Hidden strike-slip fracture identification method based on earthquake-geologic model deep learning
By obtaining seismic and drilling data to generate labels, and using deep learning models to train to identify hidden strike-slip fractures, the problem of low recognition accuracy in the existing technology is solved, efficient and accurate hidden strike-slip fracture recognition is achieved, and exploration and development efficiency is improved.
Patent Information
- Application Number
- CN202510792970.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-13
- Publication Date
- 2025-08-12
AI Technical Summary
The prior art is difficult to accurately identify hidden strike-slip fractures, with low recognition accuracy and noise interference, and lack of 3D solid seismic-geological models, resulting in low exploration and development efficiency.
By acquiring seismic data and drilling data, the strike-slip fault seismic response label and the artifact seismic response label are generated, and deep learning models are used for training, the model parameters are optimized and identification methods for hidden strike-slip faults are generated.
The identification accuracy of hidden strike-slip fractures is improved, the impact of seismic noise and geological phenomena is eliminated, the credibility and efficiency of explanations are improved, and time and labor costs are saved.
Smart Images
Figure CN120468933A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of structural geology and oil and gas exploration and development evaluation, and specifically to a method for identifying hidden strike-slip faults based on deep learning of seismic-geological models. Background Art
[0002] In oil and gas basins, faults not only play a significant role in controlling the migration, accumulation, and preservation of oil and gas, but also have a significant impact on the transformation of tight reservoirs. They are a key component of fault-related structural geology research and applied exploration and development research in oil and gas reservoirs. In sedimentary basins, faults are typically identified through seismic data. However, hidden strike-slip faults (difficult to identify with conventional seismic data) are nearly vertical and small in scale, making seismic imaging difficult. They are also easily obscured by various seismic noise and other heterogeneous geological phenomena (such as weathering crust). Therefore, conventional seismic methods and techniques struggle to accurately identify strike-slip faults. Strike-slip faults are typically identified and interpreted through a combination of multi-method seismic attributes and human-computer interaction. However, due to the lack of a unified identification standard and the significant differences in interpretations based on different individuals' experience, seismic interpretation is difficult and the accuracy of actual drilling results is poor. This leads to drilling failures and safety incidents, which are technical challenges that hinder efficient exploration and development.
[0003] To this end, artificial intelligence methods have begun to be introduced in the seismic identification of strike-slip faults in recent years. Existing technologies based on deep learning not only improve the resolution and interpretation accuracy of strike-slip faults, but also avoid the influence of personal experience, greatly improving the efficiency of interpretation.
[0004] However, unlike the simple, nearly vertical sections of classical mechanical models, hidden strike-slip fault zones have complex fault structures and large vertical and horizontal variations. The 3D fault structures of strike-slip faults in different regions are complex and diverse. The lack of a 3D physical geological model makes it difficult to even establish a seismic-geological response model for low-order hidden strike-slip faults. Therefore, hidden strike-slip faults based on deep learning often have uncertain samples and difficult to label. Labeling and sample learning are key factors that limit the effectiveness of deep learning. Existing studies often use theoretical models of vertical sections or broken lines as labels, which can also lead to certain errors, resulting in differences in the structure, spatial morphology, and number of hidden faults compared to actual drilling. Furthermore, some atypical strike-slip faults may be deleted, while some fault artifacts caused by noise may be retained.
[0005] It can be seen that due to the uncertainty of the geological model of hidden strike-slip faults and the lack of 3D physical seismic-geological model samples, the existing deep learning strike-slip fault identification still has great limitations. It is difficult to identify low-order tiny hidden strike-slip faults and the identification accuracy is low. Summary of the Invention
[0006] To solve the above problems, this application provides a hidden strike-slip fault identification method based on deep learning of seismic-geological models, aiming to solve the problems of sample uncertainty and low identification accuracy in existing hidden strike-slip fault research.
