Earthquake velocity model acquisition method and device, electronic equipment and storage medium

Abstract

The invention provides a seismic velocity model acquisition method and device, electronic equipment and a storage medium, and relates to the technical field of seismic data processing. The method comprises the steps of obtaining a spatial position, a shot point coordinate, an initial velocity model and an initial imaging result, and inputting the spatial position, the shot point coordinate, the initial velocity model and the initial imaging result into a trained velocity model neural network to obtain a target velocity model. According to the method, the space structure information in the imaging result can be extracted by using the trained speed model neural network, and the tomography speed model is obtained according to the space structure information, so that the obtained speed model is higher in resolution and more accurate in structure.

Description

Technical Field

[0001] The present invention relates to the technical field of seismic data processing, and particularly relates to a method, device, electronic device and storage medium for obtaining a seismic velocity model. Background Art

[0002] Tomography is the cornerstone of geophysical exploration methods, which helps to reveal the internal structure characteristics and dynamic mechanisms of the Earth. The development of modern geophysical scientific research has put forward higher requirements for the accuracy and resolution of tomography technology. The currently adopted methods generally obtain a seismic velocity model based on travel time information.

[0003] The existing methods for obtaining a seismic velocity model cannot obtain a high-resolution velocity result with high wavenumber components, so the resolution of the obtained velocity model is not high. Summary of the Invention

[0004] The present invention provides a method, device, electronic device and storage medium for obtaining a seismic velocity model, which is used to solve the defect that the resolution of the obtained velocity model in the prior art is not high. The trained velocity model neural network is used to extract the spatial structure information in the imaging result, and the tomographic velocity model is obtained according to the spatial structure information, so that the obtained velocity model has higher resolution and more accurate structure.

[0005] The present invention provides a method for obtaining a seismic velocity model, including the following steps: Obtain the spatial position, shot point coordinates, initial velocity model and initial imaging result, where the spatial position is used to represent the position of the seismic wave receiver, the shot point coordinates are used to represent the position of the seismic wave source, the initial velocity model is used to represent the initial velocity distribution data, and the initial imaging result is the seismic imaging data obtained based on the initial velocity model; Input the spatial position, the shot point coordinates, the initial velocity model and the initial imaging result into the trained velocity model neural network to obtain a target velocity model. The target velocity model is obtained by the trained velocity model neural network extracting the spatial structure information based on the spatial position, the shot point coordinates, the initial velocity model and the initial imaging result, and constraining and inverting the velocity structure according to the spatial structure information.

[0006] According to the method for obtaining a seismic velocity model provided by the present invention, the obtaining of the initial imaging result includes: Obtain seismic gather data, where the seismic gather data is obtained by exciting seismic waves and recording the signals transmitted and reflected back by the seismic waves in the formation; Calculate the seismic gather data and the initial velocity model through a depth migration algorithm to obtain the initial imaging result.

[0007] According to a method for obtaining a seismic velocity model provided by the present invention, the trained velocity model neural network includes a velocity generation network and a spatial structure extraction network. Inputting the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result into the trained velocity model neural network to obtain a target velocity model includes: Inputting the spatial position into the velocity generation network to obtain a preliminary velocity model; Inputting the spatial position, the initial velocity model, and the initial imaging result into the spatial structure extraction network to obtain model correction information, where the model correction information is generated according to the spatial structure information; Constraining the structure of the velocity model obtained by inversion according to the sum of the preliminary velocity model and the model correction information to obtain the target velocity model.

[0008] According to a method for obtaining a seismic velocity model provided by the present invention, the trained velocity model neural network further includes a travel time factor calculation network. Constraining the structure of the velocity model obtained by inversion according to the sum of the preliminary velocity model and the model correction information to obtain the target velocity model includes: Inputting the spatial position and the shot point coordinates into the travel time factor calculation network to obtain a travel time factor; Constraining the structure of the velocity model obtained by inversion through the travel time factor, and the sum of the preliminary velocity model and the model correction information to obtain the target velocity model.

[0009] According to a method for obtaining a seismic velocity model provided by the present invention, before obtaining the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result, the method further includes: Determining an initial velocity model neural network, where the initial velocity model neural network includes a travel time factor calculation network, a velocity generation network, and a spatial structure extraction network. The spatial structure extraction network is a convolutional neural network, and both the travel time factor calculation network and the velocity generation network are multi-layer perceptron networks; Determining a target loss function corresponding to the initial velocity model neural network, where the target loss function includes a partial differential equation loss, a data loss, a boundary loss, a gradient constraint loss, and a structural consistency loss; Training the initial velocity model neural network according to the target loss function to obtain the trained velocity model neural network.

