Inversion method, device, equipment and medium

Abstract

The invention provides an inversion method and device, equipment and a medium, and the method comprises the steps: constructing a reversible neural network model and a coupling objective function, and randomly generating a plurality of groups of Gassmann fluid item elastic parameters based on the elastic parameter characteristics of a to-be-observed region; inputting the multiple groups of Gassmann fluid item elastic parameters into a forward mapping process of the reversible neural network, and performing iterative training on the reversible neural network based on the coupling objective function until the coupling objective function converges, so as to obtain a trained reversible neural network model; and inputting actual seismic data of a to-be-observed area into a reverse mapping process of the trained reversible neural network model to obtain an elastic parameter inversion result of the to-be-observed area. The method provided by the invention can improve the precision of the inversion result.

Description

Technical Field

[0001] The present application relates to the field of geophysics, and in particular to an inversion method, device, equipment and medium. Background Art

[0002] Gassmann fluid term is an important elastic parameter in reservoir fluid identification. Inversion is a common method to obtain Gassmann fluid term in the field of oil and gas exploration, which can be divided into indirect method and direct method. Indirect inversion first obtains velocity and density based on seismic data, and then calculates Gassmann fluid term using empirical formula; direct inversion uses seismic data to directly invert the forward equation of reflection coefficient using Gassmann fluid term to obtain Gassmann fluid term. Direct inversion can avoid the cumulative error existing in the indirect inversion calculation process to a certain extent.

[0003] The current direct inversion method using Amplitude Variation with Offset (AVO) technology is too dependent on the initial low-frequency parameters and training samples in the seismic data. This will result in low accuracy in the inversion results when the initial low-frequency parameters or training samples are small. Summary of the invention

[0004] The present application provides an inversion method, device, equipment and medium, which can improve the accuracy of inversion results.

[0005] In a first aspect, the present application provides an inversion method, the method comprising:

[0006] A reversible neural network model and coupled objective function are constructed, and multiple groups of elastic parameters of Gassmann fluid terms are randomly generated based on the elastic parameter characteristics of the area to be observed;

[0007] The reversible neural network model is obtained by inputting the multiple groups of Gassmann fluid term elastic parameters into the forward mapping process of the reversible neural network, and iteratively training the reversible neural network based on the coupling objective function until the coupling objective function converges; wherein the forward mapping of the reversible neural network model reflects the mapping of elastic parameters to seismic data, and the reverse mapping of the reversible neural network model reflects the mapping of seismic data to elastic parameters;

[0008] The actual seismic data of the area to be observed is obtained, and the actual seismic data of the area to be observed is input into the reverse mapping process of the trained reversible neural network model to obtain the elastic parameter inversion result of the area to be observed.

[0009] In one example, the constructing of the reversible neural network model and the coupling objective function includes:

[0010] Constructing a reversible neural network model, wherein the loss function of the reversible neural network model includes a supervised loss function defined by the error between seismic data, a first unsupervised loss function defined by the error between the latent variable distribution and the Gaussian distribution, and a second unsupervised loss function defined by the error between the elastic parameter distributions;

[0011] The supervised loss function is defined using a zero-delay cross-correlation function, the first unsupervised loss function and the second unsupervised loss function are defined using a maximum mean difference function, and according to the set weight coefficients of each loss function, the supervised loss function, the first unsupervised loss function and the second unsupervised loss function are coupled to obtain the coupled objective function.

[0012] In one example, the method comprises: inputting the multiple groups of Gassmann fluid term elastic parameters into the forward mapping process of the reversible neural network, and iteratively training the reversible neural network based on the coupling objective function until the coupling objective function converges to obtain a trained reversible neural network model, comprising:

[0013] Inputting the elastic parameters into a forward modeling equation in which the reflection coefficient is directly characterized by the elastic parameters, to obtain synthetic seismic data corresponding to the elastic parameters;

[0014] Inputting the elastic parameters into the forward mapping process of the reversible neural network model to obtain predicted seismic data and potential variables corresponding to the elastic parameters output by the reversible neural network model in the forward mapping process; and inputting the predicted seismic data and potential variables corresponding to the elastic parameters into the reverse mapping process of the reversible neural network to obtain predicted elastic parameters corresponding to the elastic parameters output by the reversible neural network model in the reverse mapping process;

[0015] According to the error between the synthetic seismic data and the predicted seismic data corresponding to the elastic parameters, the error between the latent variable distribution and the Gaussian distribution corresponding to the elastic parameters, and the error between the data distribution of the elastic parameters and the corresponding predicted elastic parameters, the reversible neural network model is iteratively updated using a back propagation algorithm based on the coupling objective function until the coupling objective function converges to obtain the trained reversible neural network model.

[0016] In one example, the actual seismic data of the area to be observed is obtained, and the actual seismic data of the area to be observed is input into the reverse mapping process of the trained reversible neural network model to obtain the elastic parameter inversion result of the area to be observed, including:

[0017] Acquire actual seismic data and final latent variables of the area to be observed, wherein the final latent variables are the latent variables most recently output by the reversible neural network model in a forward mapping process when the coupling objective function converges;

[0018] The actual seismic data of the area to be observed and the final potential variables are input into the reverse mapping process of the trained reversible neural network model to obtain the elastic parameter inversion result of the area to be observed.

[0019] In one example, the randomly generating multiple groups of Gassmann fluid term elastic parameters based on the elastic parameter characteristics of the area to be observed includes:

[0020] According to an empirical formula, a plurality of longitudinal wave velocities are randomly generated, and the shear wave velocity and density corresponding to each longitudinal wave velocity are calculated, and a plurality of shear wave velocities and a plurality of densities are obtained by adding random numbers and / or replacing low frequencies;

[0021] Calculating and obtaining the plurality of groups of Gassmann fluid term elastic parameters according to the plurality of longitudinal wave velocities, the plurality of shear wave velocities and the plurality of densities;

[0022] Determine whether the multiple groups of Gassmann fluid item elastic parameters satisfy the set random generation rules, and remove the elastic parameters that do not satisfy the random generation rules from the multiple groups of Gassmann fluid item elastic parameters, wherein the random generation rules include at least one of the following: the value range of the randomly generated elastic parameters is not greater than the value range of the elastic parameters of the area to be observed, the thinnest layer thickness of the randomly generated elastic parameters is less than the thinnest layer thickness identified according to the seismic data of the area to be observed, and the vertical variation trend of the randomly generated elastic parameters conforms to the geological laws of the area to be observed.

[0023] In one example, the reversible neural network model includes a forward reversible block and a reverse reversible block of complementary affine coupling layers;

[0024] The forward reversible block is used to split the elastic parameter input in the forward mapping process into a first elastic parameter and a second elastic parameter; process the first elastic parameter based on a first activation function to obtain a first processing result; process the first elastic parameter based on a second activation function to obtain a second processing result, and output predicted seismic data based on the second elastic parameter, the first processing result and the second processing result by using matrix inner product calculation and summation calculation; and process the predicted seismic data based on a third activation function to obtain a third processing result; process the second elastic parameter based on a fourth activation function to obtain a fourth processing result, and output potential variables based on the first elastic parameter, the third processing result and the fourth processing result by using matrix inner product calculation and summation calculation;

[0025] The reverse reversible block is used to process the input seismic data based on the third activation function to obtain a fifth processing result during the reverse mapping process; process the input seismic data based on the fourth activation function to obtain a sixth processing result, and output a first predicted elastic parameter based on the fifth processing result, the sixth processing result and the input latent variable by using matrix inner product calculation and difference calculation; and process the first activation function and the first predicted elastic parameter to obtain a seventh processing result; process the first predicted elastic parameter based on the second activation function and the first predicted elastic parameter to obtain an eighth processing result, and output a second predicted elastic parameter based on the seventh processing result, the eighth processing result and the input seismic data by using matrix inner product calculation and difference calculation, and synthesize the first predicted elastic parameter and the second predicted elastic parameter to output as a predicted elastic parameter.

