A method for predicting the effect of horizontal slit well fracturing
By using cluster analysis and establishing an intelligent prediction model, the problem of the inability to predict the fracturing effect of horizontally fractured oil wells in existing technologies has been solved, enabling accurate prediction of post-fracturing effects and calculation of increased oil production.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DAQING OILFIELD CO LTD
- Filing Date
- 2024-12-18
- Publication Date
- 2026-06-19
AI Technical Summary
Existing technologies cannot effectively predict the effects of fracturing horizontally fractured oil wells, especially failing to consider the influence of factors such as displacement methods and fracturing processes.
By cluster analysis of fracturing layers, the main controlling factors are identified, an intelligent prediction model is established, and the post-fracturing effects are predicted using pre-fracturing dynamic and static parameters and reservoir conditions.
It enables the prediction of the effects of fracturing in horizontally fractured oil wells, reflects the production increase pattern, provides accurate prediction of increased oil production, and has a low error rate.
Smart Images
Figure CN122241327A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the fields of fracturing and related technologies, specifically a method for predicting the fracturing effect of horizontally fractured oil wells. Background Technology
[0002] Regarding methods for predicting the effects of oil well fracturing, publication number CN117195140A provides a method based on a kernel regression accelerated algorithm. Its main technical idea is to clean, fill, and transform the fracturing data for preprocessing. The preprocessed results are then log-centered to find projection vectors for data partitioning, resulting in m subsets. c columns are randomly extracted from the kernel matrix to construct a column subset matrix C. A cross matrix W is constructed based on the column subset matrix C, thus obtaining a low-rank approximation matrix and a kernel SVR model. The model obtained in step S3 is trained on each region of the m subsets. Finally, each kernel SVR model predicts the fracturing data to be identified that falls within the same region. This patented technology is a method for predicting the effects of oil well fracturing based on a kernel regression accelerated algorithm, but it does not establish a production capacity influencing factor and production capacity prediction model for horizontal fractures, thus failing to predict the post-fracturing effects of horizontal fractures.
[0003] Publication number CN117744509A provides a method for predicting the production capacity of fractured horizontal fracture wells in shallow oil reservoirs. The main idea is to establish basic assumptions for the model and, based on these assumptions, establish simplified formulas in a two-dimensional cylindrical coordinate system. However, it only makes numerous simplifications and assumptions based on physical mechanisms, without examining past data patterns or considering factors such as displacement methods and fracturing techniques.
[0004] Publication number CN118223842A provides a method for optimizing fracture parameters in horizontal fracture fracturing. The main idea is to use seismic data and existing well logging data to predict the size and orientation of the sand body in the block, determine the long axis extension direction of the horizontal fracture, obtain the permeability ratio along the long and short axes of the horizontal fracture in the target well, calculate the ratio of the maximum to minimum horizontal stress in the block, determine the ratio of the long and short axis radii of the horizontal fracture in the target well, initially determine the fracture radii along the long and short axes of the horizontal fracture in the target well, optimize the long axis radius and conductivity of the horizontal fracture in the target well, determine the optimal horizontal fracture length and short axis radius of the target well, determine the specific fracturing scale and fluid volume, and, with the optimal horizontal fracture conductivity as the optimization target, determine the specific pumping procedure and sand ratio concentration. This patented technology only designs the fracture length, sand volume, and pumping procedure for the horizontal fracture, and cannot predict production capacity. Summary of the Invention
[0005] In view of this, this disclosure provides a method for predicting the fracturing effect of horizontally fractured oil wells, which solves the problem that existing oil well fracturing effect prediction methods cannot predict the post-fracturing effect of horizontally fractured oil wells.
[0006] To address the aforementioned technical problems, this disclosure provides a method for predicting the fracturing effect of horizontal fractured oil wells, including:
[0007] Based on the sample data of fractured wells with obtained post-fracture effects, the fractured layers are clustered to obtain a classification of fractured layers.
[0008] Identify the main controlling factors affecting the post-fracturing effect of different types of fracturing layers;
[0009] Using the main controlling factors of each type of fracturing layer as independent variables, a corresponding intelligent prediction model for post-fracturing effects is established.
[0010] Based on the fractured well sample data of the target well, clustering is used to determine the fractured layer type of the target well, and the corresponding intelligent prediction model for post-fracture effect is invoked to output the post-fracture effect of the target well.
[0011] In this disclosure and possible embodiments, the method for obtaining the sample data of the fractured wells that have achieved post-fracturing effects includes:
[0012] Obtain raw parameter data of horizontal fracture fracturing and post-fracturing effects in shallow oil wells, and preprocess the raw parameter data.