[0010] A method for obtaining a seismic velocity model provided by the present invention, training the initial velocity model neural network according to the target loss function to obtain the trained velocity model neural network, includes: Combining the outputs of the travel time factor calculation network and the velocity generation network with the output of the spatial structure extraction network to optimize the partial differential equation loss; Comparing the output of the spatial structure extraction network with the input imaging profile to optimize the structural consistency loss.

[0011] The present invention also provides a device for obtaining a seismic velocity model, including the following modules: A data acquisition module, configured to acquire a spatial position, shot point coordinates, an initial velocity model, and an initial imaging result, where the spatial position is used to represent the position of a seismic wave receiver, the shot point coordinates are used to represent the position of a seismic wave source, the initial velocity model is used to represent velocity distribution data obtained according to prior information, and the initial imaging result is seismic imaging data obtained based on the initial velocity model; A model acquisition module, configured to input the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result into the trained velocity model neural network to obtain a target velocity model, where the target velocity model is obtained by the trained velocity model neural network extracting spatial structure information based on the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result, and constraining and inverting the velocity structure according to the spatial structure information.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and running on the processor, where when the processor executes the computer program, the method for obtaining a seismic velocity model as described in any one of the above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method for obtaining a seismic velocity model as described in any one of the above is implemented.

[0014] The present invention also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the method for obtaining a seismic velocity model as described in any one of the above is implemented.

[0015] The method, device, electronic device, and storage medium for obtaining a seismic velocity model provided by the present invention extract spatial structure information in an imaging result through a trained velocity model neural network, and obtain a tomographic velocity model according to the spatial structure information, and the obtained velocity model has a higher resolution and a more accurate structure. Description of the Drawings

[0016] To more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the accompanying drawings required in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0017] Figure 1 It is one of the flow schematic diagrams of the method for obtaining the seismic velocity model provided by the present invention.

[0018] Figure 2 It is the flow schematic diagram of the method for obtaining the initial imaging result provided by the present invention.

[0019] Figure 3 It is the second flow schematic diagram of the method for obtaining the seismic velocity model provided by the present invention.

[0020] Figure 4 It is the architecture diagram of the SC-PINN method provided by the present invention.

[0021] Figure 5 It is the schematic diagram of the experimental results of the present invention tested on actual seismic data.

[0022] Figure 6 It is the schematic diagram of the imaging results of the initial linear velocity model, PINN inversion velocity model, and SC-PINN inversion velocity model provided by the present invention.

[0023] Figure 7 It is the structural schematic diagram of the device for obtaining the seismic velocity model provided by the present invention.

[0024] Figure 8 It is the entity structural schematic diagram of the electronic device provided by the present invention. Detailed implementation manners

[0025] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention in conjunction with the accompanying drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0026] Tomography technology is the cornerstone of geophysical exploration methods, which helps to reveal the internal structure characteristics and dynamic mechanisms of the Earth. The development of modern geophysical scientific research has put forward higher requirements for the accuracy and resolution of tomography technology. The methods currently adopted generally obtain the seismic velocity model based on travel-time information.

[0027] Existing methods for obtaining seismic velocity models cannot obtain high-resolution velocity results with high wavenumber components, so the resolution of the obtained velocity models is not high.

[0028] Exemplarily, existing methods for obtaining seismic velocity models include using the Physics informed neural network (PINN). PINN utilizes the automatic differentiation mechanism of the deep learning framework and adds partial differential equations (PDEs) as constraints to the neural network training process. During the optimization of the loss function, the neural network follows the physical process controlled by the PDE while fitting the observed data, enabling the network to accurately model and predict complex physical systems. For example, in fluid dynamics, PINN has been used to encode the Navier-Stokes equations in a neural network, enabling the extraction of velocity and pressure fields from visualized fluids and the prediction of the Reynolds number. This method represents a new fluid dynamics modeling strategy. In solid mechanics, PINN has been used to identify elastic material parameters, design topology optimization structures, and simulate crack propagation. In addition to these fields, PINNs have also been applied to heat transfer, structural mechanics, and even more cutting-edge applications, such as designing active composite materials for 4D printing, revealing potential kinetic mechanisms, and predicting laboratory earthquakes. PINN utilizes the powerful non-linear fitting ability of neural networks and has great application potential in solving inverse problems. Therefore, preliminary applications have also begun in the geophysical field and some progress has been made. Li et al. combined the Full waveform inversion (FWI) method with the PINN method, and experiments on simulated data demonstrated higher accuracy compared to traditional FWI methods. Sun et al. further proposed an implicit FWI method based on the deep neural network expression method and added Bayesian inference methods for uncertainty analysis, providing a new idea for solving geophysical inverse problems. To further improve the accuracy of tomography methods, researchers have started applying PINN to tomography problem solving. Solving tomography problems using PINN technology mainly relies on its solution of the eikonal equation. Existing PINN methods mainly solve the factorized eikonal equation. To improve the accuracy of parameter estimation, researchers have further proposed techniques such as Bayesian PINN and PINN Fourier neuron operators, but these methods are still relatively difficult to solve practical tomography problems. It can be understood that since the current PINN method based on the Multi-layer perceptron (MLP) structure does not have the ability to capture spatial structure information, it limits the accuracy of the PINN method in solving tomography problems.

