Hydrocarbon source rock comprehensive evaluation method based on data driving

Abstract

The invention discloses a hydrocarbon source rock comprehensive evaluation method based on data driving, and relates to the technical field of oil and gas field development, and the hydrocarbon source rock comprehensive evaluation method based on data driving mainly comprises the following steps: taking rock core actual measurement data as dependent variables and corresponding logging data as independent variables; establishing different output parameter prediction models through a multi-output random forest algorithm; multi-parameter prediction from a single well to a research area is achieved through the trained model, meanwhile, the organic carbon content, the free hydrocarbon content and the pyrolytic hydrocarbon content are predicted, categories are divided according to the hydrocarbon source rock evaluation national standard, a model for predicting the hydrocarbon source rock type is further established, and hydrocarbon source rock evaluation parameter prediction is achieved through first machine learning; and the second machine learning realizes hydrocarbon source rock category prediction so as to realize comprehensive evaluation of the hydrocarbon source rocks in the whole area. By means of the hydrocarbon source rock comprehensive evaluation method based on data driving, the accuracy, reliability and efficiency of hydrocarbon source rock comprehensive evaluation can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas field development, and more specifically, to a data-driven comprehensive evaluation method for source rocks. Background Art

[0002] In the field of oil exploration, source rocks play a vital role in the formation and accumulation of oil and gas reserves. It is not only the material basis for oil and gas generation, but also the core of oil and gas resource evaluation and prediction. The quality and quantity of source rocks directly determine the potential and economic feasibility of oil and gas resources. Therefore, a comprehensive evaluation of source rocks is of great significance in oil and gas exploration and development.

[0003] In the evaluation of oil and gas resources, organic carbon content (TOC-%), free hydrocarbon content (S1-mg / g) and pyrolysis hydrocarbon content (S2-mg / g) are usually obtained through pyrolysis experiments to evaluate the organic matter abundance and hydrocarbon generation potential characteristics of source rocks. However, the evaluation of source rocks mainly relies on experimental data, which is costly, time-consuming, and is greatly limited by core samples. In the past, some scholars would predict parameters such as TOC (%) and S1 (mg / g) by establishing empirical formulas such as the ΔLogR method, the improved ΔLogR method, and the multivariate regression method. Although the calculation of parameters by establishing a model through well logging can effectively make up for the problem of limited data, this method requires experts in various fields to spend a lot of time using complex logical reasoning and mathematical and physical formulas for calculation. It is a very tedious task and may have the limitations of poor portability and inability to express nonlinear relationships. With the development of science and technology, machine learning is used to predict various parameters. However, there are many parameters involved in the evaluation of source rocks, and it is cumbersome to achieve comprehensive evaluation of source rocks through a single machine learning model. Summary of the invention

[0004] The purpose of the present invention is to provide a data-driven comprehensive evaluation method for source rocks, which can simultaneously predict the organic carbon content, free hydrocarbon content and pyrolysis hydrocarbon content and further predict the type of source rocks, thereby achieving a comprehensive evaluation of source rocks and improving the accuracy, reliability and efficiency of the comprehensive evaluation of source rocks.

[0005] The present invention provides a data-driven comprehensive evaluation method for source rocks, comprising the following steps: S1: obtaining core experimental data and corresponding well logging data; obtaining a first training data set according to the core experimental data and the corresponding well logging data; S2: constructing a first prediction model using a random forest regression model; S3: training the first prediction model using the first training data set to obtain a trained first prediction model; S4: predicting single well logging data using the trained first prediction model to obtain a predicted value of organic carbon content, a predicted value of free hydrocarbon content, and a predicted value of pyrolytic hydrocarbon content; S5: predicting the organic carbon content according to the national evaluation standard for source rocks, the predicted value of organic carbon content, and the predicted value of free hydrocarbon content. , free hydrocarbon content prediction value and pyrolysis hydrocarbon content prediction value, and obtain the single well source rock category; S6: re-collect data according to the single well source rock category and the single well logging data to obtain a second training data set, use the random forest regression model to build a second prediction model, use the second training data set to train the second prediction model to obtain a trained second prediction model; S7: use the trained first prediction model and the trained second prediction model to predict the logging data to obtain the predicted organic carbon content prediction value, free hydrocarbon content prediction value, pyrolysis hydrocarbon content prediction value and single well source rock category prediction value to complete the single well and the whole area source rock evaluation.

