Seismic velocity modeling optimization method and device, computing equipment and storage medium

Abstract

The invention discloses a seismic velocity modeling optimization method and device, computing equipment and a storage medium, and the method comprises the steps: obtaining three-dimensional seismic time domain interval velocity data based on three-dimensional seismic depth domain migration velocity data; performing learning training on the three-dimensional seismic time domain interval velocity data by using the three-dimensional seismic time domain result profile data based on a neural network model to obtain a training model; and inputting the three-dimensional earthquake time domain result profile data into the training model to obtain optimized three-dimensional earthquake time domain interval velocity data. According to the seismic velocity modeling optimization method, the existing velocity model can be optimized, the precision of the seismic velocity model is improved, human intervention and calculation time can be reduced, and the working efficiency is improved.

Description

Technical Field

[0001] The present invention relates to a seismic velocity modeling optimization method, and in particular to a seismic velocity modeling optimization method, device, computing equipment and computer storage medium. Background Art

[0002] Seismic velocity modeling is a core link in the field of geophysical exploration. It is of vital importance to oil and gas exploration, earthquake risk assessment, underground resource management, etc. With the development of science and technology, traditional seismic velocity modeling methods have gradually shown their limitations, especially when dealing with complex geological structures. Traditional seismic velocity modeling methods have shown great limitations in accuracy and efficiency. Summary of the invention

[0003] In view of the above problems, the present invention is proposed to provide a seismic velocity modeling optimization method, apparatus, computing device and computer storage medium that overcome the above problems or at least partially solve the above problems.

[0004] According to one aspect of the present invention, a seismic velocity modeling optimization method is provided, comprising:

[0005] Acquire 3D seismic time domain layer velocity data based on 3D seismic depth domain migration velocity data;

[0006] Based on the neural network model, the three-dimensional seismic time domain result profile data is used to perform learning training on the three-dimensional seismic time domain layer velocity data to obtain a training model; and

[0007] The three-dimensional seismic time domain profile data is input into the training model to obtain the optimized three-dimensional seismic time domain layer velocity data.

[0008] Optionally, the neural network model includes a long short-term memory network and a convolutional neural network, the long short-term memory network is used to extract time features from three-dimensional seismic time domain result profile data and three-dimensional seismic time domain layer velocity data, and the convolutional neural network is used to extract spatial features from three-dimensional seismic time domain result profile data and three-dimensional seismic time domain layer velocity data.

[0009] Optionally, the sampling time interval of the three-dimensional seismic time-domain layer velocity data is 10 ms, and the number of longitudinal sampling points of the three-dimensional seismic time-domain layer velocity data is 600.

[0010] Optionally, the sampling time interval of the three-dimensional seismic time domain result profile data is 10ms, and the longitudinal sampling points of the three-dimensional seismic time domain result profile data are 600. If the longitudinal sampling points of the three-dimensional seismic time domain result profile data exceed 600, only the first 600 longitudinal sampling points are selected. If the longitudinal sampling points of the three-dimensional seismic time domain result profile data are less than 600, zero values ​​are added at the end of each seismic trace until the longitudinal sampling points reach 600.

[0011] Optionally, the number of hidden units in the LSTM layer is set to 1000.

[0012] Optionally, the convolutional neural network includes an encoding structure and a decoding structure, wherein the encoding structure is used to extract spatial features from input three-dimensional seismic time domain result profile data and three-dimensional seismic time domain layer velocity data, and the decoding structure is used to reconstruct the input three-dimensional seismic time domain layer velocity data based on the extracted spatial features.

[0013] Optionally, the training of the neural network model uses a mean square error loss function to calculate the error between the predicted value and the true value, and uses an Adam optimizer to update parameters.