[0029] In view of this, an embodiment of the present invention provides a method for obtaining a seismic velocity model. By obtaining a spatial position, shot point coordinates, an initial velocity model, and an initial imaging result, and inputting the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result into a trained velocity model neural network, a target velocity model is obtained. This method can use the trained velocity model neural network to extract spatial structure information from the imaging result, and obtain a tomographic velocity model based on the spatial structure information. The obtained velocity model has higher resolution and more accurate structure.

[0030] Next, the technical solutions in the embodiments of the present invention will be described with reference to the accompanying drawings in the embodiments of the present invention.

[0031] Figure 1 FIG. is one of the flow diagrams of the method for obtaining a seismic velocity model provided by the present invention. As Figure 1 shown, the method for obtaining a seismic velocity model can be applied to an electronic device, and the electronic device can be various types of devices with information processing capabilities during implementation. For example, the electronic device can include a personal computer, a laptop computer, a palm computer, or a server, etc.; the electronic device can also be a mobile terminal. For example, the mobile terminal can include a mobile phone, an in-vehicle computer, a tablet computer, or a projector, etc. As Figure 1 shown, the method can include the following steps 101 to step 102: Step 101: Obtain a spatial position, shot point coordinates, an initial velocity model, and an initial imaging result. The spatial position is used to represent the position of a seismic wave receiver, the shot point coordinates are used to represent the position of a seismic wave source, the initial velocity model is used to represent initial velocity distribution data, and the initial imaging result is seismic imaging data obtained based on the initial velocity model.

[0032] It should be noted that the method for obtaining the spatial position, shot point coordinates, initial velocity model, and initial imaging result can be by self-reading or by forwarding and obtaining from other devices. The present invention does not limit the manner of obtaining the spatial position, shot point coordinates, initial velocity model, and initial imaging result.

[0033] Step 102: Input the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result into a trained velocity model neural network to obtain a target velocity model. The target velocity model is obtained by the trained velocity model neural network extracting spatial structure information based on the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result, and constraining and inversing the velocity structure according to the spatial structure information.

[0034] It should be noted that in the solution of seismic tomography inverse problems, a very important prior information is the migration imaging result of seismic data. Even if the initial velocity model is inaccurate, the migration imaging result can still invert the main subsurface structural information and suggest the spatial structure that a reasonable inversion result should have. Therefore, the spatial structure information can be extracted through a trained velocity model neural network, that is, the trained velocity model neural network can extract the spatial structure information based on the spatial position, the shot coordinates, the initial velocity model, and the initial imaging result to constrain the physical neural network to complete the tomography inversion task.

[0035] It can be understood that the present invention relates to a seismic velocity model acquisition technology based on spatial structure migration and seismic tomography physical information neural network. It can use the spatial structure prior information of the seismic data imaging result, extract the spatial structure features by using a trained velocity model neural network, constrain the physical neural network to complete the tomography inversion task, and the obtained velocity model has higher resolution and more accurate structure.

[0036] Figure 2 It is a schematic flow chart of the method for obtaining the initial imaging result provided by the present invention. As Figure 2 shown, obtaining the initial imaging result in step 101 may include: Step 201: Obtain seismic shot gather data, which is obtained by exciting seismic waves and recording the signals of the seismic waves transmitted and reflected back in the formation.

[0037] It should be noted that the method for obtaining seismic shot gather data can be through receiver acquisition or through other device transmission. The present invention does not limit the method for obtaining seismic shot gather data.

[0038] Step 202: Calculate the seismic shot gather data and the initial velocity model through a depth migration algorithm to obtain the initial imaging result.

[0039] It should be noted that the method for calculating the seismic shot gather data and the initial velocity model through a depth migration algorithm to obtain the initial imaging result may include selecting a suitable depth migration algorithm, setting suitable calculation parameters, and performing calculations, etc. The present invention does not limit the method for calculating the seismic shot gather data and the initial velocity model through a depth migration algorithm to obtain the initial imaging result.

