Method, device and electronic equipment for determining the position of natural gas hydrate

Abstract

The present invention provides a method, device and electronic device for determining the position of natural gas hydrates, and relates to the technical field of seismic detection. After calculating the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the natural gas hydrate-containing formation based on the background velocity model and the background quality factor model of the natural gas hydrate-containing formation, the method uses the above two imaging data as physical principle constraints, and uses the target multi-scale convolution residual network model to process the background velocity model and the background quality factor model respectively, so as to obtain the target velocity model and the target quality factor model of the natural gas hydrate-containing formation, and then determines the position data of the natural gas hydrate in the natural gas hydrate-containing formation according to the target velocity model and the target quality factor model. Since the target multi-scale convolution residual network model can extract the multi-scale information of the image, the prediction result of the method of the present invention has higher accuracy and certain generalization ability.

Description

Technical Field

[0001] The present invention relates to the technical field of earthquake detection, and in particular to a method, device and electronic equipment for determining the position of natural gas hydrate. Background Art

[0002] Traditional migration imaging methods can only determine the position of the bottom interface of natural gas hydrate (NHG) through bottom simulating reflectors (BSR), and the position of the top interface is difficult to determine. It is known that velocity and quality factor are important attributes for identifying natural gas hydrates, which can provide more basis for determining the spatial distribution of natural gas hydrates. However, in the prior art, the method for inverting velocity and quality factor usually adopts a full waveform inversion method based on full data drive, but this inversion method ignores the forward operator and adjoint operator, and directly realizes the mapping from the shot gather domain (x, t) to the (x, z) domain (x represents the horizontal position, t represents the seismic travel time, and z represents the depth). Due to the spatial inconsistency between the input data and the output data, the quality of the prediction results is seriously dependent on the diversity of the training data, and the anti-noise and generalization capabilities are poor. Summary of the invention

[0003] The object of the present invention is to provide a method, device and electronic equipment for determining the position of natural gas hydrate, so as to alleviate the technical problem of low prediction accuracy existing in the existing method for determining the position of natural gas hydrate.

[0004] In a first aspect, the present invention provides a method for determining the position of a natural gas hydrate, comprising: obtaining a background velocity model and a background quality factor model of a natural gas hydrate-containing formation; simulating forward wavefield data of a preset source in the natural gas hydrate-containing formation based on the background velocity model, the background quality factor model and a viscoacoustic wave equation containing a fractional-order Laplace operator, and obtaining seismic forward simulation data; using the seismic forward simulation data as a reverse propagation source, simulating the reverse propagation of the reverse propagation source in the natural gas hydrate-containing formation, and obtaining reverse propagation wavefield data; calculating pre-stack reverse time migration imaging data and post-stack reverse time migration imaging data of the natural gas hydrate-containing formation based on the forward wavefield data and the reverse propagation wavefield data; processing the background velocity model and the background quality factor model based on the pre-stack reverse time migration imaging data, the post-stack reverse time migration imaging data and a target multi-scale convolutional residual network model, respectively, to obtain a target velocity model and a target quality factor model of the natural gas hydrate-containing formation; and determining the position data of the natural gas hydrate in the natural gas hydrate-containing formation based on the target velocity model and the target quality factor model.

[0005] In an optional embodiment, based on the background velocity model, the background quality factor model and the viscoacoustic wave equation containing fractional-order Laplace operators, the forward wavefield data of a preset seismic source in the natural gas hydrate-containing formation are simulated, including: decoupling the viscoacoustic wave equation containing fractional-order Laplace operators to obtain a viscoacoustic wave equation with an approximate constant Q; deforming the viscoacoustic wave equation with an approximate constant Q to obtain a fractional-order partial differential equation to which absorbing boundary conditions can be applied; and bringing the background velocity model, the background quality factor model and the preset seismic source into the fractional-order partial differential equation to obtain the forward wavefield data.

[0006] In an optional embodiment, based on the forward wavefield data and the back propagation wavefield data, the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the natural gas hydrate-containing formation are calculated, including: using Q compensation imaging conditions to perform Q compensation reverse time migration processing on the forward wavefield data and the back propagation wavefield data to obtain the pre-stack reverse time migration imaging data; and superimposing the pre-stack reverse time migration imaging data to obtain the post-stack reverse time migration imaging data.

[0007] In an optional embodiment, based on the pre-stack reverse time migration imaging data, the post-stack reverse time migration imaging data and the target multi-scale convolution residual network model, the background velocity model and the background quality factor model are processed respectively, including: based on the expression Iteratively invert each background velocity in the background velocity model to obtain a target velocity corresponding to each background velocity; wherein Λ represents the target multi-scale convolutional residual network model, represents the speed of the k1th iteration, k1=0,1,2,...N1-1, N1 represents the preset number of iterations of the speed, υ0=υ back , back represents the background velocity, I1 represents the pre-stack reverse time migration imaging data, I2 represents the post-stack reverse time migration imaging data, The regularization term representing the speed The gradient of Represents the target speed; using the expression Iteratively invert each background quality factor in the background quality factor model to obtain a target quality factor corresponding to each background quality factor; wherein, represents the quality factor of the k2th iteration, k2=0,1,2,...N2-1, N2 represents the preset number of iterations of the quality factor, Q0=Q back , Q back represents the background quality factor, The regularization term representing the quality factor The gradient of Represents the target quality factor.

[0008] In an optional embodiment, the forward wavefield data and the backpropagation wavefield data are subjected to Q compensation reverse time migration processing using the Q compensation imaging condition to obtain the prestack reverse time migration imaging data, including: using the formula Calculate the prestack reverse time migration imaging data; wherein, P S (r, t) represents the forward wave field data, P R (r, t) represents the reverse propagation wavefield data, r represents the spatial coordinate position, t represents the earthquake travel time, T represents the total length of the seismic forward simulation data sampling time, and I1 represents the pre-stack reverse time migration imaging data.

[0009] In an optional embodiment, the method also includes: obtaining a training sample set; wherein the training sample set includes multiple sample formation models containing natural gas hydrates; performing filtering processing on each sample formation model in the training sample set for a specified number of times to obtain a sample background model corresponding to each sample formation model; wherein the sample background model includes: a sample background velocity model and a sample background quality factor model; based on the target sample background model and the viscoacoustic wave equation containing a fractional-order Laplace operator, calculating the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the corresponding formation to obtain training pre-stack reverse time migration imaging data and training post-stack reverse time migration imaging data; wherein the target sample background model represents any model among the multiple sample background models; based on all sample background models and their corresponding sample formation models, training pre-stack reverse time migration imaging data and training post-stack reverse time migration imaging data, iteratively training the initial multi-scale convolutional residual network model to obtain the target multi-scale convolutional residual network model.

