Inversion velocity modeling methods, devices, electronic equipment and media
By introducing a second-order optimization method using the Hessian operator into the full waveform inversion of reflected waves and updating the direction using the Gauss-Newton method, the resolution and illumination deficiencies of reflected wave information in deep model updates are solved, thereby improving the accuracy and reliability of the velocity model.
Patent Information
- Application Number
- CN202111240045.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-10-25
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2041-10-25
AI Technical Summary
Existing technologies suffer from low resolution and insufficient illumination when using reflected wave information for deep model updates, and conventional optimization methods have slow convergence speed and insufficient accuracy.
A second-order optimization method based on the Hessian operator is adopted, combined with the Gauss-Newton method, to perform full waveform inversion of the reflected wave. The second-order optimization update direction is obtained by truncating the Gauss-Newton method, which improves the convergence and resolution of the inversion process.
It improves the reconstruction accuracy of low and medium wavenumber components of the velocity model, enhances the reliability of the migration velocity model, and finely characterizes the complex fracture system and deep basement morphology within the model.
Smart Images

Figure CN116029073B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of petroleum seismic exploration technology, and more specifically, to an inversion velocity modeling method, apparatus, electronic device, and medium. Background Technology
[0002] As a wave equation inversion method based on waveform matching, Front-In-Time (FWI) can theoretically recover all model wavenumber components from low to high within a seismic frequency band. However, due to the periodicity of seismic wave signals, the objective function of waveform matching has many local extrema. That is, when the time difference between simulated data and observed data is greater than half a period, a "cycle skipping" phenomenon occurs. This requires that the inversion must have a good initial model or include low-frequency and long-offset data. The FWI method usually uses transmitted wave data to recover the low-to-medium wavenumber characteristics of the model, and then uses small-angle reflected wave data to update the high wavenumber components of the model. However, due to the limited penetration depth of transmitted waves, conventional FWI often cannot effectively update the deep background model when the offset range is not wide enough.
[0003] Even for seismic data with limited offsets, reflected waves before the critical angle provide excellent depth illumination, which can be used to recover the low and medium wavenumber components of the deep region in the model. Therefore, how to utilize reflected wave information for deep model updates is receiving increasing attention in exploration geophysics (Wu and Alkhalifah, 2015; Guo and Alkhalifah, 2016). Inspired by the MBTT method (Clement et al., 2001), Xu et al. (2012) proposed the RWI method, which uses reflected wave path information to update the low wavenumber components of the model. Essentially, it obtains a high wavenumber reflectivity or velocity perturbation model using the LSRTM method, then uses inverse migration to obtain predicted reflected wave data, and backprojects the residual between this and the observed reflected wave data onto the reflected wave path to update the background velocity.
[0004] Both FWI and RWI, first-order optimization inversion methods based on updating the model in the negative gradient direction have been widely used, but they often suffer from low accuracy, significant parameter crosstalk, and slow convergence. To address these issues, many researchers have improved the solution to the inverse problem by addressing data or gradient preconditioning, model regularization, and inversion strategies. Among these, second-order optimization methods based on the Hessian operator are theoretically more complete and have greater potential. Due to computational limitations, early methods primarily utilized the diagonal elements of the Hessian as preconditioning operators, effectively compensating for insufficient illumination during inversion and imaging (Shin et al., 2001; Operto et al., 2006). Wang et al. (2016c) replaced diagonal elements with diagonal blocks, taking into account the influence of points near the diagonal elements to some extent and appropriately mitigating parameter crosstalk. However, since these two methods do not fully consider the off-diagonal block information of the Hessian, they struggle to solve problems such as low resolution and insufficient illumination under complex media conditions (e.g., below salt domes). In recent years, more scholars have tended to introduce complete Hessian information into the inversion problem, but these works have focused on FWI (Pratt et al., 1998; Liu et al., 2015; Metivier et al., 2015), while RWI has only recently attracted attention (Cheng et al., 2020).
