Shale lithofacies intelligent identification method based on random forest algorithm
Through the random forest algorithm combined with core and logging data, the problem of low accuracy in shale stale facies characterization in traditional methods is solved, and high-precision lithophagocytic recognition is achieved, with the recognition rate reaching 78% to 85%.
Patent Information
- Application Number
- CN202310584694.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-23
- Publication Date
- 2024-11-26
AI Technical Summary
Traditional methods fail to fully consider the similarity interference between well logging data and the scale differences in core experimental data in lithophagos in organic-rich shale, resulting in low accuracy of lithophagos prediction, blurred boundaries and large identification deviation.
The shale quadrilateral intelligent recognition method based on random forest algorithm is adopted, and typical lithophagometric marks are determined through core data, combined with well logging curves for preprocessing and classification, and the lithophagometric recognition is used using random forest algorithm, which is divided into organic matter rich, organic matter and poor Organic lithophagocytic.
It improves the accuracy and accuracy of lithophagometric recognition, and can obtain better training and learning results under a small amount of data, with a recognition rate of 78% to 85%.
Smart Images

Figure CN119025952A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of oil and gas geophysical seismic data interpretation, and in particular to a shale lithofacies intelligent identification method based on a random forest algorithm. Background Art
[0002] Organic-rich shale is an important unconventional oil and gas resource. Lithofacies characterization is an important basic research before shale reservoir evaluation. The research method and precision directly affect the evaluation of shale oil reservoir quality and the formulation of development plans.
[0003] Organic-rich shales are characterized by multiple lithofacies types and rapid changes. The method of establishing lithology maps based on traditional conventional logging curves does not fully consider the interference caused by the similarity between logging data and the difference in scale with core experimental data, resulting in overlapping of sample points of different categories in the established maps, blurred boundaries, limited identification accuracy, and large deviations in lithofacies prediction. Summary of the invention
[0004] In view of the above problems, the present invention is proposed to provide a shale lithofacies intelligent identification method based on a random forest algorithm to overcome the above problems or at least partially solve the above problems.
[0005] According to one aspect of the present invention, a shale lithofacies intelligent identification method based on a random forest algorithm is provided, the identification method comprising:
[0006] Based on the core data, the typical lithofacies marks of the whole area are determined, the shale lithofacies are finely divided, and the classification standards are determined;
[0007] According to the classification standard, the lithofacies are divided into organic-rich lithofacies, organic-matter lithofacies, and organic-poor lithofacies;
[0008] Combine the logging curve data to establish the logging phase corresponding to the rock;
[0009] Preprocessing the well logging curve to obtain a smooth well logging curve;
[0010] Using the smoothed logging curve and the corresponding lithofacies as a sample set;
[0011] The sample set is subjected to lithofacies intelligent identification based on random forest algorithm to complete the identification of different lithofacies in the well to be predicted.
[0012] Optionally, the fine division of shale lithofacies specifically includes:
[0013] The shale lithofacies are finely divided into four aspects: total organic carbon content TOC, rock structure, composition characteristics and fractures.
[0014] Optionally, the total organic carbon content TOC classification method includes:
[0015] The classification is based on the organic matter content, and the organic matter content can be judged by color. Rocks rich in organic matter are mainly dark gray and gray-black, those containing organic matter are mainly gray-brown, and those poor or free of organic matter are mainly light gray.
[0016] Optionally, the rock structure division method includes: dividing the rock structure into blocky, laminar, layered and thin-layered shapes.
[0017] Optionally, the component characteristic classification method includes: according to the main component characteristics, mud, gray and sand are taken as three end members, and divided into three categories with their respective contents of 40% as the boundary.
[0018] Optionally, the fracture classification method includes: providing a classification evaluation according to the degree of fracture development.
[0019] Optionally, the preprocessing of the logging curve to obtain a smooth logging curve specifically includes: using a window function to smooth the curve, removing abnormal jitter and burrs, and ensuring that the curve is relatively smooth to avoid erroneous sample learning.