[0040] Exemplarily, for the depth migration algorithm, an appropriate depth migration algorithm can be selected according to the characteristics of the subsurface medium and imaging requirements. Common algorithms include wave equation migration algorithms (such as finite difference method, finite element method, etc.) and Kirchhoff migration algorithms. For setting calculation parameters, appropriate calculation parameters can be set according to the algorithm requirements and data characteristics, such as grid size, number of iterations, boundary conditions, etc. For performing depth migration calculation, it can be to input the seismic shot gather data and the initial velocity model into the depth migration algorithm for calculation. The algorithm will simulate the propagation process of seismic waves in the subsurface medium and perform migration processing on the seismic waves according to the initial velocity model to obtain the imaging result.

[0041] It can be understood that the method of calculating the initial imaging result by the depth migration algorithm for the seismic shot gather data and the initial velocity model can provide important prior information for obtaining the target velocity model and lay a foundation for obtaining a velocity result with higher resolution.

[0042] In some embodiments, the trained velocity model neural network includes a velocity generation network and a spatial structure extraction network. The step of inputting the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result into the trained velocity model neural network to obtain the target velocity model may include: inputting the spatial position into the velocity generation network to obtain a preliminary velocity model; inputting the spatial position, the initial velocity model, and the initial imaging result into the spatial structure extraction network to obtain model correction information, where the model correction information is generated according to the spatial structure information; and constraining the structure of the velocity model obtained by inversion according to the sum of the preliminary velocity model and the model correction information to obtain the target velocity model.

[0043] It should be noted that the velocity model neural network provided by the present invention can be a Space Correlation constrained Physics-Informed Neural Network (SC-PINN). The velocity generation network can adopt a Multi-Layer Perceptron (MLP) structure, and the spatial structure extraction network can adopt a Convolutional Neural Network (CNN).

[0044] Exemplarily, the SC-PINN network structure proposed in the present invention may include a velocity generation network based on the MLP structure (MLP-based velocity network, MVN) and a spatial structure extraction network based on the CNN (Spatial structure transfer network, SSTN). The SSTN network uses a CNN to extract the spatial structure of the observed data, enhancing the network's ability to capture high-frequency changes in the data.

[0045] Further, the trained velocity model neural network further includes a travel time factor calculation network. Constraining the velocity model structure obtained by inverting according to the sum of the preliminary velocity model and the model correction information to obtain the target velocity model may include: inputting the spatial position and the shot point coordinates into the travel time factor calculation network to obtain a travel time factor; jointly constraining the velocity model structure obtained by inverting through the travel time factor, as well as the sum of the preliminary velocity model and the model correction information, to obtain the target velocity model.

[0046] It should be noted that the travel time factor calculation network can adopt a multi-layer perceptron (MLP) structure.

[0047] Exemplarily, the SC-PINN network structure proposed in the present invention includes: a travel time factor calculation network .

[0048] It can be understood that the present invention proposes a space correlation constrained physics-informed neural network (Space Correlation constrained PINN, SC-PINN). This method introduces a convolutional neural network (CNN) to extract the spatial structure information of the observed data, and also introduces a multi-layer perceptron (MLP) to obtain spatial correlation information. Based on the texture transfer idea, the spatial structure of the seismic profile is transferred to the output velocity model, constraining the velocity model structure obtained by inversion, and greatly improving the inversion accuracy.

[0049] Further, the overall optimization objective of the SC-PINN network is: ; ; ; Wherein , and represent the trainable parameters of three networks, represents the initial velocity model, represents the seismic imaging result based on the initial velocity model, represents Wherein is the dimension, Denote the observed travel time from the shot point coordinates to the spatial position .

[0050] As can be seen from equation (1), the input of the travel time factor calculation network is the spatial position and the shot point coordinates . The input of the velocity generation network MVN based on the MLP structure is the spatial position . The input of the SSTN is the spatial position . The initial imaging velocity, i.e., the initial velocity model and the initial imaging profile, i.e., the initial imaging result . Based on the input spatial position coordinates , the SSTN samples the input velocity and imaging profile, then further vectorizes them. The vectorized result is added to the output result of the MVN to obtain the output velocity model.

[0051] It can be understood that the present invention proposes a spatially correlated constraint physics-driven neural network SC-PINN for tomographic inversion, which uses a CNN to extract the spatial structure features of the observed data, utilizes the prior information of the spatial correlation of the data to constrain the inverted velocity structure. At the same time, this method also combines the optimization objective of the PINN, which enables the SC-PINN method to not only ensure the rationality of the inversion result based on the tomographic physical rule constraints, but also improve the accuracy and resolution of the velocity inversion by using the spatial structure fitting ability of the CNN, and transfer the prior spatial structure information to the final velocity inversion result.