[0007] A first aspect of an embodiment of the present invention provides a method for identifying hidden strike-slip faults based on deep learning of earthquake-geological models, comprising: S1. Acquire seismic data and drilling information, wherein the drilling information includes well location coordinates, well trajectory, drilling depth, porosity, permeability, and formation microresistivity scanning imaging data; S2. Generate strike-slip fault seismic response labels and strike-slip fault pseudo-seismic response labels based on seismic data and drilling data; S3. Input the strike-slip fault seismic response label and the strike-slip fault pseudo-seismic response label to train a strike-slip fault identification model, and identify hidden strike-slip faults based on the trained strike-slip fault identification model.
[0008] In an optional embodiment, the specific steps of S2 are: The geometry and reflection characteristics of strike-slip faults are obtained through seismic data; Determine the existence, depth and spatial distribution of strike-slip faults through drilling data; Based on the well-seismic calibration method, the strike-slip fault location is determined, and the reflection intensity, waveform characteristics, and amplitude variation characteristics of the strike-slip fault are extracted to generate the strike-slip fault seismic response label. Areas with abnormal reflection waves and waveform disturbances but no faults in the seismic data are selected, and the reflection intensity, waveform characteristics and amplitude change characteristics are extracted to generate seismic response labels of strike-slip fault illusions.
[0009] In an optional embodiment, when the strike-slip fault identification model is trained in S3, the parameters of the strike-slip fault identification model are also optimized, wherein the stochastic gradient descent calculation formula is: (1) in, is the network weight at the current moment, is the learning rate, is the gradient of the loss function with respect to the weights.
[0010] In an optional embodiment, the step S3 further includes verifying and updating the strike-slip fault identification model: The seismic data of the selected study area are input into the strike-slip fault identification model to calculate the fault probability volume of the study area; The degree of agreement between the fault probability volume and the actual strike-slip fault position calculated based on drilling data: (2) If the degree of agreement reaches the preset threshold, the trained strike-slip fault identification model is considered a valid model; If the degree of fit does not reach the preset threshold, adjust the learning rate , retrain the strike-slip fault identification model.
[0011] In an optional embodiment, the identifying of hidden strike-slip faults further includes generating a cross-sectional and planar distribution map of the strike-slip faults based on the identified hidden strike-slip faults in the study area.
[0012] A second aspect of an embodiment of the present invention provides a method and device for identifying hidden strike-slip faults based on deep learning of earthquake-geological models, the device comprising: Data acquisition module: acquires seismic data and drilling information, including well location coordinates, well trajectory, drilling depth, porosity, permeability and formation microresistivity scanning imaging data; Label generation module: Generates strike-slip fault seismic response labels and strike-slip fault pseudo-seismic response labels based on seismic data and drilling information; Model building module: input the strike-slip fault seismic response label and the strike-slip fault pseudo-seismic response label to train the strike-slip fault identification model, and identify the hidden strike-slip fault based on the trained strike-slip fault identification model.
[0013] A third aspect of an embodiment of the present invention provides an electronic device, characterized in that it includes: a processor, a memory, and a computer program stored in the memory and runnable on the processor, wherein when the processor executes the program, it implements a hidden strike-slip fault identification method based on deep learning of earthquake-geological models.
[0014] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, a hidden strike-slip fault identification method based on deep learning of seismic-geological models is provided.