[0006] Furthermore, the core experimental data include organic carbon content data, free hydrocarbon content data and pyrolysis hydrocarbon content data; the corresponding logging data include natural gamma data, resistance data, acoustic wave time difference data, density data and neutron data.

[0007] Furthermore, the above-mentioned first training data set includes a first data set, a second data set and a third data set; the first data set includes organic carbon content data and corresponding logging data; the second data set includes free hydrocarbon content data and corresponding logging data; the third data set includes pyrolysis hydrocarbon content data and corresponding logging data.

[0008] Furthermore, the above-mentioned first prediction model includes a first parameter prediction model, a second parameter prediction model and a third parameter prediction model.

[0009] Furthermore, the above-mentioned step S3 specifically includes: taking the organic carbon content data in the first data set, the free hydrocarbon content data in the second data set and the pyrolysis hydrocarbon content data in the third data set as dependent variables, and taking the corresponding logging data as independent variables, training and predicting the first prediction model to obtain predicted organic carbon content, predicted free hydrocarbon content and predicted pyrolysis hydrocarbon content; obtaining a first determination coefficient based on the organic carbon content data and the predicted organic carbon content in the first data set; obtaining a second determination coefficient based on the free hydrocarbon content data and the predicted free hydrocarbon content in the second data set; obtaining a third determination coefficient based on the pyrolysis hydrocarbon content data and the predicted pyrolysis hydrocarbon content in the third data set; when any one of the first determination coefficient, the second determination coefficient and the third determination coefficient is less than a preset determination coefficient threshold, continuing training and prediction; when the first determination coefficient, the second determination coefficient and the third determination coefficient are all not less than the preset determination coefficient threshold, stopping training to obtain a trained first prediction model.

[0010] Furthermore, the above preset determination coefficient threshold is 0.8.

[0011] Furthermore, the trained first prediction model includes a trained first parameter prediction model, a trained second parameter prediction model and a trained third parameter prediction model; The present invention also provides a computer program product, including a computer program, which implements the steps of the above-mentioned data-driven comprehensive evaluation method for source rocks when executed by a processor.

[0012] The implementation of the data-driven comprehensive source rock evaluation method and computer program product provided by the present invention has the following beneficial effects: The present invention uses the measured data of core (total organic carbon-TOC (%), free hydrocarbon content-S1 (mg / g), pyrolysis hydrocarbon content S2 (mg / g)) as the dependent variable , , , the corresponding logging data The training set and the test set are used as independent variables, and the prediction models with different output parameters are established through the multi-output random forest algorithm. The trained model can realize the multi-parameter prediction from a single well to the study area, and can simultaneously predict TOC (%), S1 (mg / g), and S2 (mg / g). Then, the categories are divided according to the national standards for source rock evaluation, and a model is further established to predict the type of source rock. The comprehensive evaluation of the source rock potential in the whole area is realized through the secondary machine learning single parameter prediction model. This comprehensive evaluation not only takes into account the organic matter content, but also includes the hydrocarbon generation potential of the source rock, and comprehensively analyzes the source rock from multiple dimensions. The present invention realizes simultaneous prediction and output of multiple parameters, which greatly improves efficiency compared with conventional machine learning prediction methods; the present invention makes full use of existing data and solves the problem of missing data through a data-driven method. Compared with traditional interpolation or filling methods, the present invention can not only predict missing data, but also capture the complex relationship between data, thereby improving the accuracy and reliability of prediction; the present invention has strong scalability and versatility, and can provide a reference for the evaluation of different types of source rocks. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1 It is a flow chart of the data-driven comprehensive evaluation method for source rocks provided by the present invention; Figure 2 It is a comparison chart of prediction effects provided by the present invention; Figure 3 This is a comprehensive evaluation diagram of single well source rocks provided by the present invention. DETAILED DESCRIPTION

