Forward modeling method and device based on wave field information
By generating seismic wavelet spectrum and optimizing the difference coefficient, the problem of large error in the finite difference method under parameter conditions is solved, and high-precision forward simulation under all conditions is realized, showing high simulation accuracy and adaptation in complex media sex.
Patent Information
- Application Number
- CN202410832716.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-26
- Publication Date
- 2024-11-08
AI Technical Summary
The existing finite difference method will produce greater spatial simulation errors under certain parameter conditions, making it difficult to output high-precision forward simulation results under all parameter conditions.
By obtaining input parameters such as wavelet main frequency, time step, space step and difference order, the spectrum of seismic wavelet is generated, and the objective function is iteratively optimized, the optimization difference coefficient is determined, and then the finite difference solution is solved in finite difference. Consider seismic wave field information to generate high-precision forward simulation results.
More accurate simulation results can be obtained under any parameter conditions, showing high simulation accuracy and adaptability in complex media.
Smart Images

Figure CN118915135A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of geophysical exploration, and in particular to a forward modeling method and device based on wave field information. Background Art
[0002] The finite difference method is one of the main methods for simulating wave equations. By adding wave field information, the finite difference coefficients can be effectively improved to suppress spatial simulation errors.
[0003] Spatial domain difference operator optimization is a very important type of difference operator optimization method. This type of method is proposed to suppress the spatial simulation error generated during the simulation process. However, under certain parameter conditions, the optimized finite difference coefficients will produce larger spatial simulation errors during the simulation process.
[0004] Based on this, a forward modeling method is needed that can output high-precision simulation results under all parameter conditions. Summary of the invention
[0005] The embodiments of this specification provide a forward modeling method and device based on wave field information to solve the following technical problem: a forward modeling method solution that can output high-precision simulation results under all parameter conditions is required.
[0006] To solve the above technical problems, one or more embodiments of this specification are implemented as follows:
[0007] In a first aspect, an embodiment of the present specification provides a forward modeling method based on wave field information, comprising:
[0008] Get input parameters, including the sub-wave frequency f m , time step τ, space step h, average speed v ave and the difference order M; generating the spectrum U of the seismic wavelet according to the wavelet main frequency and time step w (k) Among them, ricker_wavelet represents the Ricker wavelet generating function, fft represents the fast Fourier transform, h represents, V ave Iteration optimization is performed on the objective function according to the spectrum of the seismic wavelet to determine the optimal differential coefficient c m , the objective function is in the form of Wherein, k represents, and subscript 2 represents the second norm: according to the input parameters and the differential coefficients, the equation Perform differential solution to generate forward simulation results, where x is the horizontal coordinate and z is the vertical coordinate. is the wave field value at time t at coordinate (x,z).
[0009] In a second aspect, one or more embodiments of this specification provide an electronic device, including:
[0010] at least one processor; and,
[0011] a memory communicatively connected to the at least one processor; wherein,
[0012] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to the first aspect.
[0013] At least one of the above technical solutions adopted in one or more embodiments of this specification can achieve the following beneficial effects: by obtaining input parameters, the input parameters include the sub-wave main frequency f m , time step τ, space step h, average speed v ave and the difference order M; generating the spectrum U of the seismic wavelet according to the wavelet main frequency and time step w (k) Iteratively optimize the objective function according to the spectrum of the seismic wavelet to determine the optimized differential coefficient c m , solve the equation by differential according to the input parameters and the differential coefficients to generate the forward simulation results, where x is the horizontal coordinate and z is the vertical coordinate. is the wave field value at time t at coordinate (x, z). By considering the seismic wave field information in the process of calculating the finite difference coefficient, a new finite difference coefficient can be output. Compared with the differential optimization algorithm, this method can obtain more accurate simulation results under any parameter conditions. Practical application also proves that the differential operator obtained by this method can have a good simulation accuracy under any parameter; the simulation process also proves the adaptability and high simulation accuracy of this method for complex media. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0015] Figure 1 A schematic diagram of a flow chart of a forward modeling method based on wave field information provided in an embodiment of this specification;
