A method, apparatus, electronic device and storage medium for predicting formation pore pressure

Abstract

This application provides a method, apparatus, electronic device, and storage medium for predicting formation pore pressure, relating to the field of oil and gas resource exploration and development technology. The method includes: obtaining depth, deep and shallow resistivity, formation density, and acoustic velocity; preprocessing these data to obtain preprocessed depth, deep and shallow resistivity, density, and velocity; obtaining a dynamic gating signal based on the preprocessed deep and shallow resistivity; generating an acoustic velocity-formation density scatter plot based on the preprocessed density and velocity; determining the formation category from the scatter plot and adding a causal label to the category; training a pre-built convolutional neural network based on the preprocessed depth, deep and shallow resistivity, density, velocity, gating signal, and labels to obtain a predictive convolutional neural network; and inputting the well logging data to be predicted into the predictive convolutional neural network to obtain the predicted formation pore pressure value. This achieves deep fusion of multi-source well logging data and geological parameters, and the predictive convolutional neural network can learn the coupling relationships between features, significantly improving the accuracy of pore pressure prediction.

Description

Technical Field

[0001] This application relates to the field of oil and gas resource exploration and development technology, and in particular to a method, apparatus, electronic device and storage medium for predicting formation pore pressure. Background Technology

[0002] In the process of oil exploration and development, formation pore pressure is a core parameter for well control safety. Accurate prediction of formation pore pressure can effectively prevent accidents such as well blowouts and provide important decision support for drilling and completion operations.

[0003] Currently, a common method for predicting formation pore pressure is to directly select a single physical parameter, such as acoustic velocity or compensated density, as an input feature and input it into a standard Long Short-Term Memory (LSTM) network model. This model first receives input features sequentially through the input layer, then the hidden layer uses a unique gating mechanism to extract and learn long-term dependencies and local fluctuations in the input features. Finally, the output layer integrates and processes the data to generate point-by-point predicted formation pore pressure values. However, this method uses only a single physical parameter as an input feature, failing to achieve deep fusion of multi-source data and geological parameters. This results in insufficient learning of the coupling relationships between features, ultimately leading to low prediction accuracy of the trained model for formation pore pressure. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, and storage medium for predicting formation pore pressure, in order to solve the problem of low prediction accuracy caused by existing methods for predicting formation pore pressure.

[0005] In a first aspect, this application provides a method for predicting formation pore pressure, including:

[0006] Obtain the depth, deep resistivity, shallow resistivity, formation density, and sonic velocity of the target well.

[0007] The depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity are preprocessed to obtain the preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, and preprocessed acoustic velocity.

[0008] The dynamic gating signal is obtained based on the preprocessed deep resistivity and the preprocessed shallow resistivity;

[0009] Based on the pre-processed formation density and the pre-processed acoustic velocity, a scatter plot of acoustic velocity-formation density is obtained.

[0010] A predetermined number of genetic categories are obtained based on the acoustic velocity-formation density scatter plot, and a genetic label is added to each genetic category;

[0011] Based on the preprocessed depth, the preprocessed deep resistivity, the preprocessed shallow resistivity, the preprocessed formation density, the preprocessed acoustic velocity, the dynamic gating signal, and the causal label, a pre-constructed convolutional neural network is trained to obtain a predictive convolutional neural network; wherein, the convolutional neural network includes a channel attention module and a spatial attention module;

[0012] The well logging data to be predicted is acquired and input into the predictive convolutional neural network to obtain the predicted formation pore pressure value.

[0013] In one possible design, training a pre-constructed convolutional neural network based on the pre-processed depth, the pre-processed deep resistivity, the pre-processed shallow resistivity, the pre-processed formation density, the pre-processed acoustic velocity, the dynamic gating signal, and the causal label to obtain a predictive convolutional neural network includes:

[0014] Based on the preprocessed depth, the preprocessed deep resistivity, the preprocessed shallow resistivity, the preprocessed formation density, the preprocessed acoustic velocity, and the causal label, a multi-scale feature description is obtained;

[0015] The multi-scale feature description and the dynamic gating signal are fused to obtain fused data;

[0016] The fused data is input into the fully connected layer to obtain the predicted formation pore pressure.

[0017] In one possible design, the process of obtaining a multi-scale feature description based on the preprocessed depth, the preprocessed deep resistivity, the preprocessed shallow resistivity, the preprocessed formation density, the preprocessed acoustic velocity, and the genetic label includes:

[0018] One-dimensional convolution operation is performed on the pre-processed depth, the pre-processed deep resistivity, the pre-processed shallow resistivity, the pre-processed formation density, the pre-processed acoustic velocity, and the genetic label to obtain local logging features.

[0019] The local logging features are subjected to maximum pooling and average pooling to obtain a multi-scale feature description.

[0020] In one possible design, obtaining the dynamic gating signal based on the preprocessed deep resistivity and the preprocessed shallow resistivity includes:

[0021] The resistivity ratio index is obtained based on the pretreated deep resistivity and the pretreated shallow resistivity.

[0022] The dynamic gating signal is obtained based on the resistivity ratio index.

[0023] In one possible design, obtaining the dynamic gating signal based on the resistivity ratio index includes:

[0024] The resistivity ratio index is input into the fully connected layer of the preset gated subnetwork to obtain a linear change result;

[0025] The activation function of the preset gating subnetwork is used to activate the linear change result to obtain a dynamic gating signal.

[0026] In one possible design, obtaining the acoustic velocity-formation density scatter plot based on the preprocessed formation density and the preprocessed acoustic velocity includes:

[0027] Based on the pre-processed formation density, the pre-processed acoustic velocity, and the acoustic velocity-formation density cross-plot discrimination theory, an acoustic velocity-formation density scatter plot is obtained.

[0028] In one possible design, the preprocessing of the depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity to obtain preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, and preprocessed acoustic velocity includes:

[0029] The depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity are subjected to minimum-maximum normalization to obtain the pre-processed depth, pre-processed deep resistivity, pre-processed shallow resistivity, pre-processed formation density, and pre-processed acoustic velocity.

[0030] Secondly, this application provides a formation pore pressure prediction device, comprising:

[0031] The well logging data acquisition module is used to acquire the depth, deep resistivity, shallow resistivity, formation density, and sonic velocity of the target well.