[0020] Optionally, the step of using the sample set to perform lithofacies intelligent identification based on a random forest algorithm to complete the identification of different lithofacies in the well to be predicted specifically includes:
[0021] Perform a certain number of random cuts on the sample set to divide it into two parts, one part is a sample set used for training, and the other part is a sample set used for testing, which is used to correct the results of the training sample set;
[0022] The decision tree model has the same number of training sample sets as the training set. For each decision tree model, the number of features is set. At each split, the best feature is selected for splitting based on the parameters on the test set.
[0023] The decision trees generated by the training set constitute the random forest algorithm model;
[0024] The final classification result is determined by voting among multiple tree classifiers;
[0025] Correct the final classified results according to the test sample set;
[0026] The logging curves of the wells to be predicted are entered into the model optimized by the trained random forest algorithm to complete the identification of different lithofacies in the wells to be predicted.
[0027] Optionally, the parameters on the test set specifically include: information gain, information gain ratio, and Gini coefficient.
[0028] Optionally, the correcting the final classified result according to the test sample set specifically includes: determining the final prediction result according to the mean of the prediction values of multiple trees.
[0029] The present invention provides a shale lithofacies intelligent identification method based on a random forest algorithm, the identification method comprising: based on core data, determining typical lithofacies marks of the entire area, finely dividing the shale lithofacies, and determining classification standards; dividing the lithofacies into organic-rich lithofacies, organic lithofacies, and organic-poor lithofacies according to the classification standards; establishing a logging phase corresponding to the shale in combination with well logging curve data; preprocessing the well logging curve to obtain a smooth logging curve; using the smooth logging curve and the corresponding lithofacies as a sample set; using the sample set for lithofacies intelligent identification based on a random forest algorithm to complete the identification of different lithofacies in the well to be predicted. Using a small amount of lithofacies data as a training set can achieve a better training and learning effect, thereby improving the recognition accuracy.
[0030] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.
[0032] Figure 1 A flow chart of a shale lithofacies intelligent identification method based on a random forest algorithm provided in an embodiment of the present invention;
[0033] Figure 2 A schematic diagram of lithofacies classification provided by an embodiment of the present invention;
[0034] Figure 3 A schematic diagram of well logging phases provided by an embodiment of the present invention;
[0035] Figure 4 A schematic diagram of a well logging curve before and after smoothing provided by an embodiment of the present invention;
[0036] Figure 5 A schematic diagram of the correspondence between logging phases and lithophases provided in an embodiment of the present invention;
[0037] Figure 6 A schematic diagram of sample set segmentation provided by an embodiment of the present invention;
[0038] Figure 7 A schematic diagram of a random forest algorithm provided by an embodiment of the present invention;
[0039] Figure 8 The random forest algorithm for identifying shale lithofacies in Work Area A provided by the embodiment of the present invention has a comprehensive recognition rate of 78%;
[0040] Fig. 9 The random forest algorithm for work area B provided in the embodiment of the present invention has an overall recognition rate of 85% for shale lithofacies recognition. DETAILED DESCRIPTION
[0041] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.
[0042] The terms "comprises" and "having" and any variations thereof in the description embodiments, claims and drawings of the present invention are intended to cover non-exclusive inclusions, for example, including a series of steps or units.
[0043] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.
[0044] like Figure 1 As shown, the shale lithofacies intelligent identification method based on the random forest algorithm includes the following steps:
[0045] Step 1: Based on the core data, determine the typical lithofacies markers of the entire area, and finely divide the shale lithofacies from four aspects: total organic carbon content TOC, rock structure, composition characteristics and fractures, such as Figure 2 shown.
[0046] First, the classification is based on the organic matter content. The organic matter content is judged by color. The organic matter-rich is mainly dark gray and gray-black, the organic matter-containing is mainly gray-brown, and the poor or no organic matter is mainly light gray. Then, according to the rock structure, it is divided into massive, laminar, layered and thin-layered. Then, according to the main component characteristics, mud, gray and sand are used as the three end members, and divided into three categories with 40% of their respective contents as the boundary. Finally, the classification evaluation is given considering the degree of crack development.
[0047] Step 2: Based on the above classification standards, the lithofacies are divided into three categories: organic-rich lithofacies, organic lithofacies, and organic-poor lithofacies.
