Crack identification method and device

Abstract

The invention provides a crack identification method and device, and the method comprises the steps: carrying out the word segmentation, coding and vectorization processing of obtained geology and exploration data; establishing an underground trap feature recognition model based on a pre-constructed trap knowledge graph and the vectorized geological and exploration data; extracting target geology and exploration data thereof from the obtained target document; performing word segmentation, coding and vectorization processing on the target geology and exploration data thereof; and inputting the vectorized target geology and exploration data thereof into the underground trap feature recognition model to obtain trap features corresponding to the target document. An underground trap feature recognition model is established by constructing a trap knowledge spectrogram and cooperating with geology and exploration data, and then trap feature information such as stratum, geological attributes and trap types in a target document is obtained through the underground trap feature recognition model.

Description

Technical Field

[0001] The present invention relates to the technical field of fracture prediction and treatment, and particularly to a method and device for identifying fractures. Background Art

[0002] With the advent of the "geological big data" era, the potential of artificial intelligence in the field of oil and gas exploration has been increasingly recognized. Compared with the traditional methods that rely on expert experience and prior models, intelligent algorithms represented by machine learning have advantages such as data mining, complex problem interpretation, and automated prediction.

[0003] As one of the key technologies for reducing exploration risks, the current commonly used fracture reservoir prediction methods mainly include seismic attribute analysis and inversion. However, due to the complex properties of carbonate reservoirs, the accuracy and clarity of the identification results obtained by the existing prediction methods still need to be improved. Summary of the Invention

[0004] The present invention provides a method and device for identifying fractures. By using a trained neural network to identify fractures from seismic images, it is of great significance for subsequent fracture treatment.

[0005] In a first aspect, the present invention provides a method for identifying fractures, including:

[0006] Generating simulated seismic data through the obtained horizontal reflectivity model;

[0007] Establishing a training data set according to the simulated seismic data;

[0008] Establishing a target MultiResUNet network by using the training data set;

[0009] Inputting the seismic data to be identified into the target MultiResUNet network to obtain the corresponding fracture identification result.

[0010] Optionally, generating simulated seismic data through the obtained horizontal reflectivity model includes:

[0011] Adding folding structures, plane shears, and plane faults to the horizontal reflectivity model by using random numbers in [-1, 1] to obtain a target horizontal reflectivity model;

[0012] Convolving the reflection coefficient in the target horizontal reflectivity model with a Ricker wavelet to generate initial simulated seismic data;

[0013] Adding random noise to the initial simulated seismic data to obtain the simulated seismic data.

[0014] Optionally, after convolving the reflection coefficient in the target horizontal reflectivity model with a Ricker wavelet to generate the simulated seismic data, the following steps are further included:

[0015] Flip the simulated seismic data vertically.

[0016] Optionally, the training data set includes: a training set; using the training data set to establish a target MultiResUNet network, including:

[0017] Input the simulated seismic data in the training set into the constructed MultiResUNet network to obtain the corresponding recognition category;

[0018] Based on the recognition category and recognition label of the training set, adjust the network parameters of the MultiResUNet network to obtain the target MultiResUNet network.

[0019] Optionally, the training data set includes: a validation set; based on the recognition category and recognition label of the training set, adjust the network parameters of the MultiResUNet network to obtain the target MultiResUNet network, including:

[0020] Determine the training error based on the recognition category and recognition label of the training set;

[0021] According to the training error, adjust the network parameters of the MultiResUNet network through the backpropagation algorithm, and use the validation set to measure the generalization ability of the adjusted MultiResUNet network to obtain the optimal network parameters;

[0022] Generate the target data prediction neural network using the optimal network parameters.

[0023] In a second aspect, the present invention provides a crack recognition device, including:

[0024] A simulated seismic data generation module for generating simulated seismic data through the acquired horizontal reflectivity model;

[0025] A training data set establishment module for establishing a training data set according to the simulated seismic data;

[0026] A network establishment module for establishing a target MultiResUNet network using the training data set;

[0027] A prediction module for inputting the seismic data to be recognized into the target MultiResUNet network to obtain the corresponding crack recognition result.