[0032] The preprocessing module is used to preprocess the depth, the deep resistivity, the shallow resistivity, the formation density, and the acoustic velocity to obtain the preprocessed depth, the preprocessed deep resistivity, the preprocessed shallow resistivity, the preprocessed formation density, and the preprocessed acoustic velocity.

[0033] A dynamic gating signal acquisition module is used to obtain a dynamic gating signal based on the preprocessed deep resistivity and the preprocessed shallow resistivity;

[0034] The scatter plot acquisition module is used to obtain a scatter plot of acoustic velocity-formation density based on the preprocessed formation density and the preprocessed acoustic velocity.

[0035] The causal label acquisition module is used to obtain a preset number of causal categories based on the acoustic velocity-formation density scatter plot, and add a causal label to each causal category;

[0036] The training module is used to train a pre-constructed convolutional neural network model based on the pre-processed depth, the pre-processed deep resistivity, the pre-processed shallow resistivity, the pre-processed formation density, the pre-processed acoustic velocity, the dynamic gating signal, and the causal label to obtain a predictive convolutional neural network; wherein, the convolutional neural network includes a channel attention module and a spatial attention module;

[0037] The prediction module is used to acquire the logging data to be predicted and input the logging data to be predicted into the prediction convolutional neural network to obtain the formation pore pressure prediction value.

[0038] Thirdly, this application provides an electronic device, including: a processor and a memory communicatively connected to the processor;

[0039] The memory stores computer-executed instructions;

[0040] The processor executes computer execution instructions stored in the memory to implement the formation pore pressure prediction method provided in the first aspect of this application.

[0041] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the formation pore pressure prediction method provided in the first aspect of this application.

[0042] Fifthly, this application provides a computer program product, including a computer program, which, when executed by a processor, is used to implement the formation pore pressure prediction method provided in the first aspect of this application.

[0043] This application provides a method, apparatus, electronic device, and storage medium for predicting formation pore pressure, including: acquiring the depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity of a target well; preprocessing the depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity to obtain preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, and preprocessed acoustic velocity; obtaining a dynamic gating signal based on the preprocessed deep resistivity and preprocessed shallow resistivity; and obtaining a dynamic gating signal based on the preprocessed formation density and preprocessed acoustic velocity. The acoustic velocity-formation density scatter plot is obtained. Based on the scatter plot, a predetermined number of causal categories are determined, and a causal label is added to each category. A pre-constructed convolutional neural network is trained using pre-processed depth, pre-processed deep resistivity, pre-processed shallow resistivity, pre-processed formation density, pre-processed acoustic velocity, dynamic gating signal, and causal labels to obtain a predictive convolutional neural network. The well logging data to be predicted is acquired and input into the predictive convolutional neural network to obtain the predicted formation pore pressure value. Through this method, the following technical effects are achieved: by gating the model with deep resistivity, shallow resistivity, formation density, and acoustic velocity, deep fusion of multi-source well logging data and geological parameters is realized. Simultaneously, the trained predictive convolutional neural network can better learn the coupling relationships between features, significantly improving the accuracy of formation pore pressure prediction. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 A flowchart illustrating the formation pore pressure prediction method provided in this application embodiment. Figure 1 ;

[0046] Figure 2 The relationship between acoustic velocity and formation density provided in the embodiments of this application Figure 1 ;

[0047] Figure 3 The relationship between acoustic velocity and formation density provided in the embodiments of this application Figure 2 ;

[0048] Figure 4 A schematic diagram illustrating the cause categories provided in the embodiments of this application;

[0049] Figure 5A graph showing the relationship between the loss value of the first fold and the number of training rounds in the five-fold cross-validation strategy provided in this application embodiment;

[0050] Figure 6 A graph showing the relationship between the loss value of the second fold and the number of training rounds in the five-fold cross-validation strategy provided in this application embodiment;

[0051] Figure 7 A graph showing the relationship between the loss value of the third fold and the number of training rounds in the five-fold cross-validation strategy provided in this application embodiment;

[0052] Figure 8 A graph showing the relationship between the loss value of the fourth fold and the number of training rounds in the five-fold cross-validation strategy provided in this application embodiment;

[0053] Figure 9 A graph showing the relationship between the loss value of the fifth fold and the number of training rounds in the five-fold cross-validation strategy provided in this application embodiment;

[0054] Figure 10 The relationship between the predicted formation pore pressure and the actual measured formation pore pressure provided in the embodiments of this application Figure 1 ;

[0055] Figure 11 The relationship between the predicted formation pore pressure and the actual measured formation pore pressure provided in the embodiments of this application Figure 2 ;

[0056] Figure 12 The relationship between the predicted formation pore pressure and the actual measured formation pore pressure provided in the embodiments of this application Figure 3 ;

[0057] Figure 13 The relationship between the predicted formation pore pressure and the actual measured formation pore pressure provided in the embodiments of this application Figure 4 ;

[0058] Figure 14 The relationship between the predicted formation pore pressure and the actual measured formation pore pressure provided in the embodiments of this application Figure 5 ;

[0059] Figure 15 The vertical formation pore pressure prediction results provided for the embodiments of this application;

[0060] Figure 16 A flowchart illustrating the formation pore pressure prediction method provided in this application embodiment. Figure 2 ;

[0061] Figure 17 A schematic diagram of the formation pore pressure prediction device provided in the embodiments of this application;

[0062] Figure 18 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0063] Explanation of reference numerals in the attached figures:

[0064] 901 - Processor; 902 - Memory; 903 - Communication components; 904 - Bus. Detailed Implementation

[0065] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0066] In the embodiments of this application, the terms "first" and "second" are used to distinguish identical or similar items with substantially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that "first" and "second" do not necessarily imply difference. It should be noted that in the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner. In the embodiments of this application, "at least one" refers to one or more, and "more than one" refers to two or more.

[0067] It should be noted that the formation pore pressure prediction method provided in this application embodiment is only an example, and the formation pore pressure prediction method may include more or less content.

[0068] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with relevant laws, regulations and standards, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0069] First, some terms used in the embodiments of this application will be explained.

[0070] Formation pore pressure: The pressure exerted by fluids in formation pores during sediment compaction, also known as pore fluid pressure.

[0071] In order to clearly understand the technical solution of this application, the solutions of the prior art will be described in detail.