[0052] Figure 3 is the second schematic flow chart of the method for obtaining the seismic velocity model provided by the present invention. As Figure 3 shown, the method for obtaining the seismic velocity model includes: Step 301: Determine the initial velocity model neural network, where the initial velocity model neural network includes a travel time factor calculation network, a velocity generation network, and a spatial structure extraction network. The spatial structure extraction network is a convolutional neural network, and both the travel time factor calculation network and the velocity generation network are multi-layer perceptron networks; Step 302: Determine the target loss function corresponding to the initial velocity model neural network. The target loss function includes a partial differential equation loss, a data loss, a boundary loss, a gradient constraint loss, and a structural consistency loss; Step 303: Train the initial velocity model neural network according to the target loss function to obtain the trained velocity model neural network; Step 304: Obtain the spatial position, shot point coordinates, initial velocity model, and initial imaging result. The spatial position is used to characterize the position of the seismic wave receiver, the shot point coordinates are used to characterize the position of the seismic wave source, the initial velocity model is used to characterize the initial velocity distribution data, and the initial imaging result is seismic imaging data obtained based on the initial velocity model. Step 305: Input the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result into the trained velocity model neural network to obtain a target velocity model. The target velocity model is obtained by the trained velocity model neural network extracting spatial structure information based on the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result, and constraining and inverting the velocity structure according to the spatial structure information.

[0053] It should be noted that one of the main tasks of the SC-PINN method is to fit the eikonal equation. The eikonal equation belongs to a first-order partial differential equation and has the following form: ; where represents the gradient operator, represents the spatial position at the travel time, represents the velocity value at.

[0054] To avoid singularities in the solution process, the currently common solution form is to use the eikonal equation in the following factor form: ; Substituting Equation (3) into Equation (2), we can obtain the PDE for as: ; where represents the travel time factor to be solved.

[0055] The present invention uses the factorized eikonal equation as the PDE loss term. The total loss function of the SC-PINN method includes the PDE loss, data loss, boundary loss, gradient constraint loss, and structural consistency loss. The general expression is as follows: ; where represents the output of the travel time factor calculation network of, represents the output of the SSTN, represents the output of the MVN, represents the input seismic imaging profile, Denote the hyperparameters that control each loss term. Denote the label travel time factor.

[0056] Based on equation (4), the specific form of the PDE loss term is: ; Where Denote the operation of taking the mean.

[0057] The data loss term and the boundary loss term are defined in the following form: ; The gradient constraint loss is to constrain the first-order gradient of the output to also fit the first-order gradient of the observed data, and has the following form:

[0058] The structure consistency loss is to ensure that the speed The structure of the CNN network output is consistent with the spatial structure of the imaging profile, and is defined as follows: ; Where And Denote the corresponding Directional gradient and Gradient in the The function is used to calculate the dip angle.

[0059] Furthermore, training the initial velocity model neural network according to the target loss function to obtain the trained velocity model neural network may include: combining the outputs of the travel time factor calculation network and the velocity generation network with the output of the spatial structure extraction network to optimize the partial differential equation loss; comparing the output of the spatial structure extraction network with the input imaging profile to optimize the structure consistency loss.

[0060] It should be noted that denoting the input initial velocity model and the corresponding imaging profile as And , then first sampling the input based on the input spatial position , which can be expressed as: ; Where Denote the spatial sampling operation. In implementation, Is a Sliding window, and the spatial sampling operation selects the velocity model and the imaging profile of this part and sends them to the SSTN for processing. 、 And The input and output of the three parts of the network are: , , .

[0061] During the training phase, , and are used to calculate the loss function . The Adam optimizer is used to optimize the objective function, the initial learning rate is set to 0.002, and the cosine learning rate adjustment mechanism is used for learning.

[0062] It can be understood that the present invention uses the PDE loss and the observed data loss to optimize the network parameters, which enables the SC-PINN method to not only ensure the rationality of the inversion result based on the tomographic physical rule constraints, but also improve the accuracy and resolution of the velocity inversion by utilizing the spatial structure fitting ability of the CNN, and transfer the prior spatial structure information to the final velocity inversion result.

[0063] Next, an exemplary application of the embodiments of the present invention in a practical application scenario will be described.

[0064] Figure 4 is the architecture diagram of the SC-PINN method provided by the present invention. As Figure 4 shown, the SC-PINN network structure proposed in the present invention consists of three parts: the travel time factor calculation network , the velocity generation network based on the MLP structure (MLP-based velocity network, MVN) and the spatial structure extraction network based on CNN (Spatial structure transfer network, SSTN). The SSTN network uses CNN to extract the spatial structure of the observed data and enhance the network's ability to capture high-frequency changes in the data.

