Rugged seabed depth migration profile imaging method and amplitude attenuation mechanism research method

Abstract

The invention relates to the field of seismic wave simulation and migration imaging, in particular to a rugged seabed depth migration profile imaging method and a research mechanism, and the three-dimensional depth migration imaging algorithm comprises the following steps: SA1, constructing a three-dimensional speed model according to the actual exploration seabed horizon; sA2, first wave field illumination of the three-dimensional rugged seabed is obtained through a wave equation method; sA3, performing second forward modeling on the three-dimensional velocity model by taking the first wave field illumination as a first amplitude-preserved offset weight factor, and obtaining second wave field illumination as a second amplitude-preserved offset weight factor; sA4, applying the second amplitude preservation offset weight factor to a depth offset algorithm to obtain a first imaging profile; the SA5 and the multiple first imaging profiles are combined to form three-dimensional depth migration imaging, the second amplitude preservation migration weight factor is applied to the depth migration algorithm, the application range of the depth migration algorithm is widened, the actual interface reflection amplitude can be effectively obtained, and the quality of the imaging profiles is greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of seismic wave simulation and migration imaging, and more specifically, to a method for deep migration profile imaging of a rugged seabed and a method for studying the amplitude attenuation mechanism. Background Art

[0002] At present, with the growing economic construction, the demand for energy is also increasing, and the exploitation of resources such as oil and natural gas cannot be separated from exploration physics. In the work of marine exploration of oil and gas resources, seismic exploration has always played an extremely important role. Some resource-rich submarine basins have complex terrains, spanning shallow seas, shelf slope breaks, and deep sea areas. There are often multiple potential targets near these rugged seabeds, which are key areas for discovering oil and gas reserves. However, there are problems such as amplitude attenuation and unclear influencing mechanisms in seismic data below the deep-water rugged seabed. There is an urgent need for a set of three-dimensional imaging mechanism research theories and algorithms for the signal attenuation phenomenon of the deep-water rugged seabed to study the mechanism of amplitude signal attenuation below the deep-water rugged seabed.

[0003] The lack of imaging amplitude of the underlying strata of the deep-water rugged seafloor is mainly caused by geological factors, including the degree of seafloor ruggedness (expressed by structural dip), abnormal Q value distribution of the medium below the seafloor, and abnormal low-velocity bodies. Traditional algorithms can be used to generate a large number of forward model samples, and AI deep learning algorithms can be used to link the influencing factors with the forward model and the offset imaging profile. The large sample model can be trained to relatively reliably predict the relationship between the amplitude attenuation and the main influencing factors, so as to evaluate and diagnose the influencing mechanism of the imaging amplitude attenuation of the underlying strata of the deep-water rugged seafloor.

[0004] The core of the mechanism research lies in the first forward simulation and amplitude-preserving imaging. How the imaging reflects the real reflection amplitude of the velocity interface is the key to the problem. At present, the deep migration imaging with strong practicality, relatively high imaging quality and amplitude preservation includes reverse time migration imaging and Kirchhoff prestack depth migration imaging. The basis for these two migration methods to achieve amplitude preservation depends on the selection of the migration weight factor. The migration weight characterizes the background propagation intensity of the seismic wave in the underground medium, which can be understood as the "amplitude" of the seismic wave at each point in the medium, that is, the wave field illumination. The wave field illumination calculation can not only show the difference in propagation amplitude at each point on the same horizontal interface, but also be added as a compensation term to the reverse time migration imaging based on the wave equation to improve the imaging effect. Specifically, by simulating the propagation of the wave in the shallow stratum excited by the source to obtain a part of the illumination, it is added as the denominator compensation term to the depth migration imaging, which can suppress the shallow low-frequency artifacts and thus improve the imaging quality. The ray method, a common wave field illumination calculation method in traditional depth migration imaging, adopts the Born approximation condition. However, the results obtained by this method are highly directional, and the illumination of some areas violates the laws of physics. It cannot correctly display the illumination energy of various interfaces below the seabed, which leads to strong mutation in the imaging of the depth migration algorithm and the imaging profile does not match the actual situation. Summary of the invention