[0026] On the other hand, the present application provides an inversion device, the device comprising:

[0027] A construction module is used to construct a reversible neural network model and a coupled objective function, and randomly generate multiple groups of elastic parameters of Gassmann fluid terms based on the elastic parameter characteristics of the area to be observed;

[0028] An input module is used to obtain a trained reversible neural network model by inputting the multiple groups of Gassmann fluid term elastic parameters into the forward mapping process of the reversible neural network and iteratively training the reversible neural network based on the coupling objective function until the coupling objective function converges; wherein the forward mapping of the reversible neural network model reflects the mapping of elastic parameters to seismic data, and the reverse mapping of the reversible neural network model reflects the mapping of seismic data to elastic parameters;

[0029] The acquisition module is used to acquire the actual seismic data of the area to be observed, input the actual seismic data of the area to be observed into the reverse mapping process of the trained reversible neural network model, and obtain the elastic parameter inversion result of the area to be observed.

[0030] In one example, the construction module is specifically used to construct a reversible neural network model, and the loss function of the reversible neural network model includes a supervised loss function defined by the error between seismic data, a first unsupervised loss function defined by the error between the latent variable distribution and the Gaussian distribution, and a second unsupervised loss function defined by the error between the elastic parameter distributions;

[0031] The construction module is specifically used to define the supervised loss function using a zero-delay cross-correlation function, define the first unsupervised loss function and the second unsupervised loss function using a maximum mean difference function, and couple the supervised loss function, the first unsupervised loss function and the second unsupervised loss function according to the set weight coefficients of each loss function to obtain the coupled objective function.

[0032] In yet another aspect, the present application provides an electronic device, comprising: a processor, and a memory communicatively connected to the processor;

[0033] The memory stores computer-executable instructions;

[0034] The processor executes the computer-executable instructions stored in the memory to implement the aforementioned method.

[0035] On the other hand, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method as described in any of the preceding items.

[0036] The inversion method, device, equipment and medium provided by the present application first construct a reversible neural network model and a coupling objective function, and randomly generate multiple groups of elastic parameters of Gassmann fluid terms based on the elastic parameter characteristics of the area to be observed; then, the multiple groups of elastic parameters of Gassmann fluid terms are input into the forward mapping process of the reversible neural network, and the reversible neural network is iteratively trained based on the coupling objective function until the coupling objective function converges, thereby obtaining a trained reversible neural network model; wherein the forward mapping of the reversible neural network model reflects the mapping of elastic parameters to seismic data, and the reverse mapping of the reversible neural network model reflects the mapping of seismic data to elastic parameters; finally, the actual seismic data of the area to be observed is obtained, and the actual seismic data of the area to be observed is input into the reverse mapping process of the trained reversible neural network model to obtain the inversion result of the elastic parameters of the area to be observed. This scheme constructs a bidirectional reversible neural network model and a coupling objective function, and trains the reversible neural network model, and obtains the inversion result of the elastic parameters of the area to be observed based on the trained reversible neural network model, thereby reducing the dependence on the initial low-frequency parameters and training samples in the seismic data and improving the accuracy of the inversion result. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0038] Figure 1This is a schematic diagram of an application scenario for this application example;

[0039] Figure 2 A schematic diagram of a flow chart of an inversion method provided in Example 1 of the present application;

[0040] Figure 3 A schematic diagram of a flow chart of another inversion method provided in Example 1 of the present application;

[0041] Figure 4 A schematic diagram of a flow chart of another inversion method provided in Example 1 of the present application;

[0042] Figure 5 Schematic diagram of the reversible neural network training process;

[0043] Figure 6 A schematic diagram of a flow chart of another inversion method provided in Example 1 of the present application;

[0044] Figure 7 A schematic diagram of a flow chart of another inversion method provided in Example 1 of the present application;

[0045] Figure 8 Schematic diagram of the calculation process of the reversible neural network;

[0046] Fig. 9 An inversion device provided in Embodiment 2 of the present application;

[0047] Fig.10 Schematic diagram of elastic parameters, shear modulus and density of Gassmann fluid term in Marmousi model;

[0048] Fig.11 Post-stack seismic data for example;

[0049] Fig.12 The inversion results of elastic parameters of Gassmann fluid term corresponding to the real seismic data of the example;

[0050] Fig.13 Schematic diagram of extracting data from well A and well B for example;

[0051] Fig.14 Another type of post-stack seismic data is shown as an example;

[0052] Fig.15 The inversion result of elastic parameters of Gassmann fluid term corresponding to another real seismic data is shown as an example;

[0053] Fig.16 This is another schematic diagram of the data of Well A and Well B extracted;

[0054] Fig.17This is a schematic diagram of the structure of an electronic device provided in Example 4 of the present application.

[0055] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0056] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0057] The specific application scenario of this application is the field of geophysics. Figure 1 Schematic diagram of the application scenario of the present application example, a is the current conventional neural network, b is the reversible neural network, the Gassmann fluid term is an important elastic parameter in reservoir fluid identification, and inversion is a common method for obtaining the elastic parameters of the Gassmann fluid term in the field of oil and gas exploration. Generally, the elastic parameters of the Gassmann fluid term are obtained based on the inversion of the seismic data of the exploration area to be observed.

[0058] At present, based on conventional neural networks, the methods for inverting seismic data to obtain the elastic parameters of the Gassmann fluid term are generally divided into indirect methods and direct methods. Direct inversion can avoid the cumulative errors existing in the indirect inversion calculation process to a certain extent, but the AVO direct inversion method can be divided into model-driven inversion and data-driven inversion. Model-driven inversion has a certain dependence on the initial low-frequency parameters in the seismic data, and data-driven inversion requires a large number of representative training samples. Therefore, the AVO direct inversion method is highly dependent on the initial low-frequency parameters and training samples. When there are no large number of initial low-frequency parameters or training samples, the accuracy of the Gassmann fluid term elastic parameters obtained by AVO direct inversion will be low.

[0059] However, the method provided in the present application can be based on constructing a reversible neural network model and a coupling objective function, wherein the reversible neural network is a bidirectional mapping, wherein the forward direction is a mapping of elastic parameters of the Gassmann fluid term to seismic data and potential variables, and the reverse direction is a mapping of seismic data and potential variables to elastic parameters of the Gassmann fluid term. The reversible neural network model is trained, and based on the trained reversible neural network model, the seismic data of the area to be observed is input into the reverse mapping of the trained reversible neural network model to obtain the inversion result of the elastic parameters of the area to be observed, thereby reducing the dependence on the initial low-frequency parameters and training samples in the seismic data and improving the accuracy of the inversion result.

[0060] It should be noted that the brief description of terms in this application is only for the convenience of understanding the embodiments described below, and is not intended to limit the embodiments of this application. Unless otherwise specified, these terms should be understood according to their common and usual meanings.

[0061] The technical solutions of the present application and the technical solutions of the present application are described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. In the description of the present application, unless otherwise clearly specified and limited, each term should be understood in a broad sense in the art. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0062] Embodiment 1

[0063] Figure 2 A schematic diagram of a process flow of an inversion method provided in Example 1 of the present application is shown in FIG. Figure 2 As shown, the method includes:

[0064] Step 201, constructing a reversible neural network model and a coupled objective function, and randomly generating multiple groups of Gassmann fluid term elastic parameters based on the elastic parameter characteristics of the area to be observed;

[0065] Step 202, by inputting the multiple groups of Gassmann fluid term elastic parameters into the forward mapping process of the reversible neural network, and iteratively training the reversible neural network based on the coupling objective function until the coupling objective function converges, a trained reversible neural network model is obtained; wherein the forward mapping of the reversible neural network model reflects the mapping of elastic parameters to seismic data, and the reverse mapping of the reversible neural network model reflects the mapping of seismic data to elastic parameters;

[0066] Step 203: Acquire actual seismic data of the area to be observed, input the actual seismic data of the area to be observed into the reverse mapping process of the trained reversible neural network model, and obtain the elastic parameter inversion result of the area to be observed.

[0067] The executor of this embodiment is a device for characterizing the seepage law, which can be implemented by a computer program, such as application software, etc.; or, it can also be implemented as a medium storing relevant computer programs, such as a U disk, a cloud disk, etc.; or, it can also be implemented by a physical device integrated or installed with relevant computer programs, such as a chip, etc.

