Fault and crack prediction method and device and computer equipment

Abstract

The invention provides a fault and crack prediction method and device and computer equipment, and belongs to the technical field of oil-gas exploration and development. The method comprises the following steps: acquiring seismic data of a target area; identifying a fault in the seismic data by using a pre-constructed machine learning model to obtain a first fault data volume; extracting seismic attribute data volumes sensitive to faults of different scales and seismic attribute data volumes sensitive to fractures from the seismic data; and performing multi-attribute fusion on the first fault data volume, the seismic attribute data volume sensitive to the faults with different scales and the seismic attribute data volume sensitive to the cracks, and identifying the faults and the cracks with different scales in the target area. According to the method, based on a machine learning algorithm and a technical concept of multi-seismic attribute fusion identification, accurate prediction of faults and fractures of different scales is realized.

Description

Technical Field

[0001] This application belongs to the field of oil and gas exploration and development technology, specifically relating to a fault and fracture prediction method, a fault and fracture prediction device, a computer device, and a machine-readable storage medium. Background Technology

[0002] The Songnan reservoir exhibits well-developed shallow and medium-depth faults, and these faults and associated fractures play a crucial role in hydrocarbon accumulation and enrichment. In conventional reservoirs, faults and associated fractures connect oil sources, forming hydrocarbon migration channels, and also act as hydrocarbon shielding zones, creating hydrocarbon-rich areas. In unconventional reservoirs, faults and associated fractures increase matrix porosity and, through fracturing, create artificial fractures that connect numerous natural fractures, altering the seepage system and increasing production. Analysis of multiple development areas suggests that closer proximity to faults leads to higher cumulative production, and this is also confirmed in interlayered shale oil, where production in fault zones is significantly better than in non-fault zones. In horizontal well drilling, the development of faults and fractures increases the risk of wellbore collapse, lost circulation, and casing deformation, affecting the successful drilling, completion, and fracturing of horizontal wells. Therefore, early warning and prediction of micro-fracture and fracture development are necessary before and during horizontal well drilling. Through forward modeling, under ideal conditions, faults with a vertical displacement greater than 10m can be identified at a dominant seismic frequency of 40Hz, and faults with a vertical displacement of about 5m can be distinguished at 80Hz. Currently, the dominant seismic frequency is between 40 and 50Hz. Due to the influence of signal-to-noise ratio and other factors, faults with a minimum vertical displacement of 15 to 20m can be directly identified "by the naked eye". Currently, some seismic fault prediction schemes have been proposed in the industry by utilizing the different sensitivities of different seismic attributes to faults and fractures. For example, by utilizing the characteristics of coherence volume to faults and maximum likelihood volume to fracture zones, coherence volume and maximum likelihood volume are extracted from post-stack seismic data. The extracted coherence volume is used to identify faults with a vertical displacement of 15 to 20m, and the maximum likelihood volume is used to identify fracture development. However, although coherence volume can enhance the fault response in seismic data, the identification results are affected by noise and other similar fault responses. The extracted fault planes are relatively fragmented, making combination difficult, and the fault prediction accuracy is low.

[0003] Therefore, it is urgent to improve the above-mentioned earthquake fault prediction scheme. Summary of the Invention

[0004] The purpose of this application is to provide a method for predicting faults and fractures, a device for predicting faults and fractures, a computer device, and a machine-readable storage medium to overcome the technical problems in the prior art where the prediction results obtained by using coherent volumes for fracture prediction are of low accuracy, small fracture planes, and difficult assembly.

[0005] To achieve the above objectives, a first aspect of this application provides a method for predicting faults and fractures, the method comprising:

[0006] Acquire seismic data for the target area;

[0007] The faults in the earthquake data are identified using a pre-built machine learning model to obtain the first fault data volume;

[0008] Seismic attribute data volumes sensitive to faults at different scales and seismic attribute data volumes sensitive to fractures are extracted from the seismic data.

[0009] The first fault data volume, the seismic attribute data volume sensitive to faults of different scales, and the seismic attribute data volume sensitive to fractures are fused to identify faults and fractures of different scales in the target area.

[0010] In a specific embodiment of this application, the method further includes:

[0011] The seismic data is recovered by performing low-frequency and high-frequency seismic data recovery to obtain bilaterally frequency-extended seismic data;

[0012] The seismic data after bilateral frequency extension is divided into frequencies, and the different frequency bands obtained by frequency division are denoised separately to obtain different frequency data volumes.

[0013] The different frequency data volumes obtained after denoising are reconstructed to obtain reconstructed seismic data, which triggers the identification of faults in the seismic data using a pre-built machine learning model to obtain the first fault data volume.

[0014] In a specific embodiment of this application, the method further includes:

[0015] The linear structure in the seismic attribute data volume that is sensitive to cracks is enhanced to obtain a crack planar map of the target area;

[0016] Vectorize the crack plan view of the target area and identify the location of each crack line segment;

[0017] The crack distribution characteristics of the target area are obtained based on the identified crack segment locations, and the crack distribution characteristics include at least one of crack density characteristics and crack direction characteristics.

[0018] In a specific embodiment of this application, extracting seismic attribute data volumes sensitive to faults at different scales and seismic attribute data volumes sensitive to fractures from the seismic data includes:

[0019] Extract the curvature volume from the earthquake data;

[0020] Extract the maximum likelihood from the earthquake data.

[0021] In a specific embodiment of this application, multi-attribute fusion is performed on the first fault data volume, the seismic attribute data volume sensitive to faults of different scales, and the seismic attribute data volume sensitive to fractures to identify faults and fractures of different scales in the target area, including:

[0022] The first fault data volume, curvature volume, and maximum likelihood volume are fused using RGB to obtain the fused data volume;

[0023] The first color data volume representing the first fault data volume in the fused data volume is used to identify the first scale fault within the target area;

[0024] The second color data volume, representing the curvature volume, in the fused data volume is used to identify the second-scale fault within the target area;

[0025] The third-scale fracture development zone associated with the fracture in the fracture network formed by the first-scale fault and the second-scale fault is identified by using the third color data volume representing the maximum likelihood in the fused data volume.

