A method for predicting formation granularity two-dimensional distribution by fusing Optuna-XGBoost and logging data
By combining well logging data and rock grain size analysis using the Optuna-XGBoost framework, a two-dimensional formation grain size distribution prediction model was constructed, which solved the problems of long time consumption and high cost in traditional methods, and realized the accurate characterization and widespread application of two-dimensional formation grain size distribution.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-03-29
- Publication Date
- 2026-03-17
AI Technical Summary
Traditional rock grain size measurement experiments are time-consuming, costly, and discontinuous, making it difficult to accurately characterize the two-dimensional distribution of formation grain size. Existing machine learning methods can only generate one-dimensional feature values of grain size, which are difficult to apply to the prediction of two-dimensional grain size distribution in oil and gas reservoirs.
Using the Optuna-XGBoost framework, combined with well logging data and rock grain size analysis experiments, a two-dimensional formation grain size distribution prediction model was constructed through hyperparameter optimization and the ultimate gradient boosting tree algorithm, thereby achieving continuous depth characterization of formation grain size.
It improves the accuracy and representativeness of two-dimensional distribution of stratigraphic grain size, reduces costs, and is easy to promote in different oil and gas fields, laying the foundation for the fine evaluation of the physical properties of oil and gas reservoir rocks.
Smart Images
Figure CN116522276B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of oil and gas resource geological exploration and development evaluation technology, and in particular to a two-dimensional formation grain size distribution prediction method that integrates Optuna-XGBoost and well logging data. Background Technology
[0002] Two-dimensional formation grain size distribution characterization plays an indicative role in the analysis of hydrodynamic conditions and the identification of sedimentary environments in oil and gas reservoirs. Traditional rock grain size measurement experiments, such as sieve analysis and laser grain size analysis, are not only time-consuming and costly, but also suffer from discontinuities in the two-dimensional grain size distribution data across depth due to limited core sampling yields. Therefore, rock grain size measurement experiments struggle to address the problem of insufficient representativeness in two-dimensional grain size distribution.
[0003] Well logging data can reflect the physical field response characteristics of formations at continuous depths. Traditional methods for calculating formation grain size based on well logging data mainly rely on regional empirical relationships. However, for most oil and gas reservoirs, it is difficult to directly apply well logging data to establish empirical relationships with high correlation to formation grain size. Machine learning technology is a technique that can better reveal the nonlinear relationship between formation physical field response and formation rock composition and structure. Although combining well logging data with machine learning can compensate for the shortcomings of rock grain size analysis experiments, existing results all focus on one-dimensional feature values of grain size, and no relevant results have been found that apply machine learning to characterize the overall characteristics of two-dimensional grain size distribution.
[0004] Therefore, it is urgent to study a method that fully combines the advantages of well logging data reflecting the response of the formation physical field and rock grain size analysis experiments to improve the accuracy of the two-dimensional grain size distribution of the formation at continuous depth. The method should be to use the automatic hyperparameter optimization framework (Optuna) and the extreme gradient boosting tree (XGBoost) to characterize the two-dimensional grain size distribution of the formation. Summary of the Invention
[0005] To address the aforementioned technical issues, this invention discloses a two-dimensional formation grain size distribution prediction method that integrates Optuna-XGBoost and well logging data. This method overcomes the limitation of conventional machine learning algorithms, which can only generate one-dimensional feature values of grain size, reduces the cost of two-dimensional formation grain size distribution characterization, and is easy to apply and promote in different oil and gas fields, laying the foundation for the fine evaluation of petrophysical properties of oil and gas reservoirs.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] According to a first aspect of the present invention, a method for predicting two-dimensional formation grain size distribution by fusing Optuna-XGBoost and well logging data is provided.
[0008] In some alternative embodiments, the method for predicting the two-dimensional distribution of formation grain size by fusing Optuna-XGBoost and well logging data includes the following steps:
[0009] Step S1. Well logging data preprocessing;
[0010] Step S2. Preprocessing of rock grain size analysis experimental data;
[0011] Step S3. Determine the range of the hyperparameter optimization framework Optuna and optimize the training;
[0012] Step S4. Assign the optimal parameters to XGBoost and construct a two-dimensional formation grain size distribution prediction model.