[0016] Figure 2 The spatial simulation error measurement diagram of the method provided in the embodiments of this specification and the Taylor expansion method and the Remitz exchange algorithm;
[0017] Figure 3a , Figure 3b and Figure 3c A set of wave field snapshot residual comparison diagrams provided in the embodiments of this specification, wherein: Figure 3a is the result of Remitz Exchange Algorithm (REA), Figure 3b is the result of Taylor expansion method (TES), Figure 3c The results of the methods provided in the embodiments of this specification;
[0018] Figure 4a , Figure 4b and Figure 4c Another set of wave field snapshot residual comparison diagrams provided in the embodiments of this specification, wherein: Figure 4a is the result of Remitz Exchange Algorithm (REA), Figure 4b is the result of Taylor expansion method (TES), Figure 4c The results of the methods provided in the embodiments of this specification;
[0019] Figure 5a , Figure 5b and Figure 5c A set of shot record residual comparison diagrams provided in the embodiments of this specification; wherein, Figure 5a is the result of Remitz Exchange Algorithm (REA), Figure 5b is the result of Taylor expansion method (TES), Figure 5c The method results provided by the embodiments of this specification;
[0020] Figure 6a , Figure 6b and Figure 6c Another set of shot record residual comparison diagrams provided in the embodiments of this specification, wherein: Figure 6a is the result of Remitz Exchange Algorithm (REA), Figure 6b is the result of Taylor expansion method (TES), Figure 6c The method results provided by the embodiments of this specification;
[0021] Figure 7 A schematic diagram of an electronic device provided in an embodiment of this specification. DETAILED DESCRIPTION
[0022] The embodiments of this specification provide a forward modeling method and device based on wave field information.
[0023] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.
[0024] First, the parameters that need to be obtained in the embodiments of this specification are described.
[0025] Seismic wave forward modeling is the basis of imaging. In the imaging process, the parameters required for the forward modeling of the acoustic wave equation of isotropic media are the velocity field, the size of the time step, the size of the space step, and the size of the main frequency. The main frequency is based on the main frequency of the actual field data, generally 10Hz to 40Hz, generally with f m Indicated. The velocity field is the longitudinal wave velocity value of the initial model, generally represented by v. The size of the time step determines how long the time interval is to record the wave field value, generally 0.1ms to 2ms, represented by τ. The size of the spatial step is determined by the distance of the detection points collected in the field, generally 5m to 20m, represented by h. In the process of using the model to simulate the acoustic wave equation of isotropic media, the velocity field is the model velocity field. The size of the time step is as small as possible to reduce the interference of time dispersion, so the time step is generally 0.1ms. The spatial step is determined by the grid size in the model discretization process. After determining the required parameters, it is also necessary to determine the order M of the finite difference method. The order of the finite difference method can be determined based on the computing memory of the device used for calculation. The larger the M, the larger the computing memory required, and the smaller the M, the smaller the computing memory required. After all parameters are determined, the difference coefficients can be optimized.
[0026] The optimized differential coefficients involved in the embodiments of this specification are explained.
[0027] First, the differential discretization format of the second-order spatial partial derivative is given,
[0028]
[0029] Where u is the scalar wave field function, c m is the difference coefficient, and h is the grid size.
[0030] According to the plane wave theory, we can get:
[0031] u m =e i[k(x+mh)] (2)
[0032] Substituting formula 2 into formula 1, and then substituting into Euler's formula and simplifying, we can convert the second-order spatial partial derivative to the wave number domain
[0033]
[0034] Where U(k) is the seismic wave field u in the wave number domain. When k = 0, Therefore, Formula 3 can be expressed as
[0035]
[0036] Formula 4 can be simplified to
[0037] B a U(k)≈U(k)B n (5)
[0038] Among them B a is the real wave number, which is -(kh) 2 , B n is the numerical wave number, which is So the spatial simulation error can be derived as
[0039]
[0040] Where E represents the spatial simulation error, which is the dot product integral of the wave field in the wavenumber domain and the spatial dispersion curve. max is the maximum wave value. According to the Nyquist theorem, kh is from 0 to π. So, k max equal Formula 6 can be written as
[0041]
[0042] In theory, we should minimize E to optimize the finite difference weights to reduce simulation errors. However, in previous studies, more attention was paid to minimizing the error of the dispersion relation, that is, minimizing the error of B(cm,k), rather than E to obtain the finite difference coefficients, ignoring the influence of the wave field U(k). Therefore, in some simulation experiments, the optimized finite difference weights will produce larger spatial simulation errors.