[0026] The first scale is larger than the second scale, and the second scale is larger than the third scale.

[0027] In a specific embodiment of this application, the machine learning model is a deep neural network model.

[0028] In a specific embodiment of this application, enhancing the linear structure in the crack-sensitive seismic attribute data volume to obtain a crack planar map of the target area includes:

[0029] Seismic profiles are generated using seismic attribute data volumes that are sensitive to cracks.

[0030] Construct the Hessian matrix corresponding to the seismic profile;

[0031] Construct Gaussian functions with different scale spatial factors, and calculate the second-order partial derivatives of each Gaussian function with respect to the horizontal and vertical coordinates of the seismic profile.

[0032] The second-order partial derivatives of each Gaussian function are convolved with the Hessian matrix to obtain the first matrix. The cracks and point structures of the linear structure in the seismic profile are then identified based on the eigenvalues ​​and eigenvectors in the first matrix.

[0033] A response function is constructed based on the eigenvalues ​​and eigenvectors. The response function is then used to enhance the cracks in the linear structure and filter out the point structures, resulting in a crack plan view of the target area.

[0034] In a specific embodiment of this application, the step of enhancing the linear structure in the crack-sensitive seismic attribute data volume to obtain a crack planar map of the target area further includes:

[0035] Binarize the cracks and background in the crack planar image of the target area to obtain the binarized crack planar image of the target area.

[0036] In a specific embodiment of this application, the vectorized target area crack planar diagram identifies the location of each crack line segment, including:

[0037] Hough transformation is performed on the crack plan view of the target area to identify the location of crack segments.

[0038] In a specific embodiment of this application, extracting the curvature volume from the seismic data includes:

[0039] The seismic data is scanned by azimuth to obtain seismic sub-data volumes with different azimuths;

[0040] Curvature attributes were extracted from each seismic sub-data volume, and each curvature volume was obtained by corresponding to the others.

[0041] In a specific embodiment of this application, the seismic data is post-stack narrow azimuth seismic data.

[0042] A second aspect of this application provides a fault and fracture prediction device, comprising:

[0043] The seismic data acquisition module is used to acquire seismic data for the target area.

[0044] The first fault data volume prediction module is used to identify faults in the seismic data using a pre-built machine learning model to obtain the first fault data volume.

[0045] The seismic attribute extraction module is used to extract seismic attribute data volumes that are sensitive to faults at different scales and seismic attribute data volumes that are sensitive to fractures from the seismic data.

[0046] The fault and fracture identification module is used to perform multi-attribute fusion of the first fault data volume, the seismic attribute data volume sensitive to faults of different scales, and the seismic attribute data volume sensitive to fractures, so as to identify faults and fractures of different scales in the target area.

[0047] In a specific embodiment of this application, multi-attribute fusion is performed on the first fault data volume, the seismic attribute data volume sensitive to faults of different scales, and the seismic attribute data volume sensitive to fractures to identify faults and fractures of different scales in the target area, including:

[0048] The first fault data volume, curvature volume, and maximum likelihood volume are fused using RGB to obtain the fused data volume;

[0049] The first color data volume representing the first fault data volume in the fused data volume is used to identify the first scale fault within the target area;

[0050] The second color data volume, representing the curvature volume, in the fused data volume is used to identify the second-scale fault within the target area;

[0051] The third-scale fracture development zone associated with the fracture in the fracture network formed by the first-scale fault and the second-scale fault is identified by using the third color data volume representing the maximum likelihood in the fused data volume.

[0052] The first scale is larger than the second scale, and the second scale is larger than the third scale.

[0053] A third aspect of this application provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the fault and fracture prediction method described in the first aspect of this application.

[0054] A fourth aspect of this application provides a machine-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the fault and fracture prediction method described in the first aspect of this application.

[0055] Machine learning algorithms can extract rich fault features from seismic data, effectively suppress noise interference, and exhibit superior noise resistance compared to coherent volumes. The aforementioned technical solution, based on the concept of machine learning algorithms and multi-seismic attribute fusion identification, utilizes a pre-constructed machine learning model to generate a first fault data volume. This first fault data volume, along with seismic attribute data volumes extracted from seismic data that are sensitive to faults at different scales and those sensitive to fractures, are then fused. This achieves accurate prediction of the development of faults and fractures at different scales in the target area, avoiding the shortcomings of identifying fragmented fault planes and the difficulty in combining fragmented fault planes.

[0056] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0057] The accompanying drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the following detailed description to explain the embodiments of this application, but do not constitute a limitation on the embodiments of this application. In the drawings:

[0058] Figure 1A schematic diagram illustrates a first flowchart of a fault and fracture prediction method according to an embodiment of this application;

[0059] Figure 2 A second flowchart of a fault and fracture prediction method according to an embodiment of this application is illustrated schematically;

[0060] Figure 3 A third flowchart of a fault and fracture prediction method according to an embodiment of this application is illustrated schematically.

[0061] Figure 4 This illustration schematically shows a rectangular coordinate system during Hough transformation according to an embodiment of this application;

[0062] Figure 5 A schematic diagram of the polar coordinate system in the Hough variation according to an embodiment of this application is shown.

[0063] Figure 6 This diagram illustrates the technical flowchart for predicting interrupted layers and cracks in a specific application example.

[0064] Figure 7 This diagram illustrates the structure of a deep neural network model in a specific application example.