[0013] Optionally, step S1, the well logging data preprocessing step, includes:
[0014] Step S11. Collect multiple types of logging data acquired by different logging tools and plot a histogram for each type of logging data;
[0015] Step S12. Perform Spearman correlation analysis to obtain cross plots and delete logging data types with poor correlation.
[0016] Step S13. Based on the cross plot of different logging data, remove outliers from the logging data and normalize each type of logging data according to the Min-Max processing method.
[0017] Optionally, step S2, the step of preprocessing the rock grain size analysis experimental data, includes:
[0018] Step S21. Collect the rocks drilled from the wellbore and mark the depth of each rock in the formation;
[0019] Step S22. Conduct rock particle size analysis experiments using a laser particle size analyzer, and arrange the two-dimensional distribution results of rock particle size obtained from the experiment in a uniform manner, i.e., the arrangement of the points is consistent.
[0020] Optionally, step S3, which involves determining the range of the hyperparameter optimization framework Optuna and optimizing the training, includes:
[0021] Step S31. Take the two-dimensional distribution of rock grain size obtained in step S22 and the preprocessed logging data in step S1 as inputs, and train and optimize the hyperparameter combination using the hyperparameter optimization framework Optuna.
[0022] Step S32. Optuna defines the objective function, loss function, and hyperparameter combination within a specific range. In each training round, it will output the score of the current loss function and compare it with the previous training round. If the Optuna optimization framework encounters an error before the set maximum training round, the program will report an error and the Optuna optimization framework needs to be restarted for retraining.
[0023] Step S33. After the maximum number of training rounds is completed, output the historical best score and the corresponding best hyperparameter combination.
[0024] Optionally, step S4, which involves assigning the optimal parameters to XGBoost and constructing a two-dimensional formation grain size distribution prediction model, includes:
[0025] Step S41. Assign the best hyperparameters output in step S3 to XGBoost, and retrain XGBoost. When the mean square error of the trained model is less than a certain selected cutoff value and the coefficient of determination reaches 0.6, the formation grain size two-dimensional distribution prediction model that integrates Optuna-XGBoost and well logging data is established.
[0026] Step S42. Import the field logging data into the model established in step S41 to predict the two-dimensional distribution of formation grain size at continuous depth.
[0027] Step S43. Filter the prediction results from step S42 three times;
[0028] Step S44. Verify the filtered prediction results.
[0029] According to a second aspect of the present invention, a two-dimensional formation grain size distribution prediction system that integrates Optuna-XGBoost and well logging data is provided.
[0030] In some alternative embodiments, the formation grain size two-dimensional distribution prediction system, which integrates Optuna-XGBoost and well logging data, includes:
[0031] The data preprocessing unit is used for the preprocessing of well logging data;
[0032] The experimental data preprocessing unit is used for the preprocessing of experimental data for rock grain size analysis.
[0033] An optimized framework training unit is used to determine the range of the hyperparameter optimization framework Optuna and optimize the training.
[0034] The prediction model building unit is used to assign the optimal parameters to XGBoost and build a two-dimensional formation grain size distribution prediction model.
[0035] Based on the results of the data preprocessing unit and the experimental data preprocessing unit, the optimization framework training unit trains and optimizes the hyperparameter combination. The optimal parameters are then assigned to the prediction model construction unit to obtain the prediction model for prediction.
[0036] According to a third aspect of the present invention, a computer device is provided.
[0037] In some alternative embodiments, the computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the prediction method described above.
[0038] The beneficial effects of this invention are that it achieves continuous depth characterization of two-dimensional formation grain size distribution, effectively integrates the advantages of well logging data and rock grain size analysis experiments, makes up for the lack of representativeness of grain size analysis experiments, overcomes the limitation of conventional machine learning algorithms that can only generate one-dimensional feature values of grain size, improves the understanding of hydrodynamic conditions and sedimentary environment at different depths, reduces the cost of two-dimensional formation grain size distribution characterization, is easy to apply and promote in different oil and gas fields, and lays the foundation for the fine evaluation of rock physical properties of oil and gas reservoirs. Attached Figure Description
[0039] Figure 1 This is a flowchart illustrating a two-dimensional formation grain size distribution prediction method that integrates Optuna-XGBoost and well logging data, as shown in an embodiment of the present invention.