[0005] Conventional first-order optimization methods such as steepest descent and conjugate gradient are susceptible to ambiguity caused by limitations in observation aperture and data bandwidth, and are even contaminated by multi-parameter crosstalk footprints and multiple scattering effects. Therefore, they converge slowly in the inversion process and have considerable room for improvement in accuracy.
[0006] Therefore, we look forward to an inversion velocity modeling method that can more accurately estimate the low, medium, and high wavenumber components of the medium parameters.
[0007] The information disclosed in the background section of this invention is intended only to enhance the understanding of the general background of this invention, and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art. Summary of the Invention
[0008] The purpose of this invention is to propose an inversion velocity modeling method, device, electronic equipment, and medium, which improves the reconstruction of low and medium wavenumber components of the velocity model and enhances the reliability of the migration velocity model.
[0009] In a first aspect, embodiments of this disclosure provide an inversion velocity modeling method, including: Step 1: Obtain the reflected wave observation data required for inversion, and construct the initial background velocity model and disturbance model for inversion; Step 2: Based on the current background velocity model and disturbance model, use the first-order Born forward modeling method of the acoustic wave equation to obtain the calculation data corresponding to the current velocity model; Step 3: Establish the target functional for full waveform inversion of the reflected wave, and calculate the gradient corresponding to the current background velocity model; Step 4: Based on the gradient, obtain the second-order optimization update direction of the current background velocity model by performing full waveform inversion of the reflected wave using the truncated Gauss-Newton method, so as to update the current background velocity model; Step 5: Repeat steps 2 to 4 until the calculated data and reflected wave observation data meet the set conditions.
[0010] In the optional scheme, the gradient formula corresponding to the current background velocity model is:
[0011] in, The kernel matrix is a sensitive matrix with dimension . ; The gradient is a dimension of . The vector, This is the fitting residual of the reflection data based on the updated perturbation model, caused by errors in the background velocity model. To perform the operation of taking the real part, Let be the objective function. Let be the partial derivative of the objective function with respect to the model. For the initial background velocity model, It is a first-order scattering field. To simulate wave fields, For observational data.
[0012] In the optional scheme, the formula for calculating the update direction is:
[0013] in,
[0014] in, For approximate Hessian matrix, For sampling operators, Angular frequency, For source functions, For the background Green's function field, Let s and r be the coordinates of the source point and the receiver point, respectively, representing the perturbation of the Green's function field. Let be the coordinates of any point in space. Let be the coordinates of any point in space.
[0015] Secondly, embodiments of this disclosure also provide an electronic device, including: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the above-described inversion speed modeling method.
[0016] Thirdly, embodiments of this disclosure also provide a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the aforementioned inversion speed modeling method.
[0017] Fourthly, embodiments of this disclosure also provide an inversion velocity modeling apparatus, comprising: An initial model building module is used to acquire the reflected wave observation data required for inversion and to build the initial background velocity model and disturbance model for inversion. The calculation module is used to obtain the calculation data corresponding to the current background velocity model based on the current background velocity model and the disturbance model, using the first-order Born forward modeling method of the acoustic wave equation. The gradient module is used to establish the target functional for full waveform inversion of the reflected wave and to calculate the gradient corresponding to the current background velocity model. An update module is used to obtain the second-order optimization update direction of the current background velocity model by truncating the Gauss-Newton method full waveform inversion based on the gradient, so as to update the current background velocity model until the calculated data and the observed reflected wave data meet the set conditions.
[0018] In the optional scheme, the gradient formula corresponding to the current background velocity model is:
[0019] in, The kernel matrix is a sensitive matrix with dimension . ; The gradient is a dimension of . The vector, This is the fitting residual of the reflection data based on the updated perturbation model, caused by errors in the background velocity model. To perform the operation of taking the real part, Let be the objective function. Let be the partial derivative of the objective function with respect to the model. For the initial background velocity model, It is a first-order scattering field. To simulate wave fields, For observational data.
