Migration velocity modeling method, device, electronic device and medium

Abstract

This application discloses a migration velocity modeling method, apparatus, electronic device, and medium. The method includes: establishing a prestack time migration velocity model, an initial velocity model, and a CRP gather; generating a residual velocity spectrum using the CRP gather; obtaining training data by energy-enhancing and normalizing the residual velocity spectrum; obtaining the root mean square velocity value of a CDP point in the CRP gather and normalizing the initial velocity model at that point to obtain label data; creating a migration velocity modeling system using a convolutional neural network model and determining training parameters for the migration velocity modeling system; obtaining seismic data, predicting the seismic data using the migration velocity modeling system, and outputting a root mean square velocity model. The neural network model uses the generated training data and label data as input data to the network, extracting seismic data, and creating a migration velocity modeling system. This method is applicable to the extremely large amounts of data currently available in seismic exploration, efficiently and accurately completing the establishment of a prestack time migration velocity field.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas geophysical exploration engineering, and more particularly to a migration velocity modeling method, device, electronic equipment and medium. Background Art

[0002] In the field of oil and gas exploration, the accuracy of the seismic velocity model seriously affects the quality of seismic imaging. The seismic velocity modeling technologies currently used in the industry mainly include conventional velocity analysis based on CMP gathers in the time domain, residual curvature analysis based on prestack time migration gathers, and depth domain focusing analysis technology. Conventional stacking velocity analysis technology obtains a smoothed root mean square velocity field that includes formation dip and occurrence information, which reflects a relatively vague geological meaning. Generally speaking, the migration gathers obtained by pre-time migration using it cannot be flattened. Pre-stack time migration velocity modeling technology uses CRP gathers for velocity analysis, corrects the migration velocity field according to the degree of upwarping or downwarping of the migration gathers, and can iteratively improve the accuracy of velocity analysis. The current velocity analysis method based on CRP gathers generally adopts the velocity scanning method: that is, after determining the maximum and minimum CDP numbers, CDP intervals, offset distances and other information for velocity analysis in the work area, a series of velocity values ​​are used to perform constant-velocity pre-stack time migration, or a series of values ​​in the [0, 2] interval are used to multiply the initial velocity model, and the obtained velocity results are used for pre-stack time migration; finally, the velocity spectrum of the output CIG gathers is picked and interpolated to obtain the root mean square velocity field used for pre-stack time migration.

[0003] As seismic exploration evolves from two-dimensional to three-dimensional, and with the advancement of two-width-one-height acquisition technology, traditional migration velocity modeling techniques based on velocity scanning are becoming increasingly inadequate for such massive amounts of seismic data, requiring complex calculations and multiple iterations to obtain a relatively accurate velocity field. Furthermore, their accuracy cannot be guaranteed. With the advancement of computer hardware and mathematical algorithms, and the rise of big data processing and deep learning applications in recent years, traditional migration velocity modeling techniques are no longer sufficient.

[0004] Therefore, we hope to propose a migration velocity modeling method to solve the technical problems that traditional prestack migration velocity modeling technology is difficult to apply to today's huge data volume and velocity modeling accuracy. Summary of the Invention

[0005] The present invention proposes a migration velocity modeling method, device, electronic equipment and medium, which at least solve the technical problems that traditional pre-stack migration velocity modeling technology is difficult to apply to today's huge data volume and velocity modeling accuracy.

[0006] In a first aspect, an embodiment of the present disclosure provides a migration velocity modeling method, comprising:

[0007] Establish pre-stack time migration velocity model, initial velocity model and CRP gathers;

[0008] Generate residual velocity spectrum using CRP gathers;

[0009] Training data is obtained by energy enhancement and normalization of the residual velocity spectrum;

[0010] Get the root mean square velocity value of the CDP point of the CRP gather and normalize the initial velocity model of the point to obtain the label data;

[0011] Using a convolutional neural network model, creating a migration velocity modeling system and determining training parameters for the migration velocity modeling system;

[0012] Seismic data is acquired, the seismic data is predicted by the migration velocity modeling system, and a root mean square velocity model is output.

