A machine learning-based method and apparatus for optimizing horizontal well perforation design

Abstract

This invention discloses a machine learning-based method for optimizing the design of horizontal well perforations. The method includes: acquiring fiber optic monitoring productivity data and logging and drilling data of horizontal wells, and performing quantitative analysis on both; training a productivity parameter model for the horizontal well based on the fiber optic monitoring productivity data and logging and drilling data to obtain a productivity parameter optimization prediction model; acquiring logging and drilling data of newly drilled horizontal wells that have not undergone fracturing operations, using the productivity parameter optimization prediction model to predict productivity, and sequentially optimizing the perforation positions of the horizontal wells to obtain optimized perforation positions. By utilizing an automated machine learning framework to iteratively optimize the horizontal well productivity parameter prediction model through hierarchical optimization of algorithms and parameters, the method solves the algorithm and parameter optimization problems. Based on the optimized productivity prediction model, the perforation positions are optimized, solving the problem of improving the effective perforation rate and improving the accuracy of oil and gas productivity prediction and the ratio of effective perforations.

Description

Technical Field

[0001] This invention relates to the field of unconventional oil and gas drilling and development technology, and in particular to a method and apparatus for optimizing the design of horizontal well perforation based on machine learning. Background Technology

[0002] Currently, large-section, multi-cluster horizontal well fracturing technology is the primary technology for developing unconventional reservoirs such as tight oil, tight gas, shale oil, and shale gas. After large-section, multi-cluster fracturing of horizontal wells, numerous artificial fractures are generated in the formation around the perforation point, significantly enhancing the reservoir's drainage area and permeability. Current large-section, multi-cluster perforation designs for horizontal wells often employ uniform geometric design methods; however, due to geological factors, engineering factors, and oil production systems, uneven distribution of horizontal well productivity and low effective perforation rates are common problems. In actual engineering operations and production, the rationality of perforation design determines the horizontal well's production volume, and the accuracy of horizontal well productivity prediction is a crucial factor restricting perforation design. Predicting the productivity of horizontal wells in unconventional reservoirs is a prevalent international challenge. Currently, analogy methods, empirical formula methods, and numerical simulation methods are the main methods for horizontal well productivity prediction; however, these traditional methods generally suffer from low data utilization and poor productivity prediction accuracy. While some deep learning and machine learning algorithms have been applied to the evaluation and prediction of oil and gas production capacity in horizontal wells, the problem of varying applicability to different algorithms on different datasets persists, leading to inconsistent results when these methods are widely adopted. Therefore, this paper proposes constructing a large perforation data table using fiber optic detection data and fracturing operation data to calibrate drilling and logging data. An automated machine learning framework is employed for intelligent prediction of oil and gas production capacity in horizontal wells, and a sliding window method is used for perforation design. The perforation location is intelligently optimized based on maximizing predicted production capacity. This technology plays a crucial role in improving oil and gas production capacity in horizontal wells, enhancing reservoir development, and reducing engineering operation costs. Summary of the Invention

[0003] The purpose of this invention is to provide a horizontal well perforation optimization design method and apparatus based on machine learning. Utilizing an automated machine learning framework, it iterative optimization of the horizontal well productivity parameter prediction model is achieved through hierarchical optimization of algorithms and parameters, solving algorithm and parameter optimization problems. Based on the optimized productivity prediction model, the sliding window method is used to simulate perforation points and predict productivity, optimizing perforation locations and addressing the issue of improving the effective perforation rate. This improves the accuracy of oil and gas productivity prediction and the ratio of effective perforations, playing a significant role in increasing horizontal well oil and gas productivity, enhancing oil and gas reservoir development, and reducing engineering operation costs.

[0004] To address the aforementioned technical problems, a first aspect of this invention provides a horizontal well perforation optimization design method based on machine learning, comprising the following steps:

[0005] Acquire fiber optic detection productivity data and logging and drilling data of horizontal wells, and perform quantitative analysis on both.

[0006] Based on the fiber optic detection production data and the logging and drilling data, the production parameter model of the horizontal well is trained to obtain an optimized prediction model for production parameters.