[0013] In this disclosure and possible embodiments, the original parameter data includes:
[0014] Displacement method, fracturing process, pre-fracturing water cut, effective thickness, porosity, permeability, sandstone thickness, medium-depth oil layer, current formation pressure, well spacing, daily oil production of the pre-fracturing layer, daily water production of the pre-fracturing layer, fracturing pressure, total fluid volume, sand addition volume, number of fractures and effective thickness.
[0015] In this disclosure and possible embodiments, the method for preprocessing the original parameter data includes:
[0016] The text-based feature parameters in the original parameter data are numerically processed. This numerical processing involves dummy variable encoding of the displacement method and fracturing process, as defined below:
[0017] The displacement methods are: polymer injection flooding—
[01] , water injection flooding—
[10] , ternary composite flooding—
[00] ;
[0018] The fracturing processes are: conventional fracturing—
[01] , multi-fracture fracturing—
[10] , and selective fracturing—
[00] ;
[0019] The original parameter data is normalized by uniformly transforming the feature parameter data of different dimensions linearly to the range [0,1]. The calculation formula is as follows:
[0020]
[0021] Among them, X i The original data before normalization; X min The minimum value of feature X before normalization; X max The maximum value of feature X before normalization; X i ′ This is the new data after normalization.
[0022] In this disclosure and possible embodiments, the method for clustering fracturing layers to obtain a fracturing layer classification includes:
[0023] Based on the fracturing layers and fracturing operation attribute values, determine the clustering feature parameters of the fracturing layers and establish a clustering feature dataset;
[0024] Clustering values were set to k = 2, 3, 4, 5, ..., n. The clustering feature dataset was clustered to obtain different fracturing layer categories.
[0025] Calculate the cohesion 'a' and the separation 'b' between each category of data for each K value, and determine the optimal classification result of the fracturing layer based on the cohesion 'a' and the separation 'b'.
[0026] In this disclosure and possible embodiments, the clustering feature parameters include:
[0027] Displacement method, fracturing method, pre-fracturing water cut, porosity, permeability, sandstone thickness, medium-depth oil layer, current formation pressure, well spacing, pre-fracturing daily oil production of the small layer, pre-fracturing daily water production of the small layer, fracturing pressure, total fluid volume, sand addition volume, number of fractures, and post-fracturing daily oil production of the small layer.
[0028] In this disclosure and possible embodiments, the method for determining the optimal classification result of fracturing layers by means of the cohesion a and the separation b includes:
[0029] The clustering evaluation result s for each K value is determined based on the cohesion 'a' and the separation 'b'. The larger s is, the better the clustering effect. The K value corresponding to the maximum s is the best classification result under the multidimensional feature values of the fracturing layer. The formula for calculating the clustering evaluation result s is as follows:
[0030]
[0031] In this disclosure and possible embodiments, the method for determining the main controlling factors affecting post-fracturing effects of different types of fracturing layers includes:
[0032] Based on the fracturing layer classification results, the clustering feature dataset is divided into K classes of sample data. For each class of sample data, the feature parameters X are determined. i Nonlinear correlation between the yield after compression and the output Y;
[0033] The determination of each feature parameter X i The nonlinear correlation between the yield after compression and the output Y is calculated using the following formula for the maximum correlation:
[0034]
[0035] Among them, X ij For each cluster feature data point (i = 1, 2, ..., 16; j = 1, 2, ..., k), Y j For the output after pressing, m j It is the number of cluster feature data samples in each category, I(X) ij ;Y j ) is X ij and Y j Mutual information between them, where ∈ is a threshold representing the proportion of times |XY|<∈ out of the total number of samples;
[0036] According to the formula for calculating the maximum correlation, the correlation coefficient between each feature parameter of each type of sample data and the daily oil production of the post-pressurization layer is calculated, and the feature with the larger correlation coefficient is selected as the main controlling factor affecting the post-pressurization effect.
[0037] In this disclosure and possible embodiments, the method for establishing a corresponding intelligent prediction model of post-fracturing effects using the main controlling factors of each type of fracturing layer as independent variables includes:
[0038] The input set is composed of the feature parameters corresponding to the main control factors affecting the post-pressing effect, and the daily oil production of the post-pressed layer is used as the output set. Together, they constitute the post-pressing effect dataset.