[0020] In the optional scheme, the formula for calculating the update direction is:
[0021] in,
[0022] in, For approximate Hessian matrix, For sampling operators, Angular frequency, For source functions, For the background Green's function field, Let s and r be the coordinates of the source point and the receiver point, respectively, representing the perturbation of the Green's function field. Let be the coordinates of any point in space. Let be the coordinates of any point in space.
[0023] The beneficial effects of this invention are as follows: This invention introduces second-order optimization based on the Hessian operator into the full waveform inversion method of reflected waves, proposing a second-order optimized RWI method based on the Gaussian-Newton descent direction. Compared with the commonly used conjugate gradient method, the Gaussian-Newton method using the Hessian matrix improves the convergence and broadband modeling capability of reflected waveform inversion. The resolution bottleneck of commercial software ray travel-time tomography limits the potential of wave equation reverse time migration. The second-order optimized reflected waveform inversion method of this invention improves the reconstruction of low and medium wavenumber components of the velocity model, enhances the reliability of the migration velocity model, and thus leverages the imaging resolution advantage of reverse time migration to finely characterize the complex fracture system and deep basement morphology within the model.
[0024] The methods and apparatus of the present invention have other features and advantages that will be apparent from or will be set forth in detail in the accompanying drawings and following detailed description, which together serve to explain the particular principles of the invention. Attached Figure Description
[0025] The above and other objects, features and advantages of the present invention will become more apparent from the more detailed description of exemplary embodiments of the invention in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same parts.
[0026] Figure 1 A flowchart illustrating the steps of an inversion velocity modeling method according to an embodiment of the present invention is shown.
[0027] Figure 2a The actual velocity model of the SEG / EAGE pushover model is shown.
[0028] Figure 2b The initial velocity model of the SEG / EAGE pushover model is shown.
[0029] Figure 2c The reverse time migration results of the SEG / EAGE pushover model based on the initial model are shown.
[0030] Figure 3a The background velocity model inversion results are shown: the negative gradient direction in the first iteration.
[0031] Figure 3b The background velocity model inversion results are shown: Gauss-Newton direction in the first iteration.
[0032] Figure 3c The background velocity model inversion results are shown: conjugate gradient method inversion results.
[0033] Figure 3d The background velocity model inversion results are shown: Gauss-Newton method inversion results.
[0034] Figure 4 The curve showing the decrease in residual after full waveform inversion of the reflected wave is shown.
[0035] Figure 5a The reverse-time migration image corresponding to the final velocity of the conjugate gradient method is shown.
[0036] Figure 5b The image shown is a reverse-time migration image corresponding to the final velocity in the Gauss-Newton method. Detailed Implementation
[0037] Preferred embodiments of the invention will now be described in more detail. While preferred embodiments of the invention are described below, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.
[0038] Real-world seismic data often lacks effective low-frequency components or long-shot-receiver distance signals, making it difficult for traditional full-waveform inversion methods to obtain long-wavelength components of mid-to-deep parametric models. Given the resolution bottleneck of ray-frame reflection travel-time tomography, there is increasing interest in wave equation reflection wave inversion methods and their practicality. Current research on reflection waveform inversion almost exclusively employs gradient-based first-order optimization methods, which still have significant room for improvement in convergence and resolution. This invention proposes a Hessian operator-optimized reflection wave inversion theory, revealing the defuzzifying effect of the Hessian matrix on the functional gradient and its mechanism for optimizing the background model update direction and improving inversion resolution. Compared to the commonly used conjugate gradient method, the Gauss-Newton method using an approximate Hessian matrix significantly improves the convergence and broadband modeling capabilities of reflection waveform inversion, significantly improves mid-to-deep migration velocity modeling, and enhances the imaging quality of deep basements.