[0072] In existing technologies, a common method for predicting formation pore pressure is to directly select a single physical parameter, such as acoustic velocity or compensated density, as an input feature and input it into a standard long short-term memory (LSM) network model. This model first receives input features sequentially through the input layer, then the hidden layer uses a unique gating mechanism to extract and learn long-term dependencies and local fluctuations in the input features. Finally, the output layer integrates and processes the data to generate point-by-point predicted formation pore pressure values. This method, using only a single physical parameter as an input feature, fails to achieve deep fusion of multi-source data and geological parameters, resulting in insufficient learning of the coupling relationships between features. Ultimately, this leads to low prediction accuracy of the trained model for formation pore pressure.

[0073] In summary, how to design a method that can solve the problem of low prediction accuracy caused by existing methods for predicting formation pore pressure is a problem that this application urgently needs to solve.

[0074] Therefore, in view of the above-mentioned technical problems existing in the prior art, the embodiments of this application provide a formation pore pressure prediction method, device, electronic device and storage medium, which aim to effectively improve the prediction accuracy of formation pore pressure.

[0075] The following describes the application scenarios of the formation pore pressure prediction method, apparatus, electronic device, and storage medium provided in the embodiments of this application. These application scenarios are merely examples, intended to help those skilled in the art understand the technical content of this application, but do not imply that the embodiments of this application cannot be used in other devices, systems, environments, or scenarios.

[0076] Oil and gas drilling engineering: The formation pore pressure prediction method provided in the embodiments of this application can accurately predict the pore pressure of high-pressure formations, guide the design of drilling fluid density, prevent well control accidents such as blowouts and lost circulation, and ensure drilling safety.

[0077] Figure 1 A flowchart illustrating the formation pore pressure prediction method provided in this application embodiment. Figure 1 The formation pore pressure prediction method provided in this embodiment includes the following steps:

[0078] S101. Obtain the depth, deep resistivity, shallow resistivity, formation density, and sonic velocity of the target well.

[0079] In this embodiment, basic logging data of the target well in the drilling area are collected, including depth, deep resistivity, shallow resistivity, formation density and sonic velocity. This logging data is the basic input for the formation pore pressure prediction value.

[0080] S102. Preprocess the depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity to obtain the preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, and preprocessed acoustic velocity.

[0081] In this embodiment, depth is aligned, invalid values ​​are processed, and normalized; deep resistivity and shallow resistivity are processed, smoothed, denoised, logarithmically transformed, and standardized; formation density is processed, smoothed, and standardized; and acoustic velocity is subject to physical range constraints, environmental correction, smoothing, and standardization. This embodiment does not impose specific limitations on the preprocessing method.

[0082] The pre-processed depth, pre-processed deep resistivity, pre-processed shallow resistivity, pre-processed formation density, and pre-processed acoustic velocity are used as characteristic data for predicting formation pore pressure.

[0083] S103. The dynamic gating signal is obtained based on the preprocessed deep resistivity and the preprocessed shallow resistivity.

[0084] As an optional implementation, a resistivity ratio sequence is obtained based on the preprocessed deep resistivity and the preprocessed shallow resistivity. This resistivity ratio sequence is then slidably evaluated using a depth window to obtain a local mean. The local mean is then constrained using a Sigmoid function to obtain a dynamic gating signal. This embodiment does not impose specific limitations on the method of obtaining the dynamic gating signal.

[0085] S104. Based on the pre-processed formation density and the pre-processed acoustic velocity, obtain the acoustic velocity-formation density scatter plot.

[0086] As an optional implementation, first obtain the preprocessed formation density and preprocessed acoustic velocity, ensuring that both have consistent lengths and no missing values. Then, import them into a plotting library and use Python's matplotlib or seaborn library to create a scatter plot with acoustic velocity as the horizontal axis and formation density as the vertical axis. Finally, add labels and units to the horizontal and vertical axes, set the chart title, and output the result to obtain the acoustic velocity-formation density scatter plot. This embodiment does not impose specific limitations on the method of plotting the acoustic velocity-formation density scatter plot.

[0087] S105. Obtain a preset number of genetic categories based on the acoustic velocity-formation density scatter plot, and add a genetic label to each genetic category.

[0088] In this embodiment, after obtaining the scatter plot of acoustic velocity-formation density, the normal compaction trend line is first identified as a baseline, and then the formation categories are classified according to the characteristics of deviation from the trend line. For example, data points closely distributed on or near the normal compaction trend line are classified as formation category one, normally compacted strata; data points that deviate significantly from the normal trend line, with a significant increase in acoustic velocity and a significant decrease in density, are classified as formation category two, under-compacted strata; and data points that deviate from the trend line, with abnormally low density but small changes in acoustic velocity, are classified as formation category three, hydrocarbon-generating strata.

[0089] The following example, using a typical well in a block of the Yinggehai Basin, illustrates the specific steps for obtaining a predetermined number of genetic categories based on the sonic velocity-formation density scatter plot.

[0090] Figure 2 The relationship between acoustic velocity and formation density provided in the embodiments of this application Figure 1 , Figure 3 The relationship between acoustic velocity and formation density provided in the embodiments of this application Figure 2 , Figure 4 A schematic diagram illustrating the cause categories provided in the embodiments of this application. Figure 2 and Figure 3 In the diagram, orange represents the Yellow Flow Group and blue represents the Second Stage of the Oriole Group. Figure 2 , Figure 3 and Figure 4 In the diagram, the horizontal axis represents the formation density, expressed in g / cm³. 3 The vertical axis represents the velocity of sound waves, expressed in km / s. Figure 2 and Figure 3 The trend line indicated by the red arrow in the image is a normal consolidation trend line. Compared to this normal consolidation trend line, Figure 4 The yellow line in the image represents type 1 of the formation, indicating uneven compaction. Figure 4 The red line in the image represents category two of the causes, indicating conduction or fluid expansion. Figure 4 The blue line in the image represents type three of the genesis, indicating tectonic compression. Figure 4 The purple line in the image represents the fourth genetic category, signifying the transformation of clay minerals.

[0091] After obtaining the cause categories, a cause label is added to each cause category. Table 1 is a correspondence table of cause label data provided in the embodiments of this application.

[0092] Table 1

[0093]

[0094] Add cause label 1 to cause category 1, cause label 2 to cause category 2, cause label 3 to cause category 3, and cause label 4 to cause category 4.