[0015] In the disclosed embodiment, by calibrating existing drilling data from blocks with abundant research data and a good research foundation, a seismic-geological response model of strike-slip faults of different types and characteristics is established. On this basis, deep learning of the seismic response of strike-slip faults is carried out, and then drilling correction is performed to carry out batch identification of seismic data and improve the identification accuracy of strike-slip faults. The present invention solves the problem of inconsistency between the theoretical model labels of strike-slip faults and the actual complex and diverse geological models of strike-slip faults, and can obtain more realistic images of hidden strike-slip faults and improve the resolution of strike-slip fault seismic images. It eliminates the influence of various noises and other geological phenomena, overcomes the multi-solution problem caused by seismic attributes and manual interpretation, and improves the credibility of strike-slip fault interpretation. Strike-slip faults are intelligently picked up to obtain planar and cross-sectional distribution maps of strike-slip faults, and a large amount of seismic interpretation and mapping time can be saved. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0017] Figure 1 This is a flow chart of a hidden strike-slip fault identification method based on deep learning of earthquake-geological models proposed in one embodiment of the present application; Figure 2 This is a fault structure diagram of a horizontal well crossing a strike-slip fault zone in the study area proposed in one embodiment of the present application; Figure 3 (a) is a seismic attribute profile of a horizontal well crossing a strike-slip fault zone in the study area proposed in one embodiment of the present application; Figure 3 (b) is a plan view of seismic attributes of a horizontal well crossing a strike-slip fault zone in the study area proposed in one embodiment of the present application; Figure 3 (c) is a diagram of an actual drilling pattern of a horizontal well crossing a strike-slip fault zone in the study area proposed in one embodiment of the present application; Figure 4 Schematic diagram of the seismic profile response characteristics and the pseudo-seismic profile response characteristics of the strike-slip fault in the study area proposed in one embodiment of the present application; Figure 5 Schematic diagram of the effect of the L2 regularization coefficient on training loss and validation loss proposed in one embodiment of the present application; Figure 6 (a) is a maximum likelihood attribute profile of the study area proposed in one embodiment of the present application; Figure 6(b) is a deep learning fault probability volume cross-section diagram proposed in an embodiment of the present application; Figure 7 (a) is a plane diagram of the coherent attributes of the target layer in the study area proposed in one embodiment of the present application; Figure 7 (b) is a plane diagram of the deep learning fault probability volume proposed in one embodiment of the present application; Figure 8 This is a schematic diagram of a hidden strike-slip fault identification device based on deep learning of earthquake-geological models proposed in one embodiment of the present application; Figure 9 This is a schematic diagram of an electronic device of the present application. DETAILED DESCRIPTION
[0018] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.
[0019] Please refer to Figure 1 , Figure 1 This is a flowchart of a method for identifying hidden strike-slip faults based on deep learning of earthquake-geological models proposed in one embodiment of the present application. Figure 1 As shown in FIG, a hidden strike-slip fault identification method based on deep learning of earthquake-geological models includes: S1. Acquire seismic data and drilling information, wherein the drilling information includes well location coordinates, well trajectory, drilling depth, porosity, permeability, and formation microresistivity scanning imaging data; In this embodiment, a research area is established or built upon an existing area. High-quality 3D seismic data is loaded and analyzed for resolution, signal-to-noise ratio, reflection intensity, and waveform characteristics to ensure that the seismic data clearly reflects fault information. Drilling data from existing and un-drilled strike-slip fault zones is collected and organized, primarily including well location coordinates, well trajectories, drilling depth, porosity, permeability, and FMI (formation microresistivity imaging). This drilling data is then loaded into the work area.
[0020] S2. Generate strike-slip fault seismic response labels and strike-slip fault pseudo-seismic response labels based on seismic data and drilling data; In this example, a representative block within the seismic work area (where a large number of horizontal wells encounter strike-slip faults) was selected, and the presence of strike-slip faults was confirmed through analysis of drilling and logging data. Combined with the seismic-geological characteristics of multiple wells, the different seismic response characteristics of strike-slip faults were comprehensively summarized.
[0021] Using the results of the aforementioned well-seismic calibration, we clearly identify characteristic areas of strike-slip fault zones in seismic data, such as reflection intensity and continuity, and create refined deep learning labels for strike-slip faults. We also identify artifact areas in the seismic data that exhibit reflection anomalies and waveform disturbances, but where drilling evidence indicates no fault development. These artifact labeling features are clearly identified, along with their reflection, waveform, and amplitude variation characteristics, to create a dataset of artifact labels.
[0022] S3. Input the strike-slip fault seismic response label and the strike-slip fault pseudo-seismic response label to train a strike-slip fault identification model, and identify hidden strike-slip faults based on the trained strike-slip fault identification model.
[0023] In this example, deep learning and artifact removal of strike-slip fault labels are performed. A convolutional neural network deep learning model is used to learn and train strike-slip fault labels and artifact labels. The model input is labeled seismic data (including real strike-slip fault labels and artifact labels). The training set accounts for 70% of the total labels, and the validation set accounts for 30%. The output is a model for intelligent strike-slip fault identification (used for subsequent fault probability volume calculation). The network structure includes several convolutional layers, pooling layers, and fully connected layers, and the network weights are optimized using a backpropagation algorithm.