[0010] In an optional implementation, the loss function when iteratively training the initial multi-scale convolutional residual network model is expressed as: in, represents the multi-scale convolutional residual network model at the kth iteration, θ represents the network parameters of the multi-scale convolutional residual network model; m k represents the target parameter of the kth iteration, wherein the target parameter represents: velocity or quality factor, I′1 represents the training pre-stack reverse time migration imaging data, I′2 represents the training post-stack reverse time migration imaging data, represents the regularization term R(m k ), m true Represents the true value of the target parameter in the sample formation model.

[0011] In a second aspect, the present invention provides a device for determining the position of a natural gas hydrate, comprising: a first acquisition module, used to acquire a background velocity model and a background quality factor model of a natural gas hydrate-containing formation; a first simulation module, used to simulate forward wave field data of a preset source in the natural gas hydrate-containing formation based on the background velocity model, the background quality factor model and a viscoacoustic wave equation containing a fractional-order Laplace operator, and to acquire seismic forward simulation data; a second simulation module, used to use the seismic forward simulation data as a backpropagation source, simulate the backpropagation source in the natural gas hydrate-containing formation to perform reverse propagation, and obtain backpropagation wave field data; a first calculation module, used to obtain a ... simulation module, used to obtain a backpropagation wave field data; a first simulation module, used to obtain a backpropagation wave A module is used to calculate the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the natural gas hydrate-containing formation based on the forward wave field data and the back propagation wave field data; a processing module is used to process the background velocity model and the background quality factor model based on the pre-stack reverse time migration imaging data, the post-stack reverse time migration imaging data and the target multi-scale convolutional residual network model, respectively, to obtain the target velocity model and the target quality factor model of the natural gas hydrate-containing formation; a determination module is used to determine the position data of the natural gas hydrate in the natural gas hydrate-containing formation based on the target velocity model and the target quality factor model.

[0012] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, and when the processor executes the computer program, the steps of the method for determining the location of natural gas hydrates described in any one of the aforementioned embodiments are implemented.

[0013] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the method for determining the location of natural gas hydrates described in any one of the aforementioned embodiments.

[0014] The present invention proposes a method for determining the location of natural gas hydrates. After calculating the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the natural gas hydrate-containing formation based on the background velocity model and the background quality factor model of the natural gas hydrate-containing formation, the above two types of imaging data are used as physical principle constraints, and the background velocity model and the background quality factor model are processed respectively using a target multi-scale convolution residual network model to obtain a target velocity model and a target quality factor model of the natural gas hydrate-containing formation, and then the location data of the natural gas hydrate in the natural gas hydrate-containing formation is determined according to the target velocity model and the target quality factor model. Since the target multi-scale convolution residual network model can extract multi-scale information of the image, the method has higher prediction accuracy and a certain generalization ability, and can provide more valuable information for the identification of hydrate reservoirs. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A flow chart of a method for determining the location of a natural gas hydrate provided by an embodiment of the present invention;

[0017] Figure 2 It is a schematic diagram of the forward propagation process of the seismic wave field in the viscous medium;

[0018] Figure 3 Schematic diagram of the Q-compensated reverse time migration process based on the source normalized cross-correlation imaging condition;

[0019] Figure 4 A schematic diagram of an optional multi-scale convolutional residual network model provided for an embodiment of the present invention;

[0020] Figure 5 It is a schematic diagram of some randomly generated sample stratigraphic models;

[0021] Figure 6 For Figure 5 Schematic diagram of the result after Gaussian filtering of the model in;

[0022] Figure 7 for Figure 6 Schematic diagram of pre-stack and post-stack reverse time migration imaging results of the model shown;

[0023] Figure 8 The training and test error reduction curves for speed and quality factors of different network structures;

[0024] Fig. 9 This is a schematic diagram of the speed prediction results based on ResNet with multi-scale convolution kernels and 30 residual blocks;

[0025] Fig.10 for Fig. 9 Schematic diagram of the vertical profile comparison between the predicted velocity values ​​and the actual velocity values ​​of the three models;

[0026] Fig.11 for Fig. 9 Schematic diagram of the gradient of the regularization term based on the prediction speed for the shown model;

[0027] Fig.12Schematic diagram of the quality factor prediction results based on ResNet with multi-scale convolution kernels and 10 residual blocks;

[0028] Fig.13 A method provided by an embodiment of the present invention Fig.12 Schematic diagram of the vertical profile comparison between the predicted quality factor values ​​and the true quality factor values ​​of the three models;

[0029] Fig.14 A method provided by an embodiment of the present invention Fig.12 Schematic diagram of the gradient of the regularization term based on the predicted quality factor for the shown model;

[0030] Fig.15 A functional module diagram of a natural gas hydrate location determination device provided by an embodiment of the present invention;

[0031] Fig.16 A schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0033] 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 invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] Some embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0035] Embodiment 1

[0036] Figure 1 A flow chart of a method for determining the location of a natural gas hydrate provided in an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method specifically comprises the following steps:

[0037] Step S102, obtaining a background velocity model and a background quality factor model of the natural gas hydrate-containing formation.

[0038] For a natural gas hydrate-containing formation whose location is unknown, the user can obtain a background velocity model and a background quality factor model. The background velocity model is also a fuzzy velocity model, and the background quality factor model is also a fuzzy quality factor model, which refers to a data model whose velocity / quality factor is not yet accurate. The velocity in the embodiments of the present invention refers to the velocity of seismic waves propagating in the formation.

[0039] Step S104, based on the background velocity model, the background quality factor model and the viscoacoustic wave equation containing the fractional-order Laplace operator, simulate the forward wave field data of the preset earthquake source in the natural gas hydrate-containing formation, and obtain seismic forward simulation data.

[0040] The traditional wave equation is aimed at elastic media. The natural gas hydrate reservoir is an attenuation medium for seismic waves, that is, a viscoacoustic medium. Therefore, it is necessary to use the viscoacoustic wave equation to describe the attenuation effect of the viscoacoustic medium on seismic waves. Specifically, the embodiment of the present invention uses the viscoacoustic wave equation containing the fractional-order Laplace operator as the basic equation to describe the physical mechanism of the propagation of seismic waves in the natural gas hydrate-containing formation. By adaptively transforming and adjusting the viscoacoustic wave equation containing the fractional-order Laplace operator, the embodiment of the present invention specifically transforms the fractional order into the first order (integer order), and then combines it with the background velocity model and background quality factor model that have been obtained, so as to simulate the forward wave field data generated by the preset source in the natural gas hydrate-containing formation, save the forward wave field data at each time and all spatial positions, and obtain the seismic forward simulation data (that is, seismic records).