[0065] Figure 5 is a schematic diagram of the experimental results of the present invention tested on actual seismic data. As Figure 5 shown, the present method was tested on the actual seismic data of the Viking Graben in the North Sea. (a) represents the initial velocity model, the initial imaging result and the data example, (b) represents the iterative result of the PINN method, and (c) represents the iterative result of the SC-PINN method. As can be seen from Figure 5 , the result of the PINN method is difficult to effectively update the initial velocity, but the SC-PINN method can effectively update the linear initial velocity model.

[0066] Figure 6Schematic diagram of the imaging results of the initial linear velocity model, PINN inversion velocity model, and SC-PINN inversion velocity model provided by the present invention. As Figure 6 shown, the imaging results of the initial linear velocity model, PINN inversion velocity model, and SC-PINN inversion velocity model are presented. It can be seen that the deep focusing quality of the imaging results of the initial linear velocity model and the PINN inversion velocity model is poor, and the event axes are discontinuous. However, the imaging results of the SC-PINN model have better deep focusing, clear stratification of the formation, and clear depiction of fault structures, with significantly improved imaging quality.

[0067] It can be understood that the existing tomographic velocity modeling methods are only based on travel-time information and cannot obtain high-resolution velocity results with high wavenumber components. Therefore, the present invention designs a model. After obtaining seismic shot gather data and an initial velocity model, the underground depth-domain imaging results are first obtained through a depth migration algorithm. Then, the depth-domain imaging results are combined with the velocity model and input together with the travel-time information into the proposed seismic velocity modeling module based on spatial structure migration and seismic tomography physical information neural network to obtain a high-resolution velocity model. Since this method model utilizes the spatial structure information in the imaging results, the velocity model obtained is of higher resolution and more accurate structure than the tomographic velocity modeling method that only uses travel-time.

[0068] Based on the foregoing embodiments, the embodiments of the present invention provide an apparatus for obtaining a seismic velocity model. Each module included in the apparatus, as well as each unit included in each module, can be implemented by a processor; of course, it can also be implemented by specific logic circuits. During implementation, the processor can be a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA), etc.

[0069] The apparatus for obtaining a seismic velocity model provided by the present invention will be described below. The apparatus for obtaining a seismic velocity model described below can be mutually corresponded and referred to the method for obtaining a seismic velocity model described above.

[0070] Figure 7 Schematic diagram of the structure of the apparatus for obtaining a seismic velocity model provided by the present invention. As Figure 7 shown, the apparatus 400 includes a data acquisition module 401 and a model acquisition module 402, where: The data acquisition module 401 is configured to acquire spatial positions, shot point coordinates, an initial velocity model, and initial imaging results. The spatial positions are used to represent the positions of seismic wave receivers, the shot point coordinates are used to represent the positions of the seismic wave sources that excite seismic waves, the initial velocity model is used to represent the velocity distribution data obtained based on prior information, and the initial imaging results are seismic imaging data obtained based on the initial velocity model; A model acquisition module 402, configured to input the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result into a trained velocity model neural network to obtain a target velocity model. The target velocity model is obtained by the trained velocity model neural network extracting spatial structure information based on the spatial position, the shot point coordinates, the initial velocity model, and the initial imaging result, and inversely solving the velocity structure according to the constraint of the spatial structure information.

[0071] In some embodiments, the data acquisition module 401 is specifically configured to: acquire seismic shot gather data, which is obtained by exciting seismic waves and recording the signals of the seismic waves transmitted and reflected back in the formation; calculate the initial imaging result by performing a depth migration algorithm on the seismic shot gather data and the initial velocity model.

[0072] In some embodiments, the trained velocity model neural network includes a velocity generation network and a spatial structure extraction network. The model acquisition module 402 includes a first acquisition unit, a second acquisition unit, and a third acquisition unit, where The first acquisition unit is configured to input the spatial position into the velocity generation network to obtain a preliminary velocity model; The second acquisition unit is configured to input the spatial position, the initial velocity model, and the initial imaging result into the spatial structure extraction network to obtain model correction information, where the model correction information is generated according to the spatial structure information; The third acquisition unit is configured to inversely solve the velocity model structure according to the result of adding the preliminary velocity model and the model correction information to obtain the target velocity model.

[0073] In some embodiments, the trained velocity model neural network further includes a travel time factor calculation network. The third acquisition unit is specifically configured to: input the spatial position and the shot point coordinates into the travel time factor calculation network to obtain a travel time factor; jointly constrain the inversely solved velocity model structure by the travel time factor and the result of adding the preliminary velocity model and the model correction information to obtain the target velocity model.