[0028] Optionally, the simulated seismic data generation module includes:

[0029] A target model acquisition sub-module, configured to add fold structures, plane shears, and plane faults to the horizontal reflectivity model using random numbers in [-1, 1] to obtain a target horizontal reflectivity model;

[0030] A convolution sub-module, configured to convolve the reflection coefficients in the target horizontal reflectivity model with a Ricker wavelet to generate initial simulated seismic data;

[0031] A simulated data acquisition sub-module, configured to add random noise to the initial simulated seismic data to obtain the simulated seismic data.

[0032] Optionally, the simulated seismic data generation module further includes:

[0033] A data expansion sub-module, configured to flip the simulated seismic data in the vertical direction.

[0034] In a third aspect, the present application provides an electronic device, including a processor and a memory, where the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps in the method provided in the first aspect above are run, including:

[0035] Generating simulated seismic data through the obtained horizontal reflectivity model;

[0036] Establishing a training data set according to the simulated seismic data;

[0037] Establishing a target MultiResUNet network using the training data set;

[0038] Inputting the seismic data to be recognized into the target MultiResUNet network to obtain a corresponding crack recognition result.

[0039] Optionally, generating simulated seismic data through the obtained horizontal reflectivity model includes:

[0040] Adding fold structures, plane shears, and plane faults to the horizontal reflectivity model using random numbers in [-1, 1] to obtain a target horizontal reflectivity model;

[0041] Convolving the reflection coefficients in the target horizontal reflectivity model with a Ricker wavelet to generate initial simulated seismic data;

[0042] Adding random noise to the initial simulated seismic data to obtain the simulated seismic data.

[0043] Optionally, after convolving the reflection coefficient in the target horizontal reflectivity model with a Ricker wavelet to generate the simulated seismic data, the method further includes:

[0044] Flipping the simulated seismic data in the vertical direction.

[0045] Optionally, the training data set includes a training set; using the training data set to establish a target MultiResUNet network, including:

[0046] Inputting the simulated seismic data in the training set into the constructed MultiResUNet network to obtain the corresponding recognition category;

[0047] Adjusting the network parameters of the MultiResUNet network based on the recognition category and recognition label of the training set to obtain the target MultiResUNet network.

[0048] Optionally, the training data set includes a validation set; adjusting the network parameters of the MultiResUNet network based on the recognition category and recognition label of the training set to obtain the target MultiResUNet network, including:

[0049] Determining a training error based on the recognition category and recognition label of the training set;

[0050] Adjusting the network parameters of the MultiResUNet network according to the training error through a backpropagation algorithm, and using the validation set to measure the generalization ability of the adjusted MultiResUNet network to obtain the optimal network parameters;

[0051] Generating the target data prediction neural network using the optimal network parameters.

[0052] In a fourth aspect, the present application provides a storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it runs the steps in the method provided in the first aspect as described above, including:

[0053] Generating simulated seismic data through the obtained horizontal reflectivity model;

[0054] Establishing a training data set according to the simulated seismic data;

[0055] Establishing a target MultiResUNet network using the training data set;

[0056] Inputting the seismic data to be recognized into the target MultiResUNet network to obtain the corresponding crack recognition result.

[0057] Optionally, based on the obtained horizontal reflectivity model, simulated seismic data is generated, including:

[0058] Adding folding structures, planar shears, and planar faults using random numbers in the range [-1, 1] to the horizontal reflectivity model to obtain a target horizontal reflectivity model;

[0059] Convolving the reflection coefficients in the target horizontal reflectivity model with a Ricker wavelet to generate initial simulated seismic data;

[0060] Adding random noise to the initial simulated seismic data to obtain the simulated seismic data.

[0061] Optionally, after convolving the reflection coefficients in the target horizontal reflectivity model with a Ricker wavelet to generate the simulated seismic data, it further includes:

[0062] Flipping the simulated seismic data in the vertical direction.

[0063] Optionally, the training dataset includes a training set; using the training dataset to establish a target MultiResUNet network, including:

[0064] Inputting the simulated seismic data in the training set into the constructed MultiResUNet network to obtain the corresponding recognition categories;

[0065] Adjusting the network parameters of the MultiResUNet network based on the recognition categories and recognition labels of the training set to obtain a target MultiResUNet network.