[0005] The purpose of the present invention is to overcome the deficiency of the prior art that the wave field illumination locally violates the physical laws, and to provide a depth migration profile imaging method for rugged seabed and a method for studying the amplitude attenuation mechanism, which is conducive to reducing the mutation of depth migration imaging and improving the imaging quality of the depth migration algorithm.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A method for depth migration profile imaging of a rugged seafloor is provided, comprising the following steps:

[0008] SA1. Construct a three-dimensional velocity model of the rugged seabed based on the actual surveyed seabed position;

[0009] SA2, performing a first forward modeling simulation by a wave equation method to obtain a first wave field illumination of the three-dimensional velocity model in step SA1;

[0010] SA3, performing a second forward simulation on the three-dimensional velocity model by using the first wave field illumination as a first amplitude preserving migration weight factor, and obtaining a second wave field illumination as a second amplitude preserving migration weight factor;

[0011] SA4. Apply the second amplitude-preserving migration weight factor to a depth migration algorithm to obtain a first imaging section.

[0012] The amplitude preservation of the depth migration imaging algorithm depends on the selection of the migration weight factor. Ray theory is a high-frequency approximation of wave theory. However, in practical applications, the frequency of waves cannot be infinitely high. Therefore, the imaging results obtained by using the amplitude preservation migration weight factor obtained by ray theory often have the characteristics of significant mutation. The wave equation method is applicable to waves of various frequencies, and the results are smoother and closer to reality. Through this setting, the wave equation method can calculate the first wave field illumination of the rugged seabed as the first amplitude preservation migration weight factor, and obtain the second wave field illumination as the second amplitude preservation migration weight factor based on the first amplitude preservation migration weight factor. The second wave field illumination is the first wave field illumination updated after the second forward simulation, and has the same meaning as the first wave field illumination. Obtaining the first amplitude preservation migration weight factor and the second amplitude preservation migration weight factor respectively can make the imaging profile closer to the actual reflection interface. The depth migration algorithm using the second amplitude preservation migration weight factor is applicable to any geological structure situation, which can effectively reduce the mutation of depth migration imaging, thereby obtaining the true amplitude of interface reflection and greatly improving the profile quality of depth migration imaging.

[0013] Preferably, the first forward modeling adopts a large three-dimensional acoustic wave equation based on a finite difference algorithm.

[0014] Preferably, in step SA2, the first forward modeling is provided with a source-receiver system, and the source-receiver system adopts a streamer acquisition system.

[0015] Preferably, the finite difference algorithm is calculated by CUDA on a large GPU resource. The GPU refers to a graphics processor or a graphics processing unit, and the CUDA refers to a compute unified device architecture.

[0016] This setting method is helpful to greatly improve the calculation speed of the finite difference algorithm.

[0017] Preferably, in step SA3, the second wave field illumination is obtained specifically by the following formula:

[0018]

[0019] Among them, K(x) is the second wave field illumination and also serves as the second amplitude-preserving migration weight factor,

[0020] |u(x s ,x,t)| 2 is the square value of the first wave field illumination in SA2,

[0021] It is the approximate Green's function of the receiver's reverse extended wave field obtained based on the spatial distribution of the source-receiver system.

[0022] In classical imaging theory, it is necessary to calculate the theoretical Green's function corresponding to each receiver in order to accurately calculate the second amplitude-preserving migration weight factor. However, for the case where the source-receiver system is evenly arranged and distributed, researchers can use the approximate form of the Green's function to solve the problem of solving the back-transmitted wave field of the source-receiver system. With the above setting, it is only necessary to perform the first forward simulation after each shot to obtain the forward wave field, which greatly improves the calculation efficiency. The calculation accuracy of the second amplitude-preserving migration weight factor is high and can be used to express the spatial illumination distribution.

[0023] Preferably, in step SA4, the first imaging section is obtained by the following formula:

[0024]

[0025] Wherein, K(x) is the second amplitude-preserving offset weight factor calculated in step SA3, is the result obtained by the traditional migration algorithm, and m(x) is the first imaging section.

[0026] The traditional migration algorithm refers to the reverse time migration imaging or Kirchhoff prestack depth migration imaging that has not been corrected by the second amplitude-preserving migration weight factor. Through this setting, the second wave field illumination obtained by the wave equation method is used as the second amplitude-preserving migration weight factor to correct the traditional migration algorithm, which is beneficial to suppress shallow low-frequency artifacts and thus improve imaging quality.

