Method, device and storage medium for obtaining a migration velocity model

Abstract

This application provides a method, device, and storage medium for acquiring a migration velocity model. The method includes: denoising and extracting trace sets from seismic data to acquire a first common offset trace set, wherein the first common offset trace set includes multiple common offset data; inputting the multiple common offset data from the first common offset trace set into an initial migration velocity model to acquire the migration velocity output by the migration velocity model; performing repositioning processing on the first common offset trace set according to the migration velocity to acquire a second common offset trace set; iteratively updating the migration velocity model according to the second common offset trace set and a target loss function; and acquiring the corresponding target migration velocity model after determining that the number of iterations has reached a threshold, thereby achieving the technical effect of improving the efficiency of acquiring the migration velocity model.

Description

Technical Field

[0001] This application relates to the field of drilling engineering, and in particular to a method, device and storage medium for obtaining an offset velocity model. Background Technology

[0002] With the continuous advancement of technology, the demand for petroleum geophysical exploration is increasing. In petroleum geophysical exploration, seismic imaging, by correcting raw underground seismic data, enables accurate positioning of oil and gas reservoir images, delineates corresponding reflection interfaces, and assists in identifying the storage locations of oil, gas, and water. However, imaging is highly sensitive to velocity models; inaccurate velocity models can lead to significant deviations in the imaging results of oil and gas reservoir morphology or drilling trajectories. Accurate velocity models are crucial for obtaining high-fidelity, high signal-to-noise ratio, and high-resolution imaging profiles. Therefore, the acquisition of migration velocity models has become a promising area of ​​application.

[0003] In existing technologies, the acquisition of the migration velocity model is mainly based on the velocity of a single point location picked up by a single gather. The migration velocity model is then reconstructed by forward matching the full wave field information through full waveform inversion technology.

[0004] Because existing full-waveform inversion techniques not only require a large amount of computing resources, but also rely heavily on complex forward modeling calculations, resulting in high matching difficulty, large computational load, and long processing time, they are difficult to meet the needs of rapid acquisition of migration velocity models in practical applications, and have the technical problem of low efficiency in acquiring migration velocity models. Summary of the Invention

[0005] This application provides a method, device, and storage medium for obtaining an offset velocity model, thereby achieving the technical effect of improving the efficiency of obtaining the offset velocity model.

[0006] Firstly, this application provides a method for obtaining a migration velocity model, comprising:

[0007] The seismic data is denoised and extracted to obtain the first common offset trace set, which includes multiple common offset data, each of which corresponds to an offset.

[0008] Multiple common offset data from the first common offset data set are input into the initial offset velocity model to obtain the offset velocity output by the offset velocity model. The offset velocity model includes an offset velocity solver and a fusion unit. The offset velocity solver is used to obtain the velocity feature corresponding to each common offset data, and the fusion unit is used to fuse multiple velocity features to obtain the offset velocity.

[0009] Based on the offset velocity, the first common offset gather is repositioned to obtain the second common offset gather. The repositioning process includes pre-stack time offset and reaction correction. The second common offset gather is the common offset gather after the first common offset gather has been repositioned.

[0010] The migration velocity model is iteratively updated based on the second common offset gather and the target loss function. After the number of iterations reaches a threshold, the corresponding target migration velocity model is obtained, where the target loss function is a loss function based on migration velocity.

[0011] Optionally, the offset velocity model also includes:

[0012] The reflection travel unit is used to acquire multiple travel time data corresponding to each common offset data in the first common offset trace set. The travel time data is used to indicate the time required for seismic waves to reflect from the subsurface interface back to the surface at each offset. Based on each travel time data, the multiple data reflection travel times corresponding to the first common offset trace set are determined, and based on each data reflection travel time, the corresponding hyperbolic activation function is determined.

[0013] The position transformation unit is used to perform longitudinal position transformation on each common offset data according to the travel time of each data reflection, and obtain the transformed target common offset data;

[0014] The offset velocity solver is used to obtain velocity features based on the longitudinal position transformation method corresponding to the hyperbolic activation function and the target co-offset distance data. The offset velocity solver includes a U-net convolutional neural network.

[0015] Optionally, the travel time of the data corresponding to the first common offset gather is obtained by the following method:

[0016]

[0017] Where t(t0,x,h) represents the data reflection travel time, h represents the offset distance, x represents the horizontal position in space, t0 represents the two-way travel time corresponding to zero offset distance, and v represents the velocity of the medium above the interface.

[0018] Optionally, the transformed target common offset data is obtained in the following way:

[0019]

[0020] Among them, I output (y,x) represents the target common offset data, I input represents the common offset data, y and x represent the longitudinal time and lateral spatial position of the imaging point, respectively, T represents the travel time matrix calculated based on the physical laws of travel time, dt represents the time sampling interval, and round represents the rounding function.