[0095] S106. Based on the pre-processed depth, pre-processed deep resistivity, pre-processed shallow resistivity, pre-processed formation density, pre-processed acoustic velocity, dynamic gating signal, and causal label, train the pre-constructed convolutional neural network to obtain the predictive convolutional neural network.

[0096] In this embodiment, the convolutional neural network includes a channel attention module and a spatial attention module.

[0097] In this embodiment, the pre-constructed convolutional neural network includes a channel attention module and a spatial attention module. Introducing a channel attention module in the channel dimension automatically adjusts the importance of each channel feature based on the channel importance of the feature data. This highlights features related to formation pore pressure changes while suppressing unimportant or redundant channel information. Introducing a spatial attention module in the spatial dimension further enhances the model's ability to learn features from locally sensitive areas. It allows the model to adaptively focus on more representative local areas in the spatial dimension, thereby improving the model's ability to identify key features. By adding channel and spatial attention modules, the model's adaptive learning ability for important formation features can be improved. Visual analysis of the feature weights output by the attention modules demonstrates the model's attention to different logging parameters. This further verifies the model's consistent response to physical causes, ensuring the model's physical interpretability.

[0098] As an optional implementation, the preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, preprocessed acoustic velocity, and dynamic gating signal are first aligned by depth. Before sample segmentation, the dynamic gating signal modulates other features through element-wise multiplication to form a two-dimensional feature matrix, which is then strictly matched with the causal label. A fixed depth window is then used as a sample window to slide and truncate the two-dimensional feature matrix, converting it into a three-dimensional tensor. A corresponding causal label is assigned to each sample window. Finally, the three-dimensional tensor is input into a pre-constructed convolutional neural network. Using the causal label as the supervision signal and the mean squared error as the loss function, the network parameters are optimized through backpropagation. The predicted formation pore pressure is compared with the actual measured formation pore pressure to optimize the pre-constructed convolutional neural network until the difference between the predicted and actual values ​​is within a preset threshold, resulting in a trained predictive convolutional neural network. After training, the trained predictive convolutional neural network is saved for regression prediction of formation pore pressure. This embodiment does not impose specific restrictions on the method of training the pre-built convolutional neural network. The dynamic gating signal is used to adjust the weights of the channel features in the convolutional neural network, thereby strengthening the key features related to the formation pore pressure prediction value and suppressing noise features.

[0099] The theoretical relationship between formation density and pore pressure gradient is: P grad ≈0.433×ρ b , where P grad For the pore pressure gradient, ρ b For formation density. During model training, a physical consistency auxiliary loss term based on formation density and pore pressure gradient is introduced. By jointly optimizing prediction error and physical consistency error, it is ensured that the model not only meets the data fitting requirements, but also maintains the rationality and consistency of physical origin.

[0100] This application employs a five-fold cross-validation strategy to train and evaluate the model on well logging samples under different geological conditions, ensuring the model's stability and generalization ability under diverse geological conditions. In each fold of cross-validation, the training and test sets are divided differently, which helps evaluate the model's performance on unseen data.

[0101] Figure 5 This is a graph showing the relationship between the loss value of the first fold and the number of training rounds in the five-fold cross-validation strategy provided in this application embodiment. Figure 6 This is a graph showing the relationship between the loss value of the second fold and the number of training rounds in the five-fold cross-validation strategy provided in this application embodiment. Figure 7 This is a graph showing the relationship between the loss value of the third fold and the number of training rounds in the five-fold cross-validation strategy provided in this application embodiment. Figure 8This is a graph showing the relationship between the loss value of the fourth fold and the number of training rounds in the five-fold cross-validation strategy provided in this application embodiment. Figure 9 The graph shows the relationship between the loss value of the fifth fold and the number of training rounds in the five-fold cross-validation strategy provided in the embodiments of this application. Figure 5 , Figure 6 , Figure 7 , Figure 8 and Figure 9 In the diagram, the yellow curve represents the validation loss value, the blue curve represents the training loss value, the horizontal axis represents the number of training epochs (in epochs), and the vertical axis represents the loss value (dimensionless). The number of training epochs is the number of times the entire training set is traversed once.

[0102] Figure 5 , Figure 6 , Figure 7 , Figure 8 and Figure 9 In the first 50 training epochs, both the validation and training losses dropped sharply from their initial values ​​to below 1, indicating that the model was rapidly learning features. From epoch 50 to 100, the rate of decrease in both validation and training losses slowed, gradually converging to the 0.5-0.8 range. After epoch 100, the curves for both validation and training losses became relatively stable with minimal fluctuations, indicating that the model had reached a stable state without significant overfitting or underfitting. In each epoch of the five-fold cross-validation strategy, the validation and training loss curves almost overlapped, with no significant separation throughout, indicating strong generalization ability and consistent fit to both the training and validation sets. Furthermore, the highly consistent curve shapes, convergence speeds, and final loss values ​​of the validation and training losses in each epoch of the five-fold cross-validation strategy validated the stability of the model training and the reliability of the results.

[0103] During the model evaluation phase, the linear fitting formula and the coefficient of determination R² between the predicted formation pore pressure and the actual measured formation pore pressure are used as the main evaluation indicators. R² is used to measure the model's fit and prediction performance.

[0104] Figure 10 The relationship between the predicted formation pore pressure and the actual measured formation pore pressure provided in the embodiments of this application Figure 1 , Figure 11 The relationship between the predicted formation pore pressure and the actual measured formation pore pressure provided in the embodiments of this application Figure 2 , Figure 12 The relationship between the predicted formation pore pressure and the actual measured formation pore pressure provided in the embodiments of this application Figure 3 , Figure 13 The relationship between the predicted formation pore pressure and the actual measured formation pore pressure provided in the embodiments of this application Figure 4 , Figure 14 The relationship between the predicted formation pore pressure and the actual measured formation pore pressure provided in the embodiments of this application Figure 5 . Figure 10 This is a graph showing the relationship between the predicted formation pore pressure and the actual measured formation pore pressure in the first fold of the five-fold cross-validation strategy. Figure 11 This is a graph showing the relationship between the predicted formation pore pressure and the actual measured formation pore pressure in the second step of the five-fold cross-validation strategy. Figure 12 This is a graph showing the relationship between the predicted formation pore pressure and the actual measured formation pore pressure in the third step of the five-fold cross-validation strategy. Figure 13 This is a graph showing the relationship between the predicted formation pore pressure and the actual measured formation pore pressure in the fourth step of the five-step cross-validation strategy. Figure 14 This is a graph showing the relationship between the predicted formation pore pressure and the actual measured formation pore pressure in the fifth step of the five-step cross-validation strategy.