[0027] Preferably, in step SA4, the depth migration algorithm adopts reverse time migration imaging or Kirchhoff prestack depth migration imaging.

[0028] The basis for achieving amplitude preservation in reverse time migration imaging or Kirchhoff prestack depth migration imaging depends on the selection of the migration weight factor. Through this setting, the high-precision second amplitude-preserving migration weight factor obtained by the three-dimensional acoustic wave equation is used in the depth migration imaging algorithm, which is beneficial to improve the interpretability and accuracy of imaging.

[0029] A method for studying the amplitude attenuation mechanism comprises the following steps:

[0030] SB1. Perform data dimension reduction on the 3D velocity model, and divide the 3D velocity model into several first characterized velocity models according to the characteristics of three different mechanism factors: structural inclination, abnormal Q value distribution, and abnormal low-velocity body distribution;

[0031] SB2. batch processing the first characterized velocity models through a deep generative model algorithm to generate a large number of second characterized velocity models with different quantitative mechanism factor characteristics;

[0032] SB3, obtaining a second amplitude-preserving migration weight factor of the second characterized velocity model as a theoretical reflection amplitude, and simultaneously performing a third forward simulation on the second characterized velocity model to obtain a second imaging section;

[0033] SB4. Compare the second imaging section with the theoretical reflection amplitude of the second characterized velocity model to obtain an imaging attenuation amplitude;

[0034] SB4. Taking the attenuation amplitude as input, and the different mechanism factor characteristics of the first characterized velocity model and the second characterized velocity model as output, the deep learning model is trained to obtain a large sample model;

[0035] SB5. Through the large sample model, predict and study the mechanism factor characteristics corresponding to the attenuation amplitude of a given seabed.

[0036] Among them, the structural inclination is used to characterize the ruggedness of the seabed. Through this setting, the complex three-dimensional velocity model affected by multiple factors can be decomposed into a first characterized velocity model that is only affected by a single mechanism factor characteristic. The mechanism factor characteristic covers the main geological factors of the missing imaging amplitude of the underlying strata in the deep-water rugged seabed. Then, the first characterized velocity model is expanded into a second characterized velocity model with different quantitative mechanism factor characteristics through the depth generation model. Specifically, the second characterized velocity model can be divided into several major categories with different mechanism factor characteristics. Within a single mechanism factor characteristic category, there is a second characterized velocity model with multiple mechanism factor characteristics with different numerical values. After processing, a large number of attenuation amplitudes of the second imaging profile obtained by different quantitative mechanism factor characteristics are obtained, wherein the theoretical reflection amplitude refers to the second amplitude-preserving offset weight factor. Finally, the attenuation amplitude is used as input and different quantitative mechanism factor characteristics are used as output to train a large sample model. The large sample model can predict the influencing mechanism based on the rugged seabed attenuation sample, thereby assisting researchers in conducting mechanism research on the rugged seabed.

[0037] Preferably, in step SB2, the deep generation model Diffusion is used to batch generate second characterized velocity models with different quantized structural inclinations, different quantized Q values ​​and different quantized low-velocity bodies.

[0038] This setting can greatly increase the number of samples, thereby improving the accuracy of large sample models.

[0039] Preferably, in step SB3, a third forward simulation is performed on the second characterized velocity model by using a viscoacoustic wave equation simulation algorithm to obtain synthetic seismic data, and then a second image corresponding to a different velocity model is generated by using a conventional depth migration algorithm.

[0040] Preferably, the first characterized velocity model and the second characterized velocity model are both two-dimensional velocity models.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] (1) Based on the wave equation method, the first amplitude-preserving migration weight factor is obtained, and then the second amplitude-preserving migration weight factor is calculated through the first amplitude-preserving migration weight factor. The second amplitude-preserving migration weight factor is applied to the depth migration algorithm, which is beneficial to improving the application scope of the depth migration algorithm, can effectively obtain the true amplitude of the interface reflection, and greatly improve the imaging section quality.

[0043] (2) By introducing the second amplitude-preserving offset weight factor to obtain the attenuation amplitude, and then training the large sample model with the attenuation amplitude, it is beneficial to improve the accuracy of the large sample model in predicting the corresponding mechanism factor characteristics of the attenuation amplitude of a given seabed.