[0068] Combined with the scenario example, the reversible neural network model constructed in this application is used to fit the bidirectional mapping relationship between the elastic parameters of the Gassmann fluid term and the seismic data, wherein the forward mapping is based on the mapping of the elastic parameters of the Gassmann fluid term to the seismic data, and the reverse mapping is based on the mapping of the seismic data to the elastic parameters of the Gassmann fluid term. Since conventional neural networks with unidirectional mapping usually learn the underdetermined inversion process from seismic data to the elastic parameters of the Gassmann fluid term, the training efficiency is low and the stability is not high, so based on the reversible neural network model of training bidirectional mapping, the inversion using the reversible neural network model can improve the efficiency and accuracy of the inversion. A representative area is selected as the area to be observed in this application, and the selection of the area to be observed can select an area where an earthquake has occurred, and the characteristics of the elastic parameters of the Gassmann fluid term in the area to be observed are determined based on the actual situation of the earthquake in the area to be observed, and according to the characteristics of the elastic parameters of the Gassmann fluid term, multiple groups of elastic parameters of the Gassmann fluid term that meet the characteristics of the elastic parameters of the Gassmann fluid term are randomly generated. Common elastic parameters of Gassmann fluid terms include: Lame constant (λ, μ), Young's modulus (E), Poisson's ratio (ν), bulk modulus (K), shear modulus (G), density (ρ), and wave velocity (Vp, Vs). For uniform isotropic and perfectly elastic media, only one pair of the three pairs of elastic moduli, λ-μ, E-ν, and KG, needs to be known to determine the other two pairs. Combined with the formation density, the longitudinal and transverse wave velocities Vp and Vs of the entire medium or formation can be determined.

[0069] The multiple groups of Gassmann fluid term elastic parameters generated above are input as training data into the forward mapping process of the reversible neural network to complete the training of the reversible neural network. Specifically, a coupling objective function is set in the reversible neural network, and the coupling objective function can characterize the error between the training data and the training results based on the training data at each stage of the training process during the training of the reversible neural network. When the error is gradually decreasing, it means that the coupling objective function is gradually converging. When the coupling objective function completes convergence, it can be said that the training of the reversible neural network has been completed. After the training of the reversible neural network is completed, the real seismic data when the earthquake occurs in the area to be observed is obtained. The real seismic data when the earthquake occurs in the area to be observed can be obtained by recording the seismic data when the earthquake occurs in the area to be observed, and querying the recorded seismic data. Because the reverse mapping of the reversible neural network is based on the mapping of seismic data to the elastic parameters of the Gassmann fluid term, the seismic data of the area to be observed can be input into the reverse mapping process of the trained reversible neural network, and the inversion result of the area to be observed can be obtained based on the inversion of the reversible neural network. The inversion result is the elastic parameters of the Gassmann fluid term of the area to be observed.

[0070] This example first constructs a reversible neural network model and a coupling objective function, and randomly generates multiple groups of Gassmann fluid term elastic parameters based on the elastic parameter characteristics of the area to be observed; then, the multiple groups of Gassmann fluid term elastic parameters are input into the forward mapping process of the reversible neural network, and the reversible neural network is iteratively trained based on the coupling objective function until the coupling objective function converges, thereby obtaining a trained reversible neural network model; wherein the forward mapping of the reversible neural network model reflects the mapping of elastic parameters to seismic data, and the reverse mapping of the reversible neural network model reflects the mapping of seismic data to elastic parameters; finally, the actual seismic data of the area to be observed is obtained, and the actual seismic data of the area to be observed is input into the reverse mapping process of the trained reversible neural network model to obtain the inversion result of the elastic parameters of the area to be observed. This example constructs a bidirectional reversible neural network model and a coupling objective function, and trains the reversible neural network model, and obtains the inversion result of the elastic parameters of the area to be observed based on the trained reversible neural network model, thereby reducing the dependence on the initial low-frequency parameters and training samples in the seismic data and improving the accuracy of the inversion result.

[0071] Optional, Figure 3 A schematic diagram of a flow chart of another inversion method provided in Example 1 of the present application, such as Figure 3 As shown, in step 201, a reversible neural network model and a coupling objective function are constructed, including:

[0072] Step 301, constructing a reversible neural network model, wherein the loss function of the reversible neural network model includes a supervised loss function defined by the error between seismic data, a first unsupervised loss function defined by the error between the latent variable distribution and the Gaussian distribution, and a second unsupervised loss function defined by the error between the elastic parameter distributions;

[0073] Step 302: define the supervised loss function using a zero-delay cross-correlation function, define the first unsupervised loss function and the second unsupervised loss function using a maximum mean difference function, and couple the supervised loss function, the first unsupervised loss function and the second unsupervised loss function according to the set weight coefficients of each loss function to obtain the coupled objective function.

[0074] Combined with the scenario example, when constructing the reversible neural network, since the reversible neural network needs to be trained, a loss function can be constructed in the reversible neural network. Since seismic data is the key data for obtaining the inversion result, and since the real seismic data of the area to be observed is correlated with the seismic data predicted by the model, the error between the seismic data can be defined as a supervised loss function based on the zero-delay cross-correlation function. In the process of training the reversible neural network, the reversible neural network is made to memorize and learn the seismic data. The loss function in supervised training is often used to reflect the degree of inconsistency between the true value of the sample and the model predicted value. In addition to defining the error between the seismic data as a supervised loss function, according to the principle of the maximum mean difference function and the characteristics of unsupervised learning, the error between the key parameters in the training process of the reversible neural network is defined as an unsupervised loss function. For example, the error between the latent variable distribution of the reversible neural network and the Gaussian distribution can be used as the first unsupervised loss function, and the error between the elastic parameter distributions can be used as the second unsupervised loss function. The latent variable distribution of the reversible neural network is a parameter of the reversible neural network itself. The role of the latent variable is to store important information related to the elastic parameters lost in the forward process. The Gaussian distribution can be selected as any Gaussian distribution. The weights corresponding to the supervised loss function, the first unsupervised loss function and the second unsupervised loss function are determined according to the quality of the seismic data and the elastic parameter data of the Gassmann fluid term. The objective function can be obtained by coupling the corresponding weights of the supervised loss function, the first unsupervised loss function and the second unsupervised loss function. For example, if the weights corresponding to the supervised loss function, the first unsupervised loss function and the second unsupervised loss function are: a1, a2 and a3 respectively, the supervised loss function, the first unsupervised loss function and the second unsupervised loss function are SL R 、USL z With USL m, the objective function can be defined as:

[0075] L=a1·SL R +a2·USL z +a3·USL m

[0076] Wherein L is the objective function. In this example, the error between the key parameters involved in the training of the reversible neural network is set as the loss function, and the loss function is coupled as the objective function. Based on the convergence of the objective function, it is determined whether the training of the reversible neural network is completed, thereby improving the efficiency of the training process of the reversible neural network and ensuring the accuracy of the reversible neural network.

[0077] Optional, Figure 4 A flow chart of another inversion method provided in Example 1 of the present application is shown as follows: Figure 4 As shown, in step 202, the reversible neural network model after training is obtained by inputting the multiple groups of Gassmann fluid term elastic parameters into the forward mapping process of the reversible neural network and iteratively training the reversible neural network based on the coupling objective function until the coupling objective function converges, including:

[0078] Step 401, inputting the elastic parameters into a forward modeling equation in which the elastic parameters directly characterize the reflection coefficient, and obtaining synthetic seismic data corresponding to the elastic parameters;

[0079] Step 402: input the elastic parameters into the forward mapping process of the reversible neural network model to obtain predicted seismic data and potential variables corresponding to the elastic parameters output by the reversible neural network model in the forward mapping process; and input the predicted seismic data and potential variables corresponding to the elastic parameters into the reverse mapping process of the reversible neural network to obtain predicted elastic parameters corresponding to the elastic parameters output by the reversible neural network model in the reverse mapping process;

[0080] Step 403, according to the error between the synthetic seismic data corresponding to the elastic parameters and the predicted seismic data, the error between the latent variable distribution corresponding to the elastic parameters and the Gaussian distribution, and the error between the data distribution of the elastic parameters and the corresponding predicted elastic parameters, based on the coupling objective function, the reversible neural network model is iteratively updated with a back propagation algorithm until the coupling objective function converges to obtain the trained reversible neural network model.

[0081] Combined with the scene example, Figure 5Schematic diagram of the reversible neural network training process. In the process of training the reversible neural network, the first group of Gassmann fluid term elastic parameters from the multiple groups of Gassmann fluid term elastic parameters is first input into the forward equation. The forward equation can synthesize the input elastic parameters into seismic data. Therefore, the first group of Gassmann fluid term elastic parameters is input into the forward equation to obtain the first synthetic seismic data corresponding to the first group of Gassmann fluid term elastic parameters. Then, the first group of Gassmann fluid term elastic parameters is input into the forward mapping process of the reversible neural network, so that the reversible neural network outputs the first predicted seismic parameters and the first latent variables corresponding to the first group of Gassmann fluid term elastic parameters. Then, the first predicted seismic parameters and the first latent variables are input into the reversible neural network, so that the reversible neural network outputs the first predicted elastic parameters corresponding to the first group of Gassmann fluid term elastic parameters. The process of obtaining the first predicted elastic parameters based on the first group of Gassmann fluid term elastic parameters is used as the first iterative training of the reversible neural network.