[0007] Obtain logging and drilling data of the horizontal well that has not undergone fracturing operations after new drilling, use the production capacity parameter optimization prediction model to predict production capacity, and then optimize the perforation position of the horizontal well in sequence to obtain the optimized perforation position of the horizontal well.

[0008] Furthermore, the acquisition of fiber optic detection productivity data and logging and drilling data of horizontal wells, and the quantitative analysis of both, includes:

[0009] Using the horizontal well perforation as the basic data unit, the fiber optic detection productivity data and logging and drilling data are obtained and linked to obtain a perforation data table.

[0010] The perforation data table is divided into data segments, with the logging and drilling data set as the feature dataset and the fiber optic detection productivity data set as the target dataset.

[0011] The hierarchical structure of the correlation coefficient matrix between the fiber optic detection capacity data and the logging and drilling data is analyzed using a hierarchical clustering algorithm based on correlation coefficients.

[0012] Further, the step of training the productivity parameter model for the horizontal well based on the fiber optic detection productivity data and the logging and drilling data includes:

[0013] The fiber optic testing capacity data is divided into a training set, a validation set, and a test set using cross-validation. The training set accounts for a first preset proportion of the total fiber optic testing capacity data, the validation set accounts for a second preset proportion of the total fiber optic testing capacity data, and the test set accounts for a third preset proportion of the total fiber optic testing capacity data.

[0014] The training set of the fiber optic detection capacity data was trained using a machine learning automated regression prediction framework.

[0015] Furthermore, the first preset ratio is 70%;

[0016] The second preset ratio is 20%;

[0017] The third preset ratio value is 10%.

[0018] Furthermore, the step of training the model on the training set using a machine learning automated regression prediction framework includes:

[0019] The model is trained using the training set data, the minimum prediction residual of the test set is used as the parameter for iterative optimization, and the minimum prediction residual of the test set is used as the algorithm for iterative optimization.

[0020] The parameters of the optimized prediction model for the production capacity parameters are obtained by combining the methods and parameters that result in the lowest prediction error and the highest prediction accuracy for the maximum daily oil production and maximum daily liquid production.

[0021] Furthermore, after acquiring the logging and drilling data of the horizontal well that has not undergone fracturing operations after the new well is drilled, the method further includes:

[0022] The horizontal wells that have not undergone fracturing operations after new drilling are divided into large sections with equal spacing.

[0023] The perforation design is carried out sequentially at each drilling data sampling point or logging data sampling point within the equally spaced large segments using the sliding window method.

[0024] Accordingly, a second aspect of the present invention provides a machine learning-based horizontal well perforation optimization design apparatus, comprising:

[0025] The data acquisition module is used to acquire fiber optic detection productivity data and logging and drilling data of horizontal wells, and to perform quantitative analysis on both.

[0026] The model training module is used to train the production parameter model of the horizontal well based on the fiber optic detection production data and the logging and drilling data, so as to obtain an optimized prediction model for the production parameters.

[0027] The perforation optimization module is used to acquire logging and drilling data of the horizontal well after new drilling without fracturing operations, use the production capacity parameter optimization prediction model to predict production capacity, and optimize the perforation position of the horizontal well in sequence to obtain the optimized perforation position of the horizontal well.

[0028] Furthermore, the data acquisition module includes:

[0029] The data synthesis unit is used to obtain the fiber optic detection productivity data and logging and drilling data as the basic data unit of the horizontal well perforation, and link them to obtain a perforation data table.

[0030] A data partitioning unit is used to partition the perforation data table, setting the logging and drilling data as a feature dataset and the fiber optic detection capacity data as a target dataset.

[0031] The coefficient analysis unit is used to analyze the hierarchical structure of the correlation coefficient matrix between the fiber optic detection capacity data and the logging and drilling data using a hierarchical clustering algorithm based on correlation coefficients.

[0032] Furthermore, the model training module includes:

[0033] The dataset partitioning unit is used to divide the optical fiber testing capacity data into a training set, a validation set, and a test set using a cross-validation method. The training set accounts for a first preset proportion of the total optical fiber testing capacity data, the validation set accounts for a second preset proportion of the total optical fiber testing capacity data, and the test set accounts for a third preset proportion of the total optical fiber testing capacity data.

[0034] The model training unit is used to train a model on the training set of the fiber optic detection capacity data using a machine learning automated regression prediction framework.