[0044] (3) The application of the second amplitude preservation factor can greatly improve the interpretability and rationality of traditional depth migration imaging. The application of the deep learning framework to link the imaging quality with the influencing factors can help researchers to efficiently and reasonably conduct mechanistic research on the factors affecting the amplitude attenuation of deep-water rugged seabed imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a section of a three-dimensional velocity model established according to a real rugged seabed in the embodiment;

[0046] Figure 2 is the profile of the second amplitude-preserving migration weight factor obtained by simulating the wave equation in Example 1 along the direction parallel to the seabed;

[0047] Figure 3 The second amplitude-preserving migration weight factor obtained by wave equation simulation in Example 1 is a profile perpendicular to the seabed direction;

[0048] Figure 4 This is a cross section of a three-dimensional velocity model built based on a real rugged seabed;

[0049] Figure 5 It is the first imaging section of Kirchhoff prestack depth migration imaging;

[0050] Figure 6 This is a flow chart of a method for studying the amplitude attenuation mechanism of Example 3. DETAILED DESCRIPTION

[0051] The present invention will be further described below in conjunction with specific implementation modes.

[0052] Example 1

[0053] like Figures 1 to 5 The first embodiment of the depth migration profile imaging method of a rugged seabed according to the present invention is shown, comprising the following steps:

[0054] SA1. Construct a three-dimensional velocity model of the rugged seabed based on the actual surveyed seabed position;

[0055] SA2, performing a first forward modeling simulation by a wave equation method to obtain the first wave field illumination of the three-dimensional rugged seabed of step SA1;

[0056] SA3, performing a second forward simulation on the three-dimensional velocity model by using the first wave field illumination as a first amplitude preserving migration weight factor, and obtaining a second wave field illumination as a second amplitude preserving migration weight factor;

[0057] SA4. Apply the second amplitude-preserving migration weight factor to the depth migration algorithm to obtain a first imaging section.

[0058] The amplitude preservation of the depth migration imaging algorithm depends on the selection of the migration weight factor. Ray theory is a high-frequency approximation of wave theory. However, in practical applications, the frequency of waves cannot be infinitely high. Therefore, the imaging results obtained by using the amplitude preservation migration weight factor obtained by ray theory often have the characteristics of significant mutation. The wave equation method is applicable to waves of various frequencies, and the results are smoother and closer to reality. Through this setting, the wave equation method can calculate the first wave field illumination of the rugged seabed as the first amplitude preservation migration weight factor, and obtain the second wave field illumination as the second amplitude preservation migration weight factor based on the first amplitude preservation migration weight factor. The second wave field illumination is the first wave field illumination updated after the second forward simulation, and has the same meaning as the first wave field illumination. Obtaining the first amplitude preservation migration weight factor and the second amplitude preservation migration weight factor respectively can make the imaging profile closer to the actual reflection interface. The depth migration algorithm using the second amplitude preservation migration weight factor is applicable to any geological structure situation, which can effectively reduce the mutation of depth migration imaging, thereby obtaining the true amplitude of interface reflection and greatly improving the profile quality of depth migration imaging.

[0059] As an embodiment of the present invention, in step SA2, it specifically includes: carrying out a first forward simulation of the three-dimensional acoustic wave equation based on a finite difference algorithm for a given three-dimensional velocity model, wherein the three-dimensional acoustic wave equation source-receiver system adopts a towed cable acquisition system.

[0060] As one embodiment of the present invention, the finite difference algorithm is calculated by CUDA on large GPU resources.

[0061] This setting method is helpful to greatly improve the calculation speed of the finite difference algorithm.

[0062] Preferably, in step SA3, the second wave field illumination is obtained specifically by the following formula:

[0063]

[0064] Among them, K(x) is the second wave field illumination and also serves as the second amplitude-preserving migration weight factor,

[0065] |u(x s ,x,t)| 2 is the squared value of the first wave field illumination of the first forward modeling of the wave equation method in SA2,

[0066] It is an approximate Green's function of the receiver's reverse extended wave field given by the spatial distribution of the source-receiver system.

[0067] In classical imaging theory, it is necessary to calculate the theoretical Green's function corresponding to each receiver in order to accurately calculate the second amplitude-preserving migration weight factor. However, through the above-mentioned setting, for the case where the source-receiver system is evenly arranged and distributed, we can use the approximate form of the Green's function to solve the problem of solving the back-propagation wave field of the source-receiver system. It is only necessary to perform a first forward simulation after each shot to obtain the forward wave field, which greatly improves the calculation efficiency, and the calculation accuracy of the second amplitude-preserving migration weight factor is high, which can be used to express the spatial illumination distribution.