[0024] Furthermore, the specific steps of S2 are: The geometry and reflection characteristics of strike-slip faults are obtained through seismic data; Determine the existence, depth and spatial distribution of strike-slip faults through drilling data; Based on the well-seismic calibration method, the strike-slip fault location is determined, and the reflection intensity, waveform characteristics, and amplitude variation characteristics of the strike-slip fault are extracted to generate the strike-slip fault seismic response label. Areas with abnormal reflection waves and waveform disturbances but no faults in the seismic data are selected, and the reflection intensity, waveform characteristics and amplitude change characteristics are extracted to generate seismic response labels of strike-slip fault illusions.
[0025] In this example, a representative block within the seismic work area (where a large number of horizontal wells encountered strike-slip faults) was selected. Within the representative block, there were 38 wells that encountered strike-slip faults and 45 wells that did not encounter strike-slip faults. The presence of strike-slip faults was determined by analyzing drilling and logging data. Combined with the seismic-geological characteristics of multiple wells, the different seismic response characteristics of strike-slip faults were comprehensively summarized.
[0026] In the selected representative blocks, through detailed analysis of the drilling and logging data of 38 typical horizontal wells, and through analysis of drilling and logging data such as drilling depth, fracture characteristics, porosity, permeability, FMI, etc., a single well geological model of the strike-slip fault encountered was established to determine whether the strike-slip fault exists and the specific depth of the strike-slip fault is clarified. Figure 2 , Figure 2This is a fracture structure diagram of a horizontal well crossing a strike-slip fault zone in the study area proposed in one embodiment of the present application. Figure 2 As shown, Figure 2 (a) is a statistical diagram of fracture density, aperture, and porosity parameters. The values are larger in areas where strike-slip faults are developed. Figure 2 (b) is a rose diagram of crack dips, showing that there are multiple groups of directional cracks in the area where strike-slip faults are developed; Figure 2 (c) is a diagram of the fracture distribution pattern, showing the development of fractures and cataclastic rocks in the fault core; Figure 2 (d) is a three-dimensional model diagram of the strike-slip fault structure, and the plane distribution refers to the results of seismic attributes; Figure 2 (e) is a seismic coherence plane map. Conventional seismic attributes cannot identify small strike-slip faults.
[0027] In the seismic area, high-precision well-seismic calibration technology is used to calibrate the above-mentioned single-well geological model to the seismic data, accurately identify the true location of the strike-slip fault in each well in the seismic area, analyze the seismic attributes reflecting the strike-slip fault, and verify the actual location of the fault in the area. Figure 3 (a) and 3(b), Figure 3 (a) is a seismic attribute profile of a horizontal well crossing a strike-slip fault zone in the study area proposed in one embodiment of the present application; Figure 3 (b) is a plan view of the seismic attributes of a horizontal well crossing a strike-slip fault zone in the study area proposed in one embodiment of the present application.
[0028] Although the strike-slip fault response in the target layer is weak, the seismic response of the deep fault zone is significant, compared with the coherent properties such as Figure 2 (e) The response effect is better and the actual drilling comparison is high, such as Figure 3 (c) Figure 3 (c) is a diagram of the actual drilling pattern of a horizontal well crossing a strike-slip fault zone in the study area proposed in one embodiment of the present application.
[0029] There may be some seismic responses similar to strike-slip faults in the seismic work area, but there are no strike-slip faults with fault responses in drilling and logging data. By comparing the seismic response characteristics with those of strike-slip faults, the false responses can be identified and distinguished. The seismic response analysis based on multiple wells that encountered strike-slip faults is mainly based on observation and comparison of the different seismic response characteristics of faults and false responses. The main seismic response characteristics of strike-slip faults are obvious phase axis offset: high and steep, vertical, through-and-through, multiple stratigraphic units are faulted, the upper and lower fold deformation is consistent, there are offsets, one side is continuous parallel reflection, the other side may be chaotic, etc. The main seismic response characteristics of strike-slip fault false responses are: local offset, upper and lower discontinuity, only individual reflection layers are faulted, local fold response, gentle knee folds, etc. Summarize the different seismic response modes of strike-slip faults and strike-slip fault false responses to form a complete response map of strike-slip faults and false responses. Please refer to Figure 4 , Figure 4 This is a schematic diagram of the seismic profile response characteristics and pseudo-seismic profile response characteristics of the strike-slip fault in the study area proposed in one embodiment of this application. It provides a basis and data support for deep learning model training.