[0082] The error between the first synthetic seismic data and the first predicted seismic data is taken as the first error, the error between the first latent variable and the Gaussian distribution is taken as the second error, the Gaussian distribution can be any Gaussian distribution, the error between the first elastic parameter and the first predicted elastic parameter is taken as the third error, and the value of the coupling objective function is obtained under the first iteration based on the first error, the second error and the third error. Under the first iterative training, the value of the coupling objective function is the sum of the first error, the second error and the third error.

[0083] After obtaining the value of the coupling objective function under the first iteration, the reversible neural network is updated based on the back propagation algorithm. After the reversible neural network is updated, the second group of Gassmann fluid term elastic parameters are respectively input into the forward equation and the forward mapping of the reversible neural network. The reversible neural network is iteratively trained for the second time based on the second group of Gassmann fluid term elastic parameters to obtain the value of the coupling objective function under the second iterative training of the second group of Gassmann fluid term elastic parameters, and the reversible neural network is updated again by the back propagation algorithm.

[0084] According to the above method, based on the generated multiple sets of Gassmann fluid term elastic parameters, the reversible neural network is iteratively trained in sequence to obtain the value of the coupling objective function corresponding to each iterative training, and after each iterative training is completed, the reversible neural network is updated based on the back propagation algorithm until the coupling objective function converges, and then the training of the reversible neural network is stopped. It is worth mentioning that if the coupling objective function has not converged after the reversible neural network is trained in sequence with the multiple sets of Gassmann fluid term elastic parameters, the reversible neural network is iteratively trained again starting from the first set of Gassmann fluid term elastic parameters based on the multiple sets of Gassmann fluid term elastic parameters until the multiple sets of Gassmann fluid term elastic parameters converge.

[0085] In this example, the reversible neural network is trained in sequence through the group of Gassmann fluid term elastic parameters, and the training completion of the reversible neural network is monitored by the convergence of the coupling objective function, so that the obtained reversible neural network can be made more accurate.

[0086] Optional, Figure 6 A flow chart of another inversion method provided in Example 1 of the present application is shown as follows: Figure 6 As shown, step 203 includes:

[0087] Step 601, obtaining actual seismic data and final latent variables of the area to be observed, wherein the final latent variables are the latent variables most recently output by the reversible neural network model in the forward mapping process when the coupling objective function converges;

[0088] Step 602: Input the actual seismic data of the area to be observed and the final potential variables into the reverse mapping process of the trained reversible neural network model to obtain the elastic parameter inversion result of the area to be observed.

[0089] Combined with the scenario example, if in the process of training the reversible neural network, if the iterative training of the reversible neural network is performed a total of n times before the coupling objective function converges, the potential variable obtained in the nth time of training the reversible neural network is used as the final potential variable. Based on the survey technology of seismic data, the actual seismic data generated when an earthquake occurs in the area to be observed is obtained, and the actual seismic data generated in the area to be observed and the final potential variable are used as input data in the reverse mapping process in the reversible neural network, so that the reversible neural network outputs the elastic parameter inversion result, that is, the elastic parameter distribution corresponding to the area to be observed is obtained.

[0090] This example is based on taking the final latent variable as a part of the input data in the reverse mapping process of the reversible neural network, so that the reversible neural network can obtain the inversion result more accurately during the inversion process.

[0091] Optional, Figure 7 A flow chart of another inversion method provided in Example 1 of the present application is shown as follows: Figure 7 As shown, in step 201, the multiple groups of Gassmann fluid term elastic parameters are randomly generated based on the elastic parameter characteristics of the area to be observed, including:

[0092] Step 701: randomly generate multiple longitudinal wave velocities according to an empirical formula, calculate the shear wave velocity and density corresponding to each longitudinal wave velocity, and obtain multiple shear wave velocities and multiple densities by adding random numbers and / or replacing low frequencies;

[0093] Step 702, calculating and obtaining the plurality of groups of Gassmann fluid item elastic parameters according to the plurality of longitudinal wave velocities, the plurality of shear wave velocities and the plurality of densities;

[0094] Step 703, determine whether the multiple groups of Gassmann fluid item elastic parameters meet the set random generation rules, and remove the elastic parameters that do not meet the random generation rules from the multiple groups of Gassmann fluid item elastic parameters, and the random generation rules include at least one of the following: the value range of the randomly generated elastic parameters is not greater than the value range of the elastic parameters of the area to be observed, the thinnest layer thickness of the randomly generated elastic parameters is less than the thinnest layer thickness identified according to the seismic data of the area to be observed, and the vertical change trend of the randomly generated elastic parameters conforms to the geological laws of the area to be observed.

[0095] Combined with the scenario example, in the process of generating multiple groups of Gassmann fluid item elastic parameters, the generation rules of elastic parameters need to be met, wherein the first rule is: the value range of the randomly generated Gassmann fluid item elastic parameter data is less than or equal to the value range of the Gassmann fluid item elastic parameter of the area to be observed, that is, the upper and lower limits of the Gassmann fluid item elastic parameters of the area to be observed are first determined, and the randomly generated Gassmann fluid item elastic parameters must be between the upper and lower limits of the Gassmann fluid item elastic parameters of the area to be observed; the second rule is: in the randomly generated Gassmann fluid item elastic parameter data, the thickness of the thinnest layer that can be identified must be less than the thickness of the thinnest layer that can be identified by the seismic data of the area to be observed, that is, the randomly generated Gassmann fluid item elastic parameter data must be identified before the subsequent training of the reversible neural network can be completed; the third rule is: the vertical change trend of the randomly generated Gassmann fluid item elastic parameter data conforms to the geological laws of the area to be observed. Based on the above three elastic parameter generation rules, the first empirical formula can be first used to generate multiple longitudinal wave velocities within the elastic parameter generation rules. In order to ensure the diversity of the generated Gassmann fluid term elastic parameters, different second empirical formulas can be used for each of the multiple longitudinal wave velocities to generate the shear wave velocity corresponding to each longitudinal wave velocity, and different third empirical formulas can be used for each longitudinal wave velocity to generate the density corresponding to each longitudinal wave velocity. The Gassmann fluid term elastic parameters are divided into high frequency and low frequency. In addition to obtaining multiple longitudinal wave velocities, shear wave velocities and densities through empirical formulas, multiple groups of random numbers can be added to serve as the remaining longitudinal wave velocities, shear wave velocities and densities, and the low-frequency values ​​in some parameters can also be replaced. In this way, multiple groups of Gassmann fluid term elastic parameters with diversity can be obtained.

[0096] After obtaining multiple groups of Gassmann fluid item elastic parameters, each group of Gassmann fluid item elastic parameters is compared with the generation rule of elastic parameters. If, in the process of calculating by the second empirical formula and the third empirical formula, or in the process of adding multiple groups of random numbers or replacing low-frequency values, the result of a group of Gassmann fluid item elastic parameters does not conform to the generation rule of the elastic parameters, then the group of Gassmann fluid item elastic parameters that do not conform to the generation rule of the elastic parameters is deleted. Finally, each group of Gassmann fluid item elastic parameters obtained finally conforms to the generation rule of elastic parameters. Finally, based on the longitudinal wave velocity, shear wave velocity and density in each group of Gassmann fluid item elastic parameters, the first calculation formula is used to calculate the Gassmann fluid item and shear modulus corresponding to each group of Gassmann fluid item elastic parameters. The Gassmann fluid item and shear modulus are the elastic parameters in each group of Gassmann fluid item elastic parameters, which are finally used to train the reversible neural network.

[0097] This example limits the generation of multiple groups of Gassmann fluid item elastic parameters by stipulating the generation rules of elastic parameters, so that the generated multiple groups of Gassmann fluid item elastic parameters are of higher quality and more valuable. In addition, multiple groups of Gassmann fluid item elastic parameters are generated by using multiple empirical formulas, adding multiple groups of random numbers or replacing low-frequency values, so that the generated multiple groups of Gassmann fluid item elastic parameters can be more diverse.