[0014] According to another aspect of the present invention, a seismic velocity modeling optimization device is provided, the seismic velocity modeling optimization device comprising:

[0015] A data processing module, the data processing module is suitable for obtaining three-dimensional seismic time-domain layer velocity data based on three-dimensional seismic depth-domain migration velocity data;

[0016] A network training module, the network training module is suitable for using the three-dimensional seismic time domain result profile data to the three-dimensional seismic time domain layer velocity data to perform learning training based on the neural network model to obtain a training model; and

[0017] The data optimization module is suitable for inputting the three-dimensional seismic time domain result profile data into the training model to obtain the optimized three-dimensional seismic time domain layer velocity.

[0018] According to another aspect of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0019] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned seismic velocity modeling optimization method.

[0020] According to another aspect of the present invention, a computer storage medium is provided, in which at least one executable instruction is stored, and the executable instruction enables a processor to execute operations corresponding to the above-mentioned seismic velocity modeling optimization method.

[0021] According to the seismic velocity modeling optimization method, device, computing device and computer storage medium of the present invention, three-dimensional seismic time domain layer velocity data is obtained based on three-dimensional seismic depth domain offset velocity data; three-dimensional seismic time domain result profile data is used to learn and train three-dimensional seismic time domain layer velocity data based on a neural network model to obtain a training model; and three-dimensional seismic time domain result profile data is input into the training model to obtain optimized three-dimensional seismic time domain layer velocity data. The seismic velocity modeling method based on this scheme can not only optimize the existing velocity model and improve the accuracy of the seismic velocity model, but also reduce human intervention and calculation time, and improve work efficiency.

[0022] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0024] Figure 1 A schematic diagram of a flow chart of a seismic velocity modeling optimization method according to an embodiment of the present invention is shown;

[0025] Figure 2 A schematic diagram of the structure of a neural network model according to an embodiment of the present invention is shown;

[0026] Figure 3 3D seismic depth domain migration velocity data (inline 1) according to an embodiment of the present invention is shown;

[0027] Figure 4 The 3D seismic depth domain migration velocity profile (inline 1) before optimization according to an embodiment of the present invention is shown;

[0028] Figure 5 The optimized 3D seismic depth domain migration velocity profile (inline 1) according to an embodiment of the present invention is shown;

[0029] Figure 6 3D seismic depth domain migration velocity data (inline2) according to an embodiment of the present invention is shown;

[0030] Figure 7The 3D seismic depth domain migration velocity profile (inline2) before optimization according to an embodiment of the present invention is shown;

[0031] Figure 8 The optimized 3D seismic depth domain migration velocity profile (inline2) according to an embodiment of the present invention is shown;

[0032] Fig. 9 3D seismic depth domain migration velocity data (inline3) according to an embodiment of the present invention is shown;

[0033] Fig.10 The 3D seismic depth domain migration velocity profile (inline3) before optimization according to an embodiment of the present invention is shown;

[0034] Fig.11 The optimized 3D seismic depth domain migration velocity profile (inline3) according to an embodiment of the present invention is shown;

[0035] Fig.12 A schematic diagram showing the structure of a seismic velocity modeling optimization device according to an embodiment of the present invention is shown; and

[0036] Fig.13 A schematic diagram of the structure of a computing device provided according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0037] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0038] Figure 1 A schematic flow chart of a seismic velocity modeling optimization method according to an embodiment of the present invention is shown.

[0039] like Figure 1 As shown, the method includes:

[0040] Step S110, acquiring 3D seismic time-domain layer velocity data based on 3D seismic depth-domain migration velocity data.

[0041] In step S110, the 3D seismic depth domain migration velocity data V is converted into 3D seismic time domain layer velocity data V_t. Optionally, the 3D seismic depth domain migration velocity data V is converted into 3D seismic time domain layer velocity data V_t by formula (1):

[0042]

[0043] Wherein: V_t(t) represents the 3D seismic time domain layer velocity at time t, that is, the 3D seismic time domain layer velocity at time t at depth z, V(z) represents the 3D seismic depth domain offset velocity, that is, the 3D seismic depth domain offset velocity at depth z, t is the propagation time of the seismic wave in the 3D seismic time domain layer velocity V_t, and T is the propagation time of the seismic wave in the 3D seismic depth domain offset velocity V.