[0039] An embodiment of the present invention provides an inversion velocity modeling method, the method comprising: Step 1: Obtain the reflected wave observation data required for inversion, and construct the initial background velocity model and disturbance model for inversion; Step 2: Based on the current background velocity model and disturbance model, use the first-order Born forward modeling method of the acoustic wave equation to obtain the calculation data corresponding to the current background velocity model; Step 3: Establish the target functional for full waveform inversion of the reflected wave, and calculate the gradient corresponding to the current background velocity model; Step 4: Based on the gradient, obtain the second-order optimization update direction of the current background velocity model by performing full waveform inversion of the reflected wave using the truncated Gauss-Newton method, so as to update the current background velocity model; Step 5: Repeat steps 2 to 4 until the calculated data and reflected wave observation data meet the set conditions.
[0040] In one example, the gradient formula corresponding to the current background velocity model is:
[0041] in, The kernel matrix is a sensitive matrix with dimension . ; The gradient is a dimension of . The vector, This is the fitting residual of the reflection data based on the updated perturbation model and the error caused by the background model. To perform the operation of taking the real part, Let be the objective function. Let be the partial derivative of the objective function with respect to the model. For the initial background velocity model, It is a first-order scattering field. To simulate wave fields, For observational data.
[0042] In one example, the formula for calculating the update direction is:
[0043] in,
[0044] in, For approximate Hessian matrix, For sampling operators, Angular frequency, For source functions, For the background Green's function field, Let s and r be the coordinates of the source point and the receiver point, respectively, representing the perturbation of the Green's function field. Let be the coordinates of any point in space. Let be the coordinates of any point in space.
[0045] Another embodiment of the present invention also provides an electronic device, comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the above-described inversion speed modeling method.
[0046] Another embodiment of the present invention provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the above-described inversion speed modeling method.
[0047] Another embodiment of the present invention also provides an inversion velocity modeling apparatus, comprising: An initial model building module is used to acquire the reflected wave observation data required for inversion and to build the initial background velocity model and disturbance model for inversion. The calculation module is used to obtain the calculation data corresponding to the current background velocity model based on the current background velocity model and the disturbance model, using the first-order Born forward modeling method of the acoustic wave equation. The gradient module is used to establish the target functional for full waveform inversion of the reflected wave and to calculate the gradient corresponding to the current background velocity model. An update module is used to obtain the second-order optimization update direction of the current background velocity model by truncating the Gauss-Newton method full waveform inversion based on the gradient, so as to update the current background velocity model until the calculated data and the observed reflected wave data meet the set conditions.
[0048] In one example, the gradient formula of the gradient module is:
[0049] in, The kernel matrix is a sensitive matrix with dimension . ; The gradient is a dimension of . The vector, This is the fitting residual of the reflection data based on the updated perturbation model and the error caused by the background model. To perform the operation of taking the real part, Let be the objective function. Let be the partial derivative of the objective function with respect to the model. For the initial background velocity model, It is a first-order scattering field. To simulate wave fields, For observational data.
[0050] In one example, the formula for calculating the update direction is:
[0051] in,
[0052] in, For approximate Hessian matrix, For sampling operators, Angular frequency, For source functions, For the background Green's function field, Let s and r be the coordinates of the source point and the receiver point, respectively, representing the perturbation of the Green's function field. Let be the coordinates of any point in space. Let be the coordinates of any point in space.
[0053] Conventional first-order optimization methods, such as steepest descent and conjugate gradient, are susceptible to ambiguity caused by limitations in observation aperture and data bandwidth, and are even affected by multi-parameter crosstalk footprints and multiple scattering effects. Therefore, they converge slowly during the inversion process, and there is significant room for improvement in accuracy. This invention introduces the Hessian operator to precondition the functional gradient using a second-order optimization approach, thereby improving convergence, increasing resolution, suppressing multi-parameter coupling, and more accurately estimating the low, medium, and high wavenumber components of the medium parameters, providing better support for Earth interior exploration and oil and gas exploration and development.