[0065] Figure 8 The illustration shows a seismic profile with improved quality in a specific application example;

[0066] Figure 9 The original seismic profile is illustrated in a specific application example;

[0067] Figure 10 This illustration shows a schematic diagram of layer Q1 in the first fault data volume in a specific application example;

[0068] Figure 11 This illustration schematically shows a 150° azimuth curvature body in a specific application example;

[0069] Figure 12 This illustration schematically shows a 90° azimuth curvature body in a specific application example;

[0070] Figure 13 This illustration shows a schematic diagram of a physical simulation experiment design pattern in a specific application example;

[0071] Figure 14 This illustration shows a physical simulation experiment diagram of different extrusion directions and the pre-existing fracture angle in a specific application example;

[0072] Figure 15 The diagram illustrates the crack direction obtained from physical simulation experiments of different extrusion directions and pre-existing fracture angles in a specific application example.

[0073] Figure 16 The diagram illustrates the RGB multi-attribute fusion result of the Q1 layer in a specific application example.

[0074] Figure 17 The density planar diagram of the fracture development zone in layer Q1 is schematically shown in a specific application example;

[0075] Figure 18 The schematic diagram illustrates the orientation plan view of the crack development zone in layer Q1 in a specific application example;

[0076] Figure 19 The diagram illustrates the predicted results of the interrupted layer and cracks in a specific application example.

[0077] Figure 20 The diagram illustrates the fracturing operation curve (4035-3985m) of the 12th stage of Well 1 in a specific application example.

[0078] Figure 21 The diagram illustrates the fracturing operation curve (3250-3190m) of the 26th stage of Well 1 in a specific application example.

[0079] Figure 22 A schematic block diagram of a computer device according to an embodiment of this application is shown. Detailed Implementation

[0080] The specific embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the embodiments of this application.

[0081] If the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Furthermore, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0082] Figure 1 A schematic flowchart illustrating a first method for predicting faults and fractures according to an embodiment of this application is shown. Figure 1 As shown in the embodiments of this application, a fault and fracture prediction method may include the following steps:

[0083] Step 100: Obtain seismic data for the target area.

[0084] In this application, the seismic data can be three-dimensional seismic survey data, which refers to the seismic data volume obtained after establishing a three-dimensional seismic survey and loading seismic data.

[0085] Step 102: Use a pre-built machine learning model to identify faults in the seismic data obtained in step 100 to obtain the first fault data volume.

[0086] In this application, the machine learning model can be a neural network model, etc. For example, a deep neural network model is obtained by training and optimizing the parameters of an initial deep neural network. The deep neural network model takes seismic data as its input data and the first fault data volume as its output data. The deep neural network model obtained after training and parameter optimization is used as a machine learning model for fault identification.

[0087] Step 104: Extract seismic attribute data volumes sensitive to faults at different scales and seismic attribute data volumes sensitive to fractures from the seismic data obtained in Step 100.

[0088] It is important to understand that being sensitive to faults or fractures means that, for a fault or fracture of a certain scale, the accuracy of identifying that scale of fault or fracture through a certain seismic attribute is higher. In this case, the seismic attribute is the seismic attribute that is sensitive to faults or fractures of that scale.

[0089] For example, coherence volumes can identify faults of a larger scale, such as faults with a displacement of more than 10m. Ant bodies and maximum likelihood bodies can identify the development of fractures associated with faults.

[0090] Step 106: Perform multi-attribute fusion on the first fault data volume, the seismic attribute data volume sensitive to faults of different scales obtained in step 104, and the seismic attribute data volume sensitive to fractures to identify faults and fractures of different scales in the target area.

[0091] In one comparative embodiment, coherence volumes extracted from seismic data are used for fault identification. While coherence volumes can enhance the fault response in seismic data, they are susceptible to noise and other similar fault responses, resulting in fragmented fault planes that are sometimes difficult to combine, leading to low accuracy and efficiency in fault identification. In another comparative embodiment, multiple seismic attribute data volumes extracted from seismic data are used to identify faults at different scales, but this method is less efficient than multi-attribute fusion analysis. Therefore, in the embodiments described above, the superior noise resistance of machine learning models is utilized to fuse the first fault data volume identified by the machine learning model with other seismic attribute data volumes sensitive to faults and fractures at different scales, thereby improving the accuracy of fault and fracture identification. Simultaneously, multi-attribute fusion analysis improves the efficiency of fault and fracture identification.

[0092] In one specific embodiment of this application, step 104, extracting seismic attribute data volumes sensitive to faults at different scales and seismic attribute data volumes sensitive to fractures from the seismic data obtained in step 100, may include the following steps:

[0093] Extracting curvature from seismic data;

[0094] Extract the maximum likelihood from seismic data.

[0095] In one specific embodiment of this application, step 106, which involves multi-attribute fusion of the first fault data volume, the seismic attribute data volume sensitive to faults of different scales obtained in step 104, and the seismic attribute data volume sensitive to fractures, to identify faults and fractures of different scales in the target area, may include the following steps:

[0096] The first fault data volume, curvature volume, and maximum likelihood volume are fused using RGB to obtain the fused data volume;

[0097] The first-scale fault within the target area is identified using the first color data volume representing the first fault data volume in the fused data volume;

[0098] The second-scale fault within the target region is identified using the second color data volume representing the curvature volume in the fused data volume;

[0099] The third-scale fracture development zone associated with the fracture in the fracture network formed by the first-scale fault and the second-scale fault was identified by using the third color data volume representing the maximum likelihood volume in the fused data volume.

[0100] The first scale is larger than the second scale, and the second scale is larger than the third scale.