[0014] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0015] Figure 1 A schematic diagram of a data-driven comprehensive evaluation method for source rocks in this embodiment is shown. In this embodiment, the data-driven comprehensive evaluation method for source rocks includes the following steps: S1: Acquire core test data and corresponding well logging data; obtain a first training data set based on the core test data and the corresponding well logging data; For an exemplary embodiment, the core experimental data includes organic carbon content data, free hydrocarbon content data and pyrolysis hydrocarbon content data; the corresponding logging data includes natural gamma data, resistance data, acoustic wave time difference data, density data and neutron data; For an exemplary embodiment, the first training data set includes a first data set, a second data set and a third data set; the first data set includes organic carbon content data and corresponding well logging data; the second data set includes free hydrocarbon content data and corresponding well logging data; the third data set includes pyrolysis hydrocarbon content data and corresponding well logging data; Specifically, in step S1, core test data and logging data corresponding to each quality parameter are collected; wherein the core test data include three quality parameters: organic carbon content, referred to as TOC, in units of %; free hydrocarbon content, referred to as S1, in units of mg / g; pyrolysis hydrocarbon content, referred to as S2, in units of mg / g; logging data include: depth; natural gamma, referred to as GR, in units of API; electrical resistance, referred to as RT, in units of Ω·m; acoustic time difference, referred to as AC, in units of us / m; density, referred to as DEN, in units of g / cm 3 ; Neutron, referred to as CNL, unit is %; sort the above data, check whether the data type parameters are missing, if a parameter is missing, remove all data of the same batch of the sample; if there is no missing, retain it, get the first training data set, and proceed to the next step; S2: Use the random forest regression model to build the first prediction model; For an exemplary embodiment, the first prediction model includes a first parameter prediction model, a second parameter prediction model and a third parameter prediction model; Specifically, in step S2, a multi-model parallel construction method based on Python programming software is used to implement multi-target prediction of random forests. Specifically, an independent random forest model is constructed for each target variable, and its hyperparameters are adjusted according to the characteristics of each target. This method allows each model to focus on a specific target variable, thereby achieving higher prediction accuracy and stability in a multi-target scenario. S3: training the first prediction model using the first training data set to obtain a trained first prediction model; For an exemplary embodiment, step S3 specifically includes: taking the organic carbon content data in the first data set, the free hydrocarbon content data in the second data set, and the pyrolysis hydrocarbon content data in the third data set as dependent variables, and taking the corresponding logging data as independent variables, training and predicting the first prediction model to obtain predicted organic carbon content, predicted free hydrocarbon content, and predicted pyrolysis hydrocarbon content; obtaining a first determination coefficient based on the organic carbon content data and the predicted organic carbon content in the first data set; obtaining a second determination coefficient based on the free hydrocarbon content data and the predicted free hydrocarbon content in the second data set; obtaining a third determination coefficient based on the pyrolysis hydrocarbon content data and the predicted pyrolysis hydrocarbon content in the third data set; when any one of the first determination coefficient, the second determination coefficient, and the third determination coefficient is less than a preset determination coefficient threshold, continuing training and prediction; when the first determination coefficient, the second determination coefficient, and the third determination coefficient are all not less than the preset determination coefficient threshold, stopping training to obtain a trained first prediction model; For an exemplary embodiment, the preset determination coefficient threshold is 0.8; Specifically, in step S3, the first prediction model is trained based on the random forest algorithm with the logging data as the independent variable and the experimental data as the dependent variable. The training process includes: data import, data partitioning, data normalization, model initialization, setting hyperparameters, model training and prediction; training the model multiple times and calculating the determination coefficient R 2 , until the prediction accuracy reaches a certain threshold, and the trained first prediction model is obtained; It should be noted that the trained first prediction model includes a trained first parameter prediction model, a trained second parameter prediction model and a trained third parameter prediction model; Optionally, the above step S3 further includes: drawing a scatter point fitting graph of the training set and the test set predicted dependent variables and the true value; S4: using the trained first prediction model to predict the single well logging data, to obtain a predicted value of organic carbon content, a predicted value of free hydrocarbon content, and a predicted value of pyrolysis hydrocarbon content; S5: Obtain the category of source rock of a single well according to the national evaluation standard of source rock, the predicted value of organic carbon content, the predicted value of free hydrocarbon content and the predicted value of pyrolysis hydrocarbon content; Specifically, in step S5, the trained model is used to complete the three-parameter prediction of the entire well section of a single well, and the classification of the source rock of a single well is completed according to the national standard hydrocarbon source rock classification evaluation table, as shown in Table 1, which is a standard table for hydrocarbon source rock classification, where Class I is the best and Class IV is the worst; Table 1: Source rock classification standards