[0044] In an optional implementation, the sampling time interval of the 3D seismic time domain layer velocity data can be set to 10 ms, for example. The specific sampling time interval can be set according to the specific situation and is not specifically limited here. The longitudinal sampling points of the 3D seismic time domain layer velocity data can also be set according to the specific situation, for example, to 600. The number of specific sampling points can be set by those skilled in the art according to the specific situation and is not specifically limited here.

[0045] In an optional implementation, the sampling time interval of the three-dimensional seismic time domain result profile data and the longitudinal sampling points of the three-dimensional seismic time domain result profile data can be set according to specific circumstances. Assuming that the longitudinal sampling points of the three-dimensional seismic time domain layer velocity are set to N, if the number of longitudinal sampling points of the three-dimensional seismic time domain result profile data is greater than the set value N, only the first N longitudinal sampling points are taken; if the number of longitudinal sampling points of the three-dimensional seismic time domain result profile data is less than the set value N, zero values ​​are added at the end of each seismic trace until the longitudinal sampling point is N, that is, the three-dimensional seismic time domain result profile data and the three-dimensional seismic time domain layer velocity data are consistent in the longitudinal direction.

[0046] Specifically, the sampling time interval of the three-dimensional seismic time domain result profile data can be 10ms, and the longitudinal sampling points of the three-dimensional seismic time domain result profile data can be 600. If the longitudinal sampling points of the three-dimensional seismic time domain result profile data exceed 600, only the first 600 longitudinal sampling points are selected. If the longitudinal sampling points of the three-dimensional seismic time domain result profile data are less than 600, zero values ​​are added at the end of each seismic trace until the longitudinal sampling points are 600, so that the three-dimensional seismic time domain result profile data and the three-dimensional seismic time domain layer velocity data are consistent in the longitudinal direction. Usually, the number of transverse sampling points of the three-dimensional seismic time domain layer velocity V_t (that is, the number of seismic traces of each profile) will be more than the three-dimensional seismic time domain result profile data. Therefore, the number of seismic traces of the three-dimensional seismic time domain result profile data needs to be supplemented with 0 values ​​to be consistent with the three-dimensional seismic time domain layer velocity V_t.

[0047] Step S120, based on the neural network model, the three-dimensional seismic time domain result profile data is used to perform learning training on the three-dimensional seismic time domain layer velocity data to obtain a training model.

[0048] In an optional embodiment, the neural network model includes a long short-term memory network and a convolutional neural network, the long short-term memory network is used to extract time features from three-dimensional seismic time domain result profile data and three-dimensional seismic time domain layer velocity data, and the convolutional neural network is used to extract spatial features from three-dimensional seismic time domain result profile data and three-dimensional seismic time domain layer velocity data.

[0049] Specifically, deep learning training is performed on the three-dimensional seismic time domain profile data to the three-dimensional seismic time domain layer velocity data to obtain a neural network training model, wherein the neural network model adopts a network model that combines a long short-term memory network (LSTM) and a convolutional neural network (CNN) for learning and training.

[0050] The deep learning training of the LSTM layer includes: arranging the three-dimensional seismic time domain result profile data and the three-dimensional seismic time domain layer velocity data (also called input data) into a format suitable for LSTM processing, and arranging the three-dimensional seismic time domain result profile data and the three-dimensional seismic time domain layer velocity data into DATA(a)(b)(t), where a is the number of data in the longitudinal survey line direction, b is the number of data in the transverse survey line direction, and t is the number of longitudinal sampling points.

[0051] A two-layer LSTM network is used to extract time features from 3D seismic time domain profile data and 3D seismic time domain layer velocity data. The output of the first layer is used as the input of the second layer. For the LSTM network, the data DATA(a)(b)(t), a is the number of samples, that is, the number of seismic profiles in the longitudinal survey line direction. b is the number of observed time points, that is, the number of channels in each seismic profile. t is the number of features at each time step, that is, the number of longitudinal sampling points of each seismic data channel.