[0035] Furthermore, the first preset ratio is 70%;

[0036] The second preset ratio is 20%;

[0037] The third preset ratio value is 10%.

[0038] Furthermore, the perforation optimization module includes:

[0039] The algorithm parameter optimization unit is used to train the model with the training set data, iteratively optimize the parameters with the minimum prediction residual of the test set as the parameter, and iteratively optimize the algorithm with the minimum prediction residual of the test set as the algorithm.

[0040] The parameter acquisition unit outputs a combination of methods and parameters that minimize the prediction error and maximize the prediction accuracy of the maximum daily oil production and maximum daily liquid production, thereby obtaining the parameters of the optimized prediction model for the production capacity parameters.

[0041] Furthermore, the perforation optimization module also includes:

[0042] Distance division unit, which is used to divide the horizontal well into large segments with equal spacing after new drilling without fracturing operations;

[0043] The parameter design unit is used to sequentially design perforations at drilling data sampling points or logging data sampling points within the equally spaced large segments using the sliding window method.

[0044] Accordingly, a third aspect of the present invention provides an electronic device, including: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to cause the at least one processor to perform the above-described machine learning-based horizontal well perforation optimization design method.

[0045] Accordingly, a fourth aspect of the present invention provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the above-described machine learning-based horizontal well perforation optimization design method.

[0046] The above-described technical solutions of the embodiments of the present invention have the following beneficial technical effects:

[0047] By utilizing an automated machine learning framework and employing hierarchical optimization of algorithms and parameters, the horizontal well productivity prediction model is iteratively optimized, addressing algorithm and parameter optimization issues. Based on the optimized productivity prediction model, the sliding window method is used to simulate perforation locations and predict productivity, optimizing perforation positions and addressing the issue of improving the effective perforation rate. This enhances the accuracy of oil and gas productivity prediction and the ratio of effective perforations, playing a crucial role in increasing horizontal well oil and gas productivity, improving reservoir development effectiveness, and reducing engineering operation costs. Attached Figure Description

[0048] Figure 1 This is a flowchart of the horizontal well perforation optimization design method based on machine learning provided in an embodiment of the present invention;

[0049] Figure 2 This is a technical roadmap for the machine learning-based horizontal well perforation optimization design method provided in this embodiment of the invention;

[0050] Figure 3 This is a schematic diagram of the machine learning automation technology framework provided in an embodiment of the present invention;

[0051] Figure 4 This is a schematic diagram of hierarchical clustering based on correlation coefficient between drilling data and maximum daily production data provided in an embodiment of the present invention;

[0052] Figure 5 This is a block diagram of the horizontal well perforation optimization design device module provided in an embodiment of the present invention;

[0053] Figure 6 This is a block diagram of the data acquisition module provided in an embodiment of the present invention;

[0054] Figure 7 This is a block diagram of the model training module provided in an embodiment of the present invention;

[0055] Figure 8This is a block diagram of the perforation optimization module provided in an embodiment of the present invention.

[0056] Figure label:

[0057] 1. Data Acquisition Module, 11. Data Synthesis Unit, 12. Data Partitioning Unit, 13. Coefficient Analysis Unit, 2. Model Training Module, 21. Dataset Partitioning Unit, 22. Model Training Unit, 3. Perforation Optimization Module, 31. Algorithm Parameter Optimization Unit, 32. Parameter Acquisition Unit, 33. Distance Partitioning Unit, 34. Parameter Design Unit. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0059] Figure 1 This is a flowchart of a machine learning-based horizontal well perforation optimization design method provided in an embodiment of the present invention.

[0060] Figure 2 This is a technical roadmap for a machine learning-based horizontal well perforation optimization design method provided in this embodiment of the invention.

[0061] Figure 3 This is a schematic diagram of the machine learning automation technology framework provided in an embodiment of the present invention.

[0062] Please refer to Figure 1 , Figure 2 and Figure 3 The first aspect of this invention provides a horizontal well perforation optimization design method based on machine learning, comprising the following steps:

[0063] S100 acquires fiber optic detection productivity data and logging and drilling data of horizontal wells, and performs quantitative analysis on both.

