An acoustic-elastic coupling medium parameter inversion method and system

Abstract

The application discloses an acoustic-elastic coupling medium parameter inversion method and system, and the method comprises the following steps: determining a unified wave equation of an acoustic-elastic coupling medium in an offshore oil and gas exploration area, representing an elastic medium wave equation as a first-order velocity-stress form, and degenerating into an acoustic wave equation in an acoustic medium when the shear modulus is zero; determining a fusion physical mechanism neural network structure, wherein the structure comprises a fully-connected neural network for representing particle velocity and a fully-connected neural network for representing model parameters; constructing a loss function based on particle velocity observation data, a residual point set and well logging data, and training the neural network through iterative optimization; and obtaining the parameter distribution of P-wave velocity, S-wave velocity and density based on the trained fully-connected neural network for representing model parameters. The application is suitable for offshore oil and gas exploration medium parameter inversion under the coupling condition of an acoustic medium and an elastic medium.

Description

Technical Field

[0001] This invention belongs to the field of exploration geophysics research technology, specifically relating to an acoustic-elastic coupling medium parameter inversion method and system. Background Technology

[0002] Marine seismic exploration is a crucial technique in offshore oil and gas resource exploration. In typical marine seismic exploration scenarios, seismic waves are usually generated by an air gun source located at the sea surface or in the seawater. They initially propagate as acoustic waves within the seawater layer. Upon reaching the seabed interface, part of the wave field is reflected, while the rest is transmitted into the seabed strata and continues to propagate as elastic waves in the elastic medium of the seabed. Therefore, the subsurface media encountered in offshore oil and gas exploration typically exhibit significant acoustic-elastic coupling characteristics. How to invert medium parameters such as P-wave velocity, S-wave velocity, and density under conditions of acoustic and elastic medium coupling is a key issue in marine seismic data interpretation and reservoir identification.

[0003] Currently, a series of studies have been conducted on the seismic inversion problem in acoustic-elastic coupled media. Traditional methods typically employ frequency-domain finite element methods, finite difference methods, Laplace domain waveform inversion, Gauss-Newton conjugate gradient methods, or full waveform inversion methods to handle the forward and inversion problems of acoustic-elastic coupled media. These methods usually require domain decomposition for the acoustic and elastic regions and setting coupling boundary conditions at the interface between the acoustic and elastic media. Therefore, in practical applications, traditional acoustic-elastic coupled media inversion methods usually require obtaining a relatively accurate location of the seafloor interface beforehand. However, in complex seafloor topography, salt domes, highly heterogeneous strata, or deep-water exploration scenarios, the location of the seafloor interface or acoustic-elastic coupling interface is often unclear. Predetermining this interface increases the modeling difficulty and may propagate interface location errors to subsequent medium parameter inversion results.

[0004] Furthermore, traditional numerical inversion methods, such as full-waveform inversion, are typically highly dependent on the initial model. When the initial velocity or density model differs significantly from the actual subsurface medium, the inversion process is prone to getting trapped in local extrema, leading to unstable inversion results for parameters such as P-wave velocity, S-wave velocity, and density. This is especially true in offshore oil and gas exploration, where deep geological structures are complex, observation apertures are limited, and low-frequency information is insufficient, making it difficult to accurately establish an initial model. This further limits the applicability of traditional inversion methods in complex acoustic-elastic coupling media.

[0005] With the improvement of data acquisition capabilities and computing platform performance, machine learning methods are increasingly being used for seismic parameter inversion. For example, existing methods utilize deep neural networks, convolutional neural networks, or semi-supervised learning models to establish a nonlinear mapping relationship between seismic data and velocity models. However, traditional machine learning inversion methods typically rely on a large amount of sample data for training, and the training samples need to cover different geological structures and media parameter distributions. In offshore oil and gas exploration scenarios, the cost of acquiring large-scale labeled training data is high, and there may be geological differences between the actual work area and the training samples, which limits the generalization ability of the network model. At the same time, purely data-driven models usually do not explicitly introduce the acoustic-elastic coupled medium wave equation, making it difficult to ensure that the inversion results satisfy the physical laws of wave field propagation.

[0006] To reduce reliance on large amounts of labeled data, fusion physical mechanism neural networks have been proposed and are increasingly being applied to geophysical problems. These methods are typically based on automatic differentiation techniques, incorporating the residuals of partial differential equations as physical constraints into the neural network's loss function. This allows the network training process to be constrained by both observational data and physical equations. Existing fusion physical mechanism neural network methods have been used for acoustic equation forward modeling, velocity field inversion, travel-time tomography, and acoustic medium parameter inversion based on surface seismic records. However, existing methods largely focus on single acoustic media or inversion scenarios based on second-order acoustic equations, failing to adequately consider the acoustic-elastic coupling problem in marine seismic exploration where both seawater acoustic media and seabed elastic media coexist.

[0007] Furthermore, in actual marine seismic observations, the data acquired by receivers is typically dominated by particle velocity components, while data such as stress, pressure, displacement, or complete wavefield snapshots are not readily available. Existing methods that integrate physical mechanisms with neural networks may not be compatible with the types of observation data that are actually available if they directly use governing equations containing stress or pressure variables to construct physical constraints. This leads to difficulties in unifying the observation terms and physical constraint terms in the loss function. Especially in acoustic-elastic coupled media, if the wave equation residuals cannot be reconstructed based on the particle velocity observation form, it is difficult to simultaneously constrain the wavefield propagation law and the medium parameter inversion process using only particle velocity observation data.

[0008] Furthermore, the complex wave characteristics of high-frequency wavefields, reflected waves, transmitted waves, and near acoustic-elastic coupling interfaces in marine seismic records place high demands on the expressive power of neural networks. Ordinary fully connected neural networks tend to preferentially fit low-frequency components when representing wavefields containing high-frequency components or those with drastic local changes, leading to insufficient representation of complex wavefield details and consequently affecting the accuracy of parameter inversion based on wavefield residuals and observational data errors.