[0013] As a specific implementation of the embodiment of the present disclosure, generating a residual velocity spectrum using a CRP gather specifically includes: using the CRP gather to calculate the residual velocity spectrum of each CDP point using a residual velocity spectrum calculation formula, and the residual velocity spectrum calculation formula is as follows:

[0014]

[0015] Among them, Rspec(it) represents the residual velocity spectrum at the sampling point it, R amp (it) is the intermediate variable of the calculation, Cip(it) represents the imaging gather amplitude at the sampling point it, γ represents the Gamma value, h represents the offset size, ns represents the number of sampling points, dt represents the sampling point interval, (0.25h 2 (γ 2 -1)+(it*dt) 2 ) is the calculated residual time difference, ntrace represents the number of seismic traces contained in each data set; among them, the Gamma value with an interval of 0.015 between 0.1 and 1.9 is selected, and the residual velocity spectrum of each sampling point is calculated according to the above formula.

[0016] As a specific implementation of the embodiment of the present disclosure, the training data is obtained by energy enhancement and normalization of the residual velocity spectrum, wherein the energy enhancement calculation formula is as follows:

[0017]

[0018] Max spec =max(R spec_abs )

[0019]

[0020] Among them, f abs Indicates the absolute value operation, R spec_abs is the absolute value spectrum, Max spec is the maximum value of the absolute value spectrum, R spec_hance is the enhanced residual velocity spectrum, R amp (it) is the intermediate variable of the calculation;

[0021] The standardized calculation formula is as follows:

[0022]

[0023] Among them, U′ represents the normalized residual velocity spectrum data, U represents the original residual velocity spectrum data, and U mean represents its mean, U std Represents its standard deviation, ε is a very small value.

[0024] As a specific implementation of the embodiment of the present disclosure, the step of obtaining the root mean square velocity value of the CDP point of the CRP gather and normalizing the initial velocity model of the point to obtain the label data specifically includes: obtaining the root mean square velocity value of the CDP point of the CRP gather and normalizing the initial velocity model of the point before the residual velocity analysis to obtain the label data; the normalization calculation formula is as follows:

[0025] V′=(VV min ) / (V max -V min )

[0026] Where V′ represents the normalized initial velocity model, v represents the seismic velocity model, and V min , V max Indicates the set normalized speed range.

[0027] As a specific implementation of the embodiment of the present disclosure, the formula of the convolutional neural network model is as follows:

[0028] Conv=(relu)(x·ω+b) ω, b∈random(kernal_size)

[0029] Among them, Conv represents the calculation result of the convolution layer, relu represents the activation function, x represents the input of the convolution layer, ω represents the weight of the convolution layer, b represents the bias of the convolution layer, and kernal_size is the size of the convolution kernel used;

[0030] MaxPool=max(m1,m2,…,m n )

[0031] Among them, MaxPool represents the pooling result, m1, m2, ..., m n In the MaxPool pooling layer processing, the value of each feature map of the data unit in the input layer is calculated;

[0032] (n1, n2, ..., n n )=F(x)

[0033] Among them, n1, n2, ..., n n is the value of each feature map of the data unit in the output layer after the upsampling layer in the network model, F is the nearest interpolation upsampling function, and x is the input feature to be upsampled.

[0034] As a specific implementation of the embodiment of the present disclosure, creating a migration velocity modeling system and determining training parameters of the migration velocity modeling system specifically include: setting network input parameters; setting input CRP gather training data; inputting initial velocity model training data; inputting label data; and network outputting data.

[0035] In a second aspect, an embodiment of the present disclosure further provides a migration velocity modeling device, comprising:

[0036] Establishment module for establishing pre-stack time migration velocity model, initial velocity model and CRP gather;

[0037] Generation module, which generates residual velocity spectrum using CRP gathers;

[0038] an acquisition module, which obtains training data by energy enhancement and normalization of the residual velocity spectrum;

[0039] The processing module obtains the root mean square velocity value of the CDP point of the CRP gather and normalizes the initial velocity model of the point to obtain the label data;

[0040] Creating a module to create a migration velocity modeling system using a convolutional neural network model and determine training parameters for the migration velocity modeling system;

[0041] The output module acquires seismic data, predicts the seismic data through the migration velocity modeling system, and outputs a root mean square velocity model.