[0041] Step S106, using the seismic forward modeling data as a backpropagation source, simulating the backpropagation source to propagate in reverse in the natural gas hydrate-containing formation, and obtaining backpropagation wavefield data.

[0042] After obtaining the seismic forward modeling data, it is used as a backpropagation source and simulated to propagate in reverse in the gas hydrate-containing formation, thereby obtaining the backpropagation wavefield data.

[0043] Step S108, calculating the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the natural gas hydrate-containing formation based on the forward wave field data and the back propagation wave field data.

[0044] Step S110, based on the pre-stack reverse time migration imaging data, the post-stack reverse time migration imaging data and the target multi-scale convolutional residual network model, the background velocity model and the background quality factor model are processed respectively to obtain the target velocity model and the target quality factor model of the natural gas hydrate-containing formation.

[0045] In order to improve the accuracy of natural gas hydrate location prediction, the present invention sets physical principle constraints for the prediction of the network model. The physical principle constraints selected by the embodiment of the present invention are specifically: pre-stack reverse time migration imaging data and post-stack reverse time migration imaging data of natural gas hydrate-containing formations. Therefore, after obtaining the forward wavefield data and the backpropagation wavefield data, it is necessary to further calculate the above two types of imaging data based on the forward wavefield data and the backpropagation wavefield data.

[0046] Traditional residual network models usually only include 3*3 convolutions, with a fixed receptive field for feature images. The extracted image features are relatively one-sided, resulting in limited model prediction accuracy. In order to further improve the prediction accuracy of the network model, the embodiment of the present invention selects a multi-scale convolution residual network model to extract multi-scale information of the image through multi-scale convolution, thereby obtaining higher prediction accuracy.

[0047] Therefore, after calculating the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data, the embodiment of the present invention processes the background velocity model and the background quality factor model respectively through the pre-trained target multi-scale convolutional residual network model, that is, the embodiment of the present invention predicts the velocity and quality factor in the bottom layer containing natural gas hydrates respectively. When predicting the velocity, the background velocity model, the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data are used as the basic data for the target multi-scale convolutional residual network model to predict, and the model output is the target velocity model; when predicting the quality factor, the background quality factor model, the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data are used as the basic data for the target multi-scale convolutional residual network model to predict, and the model output is the target quality factor model.

[0048] Step S112, determining the location data of the natural gas hydrate in the natural gas hydrate-containing formation based on the target velocity model and the target quality factor model.

[0049] It is known that there is a large difference between the velocity / quality factor of natural gas hydrates and the velocity / quality factor of non-natural gas hydrates. Therefore, after obtaining the target velocity model and the target quality factor model of the natural gas hydrate-containing formation, the position of the natural gas hydrate in the natural gas hydrate-containing formation can be accurately located through the boundaries formed by the numerical differences at different positions in the target velocity model and the target quality factor model.

[0050] The embodiment of the present invention proposes a method for determining the position of natural gas hydrates. After calculating the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the natural gas hydrate-containing formation based on the background velocity model and the background quality factor model of the natural gas hydrate-containing formation, the above two types of imaging data are used as physical principle constraints, and the background velocity model and the background quality factor model are processed respectively using a target multi-scale convolution residual network model to obtain a target velocity model and a target quality factor model of the natural gas hydrate-containing formation, and then the position data of the natural gas hydrate in the natural gas hydrate-containing formation is determined according to the target velocity model and the target quality factor model. Since the target multi-scale convolution residual network model can extract multi-scale information of the image, the method has higher prediction accuracy and a certain generalization ability, and can provide more valuable information for the identification of hydrate reservoirs.

[0051] In an optional embodiment, in the above step S104, based on the background velocity model, the background quality factor model and the viscoacoustic wave equation containing the fractional-order Laplace operator, the forward wave field data of the preset seismic source in the natural gas hydrate-containing formation is simulated, which specifically includes the following steps:

[0052] Step S1041, decoupling the viscoacoustic wave equation containing the fractional-order Laplace operator to obtain an approximately constant-Q viscoacoustic wave equation.

[0053] Specifically, the viscoacoustic wave equation containing the fractional-order Laplace operator is expressed as: Among them, ω0 represents the reference frequency of the formation background velocity in wave propagation, β represents the formation medium attenuation parameter, 1≤β≤2, when β=1, the viscoacoustic wave equation will degenerate into a completely elastic acoustic wave equation, when β=2, it is an infinite attenuation medium, and when describing viscoelastic media, β is generally slightly greater than 1. c0 represents the phase velocity of ω0, p represents the acoustic pressure field, s represents the earthquake source, x represents the horizontal position, and z represents the depth.

[0054] The method of decoupling the viscoacoustic wave equation containing the fractional-order Laplace operator is an existing method. Through decoupling, an approximately constant-Q viscoacoustic wave equation can be obtained: Among them, σ represents the ground stress, which refers to the distribution of forces in the rock mass inside the earth formed by the interaction of rocks and geological structural constraints, t represents the travel time of the earthquake, represents the gradient operator, γ=1-1 / β and γ=arctan(1 / Q) / π, Q represents the quality factor, and Q>0,

[0055] Step S1042, deforming the viscoacoustic wave equation with approximate constant Q to obtain a fractional-order partial differential equation to which absorbing boundary conditions can be applied.

[0056] Under normal circumstances, if the boundary interference is serious, the boundary reflection interference can be removed by applying the absorbing boundary conditions. However, the traditional absorbing boundary conditions are designed based on the second-order (integer-order) Laplace operator, and it is difficult to apply them directly to fractional-order equations. Therefore, it is necessary to deform the viscoacoustic wave equation with an approximate constant Q, specifically rewriting the equation to convert the fractional order into the first-order (integer-order) solution. Naturally, the boundary conditions are fully applicable.

[0057] In the embodiment of the present invention, the viscoacoustic wave equation with approximate constant Q is deformed, and the obtained fractional-order partial differential equation is expressed as: Among them, v represents the wave propagation velocity, r represents the spatial coordinate position (x, z), ρ represents the formation density, and e represents the strain. The strain e (strain) is the degree of deformation of the rock mass under the action of stress σ (stress). represents the divergence operator, M0(r) represents the bulk modulus at the spatial coordinate position r, and When Q approaches infinity and γ approaches 0, the viscoacoustic wave equation and fractional partial differential equation with approximate constant Q both degenerate into the acoustic wave equation of a completely elastic medium. In the simulation of seismic wave fields, only the first two first-order equations in the fractional partial differential equations are applied with perfectly matched layer (PML) absorbing boundary conditions. The fractional partial differential equations describe the physical mechanism of seismic wave propagation in natural gas hydrates. The above fractional partial differential equations are obtained in preparation for the subsequent compensation migration imaging.