[0054] This invention introduces second-order optimization based on the Hessian operator into the full waveform inversion method of reflected waves, proposing a second-order optimized RWI method based on the Gaussian-Newton descent direction. Compared with the commonly used conjugate gradient method, the Gaussian-Newton method using the Hessian matrix improves the convergence and broadband modeling capability of reflected waveform inversion. The resolution bottleneck of commercial software ray travel-time tomography limits the potential of wave equation reverse time migration. The second-order optimized reflected waveform inversion method of this invention improves the reconstruction of low and medium wavenumber components of the velocity model, enhances the reliability of the migration velocity model, and thus leverages the imaging resolution advantage of reverse time migration to finely characterize the complex fracture system and deep basement morphology within the model.
[0055] To facilitate understanding of the solutions and effects of the embodiments of the present invention, three specific application examples are given below. Those skilled in the art should understand that these examples are merely for the purpose of understanding the present invention, and any specific details therein are not intended to limit the present invention in any way.
[0056] Example 1
[0057] Figure 1 A flowchart illustrating the steps of an inversion velocity modeling method according to an embodiment of the present invention is shown.
[0058] like Figure 1 As shown, the inversion velocity modeling method includes: Step 1: Obtain the reflected wave observation data required for inversion, and construct the initial background velocity model and perturbation model for inversion; Step 2: Based on the current background velocity model and perturbation model, use the first-order Born forward modeling method of the acoustic wave equation to obtain the calculation data corresponding to the current background velocity model; Step 3: Establish the target functional for the full waveform inversion of reflected waves, and calculate the gradient corresponding to the current background velocity model; Step 4: Obtain the second-order optimization update direction of the current background velocity model based on the gradient and the full waveform inversion of reflected waves using the truncated Gauss-Newton method, so as to update the current background velocity model; Step 5: Repeat steps 2 to 4 until the calculation data and the reflected wave observation data meet the set conditions.
[0059] The first step is to obtain the calculation data corresponding to the current background velocity model based on the first-order Born forward model of the acoustic wave equation. Assuming the disturbance component is much smaller than the background component, the total wave field can be decomposed into the incident field. With first-order scattering field The sum of When underground disturbances have a certain continuity and spatial distribution, the scattered wave field The superposition of these waves forms a reflected wave, which is recorded by a limited series of detectors on the Earth's surface, forming computational data.
[0060] The second step is to establish the target functional for the full waveform inversion of the reflected wave and calculate the gradient corresponding to the current background velocity model. The conventional steepest descent (SD) method and conjugate gradient (CG) method are the most commonly used first-order optimization methods. The former updates the model along the negative gradient direction, where the gradient formula corresponding to the current background velocity model is the first-order partial derivative of the error function with respect to the model parameters, satisfying: (1) in To perform the operation of taking the real part, Based on reflection data about model components The first-order partial derivatives constitute the Fréchet derivative or sensitive kernel (matrix), with dimension 1. Functional gradient The dimension is The vector, This represents the residual of the reflection data fitting, based on the updated perturbation model and caused by errors in the background velocity model.
[0061] Step 3: Obtain the second-order optimization update direction of the current background velocity model based on the full waveform inversion of the reflected wave using the truncated Gauss-Newton method. In seismic imaging and velocity modeling, even for two-dimensional single-parameter problems, the model dimension is at least on the order of 10⁶, making the Hessian matrix extremely large and unsuitable for explicit calculation and inversion. Therefore, the update direction is generally obtained by solving Newton's equations. Generally, the Hessian matrix can be decomposed into two terms. When the data residuals are large or the inverse problem is highly nonlinear, the effect of the second term of the Hessian operator cannot be ignored. Under the assumption of local linearity, the following Gauss-Newton equation can be solved: (2) Obtain the second-order optimization update direction of the current background velocity model in the full waveform inversion of the reflected wave: (3) Because the approximate Hessian matrix H possesses positive definiteness, the Gauss-Newton method can always provide a reasonable descent direction and avoid the computational complexity introduced by the nonlinear terms of the Hessian operator. The Gauss-Newton equations are often solved iteratively using the conjugate gradient (CG) method. The inversion algorithm consists of internal iterations (solving the Newton or Gauss-Newton equations) and external iterations (constructing the functional gradient and updating the model). The number of wave equation solutions involved in one iteration of the internal conjugate gradient method and one iteration of the external method is roughly equivalent.