[0021] Optionally, the target loss function is determined based on the unsupervised loss function and the auxiliary data loss function;

[0022] Unsupervised loss function is a loss function based on offset velocity;

[0023] The auxiliary data loss function is a loss function based on the target co-offset data;

[0024] Based on the unsupervised loss function and the auxiliary data loss function, determine the corresponding target loss function.

[0025] Optionally, the unsupervised loss function is obtained in the following way:

[0026]

[0027]

[0028] Among them, L us This represents the unsupervised loss function, and v represents the offset velocity calculated by the network. Represents the Laplace operator, MSE represents the root mean square error, x i Represents a local window centered at x. and σ represents the mean of I and J, respectively. I and σ J Cov represents the standard deviations of I and J. IJ c1 and c2 represent constants, M represents the target co-offset data, which is composed of noff imaging profiles stitched together. I represents the 1st to noff-1th co-offset gathers, and J represents the 2nd to noffth co-offset gathers.

[0029] Optionally, the auxiliary data loss function is obtained in the following way:

[0030] L a =-NCC(M,A)-SSIM(M,A)

[0031] Among them, L a represents the auxiliary data loss function, where A represents the auxiliary dataset, which is composed of noff parts of superimposed profiles stitched together horizontally.

[0032] Optionally, the target loss function is obtained in the following way:

[0033] L=λL us +(1-λ)L a

[0034] Where L represents the target loss function, Lus L represents the unsupervised loss function. a λ represents the auxiliary data loss function, and λ represents the hyperparameter set to balance the weights of the two loss functions.

[0035] Secondly, this application provides a device for obtaining a displacement velocity model, comprising:

[0036] The first acquisition module is used to perform noise reduction and gather extraction processing on the seismic data to acquire the first common offset gather, wherein the first common offset gather includes multiple common offset data, and each common offset data corresponds to an offset.

[0037] The second acquisition module is used to input multiple common offset data from the first common offset data set into the initial offset velocity model, and acquire the offset velocity output by the offset velocity model. The offset velocity model includes an offset velocity solver and a fusion unit. The offset velocity solver is used to acquire the velocity feature corresponding to each common offset data, and the fusion unit is used to fuse multiple velocity features to obtain the offset velocity.

[0038] The first processing module is used to perform repositioning processing on the first common offset gather according to the offset velocity to obtain the second common offset gather. The repositioning processing includes pre-stack time offset and reaction correction. The second common offset gather is the common offset gather after the repositioning processing of the first common offset gather.

[0039] The second processing module is used to iteratively update the migration velocity model based on the second common offset gather and the target loss function. After determining that the number of iterations has reached a threshold, the corresponding target migration velocity model is obtained, wherein the target loss function is a loss function based on migration velocity.

[0040] Optionally, the second acquisition module is also used for:

[0041] The reflection travel unit is used to acquire multiple travel time data corresponding to each common offset data in the first common offset trace set. The travel time data is used to indicate the time required for seismic waves to reflect from the subsurface interface back to the surface at each offset. Based on each travel time data, the multiple data reflection travel times corresponding to the first common offset trace set are determined, and based on each data reflection travel time, the corresponding hyperbolic activation function is determined.

[0042] The position transformation unit is used to perform longitudinal position transformation on each common offset data according to the travel time of each data reflection, and obtain the transformed target common offset data;

[0043] The offset velocity solver is used to obtain velocity features based on the longitudinal position transformation method corresponding to the hyperbolic activation function and the target co-offset distance data. The offset velocity solver includes a U-net convolutional neural network.

[0044] Optionally, the second acquisition module is also used for:

[0045] The travel time of the data corresponding to the first common offset gather is obtained through the following method:

[0046]

[0047] Where t(t0,x,h) represents the data reflection travel time, h represents the offset distance, x represents the horizontal position in space, t0 represents the two-way travel time corresponding to zero offset distance, and v represents the velocity of the medium above the interface.

[0048] Optionally, the second acquisition module is also used for:

[0049] The transformed target common offset data is obtained in the following way:

[0050]

[0051] Among them, I output (y,x) represents the target common offset data, I input represents the common offset data, y and x represent the longitudinal time and lateral spatial position of the imaging point, respectively, T represents the travel time matrix calculated based on the physical laws of travel time, dt represents the time sampling interval, and round represents the rounding function.

[0052] Optionally, the second processing module is also used for:

[0053] Unsupervised loss function is a loss function based on offset velocity;

[0054] The auxiliary data loss function is a loss function based on the target co-offset data;

[0055] Based on the unsupervised loss function and the auxiliary data loss function, determine the corresponding target loss function.