[0030] Based on an analysis of the seismic response characteristics of strike-slip faults, deep learning labels are generated for strike-slip faults with different seismic response characteristics. These labels reflect the response patterns of strike-slip faults in seismic data, such as reflection intensity and waveform characteristics, and provide training data for the deep learning model. Artifact labels are generated for artifact responses in seismic data that resemble strike-slip faults but are not verified by drilling. This helps improve the deep learning model's ability to distinguish between true strike-slip faults and artifact responses.
[0031] Furthermore, when the strike-slip fault identification model is trained in S3, the parameters of the strike-slip fault identification model are also optimized, wherein the stochastic gradient descent calculation formula is: (1) in, is the network weight at the current moment, is the learning rate, is the gradient of the loss function with respect to the weights.
[0032] In this embodiment, the network weight Random initialization, no need to manually set the learning rate The initial value can be set to , if the training is unstable, it can be reduced to or smaller.
[0033] During the training process, the cross-entropy loss function is used as the optimization objective function to measure the gap between the predicted label and the true label. Its formula is as follows:
[0034] in, is the label of the i-th sample (if the sample is a strike-slip fault label, then , if the sample is a strike-slip fault pseudo label, then ), is the predicted label output by the model, and N is the number of samples.
[0035] L2 regularization and batch normalization. In order to improve the generalization ability of the model and prevent overfitting, L2 regularization (Ridge Regularization) is used to penalize the loss function. The optimized loss function is:
[0036] in, is the regularization coefficient, is the network weight, n is the number of weights. The number of weights n does not need to be manually assigned, and the regularization coefficient Typically set to arrive During training, Gradually increase it, and refer to the diagram of the impact of the L2 regularization coefficient on training loss and validation loss to select the optimal parameter until you see performance improvement on the validation set.
[0037] At the same time, batch normalization is used to speed up the training process and improve the stability of the model. The formula is as follows:
[0038] in, is the input data, and are the mean and standard deviation of the data, is a small constant to prevent division by zero errors. 、 and No manual assignment required. Typically set to .
[0039] The stochastic gradient descent algorithm with momentum is used in the model optimization process, and the initial learning rate is set to , and use L2 regularization and batch normalization, regularization coefficient Set to 、 、 and 0, verify the L2 regularization coefficient The impact on training and validation loss, four sets of curves are obtained, please refer to Figure 5 , Figure 5 This is a schematic diagram of the impact of the L2 regularization coefficient on training loss and validation loss proposed in one embodiment of the present application. = 0 (no regularization), the training loss decreases the fastest, but the validation loss decreases slowly and tends to be stable, indicating that the model is overfitting. = When , the training and validation losses decrease evenly and are close, reflecting the best generalization ability, and both overfitting and underfitting are balanced. = When , the training loss is slightly higher, but the validation loss decreases steadily and still maintains good performance. = When , the training loss decreases slowly and the validation loss is also high, indicating that the model learning ability is limited and underfitting occurs. The influence of the L2 regularization coefficient on the training loss and validation loss shows that the regularization coefficient in this work area is = When , it can prevent overfitting while maintaining the model fitting ability, reflecting the best generalization ability.
[0040] Furthermore, the S3 also includes verifying and updating the strike-slip fault identification model: The seismic data of the selected study area are input into the strike-slip fault identification model to calculate the fault probability volume of the study area; The degree of agreement between the fault probability volume and the actual strike-slip fault position calculated based on drilling data: (2) If the degree of agreement reaches the preset threshold, the trained strike-slip fault identification model is considered a valid model; If the degree of fit does not reach the preset threshold, adjust the learning rate , retrain the strike-slip fault identification model.