[0039] The post-pressure effect dataset is randomly divided into a training set and a test set. Multiple self-service sample sets are generated using a self-service sampling method. A post-pressure effect intelligent prediction model is constructed for each self-service sample set and trained. The prediction results of each post-pressure effect intelligent prediction model are aggregated to obtain the final prediction result.
[0040] In this disclosure and possible embodiments, a method for randomly dividing the post-pressure effect dataset into a training set and a test set includes:
[0041] The dataset of post-pressure effects is randomly divided into subsets, which include a training set and a test set. The ratio of the two sets is set to (10-N):N (N<5), where the subset with more samples is used as the training set and the subset with fewer samples is used as the test set.
[0042] In this disclosure and possible embodiments, the process of generating multiple bootstrap sample sets using the bootstrap sampling method is as follows:
[0043] For an original training dataset containing N samples, the bootstrap sampling process is as follows: randomly select N samples with replacement from the original dataset to form a bootstrap sample set.
[0044] In this disclosure and possible embodiments, the calculation formula for aggregating the prediction results of each intelligent prediction model for post-pressure effects to obtain the final prediction result is:
[0045]
[0046] in, This represents the prediction result of the i-th intelligent prediction model for post-pressure effects, and N represents the number of intelligent prediction models for post-pressure effects.
[0047] This disclosure has the following beneficial effects:
[0048] This disclosed method for predicting the fracturing effect of horizontal fractured oil wells establishes intelligent prediction models for the post-fracturing effects of various fracturing layers based on pre-fracturing dynamic and static parameters, reservoir conditions, and fracturing technology. These models are used to predict the post-fracturing effects of horizontal fractures, reflecting the production increase law of horizontal fracture fracturing. The method predicts the corresponding post-fracturing oil increase for different fracture lengths and can obtain the horizontal fracture length at the maximum oil increase. This effectively solves the problem that existing oil well fracturing effect prediction methods cannot predict the post-fracturing effects of horizontal fractured oil wells. Applying this prediction method, 30 wells with different displacement methods were selected for optimized design verification. Comparing the post-fracturing production, the average relative error rate of the calculated production was only 9.1%, providing accurate technical support for predicting the post-fracturing effects of horizontal fractures. Attached Figure Description
[0049] The above and other objects, features, and advantages of this disclosure will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:
[0050] Figure 1 This disclosure includes a schematic diagram of the intelligent prediction model for post-pressurization capacity;
[0051] Figure 2 This disclosure includes a schematic diagram illustrating the application principle of the intelligent model for post-pressurization capacity.
[0052] Figure 3 This disclosure presents an application example of the optimal penetration ratio for oil production during fracturing of the S2_1 layer in well C27-E39.
[0053] Figure 4 This disclosure provides an example of the optimal penetration ratio for fracturing the S2_2_1 to S2_3 layers of well C27-E39 to achieve optimal oil production.
[0054] Figure 5 This disclosure applies to the optimal penetration ratio during fracturing of the S2_4_1 to S2_4_2 layers in well C27-E39 for oil production.
[0055] Figure 6 This disclosure provides an example of the optimal penetration ratio for fracturing the S2_8_1 to S2_8_3 layers of well C27-E39 to achieve optimal oil production.
[0056] Figure 7 This disclosure presents an application example of the optimal penetration ratio for oil production during fracturing of the S2_9 layer in well C27-E39. Detailed Implementation
[0057] The present disclosure is described below based on embodiments; however, it is worth noting that the present disclosure is not limited to these embodiments. In the detailed description of the present disclosure below, certain specific details are described in detail. However, those skilled in the art will fully understand the present disclosure for the parts not described in detail.
[0058] Furthermore, unless the context explicitly requires it, the words "comprising," "including," and similar terms throughout the specification and claims should be interpreted as including rather than exclusive or exhaustive; that is, meaning "including but not limited to."
[0059] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and examples of embodiments and applications.
[0060] Figure 1 This disclosure includes a schematic diagram of the intelligent prediction model for post-pressurization capacity; Figure 2 This disclosure presents a schematic diagram illustrating the application principle of the intelligent model for post-pressurization capacity. Combined with... Figure 2 and Figure 3 As shown in this disclosure, the method for predicting the fracturing effect of horizontal fractured oil wells includes the following steps:
[0061] (1) Obtain the raw parameter data of horizontal fracture fracturing and post-fracturing effects in oil wells, and preprocess the raw parameter data as follows:
[0062] 1) In this embodiment of the disclosure, the original parameter data that needs to be obtained specifically includes: displacement method, fracturing process, pre-fracturing water cut, effective thickness, porosity, permeability, sandstone thickness, medium-depth oil layer, current formation pressure, well spacing, pre-fracturing small layer daily oil production, pre-fracturing small layer daily water production, fracture pressure, total fluid volume, sand addition volume, number of fractures, and post-fracturing small layer daily oil production.