[0105] Figure 10 , Figure 11 , Figure 12 , Figure 13 and Figure 14 In the graph, the horizontal axis represents the actual measured value of formation pore pressure in psi, and the vertical axis represents the predicted value of formation pore pressure in psi. Figure 10 In the model, the linear fitting formula between the predicted and actual measured values ​​of formation pore pressure is: y = 0.9195x + 388.29, with a coefficient of determination R² = 0.913. Figure 11 In the model, the linear fitting formula between the predicted and actual measured values ​​of formation pore pressure is: y = 0.8864x + 896.27, with a coefficient of determination R² = 0.8498. Figure 12 In the model, the linear fitting formula between the predicted and actual measured values ​​of formation pore pressure is: y = 0.9075x + 619.7, with a coefficient of determination R² = 0.8526. Figure 13 In the model, the linear fitting formula between the predicted and actual measured values ​​of formation pore pressure is: y = 0.8452x + 1136, with a coefficient of determination R² = 0.8797. Figure 14 In the model, the linear fitting formula between the predicted value of formation pore pressure and the actual measured value of formation pore pressure is: y=0.9282x+519.59, with a coefficient of determination R²=0.9433.

[0106] Figure 10 , Figure 11 , Figure 12 , Figure 13 and Figure 14In the linear fitting formula between the predicted and actual measured values ​​of formation pore pressure, the slope is close to 1, indicating a strong correlation and high consistency between the predicted and actual values, which shows that the overall performance of the predictive convolutional neural network is good. The coefficient of determination is high, close to 1, indicating that the predictive convolutional neural network has high prediction accuracy.

[0107] In practical applications, the trained predictive convolutional neural network is used to predict well logging data in new blocks and provides formation pore pressure distribution maps, which can assist in well control design and drilling fluid density planning. This method improves drilling safety and has broad engineering application value and promising prospects for widespread adoption.

[0108] S107. Obtain the logging data to be predicted and input the logging data to be predicted into the predictive convolutional neural network to obtain the formation pore pressure prediction value.

[0109] In this embodiment, after obtaining the trained predictive convolutional neural network, the well logging data to be predicted is input into the predictive convolutional neural network to obtain the formation pore pressure prediction value.

[0110] By gating the model with deep resistivity, shallow resistivity, formation density, and sonic velocity, deep fusion of multi-source logging data and geological parameters was achieved. At the same time, the trained predictive convolutional neural network can learn the coupling relationship between features well, which significantly improves the accuracy of formation pore pressure prediction.

[0111] Figure 15 The vertical formation pore pressure prediction results provided for the embodiments of this application. Figure 15 In the diagram, the vertical axis represents depth in meters (m), deep resistivity in ohms (Ω / m), shallow resistivity in ohms (Ω / m), and formation density in grams per cubic meter (g / cm³). 3 The units are as follows: acoustic velocity is expressed in μs / ft, and predicted formation pore pressure is expressed in psi. To analyze the effectiveness of formation pore pressure prediction, in practical applications, the trained predictive convolutional neural network was applied to predict formation pore pressure data. Taking Well 7 as an example, there are a total of 19 actual measured formation pore pressure data points in this well, providing a vertical distribution map of formation pore pressure. Through single-well map analysis, the trends of the predicted formation pore pressure and the actual measured formation pore pressure are highly consistent. This indicates that the model's prediction results in this well are highly consistent with the actual pressure performance, verifying the effectiveness of the model. This map will serve as a basis for well control design and drilling fluid density planning, effectively improving drilling safety.

[0112] This application provides a method for predicting formation pore pressure, comprising: acquiring the depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity of a target well; preprocessing the depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity to obtain preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, and preprocessed acoustic velocity; obtaining a dynamic gating signal based on the preprocessed deep resistivity and preprocessed shallow resistivity; and obtaining an acoustic velocity based on the preprocessed formation density and preprocessed acoustic velocity. A velocity-formation density scatter plot is generated. Based on this plot, a predetermined number of causal categories are obtained, and a causal label is added to each category. A pre-constructed convolutional neural network is trained using pre-processed depth, pre-processed deep resistivity, pre-processed shallow resistivity, pre-processed formation density, pre-processed acoustic velocity, a dynamic gating signal, and causal labels to obtain a predictive convolutional neural network. The well logging data to be predicted is acquired and input into the predictive convolutional neural network to obtain the predicted formation pore pressure value. This method achieves the following technical effects: by gating the model with deep resistivity, shallow resistivity, formation density, and acoustic velocity, deep fusion of multi-source well logging data and geological parameters is achieved. Simultaneously, the trained predictive convolutional neural network can better learn the coupling relationships between features, significantly improving the accuracy of formation pore pressure prediction.

[0113] Figure 16 A flowchart illustrating the formation pore pressure prediction method provided in this application embodiment. Figure 2 In the formation pore pressure prediction method provided in this embodiment, S106 includes the following steps:

[0114] S201. Based on the preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, preprocessed acoustic velocity, and genetic label, a multi-scale feature description is obtained.

[0115] As an optional implementation, data alignment is first performed on the preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, preprocessed sonic velocity, and causal labels to ensure that all preprocessed logging data and causal labels are strictly matched at the depth sampling points. Then, local statistics are calculated for each logging curve using different window sizes; short-time Fourier transform or wavelet transform is performed on each logging curve to extract frequency domain multi-scale energy features; macroscopic features, including mean, range, and trend slope, are statistically analyzed for the entire well section. Finally, the local statistics, frequency domain multi-scale energy features, and macroscopic features are concatenated by depth points to form a multi-scale feature description matrix; the causal labels are then mapped to this multi-scale feature description matrix at depth points to obtain the multi-scale feature description. This embodiment does not impose specific limitations on the method of obtaining the multi-scale feature description based on the preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, preprocessed sonic velocity, and causal labels.

[0116] S202. The multi-scale feature description and dynamic gating signal are fused to obtain fused data.