[0009] In summary, existing technologies suffer from at least the following problems: First, traditional acoustic-elastic coupling medium inversion methods typically require prior knowledge of the seabed interface location, making them unsuitable for offshore oil and gas exploration scenarios where the acoustic-elastic coupling interface is unknown. Second, traditional numerical inversion methods are heavily reliant on initial velocity or density models, resulting in insufficient inversion stability under complex geological conditions. Third, traditional machine learning methods rely on large amounts of labeled training data and lack constraints from acoustic-elastic coupling wave equations. Fourth, existing methods incorporating physical mechanism neural networks are mostly designed for single acoustic media and are difficult to directly apply to the inversion of coupled media where acoustic and elastic media coexist. Fifth, existing methods struggle to construct physical constraints that match the form of the observed data when using only particle velocity observation data. Sixth, the ability of ordinary fully connected neural networks to represent complex wave fields and high-frequency information still needs improvement. Therefore, a method and system are needed that can jointly invert P-wave velocity, S-wave velocity, and density under acoustic-elastic coupling conditions by combining particle velocity observation data, physical equation constraints, and well logging data. Summary of the Invention

[0010] The technical problem to be solved by the present invention is to provide an acoustic-elastic coupling medium parameter inversion method and system to address the shortcomings of the prior art. This method solves the technical problems that existing acoustic-elastic coupling medium inversion methods usually rely on the known location of the seabed interface, multiple types of observation data or a large amount of training data, and are difficult to simultaneously constrain the wave field propagation laws of the acoustic medium and the elastic medium and invert the P-wave velocity, S-wave velocity and density when only particle velocity observation data are available.

[0011] The present invention adopts the following technical solution: The method for inverting parameters of acoustic-elastic coupling media includes the following steps: S1. Determine the unified wave equation of the acoustic-elastic coupling medium in the offshore oil and gas exploration area, express the elastic wave equation in the elastic medium as a first-order velocity-stress form, and degenerate the first-order velocity-stress form into the acoustic wave equation in the acoustic medium when the shear modulus is zero. S2. Determine the fusion physical mechanism neural network structure, which includes two independent fully connected neural networks: a fully connected neural network for representing particle velocity and a fully connected neural network for representing model parameters. The fully connected neural network for representing particle velocity outputs the horizontal and vertical particle velocity components based on the spatial location, time, and source location. The fully connected neural network for representing model parameters outputs the P-wave velocity, S-wave velocity, and density at the corresponding spatial location based on the spatial location. S3. Determine the loss function and training method based on step S2, wherein the horizontal and vertical particle velocity components collected by the receiver are obtained as particle velocity observation data, the set of residual points for physical constraints is determined, and well logging data is obtained. Based on the unified wave equation, a wave equation residual containing only particle velocity and model parameters is constructed, and a loss function is constructed according to the particle velocity observation data, the wave equation residual, and the well logging data. The weight parameters and bias parameters of the fully connected neural network used to represent particle velocity and the fully connected neural network used to represent model parameters are updated by iteratively optimizing the loss function. S4. Based on step S3, obtain the parameter inversion results of the acoustic-elastic coupling medium. The parameter inversion results include the parameter distributions of P-wave velocity, S-wave velocity, and density in the offshore oil and gas exploration area.

[0012] Preferably, in step S1, the first-order velocity-stress form is expressed as:

[0013] in, and Represents two-dimensional spatial coordinates, and t represents time. This represents the horizontal velocity component of a particle. This represents the vertical velocity component of the particle. These represent the components of the stress tensor. Indicates the density of the medium. and This refers to the Lamé coefficient in elastic dynamics theory. This represents the physical strength term in the time domain.

[0014] Preferably, the P-wave velocity, S-wave velocity, Lamé parameter, and dielectric density satisfy the following relationship:

[0015] when When =0, =0, the first-order velocity-stress form degenerates into the acoustic wave equation in the acoustic medium as follows:

[0016] in, Indicates P-wave velocity. This indicates the S-wave velocity.

[0017] Preferably, in step S2, the input of the fully connected neural network used to represent the particle velocity includes spatial location, time, and source location, and the output layer includes two neurons, which are used to represent the horizontal particle velocity component and the vertical particle velocity component, respectively; the input of the fully connected neural network used to represent the model parameters is spatial location, and the output is the P-wave velocity, S-wave velocity, and density at the corresponding spatial location.

[0018] Preferably, the input coordinates of the fully connected neural network used to represent the particle velocity are obtained by Fourier feature mapping to obtain a high-dimensional feature vector, and then input into the fully connected neural network used to represent the particle velocity. The input coordinates include spatial location, time, and source location, and the number of Fourier basis functions in the Fourier feature mapping is 20.

[0019] Preferably, in step S3, when constructing the wave equation residual, the first two particle velocity equations in the first-order velocity-stress form are differentiated with respect to time, and the last three stress evolution equations in the first-order velocity-stress form are substituted into the particle velocity equations after time differentiation to eliminate stress variables and obtain the wave equation residual containing only particle velocities and model parameters.

[0020] Preferably, the loss function This includes observation data error terms, physical constraint terms, and well logging data constraint terms, specifically:

[0021] in, This represents the weighting coefficient of the physical constraint term. Represents the residual term of the wave equation. This represents the error term in the observed data. This indicates the error term in the well logging data. The weighting coefficients represent the error term in the observed data. This represents the weighting coefficient of the measurement data error term. This represents the set of weights and bias parameters of a neural network that integrates physical mechanisms. Represents the set of residual points. Represents the set of observed data points. This represents the set of well logging data points.

[0022] Preferably, the observation data error term is determined based on the error between the particle velocity observation data and the particle velocity components output by the fully connected neural network used to represent the particle velocity; the physical constraint term is determined based on the residual of the wave equation; and the well logging data constraint term is determined based on the error between the P-wave velocity, S-wave velocity, and density in the well logging data and the P-wave velocity, S-wave velocity, and density output by the fully connected neural network used to represent the model parameters.

[0023] Preferably, in step S3, when selecting training data points from the data points corresponding to the observed seismic records, 80% of the data points are randomly selected from the seismic wave region, and 20% of the data points are randomly selected from the remaining regions; and the ADAM optimization algorithm is used to iteratively minimize the loss function to update the weight parameters and bias parameters of the fully connected neural network used to represent the particle velocity and the fully connected neural network used to represent the model parameters.