[0066] Optionally, the training dataset includes a validation set; adjusting the network parameters of the MultiResUNet network based on the recognition categories and recognition labels of the training set to obtain a target MultiResUNet network, including:

[0067] Determining a training error based on the recognition categories and recognition labels of the training set;

[0068] Adjusting the network parameters of the MultiResUNet network according to the training error through backpropagation algorithm, and using the validation set to measure the generalization ability of the adjusted MultiResUNet network to obtain optimal network parameters;

[0069] Generating the target data prediction neural network using the optimal network parameters.

[0070] From the above technical solutions, it can be seen that the present invention has the following advantages:

[0071] The present invention provides a method and apparatus for crack identification. The method includes: generating simulated seismic data through the obtained horizontal reflectivity model; establishing a training data set according to the simulated seismic data; establishing a target MultiResUNet network by using the training data set; and inputting the seismic data to be identified into the target MultiResUNet network to obtain a corresponding crack identification result. Identifying cracks from seismic images through a trained neural network is of great significance for key steps in seismic structure interpretation, reservoir description, and well location arrangement. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other accompanying drawings can be obtained based on these drawings without creative efforts.

[0073] Figure 1 It is a flowchart of the first embodiment of a method for crack identification according to the present invention;

[0074] Figure 2 It is a flowchart of the second embodiment of a method for crack identification according to the present invention;

[0075] Figure 3 It is a schematic diagram of the network structure of the MultiResUNet network;

[0076] Figure 4 It is a schematic diagram of the structure of the MultiRes module;

[0077] Figure 5 It is a schematic diagram of the structure of the Res Path module;

[0078] Figure 6 It is a schematic diagram of the internal structure identification result obtained by using a method for trap feature identification according to the present invention;

[0079] Figure 7 It is a schematic diagram of the planar result of crack treatment obtained by analyzing using a method for trap feature identification according to the present invention;

[0080] Figure 8 It is a schematic diagram of the internal structure identification result obtained by analyzing using the existing method;

[0081] Figure 9 It is a schematic diagram of the planar result of crack treatment obtained by analyzing using the existing method;

[0082] Figure 10 It is a block diagram of the structure of an embodiment of a crack identification apparatus according to the present invention. Specific Embodiments

[0083] An embodiment of the present invention provides a method and device for crack identification. By using a trained neural network to identify cracks from seismic images, it is of great significance for subsequent crack processing.

[0084] In order to make the objectives, features, and advantages of the present invention more obvious and understandable, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the embodiments described below are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0085] Embodiment 1. Please refer to Figure 1 , Figure 1 which is a flowchart of Embodiment 1 of a method for crack identification according to the present invention, including:

[0086] S101. Generate simulated seismic data through the obtained horizontal reflectivity model;

[0087] In an optional embodiment, generating simulated seismic data through the obtained horizontal reflectivity model includes:

[0088] Add folding structures, planar shears, and planar faults to the horizontal reflectivity model using random numbers in [-1, 1] to obtain a target horizontal reflectivity model;

[0089] Convolve the reflection coefficients in the target horizontal reflectivity model with a Ricker wavelet to generate initial simulated seismic data;

[0090] Add random noise to the initial simulated seismic data to obtain the simulated seismic data.

[0091] It should be noted that the relevant attributes for crack prediction are all extracted from seismic data and can be used to describe the geometric, kinematic, dynamic, and statistical characteristics of the subsurface. These attributes can not only reflect the geometric characteristics of structures and reservoirs, including coherence, variance, edge detection, edge-preserving smoothing filtering, texture, structure-oriented filtering, energy gradient calculation, dip estimation, curvature, rose diagram, amplitude variation, etc. At the same time, these attributes can also reflect the attributes of different categories such as oil and gas and reservoirs, such as time-frequency analysis, single frequency, hydrocarbon detection, pre-stack / post-stack formation absorption coefficient, etc. Finally, these attributes can also be used for analysis categories, such as attribute ratio fusion, formation slicing, etc.

[0092] S102. Establish a training data set according to the simulated seismic data;

[0093] S103. Establish a target MultiResUNet network using the training data set.