[0074] In some embodiments, the apparatus further includes a model determination module, a parameter determination module, and a model training module, where The model determination module is configured to determine an initial velocity model neural network. The initial velocity model neural network includes a travel time factor calculation network, a velocity generation network, and a spatial structure extraction network. The spatial structure extraction network is a convolutional neural network, and both the travel time factor calculation network and the velocity generation network are multi-layer perceptron networks; The parameter determination module is configured to determine a target loss function corresponding to the initial velocity model neural network, where the target loss function includes a partial differential equation loss, a data loss, a boundary loss, a gradient constraint loss, and a structural consistency loss; The model training module is configured to train the initial velocity model neural network according to the target loss function to obtain the trained velocity model neural network.

[0075] In some embodiments, the model training module is specifically configured to: combine the outputs of the travel time factor calculation network and the velocity generation network with the output of the spatial structure extraction network to optimize the partial differential equation loss; compare the output of the spatial structure extraction network with the input imaging profile to optimize the structural consistency loss.

[0076] In the embodiments of the present invention, the trained velocity model neural network can be used to extract spatial structure information in the imaging result, and a tomographic velocity model can be obtained according to the spatial structure information, and the obtained velocity model has higher resolution and more accurate structure.

[0077] Figure 8 It is a schematic physical structure diagram of an electronic device provided by the present invention. As Figure 8 shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530, and a communication bus 540. Among them, the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call the logical instructions in the memory 530 to execute the method for obtaining the seismic velocity model, and the method includes: obtaining a spatial position, a shot point coordinate, an initial velocity model, and an initial imaging result, where the spatial position is used to represent the position of the seismic wave receiver, the shot point coordinate is used to represent the position of the seismic source that generates the seismic wave, the initial velocity model is used to represent the initial velocity distribution data, and the initial imaging result is seismic imaging data obtained based on the initial velocity model; inputting the spatial position, the shot point coordinate, the initial velocity model, and the initial imaging result into the trained velocity model neural network to obtain a target velocity model, where the target velocity model is obtained by the trained velocity model neural network extracting spatial structure information based on the spatial position, the shot point coordinate, the initial velocity model, and the initial imaging result, and constraining and inverting the velocity structure according to the spatial structure information.

[0078] In addition, when the logical instructions in the above-mentioned memory 530 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0079] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the method for obtaining a seismic velocity model provided by the above-mentioned various methods. The method includes: obtaining a spatial position, a shot point coordinate, an initial velocity model, and an initial imaging result. The spatial position is used to represent the position of a seismic wave receiver, the shot point coordinate is used to represent the position of a seismic wave source, the initial velocity model is used to represent initial velocity distribution data, and the initial imaging result is seismic imaging data obtained based on the initial velocity model; inputting the spatial position, the shot point coordinate, the initial velocity model, and the initial imaging result into a trained velocity model neural network to obtain a target velocity model. The target velocity model is obtained by the trained velocity model neural network extracting spatial structure information based on the spatial position, the shot point coordinate, the initial velocity model, and the initial imaging result, and constraining and inverting the velocity structure according to the spatial structure information.

[0080] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, they generate all or part of the processes or functions described in the embodiments of the present invention. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a server or data center that includes one or more integrated available media. The available medium may be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)).

[0081] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program, which when executed by a processor, implements a method for obtaining a seismic velocity model provided by the above methods. The method includes: obtaining a spatial position, a shot point coordinate, an initial velocity model, and an initial imaging result, where the spatial position is used to represent the position of a seismic wave receiver, the shot point coordinate is used to represent the position of a seismic wave source, the initial velocity model is used to represent initial velocity distribution data, and the initial imaging result is seismic imaging data obtained based on the initial velocity model; inputting the spatial position, the shot point coordinate, the initial velocity model, and the initial imaging result into a trained velocity model neural network to obtain a target velocity model, where the target velocity model is obtained by the trained velocity model neural network extracting spatial structure information based on the spatial position, the shot point coordinate, the initial velocity model, and the initial imaging result and constraining and inverting the velocity structure according to the spatial structure information.

[0082] The above computer-readable storage medium may adopt any combination of one or more computer-readable media. The computer-readable media may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (a non-exhaustive list) of the computer-readable storage medium include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), or a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0083] The computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries the computer-readable program code. Such a propagated data signal may take various forms, including - but not limited to - an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable medium other than the computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0084] The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including - but not limited to - wireless, wire, optical fiber, radio frequency (RF), etc., or any suitable combination of the above.