[0040] Figure 2 This is a comparison diagram of the two-dimensional particle size distribution prediction results and rock particle size analysis experimental results at four different depth points in well A, as shown in an embodiment of the present invention. Detailed Implementation
[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0042] Example 1
[0043] A method for predicting two-dimensional formation grain size distribution by integrating Optuna-XGBoost and well logging data, such as Figure 1 As shown, it includes the following steps:
[0044] Step S1. Well logging data preprocessing
[0045] Specifically, this includes: Step S11. Collecting multiple types of logging data acquired by different logging tools and drawing a histogram of each type of logging data;
[0046] Step S12. Perform Spearman correlation analysis to obtain cross plots and delete logging data types with poor correlation.
[0047] Step S13. Based on the cross plot of different logging data, remove outliers from the logging data and normalize each type of logging data according to the Min-Max processing method.
[0048] Step S2. Preprocessing of rock grain size analysis experimental data
[0049] Specifically, it includes:
[0050] Step S21. Collect the rocks drilled from the wellbore and mark the depth of each rock in the formation;
[0051] Step S22. Conduct rock particle size analysis experiments using a laser particle size analyzer, and arrange the two-dimensional distribution results of rock particle size obtained from the experiment in a uniform manner.
[0052] Step S3. Determine the range of the hyperparameter optimization framework Optuna and optimize the training.
[0053] Specifically, it includes:
[0054] Step S31. Take the two-dimensional distribution of rock grain size obtained in step S22 and the preprocessed logging data in step S1 as inputs, and train and optimize the hyperparameter combination using the hyperparameter optimization framework Optuna.
[0055] Step S32. Optuna defines the objective function, loss function, and hyperparameter combination within a specific range. In each training round, it will output the score of the current loss function and compare it with the previous training round. If the Optuna optimization framework encounters an error before the set maximum training round, the program will report an error and the Optuna optimization framework needs to be restarted for retraining.
[0056] Step S33. After the maximum number of training rounds is completed, output the historical best score and the corresponding best hyperparameter combination.
[0057] Step S4. Assign the optimal parameters to XGBoost and construct a two-dimensional formation grain size distribution prediction model.
[0058] Specifically, it includes:
[0059] Step S41. Assign the best hyperparameters output in step S3 to XGBoost, and retrain XGBoost. When the mean square error of the trained model is less than a certain selected cutoff value and the coefficient of determination reaches 0.6, the formation grain size two-dimensional distribution prediction model that integrates Optuna-XGBoost and well logging data is established.
[0060] Step S42. Import the field logging data into the model established in step S41 to predict the two-dimensional distribution of formation grain size at continuous depth.
[0061] Step S43. Filter the prediction results from step S42 three times;
[0062] Step S44. Verify the filtered prediction results.
[0063] Example 2
[0064] A two-dimensional formation grain size distribution prediction system integrating Optuna-XGBoost and well logging data includes:
[0065] The data preprocessing unit is used for the preprocessing of well logging data;
[0066] The experimental data preprocessing unit is used for the preprocessing of experimental data for rock grain size analysis.
[0067] An optimized framework training unit is used to determine the range of the hyperparameter optimization framework Optuna and optimize the training.
[0068] The prediction model building unit is used to assign the optimal parameters to XGBoost and build a two-dimensional formation grain size distribution prediction model.
[0069] Based on the results of the data preprocessing unit and the experimental data preprocessing unit, the optimization framework training unit trains and optimizes the hyperparameter combination. The optimal parameters are then assigned to the prediction model construction unit to obtain the prediction model for prediction.
[0070] Example 3
[0071] This embodiment provides a storage medium on which a program is stored. When the program is executed by a processor, it implements the steps in the two-dimensional formation grain size distribution prediction method that integrates Optuna-XGBoost and well logging data as provided in Embodiment 1.