[0101] For example, machine learning models can be used to identify large-scale faults and their overlapping relationships. Large-scale faults refer to faults with a displacement of 10m or more. Mesoscale faults can be identified using curvature volumes. Mesoscale faults refer to faults with a displacement of less than 10m, especially faults with a displacement of less than 5m, i.e., hidden faults. Small-scale faults and fracture development zones associated with faults in the fracture network formed by first-scale and second-scale faults can be identified using maximum likelihood volumes.

[0102] As reservoir development progresses, the need for predicting faults and fractures at smaller scales arises in the later stages of development and during horizontal well drilling. Therefore, comprehensive identification of fault and fracture zones in seismic data is essential to adapt to the entire reservoir development process. As described in the above embodiments, this application starts with the comprehensive identification of fault and fracture zones in seismic data, selecting different attributes to identify fault and fracture development at different scales. This achieves progressive fault and fracture prediction at different scales, playing a crucial role in horizontal well deployment, drilling and fracturing parameter optimization, preventing wellbore collapse, preventing lost circulation, and preventing casing deformation. It ensures normal drilling, completion, and fracturing of horizontal wells, and guides mid-to-late-stage reservoir development through fault and fracture prediction results, thereby ensuring the effectiveness of reservoir development.

[0103] In one specific embodiment of this application, extracting the curvature volume from seismic data may include the following steps:

[0104] Seismic data is scanned by azimuth to obtain seismic sub-data volumes with different azimuths;

[0105] Curvature attributes were extracted from each seismic sub-data volume to obtain each curvature volume in a one-to-one correspondence.

[0106] In the above embodiments, by scanning seismic data in different azimuths, curvature volumes in different azimuths are obtained. Fault identification is performed using the obtained different curvature volumes. Compared with curvature volumes in a single azimuth, fault characteristics can be more comprehensively and from multiple angles, thereby improving the accuracy of fault identification.

[0107] In another specific embodiment of this application, extracting the curvature volume from seismic data may include the following steps:

[0108] Seismic data is scanned by azimuth to obtain seismic sub-data volumes with different azimuths;

[0109] Curvature attributes were extracted from each seismic sub-data volume to obtain each curvature volume in a one-to-one correspondence.

[0110] Each curvature volume is used to perform tomographic identification. The curvature volume that is optimal for tomographic identification at the second scale is determined as the curvature volume for RGB fusion. This triggers the RGB fusion of the first tomographic data volume, the curvature volume, and the maximum likelihood volume to obtain the fused data volume.

[0111] As in the above embodiment, by selecting the curvature body with the optimal orientation and using the curvature body with the optimal orientation for fault identification, although it cannot fully characterize the fault features compared to using different curvature bodies for fault identification, it can exclude curvature bodies with poor orientation for fault identification and reduce interference.

[0112] In one specific embodiment of this application, constructing a machine learning model may include the following steps:

[0113] Training samples were constructed using historical earthquake data and fault labels;

[0114] The initial machine learning network is trained and its parameters are optimized using the constructed training samples to obtain a machine learning model for fault identification. The machine learning network uses seismic data as input data and fault data as output data.

[0115] For example, the machine learning model is a deep neural network model, and the deep neural network used in the deep neural network model can be a U-net network, etc.

[0116] Figure 2 A second flowchart illustrating a fault and fracture prediction method according to an embodiment of this application is shown schematically. Figure 2 As shown, the fault and fracture prediction method provided in this application, based on the above embodiments, further includes the following steps:

[0117] Step 101: Preprocess the seismic data.

[0118] In a preferred embodiment, step 101 involves preprocessing the seismic data, including the following steps:

[0119] Low-frequency and high-frequency seismic data are recovered from the seismic data to obtain bilaterally frequency-extended seismic data;

[0120] The seismic data after bilateral frequency extension is divided into frequencies, and the different frequency bands obtained by frequency division are denoised separately to obtain different frequency data volumes.

[0121] The different frequency data volumes obtained after denoising are reconstructed to obtain reconstructed seismic data, thereby triggering step 102.

[0122] As described in the above embodiment, by combining the multi-resolution characteristics of seismic data in the continuous wavelet domain, the lost low-frequency and high-frequency signals are recovered through methods such as suppressing the sidelobes of the seismic signal wavelet, thereby achieving bilateral frequency extension of the seismic signal. The wider the bandwidth of the seismic wavelet, the higher the resolution, thus improving the resolution of the seismic data after bilateral frequency extension. Specifically, the recovery of low-frequency and high-frequency seismic data can be combined with the filtering algorithm in the ordinary embodiment, which will not be elaborated on in this embodiment. In addition, considering the characteristic of the signal-to-noise ratio (SNR) changing with frequency bands, generally, the SNR of the dominant frequency wavelet profile is high, while the SNR of the low-frequency and high-frequency wavelet profiles is low. Therefore, frequency division denoising is performed on the seismic data after bilateral frequency extension to improve the SNR. Through bilateral frequency extension, frequency division denoising, and reconstruction, the quality of seismic data is improved, thereby enhancing the accuracy of fault and fracture prediction.

[0123] For example, the bandwidth of earthquake data can be widened from 10–60 Hz to 5–80 Hz.

[0124] Figure 3 A third flowchart illustrating a fault and fracture prediction method according to an embodiment of this application is shown schematically. Figure 3 As shown, the fault and fracture prediction method provided in this application, based on the above embodiments, further includes the following steps:

[0125] Step 108: Determine the crack distribution characteristics of the target area.

[0126] In a preferred embodiment, determining the crack distribution characteristics of the target area may include the following steps:

[0127] Step 1081: Enhance the linear structure in the seismic attribute data volume that is sensitive to cracks to obtain a crack planar map of the target area.

[0128] In this application, the linear structure is a crack. To enhance the linear structure in the image, image segmentation or detection methods based on Hessian matrix or Jacobi matrix, or other image segmentation algorithms, can be used.

[0129] Step 1082: Vectorize the crack plan view of the target area and identify the location of each crack line segment.