[0024] Secondly, embodiments of the present invention provide an acoustic-elastic coupling medium parameter inversion system, comprising: The equation module is used to determine the unified wave equation of the acoustic-elastic coupling medium in the offshore oil and gas exploration area. It expresses the elastic wave equation in the elastic medium as a first-order velocity-stress form, and degenerates the first-order velocity-stress form into the acoustic wave equation in the acoustic medium when the shear modulus is zero. A network module is used to determine the fusion physical mechanism neural network structure, which includes two independent fully connected neural networks: a fully connected neural network for representing the velocity of a particle and a fully connected neural network for representing model parameters. The training module is used to acquire particle velocity observation data, residual point set and well logging data, construct wave equation residual containing only particle velocity and model parameters based on the unified wave equation, and construct loss function according to the particle velocity observation data, the wave equation residual and the well logging data. The weight parameters and bias parameters of the fully connected neural network used to represent particle velocity and the fully connected neural network used to represent model parameters are updated by iteratively optimizing the loss function. The parameter inversion module is used to obtain the parameter inversion results of the acoustic-elastic coupling medium based on the fully connected neural network used to represent the model parameters after training. The parameter inversion results include the parameter distributions of P-wave velocity, S-wave velocity and density in the offshore oil and gas exploration area.

[0025] Thirdly, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described acoustic-elastic coupling medium parameter inversion method.

[0026] Fourthly, embodiments of the present invention provide a computer-readable storage medium including a computer program, which, when executed by a processor, implements the steps of the above-described acoustic-elastic coupling medium parameter inversion method.

[0027] Fifthly, a chip includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described acoustic-elastic coupling medium parameter inversion method.

[0028] In a sixth aspect, embodiments of the present invention provide an electronic device, including a computer program, which, when executed by the electronic device, implements the steps of the above-described acoustic-elastic coupling medium parameter inversion method.

[0029] Compared with the prior art, the present invention has at least the following beneficial effects: An acoustic-elastic coupling medium parameter inversion method is proposed. It establishes the physical connection between the seawater acoustic medium and the seabed elastic medium through a unified wave equation. Then, it uses a fusion physical mechanism neural network to characterize the particle velocity and model parameters respectively. The loss function is constructed by combining particle velocity observation data, wave equation residuals and well logging data, so that the network training process is simultaneously constrained by observation data and physical equations. Finally, it outputs the parameter distributions of P-wave velocity, S-wave velocity and density.

[0030] Furthermore, the first-order velocity-stress form contains both particle velocity components and stress tensor components, which can describe the propagation relationship of longitudinal and transverse waves in elastic media. Under this equation form, by adjusting the shear modulus, the equations for elastic media can be made to correspond to those for acoustic media, providing an equational basis for the subsequent construction of unified physical constraints.

[0031] Furthermore, the P-wave velocity is determined by both the Lamé parameter and density, while the S-wave velocity is determined by the shear modulus and density; when the shear modulus is zero, the S-wave velocity is zero, and the medium behaves as an acoustic medium. Using this parametric relationship, the P-wave velocity, S-wave velocity, and density output by the neural network can be substituted into the residuals of the wave equation for training.

[0032] Furthermore, the inputs and outputs of two independent fully connected neural networks are defined to represent state variables and model parameters, respectively. The fully connected neural network representing particle velocities takes spatial location, time, and source location as input, and outputs horizontal and vertical particle velocity components to describe the wavefield's state as a function of time and the source. The fully connected neural network representing model parameters takes spatial location as input and outputs P-wave velocity, S-wave velocity, and density to describe the spatial distribution of medium parameters. This separation allows state variable inversion and model parameter inversion to be trained collaboratively under the same loss function.

[0033] Furthermore, Fourier feature mapping maps input coordinates such as spatial location, time, and source location to a high-dimensional feature space containing different frequency components, enabling fully connected neural networks to include richer wave field variation information when representing particle velocity components, thereby providing a more complete state variable expression for the observation data error term and physical constraint term in the loss function.

[0034] Furthermore, by differentiating the first two particle velocity equations in the first-order velocity-stress form with respect to time and substituting the latter three stress evolution equations into them, stress variables can be eliminated, yielding a wave equation residual containing only particle velocities and model parameters. This residual can be directly calculated from the neural network output and its automatic differentiation results, and used as a physical constraint term in the loss function.

[0035] Furthermore, the observation data error term is used to ensure that the particle velocity components output by the network are consistent with the particle velocity observation data collected by the receiver; the physical constraint term is used to ensure that the network output satisfies the acoustic-elastic coupling medium wave equation; and the well logging data constraint term is used to ensure that the P-wave velocity, S-wave velocity, and density output by the network are consistent with the well logging data. Through weighted combination, a unified neural network training objective can be formed.

[0036] Furthermore, the observation data error term originates from the difference between the observed particle velocity data and the output of the particle velocity network; the physical constraint term originates from the residuals of the wave equation; and the well logging data constraint term originates from the differences between the P-wave velocity, S-wave velocity, and density in the well logging data and the output of the model parameter network. This ensures that each term of the loss function has a clear data source and physical meaning.

[0037] Furthermore, 80% of the data points are randomly selected from the seismic wave region, and 20% of the data points are randomly selected from the remaining regions, so that the training samples cover both the main wavefield region and non-main wavefield region. The ADAM optimization algorithm is used to iteratively minimize the loss function, and the weight parameters and bias parameters of the fully connected neural network used to represent the particle velocity and the fully connected neural network used to represent the model parameters can be updated according to the gradient information.

[0038] It is understood that the beneficial effects of the second to sixth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0039] In summary, this invention unifies the wave equations of acoustic and elastic media into a first-order velocity-stress form, and combines a fusion physical mechanism neural network, particle velocity observation data, well logging data, and wave equation residuals for joint training. Under acoustic-elastic coupling conditions, it inverts P-wave velocity, S-wave velocity, and density parameters, providing a technical means for modeling medium parameters in offshore oil and gas exploration.

[0040] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0041] Figure 1 To integrate the physical mechanism neural network structure diagram; Figure 2 The inversion results of the salt dome model are shown, where (a) represents the actual parameters (P-wave velocity, S-wave velocity, and density), and (b) represents the inversion parameter field. Figure 3 To integrate earthquake records predicted by physical mechanism neural networks, where (a) is the actual earthquake record ( and (b) represents the predicted earthquake record, and (c) represents the differential profile; Figure 4 This is a schematic diagram of the process of the present invention; Figure 5 A schematic diagram of a computer device provided in an embodiment of the present invention; Figure 6 This is a block diagram of a chip provided according to an embodiment of the present invention.