[0068] Example 2

[0069] The following is a second embodiment of a depth migration profile imaging method for a rugged seabed according to the present invention. This embodiment is similar to the first embodiment, except that the implementation method of step SA4 is further limited.

[0070] As an embodiment of the present invention, in step SA4, the first imaging section is obtained by the following formula:

[0071]

[0072] Wherein, K(x) is the second amplitude-preserving offset weight factor calculated in step SA3, is the result obtained by the traditional migration algorithm, and m(x) is the first imaging section.

[0073] Through this setting, the second wave field illumination obtained by the wave equation method is used as the second amplitude-preserving migration weight factor to correct the traditional migration algorithm, which is beneficial to suppress shallow low-frequency artifacts and thus improve imaging quality.

[0074] As an implementation of the present invention, in step SA4, the depth migration algorithm adopts Kirchhoff prestack depth migration imaging.

[0075] The basis for Kirchhoff prestack depth migration imaging to achieve amplitude preservation depends on the selection of the migration weight factor. Through this setting, the high-precision second amplitude-preserving migration weight factor obtained by the three-dimensional acoustic wave equation is used in the depth migration imaging algorithm, which is beneficial to improve the interpretability and accuracy of imaging.

[0076] Example 3

[0077] like Figure 6 An embodiment of a method for studying the amplitude attenuation mechanism is shown, which is based on a depth migration profile imaging method of a rugged seabed as in Embodiment 1 or Embodiment 2, and includes the following steps:

[0078] SB1. Perform data dimension reduction on the 3D velocity model, and divide the 3D velocity model into several first characterized velocity models according to the characteristics of three different mechanism factors: structural inclination, abnormal Q value distribution, and abnormal low-velocity body distribution;

[0079] SB2. batch processing the first characterized velocity models through a deep generative model algorithm to generate a large number of second characterized velocity models with different quantitative mechanism factor characteristics;

[0080] SB3, obtaining a second amplitude-preserving migration weight factor of the second characterized velocity model as a theoretical reflection amplitude, and simultaneously performing a third forward simulation on the second characterized velocity model to obtain a second imaging section;

[0081] SB4. Compare the second imaging section with the theoretical reflection amplitude of the second characterized velocity model to obtain an attenuation amplitude;

[0082] SB4. Taking the attenuation amplitude as input, and the different mechanism factor characteristics of the first characterized velocity model and the second characterized velocity model as output, the deep learning model is trained to obtain a large sample model;

[0083] SB5. Predict and study the mechanism characteristics corresponding to a given seafloor attenuation amplitude through a large sample model.

[0084] Among them, the structural inclination is used to characterize the ruggedness of the seabed. Through this setting, the complex three-dimensional velocity model affected by multiple factors can be decomposed into a first characterized velocity model that is only affected by a single mechanism factor characteristic. The mechanism factor characteristic covers the main geological factors of the missing imaging amplitude of the underlying strata in the deep-water rugged seabed. Then, the first characterized velocity model is expanded into a second characterized velocity model with different quantitative mechanism factor characteristics through the depth generation model. Specifically, the second characterized velocity model can be divided into several major categories with different mechanism factor characteristics. Within a single mechanism factor characteristic category, there is a second characterized velocity model with multiple mechanism factor characteristics with different numerical values. After processing, a large number of attenuation amplitudes of the second imaging profile obtained by different quantitative mechanism factor characteristics are obtained, wherein the theoretical reflection amplitude refers to the second amplitude-preserving offset weight factor. Finally, the attenuation amplitude is used as input and different quantitative mechanism factor characteristics are used as output to train a large sample model. The large sample model can predict the influencing mechanism based on the rugged seabed attenuation sample, thereby assisting researchers in conducting mechanism research on the rugged seabed.

[0085] As an embodiment of the present invention, in step SB2, the deep generation model Diffusion is used to batch generate second characteristic velocity models with different quantized structural inclinations, different quantized Q values ​​and different quantized low-velocity bodies.