[0058] Step S1043, bringing the background velocity model, the background quality factor model and the preset earthquake source into the fractional-order partial differential equation to obtain the forward wavefield data.

[0059] According to the expression of fractional-order partial differential equation, by substituting the specific parameter values ​​in the background velocity model and the background quality factor model into the equation, the forward wave field can be simulated and the forward wave field data can be obtained through finite difference approximation.

[0060] In an optional embodiment, the above step S108, based on the forward wavefield data and the back propagation wavefield data, calculates the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the natural gas hydrate-containing formation, and specifically includes the following steps:

[0061] Step S1081, using the Q compensation imaging condition, Q compensation reverse time migration processing is performed on the forward wavefield data and the back propagation wavefield data to obtain pre-stack reverse time migration imaging data.

[0062] The existing reverse time migration imaging method directly uses the forward wave field data and the back propagation wave field data for cross-correlation imaging to obtain the reverse time migration imaging result, namely: Among them, P S(r, t) represents the forward wave field data, P R (r, t) represents the reverse propagation wavefield data, r represents the spatial coordinate position, t represents the earthquake travel time, T represents the total length of the seismic forward simulation data sampling time, and I(r) is the seismic imaging section. Although the cross-correlation imaging condition is easy to achieve and can provide relatively high-precision reverse time migration imaging results based on the entire wavefield information, it still has the defects of low-frequency noise and poor depth illumination, and can only reflect structural information without physical interpretation.

[0063] In an attenuating medium, the amplitude of seismic waves decreases with increasing propagation distance, and the phase also gradually distorts. For ease of explanation, it is generally assumed that the amplitude is an exponential function e -αL The amplitude of the seismic wave is reduced to the original value in the process of propagating from the source to the receiver. Figure 2 Schematic diagram of the forward propagation process of seismic waves in viscous media, where L down and L up are the propagation distances of the downgoing and upgoing waves, respectively.

[0064] In view of this, in order to accurately present the imaging results, it is necessary to compensate for the attenuation. In order to avoid the amplification of high-frequency noise during the compensation process, regularization or filtering must be performed in the process of extrapolating the forward propagating source wave field (that is, the forward wave field) and the reverse propagating receiving wave field (that is, the reverse wave field), thereby causing a certain amplitude loss. Therefore, the embodiment of the present invention uses the source normalized cross-correlation imaging condition for Q compensation. Q compensation can be used to correct and compensate for the frequency attenuation effect in seismic data to restore the original spectral characteristics of the seismic signal. During the propagation of seismic waves, the energy dissipation in the earth's medium will cause the frequency attenuation of seismic waves. This attenuation factor is denoted as Q. Because it is an attenuated wave field of seismic wave propagation, it is only necessary to compensate the receiving wave field. Figure 3 The figure is a schematic diagram of the Q-compensated reverse time migration process based on the source normalized cross-correlation imaging condition. By using the source normalized cross-correlation imaging condition for Q compensation, prestack reverse time migration imaging data can be obtained.

[0065] In the embodiment of the present invention, Q compensation imaging conditions are used to perform Q compensation reverse time migration processing on forward wavefield data and back propagation wavefield data to obtain prestack reverse time migration imaging data, which specifically includes the following contents:

[0066] Using the formula Calculate prestack reverse time migration imaging data; where P S (r, t) represents the forward wave field data, P R(r, t) represents the reverse propagation wavefield data, r represents the spatial coordinate position, t represents the earthquake travel time, T represents the total length of the seismic forward simulation data sampling time, and I1 represents the pre-stack reverse time migration imaging data.

[0067] Specifically, the Q compensation imaging condition can be described as: in, After substituting into the formula, we get: Further simplification gives: That is the pre-stack reverse time migration imaging data I1.

[0068] Under this imaging condition, the source wave field is an attenuated field, and there is no need to filter the source wave field, thereby reducing the loss of reflectivity value in the imaging result. At the reflection point, the waveforms of the source wave field and the receiving wave field have good matching.

[0069] Step S1082, stacking the pre-stack reverse time migration imaging data to obtain post-stack reverse time migration imaging data.

[0070] In an optional embodiment, the above step S110 processes the background velocity model and the background quality factor model based on the pre-stack reverse time migration imaging data, the post-stack reverse time migration imaging data and the target multi-scale convolutional residual network model, respectively, and specifically includes the following contents:

[0071] Expression-based Iteratively invert each background velocity in the background velocity model to obtain the target velocity corresponding to each background velocity; where Λ represents the target multi-scale convolutional residual network model, represents the speed of the k1th iteration, k1=0,1,2,...N1-1, N1 represents the preset number of iterations of the speed, υ0=υ back , back represents the background velocity, I1 represents the pre-stack reverse time migration imaging data, I2 represents the post-stack reverse time migration imaging data, The regularization term representing the speed The gradient of Indicates the target speed.

[0072] Using expressions Each background quality factor in the background quality factor model is iteratively inverted to obtain the target quality factor corresponding to each background quality factor; wherein, represents the quality factor of the k2th iteration, k2=0,1,2,...N2-1, N2 represents the preset number of iterations of the quality factor, Q0=Q back , Q back represents the background quality factor, The regularization term representing the quality factor The gradient of Represents the target quality factor.

[0073] Specifically, the embodiment of the present invention adopts the idea of ​​iterative learning to train the initial multi-scale convolutional residual network model and obtain the target multi-scale convolutional residual network model to invert the velocity and quality factor of the natural gas hydrate-containing formation. In order to prevent overfitting and ensure the stability of the network learning process, a regularization term R(m k ), where m k represents the target parameter of the kth iteration. The target parameter represents: velocity υ or quality factor Q. When the target multi-scale convolutional residual network model is used to invert the velocity, the regularization term R(m k ) is When the quality factor is inverted using the target multi-scale convolutional residual network model, the regularization term R(m k ) is

[0074] Optionally, the regularization term can be a Dirichlet energy function: Represents the gradient operator, and the gradient of the above formula is obtained: represents the divergence operator, That is, the Laplace operator of the current model, which helps to extract boundary information. The embodiment of the present invention does not specifically limit the function of the regularization term, and the user can choose according to actual needs. For example, you can also choose to use:

[0075] According to the above iterative inversion expression, for the target multi-scale convolutional residual network model, the data processed by the model actually includes four items. The first item is the velocity / quality factor, the second item is the pre-stack reverse time migration imaging data, the third item is the post-stack reverse time migration imaging data, and the fourth item is the gradient of the velocity / quality factor regularization term. The user only needs to provide background parameters (background velocity / background quality factor) and pre-stack and post-stack reverse time migration imaging data. The model can perform a preset number of iterative inversions based on the above parameters, and use the result of the last iteration as the target velocity / target quality factor.