[0052] The specific working principle of the LSTM layer is forward propagation and back propagation. Forward propagation: During the training process, the data of each time step is input into the LSTM network through forward propagation. Each LSTM unit receives the hidden state and cell state from the previous time step as input, and receives the input data of the current time step at the same time. Backward propagation: The error between the predicted value and the true value is calculated according to the mean square error loss function, and then each weight and bias is adjusted through the back propagation algorithm to optimize the parameters of the network. The mean square error loss function expression is:

[0053]

[0054] In formula (2), L represents the loss between the actual output value and the true value, n is the number of training samples, y is the true value, and y' is the actual output value.

[0055] During the back-propagation process, the Adam (Adaptive Moment Estimation) optimizer is used to update the network parameters to minimize the loss function.

[0056] Among them, the main working method of the Adam optimizer is to use two state variables m_t and v_t to calculate the update of each neural network parameter. The calculation method is:

[0057] m_t=b1*m{t-1}+(1-b1)*g_t (3)

[0058] v_t=b2*v{t-1}+(1-b2)*g_t 2 (4)

[0059] Among them: g_t represents the gradient of the current step, m_t is the first-order moment estimate, that is, the mean of the gradient. v_t is the second-order moment estimate, b1 and b2 are decay rate parameters, usually set to values ​​close to 1, and the above variables are used to calculate the update step of the neural network parameters.

[0060]

[0061] Where: θ_t represents the parameter of the tth step, eta is the learning rate, and ε is a very small number, usually set to 0.00000001.

[0062] The number of hidden units in the LSTM layer is usually set to 1000.

[0063] Figure 2 FIG. 2 shows a schematic diagram of the structure of a neural network model according to an embodiment of the present invention. Figure 2 As shown, the neural network model of this embodiment uses a combination of LSTM and CNN. The first and second layers are long short-term memory neural network LSTM layers, and the third to eighteenth layers are convolutional neural network models. The convolutional neural network model is trained through operations such as convolution, pooling, discarding and upsampling.

[0064] In an optional embodiment, the convolutional neural network includes a coding structure and a decoding structure, wherein the coding structure is used to extract spatial features from the input three-dimensional seismic time domain result profile data and the three-dimensional seismic time domain layer velocity data, and the decoding structure is used to reconstruct the input three-dimensional seismic time domain layer velocity data according to the extracted spatial features. Specifically, the coding structure may include three convolutional layers, three maximum pooling layers and three dropout layers, and the number of convolution kernels of the convolutional layers remains consistent, usually 64. The decoding structure includes three upsampling layers and three convolutional layers. Among them, the eighteenth layer is the output layer, which is a single-core convolutional layer, used to reconstruct the optimized three-dimensional seismic depth domain offset velocity data. The convolutional neural network model of the neural network model structure of the embodiment of the present application has a more stable effect when processing actual data.

[0065] Specifically, the first and second layers of the neural network model structure of the embodiment of the present application are long short-term memory neural network LSTM layers, the third to eleventh layers belong to the CNN encoding structure, and the twelfth to eighteenth layers belong to the CNN decoding structure. Specifically as follows:

[0066] LSTM Layer:

[0067] The first layer: long short-term memory layer, dimension 1000,

[0068] Second: Long short-term memory layer, the dimension is the number of vertical sample points of the data, CNN encoding structure:

[0069] The third layer: convolution layer: 64 3x3 convolution kernels with a step size of 1.

[0070] The fourth layer is the maximum pooling layer with a pooling window of 2x2.

[0071] The fifth layer is the Dropout layer with a dropout rate of 0.1.

[0072] Sixth layer: Convolutional layer: 64 3x3 convolution kernels with a stride of 1.

[0073] The seventh layer is the maximum pooling layer with a pooling window of 2x2.

[0074] The eighth layer: Dropout layer, the dropout rate is 0.1.

