analyzed, and finally the division result of the residual oil microscopic morphology is obtained.
[0070] In summary, the residual oil micromorphology classification method based on CT scanning and machine learning provided by the present invention adopts a CT scanning method to obtain a three-dimensional image of a core containing residual oil, and the established three-dimensional digital core model containing residual oil is completely real and highly accurate; a machine learning method based on a self-organizing map neural network is adopted to classify the residual oil micromorphology, which is automatically completed by a machine without manual intervention, has high classification efficiency and good classification effect, and is suitable for large-scale promotion and application.
[0071] The applicant declares that the above is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Those skilled in the art should understand that any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention shall fall within the protection scope and disclosure scope of the present invention.
Claims
1. A method for dividing residual oil microstructure based on CT scanning and machine learning, characterized in that: The residual oil microscopic morphology classification method comprises the following steps: S1: Perform CT scanning on the core sample to establish a three-dimensional digital core model containing residual oil; S2: using a watershed method to perform image segmentation on the three-dimensional digital core model to divide the remaining oil area; S3: calculating the microscopic occurrence morphology characterization parameters of the residual oil according to the residual oil area, and forming a data set; the microscopic occurrence morphology characterization parameters of the residual oil include radius, surface area, shape factor and coordination number; S4: Using machine learning methods, cluster analysis is performed on the data set to obtain the microscopic morphology division results of the remaining oil.
2. The method for classifying residual oil microstructure according to claim 1, characterized in that: Step S1: Before CT scanning the core sample, an oil-water drive experiment is performed. When no water comes out from the outlet, water is used to drive the core sample to a water content of 98% at the outlet. After that, the size of the core sample is processed to obtain a size suitable for CT scanning.
3. The residual oil microscopic morphology classification method according to claim 1 or 2, characterized in that: After the CT scan in step S1 obtains the three-dimensional information of the entire core sample, the three-dimensional volume data of the pore structure is obtained through image denoising, enhancement and filtering, and then a three-dimensional digital core model containing residual oil is established; Preferably, according to the three-dimensional digital core model containing the remaining oil, it is concluded that the remaining oil includes four microscopic occurrence types: drop-shaped, cluster-shaped, blind-end-shaped and film-shaped.
4. The method for classifying residual oil microstructure according to any one of claims 1 to 3, characterized in that: The method of performing image segmentation on the three-dimensional digital core model using the watershed method in step S2 includes: The grayscale value of the pixel in the input image is regarded as the altitude, and then the gradient amplitude is calculated to divide the remaining oil area; Preferably, the gradient amplitude is the degree of difference between pixel gray levels.
5. The method for classifying residual oil microstructure according to any one of claims 1 to 4, characterized in that: The radius in step S3 is the radius of the residual oil area, which is the radius of the inscribed sphere and is denoted as R; Preferably, the surface area is the surface area of a sphere inscribed within the radius of the residual oil region, denoted by S.
6. The method for classifying residual oil microstructure according to any one of claims 1 to 5, characterized in that: The shape factor in step S3 is denoted as G, and its calculation formula is: Where A is the cross-sectional area of the remaining oil, and P is the perimeter of the cross-sectional graph of the remaining oil; Preferably, the coordination number is the number of connections between the residual oil and the pores or throats, denoted by C.
7. The method for classifying residual oil microstructure according to any one of claims 1 to 6, characterized in that: The machine learning method in step S4 includes a machine learning method based on a self-organizing map neural network.
8. The method for classifying residual oil microstructure according to any one of claims 1 to 7, characterized in that: The machine learning method based on the self-organizing map neural network has an input layer and a competition layer. The data set is imported into the input layer and the competition rate is set. The residual oil microscopic morphology division result is obtained through the learning operation of the competition layer.
9. The method for classifying residual oil microstructure according to any one of claims 1 to 8, characterized in that: The residual oil microscopic morphology classification method comprises the following steps: S1: Carry out oil-water driving experiment on core samples. When no water comes out from the outlet, water is used to drive the core outlet to 98% water content. Then, the size of the core samples is processed to obtain a size suitable for CT scanning, and then CT scanning is performed to obtain three-dimensional information of the entire core sample. The three-dimensional volume data of the pore structure is obtained through image denoising, enhancement and filtering, and then a three-dimensional digital core model containing residual oil is established. According to the three-dimensional digital core model containing residual oil, it is concluded that the residual oil includes four microscopic occurrence types: droplet-shaped, cluster-shaped, blind-end-shaped and film-shaped. S2: using the watershed method to perform image segmentation on the three-dimensional digital core model, treating the grayscale value of the pixel point of the input image as the altitude, and then calculating the gradient amplitude to divide the remaining oil area; The gradient amplitude is the degree of difference between pixel gray levels; S3: calculating the microscopic occurrence morphology characterization parameters of the residual oil according to the residual oil area, and forming a data set; the microscopic occurrence morphology characterization parameters of the residual oil include radius, surface area, shape factor and coordination number; The radius is the radius of the residual oil area, which is the radius of the inscribed sphere, recorded as R; the surface area is the surface area of the inscribed sphere, recorded as S; the shape factor is recorded as G, and its calculation formula is: Wherein A is the cross-sectional area of the residual oil, P is the perimeter of the cross-sectional graph of the residual oil; the coordination number is the number of connections between the residual oil and the pores or throats, denoted by C; S4: A machine learning method based on a self-organizing map neural network with an input layer and a competition layer is used to import the data set into the input layer and set the competition rate. Through the learning operation of the competition layer, the microscopic morphology division result of the residual oil is obtained.
10. A residual oil microscopic morphology classification device based on CT scanning and machine learning, characterized in that: The residual oil microscopic morphology classification device performs the residual oil microscopic morphology classification method based on CT scanning and machine learning as described in any one of claims 1 to 9; The residual oil microscopic morphology classification device comprises a core CT scanning module, a residual oil area classification module, a characterization parameter calculation module and a machine learning classification module which are connected in sequence.
Citation Information
Cited By
Deep learning quantitative characterization method and system for oiliness of CT (Computed Tomography) scanning oil layer
CN120318439A
