Seismic data matching method based on physical constraint self-supervised learning, electronic equipment and storage medium

Abstract

The invention discloses a seismic data matching method based on physical constraint self-supervised learning, electronic equipment and a storage medium, and belongs to the technical field of seismic data processing. In order to improve the matching performance of seismic data, the method comprises the following steps: performing first-order Taylor expansion on a relational expression of a matching function and a velocity ratio of a PP wave to a PS wave to obtain a discrete expression, and constructing a physical constraint to obtain a matching function matrix equation; for the matching function matrix equation, constructing a parameterized matching function by considering the boundary condition of the velocity ratio of the PP wave to the PS wave; constructing an objective function, and then substituting the parameterized matching function into the objective function to obtain a matched objective function expression; and network training solving is carried out, and matching function solving based on physical constraints is carried out by using a neural network U-net and an adaptive moment estimation Adam method. According to the method, the problem of non-stationary seismic data matching is effectively solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of seismic data processing, and in particular relates to a seismic data matching method, electronic equipment and storage medium based on physical constraint self-supervised learning. Background Art

[0002] As important energy sources for national economic development, the demand for oil and natural gas is increasing day by day. As an important means of energy exploration, seismic exploration technology plays an important role in the energy collection process.

[0003] Multi-wave exploration technology makes full use of the different characteristics of longitudinal and shear waves propagating in oil and gas-bearing media and their rich wave field information, and has advantages in gas cloud imaging and concealed oil and gas reservoir description. In order to give full play to the advantages of multi-wave exploration, the matching problem of seismic signals must be solved first. Seismic data matching is a technical means to effectively solve this problem by matching the target seismic data with the data to be matched through matching functions.

[0004] Current seismic data matching technologies can be divided into traditional methods, such as automatic seismic data matching, optimization-based matching, and machine learning-based methods. The objective function of the automatic seismic data matching method is to achieve seismic data matching based on the local cross-correlation between the target and the matched seismic data. This method cannot achieve a better matching effect for disturbances such as noise, amplitude, phase, and frequency in the seismic data. In the optimization-based matching method, the purpose of these objective functions is to minimize the difference between the target seismic data and the matched seismic data, usually using l 2 norm. These optimization problems are usually nonlinear, ill-posed, and may have multiple local minima. Generally, a multi-step approach is used to achieve matching, such as using preprocessing methods to balance amplitude and frequency differences; performing a rough match before the final match. In recent years, with the popularity of deep learning technology, data-driven methods have occupied a dominant position. This type of method directly learns a nonlinear mapping from the target seismic data and the seismic data to be matched to the matching function through a large amount of labeled data. Deep learning methods do not require mathematical and physical assumptions and have high computational efficiency, but the matching ability is more dependent on the type of training data and has poor generalization ability.

[0005] In recent years, self-supervised methods that do not rely on true matching functions have attracted increasing attention. The input is fixed noise and the output is a matching function. This type of method has improved the generalization ability of deep learning methods to a certain extent. However, deep learning methods still have a major flaw. Seismic data has non-stationary factors such as amplitude, phase and frequency disturbances.

[0006] Directly matching seismic data using neural networks may distort the waveform of the matched seismic data. Summary of the invention

[0007] The problem to be solved by the present invention is to improve the matching performance of seismic data, and propose a seismic data matching method, electronic equipment and storage medium based on physical constraint self-supervised learning.

[0008] To achieve the above object, the present invention is implemented through the following technical solutions:

[0009] A seismic data matching method based on physical constraint self-supervised learning includes the following steps:

[0010] S1. The relationship between the matching function and the velocity ratio of the PP wave to the PS wave is obtained by using the first-order Taylor expansion to obtain a discrete expression, and physical constraints are constructed to obtain the matching function matrix equation;

[0011] S2. For the matching function matrix equation obtained in step S1, a parameterized matching function is constructed by considering the boundary condition of the velocity ratio of the PP wave to the PS wave;

[0012] S3. Construct the objective function, and then bring the parameterized matching function obtained in step S2 into the objective function to obtain a matching objective function expression;

[0013] S4. Network training and solution, using the neural network U-net and the adaptive moment estimation Adam method to solve the matching function based on physical constraints.