[0041] In this embodiment, please refer to Figure 6 , Figure 6 (a) is a maximum likelihood attribute profile of the study area proposed in one embodiment of the present application; Figure 6 (b) is a cross-sectional view of the deep learning fault probability volume proposed in one embodiment of this application. The model was applied to seismic data from a selected work area with a high concentration of wells, calculating the fault probability volume. This volume was then validated against the drilling data to ensure model accuracy. Within the study area, areas with a high concentration of drilling data were selected to ensure comprehensive and representative seismic data coverage within the selected area.
[0042] The trained convolutional neural network (CNN) model is applied to the selected seismic data. The trained model is used to analyze the data and calculate the fault probability volume for the area. The calculated fault probability volume is verified. The predicted fault area is compared with the known fault locations in the drilling data to calculate the degree of fit: Goodness of fit is greater than 95%: If the goodness of fit reaches or exceeds 95%, the model is considered to be sufficiently accurate and can be determined as a valid model.
[0043] The degree of fit is less than 95%: If the degree of fit is less than 95%, it indicates that there is a certain error in the model and the model needs to be adjusted. Adjust the learning rate in step 4S. , repeat the calculation operation of S3 until the degree of agreement reaches more than 95%.
[0044] For example, a representative area within the study area with dense drilling data was selected and the trained CNN model was applied to calculate the fault probability volume. The first validation revealed that the predicted fault locations matched only 92.7% of the actual drilling fault locations, falling short of the 95% requirement. The model was optimized for the first time, with the learning rate reduced to 0. Deep learning training was then repeated and the fault probability volume recalculated. After the second validation, the agreement between the predicted and actual drilling fault locations improved to 96.3%, meeting practical standards.
[0045] Furthermore, the identifying of hidden strike-slip faults also includes generating a cross-sectional and planar distribution map of the strike-slip faults based on the identified hidden strike-slip faults in the study area.
[0046] In this embodiment, please refer to Figure 7 , Figure 7 (a) is a plane diagram of the coherent attributes of the target layer in the study area proposed in one embodiment of the present application. Figure 7 (b) is a plane diagram of the deep learning fault probability volume proposed in one embodiment of the present application. Figure 7 As shown in the figure, using a professional seismic interpretation platform, the resulting strike-slip fault probability volume is automatically converted into precise strike-slip fault profiles and planar distribution maps, visually demonstrating the spatial distribution of hidden strike-slip faults within the region and providing accurate structural evidence for oil and gas exploration and development. Comparative analysis shows that strike-slip fault identification based on deep learning of seismic and geological models takes less than 1 / 30 of the time of conventional interpretation, while improving interpretation accuracy by over 20%.
[0047] Please refer to Figure 8 , Figure 8 This is a schematic diagram of a hidden strike-slip fault identification device based on deep learning of earthquake-geological models proposed in one embodiment of the present application. Figure 8 As shown, the embodiment of the present disclosure also provides a hidden strike-slip fault identification method and device based on deep learning of earthquake-geological models, the device comprising a data acquisition module 801, a label making module 802 and a model building module 803: Data acquisition module: acquires seismic data and drilling information, including well location coordinates, well trajectory, drilling depth, porosity, permeability and formation microresistivity scanning imaging data; Label generation module: Generates strike-slip fault seismic response labels and strike-slip fault pseudo-seismic response labels based on seismic data and drilling information; Model building module: input the strike-slip fault seismic response label and the strike-slip fault pseudo-seismic response label to train the strike-slip fault identification model, and identify the hidden strike-slip fault based on the trained strike-slip fault identification model.
[0048] The present disclosure also provides an electronic device. Figure 9 , Figure 9 Schematic diagram of an electronic device according to an embodiment of the present disclosure. Figure 9 As shown, the electronic device 100 includes: a memory 110 and a processor 120. The memory 110 and the processor 120 are connected via a bus communication. A computer program is stored in the memory 110, and the computer program can be run on the processor 120 to implement the steps in the hidden strike-slip fault identification method based on deep learning of earthquake-geological models disclosed in the embodiment of the present disclosure.
[0049] The disclosed embodiment also provides a computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by a processor of a computer device, the computer device is enabled to perform the steps in the hidden strike-slip fault identification method based on deep learning of seismic-geological models as described in the disclosed embodiment.