[0076] Figure 4 A schematic diagram of an optional multi-scale convolutional residual network model provided in an embodiment of the present invention, wherein the model is a residual network and a multi-scale convolution kernel is used to construct a residual block. Figure 4In the figure, the number above the cube represents the number of feature channels. The first channel of the input model is the velocity or quality factor. The data bodies of the next 5 channels are the pre-stack reverse time migration imaging data. The data bodies of the remaining two channels are the post-stack reverse time migration imaging data and the gradient of the regularization term, respectively. Then, 3*3, 5*5 and 8*8 convolutions are performed on the input data in turn to extract information of different scales of the data. Each convolution is followed by a ReLU activation function, which is then mapped to a residual value through a 1*1 convolution and added to the input velocity or quality factor to obtain the result of this round of iterative update. Then, the iteration is continued until the corresponding preset number of iterations is reached.

[0077] The following is an introduction to a method for obtaining a target multi-scale convolutional residual network model. In one embodiment, the method of the present invention further includes the following steps:

[0078] Step S201, obtaining a training sample set.

[0079] The training sample set includes multiple sample formation models containing natural gas hydrates.

[0080] The known effective medium model is mainly used to estimate the elastic modulus and velocity parameters of hydrate-containing sediments or high-porosity, non-cemented sediments. The model not only uses statistics and rock elastic mechanics, but also considers the microstructure and macroscopic properties of the medium. The White model targets the inhomogeneous infiltration phenomenon caused by the infiltration of immiscible multiphase fluids into porous media. When seismic waves propagate in rocks, the inhomogeneous multiphase fluids will undergo irreversible micro-shifts or deformations.

[0081] In order to train the network, the embodiment of the present invention randomly generates 229 (the following specific values ​​are all exemplary values) hydrate models based on the White model and the effective medium model. The size of the model is 411×411, and the spatial sampling interval in the x and z directions is 5m. The first step of model generation is to use sin(τ×π×x i / n x / d x ) function randomly generates four stratum interfaces, where τ∈[1,3] is a random number that controls the undulation of the stratum interface, x i ∈[5,2055] is the horizontal coordinate position, n x =411 is the number of horizontal grid points, d x=5 is the horizontal grid spacing, where the first layer is the seafloor interface, controlled below 800m depth, and the depths of the remaining layers are randomly generated below the seafloor. The velocity and quality factor of each layer are also calculated by the effective medium model and the White model. The solid phase composition is still composed of 22% clay, 30% quartz, 20% muscovite, 20% calcite and 8% plagioclase. The porosity and permeability of each layer decrease as the layer depth decreases, where the layer depth is taken as the layer center depth. Then, hydrate layers and free gas layers are randomly generated. The bottom interface of the hydrate layer is consistent with the corresponding seabed interface. The bottom interface is randomly generated between 150m and 400m in depth, and the horizontal extension range is randomly generated between 250m and 1000m. The thickness is from 100m to below the seabed. The thickness of the gas layer is between 100m and 280m. The hydrate saturation is between 0.3 and 0.5, and the hydrate saturation near the upper interface of the hydrate layer decreases rapidly. The gas saturation of the gas layer is 0.05. Therefore, the effective medium model and the White model are used to assign values ​​to the hydrate layer and the gas layer. Finally, faults are randomly generated with a probability of 50%. The randomly generated partial sample formation models are as follows Figure 5 shown.

[0082] Step S202: Perform a specified number of filtering processes on each sample formation model in the training sample set to obtain a sample background model corresponding to each sample formation model.

[0083] The sample background model includes: a sample background velocity model and a sample background quality factor model.

[0084] In order to simulate the formation model obtained by the user in actual use, the embodiment of the present invention performs a specified number of filtering processes on each accurate sample formation model in the training sample set. For example, Gaussian filtering is repeated 30 times on each accurate sample formation model to obtain a sample background model corresponding to each sample formation model. The window size of the Gaussian filter is 40×40, the standard deviation is 20, Figure 6 For Figure 5 Schematic diagram of the results after Gaussian filtering of the model in.

[0085] Step S203, based on the target sample background model and the viscoacoustic wave equation containing the fractional-order Laplace operator, the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the corresponding formation are calculated to obtain the training pre-stack reverse time migration imaging data and the training post-stack reverse time migration imaging data.

[0086] The target sample background model represents any one of multiple sample background models.

[0087] After obtaining the target sample background model, the same method as steps S104 to S108 above is used to calculate the corresponding pre-stack reverse time migration imaging data and post-stack reverse time migration imaging data. Please refer to the above for the specific steps, which will not be repeated here.

[0088] In this embodiment, five seismic sources are evenly distributed in the horizontal direction, the distance between the seismic sources is about 513 meters, the number of geophones is 411, the interval is 5 meters, the seismic sources and geophones are located at a depth of 5 meters, the seismic sources are Ricker sub-waves with a central frequency of 35 Hz, the time sampling interval is 0.5 ms, and the total record length is 3.2 s. Figure 7 Shown Figure 6 Pre-stack and post-stack reverse time migration imaging results of the model shown. The first row is the post-stack result, and the second row is the pre-stack result.

[0089] Step S204, based on all sample background models and their corresponding sample formation models, training pre-stack reverse time migration imaging data and training post-stack reverse time migration imaging data, iteratively train the initial multi-scale convolutional residual network model to obtain a target multi-scale convolutional residual network model.

[0090] Optionally, the initial multi-scale convolutional residual network model can be implemented under the Baidu PaddlePaddle framework, and the optimization algorithm adopts the adaptive gradient optimization algorithm (Adagrad) and the learning rate is set to 0.01.

[0091] Among them, the loss function for iterative training of the initial multi-scale convolutional residual network model is expressed as: That is, the root mean square error. represents the multi-scale convolutional residual network model at the kth iteration, θ represents the network parameters of the multi-scale convolutional residual network model; m k represents the target parameter of the kth iteration. The target parameter represents: velocity or quality factor. I′1 represents the training pre-stack reverse time migration imaging data. I′2 represents the training post-stack reverse time migration imaging data. represents the regularization term R(m k ), m true Represents the true value of the target parameter in the sample formation model.