[0056] Optionally, the second processing module is also used for:

[0057] The unsupervised loss function is obtained in the following way:

[0058]

[0059]

[0060]

[0061] Among them, L us This represents the unsupervised loss function, and v represents the offset velocity calculated by the network. Represents the Laplace operator, MSE represents the root mean square error, xi Represents a local window centered at x. and σ represents the mean of I and J, respectively. I and σ J Cov represents the standard deviations of I and J. IJ c1 and c2 represent constants, M represents the target co-offset data, which is composed of noff imaging profiles stitched together. I represents the 1st to noff-1th co-offset gathers, and J represents the 2nd to noffth co-offset gathers.

[0062] Optionally, the second processing module is also used for:

[0063] The auxiliary data loss function is obtained in the following way:

[0064] L a =-NCC(M,A)-SSIM(M,A)

[0065] Among them, L a represents the auxiliary data loss function, where A represents the auxiliary dataset, which is composed of noff parts of superimposed profiles stitched together horizontally.

[0066] Optionally, the second processing module is also used for:

[0067] The target loss function is obtained in the following way:

[0068] L=λL us +(1-λ)L a

[0069] Where L represents the target loss function, L us L represents the unsupervised loss function. a λ represents the auxiliary data loss function, and λ represents the hyperparameter set to balance the weights of the two loss functions.

[0070] Thirdly, this application provides a device for acquiring an offset velocity model, comprising:

[0071] Processor and memory;

[0072] The memory stores the instructions that the computer executes;

[0073] The processor executes computer execution instructions stored in memory, causing the processor to perform the various possible implementations described in the first aspect above.

[0074] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, are used to implement various possible implementations as described in the first aspect.

[0075] Fifthly, this application provides a computer program product, which, when executed by a processor, is used to implement various possible implementations as described in the first aspect.

[0076] This application provides a method, device, and storage medium for acquiring a migration velocity model. The method involves denoising and extracting trace sets from seismic data to obtain a first common offset trace set, which includes multiple common offset data points. These common offset data points are then input into an initial migration velocity model to obtain the migration velocity output by the model. Based on the migration velocity, the first common offset trace set is repositioned to obtain a second common offset trace set. The migration velocity model is iteratively updated based on the second common offset trace set and a target loss function. After the number of iterations reaches a threshold, the corresponding migration velocity model is acquired. The target offset velocity model is developed, which utilizes the offset velocity solver and fusion unit in the offset velocity model to obtain the offset velocity corresponding to the first common offset distance gather, saving a significant amount of computational resources and time costs. At the same time, the offset velocity model is iteratively updated through the target loss function and the second common offset distance gather after relocation processing to obtain the optimal target offset velocity model. This reduces the complexity of obtaining the target offset velocity model and improves the convenience, accuracy, and reliability of obtaining the target offset velocity model. It solves the technical problem of low efficiency in obtaining the offset velocity model and achieves the technical effect of improving the efficiency of obtaining the offset velocity model. Attached Figure Description

[0077] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0078] Figure 1 The flowchart of the method for obtaining the offset velocity model provided in the embodiments of this application Figure 1 ;

[0079] Figure 2 A schematic diagram of the first common offset gather provided for an embodiment of this application;

[0080] Figure 3 An interactive diagram illustrating the method for obtaining the offset velocity model provided in this application embodiment;

[0081] Figure 4 A schematic diagram illustrating the timekeeping pattern provided in the embodiments of this application;

[0082] Figure 5 A schematic diagram illustrating the time-position transformation provided in the embodiments of this application;

[0083] Figure 6A schematic diagram illustrating the time-position transformation process provided in the embodiments of this application;

[0084] Figure 7 The flowchart of the method for obtaining the offset velocity model provided in the embodiments of this application Figure 2 ;

[0085] Figure 8 A schematic diagram of the architecture of the offset velocity model provided in the embodiments of this application;

[0086] Figure 9 A schematic diagram of the structure of the device for obtaining the offset velocity model provided in the embodiments of this application;

[0087] Figure 10 This is a hardware structure diagram of the device for acquiring the offset velocity model provided in an embodiment of this application.

[0088] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0089] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0090] Because existing full-waveform inversion techniques not only require a large amount of computing resources, but also rely heavily on complex forward modeling calculations, resulting in high matching difficulty, large computational load, and long processing time, they are difficult to meet the needs of rapid acquisition of migration velocity models in practical applications, and have the technical problem of low efficiency in acquiring migration velocity models.

[0091] This application provides a method, device, and storage medium for obtaining an offset velocity model. By utilizing the offset velocity solver and fusion unit in the offset velocity model, the offset velocity corresponding to the first common offset distance gather is obtained, saving a significant amount of computational resources and time costs. Simultaneously, the offset velocity model is iteratively updated using the target loss function and the repositioned second common offset distance gather to obtain the optimal target offset velocity model. This reduces the complexity of obtaining the target offset velocity model and improves the convenience, accuracy, and reliability of obtaining the target offset velocity model. It solves the technical problem of low efficiency in obtaining the offset velocity model and achieves the technical effect of improving the efficiency of offset velocity model acquisition.