[0094] In an alternative embodiment, the training data set includes a training set. Establishing a target MultiResUNet network using the training data set includes:

[0095] Input the simulated seismic data in the training set into the constructed MultiResUNet network to obtain the corresponding recognition categories.

[0096] Adjust the network parameters of the MultiResUNet network based on the recognition categories and labels of the training set to obtain the target MultiResUNet network.

[0097] S104. Input the seismic data to be recognized into the target MultiResUNet network to obtain the corresponding fracture recognition result.

[0098] In an identification method for trap features provided in an embodiment of the present invention, it includes: performing word segmentation, encoding, and vectorization processing on the obtained geology and exploration data; establishing an underground trap feature identification model based on a pre-constructed trap knowledge graph and the geology and exploration data after vectorization processing; extracting target geology and exploration data from the obtained target document; performing word segmentation, encoding, and vectorization processing on the target geology and exploration data; inputting the target geology and exploration data after vectorization processing into the underground trap feature identification model to obtain the trap features corresponding to the target document. By constructing a trap knowledge spectrum diagram and jointly establishing an underground trap feature identification model with geology and exploration data, trap feature information such as the formation, geological attributes, and trap types belonging to the target document can be obtained through the underground trap feature identification model.

[0099] Embodiment 2. Please refer to Figure 2 , Figure 2 which is a flowchart of Embodiment 2 of a fracture identification method of the present invention. The method includes:

[0100] S201. Add folding structures, planar shears, and planar faults to the horizontal reflectivity model using random numbers in [-1, 1] to obtain a target horizontal reflectivity model.

[0101] In an embodiment of the present invention, first add folding structures to the model using random numbers in [-1, 1], and then add planar shears and planar faults to the model. Among them, the folding structure utilizes a two-dimensional normal distribution function, and for the added planar faults, the dip angles and strikes on each fault are different.

[0102] S202, convolve the reflection coefficient in the target horizontal reflectivity model with a Ricker wavelet to generate initial simulated seismic data;

[0103] It should be noted that the Ricker wavelet is a commonly used waveform for simulating seismic data.

[0104] In the embodiment of the present invention, the target horizontal reflectivity model is convolved with a Ricker wavelet to simulate the waveform change when seismic waves pass through different strata, and then initial simulated seismic data is obtained.

[0105] S203, add random noise to the initial simulated seismic data to obtain the simulated seismic data;

[0106] In the embodiment of the present invention, random noise is added to the initial simulated seismic data to simulate the noise interference existing in actual exploration, thereby increasing the authenticity and complexity of the seismic data.

[0107] S204, flip the simulated seismic data in the vertical direction;

[0108] In the embodiment of the present invention, in order to expand the quantity of the simulated seismic data, the simulated seismic data obtained in step S203 is rotated by a certain angle in the numerical direction to obtain new simulated seismic data.

[0109] S205, establish a training data set according to the simulated seismic data; the training data set includes: a training set and a validation set;

[0110] In the embodiment of the present invention, using "1" and "0" as labels to represent "presence of cracks" and "absence of cracks" respectively, a training data set is formed, and then the training data set is divided into a training set and a validation set according to 10:1.

[0111] S206, input the simulated seismic data in the training set into the constructed MultiResUNet network to obtain the corresponding recognition category;

[0112] In the implementation of the present invention, Figure 3 the network structure of the MultiResUNet network is shown. Specifically, the MultiResUNet network is a new network structure proposed based on the U-Net. Similar to the U-Net network structure, the network is divided into an encoding part and a decoding part. On the basis of the U-Net, a multi-residual (MultiRes) module and a residual path (ResPath) module are introduced. The structure of the MultiRes module is as Figure 4As shown, this module draws on the idea of the Inception network, using convolutional kernels of different sizes of 3×3, 5×5, and 7×7 to extract features respectively, and splicing the output results in the channel direction to learn spatial features of different sizes. This module uses 2 3×3 convolutional layers to replace the 5×5 convolutional layer, and uses 3 3×3 convolutional layers to replace the 7×7 convolutional layer, reducing the number of network parameters and accelerating network training. At the same time, a residual structure using 1×1 convolution is introduced. This module is used to replace the convolutional layer of U-Net. In addition, the structure of the Res Path module is as Figure 5 shown. Instead of simply splicing the feature maps of the encoder and the decoder together directly, the feature maps of the encoder are passed through a series of convolutional layers. These additional non-linear operations can reduce the semantic gap between the encoder and decoder features. In addition, residual connections that make training easier and are very useful in deep convolutional networks are also introduced.