[0042] As a specific implementation of the embodiment of the present disclosure, generating a residual velocity spectrum using a CRP gather specifically includes: using the CRP gather to calculate the residual velocity spectrum of each CDP point using a residual velocity spectrum calculation formula, and the residual velocity spectrum calculation formula is as follows:

[0043]

[0044] Among them, Rspec(it) represents the residual velocity spectrum at the sampling point it, Ramp (it) is the intermediate variable of the calculation, Cip(it) represents the imaging gather amplitude at the sampling point it, γ represents the Gamma value, h represents the offset size, ns represents the number of sampling points, dt represents the sampling point interval, (0.25h 2 (γ 2 -1)+(it*dt) 2 ) is the calculated residual time difference, ntrace represents the number of seismic traces contained in each data set; among them, the Gamma value with an interval of 0.015 between 0.1 and 1.9 is selected, and the residual velocity spectrum of each sampling point is calculated according to the above formula.

[0045] In a third aspect, an embodiment of the present disclosure further provides an electronic device, the electronic device comprising:

[0046] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the migration velocity modeling method as described above.

[0047] In a fourth aspect, an embodiment of the present disclosure further provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the above-mentioned migration velocity modeling method.

[0048] The beneficial effects of the present invention are:

[0049] The present invention uses a neural network model, utilizes the generated training data and label data as input data to input the network, extracts the multi-dimensional feature information of the seismic data, and creates an offset velocity modeling system. It can be applied to the current large amount of data in seismic exploration and efficiently and accurately completes the establishment of the pre-stack time migration velocity field.

[0050] The methods and apparatus of the present invention have other features and advantages that will be apparent from or will be described in detail in the accompanying drawings and subsequent detailed descriptions incorporated herein, which together serve to explain the specific principles of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The above and other objects, features and advantages of the present invention will become more apparent through a more detailed description of exemplary embodiments of the present invention with reference to the accompanying drawings, wherein like reference numerals generally represent like components throughout the exemplary embodiments of the present invention.

[0052] Figure 1is a flow chart of a migration velocity modeling method according to embodiment 1 of the present invention;

[0053] Figure 2a 、 2b and 2c are schematic diagrams of the prestack time migration velocity model, initial model, and CRP gathers of the present invention, respectively;

[0054] Figure 3 This is a schematic diagram of the residual velocity spectrum generated by the present invention for the CRP gather;

[0055] Figure 4 is a schematic diagram of training data obtained by enhancing and standardizing the generated residual velocity spectrum according to the present invention;

[0056] Figure 5a and Figure 5b They are respectively schematic diagrams of the initial velocity model of the present invention and the corresponding generated label data and training data;

[0057] Figure 6 Schematic diagram of the neural network model of the present invention;

[0058] Figure 7a 、 7b 7c are schematic diagrams of the initial velocity model, the predicted model and the true model, respectively; 7d is a schematic diagram of a comparison curve between the inversion result obtained by the present invention and the true velocity field;

[0059] Figure 8a 、 8b and 8c are CRP gathers obtained by prestack time migration based on the true model, initial model and prediction model of the present invention;

[0060] Figure 9 This is a structural block diagram of a migration velocity modeling device according to a second embodiment of the present invention.

[0061] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. DETAILED DESCRIPTION

[0062] The preferred embodiments of the present invention will be described in more detail below. Although the preferred embodiments of the present invention are described below, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0063] Example 1

[0064] Figure 1 A flow chart showing the steps of a migration velocity modeling method according to an embodiment of the present invention is shown in FIG. Figure 1 As shown, the migration velocity modeling method includes:

[0065] S01: Establish pre-stack time migration velocity model, initial velocity model and CRP gathers;

[0066] S02: Generate residual velocity spectrum using CRP gathers;

[0067] S03: obtaining training data by performing energy enhancement and standardization on the residual velocity spectrum;

[0068] S04: Obtain the root mean square velocity value of the CDP point of the CRP gather, and normalize the initial velocity model of the point to obtain label data;

[0069] S05: Using the convolutional neural network model, create a migration velocity modeling system and determine training parameters of the migration velocity modeling system;

[0070] S06: Acquire seismic data, predict the seismic data using the migration velocity modeling system, and output a root mean square velocity model.

