Shale total organic carbon content logging prediction method and device based on random forest algorithm

Abstract

The invention provides a shale total organic carbon content logging prediction method and device based on a random forest algorithm. The method comprises the steps that logging curve data are collected and processed; constructing a TOC prediction model according to the collected logging curve data by using a random forest algorithm; training the prediction model by using a training data set of TOC values; and predicting the shale TOC by using the trained prediction model. According to the method, the mud shale TOC prediction precision is improved, the influence of human factors is eliminated, rapid processing and comprehensive analysis of a large amount of logging data can be achieved, key information related to TOC is extracted, the defects of a linear or nonlinear model and a single machine learning algorithm model are overcome, the mud shale TOC prediction precision is effectively improved, and the prediction accuracy of the mud shale TOC is improved. Powerful support is provided for exploration and development of shale oil and gas.

Description

Technical Field

[0001] The present invention belongs to the field of petroleum exploration, and in particular relates to a method and device for predicting the total organic carbon content of shale using a random forest algorithm. Background Art

[0002] The total organic carbon (TOC) content of shale is a key indicator of its organic matter abundance and plays a significant role in controlling the generation and accumulation of shale oil and gas. With the rapid development of unconventional oil and gas, shale oil and gas exploration has garnered widespread attention. While the TOC content of oil and gas-bearing shales in the United States typically ranges from 1.5% to 25%, the TOC content of Chinese shales has a wider distribution, ranging from 0.2% to 30%, though most have low abundances. Due to the complexity of geological conditions, accurate identification of TOC in shale is crucial for the exploration and development of shale oil and gas.

[0003] Chinese patent CN109543915B discloses a method for identifying the total organic carbon content of source rocks throughout a well section based on well logging information. By processing well logging and total organic carbon data and removing outliers, a regression equation is established between TOC and logging parameters to predict the total organic carbon content of source rocks throughout the well section. This method improves prediction accuracy compared to currently used methods. However, it requires multiple attribute modifications and model adjustments, resulting in limited generalizability.

[0004] Shang Yazhou et al., in their article "Research on TOC Logging Evaluation Method Based on PCA-BP Neural Network," published in the 2024 issue of Well Logging Technology, state that the current application of neural network methods is mostly limited to the method itself, without in-depth analysis and processing of logging parameters before applying the neural network. This makes it difficult for neural networks to achieve ideal prediction results when multiple logging curves are collinear. To address the shortcomings of neural networks in TOC prediction and improve TOC logging evaluation accuracy, this study used experimental analysis data and conventional logging data for the gray-dark gray mudstone of the Late Triassic Baijiantan Formation in the Mahu Sag of the Junggar Basin. Based on the logging curve response mechanism and principal component analysis, the authors eliminated the complex collinearity between multiple logging curves caused by measurement errors and wellbore irregularities. Combined with the highly adaptable and fault-tolerant BP neural network, they established a PCA-BP neural network TOC intelligent prediction model, improving TOC logging evaluation accuracy. However, this method is a single machine learning algorithm, and prediction accuracy needs to be further improved. Summary of the Invention

[0005] The embodiment of the present application provides a method and device for predicting the total organic carbon content in mud shale based on a random forest algorithm. Based on the preprocessing of original conventional logging data and core retrieval, the random forest algorithm is used to quickly process and comprehensively analyze a large amount of logging data, screen out sensitive logging curves related to TOC, and construct multiple decision trees through random sampling and random feature selection to construct a TOC prediction model, which significantly improves the stability and accuracy of the model, greatly improves the accuracy of mud shale TOC prediction, and provides strong support for the exploration and development of shale oil and gas.

[0006] In a first aspect, an embodiment of the present application provides a method for predicting the total organic carbon content of shale by well logging based on a random forest algorithm, comprising:

[0007] Well logging data are collected and processed; a TOC prediction model is constructed based on the collected well logging data using a random forest algorithm; the prediction model is trained using a training data set of TOC values; and the TOC of shale is predicted using the trained prediction model.

[0008] Among them, collecting and processing logging curve data includes: collecting multiple logging curve response value data; cleaning and preprocessing the collected data, the preprocessing includes removing outliers and noise, interpolating missing data, and normalizing the data; dividing the data set into training sets and test sets; the logging curves include density curves, U content curves, natural gamma curves, acoustic time difference curves, and resistivity curves.