[0064] S200 uses fiber optic detection productivity data and logging and drilling data to train a productivity parameter model for horizontal wells, resulting in an optimized prediction model for productivity parameters.

[0065] S300 acquires logging and drilling data of horizontal wells that have not undergone fracturing operations after new drilling, uses a production capacity parameter optimization prediction model to predict production capacity, and optimizes the perforation position of the horizontal well in sequence to obtain the optimized horizontal well perforation position.

[0066] For wells that have not undergone fracturing operations after new drilling, after designing the fracturing section, the sliding window method is used to process logging and drilling data into data units as feature parameters within the fracturing section. The final production parameter optimization prediction model is then used to predict production capacity, with the maximum predicted oil and gas production capacity as the decision condition, and progressive hierarchical location optimization is performed. The final output is the optimal combination of perforation design locations.

[0067] Further, in step S100, fiber optic detection productivity data and logging and drilling data of the horizontal well are acquired and quantitatively analyzed, including:

[0068] S110 uses horizontal well perforation as the basic data unit to obtain fiber optic detection productivity data and logging and drilling data, and links them to obtain a perforation data table.

[0069] Using horizontal well perforation points as the basic data unit, a large perforation data table is constructed by linking fiber optic detection production data, fracturing operation parameters, core test data, logging and drilling data.

[0070] S120: Divide the perforation data table into data sets, setting the logging and drilling data as the feature dataset and the fiber optic detection production data as the target dataset.

[0071] S130 uses a hierarchical clustering algorithm based on correlation coefficients to analyze the hierarchical structure of the correlation coefficient matrix between fiber optic detection capacity data and logging and drilling data.

[0072] Further, in step S200, based on fiber optic detection productivity data and logging and drilling data, a productivity parameter model for horizontal wells is trained, including:

[0073] S210, using cross-validation, the fiber optic testing capacity data is divided into a training set, a verification set, and a test set. The training set accounts for a first preset proportion of the total fiber optic testing capacity data, the verification set accounts for a second preset proportion of the total fiber optic testing capacity data, and the test set accounts for a third preset proportion of the total fiber optic testing capacity data.

[0074] S220 utilizes a machine learning-based automated regression prediction framework to train a model on the training set of fiber optic testing capacity data.

[0075] By using a machine learning automation technology framework as an algorithm and parameter optimization tool, a horizontal well oil and gas production parameter model is trained, and the optimal algorithm and parameter combination are optimized to obtain the final production parameter optimization prediction model.

[0076] Specifically, the first preset ratio is 70%; the second preset ratio is 20%; and the third preset ratio is 10%.

[0077] Further, in step S300, the training set is trained using a machine learning automated regression prediction framework, including:

[0078] S310 is a model training algorithm that uses the training set data, uses the minimum prediction residual of the test set as the parameter for iterative optimization, and uses the minimum prediction residual of the test set as the algorithm for iterative optimization.

[0079] S320 outputs the combination of methods and parameters with the lowest prediction error and highest prediction accuracy for maximum daily oil production and maximum daily liquid production, thus obtaining the parameters of the capacity parameter optimization prediction model.

[0080] Furthermore, in step S300, after obtaining the logging and drilling data of the horizontal well that has not undergone fracturing operations after the new well is drilled, the following steps are also included:

[0081] S330 is used to divide horizontal wells that have not undergone fracturing operations after new drilling into large sections with equal spacing.

[0082] S340, in sequence, uses the sliding window method to design perforations at drilling data sampling points or logging data sampling points one by one within a large segment with equal spacing.

[0083] Figure 4 This is a schematic diagram of hierarchical clustering based on correlation coefficient between drilling data and maximum daily production data provided in an embodiment of the present invention;

[0084] Please refer to Figure 6 The following is an example of a specific test project of the technical solution of this invention:

[0085] The above steps involve capacity prediction and perforation optimization, ultimately achieving a capacity prediction accuracy of 90% and an effective perforation rate of 95%. The specific procedures are as follows:

[0086] (1) Using the perforation depth of the well detected by M-fiber as the unit data, the average method is used to discretize the data, and the production capacity test results, logging data and drilling data are linked together to form a large perforation data table.