[0092] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0093] Figure 1 The flowchart of the method for obtaining the offset velocity model provided in the embodiments of this application Figure 1 .like Figure 1 As shown in the embodiments of this application, the method for obtaining the offset velocity model includes:

[0094] S101. Denoise and extract traces from the seismic data to obtain the first common offset trace, wherein the first common offset trace includes multiple common offset data.

[0095] In this embodiment, each common offset data corresponds to an offset, which refers to the distance from the seismic source to the detector.

[0096] Denoising and gather extraction are performed on the seismic data of the target geological body to obtain the first common offset gather after processing.

[0097] Figure 2 A schematic diagram of the first common offset gather provided in the embodiments of this application, as shown below. Figure 2 As shown, the first common offset gather includes multiple common offset data, each corresponding to an offset. CDP refers to the horizontal position of the underground reflection point, and Time refers to the time required for the ground surface to reach the center of the detector.

[0098] S102. Input multiple common offset data from the first common offset channel set into the initial offset velocity model, and obtain the offset velocity output by the offset velocity model.

[0099] In this embodiment, the offset velocity model includes an offset velocity solver and a fusion unit. The offset velocity solver is used to obtain the velocity features corresponding to each common offset distance data, and the fusion unit is used to fuse multiple velocity features to obtain the offset velocity.

[0100] Multiple common offset data from the first common offset data set are input into the initial offset velocity model. The offset velocity solver in the offset velocity model obtains multiple corresponding velocity features based on each common offset data, and sends each velocity feature to the fusion unit to fuse the multiple velocity features and obtain the fused offset velocity.

[0101] S103. Based on the offset speed, perform repositioning processing on the first common offset gather to obtain the second common offset gather.

[0102] In this embodiment, the repositioning process includes pre-stack time offset and reaction correction, and the second common offset gather is the common offset gather after the repositioning process of the first common offset gather.

[0103] Based on the migration velocity of the migration velocity model, pre-stack time migration and reaction correction are performed on the first common offset gather to obtain the second common offset gather after repositioning.

[0104] S104. Iteratively update the offset velocity model based on the second common offset gather and the target loss function. After determining that the number of iterations has reached a threshold, obtain the corresponding target offset velocity model.

[0105] In this embodiment, the target loss function is a loss function based on the offset velocity.

[0106] Based on the second common offset gather and the target loss function, the offset velocity model is iteratively updated multiple times. After the number of iterations of the offset velocity model reaches a preset threshold, the iterative update of the offset velocity model is stopped, and the offset velocity model after the last iteration is determined as the target offset velocity model.

[0107] This application provides a method for obtaining a migration velocity model. The method involves denoising and extracting trace sets from seismic data to obtain a first common offset trace set, which includes multiple common offset data points. These common offset data points are then input into an initial migration velocity model to obtain the migration velocity output by the model. Based on the migration velocity, the first common offset trace set is repositioned to obtain a second common offset trace set. The migration velocity model is iteratively updated based on the second common offset trace set and a target loss function. After the number of iterations reaches a threshold, the corresponding target migration is obtained. The velocity model, by utilizing the offset velocity solver and fusion unit in the offset velocity model, obtains the offset velocity corresponding to the first common offset distance gather, saving a significant amount of computational resources and time costs. Simultaneously, the offset velocity model is iteratively updated through the target loss function and the repositioned second common offset distance gather to obtain the optimal target offset velocity model. This reduces the complexity of obtaining the target offset velocity model and improves the convenience, accuracy, and reliability of obtaining the target offset velocity model, thus solving the technical problem of low efficiency in obtaining the offset velocity model and achieving the technical effect of improving the efficiency of offset velocity model acquisition.

[0108] Figure 3 An interactive diagram illustrating the method for obtaining the offset velocity model provided in this application embodiment. For example... Figure 3As shown, the migration velocity model includes a migration velocity solver, a fusion unit, a reflection travel unit, and a position transformation unit. This embodiment, based on the above embodiment, provides supplementary explanations of the interaction process of each unit in the initial migration velocity model, including:

[0109] S301. The reflection travel unit obtains multiple travel time data corresponding to each common offset data in the first common offset trace set, determines multiple data reflection travel times corresponding to the first common offset trace set based on each travel time data, and determines the corresponding hyperbolic activation function based on each data reflection travel time.

[0110] In this embodiment, travel time data is used to indicate the time required for seismic waves to reflect back to the surface from the subsurface interface at each offset.