[0098] Optionally, the reversible neural network model includes a forward reversible block and a reverse reversible block of a complementary affine coupling layer;

[0099] The forward reversible block is used to split the elastic parameter input in the forward mapping process into a first elastic parameter and a second elastic parameter; process the first elastic parameter based on a first activation function to obtain a first processing result; process the first elastic parameter based on a second activation function to obtain a second processing result, and output predicted seismic data based on the second elastic parameter, the first processing result and the second processing result by using matrix inner product calculation and summation calculation; and process the predicted seismic data based on a third activation function to obtain a third processing result; process the second elastic parameter based on a fourth activation function to obtain a fourth processing result, and output potential variables based on the first elastic parameter, the third processing result and the fourth processing result by using matrix inner product calculation and summation calculation;

[0100] The reverse reversible block is used to process the input seismic data based on the third activation function to obtain a fifth processing result during the reverse mapping process; process the input seismic data based on the fourth activation function to obtain a sixth processing result, and output a first predicted elastic parameter based on the fifth processing result, the sixth processing result and the input latent variable by using matrix inner product calculation and difference calculation; and process the first activation function and the first predicted elastic parameter to obtain a seventh processing result; process the first predicted elastic parameter based on the second activation function and the first predicted elastic parameter to obtain an eighth processing result, and output a second predicted elastic parameter based on the seventh processing result, the eighth processing result and the input seismic data by using matrix inner product calculation and difference calculation, and synthesize the first predicted elastic parameter and the second predicted elastic parameter to output as a predicted elastic parameter.

[0101] Combined with the scenario example, since the reversible neural network can perform three functions of forward mapping, reverse mapping and calculation of coupling objective function, the reversible neural network can be divided into three modules according to the function, namely, forward reversible block, reverse reversible block and complementary affine coupling layer. The forward reversible block is used to realize the forward mapping process, the reverse reversible block is used to realize the reverse mapping process, and the complementary affine coupling layer is used to realize the calculation of coupling objective function. Since the elastic parameters of Gassmann fluid term are divided into high frequency and low frequency, the Gassmann fluid term and shear modulus in each group of Gassmann fluid term elastic parameters are respectively divided into high frequency Gassmann fluid term, low frequency Gassmann fluid term, high frequency shear modulus and low frequency shear modulus, and the high frequency Gassmann fluid term and the high frequency shear modulus are used as the first elastic parameter, and the low frequency Gassmann fluid term and the low frequency shear modulus are used as the second elastic parameter. The first elastic parameter and the second elastic parameter are simultaneously input into the forward reversible block. Figure 8 It is a schematic diagram of the calculation process of the reversible neural network, such as Figure 8 As shown, the predicted earthquake data and potential variables output by the forward mapping are obtained based on the first formula and the second formula below respectively.

[0102] First formula: y1=x1⊙exp[s2(x2)]+t2(x2)

[0103] Second formula: y2=x2⊙exp[s1(x1)]+t1(y1)

[0104] Among them: y1 is the predicted earthquake data, y2 is the potential variable, x1 is the second elastic parameter, and x2 is the first elastic parameter.

[0105] Specifically, according to the first formula, the process of obtaining the predicted seismic data through the first elastic parameter and the second elastic parameter can firstly process the first elastic parameter using the first activation function s2 to obtain a first processing result s2(x2), then process the first elastic parameter based on the second activation function t2 to obtain a second processing result t2(x2), then perform matrix inner product calculation on the second elastic parameter and the first processing result to obtain a processing result of x1⊙exp[s2(x2)], and finally add the second processing result to obtain the predicted seismic data y1;

[0106] According to the second formula, the predicted seismic data y1 can be processed by the third activation function t1 to obtain the third processing result t1(y1), the second elastic parameter can be processed based on the fourth activation function s1 to obtain the fourth processing result s1(x1), and then the first elastic parameter and the fourth processing result are subjected to matrix inner product calculation to obtain x2⊙exp[s1(x1)], and finally the latent variable y2 is obtained by adding the third processing result.

[0107] After obtaining the predicted seismic data and the latent variables, the predicted seismic data and the latent variables are input into the reverse reversible block of the reversible neural network, and the first predicted elastic parameter and the second predicted elastic parameter are obtained based on the third formula and the fourth formula below, respectively.

[0108] The third formula: x2' = [y2-t1(y1)]⊙exp[-s1(y1)]

[0109] Fourth formula: x1'=[y1-t2(x2)]⊙exp[-s2(x2)]

[0110] Among them: x1' is the second elastic parameter, x2' is the first elastic parameter.

[0111] Specifically, the predicted seismic data y1 is processed by the third activation function t1 to obtain a fifth processing result t1(y1), the predicted seismic data y1 is processed based on the fourth activation function s1 to obtain a sixth processing result s1(y1), the sixth processing result is negated to obtain -s1(y1), the latent variable y2 is subtracted from the fifth processing result to obtain y2-t1(y1), and based on -s1(y1) and y2-t1(y1), the first elastic parameter x2' is calculated using the matrix inner product.

[0112] The first elastic parameter is processed using the first activation function s2 to obtain the seventh processing result s2(x2), and the first elastic parameter is processed based on the second activation function t2 to obtain the eighth processing result t2(x2), the seventh processing result is the same as the first processing result, and the eighth processing result is the same as the second processing result. The predicted seismic data y1 is subtracted from the eighth processing result to obtain y2-t1(y1), the opposite of the seventh processing result is calculated to obtain -s1(y1), and the second elastic parameter x1' is calculated based on y2-t1(y1) and -s1(y1) using the matrix inner product. It is worth mentioning that the first activation function, the second activation function, the third activation function and the fourth activation function can use the same activation function, or different activation functions.

[0113] If the reverse mapping process is the iterative process of training the reversible neural network, the first elastic parameter x2' and the second elastic parameter x1' outputted are the elastic parameter inversion results outputted in the iterative process; if the reverse mapping process is that the reversible neural network has been trained and is formally used to invert the elastic parameters, the predicted seismic data y1 is replaced by the actual seismic data of the area to be observed, and the latent variable y2 is replaced by the final latent variable, then the first elastic parameter x2' obtained by the reverse mapping is the high-frequency Gassmann fluid term and high-frequency shear modulus of the elastic parameter inversion result of the area to be observed; the second elastic parameter x1' obtained by the reverse mapping is the low-frequency Gassmann fluid term and low-frequency shear modulus of the elastic parameter inversion result of the area to be observed. This example adopts multiple formulas to be applied to the forward mapping and reverse mapping processes, so that the reversible neural network can quickly and accurately obtain the inversion results during training and use.

[0114] This embodiment first constructs a reversible neural network model and a coupling objective function, and randomly generates multiple groups of Gassmann fluid term elastic parameters based on the elastic parameter characteristics of the area to be observed; then, the multiple groups of Gassmann fluid term elastic parameters are input into the forward mapping process of the reversible neural network, and the reversible neural network is iteratively trained based on the coupling objective function until the coupling objective function converges, thereby obtaining a trained reversible neural network model; wherein the forward mapping of the reversible neural network model reflects the mapping of elastic parameters to seismic data, and the reverse mapping of the reversible neural network model reflects the mapping of seismic data to elastic parameters; finally, the actual seismic data of the area to be observed is obtained, and the actual seismic data of the area to be observed is input into the reverse mapping process of the trained reversible neural network model to obtain the inversion result of the elastic parameters of the area to be observed. This embodiment constructs a bidirectional reversible neural network model and a coupling objective function, and trains the reversible neural network model, and obtains the inversion result of the elastic parameters of the area to be observed based on the trained reversible neural network model, thereby reducing the dependence on the initial low-frequency parameters and training samples in the seismic data and improving the accuracy of the inversion result.

[0115] Embodiment 2

[0116] Fig. 9 An inversion device is provided in the second embodiment of the present application, such as Fig. 9 As shown, the device comprises:

[0117] A construction module 91 is used to construct a reversible neural network model and a coupling objective function, and randomly generate multiple groups of elastic parameters of Gassmann fluid terms based on the elastic parameter characteristics of the area to be observed;

[0118] An input module 92 is used to obtain a trained reversible neural network model by inputting the multiple groups of Gassmann fluid term elastic parameters into the forward mapping process of the reversible neural network and iteratively training the reversible neural network based on the coupling objective function until the coupling objective function converges; wherein the forward mapping of the reversible neural network model reflects the mapping of elastic parameters to seismic data, and the reverse mapping of the reversible neural network model reflects the mapping of seismic data to elastic parameters;

[0119] The acquisition module 93 is used to acquire the actual seismic data of the area to be observed, input the actual seismic data of the area to be observed into the reverse mapping process of the trained reversible neural network model, and obtain the elastic parameter inversion result of the area to be observed.