[0130] Step 1083: Obtain the crack distribution characteristics of the target area based on the identified crack line segment locations. The crack distribution characteristics include at least one of crack density characteristics and crack direction characteristics.

[0131] As in the above embodiment, by enhancing and vectorizing the linear structure of the seismic attribute data volume that is sensitive to fractures, quantitative prediction of fractures is achieved, realizing the transition from qualitative to quantitative prediction of fractures, so as to support reservoir exploration and development, deployment of horizontal wells, drilling of horizontal wells, and optimization of horizontal well fracturing parameters.

[0132] For example, in one specific embodiment, in step 1081, the Hessian matrix can be used to enhance the linear structure, and the specific process is as follows:

[0133] Seismic profiles are generated using seismic attribute data volumes that are sensitive to cracks.

[0134] Construct the Hessian matrix corresponding to this seismic profile;

[0135] Construct Gaussian functions with different scale spatial factors, and calculate the second-order partial derivatives of each Gaussian function with respect to the horizontal and vertical coordinates of the seismic profile.

[0136] The second-order partial derivatives of each Gaussian function are convolved with the Hessian matrix to obtain the first matrix. Based on the eigenvalues ​​and eigenvectors in the first matrix, cracks and point structures in the seismic profile are identified.

[0137] A response function is constructed based on eigenvalues ​​and eigenvectors. The response function is then used to enhance cracks in linear structures and filter out point structures, resulting in a crack planar map of the target area.

[0138] In this application, the dot-like structure may include noise points, etc.

[0139] For example, in one specific embodiment, step 1082, vectorizing the crack plan view of the target area and identifying the location of each crack segment, includes the following steps:

[0140] Hough transformation is performed on the crack plan view of the target area to identify the location of crack segments.

[0141] Combination Figure 4 and Figure 5 As shown, the Hough transform principle is: It transforms the problem of detecting a given straight line in the original image into finding a peak value in the parameter space. Assume there exists a straight line y = a1x + b1 in the image, as... Figure 4 As shown, the parametric equation of this line in polar coordinates is xcosθ + ysinθ = ρ. The parametric plane corresponds to the curve (θ, ρ). The line in the (x, y) plane corresponding to the point where the lines intersect most frequently in the parametric plane is the desired line. Using this method, the location of the crack segment can be determined from the crack planar map of the target area, and its coordinates can be output.

[0142] To facilitate Hough transformation and improve the efficiency of crack distribution identification, after obtaining the crack planar map of the target area, the cracks and background in the crack planar map of the target area are binarized to obtain the binarized crack planar map of the target area. Accordingly, step 1082, the crack planar map of the target area is vectorized to identify the position of each crack segment, including the following steps:

[0143] Hough transformation is performed on the crack plane map of the binarized target area to identify the location of crack segments.

[0144] Figure 6 The diagram illustrates a specific application example: a technical flowchart for predicting fault layers and cracks. For example... Figure 6 As shown, a specific application example is the application of the above embodiments in research area A. In this application example, the maximum likelihood volume, the first fault data volume identified by the machine learning model, and the azimuth curvature volume, which has the best fault identification effect for research area A, are subjected to multi-attribute fusion analysis to identify fault and fracture development zones within research area A. The fracture density and fracture direction of the fracture development zones are then quantitatively identified. The machine learning model uses a deep neural network model for three-dimensional fault identification, and the network structure within the deep neural network model is as follows: Figure 7 As shown, Figure 7 In this context, EPA (Efficient paired attention block) represents an efficient paired attention module, Downsampling represents downsampling, Upsampling represents upsampling, Conv1×1×1 represents a convolution with a 1×1×1 kernel, ConvBlock represents a convolution block, Addition represents addition, Dot-product represents dot product, softmax represents the softmax activation function, Transpose represents transpose, and SkipConnection represents a skip connection.

[0145] The prediction process for the fault and fracture development zones in study area A is as follows:

[0146] Step 1: Establish a 3D seismic study area and load the post-stack narrow azimuth seismic data of study area A, as well as the well logging data and interpretation results, to obtain 3D seismic study area data for fault and fracture identification. The well logging data includes wellhead coordinates, core elevation, well trajectory and logging curves, etc., and the interpretation results include target layer position, fault data, etc.

[0147] Step 2: Improve the quality of the 3D seismic data, including bilateral upscaling, frequency division denoising, and reconstruction. The reconstructed 3D seismic data will be used in subsequent steps. In particular, the frequency band is widened from the original 10-60Hz to 5-80Hz through bilateral upscaling. Figure 8This is the improved seismic profile. Figure 9 Original seismic profile, for comparison Figure 8 and Figure 9 As indicated by the red arrow, the improved seismic profile shows clear breakpoints, highlighting the imaging effect of the phase axis misalignment at the fracture location.

[0148] Step 3: Using the constructed deep neural network model, identify relatively large faults in the reconstructed 3D seismic data obtained in Step 2 that have no obvious fault displacement on the seismic profile but have an actual fault displacement of more than 10m. This identifies the first-scale faults (large-scale faults) and generates the first fault data volume. The Q1 layer in the generated first fault data volume is shown below. Figure 10 As shown.