[0117] In this embodiment, after obtaining the multi-scale feature description, the multi-scale feature description is fused with the dynamic gating signal to obtain fused data. The dynamic gating signal and the multi-scale feature description are modulated by element-wise multiplication, thereby enabling the model to automatically adjust features under different causes.

[0118] S203. Input the fused data into the fully connected layer to obtain the predicted formation pore pressure.

[0119] In this embodiment, fused data is input into multiple fully connected layers. A linear activation function is used to perform regression prediction of formation pore pressure. Mean squared error is used as the primary loss function to measure the difference between the predicted and actual measured values. A physical consistency loss term is introduced for end-to-end training, and the Adam optimizer is employed during optimization. The regression output is used to generate the predicted formation pore pressure values.

[0120] Based on the preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, preprocessed acoustic velocity, and causal labels, multi-scale feature descriptions are obtained. These multi-scale feature descriptions and dynamic gating signals are then used to train a pre-constructed convolutional neural network (CNN). By using multi-scale feature descriptions and dynamic gating signals to train the CNN model and fusing multi-source logging data, the model's ability to capture the response patterns of formation pore pressure at different depth scales is improved. Furthermore, the dynamic gating signals adaptively enhance key features related to overpressure genesis, effectively improving the model's adaptability to complex geological conditions, mitigating overfitting, and ultimately improving the accuracy of formation pore pressure prediction.

[0121] Based on the above embodiments, this embodiment provides a method for predicting formation pore pressure. In the formation pore pressure prediction method provided in this embodiment, step S201 includes the following steps:

[0122] S301. Perform a one-dimensional convolution operation on the pre-processed depth, pre-processed deep resistivity, pre-processed shallow resistivity, pre-processed formation density, pre-processed sonic velocity, and genetic label to obtain local logging features.

[0123] In this embodiment, the preprocessed well logging data and causal labels are processed by two one-dimensional convolutional layers to obtain local well logging features. Each one-dimensional convolutional layer extracts local features from the data through filters, and these local features can help the model learn the laws governing formation pore pressure changes.

[0124] S302. Perform maximum pooling and average pooling processing on the local features of well logging to obtain multi-scale feature descriptions.

[0125] In this embodiment, after the convolutional layer, the local logging features are subjected to max pooling and average pooling to obtain multi-scale feature descriptions. This operation allows the model to focus on information at different scales, thereby enhancing the model's robustness.

[0126] First, one-dimensional convolution operations are used to obtain local logging features. Then, max pooling and average pooling are used to obtain multi-scale feature descriptions. By extracting subtle local features from the logging data through one-dimensional convolution operations, and combining the ability of max pooling to capture significant local responses with the aggregation effect of average pooling on the overall regional trend, key information from different sensing fields can be systematically integrated to form a multi-scale feature description covering both local details and macro trends. This effectively improves the richness and robustness of feature representation, enhances the model's adaptability to complex formation conditions, reduces the risk of overfitting, and improves prediction accuracy.

[0127] Based on the above embodiments, this embodiment provides a method for predicting formation pore pressure. In the formation pore pressure prediction method provided in this embodiment, step S103 includes the following steps:

[0128] S401. Based on the pretreated deep resistivity and the pretreated shallow resistivity, obtain the resistivity ratio index.

[0129] In this embodiment, the ratio of the pretreated shallow resistivity to the pretreated deep resistivity is used as the resistivity ratio index. The resistivity ratio index is an important characterization parameter reflecting the difference in fluid permeability and the cause of formation pressure.

[0130] S402. Obtain the dynamic gating signal based on the resistivity ratio index.

[0131] As an optional implementation, after obtaining the resistivity ratio index, the resistivity ratio index is input into a pre-trained decision tree model. Starting from the root node, splitting conditions are matched layer by layer until the leaf node is reached. A dynamic gating signal is then output according to the leaf node rules. This embodiment does not impose specific limitations on the method of obtaining the dynamic gating signal from the resistivity ratio index.

[0132] This paper provides a specific method for obtaining dynamic gating signals based on preprocessed deep resistivity and preprocessed shallow resistivity. Based on the sensitivity of resistivity ratio index to formation fluids and combined with the ratio stability reflected by local standard deviation, the method adaptively generates dynamic gating signals that have both genetic orientation and spatial smoothness. This method can accurately locate key formation segments related to fluid properties, dynamically enhance effective feature weights and suppress non-target interference, and significantly improve the accuracy and robustness of the model in predicting formation pore pressure.

[0133] Based on the above embodiments, this embodiment provides a method for predicting formation pore pressure. In the formation pore pressure prediction method provided in this embodiment, step S402 includes the following steps:

[0134] S501. Input the resistivity ratio index into the fully connected layer of the preset gated subnetwork to obtain the linear change result.

[0135] In this embodiment, the preset gated subnetwork is an independent gated subnetwork. First, the resistivity ratio index is input into the fully connected layer of the preset gated subnetwork, and the linear change result is obtained using the following formula:

[0136]

[0137] Where h is the result of linear change, W is the preset weight, x is the resistivity ratio index, and b is the preset bias.

[0138] The weights and biases are initialized with random small data and automatically optimized on the training data through the backpropagation algorithm to obtain the optimized weights and biases. The optimized weights are used as the preset weights W and the optimized biases are used as the preset biases b.

[0139] S502. Using the activation function of the preset gating sub-network, the linear change result is activated to obtain the dynamic gating signal.

[0140] In this embodiment, after obtaining the linear transformation result, the linear transformation result is activated using the Sigmoid activation function to obtain the dynamic gating signal. The calculation formula is as follows:

[0141]

[0142] Where g is the dynamic gating signal and h is the linear change result.

[0143] A specific method for obtaining dynamic gating signals based on resistivity ratio is provided, which can improve the efficiency and accuracy of obtaining dynamic gating signals.

[0144] Based on the above embodiments, this embodiment provides a method for predicting formation pore pressure. In the formation pore pressure prediction method provided in this embodiment, S104 includes the following steps:

[0145] S601. Based on the pre-processed formation density, the pre-processed acoustic velocity, and the acoustic velocity-formation density cross-plot discrimination theory, an acoustic velocity-formation density scatter plot is obtained.