[0014] Furthermore, the specific implementation method of step S1 includes the following steps:

[0015] S1.1. The relationship between the matching function and the velocity ratio of the PP wave to the PS wave is set as follows:

[0016]

[0017] Where t represents time, x represents seismic trace, w(t,x) is the matching function of time t and seismic trace x, and γ(t,x) represents the velocity ratio of PP wave to PS wave at time t and seismic trace x.

[0018] S1.2. The relationship between the matching function obtained in step S1.1 and the velocity ratio of the PP wave to the PS wave is expressed by the first-order Taylor expansion as follows:

[0019]

[0020] Where Δt is the sampling time interval, t i is the time of sampling point i, t i =(i-1)Δt, i=1,2,…,N 1 , N 1 is the total number of sampling points, x jis the jth seismic trace, j = 1, 2, …, M, M is the total number of seismic traces;

[0021] S1.3. Simplified as w(t i ,x j )=w i,j ,γ(t i ,x j )=γ i,j , simplify the discrete expression obtained in step S1.2 to:

[0022]

[0023] Among them, w i+1,j is the matching value of the matching matrix at coordinate (i+1,j);

[0024] Setting w 1,j =w(0,x j )=0, so we get the following expression:

[0025]

[0026] S1.4. Convert the simplified discrete expression obtained in step S3 into a matrix form to obtain the expression:

[0027]

[0028] Let w′ be the matching function matrix, A be a sparse matrix, and p be the matrix with V P / V S The relevant matrix:

[0029]

[0030] Then we get the matching function matrix equation:

[0031] w′=Ap (6).

[0032] Furthermore, the specific implementation method of step S2 includes the following steps:

[0033] S2.1. Setting V P / V S With lower bound γ 0 ;

[0034] S2.2. Based on the setting conditions of step S2.1, we get Then use e u Fitting The matching function matrix equation obtained in step S1 is constructed as a parameterized matching function, and the expression is:

[0035]

[0036] Among them, 1 is a column vector whose elements are all 1, and u is the intermediate variable that needs to be solved.

[0037] Furthermore, the specific implementation method of step S3 includes the following steps:

[0038] S3.1. Construct an energy function to solve the seismic data matching problem. The expression of the energy function is:

[0039]

[0040] Where P and S are PP and PS waves respectively, and the matching function w is used to align the PS seismic reflection with the corresponding PP reflection; ∫∫(P(t,x)-S(w(t,x),x)) 2 dtdx is the data fidelity term used to align the PP wave with the PS wave; is a regularization term used to smooth the matching function; λ is a trade-off parameter;

[0041] S3.2. Using the parameterized matching function to construct the discrete form of the energy function, the expression of the matching objective function is obtained as follows:

[0042]

[0043] Among them, P and S are the discrete forms of PP wave and PS wave respectively, and D tt and D xx They are the second-order derivative matrices in time and space respectively;

[0044] S3.3. Then transform the matching objective function into the matching objective function expression:

[0045]

[0046] in, is the optimal intermediate variable, and argmin is the value of the variable u when it takes the minimum value.

[0047] Furthermore, the specific implementation method of step S4 is to input the random noise z into the neural network U-net, obtain the PS wave through the parameterized matching function constructed in step S2, and then continuously loop to optimize the matching objective function expression obtained in step S3 to obtain the matched PS wave, and then calculate the loss function until the set maximum number of iterations is reached to terminate the loop and obtain the final matching function.

[0048] Furthermore, the expression of the loss function RRMSE in step S4 is:

[0049]

[0050] Among them, w′ is the real matching function, w′ r is the predicted matching function.

[0051] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a seismic data matching method based on physical constraint self-supervised learning when executing the computer program.

[0052] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements a seismic data matching method based on physical constraint self-supervised learning.

[0053] Beneficial effects of the present invention:

[0054] The seismic data matching method based on physical constraint self-supervised learning described in the present invention proposes a new type of constrained self-supervised framework. Compared with traditional deep learning, the framework can realize matching intelligently, thereby improving matching performance.

[0055] The seismic data matching method based on physical constraint self-supervised learning described in the present invention proposes a new seismic data matching technology. By introducing physical constraints, the monotonicity of the warping function is guaranteed, the basic waveform of the seismic data after matching is retained, and the problem of non-stationary seismic data matching is effectively solved.