[0120] In combination with the scenario example, the input module 92 inputs the multiple groups of Gassmann fluid term elastic parameters generated above as training data into the forward mapping process of the reversible neural network to complete the training of the reversible neural network. Specifically, the construction module 91 sets a coupling objective function in the reversible neural network. The coupling objective function can characterize the error between the training data and the training results based on the training data at each stage of the training process during the training of the reversible neural network. When the error is gradually decreasing, it means that the coupling objective function is gradually converging. When the coupling objective function completes convergence, it can be said that the training of the reversible neural network has been completed. After completing the training of the reversible neural network, the acquisition module 93 obtains the real seismic data when an earthquake occurs in the area to be observed. The real seismic data when an earthquake occurs in the area to be observed can be obtained by recording the seismic data when an earthquake occurs in the area to be observed, and querying the recorded seismic data. Because the reverse mapping of the reversible neural network is based on the mapping of seismic data to the elastic parameters of the Gassmann fluid term, the seismic data of the area to be observed can be input into the reverse mapping process of the trained reversible neural network, and the inversion result of the area to be observed can be obtained based on the inversion of the reversible neural network. The inversion result is the elastic parameters of the Gassmann fluid term of the area to be observed.

[0121] This example constructs a bidirectional reversible neural network model and a coupled objective function, and trains the reversible neural network model. Based on the trained reversible neural network model, the elastic parameter inversion result of the area to be observed is obtained, which reduces the dependence on the initial low-frequency parameters and training samples in the seismic data and improves the accuracy of the inversion result. Optionally, the construction module 91 is specifically used to construct a reversible neural network model, and the loss function of the reversible neural network model includes a supervised loss function defined by the error between seismic data, a first unsupervised loss function defined by the error between the latent variable distribution and the Gaussian distribution, and a second unsupervised loss function defined by the error between the elastic parameter distributions; the construction module 91 is also specifically used to define the supervised loss function using a zero-delay cross-correlation function, define the first unsupervised loss function and the second unsupervised loss function using a maximum mean difference function, and couple the supervised loss function, the first unsupervised loss function and the second unsupervised loss function according to the weight coefficients of each loss function set to obtain the coupled objective function. Combined with the scenario example, when constructing the reversible neural network, since the reversible neural network needs to be trained, a loss function can be constructed in the reversible neural network. Since seismic data is the key data for obtaining the inversion result, and since the real seismic data of the area to be observed is related to the seismic data predicted by the model, the error between the seismic data can be defined as a supervised loss function according to the zero-delay cross-correlation function. In the training process of the reversible neural network, the reversible neural network is made to memorize and learn the seismic data. The loss function in supervised training is often used to reflect the degree of inconsistency between the true value of the sample and the model predicted value. In addition to defining the error between the seismic data as a supervised loss function, according to the principle of the maximum mean difference function and the characteristics of unsupervised learning, the error between the key parameters in the training process of the reversible neural network is defined as an unsupervised loss function. The weights corresponding to the supervised loss function, the first unsupervised loss function and the second unsupervised loss function are determined according to the quality of the seismic data and the elastic parameter data of the Gassmann fluid term. The objective function can be obtained by coupling the weights corresponding to the supervised loss function, the first unsupervised loss function and the second unsupervised loss function. This example improves the efficiency of the reversible neural network training process and ensures the accuracy of the reversible neural network by setting the errors between the key parameters involved in the training of the reversible neural network as the loss function, coupling the loss function as the objective function, and determining whether the training of the reversible neural network is completed based on the convergence of the objective function.In combination with the scenario example, in the process of training the reversible neural network, the first group of Gassmann fluid term elastic parameters from the multiple groups of Gassmann fluid term elastic parameters is first input into the forward equation. The forward equation can synthesize the input elastic parameters into seismic data, so the first group of Gassmann fluid term elastic parameters is input into the forward equation to obtain the first synthetic seismic data corresponding to the first group of Gassmann fluid term elastic parameters. Then, the first group of Gassmann fluid term elastic parameters is input into the forward mapping process of the reversible neural network, so that the reversible neural network outputs the first predicted seismic parameters and the first latent variable corresponding to the first group of Gassmann fluid term elastic parameters. Then, the first predicted seismic parameters and the first latent variable are input into the reversible neural network, so that the reversible neural network outputs the first predicted elastic parameters corresponding to the first group of Gassmann fluid term elastic parameters, and the process of obtaining the first predicted elastic parameters based on the first group of Gassmann fluid term elastic parameters is used as the first iterative training of the reversible neural network. The error between the first synthetic seismic data and the first predicted seismic data is taken as the first error, the error between the first latent variable and the Gaussian distribution is taken as the second error, the Gaussian distribution can be any Gaussian distribution, the error between the first elastic parameter and the first predicted elastic parameter is taken as the third error, the value of the coupling objective function under the first iteration is obtained based on the first error, the second error and the third error, and the value of the coupling objective function under the first iteration training is obtained by summing the first error, the second error and the third error. After obtaining the value of the coupling objective function under the first iteration, the reversible neural network is updated based on the back propagation algorithm, after the reversible neural network is updated, the second group of Gassmann fluid term elastic parameters are respectively input into the forward equation and the forward mapping of the reversible neural network, the reversible neural network is iteratively trained for the second time based on the second group of Gassmann fluid term elastic parameters, the value of the coupling objective function under the second iteration training of the second group of Gassmann fluid term elastic parameters is obtained, and the reversible neural network is updated again by the back propagation algorithm. According to the above method, based on the generated multiple groups of Gassmann fluid term elastic parameters, the reversible neural network is iteratively trained in turn to obtain the value of the coupling objective function corresponding to each iterative training, and after each iterative training is completed, the reversible neural network is updated based on the back propagation algorithm until the coupling objective function converges, then the training of the reversible neural network is stopped.It is worth mentioning that if the coupling objective function has not converged after the multiple groups of Gassmann fluid term elastic parameters have completed the training of the reversible neural network in sequence, then the iterative training of the reversible neural network is performed again in sequence starting from the first group of Gassmann fluid term elastic parameters based on the multiple groups of Gassmann fluid term elastic parameters until the multiple groups of Gassmann fluid term elastic parameters have completed convergence. Combined with the scenario example, if in the process of training the reversible neural network, if the iterative training of the reversible neural network is performed a total of n times before the coupling objective function converges, then the potential variables obtained in the nth time during the training of the reversible neural network are used as the final potential variables. Based on the survey technology of seismic data, the actual seismic data generated when an earthquake occurs in the area to be observed is obtained, and the actual seismic data generated in the area to be observed and the final potential variables are used as the input data in the reverse mapping process in the reversible neural network, so that the reversible neural network outputs the elastic parameter inversion result, that is, the elastic parameter distribution corresponding to the area to be observed is obtained. This example is based on taking the final potential variable as part of the input data in the reverse mapping process in the reversible neural network, so that the reversible neural network can obtain the inversion result more accurately during the inversion process. Combined with the scenario example, in the process of generating multiple sets of Gassmann fluid term elastic parameters, it is necessary to meet the generation rules of elastic parameters. Based on the generation rules of elastic parameters, the first empirical formula can be first used to generate multiple longitudinal wave velocities in the generation rules of elastic parameters. In order to ensure the diversity of the generated Gassmann fluid term elastic parameters, different second empirical formulas can be used for each longitudinal wave velocity on the multiple longitudinal wave velocities to generate the shear wave velocity corresponding to each longitudinal wave velocity, and different third empirical formulas can be used for each longitudinal wave velocity to generate the density corresponding to each longitudinal wave velocity. The elastic parameters of the Gassmann fluid term are divided into high frequency and low frequency. In addition to obtaining multiple longitudinal wave velocities, shear wave velocities and densities through empirical formulas, multiple sets of random numbers can be added to serve as the remaining longitudinal wave velocities, shear wave velocities and densities, and the low-frequency values ​​in some parameters can also be replaced. In this way, multiple sets of Gassmann fluid term elastic parameters with diversity are obtained.After obtaining multiple groups of Gassmann fluid item elastic parameters, each group of Gassmann fluid item elastic parameters is compared with the generation rule of elastic parameters. If, in the process of calculating by the second empirical formula and the third empirical formula, or in the process of adding multiple groups of random numbers or replacing low-frequency values, the result of a group of Gassmann fluid item elastic parameters does not conform to the generation rule of the elastic parameters, then the group of Gassmann fluid item elastic parameters that do not conform to the generation rule of the elastic parameters is deleted. Finally, each group of Gassmann fluid item elastic parameters obtained finally conforms to the generation rule of elastic parameters. Finally, based on the longitudinal wave velocity, shear wave velocity and density in each group of Gassmann fluid item elastic parameters, the first calculation formula is used to calculate the Gassmann fluid item and shear modulus corresponding to each group of Gassmann fluid item elastic parameters. The Gassmann fluid item and shear modulus are the elastic parameters in each group of Gassmann fluid item elastic parameters, which are finally used to train the reversible neural network. This example limits the generation of multiple sets of Gassmann fluid item elastic parameters by specifying the generation rules of elastic parameters, so that the generated multiple sets of Gassmann fluid item elastic parameters are of higher quality and more valuable, and multiple sets of Gassmann fluid item elastic parameters are generated by multiple empirical formulas, adding multiple sets of random numbers or replacing low-frequency values, so that the generated multiple sets of Gassmann fluid item elastic parameters are more diverse. Combined with the scenario example, since the reversible neural network can perform three functions of forward mapping, reverse mapping and calculation of coupling objective functions, the reversible neural network can be divided into three modules according to the functions, namely, forward reversible block, reverse reversible block and complementary affine coupling layer, the forward reversible block is used to realize the forward mapping process, the reverse reversible block is used to realize the reverse mapping process, and the complementary affine coupling layer is used to realize the calculation of coupling objective functions. Since the elastic parameters of Gassmann fluid items are divided into high frequency and low frequency, the Gassmann fluid items and shear modulus in each group of elastic parameters of Gassmann fluid items are respectively divided into high frequency Gassmann fluid items, low frequency Gassmann fluid items, high frequency shear modulus and low frequency shear modulus, and the high frequency Gassmann fluid items and the high frequency shear modulus are used as the first elastic parameters, and the low frequency Gassmann fluid items and the low frequency shear modulus are used as the second elastic parameters. The first elastic parameter and the second elastic parameter are simultaneously input into the forward reversible block, and after the predicted seismic data and the latent variables are obtained, the predicted seismic data and the latent variables are input into the reverse reversible block of the reversible neural network.If the reverse mapping process is the iterative process of training the reversible neural network, the first elastic parameter x2' and the second elastic parameter x1' outputted are the elastic parameter inversion results outputted in the iterative process; if the reverse mapping process is that the reversible neural network has been trained and is formally used to invert the elastic parameters, the predicted seismic data y1 is replaced by the actual seismic data of the area to be observed, and the latent variable y2 is replaced by the final latent variable, then the first elastic parameter x2' obtained by the reverse mapping is the high-frequency Gassmann fluid term and high-frequency shear modulus of the elastic parameter inversion result of the area to be observed; the second elastic parameter x1' obtained by the reverse mapping is the low-frequency Gassmann fluid term and low-frequency shear modulus of the elastic parameter inversion result of the area to be observed. This example adopts multiple formulas to be applied to the forward mapping and reverse mapping processes, so that the reversible neural network can quickly and accurately obtain the inversion results during training and use. This embodiment constructs a bidirectional reversible neural network model and a coupled objective function, and trains the reversible neural network model. Based on the trained reversible neural network model, the elastic parameter inversion result of the area to be observed is obtained, which reduces the dependence on the initial low-frequency parameters and training samples in the seismic data and improves the accuracy of the inversion result.