[0062] Step 4: Update the current background velocity model and optimize the high wavenumber perturbation model. Determine if convergence has occurred (the calculated data and the observed reflected wave data are highly equal). If convergence has occurred, the inversion process ends.
[0063] refer to Figures 2a to 5b Through model data testing, this technology can reconstruct a high-precision background velocity model and improve the quality of stratigraphic imaging. Figure 2a The real-world background velocity model was used as the benchmark for this test. Figure 2b This is the constructed initial background velocity model. Because the initial background velocity model differs significantly from the true background velocity model, the corresponding reverse time migration imaging results ( Figure 2c The depiction of the strata is very chaotic, and the stratigraphic structure is not accurate enough. Figure 3a and 3b These represent the negative gradient direction in the first iteration using the conjugate gradient method and the Gauss-Newton direction in the first iteration using the method of this patent, respectively. Figure 3a It is evident that the mid-to-deep regions of the gradient are poorly illuminated, with uneven energy distribution between the upper and lower parts, resulting in strong artifacts. Furthermore, the coupling blurring effect between spatial points is significant, leading to insufficient resolution. However, when Hessian information is incorporated using the method described in this patent, the Gaussian-Newton direction becomes more balanced across the shallow, mid-to-deep regions, and the resolution of each layer is enhanced. Figure 3c and 3d The results of the inversion of the background velocity model using the conjugate gradient method and the truncated Gauss-Newton method are shown respectively. It can be seen that the former has a good effect on the update of shallow layers, but the correction of medium and deep models is not significant. The latter has higher resolution and is closer to the real model, whether in shallow or deep layers. Figure 4 The convergence curve of the objective function is shown. It can be seen that the second-order reflected wave waveform inversion method based on Gauss-Newton proposed in this patent has better convergence. Moreover, under the same number of iterations, its data residual is smaller, which improves the modeling accuracy of the background velocity model while saving computational costs. Figure 5a and 5b These are the final velocity obtained by the conjugate gradient method and the inverse time migration image corresponding to the inversion method of this patent. It can be seen that the background velocity model obtained by the inversion method of this patent can obtain higher accuracy formation imaging results and effectively improve the ability to predict the distribution of underground oil reservoirs.
[0064] Example 2
[0065] This disclosure also provides an electronic device, which includes: At least one processor; and, A memory that is communicatively connected to at least one processor; wherein, The memory stores instructions that can be executed by at least one processor, which enables the at least one processor to perform the aforementioned inversion speed modeling method.
[0066] An electronic device according to an embodiment of the present disclosure includes a memory and a processor.
[0067] This memory is used to store non-transitory computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.
[0068] The processor may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In one embodiment of this disclosure, the processor is used to execute computer-readable instructions stored in the memory.
[0069] Those skilled in the art will understand that, in order to solve the technical problem of how to achieve a good user experience, this embodiment may also include well-known structures such as communication buses and interfaces, and these well-known structures should also be included within the protection scope of this disclosure.
[0070] This disclosure also provides a non-transitory computer-readable storage medium that stores computer instructions for causing a computer to execute the above-described inversion speed modeling method.
[0071] Example 3
[0072] This disclosure provides a non-transitory computer-readable storage medium that stores computer instructions for causing a computer to execute the aforementioned inversion speed modeling method.