[0009] The collecting and processing of logging curve data includes: performing correlation analysis on the logging curve response value and TOC using a statistical method, and screening out logging curves with absolute values ​​of correlation coefficients greater than a threshold.

[0010] The collection and processing of well logging curve data includes: using the feature importance evaluation function in the random forest algorithm to evaluate the contribution of the selected well logging curves to the prediction of TOC, and optimizing the feature set based on the feature importance evaluation results.

[0011] Among them, using the random forest algorithm, based on the collected logging curve data, constructing a TOC prediction model includes: using the random forest algorithm, taking the screened logging curve response values ​​as input features, and TOC as the output target, constructing a random forest prediction model, and adjusting the parameters of the random forest algorithm to optimize the prediction model performance. The parameters include the number of decision trees, the maximum depth, and the minimum number of samples.

[0012] Among them, training the prediction model using a training data set of TOC values ​​includes: training the prediction model using a training data set of known TOC values, evaluating the prediction accuracy and stability of the model through a validation set, and adjusting the model parameters according to the validation results until a prediction effect that meets the requirements is achieved.

[0013] The method of predicting shale TOC using the trained prediction model includes predicting new well logging data using the trained model to obtain the TOC of the shale reservoir.

[0014] In a second aspect, the present application provides a well logging prediction device for total organic carbon content in shale based on a random forest algorithm, comprising:

[0015] A collection unit, used for collecting and processing logging curve data;

[0016] A construction unit for constructing a TOC prediction model based on the collected well logging curve data using a random forest algorithm;

[0017] A training unit, configured to train the prediction model using a training data set of TOC values;

[0018] The prediction unit is used to predict shale TOC using the trained prediction model.

[0019] In a third aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any of the above methods when executed by a processor.

[0020] In a fourth aspect, an embodiment of the present application provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any of the above methods when executing the program.

[0021] The method and device for predicting the total organic carbon content of shale by logging based on the random forest algorithm in the embodiment of the present application have the following beneficial effects:

[0022] This application makes full use of the advantages of conventional logging data and random forest algorithms, can realize the rapid processing and analysis of large amounts of logging data, extract key information related to TOC, and accurately predict TOC, providing strong support for the exploration and development of shale oil and gas. This method improves the accuracy of shale TOC prediction, eliminates the influence of human factors, can realize the rapid processing and comprehensive analysis of large amounts of logging data, extract key information related to TOC, avoids the shortcomings of linear or nonlinear models and single machine learning algorithm models, and effectively improves the prediction accuracy of shale TOC, with the correlation coefficient R 2Close to 1; this prediction method not only improves the prediction accuracy, but also greatly shortens the prediction cycle and reduces exploration costs; the prediction results provide an important basis for determining the "sweet spot" area, optimizing the "sweet spot" layer and the horizontal well target window, and assist in the successful drilling of shale oil horizontal wells to meet the needs of exploration and production. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of a method for predicting total organic carbon content in shale using well logging based on a random forest algorithm according to an embodiment of the present application;

[0024] Figure 2 This is a flow chart of the principle of the random forest algorithm in this application embodiment;

[0025] Figure 3 This is a schematic diagram of core return in an embodiment of the present application;

[0026] Figure 4 This is a heat map of the correlation analysis between shale logging parameters and TOC in the embodiment of this application;

[0027] Figure 5 This is a feature importance diagram of the random forest model for shale TOC identification in an embodiment of the present application;

[0028] Figure 6 This is a prediction diagram of the shale TOC identification test data in the embodiment of this application;

[0029] Figure 7 This is a flow chart for implementing a method for predicting the total organic carbon content in shale using a random forest algorithm according to an embodiment of the present application;

[0030] Figure 8 This is a schematic diagram of the structure of a well logging prediction device for total organic carbon content in shale based on a random forest algorithm in an embodiment of the present application. DETAILED DESCRIPTION

[0031] The present application will be further described below with reference to the accompanying drawings and embodiments.

[0032] In the following description, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance. The following description provides multiple embodiments of the present invention, and different embodiments can be replaced or combined, so this application can also be considered to include all possible combinations of the same and / or different embodiments described. Therefore, if one embodiment includes features A, B, and C, and another embodiment includes features B and D, then this application should also be considered to include embodiments containing one or more of all other possible combinations of features A, B, C, and D, even though such embodiments may not be explicitly described in the following text.