[0063] 2) The text-based feature parameters in the original parameter data are numerically processed, specifically by dummy encoding the displacement method and fracturing process, as defined below:
[0064] The displacement methods are: polymer injection flooding—
[01] , water injection flooding—
[10] , ternary composite flooding—
[00] ;
[0065] The fracturing processes are: conventional fracturing—
[01] , multi-fracture fracturing—
[10] , and selective fracturing—
[00] ;
[0066] 3) The original parameter data is normalized, specifically by uniformly linearly transforming the feature parameter data of different dimensions to the range [0,1]. The calculation formula is as follows:
[0067]
[0068] Among them, X i The original data before normalization; X min The minimum value of feature X before normalization; X max The maximum value of feature X before normalization; X i ′ This is the new data after normalization.
[0069] (2) Taking the fracturing layers as the object, the fracturing layers are classified based on multi-dimensional characteristics, as follows:
[0070] 1) Based on the fracturing layers and fracturing operation attribute values, determine the clustering characteristic parameters of the fracturing layers and establish a clustering characteristic dataset. In this embodiment of the disclosure, the clustering characteristic parameters include: displacement method, fracturing method, pre-fracturing water cut, porosity, permeability, sandstone thickness, oil layer depth, current formation pressure, well spacing, pre-fracturing small layer daily oil production, pre-fracturing small layer daily water production, fracturing pressure, total fluid volume, sand addition volume, number of fractures, and effective thickness.
[0071] 2) Set the clustering values to k = 2, 3, 4, 5, ..., n, and cluster the clustering feature dataset to obtain different fracturing layer categories;
[0072] 3) Calculate the cohesion *a* and separation *b* between each class of data for each K value. Based on the cohesion *a* and separation *b*, determine the clustering evaluation result *s* for each K value. A larger *s* corresponds to a better clustering effect. The K value corresponding to the largest *s* is the optimal classification result under the multidimensional feature values of the fracturing layer. The formula for calculating the clustering evaluation result *s* is:
[0073]
[0074] (3) Identify the main controlling factors affecting the post-fracturing effect of different types of fracturing layers, as follows:
[0075] Based on the fracturing layer classification results in step (2), the clustering feature dataset is divided into K classes of sample data. For each class of sample data, the feature parameters X are determined. i The nonlinear correlation between the yield after compression and the output Y.
[0076] In this embodiment of the disclosure, the feature parameter X i Specifically, this includes displacement method, fracturing method, pre-fracturing water cut, effective thickness, porosity, permeability, sandstone thickness, oil layer depth, current formation pressure, well spacing, pre-fracturing daily oil production, pre-fracturing daily water production, fracturing pressure, total fluid volume, sand addition volume, number of fractures, and post-fracturing daily oil production.
[0077] In this embodiment of the disclosure, each characteristic parameter X is determined. i The nonlinear correlation between the yield after compression and the output Y is calculated using the following formula for the maximum correlation:
[0078]
[0079] Among them, X ij For each cluster feature data point (i = 1, 2, ..., 16; j = 1, 2, ..., k), Y j For the output after pressing, m j It is the number of cluster feature data samples in each category, I(X) ij ;Y j ) is X ij and Y j The mutual information between them, ∈ is the threshold, representing the proportion of times |XY|<∈ out of the total number of samples.
[0080] According to the above formula for calculating the maximum correlation, the correlation coefficient between each feature parameter of each type of sample data and the daily oil production of the post-pressurization layer is calculated. The feature with a larger correlation coefficient is selected as the main controlling factor affecting the post-pressurization effect. The corresponding feature parameters constitute the input set, and the daily oil production of the post-pressurization layer is used as the output set. The two together constitute the sample set as the post-pressurization effect dataset.
[0081] (4) Using the main controlling factors of each type of fracturing layer as independent variables, an intelligent prediction model for post-fracturing effects is established, as follows:
[0082] 1) Randomly divide the post-compression effect dataset into two mutually exclusive subsets, namely the training set R and the test set T, with the ratio of (10-N):N (N<5). The subset with more samples is used as the training set, and the subset with fewer samples is used as the test set.