[0048] The cores of organic-rich lithofacies are mainly gray and gray-black, with a high degree of foliation. The main components are gray and mud. The structure is mainly laminated. There are also small cracks on the foliation surface. The total organic carbon content is high, and oil film can be seen. The cores of organic-containing lithofacies are mainly gray-brown and brown-brown, with relatively developed foliation. The main components are gray and mud. The structure is mainly layered and thin-layered. Dissolution pores, faults, high-angle sections and reticular fractures are all developed. The fracture fillings are mud and cryptocrystalline dolomite. Oil film can be observed, and the total organic carbon content TOC content is medium and high. The cores of organic-poor lithofacies are mainly light gray and gray, with a low degree of foliation. The main components are gray and mud. The structure is mainly blocky and layered. There are a small number of parallel fractures and low-angle fractures, and the total organic carbon content TOC content is low.
[0049] Step 3: Combine the logging curve data to establish the logging phase corresponding to the rock, such as Figure 3 As shown. The difference in total organic carbon content TOC under different algorithms is too large, and it can be evaluated in combination with core and cuttings testing methods. The GR value of the logging curve is low when the ash content is not affected by sandstone, reflecting the high content of carbonate rock and the abundance of brittle minerals. In the comparison of core and logging, the GR is more brittle when it is below 40-60API. When the lamina is well developed, the logging response characteristics are manifested as a jagged porosity curve, and a high degree of AC, RT and microelectrode toothing. The logging response characteristics of the microfracture development section are diameter expansion, high acoustic wave time difference, high neutron porosity, high resistivity, low density, and low natural gamma. Characteristics of favorable shale oil layer sections: increased acoustic waves, decreased neutrons, reduced gamma, and increased resistivity.
[0050] Step 4: Preprocess all the above logging curves, use the window function to smooth the curves, remove abnormal jitter and burrs, and ensure that the curves are relatively smooth, such as Figure 4 As shown, the wrong sample learning is avoided.
[0051] Step 5: Use these smoothed logging curves and the corresponding lithofacies as sample sets. The logging phases and lithofacies correspond to each other as shown in the following figure. Figure 5 shown.
[0052] Write the lithofacies intelligent identification code based on random forest algorithm on the Python platform.
[0053] Step 6: Based on the random forest algorithm, the sample set is first randomly cut a certain number of times, which is divided into two parts, one is the sample set used for training, and the other is the sample set used for testing, as a correction to the training sample set results, such as Figure 6 shown.
[0054] Step 7: The number of training sample sets and decision tree models is the same as that of the training set. For each decision tree model, set the number of features. Each time the split occurs, select the best feature for splitting based on the information gain, information gain ratio, and Gini coefficient on the test set. The decision trees generated by the training set constitute the random forest algorithm model.
[0055] Step 8: Voting among multiple tree classifiers determines the final classification result.
[0056] Step 9: Use the test sample set to calibrate the final classification results, and the final prediction result is determined by the mean of the prediction values of multiple trees, such as Figure 7 shown.
[0057] Step 10: Enter the logging curve of the well to be predicted into the model optimized by the trained random forest algorithm to complete the identification of different lithofacies in the well to be predicted.
[0058] Step 11: Application Examples
[0059] Example 1: Figure 8 The lithofacies identification effect map obtained by using this method for the actual data work area A has a comprehensive identification rate of 78%.
[0060] Example 2: Fig. 9 The lithofacies identification effect map obtained by using this method for the actual data work area B has a comprehensive identification rate of 85%.
[0061] Beneficial effects: On the one hand, it breaks the barrier of using third-party commercial software, and the random forest algorithm can be realized through independent programming, which is simple to operate, highly practical, and has a wide range of applications; on the other hand, the present invention can better classify shale lithofacies, and associate the well logging phase with the lithofacies, and then carry out artificial intelligence calculations based on the random forest algorithm on the well logging curve, so as to efficiently integrate artificial intelligence with lithofacies identification, which has good foresight and practicality, and can especially promote the realization of intelligent identification of lithofacies in unconventional shale reservoirs. It effectively solves the quantitative identification of lithofacies in current unconventional basic geological research, and has a good prospect for application and promotion.