[0056] The seismic data matching method based on physical constraint self-supervised learning described in the present invention introduces prior knowledge of the minimum velocity ratio when solving the matching function, so that the matching result is more in line with the actual situation, thereby improving the matching accuracy and the matching effect of the seismic data. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A flowchart of a seismic data matching method based on physical constraint self-supervised learning according to the present invention;

[0058] Figure 2 A network architecture diagram of the neural network U-net of the present invention;

[0059] Figure 3 A synthetic seismic record used in the test of Example 1;

[0060] Figure 4 The synthesized PS wave, matching function and PP wave used in the test of Example 1, wherein (a) is the PS wave, (b) is the matching matrix, and (c) is the PP wave;

[0061] Figure 5 The synthesized PS wave and PP wave after adding noise in Example 1, wherein (a) is the PS wave after adding noise, and (b) is the PP wave after adding noise;

[0062] Figure 6 Example 1 is based on the deep learning algorithm Figure 5 The matching results, where (a) is the PS wave after matching obtained by the DLR method, and (b) is the matching matrix obtained by the DLR method;

[0063] Figure 7 Example 1 Dynamic Image Warping Algorithm Figure 5 The matching results, where (a) is the PS wave after matching obtained by the dynamic image warping algorithm, and (b) is the matching matrix obtained by the dynamic image warping algorithm;

[0064] Figure 8 Example 1 The method of the present invention is Figure 5 The matching result, wherein (a) is the PS wave after matching obtained by the method of the present invention, and (b) is the matching matrix obtained by the method of the present invention;

[0065] Fig. 9 The actual seismic data used in the test of Example 2, wherein (a) is the PP wave, (b) is the PS wave, and (c) is the average frequency spectrum of the PP wave and PS wave data;

[0066] Fig.10 is the actual seismic data after preprocessing in Example 2, wherein (a) is the preprocessed PP wave, (b) is the preprocessed PS wave, and (c) is the average frequency spectrum of the preprocessed PP wave and PS wave data;

[0067] Fig.11 Example 2 Dynamic image warping and the method of the present invention Fig.10 The matching results are shown in Figure 2, where (a) is the matched PS wave obtained by directly downsampling the PS wave by one time, (b) is the matched PS wave obtained by the dynamic image warping algorithm, and (c) is the matched PS wave obtained by the method of the present invention;

[0068] Fig.12 For Example 2, the method of the present invention is Fig.10 Matched spectrum and single channel comparison diagram, where (a) is the average spectrum of the PS wave and PP wave after matching, and (b) is the average spectrum of the PS wave and PP wave after matching. Fig.11 (c) Details show that the PP wave and PS wave traces 300, 340, and 380 are extracted for comparison;

[0069] Fig.13 Flowchart of network training in step S4. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solution and advantages of the present invention more clear, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the specific embodiments described are only part of the embodiments of the present invention, rather than all of the specific embodiments. The components of the specific embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations, and the present invention can also have other embodiments.

[0071] Therefore, the following detailed description of the specific embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents the selected specific embodiments of the present invention. Based on the specific embodiments of the present invention, all other specific embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0072] In order to further understand the content, features and effects of the present invention, the following specific implementation methods are given as examples, and the attached Figure 1 -Attached Fig.13 The detailed instructions are as follows:

[0073] Embodiment 1:

[0074] A seismic data matching method based on physical constraint self-supervised learning includes the following steps:

[0075] S1. The relationship between the matching function and the velocity ratio of the PP wave to the PS wave is obtained by using the first-order Taylor expansion to obtain a discrete expression, and physical constraints are constructed to obtain the matching function matrix equation;

[0076] Furthermore, the specific implementation method of step S1 includes the following steps:

[0077] S1.1. The relationship between the matching function and the velocity ratio of the PP wave to the PS wave is set as follows:

[0078]

[0079] Where t represents time, x represents seismic trace, w(t,x) is the matching function of time t and seismic trace x, and γ(t,x) represents the velocity ratio of PP wave to PS wave at time t and seismic trace x.