[0083] 2) Generate multiple bootstrap sample sets using the bootstrap sampling method. For an original training dataset containing N samples, the bootstrap sampling process can be represented as: randomly selecting N samples with replacement from the original dataset to form a bootstrap sample set.
[0084] 3) Construct and train an intelligent prediction model for the post-pressure effect for each self-help sample set. Each intelligent prediction model for the post-pressure effect can be trained on the self-help sample set and obtain the corresponding prediction results.
[0085] 4) The prediction results of each intelligent prediction model for post-pressure effects are aggregated to obtain the final prediction result.
[0086]
[0087] in, This represents the prediction result of the i-th intelligent prediction model for post-pressure effects, and N represents the number of intelligent prediction models for post-pressure effects.
[0088] The above calculations disclosed herein can be implemented using mathematical models and computational software, specifically based on commonly used fracturing effect analysis and process optimization design platforms in this field.
[0089] Based on the method for predicting the fracturing effect of horizontal fractured oil wells, this disclosure selects 30 wells with different displacement methods for optimization design verification. The results are shown in Table 1. Compared with the production after fracturing, the average relative error rate of production is calculated to be 9.1%.
[0090] Table 1 Comparison of daily oil production after verification well pressure with calculated values.
[0091]
[0092]
[0093]
[0094] Application examples
[0095] Using the "Fracturing Effect Analysis and Process Optimization Design Platform," parameters such as sandstone thickness, effective thickness, and permeability of each fracturing layer in wells C27-E39 were input. Based on the planned oil production at different penetration ratios according to the model, the penetration ratio with the maximum oil production was selected for construction. The results are shown in Table 2. Figures 3-7 As shown.
[0096] Table 2 Reservoir properties and design penetration ratio of fracturing layers C27-E39
[0097]
[0098] The C27-E39 well produced 12.32 tons of oil per day after fracturing, which was only 7.3% lower than the predicted daily production of 11.48 tons.
[0099] The embodiments described above are merely illustrative of implementation methods of this disclosure, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent disclosure. It should be noted that those skilled in the art can make various modifications, equivalent substitutions, and improvements without departing from the concept of this disclosure, and these all fall within the protection scope of this disclosure. Therefore, the protection scope of this patent disclosure should be determined by the appended claims.
Claims
1. A method for predicting the fracturing effect of horizontally fractured oil wells, characterized in that, include: Based on the sample data of fractured wells with obtained post-fracture effects, the fractured layers are clustered to obtain a classification of fractured layers. Identify the main controlling factors affecting the post-fracturing effect of different types of fracturing layers; Using the main controlling factors of each type of fracturing layer as independent variables, a corresponding intelligent prediction model for post-fracturing effects is established. Based on the fractured well sample data of the target well, clustering is used to determine the fractured layer type of the target well, and the corresponding intelligent prediction model for post-fracture effect is invoked to output the post-fracture effect of the target well.
2. The method for predicting the fracturing effect of horizontal fractured oil wells according to claim 1, characterized in that, The method for obtaining the sample data of fractured wells that have achieved post-fracturing effects includes: Obtain raw parameter data of horizontal fracture fracturing and post-fracturing effects in shallow oil wells, and preprocess the raw parameter data.
3. The method for predicting the fracturing effect of horizontal fractured oil wells according to claim 2, characterized in that, The original parameter data includes: Displacement method, fracturing process, pre-fracturing water cut, effective thickness, porosity, permeability, sandstone thickness, medium-depth oil layer, current formation pressure, well spacing, daily oil production of the pre-fracturing layer, daily water production of the pre-fracturing layer, fracturing pressure, total fluid volume, sand addition volume, number of fractures and effective thickness.
4. The method for predicting the fracturing effect of horizontal fractured oil wells according to claim 3, characterized in that, The method for preprocessing the original parameter data includes: The text-based feature parameters in the original parameter data are numerically processed. This numerical processing involves dummy variable encoding of the displacement method and fracturing process, as defined below: The displacement methods are: polymer injection flooding—[01], water injection flooding—[10], ternary composite flooding—[00]; The fracturing processes are: conventional fracturing—[01], multi-fracture fracturing—[10], and selective fracturing—[00]; The original parameter data is normalized by uniformly transforming the feature parameter data of different dimensions linearly to the range [0,1]. The calculation formula is as follows: Among them, X i The original data before normalization; X min The minimum value of feature X before normalization; X max The maximum value of feature X before normalization; X i ′ This is the new data after normalization.
