Wave impedance inversion model training method, wave impedance inversion method and electronic equipment

Abstract

The invention relates to the technical field of oil and gas geophysical exploration, and discloses a wave impedance inversion model training method, a wave impedance inversion method and electronic equipment, and the method comprises the steps: constructing a broadband wavelet library through broadband Ricker wavelets according to post-stack seismic data and logging data of each well; establishing a logging interpolation impedance model according to the logging data of the target layer section; according to the logging interpolation impedance model and the broadband wavelet library, constructing a noiseless broadband high-resolution data set and a noisy broadband high-resolution data set; constructing a training sample set according to the noise-free broadband high-resolution data set, the noise-containing broadband high-resolution data set and a logging interpolation impedance model; constructing a generative adversarial network for wave impedance inversion; determining an objective function of wave impedance inversion; and training the generative adversarial network according to the training sample set and the objective function to obtain a wave impedance inversion model. According to the invention, high-resolution impedance inversion of the thin reservoir can be carried out, and the inversion precision of the thin reservoir is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas geophysical exploration, and in particular to a wave impedance inversion model training method, a wave impedance inversion method and electronic equipment. Background Art

[0002] Tight sandstone reservoirs are characterized by obvious stratification and thin reservoir thickness. With the continuous deepening of oil exploration and development, the use of wave impedance inversion to improve the ability to identify thin layers has become an important means of reservoir prediction. Conventional wave impedance inversion methods include two categories: one is deterministic methods, including sparse pulse inversion and other methods; the other is statistical methods, including methods such as statistical sampling of posterior distribution based on the Bayesian theoretical framework.

[0003] Since seismic data are band-limited, impedance inversion has multiple solutions. Both deterministic and statistical methods require prior information to regularize the inversion process, and usually make some simple prior assumptions, such as sparse assumptions or minimum length assumptions. However, these prior assumptions lead to overly smooth impedance inversion results and limited resolution of thin reservoirs.

[0004] In recent years, deep learning methods have developed rapidly, providing new ideas for solving data-driven seismic inversion. They can directly learn the mapping function between seismic data and target parameters end-to-end from the training data set, and have achieved good results in some application examples. However, data-driven deep learning methods are highly dependent on training sample sets, and require the collection of a large amount of observation data and high-resolution data pairs in advance to train the network.

[0005] Therefore, there is an urgent need for a wave impedance inversion model training method and a wave impedance inversion method to solve the above technical problems. Summary of the invention

[0006] In view of the above problems, the present invention provides a wave impedance inversion model training method, a wave impedance inversion method and an electronic device, which can train a wave impedance inversion model capable of performing high-resolution impedance inversion on thin reservoirs without pre-collecting a large amount of observation data and high-resolution data pairs, thereby improving the inversion accuracy of thin reservoirs.

[0007] The present invention provides a wave impedance inversion model training method, the method comprising:

[0008] According to the post-stack seismic data and the logging data of each well, a broadband wavelet library is constructed through broadband Ricker wavelets; wherein the broadband wavelet library includes broadband wavelets of each well and comprehensive broadband wavelets;

[0009] Obtaining the target layer logging data from the logging data of each well, and establishing a logging interpolation impedance model based on the target layer logging data;

[0010] According to the well logging interpolation impedance model and broadband wavelet library, a noise-free broadband high-resolution data set and a noisy broadband high-resolution data set are constructed;

[0011] A training sample set for training the wave impedance inversion model is constructed based on a noise-free broadband high-resolution data set, a noisy broadband high-resolution data set, and a well logging interpolation impedance model;

[0012] According to the data structure characteristics of the training sample set and post-stack seismic data, a generative adversarial network for wave impedance inversion is constructed;

[0013] According to the preset wave impedance inversion requirements, determine the objective function of wave impedance inversion;

[0014] The generative adversarial network is trained according to the training sample set and the objective function to obtain a trained wave impedance inversion model.

[0015] Furthermore, based on the post-stack seismic data and the logging data of each well, a broadband wavelet library is constructed by broadband Ricker wavelet, including:

[0016] Based on the post-stack seismic data and the logging data of each well, each well is calibrated by broadband Ricker wavelet to determine the broadband wavelet of each well. Based on the broadband wavelet of each well, a comprehensive broadband wavelet is determined to construct a broadband wavelet library.

[0017] Furthermore, a logging interpolation impedance model is established based on the logging data of the target layer, including:

[0018] Perform horizon interpretation based on post-stack seismic data to obtain seismic horizons;

[0019] A stratigraphic framework is constructed according to the seismic horizon, and the logging data of the target layer is interpolated along the stratigraphic framework using the Kriging method to obtain the logging interpolation impedance model at the preset sampling time interval.