[0016] S6: re-collecting data according to the single well source rock category and the single well logging data to obtain a second training data set, constructing a second prediction model using a random forest regression model, and training the second prediction model using the second training data set to obtain a trained second prediction model; Specifically, in step S6, data is collected again for the single wells that have completed the classification of source rock categories, and the source rock category is used as the dependent variable , the corresponding logging data is the independent variable , establish a random forest source rock category prediction model, divide the training set and test set, import data, and conduct model training; draw a map based on the training results and verification results to show the accuracy, such as Figure 2 (a) shows a comparison chart of the prediction effect of the training set, and the model evaluation is performed to train the optimal model and obtain the trained second prediction model; S7: using the trained first prediction model and the trained second prediction model to predict the well logging data to be predicted, obtaining the predicted value of the organic carbon content, the free hydrocarbon content, the pyrolysis hydrocarbon content and the single well source rock category prediction value to be predicted, and completing the single well and the whole area source rock evaluation; It should be noted that in step S7, by establishing the above two models, the prediction of necessary parameters is completed first, and then the source rock category is predicted, so as to realize the comprehensive evaluation and analysis of the source rocks in the whole area; Figure 3 The figure shows the comprehensive evaluation diagram of source rocks in a single well. The accuracy of the model can be verified by predicting source rock evaluation parameters and source rock categories.

[0017] In an exemplary embodiment, the data-driven comprehensive evaluation method for source rocks is implemented in the following manner. In this embodiment, the data-driven comprehensive evaluation method for source rocks includes the following steps: Obtain core test data and corresponding well logging data; obtain a training data set based on the core test data and the corresponding well logging data; The three-parameter prediction model was established using Python programming software, including: (1) Import necessary libraries, including pandas, numpy, and related modules in sklearn; (2) Import data: read the Excel file from the specified path and store it in the data variable; (3) Define independent and dependent variables: Divide the data into three parts, each of which contains a set of independent variables (X) and a corresponding dependent variable (Y); (4) Use the train_test_split function to divide each group of independent variables and dependent variables into a training set and a test set, with the test set accounting for 20%, and then normalize them; (5) Initialize three RandomForestRegressor random forest regression models and set different hyperparameters, including the number of trees, maximum depth, minimum number of sample leaves, and minimum number of sample splits; Training model: Use the training data set to train each model in the three-parameter prediction model; Make predictions: Make predictions on the training set and the test set respectively to obtain the prediction results; Calculate and display R 2 Value: Calculate the R of each model on the training set and test set 2 value; Save prediction results: Create two DataFrames to store the true values ​​and predicted values ​​of the training set and test set respectively; (6) First, clear the environment variables and import the data; (7) Randomly shuffle the data and divide the source rock types and corresponding logging data into training sets and test sets; (8) Adjust the number of trees, maximum depth, minimum number of sample leaves, and minimum number of sample splits, and enable OOB error prediction to calculate the accuracy of the random forest model; (9) Use the training set to train the model, then use the trained model to predict the training set and test set, and sort the data to facilitate plotting and displaying the accuracy; Through the establishment of the above two models, the necessary parameters are predicted first, and then the source rock category is predicted, thus realizing the comprehensive evaluation and analysis of source rocks in the whole area.