[0111] Figure 4 A schematic diagram of the timekeeping pattern provided in the embodiments of this application, such as Figure 4 As shown, the hyperbola represents the trajectory of the same geological body at different offsets. In the migration velocity model, the reflection travel unit, based on each common offset data point in the first common offset gather, acquires multiple corresponding travel time data to determine the corresponding multiple data reflection travel times. Based on the assumption of a horizontally layered medium, the hyperbolic activation function corresponding to the data reflection travel time is determined. This function can be expressed as:

[0112]

[0113] Where t(t0,x,h) represents the data reflection travel time, h represents the offset distance, x represents the horizontal position in space, t0 represents the two-way travel time corresponding to zero offset distance, and v represents the velocity of the medium above the interface. The reflection travel time and the offset distance maintain a hyperbolic relationship, meaning that imaging structures at different offset distances should be matched according to the hyperbolic travel time law.

[0114] S302, The reflection travel unit sends the reflection travel time of multiple data corresponding to the first common offset gather to the position transformation unit;

[0115] S303, the position transformation unit performs longitudinal position transformation on each common offset data according to the travel time of each data reflection, and obtains the transformed target common offset data;

[0116] In the offset velocity model, the position transformation unit performs a longitudinal position transformation on each common offset data point based on the reflection travel time of each data point, obtaining the transformed target common offset data, which can be expressed as:

[0117]

[0118] Among them, I output (y,x) represents the target common offset data, Iinput represents the common offset data, y and x represent the longitudinal time and lateral spatial position of the imaging point, respectively, T represents the travel time matrix calculated based on the physical laws of travel time, dt represents the time sampling interval, and round represents the rounding function.

[0119] Figure 5 A schematic diagram of time position changes provided in the embodiments of this application, such as Figure 5 As shown, the position transformation unit in the offset velocity model is implemented through time grid sampling, based on matrix I. output The location of the red imaging point is found in matrix T, along with its corresponding travel times t1 and t2. From this, matrix I can be calculated. input The position of the blue imaging point is determined, and then it is moved to matrix I. output The position of the red imaging point in the middle can be used to complete the imaging correction of the common offset data; when T = t0, no correction is performed on the common offset data.

[0120] Figure 6 A schematic diagram of the time position transformation process provided in the embodiments of this application, as shown below. Figure 6 As shown, (a) is the data reflection travel time corresponding to the common offset gather, (b) is the common offset data before longitudinal position transformation, and (c) is the target common offset data after longitudinal position transformation. The imaging structure of the common offset gather with different offsets has longitudinal displacement, which is mainly reflected in the travel time difference. When the network calculates the correct data reflection travel time, the common offset data can be corrected based on the longitudinal position transformation to obtain the target common offset data with consistent imaging structure.

[0121] S304, The position transformation unit sends the target common offset data to the offset velocity solver;

[0122] S305, the reflection travel unit sends the hyperbolic activation function to the offset velocity solver;

[0123] S306. The offset velocity solver obtains the velocity characteristics corresponding to each co-offset data point based on the hyperbolic activation function and the longitudinal position transformation method corresponding to the target co-offset data.

[0124] In this embodiment, the offset velocity solver includes a U-net convolutional neural network.

[0125] The hyperbolic activation function and the longitudinal position transformation method corresponding to the target co-offset data are introduced into the U-net convolutional neural network corresponding to the offset velocity solver in the offset velocity model, so that the offset velocity solver can obtain the velocity feature corresponding to each co-offset data according to the hyperbolic activation function and the longitudinal position transformation method corresponding to the target co-offset data.

[0126] S307, The offset velocity solver sends the velocity characteristics corresponding to each common offset data point to the fusion unit;

[0127] S308, the fusion unit fuses multiple velocity features to obtain the offset velocity.

[0128] The velocity features corresponding to each common offset data obtained in the offset velocity solver are input into the fusion unit in the offset velocity model so that the fusion unit fuses multiple velocity features to obtain the fused offset velocity.

[0129] This application provides a method for obtaining a migration velocity model. By establishing a migration velocity model including a migration velocity solver, fusion unit, reflection travel unit, and position transformation unit, the corresponding migration velocity is determined based on the first common offset gather, thereby saving a significant amount of computational resources and time. Furthermore, under the constraint of the hyperbolic activation function in the reflection travel unit, the first common offset data is matched to the program structure based on hyperbolic laws, ensuring the integrity and continuity of the in-phase axis structure and guaranteeing the accuracy and reliability of the migration velocity acquisition. In addition, by applying a longitudinal position transformation method, it is not necessary to consider seismic wavelets and forward modeling methods, further saving significant computational resources and time in the migration velocity acquisition process, improving the speed and convenience of migration velocity acquisition, solving the technical problem of low efficiency in acquiring migration velocity models, and achieving the technical effect of improving the efficiency of migration velocity model acquisition.