[0033] Example 1

[0034] like Figure 1 As shown, an embodiment of the present application provides a method for predicting the total organic carbon content of mud shale by logging based on a random forest algorithm, including: S101, collecting and processing logging curve data; S103, using a random forest algorithm to construct a TOC prediction model based on the collected logging curve data; S105, using a training data set of TOC values ​​to train the prediction model; S107, using the trained prediction model to predict the TOC of mud shale.

[0035] This application fully utilizes the advantages of conventional logging data and random forest algorithm, can realize the rapid processing and analysis of large amounts of logging data, extract key information related to TOC, realize accurate prediction of shale TOC, and provide strong support for the exploration and development of shale oil and gas.

[0036] Example 2

[0037] The method for predicting the total organic carbon content of shale using well logging based on the random forest algorithm in the embodiment of the present application includes:

[0038] Step 1: Collect various well logging response data, including but not limited to density curves (DEN), U content curves (U), natural gamma ray curves (GR), acoustic transit time curves (AC), and resistivity curves (RT). These curves can reflect the physical and chemical properties of shale reservoirs. Clean and preprocess the collected data to remove outliers and noise to ensure data quality. Interpolate missing data to ensure data integrity and continuity. Normalize the data and split the dataset into training and test sets. For example, the ratio is 70% training set + 30% test set. During the split, each sample's characteristics (logging parameters) must correspond to the target variable (TOC). The order of the dataset should be randomly shuffled (see Appendix 1) to avoid depth dependence. Normalize the well logging data to eliminate dimensional differences between different well logging curves.

[0039] Step 2: Use statistical methods to analyze the correlation between the well log response values ​​and TOC, and select well logs with strong correlation with TOC. The purpose of this step is to determine which well logs have a significant impact on TOC prediction, thereby optimizing the input feature set.

[0040] Step 3: Based on the results of the correlation analysis, well logs with strong correlation are selected as input features. At the same time, the feature set is further optimized to improve the model's predictive performance by considering the redundancy and interaction between features.

[0041] Step 4: If Figure 2As shown in the figure, a prediction model was constructed using the random forest algorithm, using the selected well logging curve response values ​​as input features and TOC as the output target. During the model construction process, parameters of the random forest algorithm, such as the number of decision trees, maximum depth, and minimum number of samples, were adjusted to optimize model performance. The random forest algorithm is a decision tree-based ensemble learning method that improves model stability and accuracy by constructing multiple decision trees and integrating their prediction results. It excels in handling classification and regression problems and is particularly suitable for processing large, high-dimensional datasets. This method fully leverages the advantages of machine learning algorithms and has strong generalization capabilities. It can effectively improve the accuracy of the TOC prediction model based on well logging information, shorten the prediction cycle, and provide strong support for shale oil and gas exploration and development.

[0042] Step 5: Use a training dataset with known TOC values ​​to train the model and evaluate its prediction accuracy and stability through methods such as cross-validation. Adjust the model parameters based on the validation results until satisfactory prediction results are achieved.

[0043] Step 6: Use the trained model to predict the TOC of the new dataset and obtain the TOC of the shale reservoir. Compare and analyze the predicted results with the actual situation to verify the reliability and accuracy of the model. Interpret and evaluate the predicted results to provide a scientific basis for shale oil and gas exploration and development.

[0044] This application collects and preprocesses conventional logging curve data, performs correlation analysis, selects and optimizes features, uses the random forest algorithm to screen sensitive logging curves, and constructs a TOC prediction model. On this basis, the model is trained using a training data set with known TOC values, and the trained model is used to predict new data sets, thereby realizing the prediction of shale TOC using conventional logging parameters.

[0045] Based on the preprocessing of original conventional logging data and core retrieval, this application uses the random forest algorithm to quickly process and comprehensively analyze a large amount of logging data, screen out sensitive logging curves related to TOC, and construct multiple decision trees through random sampling and random feature selection to build a TOC prediction model, which can significantly improve the stability and accuracy of the model and provide strong support for the exploration and development of shale oil and gas.