[0071] It should be noted that S0N does not represent the order of precedence.

[0072] refer to Figure 2a 、 2b and 2c, establishing the prestack time migration velocity model, initial velocity model, and CRP gathers.

[0073] Specifically, a training set is established by using the pre-stack time migration velocity model that can be flattened in the migration gather in the work area, the initial model before residual velocity analysis, and the CRP gather; the training set results are generated as follows: Figure 2a 、 2b and 2c.

[0074] refer to Figure 3 , and generate the residual velocity spectrum using the CRP gather.

[0075] Specifically, the residual velocity spectrum of each CDP point is calculated according to the residual velocity spectrum calculation formula using the CRP gathers obtained in step S01; Figure 3 As shown; the residual velocity spectrum calculation formula is as follows:

[0076]

[0077] Among them, Rspec(it) represents the residual velocity spectrum at the sampling point it, R amp (it) is the intermediate variable of the calculation, Cip(it) represents the imaging gather amplitude at the sampling point it, γ represents the Gamma value, h represents the offset size, ns represents the number of sampling points, dt represents the sampling point interval, (0.25h 2 (γ 2 -1)+(it*dt) 2) is the calculated residual time difference, ntrace represents the number of seismic traces contained in each data set; among them, the Gamma value with an interval of 0.015 between 0.1 and 1.9 is selected, and the residual velocity spectrum of each sampling point is calculated according to the above formula.

[0078] refer to Figure 4 , training data is obtained by energy enhancement and normalization of the residual velocity spectrum.

[0079] The residual velocity spectrum calculated in step S02 is energy-enhanced and normalized as input training data for the deep learning migration velocity modeling system, as shown in Figure 4 During the implementation process, the residual velocity spectrum data obtained in S02 is normalized row by row; the energy enhancement calculation formula is as follows:

[0080]

[0081] Max spec =max(R spec_abs )

[0082]

[0083] Among them, f abs Indicates the absolute value operation, R spec_abs is the absolute value spectrum, Max spec is the maximum value of the absolute value spectrum, R spec_hance is the enhanced residual velocity spectrum, R amp (it) is an intermediate variable in the calculation; the standardized calculation formula is as follows:

[0084]

[0085] Among them, U′ represents the normalized residual velocity spectrum data, U represents the original residual velocity spectrum data, and U mean represents its mean, U std Represents its standard deviation, and ε is a very small value.

[0086] refer to Figure 5a and 5b , obtain the root mean square velocity value of the CDP point of the CRP gather, and normalize the initial velocity model of the point to obtain the label data.

[0087] Specifically, the label data is selected as the root mean square velocity value of the CDP point corresponding to the CRP gather in S02. The initial velocity model before the residual velocity analysis of this point is also specified to perform maximum and minimum value normalization as the input label data and training data of the deep learning migration velocity modeling system, as shown in Figure 5a and 5b As shown; the normalized calculation formula is as follows:

[0088] V′=(VV min ) / (V max -V min )

[0089] Where v′ represents the normalized initial velocity model, v represents the seismic velocity model, and v min 、v max Indicates the set normalized speed range.

[0090] refer to Figure 6 , using a convolutional neural network model, create an offset velocity modeling system and determine training parameters of the offset velocity modeling system.