[0080] S1.2. The relationship between the matching function obtained in step S1.1 and the velocity ratio of the PP wave to the PS wave is expressed by the first-order Taylor expansion as follows:

[0081]

[0082] Where Δt is the sampling time interval, t i is the time of sampling point i, t i =(i-1)Δt, i=1,2,…,N 1 , N 1 is the total number of sampling points, x j is the jth seismic trace, j = 1, 2, …, M, M is the total number of seismic traces;

[0083] S1.3. Simplified as w(t i ,x j )=w i,j ,γ(t i ,x j )=γ i,j , simplify the discrete expression obtained in step S1.2 to:

[0084]

[0085] Among them, w i+1,j is the matching value of the matching matrix at coordinate (i+1,j);

[0086] Setting w 1,j =w(0,x j )=0, so we get the following expression:

[0087]

[0088] S1.4. Convert the simplified discrete expression obtained in step S3 into a matrix form to obtain the expression:

[0089]

[0090] Let w′ be the matching function matrix, A be a sparse matrix, and p be the matrix with V P / V S The relevant matrix:

[0091]

[0092] Then we get the matching function matrix equation:

[0093] w′=Ap (6).

[0094] Furthermore, in order to avoid unwanted folding of the waveform of the data to be matched, the monotonicity constraint of the warping function time needs to be satisfied. The matching function obtained by the right side of equation (6) is monotonic. Since the speed of the PP wave is faster than that of the PS wave (γ = V P / V S>1), so the PS wave needs more travel time to reach the same stratum, which can be achieved by equation (6). Secondly, seismic data matching requires mapping the PS wave to the time domain of the corresponding PP wave. The matching function causes the PS wave's event axis to shift. Since the event axis of seismic data is piecewise continuous, the matching function smoothness constraint needs to be satisfied. Multiplying the vector p by the lower triangular matrix and using the regularization term can obtain a smooth matching function.

[0095] S2. For the matching function matrix equation obtained in step S1, a parameterized matching function is constructed by considering the boundary condition of the velocity ratio of the PP wave to the PS wave;

[0096] Furthermore, the specific implementation method of step S2 includes the following steps:

[0097] S2.1. Setting V P / V S With lower bound γ 0 ;

[0098] S2.2. Based on the setting conditions of step S2.1, we get Then use e u Fitting The matching function matrix equation obtained in step S1 is constructed as a parameterized matching function, and the expression is:

[0099]

[0100] Among them, 1 is a column vector whose elements are all 1, and u is the intermediate variable that needs to be solved.

[0101] S3. Construct the objective function, and then bring the parameterized matching function obtained in step S2 into the objective function to obtain a matching objective function expression;

[0102] Assume that the matching only deals with vertical mismatches in the time domain or depth domain. The seismic data are correctly positioned in the horizontal space. In this case, the seismic data matching problem is solved by optimizing the following energy function;

[0103] Furthermore, the specific implementation method of step S3 includes the following steps:

[0104] S3.1. Construct an energy function to solve the seismic data matching problem. The expression of the energy function is:

[0105]

[0106] Where P and S are PP and PS waves respectively, and the matching function w is used to align the PS seismic reflection with the corresponding PP reflection; ∫∫(P(t,x)-S(w(t,x),x)) 2dtdx is the data fidelity term used to align the PP wave with the PS wave; is a regularization term used to smooth the matching function; λ is a trade-off parameter;

[0107] S3.2. Using the parameterized matching function to construct the discrete form of the energy function, the expression of the matching objective function is obtained as follows:

[0108]

[0109] Among them, P and S are the discrete forms of PP wave and PS wave respectively, and D tt and D xx They are the second-order derivative matrices in time and space respectively;

[0110] S3.3. Then transform the matching objective function into the matching objective function expression:

[0111]

[0112] in, is the optimal intermediate variable, and argmin is the value of the variable u when it takes the minimum value.

[0113] S4. Network training and solution, using the neural network U-net and the adaptive moment estimation Adam method to solve the matching function based on physical constraints.

[0114] Furthermore, the specific implementation method of step S4 is to input the random noise z into the neural network U-net, obtain the PS wave through the parameterized matching function constructed in step S2, and then continuously loop to optimize the matching objective function expression obtained in step S3 to obtain the matched PS wave, and then calculate the loss function until the set maximum number of iterations is reached to terminate the loop and obtain the final matching function.

[0115] Furthermore, the expression of the loss function RRMSE in step S4 is:

[0116]

[0117] Among them, w′ is the real matching function, w′ r is the predicted matching function.