[0146] In this embodiment, based on the acoustic velocity-formation density cross-plot discrimination theory proposed by Bowers' method, and using pre-processed formation density and pre-processed acoustic velocity, an acoustic velocity-formation density scatter plot is generated, which can analyze the trend characteristics corresponding to different formation pressure causes. This scatter plot analysis can help the model accurately capture sensitive features, thereby improving the physical interpretability and accuracy stability of formation pore pressure prediction.

[0147] Based on the above embodiments, this embodiment provides a method for predicting formation pore pressure. In the formation pore pressure prediction method provided in this embodiment, S102 includes the following steps:

[0148] S701. Perform minimum-maximum normalization on depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity to obtain pre-processed depth, pre-processed deep resistivity, pre-processed shallow resistivity, pre-processed formation density, and pre-processed acoustic velocity.

[0149] In this embodiment, the minimum-maximum normalization method is used to standardize the depth, deep resistivity, shallow resistivity, formation density, and sonic velocity, standardizing all feature data to the [0,1] interval. This ensures that the contribution of logging data to model training is equal, thereby eliminating dimensional differences between different logging data.

[0150] Figure 17 This is a schematic diagram of the formation pore pressure prediction device provided in an embodiment of this application. Figure 17 As shown, in this embodiment, the formation pore pressure prediction device can be located in an electronic device. The formation pore pressure prediction device includes:

[0151] The well logging data acquisition module 801 is used to acquire the depth, deep resistivity, shallow resistivity, formation density, and sonic velocity of the target well.

[0152] The preprocessing module 802 is used to preprocess the depth, deep resistivity, shallow resistivity, formation density and acoustic velocity to obtain the preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density and preprocessed acoustic velocity.

[0153] The dynamic gating signal acquisition module 803 is used to obtain the dynamic gating signal based on the preprocessed deep resistivity and the preprocessed shallow resistivity.

[0154] The scatter plot acquisition module 804 is used to obtain a scatter plot of acoustic velocity-formation density based on the preprocessed formation density and the preprocessed acoustic velocity.

[0155] Genetic label acquisition module 805 is used to obtain a preset number of genetic categories based on the sonic velocity-formation density scatter plot, and add genetic labels to each genetic category;

[0156] Training module 806 is used to train a pre-constructed convolutional neural network model based on the pre-processed depth, pre-processed deep resistivity, pre-processed shallow resistivity, pre-processed formation density, pre-processed acoustic velocity, dynamic gating signal and causal label, to obtain a predictive convolutional neural network; wherein, the convolutional neural network includes a channel attention module and a spatial attention module.

[0157] The prediction module 807 is used to acquire the logging data to be predicted and input the logging data to be predicted into the prediction convolutional neural network to obtain the formation pore pressure prediction value.

[0158] The formation pore pressure prediction device provided in this embodiment can perform... Figure 1 The technical solution of the formation pore pressure prediction method embodiment shown herein, its implementation principle and technical effect are similar to Figure 1 The examples of formation pore pressure prediction methods shown are similar and will not be described in detail here.

[0159] Meanwhile, the formation pore pressure prediction device provided by the present invention further refines the formation pore pressure prediction device based on the formation pore pressure prediction device provided in the previous embodiment.

[0160] Optionally, in this embodiment, the training module 806 is further configured to:

[0161] Based on the preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, preprocessed acoustic velocity, and genetic label, a multi-scale feature description is obtained.

[0162] Multi-scale feature descriptions and dynamic gating signals are fused to obtain fused data.

[0163] The fused data is input into the fully connected layer to obtain the predicted formation pore pressure.

[0164] Optionally, in this embodiment, the training module 806 is further configured to:

[0165] One-dimensional convolution operation was performed on the pre-processed depth, pre-processed deep resistivity, pre-processed shallow resistivity, pre-processed formation density, pre-processed sonic velocity, and genetic label to obtain local logging features.

[0166] The local logging features are subjected to maximum pooling and average pooling to obtain multi-scale feature descriptions.

[0167] Optionally, in this embodiment, the dynamic gating signal acquisition module 803 is further used for:

[0168] The resistivity ratio index is obtained based on the pretreated deep resistivity and the pretreated shallow resistivity.

[0169] The dynamic gating signal is obtained based on the resistivity ratio index.

[0170] Optionally, in this embodiment, the dynamic gating signal acquisition module 803 is further used for:

[0171] The resistivity ratio index is input into the fully connected layer of the preset gated subnetwork to obtain a linear change result;

[0172] By using the activation function of a pre-defined gated subnetwork, the linear transformation result is activated to obtain a dynamic gated signal.

[0173] Optionally, in this embodiment, the scatter plot acquisition module 804 is further used for:

[0174] Based on the pre-processed formation density, the pre-processed acoustic velocity, and the discrimination theory of acoustic velocity-formation density cross-plot, a scatter plot of acoustic velocity-formation density is obtained.

[0175] Optionally, in this embodiment, the preprocessing module 802 is further configured to:

[0176] Minimum-maximum normalization was performed on depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity to obtain the pre-processed depth, pre-processed deep resistivity, pre-processed shallow resistivity, pre-processed formation density, and pre-processed acoustic velocity.

[0177] Figure 18This is a schematic diagram of the electronic device provided in an embodiment of this application. The electronic device is intended for use with various electronic devices capable of performing formation pore pressure prediction methods, such as microcomputers, microcontrollers, and other suitable computers. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0178] like Figure 18 As shown, the electronic device includes at least one processor 901 and a memory 902. The electronic device also includes a communication component 903. The processor 901, memory 902, and communication component 903 are connected via a bus 904.

[0179] In the specific implementation process, at least one processor 901 executes computer execution instructions stored in memory 902, causing at least one processor 901 to execute the formation pore pressure prediction method executed on the electronic device side as described above.

[0180] The specific implementation process of processor 901 can be found in the above-described embodiment of the formation pore pressure prediction method. Its implementation principle and technical effect are similar, and will not be repeated here.

[0181] In the above embodiments, it should be understood that the processor 901 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor 901 can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0182] The memory 902 may include high-speed RAM memory, and may also include non-volatile memory (NVM), such as at least one disk storage.

[0183] Bus 904 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Bus 904 can be divided into address bus, data bus, control bus, etc. For ease of illustration, the bus 904 in the accompanying drawings of this application is not limited to only one bus or one type of bus.