[0091] Specifically, a deep learning offset speed modeling system is built using a convolutional neural network model: it includes 28 convolutional layers, 7 pooling layers, 7 upsampling layers, and 28 BN layers: the BN layer performs batch normalization on the input feature maps according to the standardization method in S03; the built deep learning offset speed modeling system is as follows Figure 6 As shown; the formula of the convolutional neural network model is as follows:

[0092] Conv=(relu)(x·ω+b) ω, b∈random(kernal_size)

[0093] Among them, Conv represents the calculation result of the convolution layer, relu represents the activation function, x represents the input of the convolution layer, ω represents the weight of the convolution layer, b represents the bias of the convolution layer, and kernal_size is the size of the convolution kernel used;

[0094] MaxPool=max(m1,m2,…,m n )

[0095] Among them, MaxPool represents the pooling result, m1, m2, ..., m n In the MaxPool pooling layer processing, the value of each feature map of the data unit in the input layer is calculated;

[0096] (n1, n2, ..., n n )=F(x)

[0097] Among them, n1, n2, ..., n n is the value of each feature map of the data unit in the output layer after the upsampling layer in the network model, F is the nearest interpolation upsampling function, and x is the input feature to be upsampled.

[0098] Determine the training parameters of the deep learning offset velocity modeling system and perform network training until the network converges to the optimal state. During the implementation process, the specific process of determining the training parameters of the offset velocity modeling system includes: setting the network input parameters: the input format is [Batchsize, height, width, channel], setting the input CRP training data shape = [Batchsize, 251, 251, 1], the input initial velocity model training data shape = [Batchsize, 1, 251, 1], the input label data shape = [Batchsize, 251], and the network output data shape = [Batchsize, 251]. Among them, Batchsize represents the size specified for small batch training. The optimizer selected during the network training process is the Adam optimizer, the training function is RMSE (root mean square error), the number of training times of the deep neural network model is set to 200, and the Batchsize used in training is 8.

[0099] refer to Figure 7a 、 7b , 7c, 7d, and Figure 8a 、 8b 8c, acquiring seismic data, predicting the seismic data through the migration velocity modeling system, and outputting a root mean square velocity model.

[0100] Specifically, the created migration velocity modeling system is used to predict the seismic data and output the predicted root mean square velocity model, such as Figure 7a 、 7b As shown in 7c, Figure 7a The initial speed model shown in the figure is trained and then denormalized using the maximum speed value of 6500 and the minimum speed value of 1500 to obtain the final prediction model. Figure 7b As shown, the real model is Figure 7c The effectiveness of the present invention can be seen from the comparison of the single-track speed curve of 7d. Figure 8a 、 8b Figures 8a and 8b are the CRP gathers obtained from prestack time migration using the true model, initial model, and predicted model of the present invention. The comparison shows that the originally curved gathers have been flattened, demonstrating the effectiveness of the present invention. In terms of prediction time, the present invention, using a trained network, achieves nearly instantaneous results, significantly exceeding the efficiency of traditional prestack time migration velocity modeling methods.

[0101] Example 2

[0102] Figure 9 A migration velocity modeling device according to an embodiment of the present invention is shown.

[0103] like Figure 9 As shown, the migration velocity modeling device includes:

[0104] Establishment module for establishing pre-stack time migration velocity model, initial velocity model and CRP gather;

[0105] Generation module, which generates residual velocity spectrum using CRP gathers;

[0106] an acquisition module, which obtains training data by energy enhancement and normalization of the residual velocity spectrum;

[0107] The processing module obtains the root mean square velocity value of the CDP point of the CRP gather and normalizes the initial velocity model of the point to obtain the label data;

[0108] Creating a module to create a migration velocity modeling system using a convolutional neural network model and determine training parameters for the migration velocity modeling system;

[0109] The output module acquires seismic data, predicts the seismic data through the migration velocity modeling system, and outputs a root mean square velocity model.

[0110] As a specific implementation of the embodiment of the present disclosure, generating a residual velocity spectrum using a CRP gather specifically includes: using the CRP gather to calculate the residual velocity spectrum of each CDP point using a residual velocity spectrum calculation formula, and the residual velocity spectrum calculation formula is as follows:

[0111]

[0112] Among them, Rspec(it) represents the residual velocity spectrum at the sampling point it, R amp (it) is the intermediate variable of the calculation, Cip(it) represents the imaging gather amplitude at the sampling point it, γ represents the Gamma value, h represents the offset size, ns represents the number of sampling points, dt represents the sampling point interval, (0.25h 2 (γ 2 -1)+(it*dt) 2 ) is the calculated residual time difference, ntrace represents the number of seismic traces contained in each data set; among them, the Gamma value with an interval of 0.015 between 0.1 and 1.9 is selected, and the residual velocity spectrum of each sampling point is calculated according to the above formula.