[0130] Figure 7 The flowchart of the method for obtaining the offset velocity model provided in the embodiments of this application Figure 2 .like Figure 7 As shown, this embodiment, based on the above embodiment, provides supplementary explanations regarding the process of obtaining the target loss function, including:

[0131] S701. Determine the corresponding unsupervised loss function based on the offset velocity;

[0132] Based on the offset velocity obtained from the offset velocity model, the corresponding unsupervised loss function is determined. This loss function includes a common offset image similarity loss function and an offset velocity smoothing constraint function. The image similarity loss function adopts Local Normalized Cross-Correlation (NCC) and Structural Similarity Measure (SSIM). NCC and SSIM can be expressed as follows:

[0133]

[0134]

[0135] Where, x i Represents a local window centered at x. and σ represents the mean of I and J, respectively. I and σ J Cov represents the standard deviations of I and J. IJ c1 and c2 represent constants, M represents the target co-offset data, which is composed of noff imaging profiles stitched together. I represents the 1st to noff-1th co-offset gathers, and J represents the 2nd to noffth co-offset gathers.

[0136] The similarity between I and J represents the degree of matching of the imaging structures between adjacent offsets. A higher degree of matching results in NCC and SSIM values ​​closer to 1. Considering that the actual offset velocity should be smooth and continuous, a smoothing constraint is applied to the offset velocity estimated by the network. The unsupervised loss function can be expressed as:

[0137]

[0138] Among them, L us This represents the unsupervised loss function, and v represents the offset velocity calculated by the network. represents the Laplace operator, and MSE represents the root mean square error.

[0139] S702. Based on the target common offset data, determine the corresponding auxiliary data loss function;

[0140] Based on the target common offset data after longitudinal time position transformation, the corresponding auxiliary data loss function is determined, which can be expressed as:

[0141] L a =-NCC(M,A)-SSIM(M,A)

[0142] Among them, L a represents the auxiliary data loss function, where A represents the auxiliary dataset, which is composed of noff parts of superimposed profiles stitched together horizontally.

[0143] S703. Determine the corresponding target loss function based on the unsupervised loss function and the auxiliary data loss function.

[0144] Based on the unsupervised loss function and the auxiliary data loss function, the final target loss function is determined, which can be expressed as:

[0145] L=λL us +(1-λ)L a

[0146] Where L represents the target loss function, L us L represents the unsupervised loss function. aλ represents the auxiliary data loss function, and λ represents the hyperparameter set to balance the weights of the two loss functions.

[0147] Figure 8 This is a schematic diagram of the architecture of the offset velocity model provided in the embodiments of this application, as shown below. Figure 8 As shown, in conjunction with the above embodiment, the horizontal concatenation of the co-offset data corresponding to different offsets in the first co-offset gather is input into the offset velocity solver in the offset velocity model. Its size is (nt, nx × noff), where nt and nx represent the time length and spatial length, respectively, and noff is the number of offsets. The U-net network layer is used to preserve the co-offset gather structure as an example feature. U-net extracts the corresponding velocity features from different co-offset data, and then the Reshape layer reshapes the velocity features of size (nt, nx × noff) into (nt, nx, noff). The convolutional layer (i.e., the fusion unit) is used to fuse the noff velocity features to obtain the offset velocity of size (nt, nx). Then, the travel time corresponding to different offsets is calculated using the hyperbolic activation function in the reflection travel unit. Its magnitude is the same as that of the input co-offset gather, which is (nt, nx × noff). Based on the data reflection travel time, the co-offset data corresponding to the input co-offset gather is calibrated in the position transformation unit using time-position transformation to obtain the calibrated target co-offset data. According to the geological model consistency criterion, the imaging structure between adjacent offsets of the co-offset gather corresponding to the calibrated target co-offset data should be consistent. Therefore, the structural similarity between adjacent co-offset gathers after calibration is used as the loss function (i.e., the target loss function mentioned above) to update the parameters of the U-net network layers and convolutional layers, thereby achieving unsupervised training and predicting the migration velocity.

[0148] This application provides a method for obtaining an offset velocity model. By acquiring corresponding unsupervised loss functions and auxiliary data loss functions based on offset velocity and target co-offset data, and then determining the final target loss function based on these two functions, the accuracy and reliability of the target offset velocity model acquisition are improved during subsequent iterative training and target offset velocity model acquisition. This reduces the complexity of the overall computation process, saves time and resources, and solves the technical problem of low efficiency in acquiring the target offset velocity model, thus achieving the technical effect of improving the efficiency of target offset velocity model acquisition.

[0149] Figure 9 This is a schematic diagram of the structure of the device for acquiring the offset velocity model provided in an embodiment of this application. The device in this embodiment can be in the form of software and / or hardware. Figure 9As shown in the embodiment of this application, the offset velocity model acquisition device 900 includes: a first acquisition module 901, a second acquisition module 902, a first processing module 903, and a second processing module 904.