[0092] Optionally, 195 of the 229 generated hydrate models are used as training samples, and the remaining 34 models are used as test samples. In order to illustrate the influence of multi-scale convolution kernels and the number of iterative residual blocks on the prediction results, this embodiment trains and predicts ResNets with different numbers of residual blocks and network structures. Figure 8 The training and test error reduction curves of speed and quality factor of different network structures are shown. Figure 8In the figure, (a) is the speed prediction loss curve based on the original ResNet with 30 and 10 residual blocks respectively, (b) is the speed prediction loss curve based on the multi-scale convolutional ResNet with 30 and 10 residual blocks respectively, (c) is the quality factor prediction loss curve based on the original ResNet with 10 and 5 residual blocks respectively, (d) is the quality factor prediction loss curve based on the multi-scale convolutional ResNet with 10 and 5 residual blocks respectively.

[0093] By comparison, it can be found that when the residual blocks are the same, whether in terms of speed or quality factor prediction, the training and test errors of the ResNet network structure based on the multi-scale convolution kernel are lower than those of the original residual network structure with 3*3 convolution kernels. In addition, the loss curve of the original residual network structure has multiple jumps on the test set, showing instability. The ResNet network structure based on the multi-scale convolution kernel is more stable on the test set, and the test error curve is smoother.

[0094] Fig. 9 The velocity prediction results of ResNet based on multi-scale convolution kernels and 30 residual blocks are shown. The first row is the stacked migration imaging result, from which obvious BSR reflection features can be seen. The imaging quality below the hydrate is poor, and the reflection interface is unclear. The second row is the smooth background velocity. The third row is the gradient of the regularization term based on the background velocity, that is, the Laplace of the velocity. It can be seen that the boundary deviates greatly from the actual position. The fourth row is the actual velocity model, and the fifth row is the predicted velocity model. Good prediction results are achieved for both smaller hydrate reservoirs and complex hydrate reservoirs divided by faults.

[0095] Fig.10 Shown Fig. 9 The vertical profile comparison between the predicted velocity values ​​and the actual velocity values ​​of the three models. Fig.10 In the figure, (a) is a single-channel comparison of sediments with hydrates, and (b) is a single-channel comparison of sediments without hydrates. Fig.10 In the cross section through the hydrate reservoir in view (a), it can be seen that the interface information between the hydrate and free gas layers is accurately predicted, the velocity value has a slight deviation, and the velocity value below the hydrate layer has a large deviation. Fig.10 In the cross section that does not pass through the hydrate reservoir in view (b), the formation interface and velocity numerical predictions are accurate. The main reason is that the reflection amplitude of the interface below the gas layer is not fully compensated and the number of seismic sources is small. Fig.11 for Fig. 9 The gradient of the regularization term based on the predicted velocity of the model shown in Figure 2 shows that the hydrate boundary position is related to Fig. 9 The third row has recovered significantly.

[0096] Fig.12 The quality factor prediction results of ResNet based on multi-scale convolution kernels and 10 residual blocks are shown. The first line is the superposition offset imaging result, the second line is the smooth background quality factor model, the third line is the gradient of the regularization term based on the background quality factor. Similarly, the quality factor boundary has a large offset from the actual position. The fourth line is the real quality factor model, and the fifth line is the predicted quality factor model. It has a relatively good prediction effect on the spatial distribution of hydrates and gas layers, which is basically consistent with the real model.

[0097] Fig.13 Shown Fig.12 The vertical profile comparison between the predicted quality factor values ​​and the true quality factor values ​​of the three models. Fig.13 In the figure, (a) shows a single-track comparison of sediments that have hydrates, and (b) shows a single-track comparison of sediments that have not hydrates. The cross-section comparison through the hydrate reservoir shows that the quality factor values ​​and interface positions of the hydrate and free gas layers are accurately predicted ( Fig.13 (a) view), in Fig.13 In the cross section that does not pass through the hydrate reservoir in view (b), the formation interface and quality factor values ​​are also predicted more accurately. Fig.14 for Fig.12 The gradient of the regularization term based on the predicted quality factor of the model shown, showing the boundary between the hydrate and gas layer and Fig.12 The third row is obviously restored to the real position. According to the above numerical experiments, the residual network structure based on multi-scale convolution has better prediction accuracy and certain generalization ability, which can provide more valuable information for the identification of hydrate reservoirs.

[0098] The embodiment of the present invention mainly proposes a velocity and quality factor inversion technology for viscoacoustic medium hydrate-containing formations based on multi-scale convolutional residual network deep learning constrained by physical principles, rather than the improvement of the CNN architecture itself. The Resnet network structure can be used, or other CNN architectures can be used supplemented by physical principle constraints.

[0099] In summary, the embodiment of the present invention proposes a velocity and quality factor inversion technology for natural gas hydrate-containing formations based on multi-scale residual network deep learning constrained by physical principles. Since the target multi-scale convolutional residual network model can extract multi-scale information of the image, it has better prediction accuracy and a certain generalization ability, which can provide more valuable information for the identification of hydrate reservoirs.

[0100] Embodiment 2

[0101] An embodiment of the present invention further provides a natural gas hydrate position determination device, which is mainly used to execute the natural gas hydrate position determination method provided in the above-mentioned embodiment 1. The natural gas hydrate position determination device provided in the embodiment of the present invention is specifically introduced below.

[0102] Fig.15 is a functional module diagram of a natural gas hydrate location determination device provided by an embodiment of the present invention, such as Fig.15 As shown, the device mainly includes: a first acquisition module 11, a first simulation module 12, a second simulation module 13, a first calculation module 14, a processing module 15, and a determination module 16, wherein:

[0103] The first acquisition module 11 is used to acquire a background velocity model and a background quality factor model of a natural gas hydrate-containing formation.

[0104] The first simulation module 12 is used to simulate the forward wave field data of a preset earthquake source in a natural gas hydrate-containing formation based on a background velocity model, a background quality factor model and a viscoacoustic wave equation containing a fractional-order Laplace operator, and to obtain seismic forward simulation data.

[0105] The second simulation module 13 is used to use the seismic forward modeling data as a backpropagation source, simulate the backpropagation source to propagate in reverse in the natural gas hydrate-containing formation, and obtain backpropagation wavefield data.

[0106] The first calculation module 14 is used to calculate the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the natural gas hydrate-containing formation based on the forward wave field data and the back propagation wave field data.

[0107] The processing module 15 is used to process the background velocity model and the background quality factor model based on the pre-stack reverse time migration imaging data, the post-stack reverse time migration imaging data and the target multi-scale convolution residual network model, respectively, to obtain the target velocity model and the target quality factor model of the natural gas hydrate-containing formation.