5. The method for predicting the fracturing effect of horizontal fractured oil wells according to any one of claims 1-4, characterized in that, The method for clustering fracturing layers to obtain fracturing layer classification includes: Based on the fracturing layers and fracturing operation attribute values, determine the clustering feature parameters of the fracturing layers and establish a clustering feature dataset; Clustering values were set to k = 2, 3, 4, 5, ..., n. The clustering feature dataset was clustered to obtain different fracturing layer categories. Calculate the cohesion 'a' and the separation 'b' between each category of data for each K value, and determine the optimal classification result of the fracturing layer based on the cohesion 'a' and the separation 'b'.
6. The method for predicting the fracturing effect of horizontal fractured oil wells according to claim 5, characterized in that, The clustering feature parameters include: Displacement method, fracturing method, pre-fracturing water cut, porosity, permeability, sandstone thickness, medium-depth oil layer, current formation pressure, well spacing, pre-fracturing daily oil production of the small layer, pre-fracturing daily water production of the small layer, fracturing pressure, total fluid volume, sand addition volume, number of fractures, and post-fracturing daily oil production of the small layer.
7. The method for predicting the fracturing effect of horizontal fractured oil wells according to claim 6, characterized in that, The method for determining the optimal classification result of fracturing layers using the cohesion a and the separation degree b includes: The clustering evaluation result s for each K value is determined based on the cohesion 'a' and the separation 'b'. The larger s is, the better the clustering effect. The K value corresponding to the maximum s is the best classification result under the multidimensional feature values of the fracturing layer. The formula for calculating the clustering evaluation result s is as follows:
8. The method for predicting the fracturing effect of horizontal fractured oil wells according to claim 6 or 7, characterized in that, The method for determining the main controlling factors affecting the post-fracturing effect of different types of fracturing layers includes: Based on the fracturing layer classification results, the clustering feature dataset is divided into K classes of sample data. For each class of sample data, the feature parameters X are determined. i Nonlinear correlation between the yield after compression and the output Y; The determination of each feature parameter X i The nonlinear correlation between the yield after compression and the output Y is calculated using the following formula for the maximum correlation: Among them, X ij For each cluster feature data point (i = 1, 2, ..., 16; j = 1, 2, ..., k), Y j For the output after pressing, m j It is the number of cluster feature data samples in each category, I(X) ij ;Y j ) is X ij and Y j Mutual information between them, where ∈ is a threshold representing the proportion of times |XY|<∈ out of the total number of samples; According to the formula for calculating the maximum correlation, the correlation coefficient between each feature parameter of each type of sample data and the daily oil production of the post-pressurization layer is calculated, and the feature with the larger correlation coefficient is selected as the main controlling factor affecting the post-pressurization effect.
9. The method for predicting the fracturing effect of horizontal fractured oil wells according to claim 8, characterized in that, The method for establishing a corresponding intelligent prediction model for post-fracturing effects, using the main controlling factors of each type of fracturing layer as independent variables, includes: The input set is composed of the feature parameters corresponding to the main control factors affecting the post-pressing effect, and the daily oil production of the post-pressed layer is used as the output set. Together, they constitute the post-pressing effect dataset. The post-pressure effect dataset is randomly divided into a training set and a test set. Multiple self-service sample sets are generated using a self-service sampling method. A post-pressure effect intelligent prediction model is constructed for each self-service sample set and trained. The prediction results of each post-pressure effect intelligent prediction model are aggregated to obtain the final prediction result.
10. The method for predicting the fracturing effect of horizontal fractured oil wells according to claim 9, characterized in that, The method for randomly dividing the post-pressure effect dataset into training and test sets includes: The dataset of post-pressure effects is randomly divided into subsets, which include a training set and a test set. The ratio of the two sets is set to (10-N):N (N<5), where the subset with more samples is used as the training set and the subset with fewer samples is used as the test set.
11. The method for predicting the fracturing effect of horizontal fractured oil wells according to claim 9, characterized in that, The process of generating multiple bootstrap sample sets using the bootstrap sampling method is as follows: For an original training dataset containing N samples, the bootstrap sampling process is as follows: randomly select N samples with replacement from the original dataset to form a bootstrap sample set.
12. The method for predicting the fracturing effect of horizontal fractured oil wells according to any one of claims 9-11, characterized in that, The formula for calculating the final prediction result by aggregating the prediction results of each intelligent prediction model for post-pressure effects is as follows: in, This represents the prediction result of the i-th intelligent prediction model for post-pressure effects, and N represents the number of intelligent prediction models for post-pressure effects.