[0043] To minimize E, we construct a two-norm objective function
[0044]
[0045] U(k) is the spectrum of the seismic wave field. Iteratively updating U(k) means performing multiple forward simulations. Even with the fastest forward simulation algorithm, the computational cost is still unacceptable. In actual numerical modeling, the spectrum of the seismic wavelet is very close to the spectrum of the seismic wave field. In order to solve this optimization problem, we replace the spectrum of the seismic wavelet U w (k) is replaced by the spectrum U(k) of the seismic wave field. Therefore, when the simulation parameters are determined, U w (k) becomes a known vector instead of a variable. Therefore, the new optimization problem can be reformulated as an optimization based on the spectrum of the seismic wavelet:
[0046]
[0047] For the iterative optimization of the aforementioned objective function, a conventional sampling method can be used to iterate the optimized differential coefficients to minimize the simulation error.
[0048] This application also provides an approximate sampling method to iterate the optimization differential coefficients. We use the following pseudo code process to solve the optimization problem
[0049] Algorithm 1.
[0050]
[0051] where f m is the main frequency of the wavelet, τ is the time step, and M is the difference order, k is the wave number, f is the frequency, fft represents the fast Fourier transform, ricker_wavelet represents the program for generating Ricker wavelets, E represents the spatial simulation error, B represents the spatial dispersion relation, and SAM represents a sampling approximation method. Next, we will introduce the detailed process of SAM.
[0052] The special M wave number is required in SAM sampling, which can be regarded as the sampling of SAM. The special M wave number can be expressed as
[0053] b=b i -p(10)
[0054] Among them, b, b j =k i h(j=1, 2, 3, ..., M) represents the special M wave number, b j are different from each other, b i is the initial wave number obtained by conventional optimization algorithms (such as REA, LSM and ADMM), and p is a positive value. By substituting the special M wave number into the spatial dispersion relation (B(cm, k)), we can express the resulting linear equations as:
[0055]
[0056] Formula 11 can be rewritten as:
[0057]
[0058] Formula 12 can be written as
[0059] A(b)c=d (13)
[0060] Where A∈R^(M×M). There is a fact about the rank of A. If the rank of A and the rank of the augmented matrix (Ad) are both equal to M, then according to the theory of linear algebra, the system of linear equations (Equation 13) has one and only one solution. Therefore, a set of FD weights can only be calculated by SAM for a special set of M wave numbers. The solution of SAM is analytical in nature, unlike the solutions of traditional optimization algorithms (REA, LSM, ADMM) which are affected by the initial parameters or the number of iterations. The FD weights of SAM can be obtained by the following formula:
[0061] C=A- 1 (b) D(14),
[0062] Among them, C is the number containing c 1 to c m is an m*1 vector of value A, and D is an m*1 vector of value -1.
[0063] The sampling method in the present application can provide more sets of differential coefficients under the condition of fixed error limit by adjusting bi and p in formula 10, while the traditional method can only provide one set of differential coefficients under the condition of fixed error limit.
[0064] The above process explains the working principle, objective function and sampling method of the present application. Based on the above, in the first aspect, the present application provides a forward modeling method based on wave field information, such as Figure 1 As shown, Figure 1 A schematic flow chart of a forward modeling method based on wave field information provided in one or more embodiments of this specification.
[0065] Figure 1 The process in may include the following steps:
[0066] S101: Obtain input parameters, which include the wavelet main frequency f m , time step τ, space step h, average speed v ave and the difference order M.
[0067] S103, generating a spectrum U of the seismic wavelet according to the wavelet main frequency and time step w (k).
[0068] At this time, we can use the above algorithm 1 to generate the spectrum of the seismic wavelet using the following formula: Among them, ricker_wavelet represents the Ricker wavelet generating function, and fft represents the fast Fourier transform. Since the computer cannot calculate continuous variables or continuous functions during the calculation process, the continuous variables and continuous functions must be discretized. The finite difference method replaces the differential quotient with the difference quotient, and the discrete function value with the continuous function value, so that the continuous well-defined problem is discretized so that it can be simulated and calculated in the computer. In order to make the continuous acoustic wave equation simulated and calculated in the computer, the finite difference method is used to discretize the second-order continuous derivative in the acoustic wave equation.