[0062] The above specific implementation methods further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A shale lithofacies intelligent identification method based on random forest algorithm, characterized in that: The identification method comprises: Based on the core data, the typical lithofacies marks of the whole area are determined, the shale lithofacies are finely divided, and the classification standards are determined; According to the classification standard, the lithofacies are divided into organic-rich lithofacies, organic-matter lithofacies, and organic-poor lithofacies; Combine logging curve data to establish logging phases corresponding to the rock; Preprocessing the well logging curve to obtain a smooth well logging curve; Using the smoothed logging curve and the corresponding lithofacies as a sample set; The sample set is used for lithofacies intelligent recognition based on random forest algorithm to complete the identification of different lithofacies in the well to be predicted.
2. The shale lithofacies intelligent identification method based on random forest algorithm according to claim 1 is characterized in that: The detailed division of shale lithofacies specifically includes: The shale lithofacies are finely divided into four aspects: total organic carbon content TOC, rock structure, composition characteristics and fractures.
3. The shale lithofacies intelligent identification method based on random forest algorithm according to claim 2 is characterized in that: The total organic carbon content TOC classification method includes: The classification is based on the organic matter content, and the organic matter content can be judged by color. Rocks rich in organic matter are mainly dark gray and gray-black, those containing organic matter are mainly gray-brown, and those poor or free of organic matter are mainly light gray.
4. The shale lithofacies intelligent identification method based on random forest algorithm according to claim 2 is characterized in that: The rock structure classification method includes: classifying the rock structure into block, laminar, layered and thin layered.
5. The shale lithofacies intelligent identification method based on random forest algorithm according to claim 2 is characterized in that: The component characteristic classification method includes: according to the main component characteristics, mud, gray and sand are taken as three end members, and divided into three categories with their respective contents of 40% as the boundary.
6. The shale lithofacies intelligent identification method based on random forest algorithm according to claim 2 is characterized in that: The fracture classification method includes: providing classification evaluation according to the degree of fracture development.
7. The shale lithofacies intelligent identification method based on random forest algorithm according to claim 1 is characterized in that: The preprocessing of the logging curve to obtain a smooth logging curve specifically includes: using a window function to smooth the curve, removing abnormal jitter and burrs, and ensuring that the curve is relatively smooth to avoid erroneous sample learning.
8. The shale lithofacies intelligent identification method based on random forest algorithm according to claim 1 is characterized in that: The method of using the sample set to intelligently identify lithofacies based on a random forest algorithm to complete the identification of different lithofacies in the well to be predicted specifically includes: Perform a certain number of random cuts on the sample set to divide it into two parts, one part is a sample set used for training, and the other part is a sample set used for testing, which is used to correct the results of the training sample set; The decision tree model has the same number of training sample sets as the training set. For each decision tree model, the number of features is set. At each split, the best feature is selected for splitting based on the parameters on the test set. The decision trees generated by the training set constitute the random forest algorithm model; The final classification result is determined by voting among multiple tree classifiers; Correct the final classified results according to the test sample set; The logging curves of the wells to be predicted are entered into the model optimized by the trained random forest algorithm to complete the identification of different lithofacies in the wells to be predicted.
9. The method for intelligent identification of shale lithofacies based on random forest algorithm according to claim 8, characterized in that: The parameters on the test set specifically include: information gain, information gain ratio, and Gini coefficient.
10. The method for intelligent identification of shale lithofacies based on random forest algorithm according to claim 8, characterized in that: The correcting of the final classified result according to the test sample set specifically includes: determining the final prediction result according to the mean of the prediction values of multiple trees.
Citation Information
Patent Citations
Electrical imaging logging image stratum attitude identification method
CN104849766A
Formation occurrence automatic detection method based on directional filtering
CN110161577A
Shale gas reservoir logging porosity rapid prediction method and system
CN110717249A
Image enhancement method and system for micro-resistivity scanning logging instrument
CN112561806A
Cited By
Shale lithofacies combination seismic scale identification and plane distribution prediction method
CN120779472A
A seismic-scale identification and planar distribution prediction method for shale lithofacies assemblages
CN120779472B
Lithofacies combination identification method, apparatus and device, and storage medium
CN121051611A