[0113] Example 3

[0114] The present disclosure provides an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein,

[0115] The memory stores instructions that can be executed by the at least one processor. The instructions are executed by the at least one processor to enable the at least one processor to perform the above-mentioned migration velocity modeling method.

[0116] An electronic device according to an embodiment of the present disclosure includes a memory and a processor.

[0117] The memory is used to store non-transitory computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), a hard disk, flash memory, etc.

[0118] The processor may be a central processing unit (CPU) or other form of processing unit having data processing capability and / or instruction execution capability, and may control other components in the electronic device to perform desired functions. In one embodiment of the present disclosure, the processor is used to execute the computer-readable instructions stored in the memory.

[0119] Those skilled in the art should understand that in order to solve the technical problem of how to obtain a good user experience, this embodiment may also include well-known structures such as a communication bus and an interface, and these well-known structures should also be included in the scope of protection of this disclosure.

[0120] For detailed description of this embodiment, please refer to the corresponding description in the aforementioned embodiments, which will not be repeated here.

[0121] Example 4

[0122] An embodiment of the present disclosure provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the above-described migration velocity modeling apparatus.

[0123] According to the non-transitory computer-readable storage medium of the embodiment of the present disclosure, when the non-transitory computer-readable instructions are executed by the processor, all or part of the steps of the aforementioned methods of the various embodiments of the present disclosure are executed.

[0124] The above-mentioned computer-readable storage media include, but are not limited to, optical storage media (e.g., CD-ROMs and DVDs), magneto-optical storage media (e.g., MOs), magnetic storage media (e.g., magnetic tapes or mobile hard disks), media with built-in rewritable non-volatile memory (e.g., memory cards), and media with built-in ROM (e.g., ROM cartridges).

[0125] Those skilled in the art should understand that the above description of the embodiments of the present invention is only for the purpose of illustrative purposes only to illustrate the beneficial effects of the embodiments of the present invention, and is not intended to limit the embodiments of the present invention to any given examples.

[0126] While various embodiments of the present invention have been described above, the above description is intended to be illustrative, not exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. A migration velocity modeling method, characterized in that: include: Establish pre-stack time migration velocity model, initial velocity model and CRP gathers; Generate residual velocity spectrum using CRP gathers; Training data is obtained by energy enhancement and normalization of the residual velocity spectrum; Get the root mean square velocity value of the CDP point of the CRP gather and normalize the initial velocity model of the point to obtain the label data; Using a convolutional neural network model, creating a migration velocity modeling system and determining training parameters for the migration velocity modeling system; Seismic data is acquired, the seismic data is predicted by the migration velocity modeling system, and a root mean square velocity model is output.

2. The migration velocity modeling method according to claim 1, characterized in that: The method of generating the residual velocity spectrum using the CRP gather specifically includes: using the CRP gather to calculate the residual velocity spectrum of each CDP point using a residual velocity spectrum calculation formula, and the residual velocity spectrum calculation formula is as follows: Among them, Rspec(it) represents the residual velocity spectrum at the sampling point it, R amp (it) is the intermediate variable of the calculation, Cip(it) represents the imaging gather amplitude at the sampling point it, γ represents the Gamma value, h represents the offset size, ns represents the number of sampling points, dt represents the sampling point interval, (0.25h 2 (γ 2 -1)+(it*dt) 2 ) is the calculated residual time difference, ntrace represents the number of seismic traces contained in each data set; among them, the Gamma value with an interval of 0.015 between 0.1 and 1.9 is selected, and the residual velocity spectrum of each sampling point is calculated according to the above formula.