[0069] The two-dimensional acoustic wave equation is
[0070]
[0071] Where ricker_wavelet is the Ricker wavelet function and v is the velocity field.
[0072] Discretize the second-order continuous derivative in the two-dimensional acoustic wave equation into:
[0073]
[0074] in is the wave field value at time t at coordinate (x,z).
[0075] S105, iteratively optimizing the objective function according to the spectrum of the seismic wavelet to determine the optimized differential coefficient c m .
[0076] The objective function is in the form of Wherein, k represents, and the subscript 2 represents the second norm. The specific calculation method for cm can adopt the conventional method, or the method of Algorithm 1 provided in the aforementioned application. The traditional method can only provide a set of differential coefficients under a fixed error limit, and the method of the present application can provide more sets of differential coefficients under the condition of a fixed error limit by adjusting bi and p in Formula 10, and provide more optimized differential coefficient options.
[0077] S107, performing differential solution on the equation according to the input parameters and the differential coefficients to generate a forward simulation result.
[0078] According to the input parameters obtained above and the differential coefficients obtained, formula (2) can be solved to obtain the wave field values at different times and locations. Combining these values in a specific way can obtain different forms of observation data. For example, taking the wave field values of all locations at a certain time is the wave field snapshot; taking the wave field values of a certain location at all times is the waveform diagram of that point.m It is the differential coefficient. The wave field information is very important for the selection of the differential coefficient. The wave field information is introduced in the embodiment of this specification. By obtaining the differential coefficient and reconstructing the forward modeling scheme, a more accurate simulation of the seismic wave field can be achieved.
[0079] Figure 2 A schematic diagram of the comparison of the spatial simulation error calculation diagram of the method (NEW) provided in the embodiment of this specification with the Taylor expansion method (TES) and the Remitz exchange algorithm (REA), the parameters of the simulation are the main frequency of 40Hz, the spatial step length of 5m, the time step length of 0.1ms, and the speed of 2000m / s. By comparing with the existing methods, it can be found that the spatial simulation error of the method provided in this application is the smallest. The spatial simulation error of the Remitz exchange algorithm is larger than that of the traditional Taylor expansion method, which means that the result after the optimization algorithm will produce a larger error. This is because the Remitz exchange algorithm only calculates the differential coefficient by minimizing the dispersion relation error, but does not minimize the spatial simulation error.
[0080] Figure 3a , Figure 3b and Figure 3c A set of wave field snapshot residual comparison diagrams of the method provided in the embodiment of this specification, the Remez exchange algorithm (REA) and the traditional Taylor expansion method (TES), wherein: Figure 3a is the result of Remitz Exchange Algorithm (REA), Figure 3b is the result of Taylor expansion method (TES), Figure 3c The results of the method provided in the embodiment of this specification. The parameters of this simulation are the main frequency of 40Hz, the spatial step size of 5m, the time step size of 0.1ms, the velocity model is a velocity of 2000m / s, the size is 1000*1000 grid points. The propagation time is 1s. The wavelet injection point is (2500m, 2500m). The parameters of this simulation are the same as Figure 2 The simulation parameters in are the same. By comparing with the existing methods, it can be found that the simulation error of the method provided by this application is the smallest, while the spatial simulation error of the Remitz exchange algorithm is larger than that of the traditional Taylor expansion method, which is consistent with our Figure 2 This is consistent with the observations in , indicating that minimizing the spatial simulation error (E) is an effective way to reduce the simulation error.
[0081] Figure 4a , Figure 4b and Figure 4c Another set of wavefield snapshot residual comparison diagrams of the method provided in the embodiment of this specification, the Remez exchange algorithm (REA) and the traditional Taylor expansion method (TES), wherein: Figure 4a is the result of Remitz Exchange Algorithm (REA), Figure 4b is the result of Taylor expansion method (TES), Figure 4cThe results of the method provided in the embodiment of this specification. The parameters of this simulation are the main frequency of 40Hz, the spatial step size of 6m, the time step size of 0.1ms, the velocity model is a velocity of 2000m / s, the size is 1000*1000 grid points. The propagation time is 1s. The wavelet injection point is (3000m, 3000m). Most of the parameters of this simulation are the same as Figure 2 The simulation parameters in are the same, except that the spatial step size is changed to 6m. By comparing with the existing methods, it can be found that the simulation error of the method provided in this application is still the smallest. These two sets of simulation results show that the new method can stably give the best simulation results in the uniform model, indicating that it can output the optimal results under various parameters.