[0086] As an embodiment of the present invention, in step SB3, a first forward simulation is performed on the second characterized velocity model by a viscoacoustic wave equation simulation algorithm to obtain synthetic seismic data, and then a second imaging corresponding to a different velocity model is generated by a traditional depth migration algorithm.

[0087] As an embodiment of the present invention, the first characterized velocity model and the second characterized velocity model are both two-dimensional velocity models.

[0088] This setting can greatly increase the number of samples, thereby improving the accuracy of large sample models.

[0089] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A method for depth migration profile imaging of a rugged seafloor, characterized in that: The following steps are involved: SA1. Construct a three-dimensional velocity model of the rugged seabed based on the actual surveyed seabed position; SA2, performing a first forward modeling simulation by a wave equation method to obtain a first wave field illumination of the three-dimensional velocity model in step SA1; SA3, performing a second forward simulation on the three-dimensional velocity model by using the first wave field illumination as a first amplitude preserving migration weight factor, and obtaining a second wave field illumination as a second amplitude preserving migration weight factor; SA4. Apply the second amplitude-preserving migration weight factor to a depth migration algorithm to obtain a first imaging section.

2. The depth migration profile imaging method for rugged seabed according to claim 1, characterized in that: The first forward modeling uses a large three-dimensional acoustic wave equation based on a finite difference algorithm.

3. The depth migration profile imaging method for rugged seabed according to claim 2, characterized in that: In step SA2, a first forward modeling is performed with a source-receiver system, wherein the source-receiver system adopts a streamer acquisition system.

4. The depth migration profile imaging method for rugged seabed according to claim 3, characterized in that: The finite difference algorithm is computed via CUDA on large GPU resources.

5. The depth migration profile imaging method for rugged seabed according to any one of claims 1 to 4, characterized in that: In step SA3, the second wave field illumination condition is obtained specifically by the following formula: Among them, K(x) is the illumination condition of the second wave field and also serves as the second amplitude-preserving migration weight factor. |u(x s ,x,t)| 2 is the square value of the illumination of the first wave field in SA2, It is the approximate Green's function of the receiver's reverse extended wave field obtained based on the spatial distribution of the source-receiver system.

6. The depth migration profile imaging method for rugged seabed according to claim 5, characterized in that: In step SA4, the first imaging section is obtained specifically by the following formula: Wherein, K(x) is the second amplitude-preserving offset weight factor calculated in step SA3, is the result obtained by the traditional migration algorithm, and m(x) is the first imaging section.

7. The depth migration profile imaging method for rugged seabed according to claim 1, characterized in that: In step SA4, the depth migration algorithm adopts a reverse time migration imaging method or a Kirchhoff prestack depth migration imaging method.

8. A method for studying the amplitude attenuation mechanism, characterized in that: The steps include: SB1. Perform data dimension reduction on the 3D velocity model, and divide the 3D velocity model into several first characterized velocity models according to the characteristics of three different mechanism factors: structural inclination, abnormal Q value distribution, and abnormal low-velocity body distribution; SB2. batch processing the first characterized velocity models through a deep generative model algorithm to generate a large number of second characterized velocity models with different quantitative mechanism factor characteristics; SB3, obtaining a second amplitude-preserving migration weight factor of the second characterized velocity model as a theoretical reflection amplitude, and simultaneously performing a third forward simulation on the second characterized velocity model to obtain a second imaging section; SB4. Compare the second imaging section with the theoretical reflection amplitude of the second characterized velocity model to obtain an attenuation amplitude; SB5. Taking the attenuation amplitude as input, and taking the different mechanism factor characteristics of the first characterized velocity model and the second characterized velocity model as output, training the deep learning model to obtain a large sample model; SB6. Through the large sample model, predict and study the corresponding mechanism factor characteristics of the attenuation amplitude of a given seabed.

9. The method for studying the amplitude attenuation mechanism according to claim 8, characterized in that: In step SB2, the deep generation model Diffusion is used to batch generate second characteristic velocity models with different quantized structural inclinations, different quantized Q values ​​and different quantized low-velocity bodies.

10. The amplitude attenuation mechanism research method according to claim 8, characterized in that: In step SB3, a third forward simulation is performed on the second characterized velocity model using the viscoacoustic wave equation simulation algorithm to obtain synthetic seismic data, and then a second imaging section corresponding to a different velocity model is generated using a conventional depth migration algorithm.

Images