[0085] Computer program code for performing the operations of this specification may be written in one or more programming languages or combinations thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and also including conventional procedural programming languages such as the "C" language or similar programming languages. The program code may execute entirely on the user's computer, partially on the user's computer, execute as a stand-alone software package, execute partially on the user's computer and partially on a remote computer, or execute entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0086] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place or distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the objectives of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative effort.

[0087] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for obtaining a seismic velocity model, characterized in that: include: Acquire a spatial position, shot point coordinates, an initial velocity model, and an initial imaging result, wherein the spatial position is used to characterize the position of a seismic wave receiver, the shot point coordinates are used to characterize the position of a source that excites seismic waves, the initial velocity model is used to characterize initial velocity distribution data, and the initial imaging result is seismic imaging data obtained based on the initial velocity model; The spatial position, the shot point coordinates, the initial velocity model and the initial imaging result are input into a trained velocity model neural network to obtain a target velocity model, wherein the target velocity model is obtained by extracting spatial structure information from the trained velocity model neural network based on the spatial position, the shot point coordinates, the initial velocity model and the initial imaging result, and constraining the inverted velocity structure according to the spatial structure information.

2. The method for obtaining a seismic velocity model according to claim 1, characterized in that: The obtaining of the initial imaging result comprises: Acquiring seismic shot data, wherein the seismic shot data is obtained by exciting seismic waves and recording signals transmitted and reflected in the strata; The seismic shot gather data and the initial velocity model are calculated by a depth migration algorithm to obtain the initial imaging result.

3. The method for obtaining a seismic velocity model according to claim 1, characterized in that: The trained velocity model neural network includes a velocity generation network and a spatial structure extraction network, and the spatial position, the shot point coordinates, the initial velocity model and the initial imaging result are input into the trained velocity model neural network to obtain a target velocity model, including: Inputting the spatial position into the velocity generation network to obtain a preliminary velocity model; Inputting the spatial position, the initial velocity model and the initial imaging result into the spatial structure extraction network to obtain model correction information, wherein the model correction information is generated according to the spatial structure information; The inverted velocity model structure is constrained according to the result of adding the preliminary velocity model and the model correction information to obtain the target velocity model.

4. The method for obtaining a seismic velocity model according to claim 3, characterized in that: The trained velocity model neural network also includes a travel time factor calculation network. The velocity model structure constrained by the inversion result of adding the preliminary velocity model and the model correction information to obtain the target velocity model includes: Inputting the spatial position and the shot point coordinates into the travel time factor calculation network to obtain the travel time factor; The target velocity model is obtained by constraining the inverted velocity model structure through the travel time factor and the result of adding the preliminary velocity model and the model correction information.

5. The method for obtaining a seismic velocity model according to claim 1, characterized in that: Before acquiring the spatial position, shot point coordinates, initial velocity model and initial imaging results, the method further includes: Determine an initial speed model neural network, wherein the initial speed model neural network includes a travel time factor calculation network, a speed generation network and a spatial structure extraction network, wherein the spatial structure extraction network is a convolutional neural network, and the travel time factor calculation network and the speed generation network are both multi-layer perceptron networks; Determine a target loss function corresponding to the initial velocity model neural network, wherein the target loss function includes partial differential equation loss, data loss, boundary loss, gradient constraint loss and structural consistency loss; The initial speed model neural network is trained according to the target loss function to obtain the trained speed model neural network.

6. The method for obtaining a seismic velocity model according to claim 5, characterized in that: The step of training the initial speed model neural network according to the target loss function to obtain the trained speed model neural network includes: Combining the outputs of the travel time factor calculation network and the velocity generation network with the output of the spatial structure extraction network to optimize the partial differential equation loss; The spatial structure extraction network output is compared with the input imaging profile and the structural consistency loss is optimized.

7. A device for obtaining a seismic velocity model, characterized in that: include: A data acquisition module, used to acquire a spatial position, shot point coordinates, an initial velocity model and an initial imaging result, wherein the spatial position is used to characterize the position of a seismic wave receiver, the shot point coordinates are used to characterize the position of a source that excites seismic waves, the initial velocity model is used to characterize velocity distribution data obtained according to prior information, and the initial imaging result is seismic imaging data obtained based on the initial velocity model; The model acquisition module is used to input the spatial position, the shot point coordinates, the initial velocity model and the initial imaging result into a trained velocity model neural network to obtain a target velocity model, wherein the target velocity model is obtained by extracting spatial structure information from the trained velocity model neural network based on the spatial position, the shot point coordinates, the initial velocity model and the initial imaging result, and constraining the inverted velocity structure according to the spatial structure information.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the method for obtaining the seismic velocity model according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for obtaining a seismic velocity model according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for obtaining a seismic velocity model according to any one of claims 1 to 6 is implemented.

Images