[0042] Among them, 60. Computer equipment; 61. Processor; 62. Memory; 63. Computer program; 600. Electronic device; 610. Processing unit; 620. Storage unit; 6201. Random access memory unit; 6202. Cache memory unit; 6203. Read-only memory unit; 6204. Program / utility; 6205. Program module; 630. Bus; 640. Display unit; 650. Input / output interface; 660. Network adapter; 700. External device. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0045] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0046] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0047] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0048] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0049] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0050] This invention provides a method for inverting parameters of acoustic-elastic coupling media. It unifies the acoustic wave equation and the elastic wave equation into a first-order velocity-stress form, embedding this form as a physical constraint regularization term into the neural network loss function. Simultaneously, it uses only particle vibration velocities as observation data to train the neural network that integrates physical mechanisms, reducing dependence on multiple types of observation data. To improve the neural network's ability to represent high-frequency wave fields and complex wave characteristics, this invention introduces Fourier feature maps into the neural network representing particle vibration velocities. During the inversion process, the initial model is randomly generated by the initial weights and biases of the neural network, avoiding strong dependence on a given initial velocity model. Numerical experimental results show that, even when the location of the acoustic-elastic coupling medium interface is unknown, this method can accurately predict P-wave velocity, S-wave velocity, and density parameters, and can identify the acoustic-elastic medium interface, i.e., the seabed location. This invention provides a new technical approach for inverting parameters of complex acoustic-elastic coupling media, improving the accuracy, stability, and applicability of the inversion.

[0051] Inspired by research progress in fusion physical mechanism neural networks, this invention focuses on the problem of model parameter inversion in acoustic-elastic coupling media, and studies the feasibility and effectiveness of parameter inversion in acoustic-elastic coupling media based on fusion physical mechanism neural networks. Compared with existing methods, this invention has at least the following characteristics: First, this invention uses only particle velocity as observation data to train a physical information neural network, and for this form of observation data, re-derives the physical constraint relationship embedded in the loss function from the elastic wave equation; Second, the present invention introduces Fourier feature mapping into the neural network used to represent particle velocity in order to improve the accuracy of the neural network in representing complex wave fields and high-frequency information. Third, even when the location on the seabed, i.e. the interface of the acoustic-elastic coupling medium, is unknown, the present invention can still predict P-wave velocity, S-wave velocity, and density parameters relatively accurately, and reduce the dependence of the inversion results on the initial model.

[0052] Please see Figure 4 This invention provides a method for inverting parameters of an acoustic-elastic coupling medium. This method addresses scenarios where seawater acoustic media and seabed elastic media coexist in offshore oil and gas exploration. By unifying the acoustic wave equation in the acoustic medium and the elastic wave equation in the elastic medium, and using the unified wave equation as a physical constraint in the training process of a neural network that integrates physical mechanisms, the method obtains the parameter distributions of P-wave velocity, S-wave velocity, and medium density in the offshore oil and gas exploration area.

[0053] The method includes the following steps: S1. Determine the unified wave equation for acoustic-elastic coupling media in offshore oil and gas exploration areas; In offshore oil and gas exploration scenarios, seismic waves are typically generated by an air gun source located on the sea surface or in the seawater. They initially propagate as sound waves within the seawater layer. When the wave field reaches the seabed interface, part of the wave field is reflected, while the rest is transmitted into the elastic medium of the seabed, where it continues to propagate as elastic waves. Therefore, the medium in offshore oil and gas exploration areas can include both acoustic and elastic media, forming an acoustic-elastic coupling medium.

[0054] In one embodiment of the present invention, a wave control equation in a two-dimensional elastic medium is first constructed, and the elastic wave equation in the elastic medium is expressed in a first-order velocity-stress form. The first-order velocity-stress form includes a particle velocity equation and a stress evolution equation, wherein the particle velocity equation describes the change of the particle velocity components with time, and the stress evolution equation describes the change of the stress tensor components with time.

[0055] The general form of the two-dimensional elastic wave equation is expressed as follows:

[0056] in, , Represents a two-dimensional spatial region; It represents the components of a particle's velocity in the horizontal and vertical directions; These represent the components of the stress tensor, respectively. Indicates the density of the medium; and is the Lamé coefficient in elastic dynamics theory, where Shear modulus, used to characterize the proportional relationship between shear stress and shear strain. Related to bulk modulus; This represents the physical activity term in the time domain. P-wave velocity and S-wave velocity are defined as follows:

[0057] When shear modulus Sometimes, .make The elastic wave equation can then be transformed into the following sound wave equation:

[0058] in, For pressure.

[0059] Therefore, this invention unifies the acoustic wave equation in acoustic media and the elastic wave equation in elastic media into a first-order velocity-stress elastic medium wave equation. When the shear modulus... In this case, the wave equation of the elastic medium can degenerate into the acoustic wave equation. The present invention aims to use observation data deployed at specific locations to invert the model parameters in the acoustic-elastic coupling medium, wherein the model parameters include the Lamé parameters or their equivalents, the P-wave velocity, the S-wave velocity, and the medium density.

[0060] Through the above processing, this embodiment does not require establishing completely independent control equations for the seawater acoustic medium and the seabed elastic medium, but instead incorporates the acoustic medium and the elastic medium into a unified wave equation framework, providing a foundation for the subsequent construction of physical constraint terms in the fusion physical mechanism neural network.

[0061] S2. Determine the structure of the neural network that integrates physical mechanisms; Since in practical applications, particle velocity components can usually only be obtained through multi-component detectors, the purpose of this invention is to obtain particle velocity components based on observations. and The P-wave velocity, S-wave velocity, and density are estimated.

[0062] After determining the unified wave equation for the acoustic-elastic coupling medium, a fusion physical mechanism neural network structure is constructed. This fusion physical mechanism neural network structure comprises two independent fully connected neural networks: one for representing particle velocities and the other for representing model parameters.

[0063] The fully connected neural network used to represent particle velocities is employed to characterize the wavefield state variables in the acoustic-elastic coupling medium. The inputs to this fully connected neural network include spatial location, time, and source location, while the outputs include horizontal and vertical particle velocity components. In other words, for a given spatial location, time, and source location, the fully connected neural network outputs two particle velocity components at the corresponding location and time.

[0064] In one embodiment, the seismic source is located on the Earth's surface or in the seawater layer, and the location of the seismic source can be represented by its horizontal coordinates; when it is necessary to describe the depth of the seismic source simultaneously, the depth of the seismic source can also be used as part of the input coordinates. The output layer of the fully connected neural network used to represent the particle velocity includes two neurons, corresponding to the horizontal particle velocity component and the vertical particle velocity component, respectively.