[0073] A computer-readable storage medium according to embodiments of the present disclosure stores non-transitory computer-readable instructions. When these non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the methods described in the foregoing embodiments of the present disclosure are performed.
[0074] The aforementioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or portable hard drive), media with built-in rewritable non-volatile memory (e.g., memory card), and media with built-in ROM (e.g., ROM cartridge).
[0075] Those skilled in the art should understand that the above description of the embodiments of the present invention is only intended to illustrate the beneficial effects of the embodiments of the present invention, and is not intended to limit the embodiments of the present invention to any of the examples given.
[0076] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.
Claims
1. A method for inverting velocity modeling, characterized in that, include: Step 1: Obtain the reflected wave observation data required for inversion, and construct the initial background velocity model and disturbance model for inversion; Step 2: Based on the current background velocity model and disturbance model, use the first-order Born forward modeling method of the acoustic wave equation to obtain the calculation data corresponding to the current background velocity model; Step 3: Establish the target functional for full waveform inversion of the reflected wave, and calculate the gradient corresponding to the current background velocity model; the gradient formula corresponding to the current background velocity model is: in, To perform the operation of taking the real part, The kernel matrix is a sensitive matrix with dimension . ; The gradient is a dimension of . The vector, This is the fitting residual of the reflection data based on the updated perturbation model, caused by errors in the background velocity model. Let be the objective function. Let be the partial derivative of the objective function with respect to the model. For the initial background velocity model, It is a first-order scattering field. To simulate wave fields, For observational data; Step 4: Based on the gradient, obtain the second-order optimization update direction of the current background velocity model through full waveform inversion using the truncated Gauss-Newton method, and update the current background velocity model; the calculation formula for the update direction is: in, in, For approximate Hessian matrix, For sampling operators, Angular frequency, For source functions, For the background Green's function field, Let s and r be the coordinates of the source point and the receiver point, respectively, representing the perturbation of the Green's function field. Let be the coordinates of any point in space. Let the coordinates be any point in space; Step 5: Repeat steps 2 to 4 until the calculated data and reflected wave observation data meet the set conditions.
2. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the inversion velocity modeling method of claim 1.
3. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing a computer to execute the inversion velocity modeling method of claim 1.
4. An inversion velocity modeling device, characterized in that, include: An initial model building module is used to acquire the reflected wave observation data required for inversion and to build the initial background velocity model and disturbance model for inversion. The calculation module is used to obtain the calculation data corresponding to the current background velocity model based on the current background velocity model and the disturbance model, using the first-order Born forward modeling method of the acoustic wave equation. The gradient module is used to establish the target functional for full waveform inversion of the reflected wave and calculate the gradient corresponding to the current background velocity model; the gradient formula corresponding to the current background velocity model is: in, To perform the operation of taking the real part, The kernel matrix is a sensitive matrix with dimension . ; The gradient is a dimension of . The vector, This is the fitting residual of the reflection data based on the updated perturbation model, caused by errors in the background velocity model. Let be the objective function. Let be the partial derivative of the objective function with respect to the model. For the initial background velocity model, It is a first-order scattering field. To simulate wave fields, For observational data; The update module is used to obtain the second-order optimization update direction of the current background velocity model by truncating the full waveform inversion of the reflected wave using the Gauss-Newton method according to the gradient, so as to update the current background velocity model; until the calculated data and the observed reflected wave data meet the set conditions; The formula for calculating the update direction is: in, in, For approximate Hessian matrix, For sampling operators, Angular frequency, For source functions, For the background Green's function field, Let s and r be the coordinates of the source point and the receiver point, respectively, representing the perturbation of the Green's function field. Let be the coordinates of any point in space. Let be the coordinates of any point in space.
Citation Information
Patent Citations
Method for inverting low- and medium-wave number components in velocity field through reflection wave information
CN104391323A
Pre-stack seismic multi-parameter inversion method of shale reservoirs
CN111025387A