[0018] like Figure 2 (b) shows the comparison of the prediction effect of the test set. Figure 3 The figure shows the comprehensive evaluation diagram of single well source rock. Figure 2 (b) and Figure 3 The results show that the model is more effective.

[0019] This embodiment provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned data-driven comprehensive evaluation method for source rocks.

[0020] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A data-driven comprehensive evaluation method for source rocks, characterized in that: The following steps are involved: S1: Acquire core test data and corresponding well logging data; obtain a first training data set according to the core test data and the corresponding well logging data; S2: Use the random forest regression model to build the first prediction model; S3: Using the first training data set to train the first prediction model to obtain a trained first prediction model; S4: using the trained first prediction model to predict the single well logging data, to obtain a predicted value of organic carbon content, a predicted value of free hydrocarbon content, and a predicted value of pyrolysis hydrocarbon content; S5: Obtaining the category of the single well source rock according to the national evaluation standard for source rock, the predicted value of organic carbon content, the predicted value of free hydrocarbon content and the predicted value of pyrolysis hydrocarbon content; S6: re-collecting data according to the single well source rock category and the single well logging data to obtain a second training data set, constructing a second prediction model using a random forest regression model, and training the second prediction model using the second training data set to obtain a trained second prediction model; S7: Use the trained first prediction model and the trained second prediction model to predict the logging data to be predicted, obtain the predicted value of the organic carbon content, the free hydrocarbon content, the pyrolysis hydrocarbon content and the single well source rock category prediction value to be predicted, and complete the single well and the whole area source rock evaluation.

2. The data-driven comprehensive evaluation method for source rocks according to claim 1 is characterized in that: The core experimental data include organic carbon content data, free hydrocarbon content data and pyrolysis hydrocarbon content data; the corresponding logging data include natural gamma data, resistance data, acoustic wave time difference data, density data and neutron data.

3. The data-driven comprehensive evaluation method for source rocks according to claim 1 is characterized in that: The first training data set includes a first data set, a second data set and a third data set; the first data set includes organic carbon content data and corresponding well logging data; the second data set includes free hydrocarbon content data and corresponding well logging data; the third data set includes pyrolysis hydrocarbon content data and corresponding well logging data.

4. The data-driven comprehensive evaluation method for source rocks according to claim 1, characterized in that: The first prediction model includes a first parameter prediction model, a second parameter prediction model and a third parameter prediction model.

5. The data-driven comprehensive evaluation method for source rocks according to claim 3 is characterized in that: Step S3 specifically includes: taking the organic carbon content data in the first data set, the free hydrocarbon content data in the second data set and the pyrolysis hydrocarbon content data in the third data set as dependent variables and the corresponding logging data as independent variables, training and predicting the first prediction model to obtain predicted organic carbon content, predicted free hydrocarbon content and predicted pyrolysis hydrocarbon content; obtaining a first determination coefficient based on the organic carbon content data and the predicted organic carbon content in the first data set; obtaining a second determination coefficient based on the free hydrocarbon content data and the predicted free hydrocarbon content in the second data set; obtaining a third determination coefficient based on the pyrolysis hydrocarbon content data and the predicted pyrolysis hydrocarbon content in the third data set; when any one of the first determination coefficient, the second determination coefficient and the third determination coefficient is less than a preset determination coefficient threshold, continuing training and prediction; when the first determination coefficient, the second determination coefficient and the third determination coefficient are all not less than the preset determination coefficient threshold, stopping training to obtain a trained first prediction model.

6. The data-driven comprehensive evaluation method for source rocks according to claim 5 is characterized in that: The preset determination coefficient threshold is 0.

8.

7. The data-driven comprehensive evaluation method for source rocks according to claim 1, characterized in that: The trained first prediction model includes a trained first parameter prediction model, a trained second parameter prediction model and a trained third parameter prediction model.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the data-driven comprehensive evaluation method for source rocks described in any one of claims 1 to 7 are implemented.

Images