[0149] Step 4: Perform azimuth scanning on the reconstructed 3D seismic data obtained in Step 2 to obtain seismic sub-data volumes with different azimuths, called azimuth data volumes. Extract curvature attributes from the azimuth data volumes to obtain curvature volumes with different azimuths, called azimuth curvature volumes. Select the azimuth curvature volume with the best effect on identifying hidden faults. Hidden faults refer to second-scale faults (mesoscale faults) with a fault displacement of less than 5m. Figure 11 and Figure 12 The Q1 layers in the 150° azimuth curvature volume and the 90° azimuth curvature volume are compared. Figure 11 and Figure 12 It can be seen that, Figure 11 The 150° azimuth curvature body shown can identify northeast-trending faults, i.e., hidden faults, using the original seismic profile generated from un-qualityd 3D seismic survey data (e.g.) Figure 9 The cross-sectional interpretation failed to identify these northeast-trending faults. Coherence volume extraction was performed using 3D seismic data from the work area without quality improvement, but the extracted coherence volume still failed to identify these northeast-trending faults. Figure 12 Since the 90° azimuth curvature body shown cannot identify northeast-trending faults, the 150° azimuth curvature body can be used as the azimuth curvature body for identifying hidden faults in subsequent steps.

[0150] Step 5: Conduct a physical simulation experiment to verify the formation mechanism of concealed faults and confirm the effectiveness of using the 150° azimuth curvature body obtained in Step 4 for concealed fault identification. The physical simulation experiment design is as follows: Figure 13 As shown, inelastic canvas was used as the base, rubber cloth as the weakest point of the base, the extrusion speed was 0.1 cm / min, dry quartz sand was used as the experimental material, the sand body thickness was 30 mm, and a thin layer of clay was sprinkled on top with a thickness of 1 mm. Combined with... Figure 14As shown, physical simulation experiments were conducted to demonstrate the fracture formation characteristics under different angles between the late-stage compression direction and the pre-existing fracture structure. The simulation results show that different angles between the compression direction and the pre-existing fracture result in different crack directions. Combined with... Figure 15 As shown, based on the actual data of study area A, study area A is mainly subjected to late-stage compressive forces with a direction of 45° to the pre-existing fault, resulting in NEE-trending (northeast-east) faults, which are late-stage active faults. The fractures associated with the NEE-trending faults can intersect with the fractures formed by the NW-trending faults, forming a fracture network together. The faults identified based on the fault prediction schemes mentioned in the seismic profile or background technology are mainly northwest-trending.

[0151] Step 6: Extract the maximum likelihood from the reconstructed 3D seismic data obtained in Step 2. The extracted maximum likelihood can identify the fracture development zone associated with the fracture within the fracture network formed by large-scale and medium-scale faults.

[0152] Step 7: Perform RGB multi-attribute fusion on the first fault data volume, the 150° azimuth curvature volume, and the maximum likelihood volume. The RGB multi-attribute fusion result of layer Q1 is as follows: Figure 16 As shown, by Figure 16 As can be seen, the darker colored faults with better continuity in the figure are large-scale faults that can be identified from the seismic profile, the pink ones are the cracks that accompany the faults, and the blue ones are the possible locations where cracks that do not respond to earthquakes may develop.

[0153] Step 8: The linear structure in the maximum likelihood volume extracted in Step 6 is enhanced using a multi-scale Hessian matrix constructed with Gaussian functions. Point structures, including noise points, are filtered out, resulting in a crack planar map of the target region. This map is then binarized, and a Hough transform is applied to determine the positions of discrete crack segments. This completes the crack segment extraction process, yielding the orientation and density planar maps of the crack development zone. The density planar map of the crack development zone in layer Q1 is shown below. Figure 17 As shown, the directional plan view of the fracture development zone in layer Q1 is as follows. Figure 18 As shown.

[0154] To verify the effectiveness of the above embodiments in optimizing horizontal well deployment, drilling, and fracturing parameters, in a specific application, three wells were deployed. Using the fault and fracture prediction method provided in the above embodiments, a fault with a displacement of approximately 1 meter was predicted at a depth of 4120m. Figure 19As shown, the fault location was slightly uplifted, and this changed during drilling. The formation dip angle was adjusted from 89.3 degrees to 88.6 degrees to ensure a sweet spot encounter rate of over 90% for the horizontal well. Regarding the optimization of fracturing parameters and prevention of casing deformation, Well 1 indicated three risk zones, two of which showed significant responses during fracturing. See details below. Figure 20 The discharge rate of the 12th section (4035-3985m) of Well 1 is 16-11m³. 3 / min, sand volume 130m 3 , liquid volume 1610m 3 The pump stop pressure was 44 MPa, the fracture development was weak, and the sand addition was successful. The discharge rate of the 26th section (3250-3190m) of Well 1 was 14-12 m³ / s. 3 / min, sand volume 106m 3 , liquid volume 1648m 3 The pump was shut down at 50 MPa. During the gel-sand addition stage, when the sand velocity was 1.2, the formation pressure rose rapidly, leading to overpressure shutdown. Analysis indicated strong natural fracture development, significant fluid loss, and inability to form a main fracture. Sand bridge accumulation caused sand blockage, making fracturing operations extremely difficult. (See details...) Figure 21 This confirmed the existence and relatively high level of small-to-medium scale fractures. Wells 2 and 3 each indicated one risk zone, and abnormal responses were observed during fracturing in both. The low construction and shutdown pressures and ease of sand addition in the 21st section of Well 3 and the 28th section of Well 2 confirmed the existence of small-to-medium scale fractures, although their level was low. Based on the fault and fracture prediction method provided in the above embodiments of this application, the fault and fracture conditions in the three risk zones were predicted, and fracturing was performed in advance using a pressure reduction method. No casing deformation or pressure channeling occurred in any of these wells.

[0155] This application also provides a fault and fracture prediction device, including:

[0156] The seismic data acquisition module is used to acquire seismic data for the target area.

[0157] The first fault data volume prediction module is used to identify faults in the seismic data using a pre-built machine learning model to obtain the first fault data volume.

[0158] The seismic attribute extraction module is used to extract seismic attribute data volumes that are sensitive to faults at different scales and seismic attribute data volumes that are sensitive to fractures from the seismic data.