[0072] The detailed steps of the above-described method are the same as those of the two-dimensional formation grain size distribution prediction method that integrates Optuna-XGBoost and well logging data provided in Example 1, and will not be repeated here.
[0073] Example 4
[0074] This embodiment provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the two-dimensional formation grain size distribution prediction method that integrates Optuna-XGBoost and well logging data as provided in Embodiment 1.
[0075] The detailed steps of the above-described procedure are the same as those of the gear fault diagnosis method based on nonlinear fast Hoyer spectrum provided in Example 1, and will not be repeated here.
[0076] Those skilled in the art will understand that the embodiments disclosed in this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage) containing computer-usable program code.
[0077] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0078] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0079] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0080] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0081] Application examples
[0082] Taking Well A in a Chinese oilfield as an example, this invention provides a two-dimensional formation grain size distribution prediction method that integrates Optuna-XGBoost and well logging data, comprising the following steps:
[0083] Step S1. Well logging data preprocessing
[0084] Multiple types of logging data were collected from well A using different logging tools, including spontaneous potential (SP), acoustic (AC), flushed zone formation resistivity (RXO), shallow lateral resistivity (RS), micro lateral resistivity (RMLL), true formation resistivity (RT), deep lateral resistivity (RD), density logging (DEN), caliper logging (CAL), natural gamma logging (GR), and compensated neutron logging (CNL). Histograms of each type of logging data were plotted. After Spearman correlation analysis and data integrity checks, RD, RS, and RMLL logging data were removed. Outliers were then eliminated based on cross-plots of different logging data. Finally, each type of logging data was normalized using the Min-Max method.
[0085] Step S2. Preprocessing of rock grain size analysis experimental data
[0086] Rock samples were collected from Well A, and the depth of each rock sample within the formation was marked at 355 points. A laser particle size analyzer was used to analyze the particle size of each of the 355 samples. The resulting rock particle sizes were ranked from smallest to largest as follows: 0.004 mm, 0.008 mm, 0.016 mm, 0.022 mm, 0.031 mm, 0.044 mm, 0.053 mm, 0.063 mm, 0.074 mm, and 0.088 mm. A total of 28 points with the following dimensions are arranged: 0.105mm, 0.125mm, 0.149mm, 0.177mm, 0.210mm, 0.250mm, 0.297mm, 0.354mm, 0.420mm, 0.505mm, 0.595mm, 0.707mm, 0.841mm, 1.000mm, 1.189mm, 1.414mm, 1.618mm, and 2.000mm.
[0087] Step S3. Determine the range of the hyperparameter optimization framework Optuna and optimize the training.
[0088] The two-dimensional distribution results of rock grain size in step S2 and the preprocessed logging data at the corresponding depth points in step S1 are used as inputs and fed into the optimization framework Optuna to train and optimize the hyperparameter combination. Since Optuna defines an objective function, a loss function, and a specific range of hyperparameter combinations, each training round will output the score of the current loss function and compare it with the previous training round. If the Optuna optimization framework encounters an error before the set maximum training rounds, the program will report an error and the Optuna optimization framework needs to be restarted for retraining. After the set maximum training rounds are completed, the historical best score and the corresponding best hyperparameter combination will be output.
[0089] S4. Assign the optimal parameters to XGBoost and construct a two-dimensional formation grain size distribution prediction model.
[0090] Assign the optimal hyperparameters output in step S3 to XGBoost, and retrain XGBoost. Select a mean squared error cutoff value of 10. When the mean squared error of the trained model is less than this cutoff value and the coefficient of determination reaches 0.6, the formation grain size two-dimensional distribution prediction model integrating Optuna-XGBoost and well logging data is established. Then, import the field well logging data into the model to predict the two-dimensional distribution of formation grain size at continuous depth. Perform three filtering operations on the prediction results, and finally verify the prediction results. Figure 2 As shown, the average coefficient of determination of the predicted two-dimensional formation grain size distribution reached 0.81, which verifies the accuracy of the two-dimensional formation grain size distribution prediction method that integrates Optuna-XGBoost and well logging data.