[0065] Furthermore, since marine seismic records contain complex wavefield information, including direct waves, reflected waves, transmitted waves, and the area near the acoustic-elastic coupling interface, ordinary fully connected neural networks may be insufficient in representing high-frequency wavefields and locally drastically changing wavefields. Therefore, in this embodiment, the input coordinates of the fully connected neural network used to represent particle velocity are processed through Fourier feature mapping to obtain a high-dimensional feature vector, which is then input into the fully connected neural network used to represent particle velocity.

[0066] The input coordinates include spatial location, time, and source location; the Fourier feature mapping maps the input coordinates into a high-dimensional feature vector containing different frequency components based on the wavenumber parameters in the Fourier basis functions. In one embodiment, the number of Fourier basis functions is 20. With this setting, the fully connected neural network used to represent particle velocity can output particle velocity components based on the mapped high-dimensional feature vector.

[0067] Fully connected neural networks used to represent model parameters characterize the medium parameters to be inverted. Unlike fully connected neural networks used to represent particle velocities, the input to a fully connected neural network used to represent model parameters is a spatial location, and the output is the P-wave velocity, S-wave velocity, and medium density at that location. In other words, for a given spatial location, this fully connected neural network outputs the model parameters at that location.

[0068] By setting up two independent fully connected neural networks, the state variables and model parameters are represented by different networks, enabling the particle velocity components and medium parameters to be trained and updated collaboratively under the same loss function constraint.

[0069] First, the state variables This is represented as the output of a fully connected neural network, i.e. The input to the neural network includes spatial location. ,time and the location of the epicenter .

[0070] It should be noted that in this invention, the seismic source is located on the Earth's surface, therefore the source depth satisfies... The location of the seismic source may be determined using only the horizontal coordinates of the seismic source. Indicates. Parameters This represents the weights and bias parameters of the neural network. The output layer of this neural network consists of two neurons, each representing a particle velocity component. and .

[0071] Since the observation data in this invention are seismic records received from the Earth's surface, the information contained in these seismic records is relatively sparse. To improve the neural network... To improve the ability to fit wave fields and alleviate the spectral bias problem during neural network training—where low-frequency components are preferentially fitted while high-frequency components are difficult to represent—this invention introduces Fourier feature mapping. Based on this Fourier feature mapping, the input coordinates of the neural network are... Mapped to high-order feature vectors:

[0072] Where Φ represents the wavenumber parameter in the Fourier basis, which is used to control the frequency component in the Fourier eigenmap.

[0073] Furthermore, the present invention expresses the model parameters to be inverted as follows: And using neural networks To characterize it, that is .in, This represents the weights and bias parameters of the neural network. Unlike neural networks used to characterize the velocity of a particle, the input to this model parameter neural network is only spatial position. It is used to output the P-wave velocity, S-wave velocity, and density at this spatial location.

[0074] S3. Determine the loss function based on step S2 and the training method for the network; After determining the structure of the fusion physical mechanism neural network, the data and sampling points required for training are obtained, and a loss function is constructed. To train the physical information neural network, this invention constructs a loss function that includes an observation data error term and a physical constraint term.

[0075] First, in the entire computational domain Randomly select several points as residual points, and denote the set of residual points as follows: At the residual point, the state variable predicted by the neural network... and model parameters The wave equation constraint should be satisfied. This wave equation constraint, as a regularization term in the loss function, is used to ensure that the neural network output conforms to the physical laws of acoustic-elastic coupled wave propagation, and to penalize predicted state variables and model parameters that do not satisfy the wave equation.

[0076] Observational earthquake records As training data, it is used to constrain the state variables predicted by the neural network. And assist in predicting model parameters The sampling points corresponding to the observed seismic records are called data points, and their set is denoted as . .

[0077] In this embodiment, the observed seismic records are acquired by receivers deployed on the Earth's surface, therefore the reception depth is... For observed seismic records corresponding to a specific epicenter, the data is located in... Within the region; due to seismic waves in The seismic wavefield information is sparsely distributed within the seismic domain. To increase the proportion of effective wavefield information in the training data, this invention employs an effective sampling strategy to select data points: 80% of the data points are randomly selected from the region where the seismic waves are located, and 20% of the data points are randomly selected from the remaining regions. The region where the seismic waves are located is defined as follows:

[0078] To alleviate the ill-posedness of the seismic inversion problem, this invention further introduces well logging data as constraint information during the training process.

[0079] The logging data is used to provide direct observations of medium parameters at the borehole location, including P-wave velocity, S-wave velocity, and density. The sampling points corresponding to the logging data are called logging data points, and their set is denoted as [details omitted]. By introducing well logging data constraints, the stability and reliability of the model parameter inversion results can be improved.

[0080] Based on the above definition, this invention constructs the physical information neural network loss function for seismic inversion of acoustic-elastic coupled media as a weighted sum of multiple loss terms, wherein each loss term adopts... The loss is measured in norm form. The loss function is defined as:

[0081] in, and

[0082] in, ,parameter , and This can be obtained through an adaptive method. Among them, This represents the residual of the wave equation.

[0083] It should be noted that this invention only considers the particle velocity component. and As observational data.

[0084] To eliminate the stress variable in the first-order elastic medium wave equation, this invention differentiates the first two equations of the first-order elastic medium wave equation with respect to time, and substitutes the last three equations of the first-order elastic medium wave equation into the first two equations, thereby reconstructing a wave equation residual containing only particle velocity and medium parameters. .

[0085] The residual is expressed as:

[0086] During the training phase, this invention employs the ADAM optimization algorithm to iteratively minimize the loss function in order to update the neural network. and The weight parameters and bias parameters in the network. The parameters of the neural network are denoted as...

[0087]

[0088] S4. Based on step S3, obtain the parameter inversion results of the acoustic-elastic coupling medium; verify the inversion capability and effectiveness of this method through the physical property parameters and seismic record results output by the network.

[0089] In one embodiment of the present invention, two independent fully connected neural networks are used to represent the state variables respectively. and model parameters Among them, the neural network used to represent state variables It consists of 5 hidden layers, each containing 80 neurons, with the tanh activation function; a neural network used to represent model parameters. It includes 5 hidden layers, each containing 40 neurons, with the activation function being the sine function. During training, the learning rate decays exponentially from 0.001 to 0.00001, the batch size is set to 4096, and the number of training epochs is set to 20000.