[0113] S207, based on the recognition categories and recognition labels of the training set, determine the training error;

[0114] S208, according to the training error, adjust the network parameters of the MultiResUNet network through the backpropagation algorithm, and use the validation set to measure the generalization ability of the adjusted MultiResUNet network to obtain the optimal network parameters;

[0115] S209, generate the target data prediction neural network using the optimal network parameters;

[0116] S210, input the seismic data to be recognized into the target MultiResUNet network to obtain the corresponding fracture recognition result.

[0117] A method for recognizing fractures in an embodiment of the present invention generates simulated seismic data through the obtained horizontal reflectivity model; establishes a training data set according to the simulated seismic data; establishes a target MultiResUNet network using the training data set; inputs the seismic data to be recognized into the target MultiResUNet network to obtain the corresponding fracture recognition result. By synthesizing simulated seismic data through the horizontal reflectivity model, training the MultiResUNet network, and then realizing fracture recognition through the trained MultiResUNet network, a fracture recognition result that is more accurate and clearer than the prior art is obtained. As a key step in seismic structure interpretation, reservoir description, and well location arrangement, the embodiment of the present invention has important significance.

[0118] For the convenience of those skilled in the art to understand the beneficial effects of the present invention, the following gives examples of the crack recognition results obtained by analyzing the present invention and the crack recognition results obtained by using the prior art for analysis.

[0119] Example 1: Using a crack recognition method provided by an embodiment of the present invention to perform internal structure recognition on the fracture-vug body of carbonate rock in Shunbei. The specific recognition results are as Figure 6 shown, and the planar result of crack processing is as Figure 7 shown. The recognition result of using the existing crack recognition method carried on software to perform internal structure recognition on the fracture-vug body of carbonate rock in Shunbei is as Figure 8 shown, and the planar result of crack processing is as Figure 9 shown. It can be seen that compared with the fracture-vug body slices obtained by the prior art, the recognition results obtained by using the method provided by the embodiment of the present invention are relatively clearer in the internal structure, relatively consistent with the logging data, and have better effects. To sum up, a crack recognition method provided by an embodiment of the present invention can improve the crack prediction result, make the anomaly more focused to a certain extent, and has a rigorous and reliable theory and a simple and practical operation process.

[0120] Example 3: Please refer to Figure 10 , Figure 10 which is a structural block diagram of an embodiment of a crack recognition device of the present invention, including:

[0121] A simulated seismic data generation module 301, configured to generate simulated seismic data through the acquired horizontal reflectivity model;

[0122] A training data set establishment module 302, configured to establish a training data set according to the simulated seismic data;

[0123] A network establishment module 303, configured to establish a target MultiResUNet network by using the training data set;

[0124] A prediction module 304, configured to input the seismic data to be recognized into the target MultiResUNet network to obtain the corresponding crack recognition result.

[0125] In an optional embodiment, the simulated seismic data generation module 301 includes:

[0126] A target model acquisition sub-module, configured to add a folding structure, a planar shear, and a planar fault to the horizontal reflectivity model by using random numbers in [-1, 1] to obtain a target horizontal reflectivity model;

[0127] A convolution sub-module, configured to convolve the reflection coefficient in the target horizontal reflectivity model with a Ricker wavelet to generate initial simulated seismic data;

[0128] A simulated data acquisition sub-module, configured to add random noise to the initial simulated seismic data to obtain the simulated seismic data.

[0129] In an alternative embodiment, the simulated seismic data generation module 301 further includes:

[0130] A data expansion sub-module, configured to flip the simulated seismic data in the vertical direction.

[0131] In an alternative embodiment, the training data set includes: a training set; the network establishment module 303 includes:

[0132] An input sub-module, configured to input the simulated seismic data in the training set into the constructed MultiResUNet network to obtain corresponding recognition categories;

[0133] A network establishment sub-module, configured to adjust the network parameters of the MultiResUNet network based on the recognition categories and recognition labels of the training set to obtain a target MultiResUNet network.