[0150] The first acquisition module 901 is used to perform denoising and gather extraction processing on the seismic data to acquire the first common offset gather, wherein the first common offset gather includes multiple common offset data, and each common offset data corresponds to an offset.

[0151] The second acquisition module 902 is used to input multiple common offset data from the first common offset data set into the initial offset velocity model, and acquire the offset velocity output by the offset velocity model. The offset velocity model includes an offset velocity solver and a fusion unit. The offset velocity solver is used to acquire the velocity feature corresponding to each common offset data, and the fusion unit is used to fuse multiple velocity features to obtain the offset velocity.

[0152] The first processing module 903 is used to perform repositioning processing on the first common offset gather according to the offset speed to obtain the second common offset gather. The repositioning processing includes pre-stack time offset and reaction correction. The second common offset gather is the common offset gather after the repositioning processing of the first common offset gather.

[0153] The second processing module 904 is used to iteratively update the migration velocity model based on the second common offset gather and the target loss function. After determining that the number of iterations has reached a threshold, the corresponding target migration velocity model is obtained, wherein the target loss function is a loss function based on migration velocity.

[0154] In one possible implementation, the second acquisition module 902 is further used for:

[0155] The reflection travel unit is used to acquire multiple travel time data corresponding to each common offset data in the first common offset trace set. The travel time data is used to indicate the time required for seismic waves to reflect from the subsurface interface back to the surface at each offset. Based on each travel time data, the multiple data reflection travel times corresponding to the first common offset trace set are determined, and based on each data reflection travel time, the corresponding hyperbolic activation function is determined.

[0156] The position transformation unit is used to perform longitudinal position transformation on each common offset data according to the travel time of each data reflection, and obtain the transformed target common offset data;

[0157] The offset velocity solver is used to obtain velocity features based on the longitudinal position transformation method corresponding to the hyperbolic activation function and the target co-offset distance data. The offset velocity solver includes a U-net convolutional neural network.

[0158] In one possible implementation, the second acquisition module 902 is further used for:

[0159] The travel time of the data corresponding to the first common offset gather is obtained through the following method:

[0160]

[0161] Where t(t0,x,h) represents the data reflection travel time, h represents the offset distance, x represents the horizontal position in space, t0 represents the two-way travel time corresponding to zero offset distance, and v represents the velocity of the medium above the interface.

[0162] In one possible implementation, the second acquisition module 902 is further used for:

[0163] The transformed target common offset data is obtained in the following way:

[0164]

[0165] Among them, I output (y,x) represents the target common offset data, I input represents the common offset data, y and x represent the longitudinal time and lateral spatial position of the imaging point, respectively, T represents the travel time matrix calculated based on the physical laws of travel time, dt represents the time sampling interval, and round represents the rounding function.

[0166] In one possible implementation, the second processing module 904 is further used for:

[0167] Unsupervised loss function is a loss function based on offset velocity;

[0168] The auxiliary data loss function is a loss function based on the target co-offset data;

[0169] Based on the unsupervised loss function and the auxiliary data loss function, determine the corresponding target loss function.

[0170] In one possible implementation, the second processing module 904 is further used for:

[0171] The unsupervised loss function is obtained in the following way:

[0172]

[0173]

[0174]

[0175] Among them, L us This represents the unsupervised loss function, and v represents the offset velocity calculated by the network. Represents the Laplace operator, MSE represents the root mean square error, x i Represents a local window centered at x. and σ represents the mean of I and J, respectively. I and σ J Cov represents the standard deviations of I and J. IJ c1 and c2 represent constants, M represents the target co-offset data, which is composed of noff imaging profiles stitched together. I represents the 1st to noff-1th co-offset gathers, and J represents the 2nd to noffth co-offset gathers.

[0176] In one possible implementation, the second processing module 904 is further used for:

[0177] The auxiliary data loss function is obtained in the following way:

[0178] L a =-NCC(M,A)-SSIM(M,A)

[0179] Among them, L a represents the auxiliary data loss function, where A represents the auxiliary dataset, which is composed of noff parts of superimposed profiles stitched together horizontally.

[0180] In one possible implementation, the second processing module 904 is further used for:

[0181] The target loss function is obtained in the following way:

[0182] L=λL us +(1-λ)L a

[0183] Where L represents the target loss function, L us L represents the unsupervised loss function. a λ represents the auxiliary data loss function, and λ represents the hyperparameter set to balance the weights of the two loss functions.

[0184] The device for obtaining the offset velocity model provided in this application can implement the above-described method embodiments. Its implementation principle and technical effect are similar, and will not be described again here.

[0185] Figure 10 This is a hardware structure diagram of the device for acquiring the offset velocity model provided in an embodiment of this application. Figure 10 As shown, the device 1000 for acquiring the offset velocity model includes:

[0186] Processor 1001 and memory 1002;

[0187] The memory stores the instructions that the computer executes;

[0188] The processor executes the computer execution instructions stored in memory 1002, causing the offset velocity model acquisition device to perform the offset velocity model acquisition method as described above.