[0090] The neural network representing the parameters to be inverted is trained on the network. The inversion result is compared with the actual velocity model to verify the effectiveness and practicality of the invention.

[0091] In another embodiment of the present invention, an acoustic-elastic coupling medium parameter inversion system is provided. This system can be used to implement the above-mentioned acoustic-elastic coupling medium parameter inversion method. Specifically, the acoustic-elastic coupling medium parameter inversion system includes an equation module, a network module, a training module, and a parameter inversion module.

[0092] Among them, the equation module is used to determine the unified wave equation of the acoustic-elastic coupling medium in the offshore oil and gas exploration area, express the elastic wave equation in the elastic medium as a first-order velocity-stress form, and degenerate the first-order velocity-stress form into the acoustic wave equation in the acoustic medium when the shear modulus is zero. A network module is used to determine the fusion physical mechanism neural network structure, which includes two independent fully connected neural networks: a fully connected neural network for representing the velocity of a particle and a fully connected neural network for representing model parameters. The training module is used to acquire particle velocity observation data, residual point set and well logging data, construct wave equation residual containing only particle velocity and model parameters based on the unified wave equation, and construct loss function according to the particle velocity observation data, the wave equation residual and the well logging data. The weight parameters and bias parameters of the fully connected neural network used to represent particle velocity and the fully connected neural network used to represent model parameters are updated by iteratively optimizing the loss function. The parameter inversion module is used to obtain the parameter inversion results of the acoustic-elastic coupling medium based on the fully connected neural network used to represent the model parameters after training. The parameter inversion results include the parameter distributions of P-wave velocity, S-wave velocity and density in the offshore oil and gas exploration area.

[0093] This invention provides a terminal device comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or corresponding function. The processor described in this embodiment can be used for the operation of an acoustic-elastic coupling medium parameter inversion method, including: A unified wave equation for acoustic-elastic coupled media in offshore oil and gas exploration areas is determined. The elastic wave equation in the elastic medium is expressed as a first-order velocity-stress form, and when the shear modulus is zero, this first-order velocity-stress form degenerates into the acoustic wave equation in the acoustic medium. A fusion physical mechanism neural network structure is determined, comprising two independent fully connected neural networks: one for representing particle velocity and one for representing model parameters. The fully connected neural network for representing particle velocity outputs horizontal and vertical particle velocity components based on spatial location, time, and source location. The fully connected neural network for representing model parameters outputs the P-wave velocity, S-wave velocity, and [other components] at the corresponding spatial location. Density; Determine the loss function and training method, wherein, the horizontal and vertical particle velocity components collected by the receiver are acquired as particle velocity observation data, the set of residual points for physical constraints is determined, and well logging data is acquired, a wave equation residual containing only particle velocity and model parameters is constructed based on the unified wave equation, and a loss function is constructed based on the particle velocity observation data, the wave equation residual, and the well logging data, and the weight parameters and bias parameters of the fully connected neural network used to represent particle velocity and the fully connected neural network used to represent model parameters are updated by iteratively optimizing the loss function; Based on the obtained parameter inversion results of the acoustic-elastic coupling medium, the parameter inversion results include the parameter distribution of P-wave velocity, S-wave velocity, and density of the offshore oil and gas exploration area.

[0094] Please see Figure 5 The terminal device is a computer device. In this embodiment, the computer device 60 includes a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When executed by the processor 61, the computer program 63 implements the acoustic-elastic coupling medium parameter inversion method of this embodiment. To avoid repetition, details are omitted here. Alternatively, when executed by the processor 61, the computer program 63 implements the functions of each model / unit in the acoustic-elastic coupling medium parameter inversion system of this embodiment. To avoid repetition, details are omitted here.

[0095] Computer device 60 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. Computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will understand that... Figure 5 This is merely an example of computer device 60 and does not constitute a limitation on computer device 60. It may include more or fewer components than shown, or combine certain components, or different components. For example, computer device may also include input / output devices, network access devices, buses, etc.

[0096] The processor 61 may be a Central Processing Unit (CPU), or other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0097] The memory 62 can be an internal storage unit of the computer device 60, such as a hard disk or memory of the computer device 60. The memory 62 can also be an external storage device of the computer device 60, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the computer device 60.

[0098] Furthermore, the memory 62 may include both internal storage units of the computer device 60 and external storage devices. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 can also be used to temporarily store data that has been output or will be output.

[0099] Please see Figure 6 The terminal device is an electronic device 600, which is manifested in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including storage unit 620 and processing unit 610), a display unit 640, etc.

[0100] The storage unit stores program code, which can be executed by the processing unit 610 to perform the steps described in the method section of this specification according to various exemplary embodiments of the present invention. For example, the processing unit 610 can perform actions such as... Figure 4 The steps are shown in the figure.

[0101] Storage unit 620 may include readable media in the form of volatile storage units, such as random access memory (RAM) 6201 and / or cache memory 6202, and may further include read-only memory (ROM) 6203.

[0102] Storage unit 620 may also include a program / utility 6204 having a set (at least one) program module 6205, such program module 6205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0103] Bus 630 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the multiple bus structures.

[0104] Electronic device 600 can also communicate with one or more external devices 700 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 600, and / or with any device that enables electronic device 600 to communicate with one or more other computing devices (e.g., router, modem). This communication can be performed via input / output interface 650. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area network, wide area network, and / or public network, such as the Internet) via network adapter 660. Network adapter 660 can communicate with other modules of electronic device 600 via bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0105] This invention also provides a storage medium, specifically a computer-readable storage medium, which is a memory device in a terminal device for storing programs and data. It is understood that the computer-readable storage medium here can include both built-in storage media in the terminal device and extended storage media supported by the terminal device; it can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor, which can be one or more computer programs (including program code). More specific examples of the computer-readable storage medium include: an electrical connection with one or more wires, a portable disk, a hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical fiber, portable compact disk read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0106] Computer-readable storage media also include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium can also be any readable medium other than a readable storage medium that can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, radio frequency, etc., or any suitable combination thereof.