[0046] Example 3

[0047] Take the HY1 well in a certain sag in a certain basin as an example. Figure 7 As shown, the present application's method for predicting total organic carbon content in shale using well logging based on the random forest algorithm includes:

[0048] Step 1: Taking a basin that has achieved a major shale oil exploration breakthrough as an example, we collected logging data from multiple typical wells in the block, including curve types such as CAL, SP, R25, AC, CNL, DEN, GR, and U. At the same time, we collected core analysis data from these wells, including laboratory values ​​of TOC, as training data and validation data for the model (Appendix Table 1):

[0049] Appendix 1 Characteristic parameters of learning samples and laboratory measured TOC

[0050]

[0051]

[0052] Clean the collected logging data to remove outliers and noise. Standardize the logging data to eliminate the dimensional differences between different logging curves and improve the training efficiency of the model. Figure 3 As shown in the figure, if there is a problem that the depth of the logging curve does not match the core depth, the core must be returned first.

[0053] The commonly used normalization formula is:

[0054]

[0055] Where X is the original data, X min is the minimum value of the feature, X max is the maximum value of the feature, X norm is the normalized data.

[0056] Step 2: If Figure 4 As shown, the correlation coefficient between the response value of each logging curve and TOC is calculated, and the logging curves with strong correlation with TOC (correlation coefficient greater than a threshold) are screened out according to the absolute value of the correlation coefficient.

[0057] Step 3: Consider the above correlation analysis results and use the feature importance evaluation function in the random forest algorithm, such as Figure 5 As shown in Figure 3, the contribution of the selected well logging curves to TOC prediction is further evaluated. Based on the feature importance evaluation results, the feature set can be optimized to improve the prediction performance of the model.

[0058] Step 4: Refer to Appendix 2 to determine the key parameters of the random forest algorithm, such as the number of decision trees, maximum depth, and minimum number of samples. Use the selected well logging curve response values ​​as input features and TOC as the output target to construct a random forest prediction model.

[0059] Step 5: Figure 6As shown in Figure 1, the collected data is divided into a training set and a validation set, where the training set is used to train the model and the validation set is used to evaluate the prediction accuracy and stability of the model. The training set data is used to train the random forest model and adjust the model parameters to optimize the prediction performance. The validation set data is used to verify the trained model and evaluate the prediction accuracy of the model (such as mean square error MSE, determination coefficient R 2 wait).

[0060] Appendix 2 Random forest model parameters for shale TOC identification

[0061]

[0062] Appendix 3 Evaluation results of random forest model for shale TOC identification

[0063]

[0064] The above table shows the prediction evaluation indicators of the training set and test set, and uses quantitative indicators to measure the prediction effect of random forest.

[0065] MSE (mean squared error): The expected value of the square of the difference between the predicted value and the actual value. The smaller the value, the more accurate the model.

[0066] RMSE (Root Mean Square Error): It is the square root of MSE. The smaller the value, the higher the model accuracy.

[0067] MAE (Mean Absolute Error): The average of the absolute errors, which reflects the actual situation of the prediction error. The smaller the value, the higher the model accuracy.

[0068] MAPE (Mean Absolute Percentage Error): A variation of MAE, it is a percentage value. The smaller the value, the higher the model accuracy.

[0069] R 2 : Compare the predicted value with the case of using only the mean. The closer the result is to 1, the higher the accuracy of the model.

[0070] Step 6: Use the trained random forest model to predict the new logging data and obtain the TOC of the shale reservoir (TOC). Compare the prediction results with the actual situation to verify the reliability and accuracy of the model. In this example, by comparing the predicted values ​​with the laboratory measured values, it is found that the prediction accuracy of the model is high, the mean square error (MSE) is small, and the determination coefficient R is high. 2 A value close to 1 indicates that the model has good prediction performance. Based on the prediction results, a scientific basis and decision support are provided for the exploration and development of shale oil and gas.

[0071] The method of this application improves the accuracy of shale TOC prediction, eliminates the influence of human factors, can realize the rapid processing and comprehensive analysis of a large amount of logging data, extracts key information related to TOC, avoids the shortcomings of linear or nonlinear models and single machine learning algorithm models, and effectively improves the prediction accuracy of shale TOC. The correlation coefficient R 2 Close to 1; This prediction method not only improves the prediction accuracy, but also greatly shortens the prediction cycle and reduces exploration costs; The prediction results provide an important basis for determining the "sweet spot" area, optimizing the "sweet spot" layer and the horizontal well target window, and help shale oil horizontal wells to be successfully drilled and meet the needs of exploration and production. This application method has been applied to 11 wells in the XX Basin, with a cumulative oil and gas production of more than 14×10 4 t, the confirmed geological resources of shale oil exceed 10×10 8 t.