[0122] Embodiment 3

[0123] Combined with the scenario example, we first use part of the Marmousi model data to test the model data. The model data includes 3001 common depth point gathers (Common Depth Point, referred to as CDP), each CDP includes 751 sampling points. Fig.10 The figure is a schematic diagram of the elastic parameters, shear modulus and density of the Gassmann fluid term in the Marmousi model, where a is the Gassmann fluid term, b is the shear modulus and c is the density. The reflection coefficient is calculated using the forward equation that directly characterizes the reflection coefficient using the elastic parameters of the Gassmann fluid term and convolved with the Ricker wavelet with a main frequency of 30 Hz and a length of 128 ms to synthesize the pre-stack seismic data as the real seismic data. Fig.11The post-stack seismic data is used as an example. The data of the 1001th and 2001th CDP are extracted as pseudo well data, which can be named Well 1 and Well 2, to verify the accuracy of the inversion results of the method of the present invention. For example, 1000 groups of Gassmann fluid term elastic parameter data are randomly generated based on the Gassmann fluid term elastic parameter data generation rule, and input into the forward mapping of the constructed reversible neural network model for training. After 300 epochs of iteration, the objective function converges. Among them, the activation function is the Sigmoid function, the learning rate is 0.01, and the weight coefficients in the objective function are 0.6, 0.3 and 0.1 respectively. The real seismic data is input into the reverse mapping of the trained reversible neural network model to obtain the inversion result of the Gassmann fluid term elastic parameter. Fig.12 : The inversion results of elastic parameters of Gassmann fluid term corresponding to the real seismic data of the example, where a is the Gassmann fluid term; b is the shear modulus; and c is the density. Fig.13 The schematic diagram of extracting data from well A and well B is shown as an example. Fig.13 It can be seen that a is well 1, b is well 2, and the elastic parameter inversion results obtained by the method of this application are Fig.10 The data shown are generally consistent and have good agreement with the pseudo-well data. For Well 1, the relative errors of the Gassmann fluid term, shear modulus and density inversion results and the true values ​​are 0.0342, 0.0351 and 0.0347, respectively; for Well 2, the values ​​are 0.0349, 0.353 and 0.0356, respectively. The above analysis shows that the method of the present invention can invert the elastic parameters of the Gassmann fluid term with higher accuracy without relying on the initial low-frequency parameters and training samples. It reduces the dependence on the initial low-frequency parameters and training samples in the seismic data and improves the accuracy of the inversion results.

[0124] In order to further verify the effectiveness of the method proposed in this application, this specification uses some F3 open source actual data for trial calculation. The actual data passes through two wells (well A and well B), including 691 CDPs, each CDP includes 400 sampling points, Fig.14 Taking another type of post-stack seismic data as an example, the recorded seismic wavelet is extracted. Based on the Gassmann fluid term elastic parameter data generation rule, 1500 sets of Gassmann fluid term elastic parameter data are randomly generated, and input into the forward mapping of the constructed reversible neural network model for training. After 400 epochs, the objective function converges. Among them, the activation function is the Sigmoid function, the learning rate is 0.01, and the weight coefficients in the objective function are 0.5, 0.4 and 0.1 respectively. The real seismic data is input into the reverse mapping of the trained reversible neural network model to invert the elastic parameters of the Gassmann fluid term. Fig.15The inversion result of elastic parameters of Gassmann fluid term corresponding to another real seismic data is shown as an example. Fig.16 Another schematic diagram of the data of well A and well B extracted, a is well 1, b is well 2, Fig.16 It can be seen that the inversion results of the method of the present invention are in good agreement with the logging data. For Well A, the relative errors of the inverted Gassmann fluid term, shear modulus and density with the logging data are 0.0875, 0.0879 and 0.0881 respectively; for Well B, the relative errors are 0.893, 0.891 and 0.898 respectively. The above analysis shows that the method of the present application can invert the elastic parameters of the Gassmann fluid term with high accuracy using only randomly generated elastic parameter data. It reduces the dependence on the initial low-frequency parameters and training samples in the seismic data, and improves the accuracy of the inversion results.

[0125] Embodiment 4

[0126] Fig.17 This is a schematic diagram of the structure of an electronic device provided in Embodiment 4 of the present application, such as Fig.17 As shown, the electronic device includes:

[0127] The server includes a processor 291 and a memory 292; it may also include a communication interface 293 and a bus 294. The processor 291, the memory 292, and the communication interface 293 may communicate with each other through the bus 294. The communication interface 293 may be used for information transmission. The processor 291 may call the logic instructions in the memory 292 to execute the method of the first embodiment.