[0134] In an alternative embodiment, the training data set includes: a validation set; the network establishment sub-module includes:

[0135] A training error determination unit, configured to determine a training error based on the recognition categories and recognition labels of the training set;

[0136] A parameter adjustment unit, configured to adjust the network parameters of the MultiResUNet network by a backpropagation algorithm according to the training error, and use the validation set to measure the generalization ability of the adjusted MultiResUNet network to obtain optimal network parameters;

[0137] A network establishment unit, configured to generate the target data prediction neural network by using the optimal network parameters.

[0138] Embodiment 4, The embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor is caused to execute the steps of a method for identifying a crack as described in any one of the above embodiments, including:

[0139] Generate simulated seismic data through the obtained horizontal reflectivity model;

[0140] Establish a training data set according to the simulated seismic data;

[0141] Establish a target MultiResUNet network by using the training data set;

[0142] Input the seismic data to be recognized into the target MultiResUNet network to obtain the corresponding fracture recognition result.

[0143] In an alternative embodiment, generate simulated seismic data through the obtained horizontal reflectivity model, including:

[0144] Add folding structures, planar shears, and planar faults to the horizontal reflectivity model using random numbers in [-1, 1] to obtain the target horizontal reflectivity model;

[0145] Convolve the reflection coefficients in the target horizontal reflectivity model with the Ricker wavelet to generate initial simulated seismic data;

[0146] Add random noise to the initial simulated seismic data to obtain the simulated seismic data.

[0147] In an alternative embodiment, after convolving the reflection coefficients in the target horizontal reflectivity model with the Ricker wavelet to generate the simulated seismic data, it further includes:

[0148] Flip the simulated seismic data vertically.

[0149] In an alternative embodiment, the training data set includes: a training set; use the training data set to establish a target MultiResUNet network, including:

[0150] Input the simulated seismic data in the training set into the constructed MultiResUNet network to obtain the corresponding recognition categories;

[0151] Based on the recognition categories and recognition labels of the training set, adjust the network parameters of the MultiResUNet network to obtain the target MultiResUNet network.

[0152] In an alternative embodiment, the training data set includes: a validation set; based on the recognition categories and recognition labels of the training set, adjust the network parameters of the MultiResUNet network to obtain the target MultiResUNet network, including:

[0153] Determine the training error based on the recognition categories and recognition labels of the training set;

[0154] According to the training error, adjust the network parameters of the MultiResUNet network through the backpropagation algorithm, and use the validation set to measure the generalization ability of the adjusted MultiResUNet network to obtain the optimal network parameters;

[0155] Generate the target data prediction neural network by using the optimal network parameters.

[0156] Example 5. The embodiment of the present invention also provides a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of a crack recognition method as described in any one of the above embodiments are implemented, including:

[0157] Generate simulated seismic data through the obtained horizontal reflectivity model;

[0158] Establish a training data set according to the simulated seismic data;

[0159] Establish a target MultiResUNet network by using the training data set;

[0160] Input the seismic data to be recognized into the target MultiResUNet network to obtain the corresponding crack recognition result.

[0161] In an optional embodiment, generating simulated seismic data through the obtained horizontal reflectivity model includes:

[0162] Add folding structures, plane shears, and plane faults to the horizontal reflectivity model by using random numbers in [-1, 1] to obtain a target horizontal reflectivity model;

[0163] Convolve the reflection coefficient in the target horizontal reflectivity model with a Ricker wavelet to generate initial simulated seismic data;

[0164] Add random noise to the initial simulated seismic data to obtain the simulated seismic data.

[0165] In an optional embodiment, after convolving the reflection coefficient in the target horizontal reflectivity model with a Ricker wavelet to generate the simulated seismic data, it further includes:

[0166] Flip the simulated seismic data in the vertical direction.

[0167] In an optional embodiment, the training data set includes a training set. Establishing a target MultiResUNet network by using the training data set includes:

[0168] Input the simulated seismic data in the training set into the constructed MultiResUNet network to obtain the corresponding recognition categories;

[0169] Adjust the network parameters of the MultiResUNet network based on the recognition categories and recognition labels of the training set to obtain a target MultiResUNet network.