[0107] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0108] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the acoustic-elastic coupling medium parameter inversion method in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor in the following steps: A unified wave equation for acoustic-elastic coupled media in offshore oil and gas exploration areas is determined. The elastic wave equation in the elastic medium is expressed as a first-order velocity-stress form, and when the shear modulus is zero, this first-order velocity-stress form degenerates into the acoustic wave equation in the acoustic medium. A fusion physical mechanism neural network structure is determined, comprising two independent fully connected neural networks: one for representing particle velocity and one for representing model parameters. The fully connected neural network for representing particle velocity outputs horizontal and vertical particle velocity components based on spatial location, time, and source location. The fully connected neural network for representing model parameters outputs the P-wave velocity, S-wave velocity, and [other components] at the corresponding spatial location. Density; Determine the loss function and training method, wherein, the horizontal and vertical particle velocity components collected by the receiver are acquired as particle velocity observation data, the set of residual points for physical constraints is determined, and well logging data is acquired, a wave equation residual containing only particle velocity and model parameters is constructed based on the unified wave equation, and a loss function is constructed based on the particle velocity observation data, the wave equation residual, and the well logging data, and the weight parameters and bias parameters of the fully connected neural network used to represent particle velocity and the fully connected neural network used to represent model parameters are updated by iteratively optimizing the loss function; Based on the obtained parameter inversion results of the acoustic-elastic coupling medium, the parameter inversion results include the parameter distribution of P-wave velocity, S-wave velocity, and density of the offshore oil and gas exploration area.

[0109] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0110] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0111] Example This embodiment uses a salt dome model as an example to illustrate the acoustic-elastic coupling medium parameter inversion method provided by the present invention.

[0112] Please see Figure 1 , Figure 1 This is a schematic diagram illustrating the parameter inversion of an acoustic-elastic coupling medium based on a fusion of physical mechanisms. Two independent fully connected neural networks are used to represent the state variables. and model parameters Among them, the neural network used to represent state variables It consists of 5 hidden layers, each containing 80 neurons, with the tanh activation function; a neural network used to represent model parameters. It includes 5 hidden layers, each containing 40 neurons, with the activation function being the sine function.

[0113] Please see Figure 2 , Figure 2 (a) represents the actual parameters. Figure 2 (b) shows the inversion parameter field. The salt dome model is derived from a salt dome medium model of a typical tectonic local area in the North Sea region, with a water layer of 200m thickness set on top of the model. The model includes an acoustic medium region and an elastic medium region, with the acoustic medium region corresponding to the seawater layer and the elastic medium region corresponding to the seafloor strata.

[0114] In this embodiment, observational seismic records are generated using five different seismic sources. The sources are spaced horizontally at intervals ranging from 10m to 970m, with a distance of 240m between adjacent sources. The observational seismic records include particle velocity components. During training, 80,000 residual points are selected, and 40,000 data points are chosen from the observational seismic records. For each data point corresponding to an observational seismic record, 80% of the data points are randomly selected from the region containing the seismic wave, and 20% are randomly selected from the remaining regions.

[0115] Due to the complex structure of the salt dome medium model, this embodiment further sets multiple logging curves to obtain velocity and density information. The logging locations are 0.05km, 0.13km, 0.29km, 0.32km, 0.41km, 0.44km, 0.55km, and 0.92km. The logging data includes the P-wave velocity, S-wave velocity, and medium density at the corresponding logging locations.

[0116] During training, a loss function is constructed based on particle velocity observation data, residual point set, and well logging data. The loss function includes an observation data error term, a physical constraint term, and a well logging data constraint term. Specifically, the observation data error term constrains the particle velocity components output by the fully connected neural network representing particle velocity; the physical constraint term constrains the network output to satisfy the unified wave equation for acoustic-elastic coupling media; and the well logging data constraint term constrains the P-wave velocity, S-wave velocity, and medium density output by the fully connected neural network representing model parameters.

[0117] During the training phase, the ADAM optimization algorithm is used to iteratively minimize the loss function. The learning rate decays exponentially from 0.001 to 0.00001, the batch size is set to 4096, and the number of training epochs is set to 20000. Through iterative training, the weight and bias parameters of the fully connected neural network representing the particle velocity and the fully connected neural network representing the model parameters are updated.

[0118] After training, the spatial locations in the offshore oil and gas exploration area are input into a fully connected neural network to represent the model parameters, yielding the P-wave velocity, S-wave velocity, and medium density at the corresponding spatial locations. Performing this process on multiple spatial locations allows for the determination of the spatial distribution of P-wave velocity, S-wave velocity, and medium density.

[0119] Please see Figure 3 , Figure 3 This is an earthquake record predicted by integrating physical mechanism neural networks. Among them, Figure 3 (a) is a real earthquake record. Figure 3 (b) is the predicted earthquake record. Figure 3 (c) is the differential profile. Since the seismic source is located in the acoustic medium, the observed seismic records only include the particle velocity components. and The observed seismic records mainly contain P-wave responses. By comparing the predicted seismic records with the actual seismic records, the fit of the fully connected neural network used to represent particle velocities to the wavefield state variables can be examined; by comparing the inverted parameter field with the actual parameters, the inversion of P-wave velocities, S-wave velocities, and medium density by the fully connected neural network used to represent the model parameters can be examined.

[0120] As can be seen from the above embodiments, the present invention achieves the inversion of P-wave velocity, S-wave velocity and medium density parameters in acoustic-elastic coupling media by using a unified wave equation, two independent fully connected neural networks, wave equation residuals containing only particle velocity and model parameters, particle velocity observation data, well logging data and ADAM optimization algorithm.

[0121] In summary, this invention provides a method and system for inverting parameters in an acoustic-elastic coupled medium. By constructing a unified wave equation for the acoustic-elastic coupled medium, it incorporates the wave field propagation process in both seawater acoustic and seabed elastic media into the same mathematical framework, avoiding the model fragmentation problem caused by separately handling the acoustic and elastic regions. By setting up two independent fully connected neural networks to represent particle velocity and model parameters respectively, state variables and medium parameters such as P-wave velocity, S-wave velocity, and density can be updated collaboratively during the same training process. Addressing the situation in actual marine seismic exploration where only particle velocity observation data is typically available, this invention constructs a wave equation residual containing only particle velocity and model parameters by time-derived from the first-order velocity-stress equation and substituted into the stress evolution equation, thus ensuring that physical constraints match the form of the observation data. Furthermore, by combining well logging data constraints and the ADAM optimization algorithm, the network weight parameters and bias parameters can be iteratively updated to obtain the distribution of P-wave velocity, S-wave velocity, and density parameters in the marine oil and gas exploration area.