[0118] In this embodiment, the matching method based on deep learning (DLR method) and the dynamic image matching method (DIW method) are used to match the synthetic simulated PP wave and PS wave seismic data with the matching method of the present invention, and the matched seismic data are obtained respectively. λ is set to 0.97, the maximum number of iterations is set to 1500, and the learning rate of the network is 0.01.

[0119] Figure 3 is a simulated earthquake data (time sampling interval = 0.004s), Figure 4 They are the PS wave formed by low-pass filtering, the matching function, and the application of the matching function to Figure 3 The red arrow indicates that the same formation is formed in the PS wave at about 0.57s. However, it can be seen that the PP wave is formed at about 0.35s. Figure 5 is Figure 4 PS wave and PP wave seismic data after adding Gaussian noise. Figure 6 , Figure 7 , Figure 8 The matching functions obtained by using the DLR, DIW algorithms and the method of the present invention are shown respectively. The DLR algorithm fails to achieve a good match in this example, while the DIW algorithm basically achieves the match. However, the predicted matching function value is generally small and easily affected by noise. It can be seen that the method of the present invention has a better effect.

[0120] The matching quality is quantified according to the relative root mean square error, which is calculated as follows:

[0121]

[0122] Among them, w′ is the real matching function, w′ r is the predicted matching function. The relative root mean square errors of the matching functions predicted by the three methods are shown in Table 1.

[0123] Table 1

[0124]

[0125] It can be seen from Table 1 that the relative root mean square error of the method of the present invention is the smallest, and the matching result is obviously better than that of the other two methods.

[0126] Embodiment 2:

[0127] The difference between this embodiment and the embodiment is that this embodiment uses two-dimensional actual data for matching operation to obtain matched seismic data respectively. λ is set to 0.97, the maximum number of iterations is set to 1500, and the learning rate of the network is 0.01.

[0128] Fig. 9 They are respectively actual original PP wave seismic data of size 1000*1200 (time sampling interval = 0.001s), actual original PS wave seismic data of size 2000*1200 (time sampling interval = 0.001s), and an average spectrum diagram. Fig.10 They are the PP wave obtained by low-pass filtering the original PP wave, the PS wave obtained by amplitude compensating the original PS wave, and the average spectrum diagram. Fig.11 Composite seismic data obtained by the DIW algorithm and the method of the present invention, respectively, where the first 340 traces are PP waves and the remaining traces are matched PS waves. The red dotted line represents the seismic event axis. The arrows represent the same point underground. The main event axes of the DIW algorithm failed to match together, and the method of the present invention achieved a better matching effect. Fig.12 This is the average spectrum diagram and single channel comparison diagram obtained by the present invention. It can be seen from the figure that the method of the present invention has achieved a better matching effect.

[0129] Embodiment 3:

[0130] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a seismic data matching method based on physical constraint self-supervised learning described in Example 1 are implemented.

[0131] The computer device of the present invention may be a device including a processor and a memory, such as a single chip microcomputer including a central processing unit, etc. Moreover, the processor is used to implement the steps of the above-mentioned seismic data matching method based on physical constraint self-supervised learning when executing the computer program stored in the memory.

[0132] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0133] The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0134] Embodiment 4:

[0135] A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the seismic data matching method based on physical constraint self-supervised learning described in Example 1 is implemented.

[0136] The computer-readable storage medium of the present invention can be any form of storage medium that can be read by a processor of a computer device, including but not limited to non-volatile memory, volatile memory, ferroelectric memory, etc. A computer program is stored on the computer-readable storage medium. When the processor of the computer device reads and executes the computer program stored in the memory, the steps of the above-mentioned seismic data matching method based on physical constraint self-supervised learning can be implemented.

[0137] The computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer readable media do not include electric carrier signals and telecommunication signals.

[0138] It should be noted that relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0139] Although the present application has been described above with reference to specific embodiments, various modifications may be made thereto and parts thereof may be replaced with equivalents without departing from the scope of the present application. In particular, as long as there is no structural conflict, the various features in the specific embodiments disclosed in the present application may be used in combination with each other in any manner, and the fact that these combinations are not exhaustively described in this specification is only for the sake of omitting space and saving resources. Therefore, the present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A seismic data matching method based on physical constraint self-supervised learning, characterized in that: The steps include: S1. The relationship between the matching function and the velocity ratio of the PP wave to the PS wave is expanded by the first-order Taylor expansion to obtain a discrete expression, and physical constraints are constructed to obtain the matching function matrix equation; S2. For the matching function matrix equation obtained in step S1, a parameterized matching function is constructed by considering the boundary condition of the velocity ratio of the PP wave to the PS wave; S3. Construct the objective function, and then bring the parameterized matching function obtained in step S2 into the objective function to obtain a matching objective function expression; S4. Network training and solution, using the neural network U-net and the adaptive moment estimation Adam method to solve the matching function based on physical constraints.