3. The migration velocity modeling method according to claim 2, characterized in that: The training data is obtained by energy enhancement and normalization of the residual velocity spectrum, wherein the energy enhancement calculation formula is as follows: Max spec =max(R spec_abs ) Among them, fabs represents the absolute value operation, R spec_abs is the absolute value spectrum, Max spec is the maximum value of the absolute value spectrum, R spec_hance is the enhanced residual velocity spectrum, R amp (it) is the intermediate variable of the calculation; The standardized calculation formula is as follows: Among them, U′ represents the normalized residual velocity spectrum data, U represents the original residual velocity spectrum data, and U mean represents its mean, U std Represents its standard deviation, ε is a very small value.

4. The migration velocity modeling method according to claim 1, characterized in that: The step of obtaining the root mean square velocity value of the CDP point of the CRP gather and normalizing the initial velocity model of the point to obtain label data specifically includes: The root mean square velocity value of the CDP point of the CRP gather is obtained, and the initial velocity model before the residual velocity analysis of the point is normalized to the maximum and minimum values ​​to obtain the label data; the normalization calculation formula is as follows: V′=(VV min ) / (V max -V min ) Where V′ represents the normalized initial velocity model, V represents the seismic velocity model, and V min 、V max Indicates the set normalized speed range.

5. The migration velocity modeling method according to claim 1, characterized in that: The formula of the convolutional neural network model is as follows: Conv=(relu)(x·ω+b)ω, b∈random(kernal_size) Among them, Conv represents the calculation result of the convolution layer, relu represents the activation function, x represents the input of the convolution layer, ω represents the weight of the convolution layer, b represents the bias of the convolution layer, and kernal_size is the size of the convolution kernel used; MaxPool=max(m1,m2,…,m n ) Among them, MaxPool represents the pooling result, m1, m2, ..., m n In the MaxPool pooling layer processing, the value of each feature map of the data unit in the input layer is calculated; (n1,n2,…,n n )=F(x) Among them, n1, n2, ..., n n is the value of each feature map of the data unit in the output layer after the upsampling layer in the network model, F is the nearest interpolation upsampling function, and x is the input feature to be upsampled.

6. The migration velocity modeling method according to claim 1, characterized in that: The step of creating a migration velocity modeling system and determining training parameters of the migration velocity modeling system specifically includes: Set network input parameters; set input CRP gather training data; input initial velocity model training data; input label data; network output data.

7. A migration velocity modeling device, characterized in that: include: Establishment module for establishing pre-stack time migration velocity model, initial velocity model and CRP gather; Generation module, which generates residual velocity spectrum using CRP gathers; an acquisition module, which obtains training data by energy enhancement and normalization of the residual velocity spectrum; The processing module obtains the root mean square velocity value of the CDP point of the CRP gather and normalizes the initial velocity model of the point to obtain the label data; Creating a module to create a migration velocity modeling system using a convolutional neural network model and determine training parameters for the migration velocity modeling system; The output module acquires seismic data, predicts the seismic data through the migration velocity modeling system, and outputs a root mean square velocity model.

8. The migration velocity modeling device according to claim 7, characterized in that: The method of generating the residual velocity spectrum using the CRP gather specifically includes: using the CRP gather to calculate the residual velocity spectrum of each CDP point using a residual velocity spectrum calculation formula, and the residual velocity spectrum calculation formula is as follows: Among them, Rspec(it) represents the residual velocity spectrum at the sampling point it, R amp (it) is the intermediate variable of the calculation, Cip(it) represents the imaging gather amplitude at the sampling point it, γ represents the Gamma value, h represents the offset size, ns represents the number of sampling points, dt represents the sampling point interval, (0.25h 2 (γ 2 -1)+(it*dt) 2 ) is the calculated residual time difference, ntrace represents the number of seismic traces contained in each data set; among them, the Gamma value with an interval of 0.015 between 0.1 and 1.9 is selected, and the residual velocity spectrum of each sampling point is calculated according to the above formula.

9. An electronic device, characterized in that: The electronic device comprises: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the migration velocity modeling method according to any one of claims 1 to 6.

10. A non-transitory computer-readable storage medium, characterized in that The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the migration velocity modeling method according to any one of claims 1 to 6.

Images