[0184] The above description addresses the functions implemented by electronic devices and main control devices, and introduces the solutions provided in the embodiments of this application. It is understood that, in order to achieve the above functions, the electronic device or main control device includes hardware structures and / or software modules corresponding to the execution of each function. By combining the units and algorithm steps of the various examples described in the embodiments disclosed in this application, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed by hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the technical solutions of the embodiments of this application.

[0185] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-mentioned formation pore pressure prediction method.

[0186] The aforementioned computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0187] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. The readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in application-specific integrated circuits (ASICs). Alternatively, the processor and the readable storage medium can exist as discrete components in an electronic device or a host device.

[0188] The memory 902 is the non-transitory computer-readable storage medium provided by this invention. The non-transitory computer-readable storage medium of this invention stores computer instructions that cause a computer to execute the formation pore pressure prediction method provided by this invention.

[0189] Memory 902, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the formation pore pressure prediction method in the embodiments of this application (e.g., Figure 17 The well logging data acquisition module 801, preprocessing module 802, dynamic gating signal acquisition module 803, scatter plot acquisition module 804, causal label acquisition module 805, training module 806, and prediction module 807 are shown. The processor 901 executes various functional applications and data processing by running non-transient software programs, instructions, and modules stored in the memory 902, thereby realizing the formation pore pressure prediction method in the above method embodiments.

[0190] In addition, this embodiment also provides a computer program product, including a computer program, which, when executed by a processor, is used to implement the formation pore pressure prediction method of the above embodiment.

[0191] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0192] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0193] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0194] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0195] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for predicting formation pore pressure, characterized in that, include: Obtain the depth, deep resistivity, shallow resistivity, formation density, and sonic velocity of the target well. The depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity are preprocessed to obtain the preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, and preprocessed acoustic velocity. The resistivity ratio index is obtained based on the pretreated deep resistivity and the pretreated shallow resistivity. The resistivity ratio index is input into the fully connected layer of the preset gated subnetwork to obtain a linear change result; The linear change result is activated using the activation function of the preset gated subnetwork to obtain a dynamic gated signal; a scatter plot of acoustic velocity-formation density is obtained based on the preprocessed formation density and the preprocessed acoustic velocity. After obtaining the sonic velocity-formation density scatter plot, the normal compaction trend line is determined. Based on the characteristics of deviation from the normal compaction trend line, a preset number of genetic categories are obtained, and a genetic label is added to each genetic category. One-dimensional convolution operation is performed on the pre-processed depth, the pre-processed deep resistivity, the pre-processed shallow resistivity, the pre-processed formation density, the pre-processed acoustic velocity, and the genetic label to obtain local logging features. The local logging features are subjected to maximum pooling and average pooling to obtain a multi-scale feature description; The multi-scale feature description and the dynamic gating signal are fused to obtain fused data; The fused data is input into the fully connected layer of a pre-constructed convolutional neural network to obtain the predicted value of formation pore pressure. Based on the predicted formation pore pressure and the actual measured formation pore pressure, a pre-constructed convolutional neural network is optimized to obtain a trained predictive convolutional neural network; wherein, the convolutional neural network includes a channel attention module and a spatial attention module; The well logging data to be predicted is acquired and input into the predictive convolutional neural network to obtain the predicted formation pore pressure value.

2. The formation pore pressure prediction method according to claim 1, characterized in that, The step of obtaining a scatter plot of acoustic velocity-formation density based on the preprocessed formation density and the preprocessed acoustic velocity includes: Based on the pre-processed formation density, the pre-processed acoustic velocity, and the acoustic velocity-formation density cross-plot discrimination theory, an acoustic velocity-formation density scatter plot is obtained.

3. The formation pore pressure prediction method according to claim 1, characterized in that, The preprocessing of the depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity to obtain preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, and preprocessed acoustic velocity includes: The depth, deep resistivity, shallow resistivity, formation density, and acoustic velocity are subjected to minimum-maximum normalization to obtain the pre-processed depth, pre-processed deep resistivity, pre-processed shallow resistivity, pre-processed formation density, and pre-processed acoustic velocity.

4. A formation pore pressure prediction device, characterized in that, include: The well logging data acquisition module is used to acquire the depth, deep resistivity, shallow resistivity, formation density, and sonic velocity of the target well. The preprocessing module is used to preprocess the depth, the deep resistivity, the shallow resistivity, the formation density, and the acoustic velocity to obtain the preprocessed depth, the preprocessed deep resistivity, the preprocessed shallow resistivity, the preprocessed formation density, and the preprocessed acoustic velocity. The dynamic gating signal acquisition module is used to obtain the resistivity ratio index based on the preprocessed deep resistivity and the preprocessed shallow resistivity. The resistivity ratio index is input into the fully connected layer of the preset gated subnetwork to obtain a linear change result; the activation function of the preset gated subnetwork is used to activate the linear change result to obtain a dynamic gated signal. The scatter plot acquisition module is used to obtain a scatter plot of acoustic velocity-formation density based on the preprocessed formation density and the preprocessed acoustic velocity. The causal label acquisition module is used to determine the normal compaction trend line after obtaining the sonic velocity-formation density scatter plot, divide the causal categories according to the characteristics of deviation from the normal compaction trend line, obtain a preset number of causal categories, and add causal labels to each causal category. The training module performs one-dimensional convolution operations on the preprocessed depth, preprocessed deep resistivity, preprocessed shallow resistivity, preprocessed formation density, preprocessed acoustic velocity, and the genetic label to obtain local logging features; it then performs max pooling and average pooling on the local logging features to obtain multi-scale feature descriptions; it fuses the multi-scale feature descriptions and the dynamic gating signal to obtain fused data; it inputs the fused data into the fully connected layer of a pre-constructed convolutional neural network to obtain predicted formation pore pressure values; and it optimizes the pre-constructed convolutional neural network based on the predicted formation pore pressure values ​​and the actual measured formation pore pressure values ​​to obtain a trained predictive convolutional neural network; wherein the convolutional neural network includes a channel attention module and a spatial attention module. The prediction module is used to acquire the logging data to be predicted and input the logging data to be predicted into the prediction convolutional neural network to obtain the formation pore pressure prediction value.

5. An electronic device, characterized in that, include: A processor and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the formation pore pressure prediction method as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the formation pore pressure prediction method as described in any one of claims 1 to 3.

Images