[0075] Ninth layer: Convolutional layer: 64 3x3 convolution kernels with a step size of 1.

[0076] The tenth layer: the maximum pooling layer, the pooling window is 2x2.

[0077] The eleventh layer is a Dropout layer with a dropout rate of 0.1.

[0078] CNN decoding structure:

[0079] Layer 12: Upsampling, size is 2x2.

[0080] Thirteenth layer: Convolutional layer: 64 3x3 convolution kernels with a step size of 1.

[0081] Layer 14: Upsampling, size is 2x2.

[0082] Fifteenth layer: Convolutional layer: 64 3x3 convolution kernels with a step size of 1.

[0083] The 16th layer: upsampling, size is 2x2.

[0084] Seventeenth layer: Convolutional layer: 64 3x3 convolution kernels with a step size of 1.

[0085] Eighteenth layer: Convolutional layer: 1 3x3 convolution kernel with a stride of 1.

[0086] Step S130, inputting the three-dimensional seismic time domain result profile data into the training model to obtain optimized three-dimensional seismic time domain layer velocity data.

[0087] Specifically, the three-dimensional seismic time domain result profile data is input, the above-mentioned trained neural network model is loaded, and the input three-dimensional seismic time domain result profile data is predicted. The output obtained is the optimized three-dimensional seismic time domain layer velocity data. In this process, self-supervised deep learning velocity model optimization is realized.

[0088] Specifically, Figure 3 is the 3D seismic depth domain migration velocity imaging profile inline1 of a certain work area. Figure 3 It can be seen that the structural fluctuations of the section are drastic (e.g. Figure 3 There is a clear structural change. Figure 4 Display Figure 3 The 3D seismic depth domain migration velocity imaging section corresponds to the 3D seismic depth domain migration velocity section before optimization ( Figure 4 The speed is relatively uniform, forming a block, with almost no changes in the horizontal and vertical directions. Figure 3 The corresponding tectonic changes are obviously inconsistent, indicating that the resolution of the 3D seismic depth-domain migration velocity imaging profile is insufficient. Figure 5 It is shown that Figure 1 The seismic profile corresponds to the three-dimensional seismic depth domain migration velocity imaging profile optimized by the seismic modeling data optimization method of this embodiment, from Figure 5 From the circled area, it can be clearly seen that the 3D seismic depth domain migration velocity imaging profile is no longer a single block, but has lateral and vertical changes, showing dramatic structural changes, that is, the optimized velocity resolution has been improved, which is more consistent with the structure of the seismic profile.

[0089] In order to verify the effects of different positions in the same work area, two more side lines, inline2 and inline3, were selected. Figure 6 Inline2 is another 3D seismic depth domain migration velocity imaging profile in this work area. Figure 7 and Figure 8 Shows Figure 6 Corresponding 3D seismic depth domain migration velocity imaging profiles before and after optimization. Fig. 9 Inline3 is another 3D seismic depth domain migration velocity imaging profile in the work area. Fig.10 and Fig.11 Shows Fig. 9 The corresponding optimized 3D seismic depth domain migration velocity imaging profile. From the comparison, it can be seen that it is consistent with the situation of inline1. It can be clearly seen that the accuracy of the optimized velocity model has been greatly improved compared with before, and the velocity resolution has been significantly improved.

[0090] The seismic velocity modeling optimization method according to the present embodiment utilizes an advanced deep learning architecture, combined with a long short-term memory neural network and a convolutional neural network, to improve the accuracy and efficiency of seismic velocity modeling. By learning and extracting features from seismic profile data, the software can automatically optimize the velocity model, thereby providing more accurate information for geological exploration and seismic interpretation. The present embodiment can optimize the existing velocity model and improve the accuracy of the seismic velocity model. The present embodiment aims to improve work efficiency by reducing human intervention and computing time. The core algorithm structure of the present embodiment includes a multi-layer LSTM network and a CNN network. The LSTM network is responsible for processing serialized seismic data and capturing the time-dependent features therein, while the CNN network is responsible for extracting spatial features. By combining the two networks, the network model of the software can understand the seismic data more comprehensively, thereby providing a more accurate velocity model.