[0050] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, apparatuses, electronic devices, and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0051] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.
[0052] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0053] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.
[0054] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.
[0055] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.
[0056] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or terminal device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements that are inherent to such process, method, article, or terminal device. In the absence of further restrictions, an element defined by the phrase "comprises a ..." does not exclude the presence of additional identical elements in the process, method, article, or terminal device that includes the element.
[0057] The above is a detailed introduction to the hidden strike-slip fault identification method based on deep learning of seismic-geological models provided by this application. This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method of this application and its core idea; at the same time, for general technical personnel in this field, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on this application.
Claims
1. A method for identifying hidden strike-slip faults based on deep learning of earthquake-geological models, characterized by: include: S1. Acquire seismic data and drilling information, wherein the drilling information includes well location coordinates, well trajectory, drilling depth, porosity, permeability, and formation microresistivity scanning imaging data; S2. Generate strike-slip fault seismic response labels and strike-slip fault pseudo-seismic response labels based on seismic data and drilling data; S3. Input the strike-slip fault seismic response label and the strike-slip fault pseudo-seismic response label to train a strike-slip fault identification model, and identify hidden strike-slip faults based on the trained strike-slip fault identification model.
2. The method for identifying hidden strike-slip faults based on deep learning of earthquake-geological models according to claim 1, characterized in that: The specific steps of S2 are: The geometry and reflection characteristics of strike-slip faults are obtained through seismic data; Determine the existence, depth and spatial distribution of strike-slip faults through drilling data; Based on the well-seismic calibration method, the strike-slip fault location is determined, and the reflection intensity, waveform characteristics, and amplitude variation characteristics of the strike-slip fault are extracted to generate the strike-slip fault seismic response label. Areas with abnormal reflection waves and waveform disturbances but no faults in the seismic data are selected, and the reflection intensity, waveform characteristics and amplitude change characteristics are extracted to generate seismic response labels of strike-slip fault illusions.
3. The method for identifying hidden strike-slip faults based on deep learning of earthquake-geological models according to claim 1, characterized in that: When the strike-slip fault identification model is trained in S3, the parameters of the strike-slip fault identification model are also optimized, wherein the stochastic gradient descent calculation formula is: （1） in, is the network weight at the current moment, is the learning rate, is the gradient of the loss function with respect to the weights.
4. The method for identifying hidden strike-slip faults based on deep learning of earthquake-geological models according to claim 3 is characterized in that: S3 also includes verification and update of the strike-slip fault identification model: The seismic data of the selected study area are input into the strike-slip fault identification model to calculate the fault probability volume of the study area; The degree of agreement between the fault probability volume and the actual strike-slip fault position calculated based on drilling data: （2） If the degree of agreement reaches the preset threshold, the trained strike-slip fault identification model is considered a valid model; If the degree of fit does not reach the preset threshold, adjust the learning rate , retrain the strike-slip fault identification model.
5. The method for identifying hidden strike-slip faults based on deep learning of earthquake-geological models according to claim 1, characterized in that: The identifying of hidden strike-slip faults further includes generating a cross-sectional and planar distribution map of the strike-slip faults based on the identified hidden strike-slip faults in the study area.
6. The method and device for identifying hidden strike-slip faults based on deep learning of earthquake-geological models according to any one of claims 1 to 5, characterized in that: The device comprises: Data acquisition module: acquires seismic data and drilling information, including well location coordinates, well trajectory, drilling depth, porosity, permeability and formation microresistivity scanning imaging data; Label generation module: Generates strike-slip fault seismic response labels and strike-slip fault pseudo-seismic response labels based on seismic data and drilling information; Model building module: input the strike-slip fault seismic response label and the strike-slip fault pseudo-seismic response label to train the strike-slip fault identification model, and identify the hidden strike-slip fault based on the trained strike-slip fault identification model.
7. An electronic device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for identifying hidden strike-slip faults based on deep learning of earthquake-geological models according to any one of claims 1 to 5 is implemented.
8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for identifying hidden strike-slip faults based on deep learning of earthquake-geological models as described in any one of claims 1 to 5 is implemented.