[0087] (2) The perforation data table was partitioned, with drilling and logging data designated as the feature dataset and fiber optic detection data as the target dataset. Production parameters were reconstructed from the fiber optic detection target dataset, and the maximum daily oil production and maximum daily liquid production were reconstructed as effective production prediction and evaluation parameters. Cross-validation was used to randomly partition the entire fiber optic detection dataset into a training set (70%), a validation set (20%), and a test set (10%).

[0088] (3) Utilizing a machine learning-automated intelligent regression prediction framework, the model is trained using the training set data. The parameters are iteratively optimized using the minimum prediction residual on the test set, and the algorithm is iteratively optimized using the minimum prediction residual on the test set. Finally, the combination of methods and parameters with the lowest prediction error and highest prediction accuracy for parameters such as maximum daily oil production and maximum daily liquid production is output as the model parameters.

[0089] (4) For the newly drilled unfractured well M, first divide it into large sections with equal spacing. In each large section, use the sliding window method to design perforations at each drilling data sampling point (1m) or logging data sampling point (0.125m). The perforation length is 1m. Use the production capacity prediction optimization model of the machine learning automated intelligent regression prediction framework to predict the production capacity parameters.

[0090] (5) Considering the necessary interval between perforation clusters, the perforation positions are optimized one by one based on the production capacity prediction results, and the perforation optimization design results are output.

[0091] The above technical solution effectively solves the problems of high-precision prediction of post-compression production capacity and optimized design of effective perforations in shale oil reservoirs. It has practical significance for deeply exploring the factors controlling oil and gas production capacity and maximizing the accuracy of production capacity prediction. It also has a clear guiding role in improving the effective perforation ratio and reducing production operating costs.

[0092] Figure 5 This is a block diagram of the horizontal well perforation optimization design device module provided in an embodiment of the present invention.

[0093] Accordingly, please refer to Figure 5 A second aspect of this invention provides a machine learning-based horizontal well perforation optimization design device, comprising: a data acquisition module 1, a model training module 2, and a perforation optimization module 3. The data acquisition module 1 acquires fiber optic detection productivity data and logging and drilling data of the horizontal well, and performs quantitative analysis on both. The model training module 2 trains a productivity parameter model for the horizontal well based on the fiber optic detection productivity data and logging and drilling data to obtain a productivity parameter optimization prediction model. The perforation optimization module 3 acquires logging and drilling data of a newly drilled horizontal well that has not undergone fracturing operations, uses the productivity parameter optimization prediction model to predict productivity, and sequentially optimizes the perforation positions of the horizontal well to obtain the optimized perforation positions.

[0094] Figure 6 This is a block diagram of the data acquisition module provided in an embodiment of the present invention.

[0095] For details, please refer to Figure 6The data acquisition module 1 includes a data synthesis unit 11, a data partitioning unit 12, and a coefficient analysis unit 13. Specifically, the data synthesis unit 11 uses horizontal well perforation as the basic data unit to acquire fiber optic detection productivity data and logging and drilling data, linking them to obtain a perforation data table. The data partitioning unit 12 partitions the perforation data table, setting logging and drilling data as the feature dataset and fiber optic detection productivity data as the target dataset. The coefficient analysis unit 13 uses a hierarchical clustering algorithm based on correlation coefficients to analyze the hierarchical structure of the correlation coefficient matrix between fiber optic detection productivity data and logging and drilling data.

[0096] Figure 7 This is a block diagram of the model training module provided in an embodiment of the present invention.

[0097] For details, please refer to Figure 7 The model training module 2 includes a dataset partitioning unit 21 and a model training unit 22. The dataset partitioning unit 21 uses cross-validation to divide the fiber optic testing capacity data into a training set, a validation set, and a test set. The training set accounts for a first preset proportion of the total fiber optic testing capacity data, the validation set accounts for a second preset proportion, and the test set accounts for a third preset proportion. The model training unit 22 uses a machine learning automated regression prediction framework to train a model on the training set of the fiber optic testing capacity data.

[0098] Furthermore, the first preset ratio is 70%; the second preset ratio is 20%; and the third preset ratio is 10%.

[0099] Figure 8 This is a block diagram of the perforation optimization module provided in an embodiment of the present invention.