[0189] It should be understood that the aforementioned processor 1001 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0190] The memory 1002 may include high-speed random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0191] This application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method for obtaining the offset velocity model as described above.

[0192] This application also provides a computer program product, which, when executed by a processor, is used to implement the method for obtaining the offset velocity model as described above.

[0193] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0194] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0195] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.

[0196] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.

[0197] When integrated units / modules are implemented in hardware, the hardware can be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. Unless otherwise specified, the processor can be any suitable hardware processor, such as a CPU, GPU, FPGA, DSP, and ASIC, etc. Unless otherwise specified, the storage unit can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc.

[0198] If the integrated unit / module is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, 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. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0199] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0200] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0201] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for obtaining an offset velocity model, characterized in that, include: The seismic data is denoised and extracted to obtain a first common offset trace set, wherein the first common offset trace set includes multiple common offset data, and each common offset data corresponds to an offset. Multiple common offset data from the first common offset channel set are input into an initial offset velocity model to obtain the offset velocity output by the offset velocity model. The offset velocity model includes an offset velocity solver and a fusion unit. The offset velocity solver is used to obtain the velocity feature corresponding to each common offset data, and the fusion unit is used to fuse multiple velocity features to obtain the offset velocity. According to the offset velocity, the first common offset gather is repositioned to obtain the second common offset gather. The repositioning process includes pre-stack time offset and reaction correction. The second common offset gather is the common offset gather after the first common offset gather has been repositioned. The offset velocity model is iteratively updated based on the second common offset gather and the target loss function. After determining that the number of iterations has reached a threshold, the corresponding target offset velocity model is obtained, wherein the target loss function is a loss function based on offset velocity.

2. The method according to claim 1, characterized in that, The offset velocity model also includes: The reflection travel unit is used to acquire multiple travel time data corresponding to each of the common offset data in the first common offset trace set, wherein the travel time data is used to indicate the time required for seismic waves to reflect from the subsurface interface back to the surface at each offset; and to determine multiple data reflection travel times corresponding to the first common offset trace set based on each travel time data, and to determine a corresponding hyperbolic activation function based on each data reflection travel time. The position transformation unit is used to perform longitudinal position transformation on each of the common offset data according to the travel time of each of the data reflections, so as to obtain the transformed target common offset data; The offset velocity solver is used to obtain the velocity features based on the hyperbolic activation function and the longitudinal position transformation method corresponding to the target co-offset distance data. The offset velocity solver includes a U-net convolutional neural network.

3. The method according to claim 2, characterized in that, The travel time of the data reflection corresponding to the first common offset gather is obtained in the following way: Where t(t0, x, h) represents the data reflection travel time, h represents the offset distance, x represents the spatial horizontal position, t0 represents the two-way travel time corresponding to zero offset distance, and v represents the velocity of the medium above the interface.

4. The method according to claim 3, characterized in that, The transformed target common offset data is obtained in the following way: Among them, I output (y, x) represents the target common offset data, I input The common offset data is represented by y and x, which represent the longitudinal time and spatial horizontal position of the imaging point, respectively. T represents the travel time matrix calculated based on the physical laws of travel time, dt represents the time sampling interval, and round represents the rounding function.

5. The method according to claim 1, characterized in that, The target loss function is determined based on the unsupervised loss function and the auxiliary data loss function; The unsupervised loss function is a loss function based on offset velocity; The auxiliary data loss function is a loss function based on the target co-offset data; Based on the unsupervised loss function and the auxiliary data loss function, determine the corresponding target loss function.

6. The method according to claim 5, characterized in that, The unsupervised loss function is obtained in the following way: Among them, L us This represents the unsupervised loss function, where v represents the offset velocity calculated by the network. Represents the Laplace operator, MSE represents the root mean square error, x i Represents a local window centered at x. and These represent the means of I and J, respectively. and Cov represents the standard deviations of I and J. IJ c1 and c2 represent the covariance between I and J, c1 and c2 represent constants, M represents the target co-offset data, which is composed of noff imaging profiles stitched together, I represents the 1st to noff-1th co-offset gathers, and J represents the 2nd to noffth co-offset gathers.

7. The method according to claim 6, characterized in that, The auxiliary data loss function is obtained in the following way: Among them, L a The auxiliary data loss function is represented by A, which represents the auxiliary dataset, consisting of noff parts of the superimposed profile stitched together horizontally.

8. The method according to claim 7, characterized in that, The target loss function is obtained in the following way: Where L represents the target loss function, L us L represents the unsupervised loss function. a λ represents the auxiliary data loss function, and λ represents the hyperparameter set to balance the weights of the two loss functions.

9. A device for acquiring an offset velocity model, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-8.

Images