[0091] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.
Claims
1. A method for predicting formation-granularity two-dimensional distribution by fusing Optuna-XGBoost and logging data, characterized in that, Comprising the following steps: Step S1. Well logging data preprocessing; Step S2. Rock grain size analysis experimental data preprocessing; Step S3. Determine the hyperparameter optimization framework Optuna range and optimize training; Step S4. Assign the best parameters to XGBoost and build a formation grain size two-dimensional distribution prediction model; In step S2, the step of rock grain size analysis experimental data preprocessing, comprising: Step S21. Collect the rock drilled in the wellbore, and mark the depth of each rock in the formation; Step S22. Use a laser particle size tester to carry out rock grain size analysis experiment, and distribute the rock grain size two-dimensional distribution results obtained by experiment in a unified manner; In step S3, the step of determining the hyperparameter optimization framework Optuna range and optimizing training, comprising: Step S31. Take the rock grain size two-dimensional distribution results obtained in step S22 and the preprocessed logging data in step S1 as input, and bring them into the hyperparameter optimization framework Optuna to train and optimize the hyperparameter combination; Step S32. Optuna defines a target function loss function and a specific range of hyperparameter combination. Each round of training outputs the score of the current loss function, which is compared with the last round of training. If the Optuna optimization framework experiment is wrong before the maximum training rounds are set, the program will report an error and need to restart the Optuna optimization framework for training again; Step S33. When the maximum training rounds are set, the historical best score and the corresponding best hyperparameter combination are output; In step S4, the step of assigning the best parameters to XGBoost and building a formation grain size two-dimensional distribution prediction model, comprising: Step S41. Assign the best hyperparameters output in step S3 to XGBoost, and train XGBoost again. When the mean square error of the trained model is less than a selected cutoff value and the determination coefficient reaches 0.6, the establishment of the formation grain size two-dimensional distribution prediction model is realized by fusing Optuna-XGBoost and well logging data; Step S42. Import the field logging data into the model established in step S41 to predict the formation continuous depth grain size two-dimensional distribution; Step S43. Filter the prediction results of step S42 three times; Step S44. Verify the filtered prediction results.
2. The method of claim 1, wherein the method of predicting a formation-granularity two-dimensional distribution by fusing Optuna-XGBoost and well logging data is characterized by, In step S1, the step of well logging data preprocessing, comprising: Step S11. Collect multiple types of logging data collected by different logging instruments, and draw a histogram for each type of logging data; Step S12. Perform Spearman correlation analysis to obtain a cross plot, and delete logging data types with poor correlation; Step S13. According to the cross plot of different logging data, remove the abnormal values of the logging data, and normalize each type of logging data according to the Min-Max processing method.
3. A prediction system for use in a prediction method as claimed in claim 2, characterized in that Comprising: A data preprocessing unit for preprocessing well logging data; An experimental data preprocessing unit for preprocessing rock grain size analysis experimental data; An optimization framework training unit for determining the hyperparameter optimization framework Optuna range and optimizing training; An optimization framework training unit for determining the hyperparameter optimization framework Optuna range and optimizing training; A prediction model construction unit is configured to assign the optimal parameters to XGBoost and construct a formation-granularity two-dimensional distribution prediction model; Based on the results of the data preprocessing unit and the experimental data preprocessing unit, the optimization framework training unit trains the optimization hyperparameter combination, and the optimal parameters obtained are assigned to the prediction model construction unit to obtain the prediction model for prediction.
4. A computer-readable storage medium having stored thereon a program, characterized in that, The program is executed by the processor to implement the steps in the method for predicting formation-granularity two-dimensional distribution by fusing Optuna-XGBoost and logging data according to claim 2.
5. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized by The processor executes the program to implement the steps in the method for predicting formation-granularity two-dimensional distribution by fusing Optuna-XGBoost and logging data according to claim 2.
Citation Information
Patent Citations
Space-time estimation and prediction method for PM2.5 concentration distribution
CN111859800A
DNA recombination site prediction method based on XGBoost
CN114093420A