[0100] For details, please refer to Figure 8 The perforation optimization module 3 includes an algorithm parameter optimization unit 31 and a parameter acquisition unit 32. The algorithm parameter optimization unit 31 trains the model using training set data, iteratively optimizes parameters using the minimum prediction residual on the test set as the parameter, and iteratively optimizes the algorithm using the minimum prediction residual on the test set as the algorithm. The parameter acquisition unit 32 outputs a combination of methods and parameters that minimizes the prediction error and maximizes the prediction accuracy for maximum daily oil production and maximum daily liquid production, thus obtaining the parameters of the capacity parameter optimization prediction model.

[0101] Furthermore, the perforation optimization module 3 also includes a distance division unit 33 and a parameter design unit 34. The distance division unit 33 is used to divide horizontal wells that have not undergone fracturing operations after new drilling into equally spaced large sections; the parameter design unit 34 is used to sequentially design perforations within the equally spaced large sections by sliding the window method at drilling data sampling points or logging data sampling points.

[0102] Accordingly, a third aspect of the present invention provides an electronic device, including: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by a processor, the instructions being executed by the processor to cause the at least one processor to perform the above-described machine learning-based horizontal well perforation optimization design method.

[0103] Accordingly, a fourth aspect of the present invention provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the above-described machine learning-based horizontal well perforation optimization design method.

[0104] This invention aims to protect a machine learning-based method for optimizing the design of horizontal well perforations, comprising the following steps: acquiring fiber optic monitoring productivity data and logging and drilling data of the horizontal well, and performing quantitative analysis on both; training a productivity parameter model for the horizontal well based on the fiber optic monitoring productivity data and logging and drilling data to obtain a productivity parameter optimization prediction model; acquiring logging and drilling data of a newly drilled horizontal well that has not undergone fracturing operations, using the productivity parameter optimization prediction model to predict productivity, and sequentially optimizing the horizontal well perforation positions to obtain the optimized horizontal well perforation positions. The above technical solution has the following advantages:

[0105] By utilizing an automated machine learning framework and employing hierarchical optimization of algorithms and parameters, the horizontal well productivity prediction model is iteratively optimized, addressing algorithm and parameter optimization issues. Based on the optimized productivity prediction model, the sliding window method is used to simulate perforation locations and predict productivity, optimizing perforation positions and addressing the issue of improving the effective perforation rate. This enhances the accuracy of oil and gas productivity prediction and the ratio of effective perforations, playing a crucial role in increasing horizontal well oil and gas productivity, improving reservoir development effectiveness, and reducing engineering operation costs.

[0106] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0107] 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 application. 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, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0108] 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.

[0109] 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.

Claims

1. A machine learning-based method for optimizing the design of horizontal well perforation, characterized in that, Bag Includes the following steps: Acquire fiber optic monitoring productivity data and logging and drilling data for horizontal wells, and perform qualitative analysis on both. Quantitative analysis; Based on the fiber optic detection capacity data and the logging and drilling data, the horizontal well is... The capacity parameter model is trained to obtain an optimized prediction model for capacity parameters. Obtain logging and drilling data for the horizontal wells that have not undergone fracturing operations after new drilling. The production capacity prediction model is optimized using the aforementioned production capacity parameters to predict production capacity, and the horizontal well perforation is performed sequentially. The perforation position is optimized to obtain the optimized horizontal well perforation position; The process involves acquiring fiber optic detection productivity data and logging and drilling data for horizontal wells and analyzing them. Quantitative analysis of the two was conducted, including: Using the horizontal well perforation as the basic data unit, the fiber optic detection capacity data and measurement data are obtained. Link well and drilling data to obtain a perforation data table; The perforation data table is segmented, and the logging and drilling data are set as features. The dataset is defined as the fiber optic testing capacity data; The optical fiber detection capacity data and the data were analyzed using a hierarchical clustering algorithm based on correlation coefficients. The hierarchical structure of the correlation coefficient matrix between well logging and drilling data was analyzed. The process is based on the fiber optic detection capacity data and the logging and drilling data. Training the productivity parameter model for the horizontal well includes: The fiber optic testing capacity data was divided into a training set, a validation set, and a cross-validation set. The test set, wherein the training set accounts for a first preset proportion of the total fiber optic testing capacity data. The value, the second preset proportion of the test set to the total amount of fiber optic testing capacity data, the measurement The third preset proportion of the test set to the total amount of fiber optic testing capacity data; The fiber optic testing capacity data were trained using a machine learning-based automated regression prediction framework. The set is used for model training; The training set is modeled using a machine learning automated regression prediction framework. Type training, including: The model is trained using the training set data, and the minimum prediction residual on the test set is used as the parameter. The iterative optimization parameters are determined by minimizing the predicted residual on the test set. Law; The prediction error for maximum daily oil production and maximum daily liquid production is the lowest, and the prediction accuracy is the highest. The combination of methods and parameters yields the parameters of the capacity parameter optimization prediction model; The logging of the horizontal well that was not subjected to fracturing operations after being newly drilled. Following drilling data, it also includes: For newly drilled horizontal wells that have not undergone fracturing operations, perform large-section divisions at equal intervals. point; Sequentially, within the equally spaced large segments, the sliding window method is used to sample drilling data points one by one. Sliding perforation design at well logging data sampling points.