2. A seismic data matching method based on physical constraint self-supervised learning according to claim 1, characterized in that: The specific implementation method of step S1 includes the following steps: S1.

1. The relationship between the matching function and the velocity ratio of the PP wave to the PS wave is set as follows: Where t represents time, x represents seismic trace, w(t,x) is the matching function of time t and seismic trace x, and γ(t,x) represents the relationship between the velocity ratio of PP wave and PS wave at time t and seismic trace x. S1.

2. The relationship between the matching function obtained in step S1.1 and the velocity ratio of the PP wave to the PS wave is expressed by the first-order Taylor expansion as follows: Where Δt is the sampling time interval, t i is the time of sampling point i, t i =(i-1)Δt, i = 1, 2, ..., N1, N1 is the total number of sampling points, x j is the jth seismic trace, j = 1, 2, …, M, M is the total number of seismic traces; S1.

3. Simplified as w(t i ,x j )=w i,j ,γ(t i ,x j )=γ i,j , simplify the discrete expression obtained in step S1.2 to: Among them, w i+1,j is the matching value of the matching matrix at coordinate (i+1,j); Setting w 1,j =w(0,x j )=0, so we get the following expression: S1.

4. Convert the simplified discrete expression obtained in step S3 into a matrix form to obtain the expression: Let w′ be the matching function matrix, A be a sparse matrix, and p be the matrix with V P / V S The relevant matrix: Then we get the matching function matrix equation: w′=Ap (6).

3. A seismic data matching method based on physical constraint self-supervised learning according to claim 2, characterized in that: The specific implementation method of step S2 includes the following steps: S2.

1. Setting V P / V S has a lower bound γ0; S2.

2. Based on the setting conditions of step S2.1, we get Then use e u Fitting The matching function matrix equation obtained in step S1 is constructed as a parameterized matching function, and the expression is: Among them, 1 is a column vector whose elements are all 1, and u is the intermediate variable that needs to be solved.

4. A seismic data matching method based on physical constraint self-supervised learning according to claim 3, characterized in that: The specific implementation method of step S3 includes the following steps: S3.

1. Construct an energy function to solve the seismic data matching problem. The expression of the energy function is: Where P and S are PP and PS waves respectively, and the matching function w is used to align the PS seismic reflection with the corresponding PP reflection; ∫∫(P(t,x)-S(w(t,x),x)) 2 dtdx is the data fidelity term used to align the PP wave with the PS wave; is a regularization term used to smooth the matching function; λ is a trade-off parameter; S3.

2. Using the parameterized matching function to construct the discrete form of the energy function, the expression of the matching objective function is obtained as follows: Among them, P and S are the discrete forms of PP wave and PS wave respectively, and D tt and D xx They are the second-order derivative matrices in time and space respectively; S3.

3. Then transform the matching objective function into the matching objective function expression: in, is the optimal intermediate variable, and argmin is the value of the variable u when it takes the minimum value.

5. A seismic data matching method based on physical constraint self-supervised learning according to claim 4, characterized in that: The specific implementation method of step S4 is to input the random noise z into the neural network U-net, obtain the PS wave through the parameterized matching function constructed in step S2, and then continuously optimize the matching objective function expression obtained in step S3 to obtain the matched PS wave, and then calculate the loss function until the set maximum number of iterations is reached, then terminate the loop and obtain the final matching function.

6. A seismic data matching method based on physical constraint self-supervised learning according to claim 5, characterized in that: The expression of the loss function RRMSE in step S4 is: Among them, w′ is the real matching function, w′ r is the predicted matching function.

7. An electronic device, characterized in that: It comprises a memory and a processor, the memory stores a computer program, and the processor implements the steps of a seismic data matching method based on physical constraint self-supervised learning as described in any one of claims 1-6 when executing the computer program.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the seismic data matching method based on physical constraint self-supervised learning as described in any one of claims 1 to 6 is implemented.

Images