[0122] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0123] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0124] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0125] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0126] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0127] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0128] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random-access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0129] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0130] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0131] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0132] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for inverting parameters of an acoustic-elastic coupling medium, characterized in that, Includes the following steps: S1. Determine the unified wave equation of the acoustic-elastic coupling medium in the offshore oil and gas exploration area, express the elastic wave equation in the elastic medium as a first-order velocity-stress form, and degenerate the first-order velocity-stress form into the acoustic wave equation in the acoustic medium when the shear modulus is zero. S2. Determine the fusion physical mechanism neural network structure, which includes two independent fully connected neural networks: a fully connected neural network for representing particle velocity and a fully connected neural network for representing model parameters. The fully connected neural network for representing particle velocity outputs the horizontal and vertical particle velocity components based on the spatial location, time, and source location. The fully connected neural network for representing model parameters outputs the P-wave velocity, S-wave velocity, and density at the corresponding spatial location based on the spatial location. S3. Determine the loss function and training method based on step S2, wherein the horizontal and vertical particle velocity components collected by the receiver are obtained as particle velocity observation data, the set of residual points for physical constraints is determined, and well logging data is obtained. Based on the unified wave equation, a wave equation residual containing only particle velocity and model parameters is constructed, and a loss function is constructed according to the particle velocity observation data, the wave equation residual, and the well logging data. The weight parameters and bias parameters of the fully connected neural network used to represent particle velocity and the fully connected neural network used to represent model parameters are updated by iteratively optimizing the loss function. S4. Based on step S3, obtain the parameter inversion results of the acoustic-elastic coupling medium. The parameter inversion results include the parameter distributions of P-wave velocity, S-wave velocity, and density in the offshore oil and gas exploration area.

2. The method for inverting acoustic-elastic coupling medium parameters according to claim 1, characterized in that, In step S1, the first-order velocity-stress form is expressed as: in, and Represents two-dimensional spatial coordinates, and t represents time. This represents the horizontal velocity component of a particle. This represents the vertical velocity component of the particle. These represent the components of the stress tensor. Indicates the density of the medium. and This refers to the Lamé coefficient in elastic dynamics theory. This represents the physical strength term in the time domain.

3. The method for inverting acoustic-elastic coupling medium parameters according to claim 2, characterized in that, The P-wave velocity, S-wave velocity, Lamé parameter, and dielectric density satisfy the following relationship: when When =0, =0, the first-order velocity-stress form degenerates into the acoustic wave equation in the acoustic medium as follows: in, Indicates P-wave velocity. This indicates the S-wave velocity.

4. The method for inverting acoustic-elastic coupling medium parameters according to claim 1, characterized in that, In step S2, the input of the fully connected neural network used to represent the particle velocity includes spatial location, time, and source location. The output layer includes two neurons, which are used to represent the horizontal particle velocity component and the vertical particle velocity component, respectively. The input of the fully connected neural network used to represent the model parameters is spatial location, and the output is the P-wave velocity, S-wave velocity, and density at the corresponding spatial location.

5. The method for inverting acoustic-elastic coupling medium parameters according to claim 4, characterized in that, The input coordinates of the fully connected neural network used to represent the particle velocity are processed through Fourier feature mapping to obtain a high-dimensional feature vector, which is then input into the fully connected neural network used to represent the particle velocity. The input coordinates include spatial location, time, and source location. The number of Fourier basis functions in the Fourier feature mapping is 20.

6. The method for inverting acoustic-elastic coupling medium parameters according to claim 1, characterized in that, In step S3, when constructing the wave equation residual, the first two particle velocity equations in the first-order velocity-stress form are differentiated with respect to time, and the last three stress evolution equations in the first-order velocity-stress form are substituted into the particle velocity equations after time differentiation to eliminate stress variables and obtain the wave equation residual containing only particle velocities and model parameters.

7. The method for inverting acoustic-elastic coupling medium parameters according to claim 6, characterized in that, The loss function This includes observation data error terms, physical constraint terms, and well logging data constraint terms, specifically: in, This represents the weighting coefficient of the physical constraint term. Represents the residual term of the wave equation. This represents the error term in the observed data. This indicates the error term in the well logging data. The weighting coefficients represent the error term in the observed data. This represents the weighting coefficient of the measurement data error term. This represents the set of weights and bias parameters of a neural network that integrates physical mechanisms. Represents the set of residual points. Represents the set of observed data points. This represents the set of well logging data points.

8. The method for inverting acoustic-elastic coupling medium parameters according to claim 7, characterized in that, The observation data error term is determined based on the error between the particle velocity observation data and the particle velocity components output by the fully connected neural network used to represent the particle velocity; the physical constraint term is determined based on the residual of the wave equation; the well logging data constraint term is determined based on the error between the P-wave velocity, S-wave velocity, and density in the well logging data and the P-wave velocity, S-wave velocity, and density output by the fully connected neural network used to represent the model parameters.

9. The method for inverting parameters of an acoustic-elastic coupling medium according to claim 1, characterized in that, In step S3, when selecting training data points from the data points corresponding to the observed seismic records, 80% of the data points are randomly selected from the seismic wave region, and 20% of the data points are randomly selected from the remaining regions. Furthermore, the ADAM optimization algorithm is used to iteratively minimize the loss function to update the weight parameters and bias parameters of the fully connected neural network used to represent the particle velocity and the fully connected neural network used to represent the model parameters.

10. An acoustic-elastic coupling medium parameter inversion system, characterized in that, include: The equation module is used to determine the unified wave equation of the acoustic-elastic coupling medium in the offshore oil and gas exploration area. It expresses the elastic wave equation in the elastic medium as a first-order velocity-stress form, and degenerates the first-order velocity-stress form into the acoustic wave equation in the acoustic medium when the shear modulus is zero. A network module is used to determine the fusion physical mechanism neural network structure, which includes two independent fully connected neural networks: a fully connected neural network for representing the velocity of a particle and a fully connected neural network for representing model parameters. The training module is used to acquire particle velocity observation data, residual point set and well logging data, construct wave equation residual containing only particle velocity and model parameters based on the unified wave equation, and construct loss function according to the particle velocity observation data, the wave equation residual and the well logging data. The weight parameters and bias parameters of the fully connected neural network used to represent particle velocity and the fully connected neural network used to represent model parameters are updated by iteratively optimizing the loss function. The parameter inversion module is used to obtain the parameter inversion results of the acoustic-elastic coupling medium based on the fully connected neural network used to represent the model parameters after training. The parameter inversion results include the parameter distributions of P-wave velocity, S-wave velocity and density in the offshore oil and gas exploration area.

Images