[0072] Example 4

[0073] like Figure 8 As shown, the present application's mud shale total organic carbon content logging prediction device based on the random forest algorithm includes: a collection unit 201, used to collect and process logging curve data; a construction unit 202, used to use the random forest algorithm to construct a TOC prediction model based on the collected logging curve data; a training unit 203, used to train the prediction model using a training data set of TOC values; and a prediction unit 204, used to predict mud shale TOC using the trained prediction model.

[0074] In this application, the embodiment of the mud shale total organic carbon content logging prediction device based on the random forest algorithm is basically similar to the embodiment of the mud shale total organic carbon content logging prediction method based on the random forest algorithm. For relevant matters, please refer to the introduction of the mud shale total organic carbon content logging prediction method based on the random forest algorithm.

[0075] The present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, any one of the above-mentioned steps of the method for predicting the total organic carbon content of shale by logging based on the random forest algorithm is implemented.

[0076] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the steps of the above-described method for predicting total organic carbon content in shale using well logging based on a random forest algorithm. The computer-readable storage medium may include, but is not limited to, any type of disk, including a floppy disk, an optical disk, a DVD, a CD-ROM, a microdrive, a magneto-optical disk, a ROM, a RAM, an EPROM, an EEPROM, a DRAM, a VRAM, a flash memory device, a magnetic card or an optical card, a nanosystem (including a molecular memory IC), or any type of medium or device suitable for storing instructions and / or data.

[0077] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for predicting total organic carbon content in shale using well logging based on random forest algorithm, characterized in that: include: Collect and process well logging data; A TOC prediction model is constructed based on the collected well logging data using a random forest algorithm; the prediction model is trained using a training data set of TOC values; and the trained prediction model is used to predict shale TOC.

2. The method for predicting total organic carbon content in shale using well logging based on random forest algorithm according to claim 1, characterized in that: Collecting and processing logging curve data includes: collecting multiple logging curve response value data; cleaning and preprocessing the collected data, wherein the preprocessing includes removing outliers and noise, interpolating missing data, and normalizing the data; and dividing the data set into a training set and a test set; the logging curves include density curves, U content curves, natural gamma curves, acoustic time difference curves, and resistivity curves.

3. The method for predicting total organic carbon content in shale by logging based on random forest algorithm according to claim 1 or 2, characterized in that: Collecting and processing the logging curve data includes: using statistical methods to perform correlation analysis on the logging curve response value and TOC, and screening out the logging curves whose absolute value of the correlation coefficient is greater than a threshold.

4. The method for predicting total organic carbon content in shale using well logging based on random forest algorithm according to claim 3, characterized in that: Collecting and processing well logging curve data includes: using the feature importance evaluation function in the random forest algorithm to evaluate the contribution of the screened well logging curves to TOC prediction, and optimizing the feature set based on the feature importance evaluation results.

5. The method for predicting total organic carbon content in shale by logging based on random forest algorithm according to claim 1 or 2, characterized in that: Using the random forest algorithm, a TOC prediction model is constructed based on the collected logging curve data, including: using the random forest algorithm, taking the screened logging curve response values ​​as input features and TOC as the output target, constructing the random forest prediction model, and adjusting the parameters of the random forest algorithm to optimize the prediction model performance. The parameters include the number of decision trees, the maximum depth, and the minimum number of samples.

6. The method for predicting total organic carbon content in shale by well logging based on random forest algorithm according to claim 1 or 2, characterized in that: Training the prediction model using a training data set of TOC values ​​includes: training the prediction model using a training data set of known TOC values, evaluating the prediction accuracy and stability of the model through a validation set, and adjusting the model parameters according to the validation results until a prediction effect that meets the requirements is achieved.

7. The method for predicting total organic carbon content in shale by well logging based on random forest algorithm according to claim 1 or 2, characterized in that: Predicting shale TOC using the trained prediction model includes: predicting new well logging data using the trained model to obtain the TOC of the shale reservoir.

8. A device for predicting total organic carbon content in shale using a random forest algorithm, characterized in that: include: A collection unit, used for collecting and processing logging curve data; A construction unit, for constructing a TOC prediction model based on the collected well logging curve data using a random forest algorithm; A training unit, configured to train the prediction model using a training data set of TOC values; The prediction unit is used to predict shale TOC using the trained prediction model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images