[0170] In an alternative embodiment, the training data set includes: a validation set; based on the recognition categories and recognition labels of the training set, adjusting the network parameters of the MultiResUNet network to obtain a target MultiResUNet network, including:

[0171] Determining a training error based on the recognition categories and recognition labels of the training set;

[0172] According to the training error, adjusting the network parameters of the MultiResUNet network through the backpropagation algorithm, and using the validation set to measure the generalization ability of the adjusted MultiResUNet network to obtain optimal network parameters;

[0173] Generating the target data prediction neural network using the optimal network parameters.

[0174] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0175] In several embodiments provided in the present application, it should be understood that the methods, devices, electronic devices, and storage media disclosed by the present invention can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0176] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0177] In addition, each functional unit in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0178] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a readable storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned readable storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs.

[0179] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for identifying cracks, characterized in that, Including: Generating simulated seismic data through the obtained horizontal reflectivity model; Establishing a training data set according to the simulated seismic data; Establishing a target MultiResUNet network using the training data set; Inputting the seismic data to be recognized into the target MultiResUNet network to obtain the corresponding fracture recognition result.

2. The method for identifying cracks according to claim 1, characterized in that, Generating simulated seismic data through the obtained horizontal reflectivity model, including: Adding fold structures, plane shears, and plane faults using random numbers in the range of [-1, 1] to the horizontal reflectivity model to obtain a target horizontal reflectivity model; Convolving the reflection coefficients in the target horizontal reflectivity model with a Ricker wavelet to generate initial simulated seismic data; Adding random noise to the initial simulated seismic data to obtain the simulated seismic data.

3. The method for identifying cracks according to claim 2, characterized in that, After convolving the reflection coefficients in the target horizontal reflectivity model with a Ricker wavelet to generate the simulated seismic data, it further includes: Flipping the simulated seismic data vertically.

4. The method for identifying cracks according to claim 1, characterized in that, The training data set includes a training set; establishing a target MultiResUNet network using the training data set, including: Inputting the simulated seismic data in the training set into the constructed MultiResUNet network to obtain the corresponding recognition categories; Adjusting the network parameters of the MultiResUNet network based on the recognition categories and recognition labels of the training set to obtain a target MultiResUNet network.

5. The method for identifying cracks according to claim 4, characterized in that, The training data set includes a validation set; adjusting the network parameters of the MultiResUNet network based on the recognition categories and recognition labels of the training set to obtain a target MultiResUNet network, including: Determining the training error based on the recognition categories and recognition labels of the training set; Adjusting the network parameters of the MultiResUNet network through backpropagation algorithm according to the training error, and using the validation set to measure the generalization ability of the adjusted MultiResUNet network to obtain the optimal network parameters; Generating the target data prediction neural network using the optimal network parameters.

6. A device for identifying cracks, characterized in that, Including: A simulated seismic data generation module for generating simulated seismic data through the obtained horizontal reflectivity model; A training data set establishment module for establishing a training data set according to the simulated seismic data; A network establishment module for establishing a target MultiResUNet network using the training data set; A prediction module for inputting the seismic data to be recognized into the target MultiResUNet network to obtain the corresponding fracture recognition result.

7. The device for identifying cracks according to claim 6, characterized in that, The simulated seismic data generation module includes: A target model acquisition sub-module for adding fold structures, plane shears, and plane faults using random numbers in the range of [-1, 1] to the horizontal reflectivity model to obtain a target horizontal reflectivity model; A convolution sub-module for convolving the reflection coefficients in the target horizontal reflectivity model with a Ricker wavelet to generate initial simulated seismic data; The simulated data acquisition sub-module is used to add random noise to the initial simulated seismic data to obtain the simulated seismic data.

8. The device for identifying cracks according to claim 7, characterized in that, The simulated seismic data generation module further includes: The data extension sub-module is used to flip the simulated seismic data in the vertical direction.

9. An electronic device, characterized in that, It includes a processor and a memory, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the method described in any one of claims 1-5 is run.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the method described in any one of claims 1-5 is run.

Images