[0128] In addition, the logic instructions in the above-mentioned memory 292 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product.

[0129] The memory 292 is a computer-readable storage medium that can be used to store software programs and computer executable programs, such as program instructions / modules corresponding to the method in the embodiment of the present application. The processor 291 executes the functional application and data processing by running the software programs, instructions and modules stored in the memory 292, that is, implementing the method of the above-mentioned embodiment 1.

[0130] The memory 292 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required for at least one function; the data storage area may store data created according to the use of the terminal device, etc. In addition, the memory 292 may include a high-speed random access memory and may also include a non-volatile memory.

[0131] An embodiment of the present application provides a non-temporary computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the method described in the above embodiment.

[0132] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the inventions claimed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary techniques in the art that are not claimed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the claims below.

[0133] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. An inversion method, characterized in that: The method comprises: A reversible neural network model and coupled objective function are constructed, and multiple groups of elastic parameters of Gassmann fluid terms are randomly generated based on the elastic parameter characteristics of the area to be observed; The reversible neural network model is obtained by inputting the multiple groups of Gassmann fluid term elastic parameters into the forward mapping process of the reversible neural network, and iteratively training the reversible neural network based on the coupling objective function until the coupling objective function converges; wherein the forward mapping of the reversible neural network model reflects the mapping of elastic parameters to seismic data, and the reverse mapping of the reversible neural network model reflects the mapping of seismic data to elastic parameters; The actual seismic data of the area to be observed is obtained, and the actual seismic data of the area to be observed is input into the reverse mapping process of the trained reversible neural network model to obtain the elastic parameter inversion result of the area to be observed.

2. The method according to claim 1, characterized in that The construction of the reversible neural network model and the coupling objective function includes: Constructing a reversible neural network model, wherein the loss function of the reversible neural network model includes a supervised loss function defined by the error between seismic data, a first unsupervised loss function defined by the error between the latent variable distribution and the Gaussian distribution, and a second unsupervised loss function defined by the error between the elastic parameter distributions; The supervised loss function is defined using a zero-delay cross-correlation function, the first unsupervised loss function and the second unsupervised loss function are defined using a maximum mean difference function, and according to the set weight coefficients of each loss function, the supervised loss function, the first unsupervised loss function and the second unsupervised loss function are coupled to obtain the coupled objective function.

3. The method according to claim 2, characterized in that The method comprises: inputting the multiple groups of Gassmann fluid term elastic parameters into the forward mapping process of the reversible neural network, and iteratively training the reversible neural network based on the coupling objective function until the coupling objective function converges to obtain a trained reversible neural network model, comprising: Inputting the elastic parameters into a forward modeling equation in which the reflection coefficient is directly characterized by the elastic parameters, to obtain synthetic seismic data corresponding to the elastic parameters; Inputting the elastic parameters into the forward mapping process of the reversible neural network model to obtain predicted seismic data and potential variables corresponding to the elastic parameters output by the reversible neural network model in the forward mapping process; and inputting the predicted seismic data and potential variables corresponding to the elastic parameters into the reverse mapping process of the reversible neural network to obtain predicted elastic parameters corresponding to the elastic parameters output by the reversible neural network model in the reverse mapping process; According to the error between the synthetic seismic data and the predicted seismic data corresponding to the elastic parameters, the error between the latent variable distribution and the Gaussian distribution corresponding to the elastic parameters, and the error between the data distribution of the elastic parameters and the corresponding predicted elastic parameters, the reversible neural network model is iteratively updated using a back propagation algorithm based on the coupling objective function until the coupling objective function converges to obtain the trained reversible neural network model.

4. The method according to claim 3, characterized in that The step of acquiring actual seismic data of the area to be observed, inputting the actual seismic data of the area to be observed into the reverse mapping process of the trained reversible neural network model, and obtaining the elastic parameter inversion result of the area to be observed includes: Acquire actual seismic data and final latent variables of the area to be observed, wherein the final latent variables are the latent variables most recently output by the reversible neural network model in a forward mapping process when the coupling objective function converges; The actual seismic data of the area to be observed and the final potential variables are input into the reverse mapping process of the trained reversible neural network model to obtain the elastic parameter inversion result of the area to be observed.

5. The method according to claim 1, characterized in that The method of randomly generating multiple groups of Gassmann fluid term elastic parameters based on the elastic parameter characteristics of the area to be observed includes: According to an empirical formula, a plurality of longitudinal wave velocities are randomly generated, and the shear wave velocity and density corresponding to each longitudinal wave velocity are calculated, and a plurality of shear wave velocities and a plurality of densities are obtained by adding random numbers and / or replacing low frequencies; Calculating and obtaining the plurality of groups of Gassmann fluid term elastic parameters according to the plurality of longitudinal wave velocities, the plurality of shear wave velocities and the plurality of densities; Determine whether the multiple groups of Gassmann fluid item elastic parameters satisfy the set random generation rules, and remove the elastic parameters that do not satisfy the random generation rules from the multiple groups of Gassmann fluid item elastic parameters, wherein the random generation rules include at least one of the following: the value range of the randomly generated elastic parameters is not greater than the value range of the elastic parameters of the area to be observed, the thinnest layer thickness of the randomly generated elastic parameters is less than the thinnest layer thickness identified according to the seismic data of the area to be observed, and the vertical variation trend of the randomly generated elastic parameters conforms to the geological laws of the area to be observed.

6. The method according to any one of claims 1 to 5, characterized in that: The reversible neural network model includes a forward reversible block and a reverse reversible block of a complementary affine coupling layer; The forward reversible block is used to split the elastic parameter input in the forward mapping process into a first elastic parameter and a second elastic parameter; and process the first elastic parameter based on a first activation function to obtain a first processing result; Processing the first elastic parameter based on a second activation function to obtain a second processing result, and outputting predicted earthquake data by using matrix inner product calculation and summation calculation according to the second elastic parameter, the first processing result and the second processing result; and, processing the predicted earthquake data based on a third activation function to obtain a third processing result; Processing the second elastic parameter based on a fourth activation function to obtain a fourth processing result, and outputting a potential variable by using matrix inner product calculation and summation calculation according to the first elastic parameter, the third processing result and the fourth processing result; The reverse reversible block is used to process the input seismic data based on the third activation function to obtain a fifth processing result during the reverse mapping process; process the input seismic data based on the fourth activation function to obtain a sixth processing result, and output a first predicted elastic parameter based on the fifth processing result, the sixth processing result and the input latent variable by using matrix inner product calculation and difference calculation; and process based on the first activation function and the first predicted elastic parameter to obtain a seventh processing result; An eighth processing result is obtained based on the second activation function and the first predicted elastic parameter. According to the seventh processing result, the eighth processing result and the input seismic data, a second predicted elastic parameter is output by using matrix inner product calculation and difference calculation, and the first predicted elastic parameter and the second predicted elastic parameter are synthesized and output as a predicted elastic parameter.

7. An inversion device, characterized in that: The device comprises: A construction module is used to construct a reversible neural network model and a coupled objective function, and randomly generate multiple groups of elastic parameters of Gassmann fluid terms based on the elastic parameter characteristics of the area to be observed; An input module is used to obtain a trained reversible neural network model by inputting the multiple groups of Gassmann fluid term elastic parameters into the forward mapping process of the reversible neural network and iteratively training the reversible neural network based on the coupling objective function until the coupling objective function converges; wherein the forward mapping of the reversible neural network model reflects the mapping of elastic parameters to seismic data, and the reverse mapping of the reversible neural network model reflects the mapping of seismic data to elastic parameters; The acquisition module is used to acquire the actual seismic data of the area to be observed, input the actual seismic data of the area to be observed into the reverse mapping process of the trained reversible neural network model, and obtain the elastic parameter inversion result of the area to be observed.

8. The device according to claim 7, characterized in that The construction module is specifically used to construct a reversible neural network model, wherein the loss function of the reversible neural network model includes a supervised loss function defined by an error between seismic data, a first unsupervised loss function defined by an error between a latent variable distribution and a Gaussian distribution, and a second unsupervised loss function defined by an error between elastic parameter distributions; The construction module is specifically used to define the supervised loss function using a zero-delay cross-correlation function, define the first unsupervised loss function and the second unsupervised loss function using a maximum mean difference function, and couple the supervised loss function, the first unsupervised loss function and the second unsupervised loss function according to the set weight coefficients of each loss function to obtain the coupled objective function.

9. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

Images