[0108] The determination module 16 is used to determine the location data of the natural gas hydrate in the natural gas hydrate-containing formation based on the target velocity model and the target quality factor model.

[0109] The embodiment of the present invention proposes a device for determining the position of natural gas hydrates. After calculating the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the natural gas hydrate-containing formation based on the background velocity model and the background quality factor model of the natural gas hydrate-containing formation, the above two types of imaging data are used as physical principle constraints, and the background velocity model and the background quality factor model are processed respectively using a target multi-scale convolution residual network model to obtain a target velocity model and a target quality factor model of the natural gas hydrate-containing formation, and then the position data of the natural gas hydrate in the natural gas hydrate-containing formation is determined according to the target velocity model and the target quality factor model. Since the target multi-scale convolution residual network model can extract multi-scale information of the image, the device has higher prediction accuracy and a certain generalization ability, and can provide more valuable information for the identification of hydrate reservoirs.

[0110] Optionally, the first simulation module 12 is specifically used for:

[0111] The viscoacoustic wave equation with fractional-order Laplace operator is decoupled to obtain an approximately constant-Q viscoacoustic wave equation.

[0112] The viscoacoustic wave equation with approximately constant Q is transformed to obtain a fractional-order partial differential equation to which absorbing boundary conditions can be applied.

[0113] The background velocity model, background quality factor model and preset seismic source are introduced into the fractional-order partial differential equation to obtain the forward wavefield data.

[0114] Optionally, the first calculation module 14 includes:

[0115] The Q compensation unit is used to perform Q compensation reverse time migration processing on the forward wavefield data and the back propagation wavefield data using the Q compensation imaging condition to obtain pre-stack reverse time migration imaging data.

[0116] The stacking unit is used to stack the pre-stack reverse time migration imaging data to obtain the post-stack reverse time migration imaging data.

[0117] Optionally, the processing module 15 is specifically used for:

[0118] Expression-based Iteratively invert each background velocity in the background velocity model to obtain the target velocity corresponding to each background velocity; where Λ represents the target multi-scale convolutional residual network model, represents the speed of the k1th iteration, k1=0,1,2,...N1-1, N1 represents the preset number of iterations of the speed, υ0=υ back , back represents the background velocity, I1 represents the pre-stack reverse time migration imaging data, I2 represents the post-stack reverse time migration imaging data, The regularization term representing the speed The gradient of Indicates the target speed.

[0119] Using expressions Each background quality factor in the background quality factor model is iteratively inverted to obtain the target quality factor corresponding to each background quality factor; wherein, represents the quality factor of the k2th iteration, k2=0,1,2,...N2-1, N2 represents the preset number of iterations of the quality factor, Q0=Q back , Q back represents the background quality factor, The regularization term representing the quality factor The gradient of Represents the target quality factor.

[0120] Optionally, the Q compensation unit is specifically used for:

[0121] Using the formula Calculate prestack reverse time migration imaging data; where P S (r, t) represents the forward wave field data, P R (r, t) represents the reverse propagation wavefield data, r represents the spatial coordinate position, t represents the earthquake travel time, T represents the total length of the seismic forward simulation data sampling time, and I1 represents the pre-stack reverse time migration imaging data.

[0122] Optionally, the device further comprises:

[0123] The second acquisition module is used to acquire a training sample set; wherein the training sample set includes a plurality of sample formation models containing natural gas hydrates.

[0124] The filtering module is used to perform a specified number of filtering processes on each sample formation model in the training sample set to obtain a sample background model corresponding to each sample formation model; wherein the sample background model includes: a sample background velocity model and a sample background quality factor model.

[0125] The second calculation module is used to calculate the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the corresponding stratum based on the target sample background model and the viscoacoustic wave equation containing the fractional-order Laplace operator, so as to obtain the training pre-stack reverse time migration imaging data and the training post-stack reverse time migration imaging data; wherein the target sample background model represents any one of the multiple sample background models.

[0126] The training module is used to iteratively train the initial multi-scale convolutional residual network model based on all sample background models and their corresponding sample formation models, training pre-stack reverse time migration imaging data and training post-stack reverse time migration imaging data to obtain a target multi-scale convolutional residual network model.

[0127] Optionally, the loss function for iterative training of the initial multi-scale convolutional residual network model is expressed as: in, represents the multi-scale convolutional residual network model at the kth iteration, θ represents the network parameters of the multi-scale convolutional residual network model; m k represents the target parameter of the kth iteration. The target parameter represents: velocity or quality factor. I′1 represents the training pre-stack reverse time migration imaging data. I′2 represents the training post-stack reverse time migration imaging data. represents the regularization term R(m k ), m true Represents the true value of the target parameter in the sample formation model.

[0128] Embodiment 3

[0129] See also Fig.16 An embodiment of the present invention provides an electronic device, which includes: a processor 60, a memory 61, a bus 62 and a communication interface 63, wherein the processor 60, the communication interface 63 and the memory 61 are connected via the bus 62; the processor 60 is used to execute an executable module stored in the memory 61, such as a computer program.

[0130] The memory 61 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 63 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0131] The bus 62 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.16 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0132] Among them, the memory 61 is used to store programs, and the processor 60 executes the program after receiving the execution instruction. The method executed by the device defined by the process disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 60 or implemented by the processor 60.

[0133] The processor 60 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 60. The above processor 60 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor to be executed, or the hardware and software modules in the decoding processor can be executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 61, and the processor 60 reads the information in the memory 61 and completes the steps of the above method in combination with its hardware.

[0134] A computer program product of a method, device and electronic device for determining the location of natural gas hydrate provided in an embodiment of the present invention includes a computer-readable storage medium storing non-volatile program code executable by a processor, and the instructions included in the program code can be used to execute the methods described in the previous method embodiments. The specific implementation can be found in the method embodiments, which will not be repeated here.

[0135] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0136] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0137] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0138] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, or the positions or positional relationships in which the product of the invention is usually placed when in use. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific position, be constructed and operated in a specific position, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0139] In addition, the terms "horizontal", "vertical", "overhanging" and the like do not mean that the components are required to be absolutely horizontal or overhanging, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0140] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining the location of a natural gas hydrate, characterized in that: include: Obtaining background velocity model and background quality factor model of gas hydrate-containing formations; The background speed model represents an inaccurate data model of speed; The background quality factor model represents a data model in which the quality factor is not yet accurate; Based on the background velocity model, the background quality factor model and the viscoacoustic wave equation containing the fractional-order Laplace operator, simulate the forward wave field data of the preset earthquake source in the natural gas hydrate-containing formation, and obtain seismic forward simulation data; Using the seismic forward modeling data as a backpropagation source, simulating the backpropagation source to propagate in reverse in the natural gas hydrate-containing formation, and obtaining backpropagation wavefield data; Based on the forward wavefield data and the backpropagation wavefield data, calculating pre-stack reverse time migration imaging data and post-stack reverse time migration imaging data of the natural gas hydrate-containing formation; Based on the pre-stack reverse time migration imaging data, the post-stack reverse time migration imaging data and the target multi-scale convolution residual network model, the background velocity model and the background quality factor model are processed respectively to obtain a target velocity model and a target quality factor model of the natural gas hydrate-containing formation; The location data of the natural gas hydrate in the natural gas hydrate-containing formation is determined based on the target velocity model and the target quality factor model.