[0159] The fault and fracture identification module is used to perform multi-attribute fusion of the first fault data volume, the seismic attribute data volume sensitive to faults of different scales, and the seismic attribute data volume sensitive to fractures, so as to identify faults and fractures of different scales in the target area.

[0160] Optionally, the device further includes a frequency extension module, a frequency division noise reduction module, and a reconstruction module, wherein:

[0161] The frequency extension module is used to recover low-frequency and high-frequency seismic data from the seismic data to obtain bilaterally frequency-extended seismic data.

[0162] The frequency division and denoising module is used to divide the seismic data after bilateral frequency extension into frequencies, and to denoise the different frequency bands obtained by frequency division, so as to obtain different frequency data volumes.

[0163] The reconstruction module is used to reconstruct the different frequency data volumes obtained after denoising to obtain reconstructed seismic data, so as to trigger the identification of faults in the seismic data using a pre-built machine learning model to obtain the first fault data volume.

[0164] Optionally, the device further includes a linear structure enhancement module, a vectorization module, and a crack distribution feature determination module, wherein:

[0165] The linear structure enhancement module is used to enhance the linear structure in the seismic attribute data volume that is sensitive to cracks, so as to obtain a crack plan view of the target area.

[0166] The vectorization module is used to vectorize the crack plan view of the target area and identify the location of each crack line segment;

[0167] The crack distribution feature determination module is used to obtain the crack distribution features of the target area based on the identified crack line segment positions, wherein the crack distribution features include at least one of crack density features and crack direction features.

[0168] Optionally, seismic attribute data volumes sensitive to faults at different scales and seismic attribute data volumes sensitive to fractures are extracted from the seismic data, including:

[0169] Extract the curvature volume from the earthquake data;

[0170] Extract the maximum likelihood from the earthquake data.

[0171] Optionally, multi-attribute fusion is performed on the first fault data volume, the seismic attribute data volume sensitive to faults at different scales, and the seismic attribute data volume sensitive to fractures to identify faults and fractures at different scales in the target area, including:

[0172] The first fault data volume, curvature volume, and maximum likelihood volume are fused using RGB to obtain the fused data volume;

[0173] The first color data volume representing the first fault data volume in the fused data volume is used to identify the first scale fault within the target area;

[0174] The second color data volume, representing the curvature volume, in the fused data volume is used to identify the second-scale fault within the target area;

[0175] The third-scale fracture development zone associated with the fracture in the fracture network formed by the first-scale fault and the second-scale fault is identified by using the third color data volume representing the maximum likelihood in the fused data volume.

[0176] The first scale is larger than the second scale, and the second scale is larger than the third scale.

[0177] Optionally, the machine learning model is a deep neural network model.

[0178] Optionally, the step of enhancing the linear structure in the crack-sensitive seismic attribute data volume to obtain a crack planar map of the target area includes:

[0179] Seismic profiles are generated using seismic attribute data volumes that are sensitive to cracks.

[0180] Construct the Hessian matrix corresponding to the seismic profile;

[0181] Construct Gaussian functions with different scale spatial factors, and calculate the second-order partial derivatives of each Gaussian function with respect to the horizontal and vertical coordinates of the seismic profile.

[0182] The second-order partial derivatives of each Gaussian function are convolved with the Hessian matrix to obtain the first matrix. The cracks and point structures of the linear structure in the seismic profile are then identified based on the eigenvalues ​​and eigenvectors in the first matrix.

[0183] A response function is constructed based on the eigenvalues ​​and eigenvectors. The response function is then used to enhance the cracks in the linear structure and filter out the point structures, resulting in a crack plan view of the target area.

[0184] Optionally, the step of enhancing the linear structure in the crack-sensitive seismic attribute data volume to obtain a crack plan view of the target area further includes:

[0185] Binarize the cracks and background in the crack planar image of the target area to obtain the binarized crack planar image of the target area.

[0186] Optionally, the vectorized target area crack planar map identifies the location of each crack segment, including:

[0187] Hough transformation is performed on the crack plan view of the target area to identify the location of crack segments.

[0188] Optionally, extracting the curvature volume from the seismic data includes:

[0189] The seismic data is scanned by azimuth to obtain seismic sub-data volumes with different azimuths;

[0190] Curvature attributes were extracted from each seismic sub-data volume, and each curvature volume was obtained by corresponding to the others.

[0191] Optionally, the seismic data is post-stack narrow azimuth seismic data.

[0192] Figure 22 A schematic block diagram of a computer device according to an embodiment of the present application is shown. In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown below. Figure 22 As shown in the figure, the computer device includes a processor A01, a network interface A02, a display screen A04, an input device A05, and a memory (not shown) connected via a system bus. The processor A01 provides computing and control capabilities. The memory includes internal memory A03 and a non-volatile storage medium A06. The non-volatile storage medium A06 stores an operating system B01 and a computer program B02. The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 stored in the non-volatile storage medium A06. The network interface A02 is used for communication with external terminals via a network connection. When the computer program is executed by the processor A01, it implements a method for predicting faults and cracks. The display screen A04 can be a liquid crystal display (LCD) or an e-ink display. The input device A05 can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0193] Those skilled in the art will understand that Figure 22 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0194] In one embodiment, the fault and fracture prediction device provided in this application can be implemented as a computer program, which can be implemented in the form of, for example, Figure 22 The device operates on the computer shown. The computer's memory can store various program modules that make up the fault and fracture prediction device. The computer program, composed of these program modules, causes the processor to execute the steps in the fault and fracture prediction methods of the various embodiments of this application described in this specification.

[0195] In one embodiment, this application also provides a machine-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the fault and fracture prediction method in the above embodiments.