[0091] Fig.12 FIG. 2 shows a functional structure diagram of a seismic velocity modeling optimization device provided according to an embodiment of the present invention. Fig.12 As shown, the device 200 includes: a data processing module 210, a network training module 220 and a data optimization module 230.

[0092] The data processing module 210 is adapted to obtain 3D seismic time domain layer velocity data based on 3D seismic depth domain migration velocity data;

[0093] The network training module 220 is adapted to perform learning training on the three-dimensional seismic time domain layer velocity data based on the neural network model using the three-dimensional seismic time domain result profile data to obtain a training model; and

[0094] The data optimization module 230 is suitable for inputting the three-dimensional seismic time domain result profile data into the training model to obtain the optimized three-dimensional seismic time domain layer velocity.

[0095] In an optional embodiment, the neural network model includes a long short-term memory network and a convolutional neural network, the long short-term memory network is used to extract time features from three-dimensional seismic time domain result profile data and three-dimensional seismic time domain layer velocity data, and the convolutional neural network is used to extract spatial features from three-dimensional seismic time domain result profile data and three-dimensional seismic time domain layer velocity data.

[0096] In an optional implementation, the sampling time interval of the three-dimensional seismic time-domain layer velocity data is 10 ms, and the number of longitudinal sampling points of the three-dimensional seismic time-domain layer velocity data is 600.

[0097] In an optional implementation, the sampling time interval of the three-dimensional seismic time domain result profile data is 10ms, and the longitudinal sampling points of the three-dimensional seismic time domain result profile data are 600. If the longitudinal sampling points of the three-dimensional seismic time domain result profile data exceed 600, only the first 600 longitudinal sampling points are selected. If the longitudinal sampling points of the three-dimensional seismic time domain result profile data are less than 600, zero values ​​are added at the end of each seismic trace until the longitudinal sampling points reach 600.

[0098] In an optional implementation, the number of hidden units in each layer of the long short-term memory network is set to 1000.

[0099] In an optional embodiment, the convolutional neural network includes an encoding structure and a decoding structure, wherein the encoding structure is used to extract spatial features from input three-dimensional seismic time domain result profile data and three-dimensional seismic time domain layer velocity data, and the decoding structure is used to reconstruct the input three-dimensional seismic time domain layer velocity data based on the extracted spatial features.

[0100] In an optional embodiment, the training of the neural network model uses a mean square error loss function to calculate the error between the predicted value and the true value, and uses an Adam optimizer to update parameters.

[0101] It can be seen that the seismic velocity modeling optimization device according to this embodiment utilizes an advanced deep learning architecture, combined with a long short-term memory neural network and a convolutional neural network, which not only improves the accuracy and efficiency of seismic velocity modeling, but also reduces human intervention and calculation time, thereby improving work efficiency.

[0102] An embodiment of the present invention provides a non-volatile computer storage medium, wherein the computer storage medium stores at least one executable instruction, and the computer executable instruction can execute the seismic velocity modeling optimization method in any of the above method embodiments.

[0103] Fig.13 A schematic diagram of the structure of a computing device provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0104] like Fig.13 As shown, the computing device may include: a processor (processor) 302 , a communication interface (Communications Interface) 304 , a memory (memory) 306 , and a communication bus 308 .

[0105] The processor 302, the communication interface 304, and the memory 306 communicate with each other via the communication bus 308. The communication interface 304 is used to communicate with other devices such as client or other server network elements. The processor 302 is used to execute the program 310, which can specifically execute the relevant steps in the above method embodiment.

[0106] Specifically, the program 310 may include program codes, which include computer operation instructions.