[0082] Figure 5a , Figure 5b and Figure 5c A comparison diagram of a set of shot record residuals of the method provided in the embodiment of this specification, the Remitz exchange algorithm (REA) and the traditional Taylor expansion method (TES). Among them, Figure 5a is the result of Remitz Exchange Algorithm (REA), Figure 5b is the result of Taylor expansion method (TES), Figure 5c The method results provided in the embodiments of this specification. The parameters of the simulation are main frequency 40Hz, spatial step size 5m, time step size 0.1ms, and the velocity model is a marmousi model with a size of 1360*700 grid points. The propagation time is 1.2s. The wavelet injection point is (680m, 1m). The spatial simulation error of the Remitz exchange algorithm is larger than that of the traditional Taylor expansion method. Among the three methods, the spatial simulation error of the method provided in this application is the smallest, indicating that minimizing E can also obtain a more accurate simulated wave field in complex media.
[0083] Figure 6a , Figure 6b and Figure 6c Another set of shot record residual comparison diagrams of the method provided in the embodiment of this specification, the Remitz exchange algorithm (REA) and the traditional Taylor expansion method (TES), wherein: Figure 6a is the result of Remitz Exchange Algorithm (REA), Figure 6b is the result of Taylor expansion method (TES), Figure 6cThe method results provided by the embodiments of this specification. The parameters of the simulation are a main frequency of 40Hz, a spatial step of 5m, a time step of 0.1ms, and a velocity model of a marmousi model with a size of 1360*700 grid points. The propagation time is 1.2s. The wavelet injection point is (680m, 1m). Most of the parameters of this simulation are the same as the simulation parameters in Figure 5, except that the main frequency is changed to 50Hz. By comparing with the existing methods, it can be found that the simulation error of the method provided by the present application is still the smallest. These two groups of simulation results show that the method provided by the present application can stably give the best simulation results in complex models, indicating that it can output the optimal results under multiple parameters.
[0084] In a second aspect, the present application also provides a computer device. Figure 7 As shown, Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of this specification, the device includes:
[0085] at least one processor; and,
[0086] a memory communicatively connected to the at least one processor; wherein,
[0087] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to the first aspect.
[0088] Based on the same idea, an embodiment of this specification also provides a non-volatile computer storage medium corresponding to the above method, which stores computer executable instructions. When a computer reads the computer executable instructions in the storage medium, the instructions enable one or more processors to execute the method described in the first aspect.
[0089] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0090] The above description is only one or more embodiments of this specification and is not intended to limit this specification. For those skilled in the art, one or more embodiments of this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of this specification shall be included in the scope of the claims of this specification.
Claims
1. A forward modeling method based on wave field information, comprising: Get input parameters, including the sub-wave frequency f m , time step τ, space step h, average speed v ave and the difference order M; Generate the spectrum U of the seismic wavelet according to the wavelet main frequency and time step w (k) Among them, ricker_wavelet represents the Ricker wavelet generating function, fft represents the fast Fourier transform, h represents, V ave express The objective function is iteratively optimized according to the spectrum of the seismic wavelet to determine the optimized differential coefficient c m , the objective function is in the form of Among them, k represents and the subscript 2 represents the second norm: According to the input parameters and the differential coefficients, the equation Perform differential solution to generate forward simulation results, where x is the horizontal coordinate and z is the vertical coordinate. is the wave field value at time t at coordinate (x, z), v is the velocity field, m is the counter, and h is the spatial distance of the grid.
2. The method of claim 1, wherein: The objective function is iteratively optimized according to the spectrum of the seismic wavelet to determine the optimized differential coefficient c m ,include: Calculate the spatial simulation error E of the objective function, and when E does not meet the preset error range, update the optimized differential coefficient c m , the update method is C = A -1 (b) D, D is an m*1 vector with a value of -1, in, k j is the wave number.
3. An electronic device, comprising: at least one processor; as well as, 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, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 2.