[0196] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0197] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0198] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0199] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for predicting faults and fractures, characterized in that, include: Acquire seismic data for the target area; The faults in the earthquake data are identified using a pre-built machine learning model to obtain the first fault data volume; Seismic attribute data volumes sensitive to faults at different scales and seismic attribute data volumes sensitive to fractures are extracted from the seismic data. The first fault data volume, the seismic attribute data volume sensitive to faults of different scales, and the seismic attribute data volume sensitive to fractures are fused to identify faults and fractures of different scales in the target area.

2. The method according to claim 1, characterized in that, Also includes: The seismic data is recovered by performing low-frequency and high-frequency seismic data recovery to obtain bilaterally frequency-extended seismic data; The seismic data after bilateral frequency extension is divided into frequencies, and the different frequency bands obtained by frequency division are denoised separately to obtain different frequency data volumes. The different frequency data volumes obtained after denoising are reconstructed to obtain reconstructed seismic data, which triggers the identification of faults in the seismic data using a pre-built machine learning model to obtain the first fault data volume.

3. The method according to claim 1, characterized in that, Also includes: The linear structure in the seismic attribute data volume that is sensitive to cracks is enhanced to obtain a crack planar map of the target area; Vectorize the crack plan view of the target area and identify the location of each crack line segment; The crack distribution characteristics of the target area are obtained based on the identified crack segment locations, and the crack distribution characteristics include at least one of crack density characteristics and crack direction characteristics.

4. The method according to claim 1, characterized in that, Extracting seismic attribute data volumes sensitive to faults at different scales and seismic attribute data volumes sensitive to fractures from the seismic data includes: Extract the curvature volume from the earthquake data; Extract the maximum likelihood from the earthquake data.

5. The method according to claim 4, characterized in that, The first fault data volume, seismic attribute data volumes sensitive to faults at different scales, and seismic attribute data volumes sensitive to fractures are fused using multi-attribute fusion to identify faults and fractures at different scales in the target area, including: The first fault data volume, curvature volume, and maximum likelihood volume are fused using RGB to obtain the fused data volume; The first color data volume representing the first fault data volume in the fused data volume is used to identify the first scale fault within the target area; The second color data volume, representing the curvature volume, in the fused data volume is used to identify the second-scale fault within the target area; The third-scale fracture development zone associated with the fracture in the fracture network formed by the first-scale fault and the second-scale fault is identified by using the third color data volume representing the maximum likelihood in the fused data volume. The first scale is larger than the second scale, and the second scale is larger than the third scale.

6. The method according to claim 1, characterized in that, The machine learning model is a deep neural network model.

7. The method according to claim 3, characterized in that, The process of enhancing the linear structure in the seismic attribute data volume sensitive to cracks to obtain a crack planar map of the target area includes: Seismic profiles are generated using seismic attribute data volumes that are sensitive to cracks. Construct the Hessian matrix corresponding to the seismic profile; Construct Gaussian functions with different scale spatial factors, and calculate the second-order partial derivatives of each Gaussian function with respect to the horizontal and vertical coordinates of the seismic profile. The second-order partial derivatives of each Gaussian function are convolved with the Hessian matrix to obtain the first matrix. The cracks and point structures of the linear structure in the seismic profile are then identified based on the eigenvalues ​​and eigenvectors in the first matrix. A response function is constructed based on the eigenvalues ​​and eigenvectors. The response function is then used to enhance the cracks in the linear structure and filter out the point structures, resulting in a crack plan view of the target area.

8. The method according to claim 7, characterized in that, The method of enhancing the linear structure in the seismic attribute data volume sensitive to cracks to obtain a crack planar map of the target area also includes: Binarize the cracks and background in the crack planar image of the target area to obtain the binarized crack planar image of the target area.

9. The method according to any one of claims 3, 7, and 8, characterized in that, The vectorized target area crack planar map identifies the location of each crack line segment, including: Hough transformation is performed on the crack plan view of the target area to identify the location of crack segments.

10. The method according to claim 4, characterized in that, Extracting the curvature volume from the seismic data includes: The seismic data is scanned by azimuth to obtain seismic sub-data volumes with different azimuths; Curvature attributes were extracted from each seismic sub-data volume, and each curvature volume was obtained by corresponding to the others.

11. The method according to claim 1, characterized in that, The earthquake data mentioned are post-stack narrow azimuth earthquake data.

12. A fault and fracture prediction device, characterized in that, include: The seismic data acquisition module is used to acquire seismic data for the target area. The first fault data volume prediction module is used to identify faults in the seismic data using a pre-built machine learning model to obtain the first fault data volume. The seismic attribute extraction module is used to extract seismic attribute data volumes that are sensitive to faults at different scales and seismic attribute data volumes that are sensitive to fractures from the seismic data. The fault and fracture identification module is used to perform multi-attribute fusion of the first fault data volume, the seismic attribute data volume sensitive to faults of different scales, and the seismic attribute data volume sensitive to fractures, so as to identify faults and fractures of different scales in the target area.

13. The apparatus according to claim 12, characterized in that, The first fault data volume, seismic attribute data volumes sensitive to faults at different scales, and seismic attribute data volumes sensitive to fractures are fused using multi-attribute fusion to identify faults and fractures at different scales in the target area, including: The first fault data volume, curvature volume, and maximum likelihood volume are fused using RGB to obtain the fused data volume; The first color data volume representing the first fault data volume in the fused data volume is used to identify the first scale fault within the target area; The second color data volume, representing the curvature volume, in the fused data volume is used to identify the second-scale fault within the target area; The third-scale fracture development zone associated with the fracture in the fracture network formed by the first-scale fault and the second-scale fault is identified by using the third color data volume representing the maximum likelihood in the fused data volume. The first scale is larger than the second scale, and the second scale is larger than the third scale.

14. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the fault and fracture prediction method according to any one of claims 1-11.

15. A machine-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the fault and fracture prediction method according to any one of claims 1-11.

Images