[0107] The processor 302 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computing device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0108] The memory 306 is used to store the program 310. The memory 306 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0109] The program 310 can be specifically used to enable the processor 402 to execute the seismic velocity modeling optimization method in any of the above-mentioned method embodiments. The specific implementation of each step in the program 310 can refer to the corresponding descriptions in the corresponding steps and units in the above-mentioned seismic velocity modeling optimization method embodiments, which will not be repeated here. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-mentioned devices and modules can refer to the corresponding process description in the above-mentioned method embodiments, which will not be repeated here.

[0110] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0111] Similarly, it should be understood that in order to streamline the present invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, the various features of the embodiments of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the following intention: that the claimed invention requires more features than the features explicitly recited in each claim. More specifically, as reflected in the claims below, inventive aspects lie in less than all the features of the individual embodiments disclosed above. Therefore, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, with each claim itself serving as a separate embodiment of the present invention.

[0112] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0113] In addition, those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims below, any one of the claimed embodiments may be used in any combination.

[0114] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim that lists a number of devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be understood as limitations on the order of execution.

Claims

1. A seismic velocity modeling optimization method, characterized in that: The method comprises: Acquire 3D seismic time domain layer velocity data based on 3D seismic depth domain migration velocity data; Based on the neural network model, the three-dimensional seismic time domain result profile data is used to perform learning training on the three-dimensional seismic time domain layer velocity data to obtain a training model; and The three-dimensional seismic time domain result profile data is input into the training model to obtain optimized three-dimensional seismic time domain layer velocity data.

2. The method according to claim 1, characterized in that The neural network model includes a long short-term memory network and a convolutional neural network. The long short-term memory network is used to extract time features from the three-dimensional seismic time domain result profile data and the three-dimensional seismic time domain layer velocity data. The convolutional neural network is used to extract spatial features from the three-dimensional seismic time domain result profile data and the three-dimensional seismic time domain layer velocity data.

3. The method according to claim 1, characterized in that: The sampling time interval of the three-dimensional seismic time-domain layer velocity data is 10 ms, and the number of longitudinal sampling points of the three-dimensional seismic time-domain layer velocity data is 600.

4. The method according to claim 1, characterized in that The sampling time interval of the three-dimensional seismic time domain result profile data is 10ms, and the number of longitudinal sampling points of the three-dimensional seismic time domain result profile data is 600. If the number of longitudinal sampling points of the three-dimensional seismic time domain result profile data exceeds 600, only the first 600 longitudinal sampling points are selected. If the number of longitudinal sampling points of the three-dimensional seismic time domain result profile data is less than 600, zero values ​​are added at the end of each seismic trace until the number of longitudinal sampling points reaches 600.

5. The method according to claim 2, characterized in that: The number of hidden units in each layer of the LSTM network is set to 1000.

6. The method according to claim 2, characterized in that The convolutional neural network includes an encoding structure and a decoding structure, wherein the encoding structure is used to extract spatial features from input three-dimensional seismic time domain result profile data and three-dimensional seismic time domain layer velocity data, and the decoding structure is used to reconstruct the input three-dimensional seismic time domain layer velocity data based on the extracted spatial features.

7. The method according to claim 1, characterized in that The training of the neural network model uses a mean square error loss function to calculate the error between the predicted value and the true value, and uses an Adam optimizer to update parameters.

8. A seismic velocity modeling optimization device, characterized in that: The seismic velocity modeling optimization device comprises: A data processing module, wherein the data processing module is adapted to obtain three-dimensional seismic time-domain layer velocity data based on three-dimensional seismic depth-domain migration velocity data; A network training module, wherein the network training module is adapted to perform learning training on the three-dimensional seismic time domain layer velocity data based on a neural network model using the three-dimensional seismic time domain result profile data to obtain a training model; and A data optimization module, wherein the data optimization module is adapted to input the three-dimensional seismic time domain result profile data into the training model to obtain optimized three-dimensional seismic time domain layer velocity data.

9. A computing device comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the seismic velocity modeling optimization method according to any one of claims 1-7.

10. A computer storage medium, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to execute operations corresponding to the seismic velocity modeling optimization method according to any one of claims 1 to 7.

Images