2. The machine learning-based horizontal well perforation optimization design method according to claim 1, Its features are, The first preset ratio is 70%; The second preset ratio is 20%; The third preset ratio value is 10%.

3. A machine learning-based horizontal well perforation optimization design device, characterized in that, For implementing the machine learning-based horizontal well perforation optimization design method as described in claim 1 or 2, the machine learning-based horizontal well perforation optimization design apparatus includes: The data acquisition module is used to acquire fiber optic monitoring productivity data and logging and drilling data for horizontal wells. Data, and quantitative analysis of both; The model training module is used to train the production parameter model of the horizontal well based on the fiber optic detection production data and the logging and drilling data, so as to obtain an optimized prediction model for the production parameters. A perforation optimization module is used to obtain the level of a new well after which no fracturing operations have been performed. Well logging and drilling data are used to optimize the prediction model based on the aforementioned production parameters for production capacity prediction. The horizontal well perforation positions are optimized sequentially to obtain the optimized horizontal well perforation positions. Place; The data acquisition module includes: The data synthesis unit is used to acquire the data based on the horizontal well perforation as the data unit. The fiber optic detection capacity data is linked with logging and drilling data to obtain a perforation data table; The data partitioning unit is used to partition the perforation data table and divide the logging data. The drilling data is set as the feature dataset, and the fiber optic detection capacity data is set as the target data. set; A coefficient analysis unit is used to analyze the optical fiber using a hierarchical clustering algorithm based on correlation coefficients. The hierarchical structure of the correlation coefficient matrix between the detection capacity data and the well logging and drilling data is used to analyze the relationship between the detection capacity data and the data. analyze; The model training module includes: Data set partitioning unit, which is used to partition the fiber optic testing capacity data using cross-validation. The fiber optic detection product is divided into a training set, a validation set, and a test set, wherein the training set accounts for a portion of the total fiber optic detection output. The first preset proportion of the total amount of data, wherein the test set accounts for a certain percentage of the total amount of fiber optic testing capacity data. The second preset ratio value, and the third preset ratio value of the test set to the total amount of fiber optic testing capacity data. Proportion value; The model training unit is used to train the optical fiber using a machine learning automated regression prediction framework. The model is trained using the training set of production capacity data. The perforation optimization module includes: The algorithm parameter optimization unit is used to train the model using the training set data. The minimum predicted residual of the test set is used as the iterative optimization parameter, and the minimum predicted residual of the test set is used as the parameter. Small as an iterative optimization algorithm; The parameter acquisition unit outputs the lowest prediction error for maximum daily oil production and maximum daily liquid production. The combination of methods and parameters that yields the highest prediction accuracy is used to obtain the parameters of the optimized prediction model for the production capacity parameters. number; The perforation optimization module also includes: Distance division unit, used for horizontal wells that have not undergone fracturing operations after new drilling. Divide the large segments into equally spaced sections; The parameter design unit is used to sequentially apply the sliding window method to each of the equally spaced large segments. Slide perforation design at drilling data sampling points or logging data sampling points.

4. The machine learning-based horizontal well perforation optimization design device according to claim 3, Its features are, The first preset ratio is 70%; The second preset ratio is 20%; The third preset ratio value is 10%.

Images