2. The method for determining the location of natural gas hydrate according to claim 1, characterized in that: Based on the background velocity model, the background quality factor model and the viscoacoustic wave equation containing a fractional-order Laplace operator, forward wave field data of a preset earthquake source in the natural gas hydrate-containing formation are simulated, including: Decoupling the viscoacoustic wave equation containing the fractional-order Laplace operator to obtain an approximately constant-Q viscoacoustic wave equation; The viscoacoustic wave equation with approximate constant Q is deformed to obtain a fractional-order partial differential equation to which absorbing boundary conditions can be applied; The background velocity model, the background quality factor model and the preset earthquake source are introduced into the fractional-order partial differential equation to obtain the forward wavefield data.

3. The method for determining the location of natural gas hydrate according to claim 1, characterized in that: Based on the forward wavefield data and the backpropagation wavefield data, calculating the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the natural gas hydrate-containing formation, including: Performing Q-compensation reverse time migration processing on the forward wavefield data and the backpropagation wavefield data using Q-compensation imaging conditions to obtain the prestack reverse time migration imaging data; The pre-stack reverse time migration imaging data are superimposed to obtain the post-stack reverse time migration imaging data.

4. The method for determining the location of natural gas hydrate according to claim 1, characterized in that: Based on the pre-stack reverse time migration imaging data, the post-stack reverse time migration imaging data and the target multi-scale convolution residual network model, the background velocity model and the background quality factor model are processed respectively, including: Expression-based Iteratively invert each background velocity in the background velocity model to obtain a target velocity corresponding to each background velocity; wherein Λ represents the target multi-scale convolutional residual network model, represents the speed of the k1th iteration, k1=0,1,2,...N1-1, N1 represents the preset number of iterations of the speed, υ0=υ back , back represents the background velocity, I1 represents the pre-stack reverse time migration imaging data, I2 represents the post-stack reverse time migration imaging data, The regularization term representing the speed The gradient of Indicates the target speed; Using expressions Iteratively invert each background quality factor in the background quality factor model to obtain a target quality factor corresponding to each background quality factor; wherein, represents the quality factor of the k2th iteration, k2=0,1,2,...N2-1, N2 represents the preset number of iterations of the quality factor, Q0=Q back , Q back represents the background quality factor, The regularization term representing the quality factor The gradient of Represents the target quality factor.

5. The method for determining the location of natural gas hydrate according to claim 3, characterized in that: Performing Q compensation reverse time migration processing on the forward wavefield data and the back propagation wavefield data using the Q compensation imaging condition to obtain the prestack reverse time migration imaging data, including: Using the formula Calculate the prestack reverse time migration imaging data; wherein, P S (r, t) represents the forward wave field data, P R (r, t) represents the reverse propagation wavefield data, r represents the spatial coordinate position, t represents the earthquake travel time, T represents the total length of the seismic forward simulation data sampling time, and I1 represents the pre-stack reverse time migration imaging data.

6. The method for determining the location of natural gas hydrate according to claim 1, characterized in that: The method further comprises: Acquire a training sample set; wherein the training sample set includes a plurality of sample formation models containing natural gas hydrates; Performing a specified number of filtering processes on each sample formation model in the training sample set to obtain a sample background model corresponding to each sample formation model; wherein the sample background model includes: a sample background velocity model and a sample background quality factor model; Based on the target sample background model and the viscoacoustic wave equation containing the fractional-order Laplace operator, the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the corresponding formation are calculated to obtain the training pre-stack reverse time migration imaging data and the training post-stack reverse time migration imaging data; wherein the target sample background model represents any one of the multiple sample background models; Based on all sample background models and their corresponding sample formation models, training pre-stack reverse time migration imaging data and training post-stack reverse time migration imaging data, the initial multi-scale convolutional residual network model is iteratively trained to obtain the target multi-scale convolutional residual network model.

7. The method for determining the location of natural gas hydrate according to claim 6, characterized in that: The loss function when iteratively training the initial multi-scale convolutional residual network model is expressed as: in, represents the multi-scale convolutional residual network model at the kth iteration, θ represents the network parameters of the multi-scale convolutional residual network model; m k represents the target parameter of the kth iteration, wherein the target parameter represents: velocity or quality factor, I1′ represents the training pre-stack reverse time migration imaging data, I′2 represents the training post-stack reverse time migration imaging data, represents the regularization term R(m k ), m true Represents the true value of the target parameter in the sample formation model.

8. A device for determining the position of natural gas hydrate, characterized in that: include: A first acquisition module is used to acquire a background velocity model and a background quality factor model of a natural gas hydrate-containing formation; The background speed model represents an inaccurate data model of speed; the background quality factor model represents an inaccurate data model of quality factor; A first simulation module is used to simulate the forward wave field data of a preset earthquake source in the natural gas hydrate-containing formation based on the background velocity model, the background quality factor model and the viscoacoustic wave equation containing a fractional-order Laplace operator, and obtain seismic forward simulation data; A second simulation module is used to use the seismic forward modeling data as a backpropagation source, simulate the backpropagation source in the natural gas hydrate-containing formation to propagate in reverse, and obtain backpropagation wavefield data; A first calculation module is used to calculate the pre-stack reverse time migration imaging data and the post-stack reverse time migration imaging data of the natural gas hydrate-containing formation based on the forward wave field data and the back propagation wave field data; A processing module, for processing the background velocity model and the background quality factor model respectively based on the pre-stack reverse time migration imaging data, the post-stack reverse time migration imaging data and the target multi-scale convolution residual network model, so as to obtain a target velocity model and a target quality factor model of the natural gas hydrate-containing formation; A determination module is used to determine the location data of natural gas hydrate in the natural gas hydrate-containing formation based on the target velocity model and the target quality factor model.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, wherein: When the processor executes the computer program, the steps of the method for determining the location of the natural gas hydrate according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method for determining the location of natural gas hydrate according to any one of claims 1 to 7 is implemented.

Images