[0020] Furthermore, based on the well logging interpolation impedance model and the broadband wavelet library, a noise-free broadband high-resolution data set and a noisy broadband high-resolution data set are constructed, including:

[0021] Generate a reflection coefficient model based on the logging interpolation impedance model, and convolve each wavelet in the broadband wavelet library with the reflection coefficient model to generate noise-free broadband high-resolution data corresponding to each wavelet, so as to construct a noise-free broadband high-resolution data set;

[0022] Adding preset Gaussian noise to the noise-free broadband high-resolution data to generate noisy broadband high-resolution data corresponding to each wavelet, so as to construct a noisy broadband high-resolution data set;

[0023] The preset Gaussian noise includes Gaussian noise having the same signal-to-noise ratio as the post-stack seismic data.

[0024] Furthermore, a training sample set is constructed based on the noise-free broadband high-resolution data set, the noisy broadband high-resolution data set and the logging interpolation impedance model, including:

[0025] Each piece of broadband high-resolution data in the noise-free broadband high-resolution data set and the noisy broadband high-resolution data set is normalized to obtain multiple input data, and the logging interpolation impedance model is used as label data to construct a training sample set.

[0026] The present invention also provides a wave impedance inversion method, the method comprising:

[0027] Normalizing the post-stack seismic data, and inputting the normalized post-stack seismic data into a preset wave impedance inversion model to perform wave impedance inversion, and obtaining wave impedance inversion results;

[0028] Wherein, the preset wave impedance inversion model is obtained according to the above-mentioned wave impedance inversion model training method.

[0029] The present invention also provides a wave impedance inversion model training device, the device comprising:

[0030] A broadband wavelet library construction module is used to construct a broadband wavelet library through broadband Ricker wavelets according to post-stack seismic data and logging data of each well; wherein the broadband wavelet library includes broadband wavelets of each well and comprehensive broadband wavelets;

[0031] A well logging interpolation impedance model building module is used to obtain the well logging data of the target layer from the well logging data of each well, and to establish a well logging interpolation impedance model based on the well logging data of the target layer;

[0032] A broadband high-resolution data generation module is used to construct a noise-free broadband high-resolution data set and a noisy broadband high-resolution data set based on the well logging interpolation impedance model and the broadband wavelet library;

[0033] A training sample set construction module is used to construct a training sample set for training a wave impedance inversion model based on a noise-free broadband high-resolution data set, a noisy broadband high-resolution data set, and a well logging interpolation impedance model;

[0034] A generative adversarial network building module is used to build a generative adversarial network for wave impedance inversion based on the data structure characteristics of the training sample set and post-stack seismic data;

[0035] An objective function determination module is used to determine the objective function of wave impedance inversion according to preset wave impedance inversion requirements;

[0036] The model training module is used to train the generative adversarial network according to the training sample set and the objective function to obtain a trained wave impedance inversion model.

[0037] The present invention also provides a wave impedance inversion device, which comprises:

[0038] Normalization processing module, used for normalization processing of post-stack seismic data;

[0039] The wave impedance inversion module is used to input the normalized post-stack seismic data into a preset wave impedance inversion model for wave impedance inversion to obtain a wave impedance inversion result; wherein the preset wave impedance inversion model is obtained according to the above-mentioned wave impedance inversion model training method.

[0040] The present invention also provides a computer-readable storage medium, wherein the computer program stored in the computer-readable storage medium implements the steps of the above method when executed by one or more processors.

[0041] The present invention also provides an electronic device, comprising a memory and one or more processors, wherein a computer program is stored in the memory, and when the computer program is executed by the one or more processors, the steps of the above method are executed.

[0042] The wave impedance inversion model training method, wave impedance inversion method and electronic device provided by the present invention have at least the following beneficial effects:

[0043] (1) The wave impedance inversion model training method provided by the present invention does not need to collect a large amount of observation data and high-resolution data pairs in advance during training. It only needs to construct a broadband wavelet library based on post-stack seismic data and logging data of each well, and then obtain a noise-free broadband high-resolution data set, a noisy broadband high-resolution data set and a logging interpolation impedance model. Then, a training sample set can be constructed to train the generative adversarial network and obtain a wave impedance inversion model with higher accuracy.

[0044] (2) The wave impedance inversion method provided by the present invention performs high-resolution impedance inversion of thin reservoirs according to the wave impedance inversion model obtained through training, which can effectively improve the inversion accuracy of thin reservoirs. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings. The drawings constituting a part of the specification of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments and descriptions in the present invention are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0047] Figure 1 This is a flow chart of the steps of the wave impedance inversion model training method provided in the first embodiment of the present invention;

[0048] Figure 2 This is a schematic diagram of the structure of the wave impedance inversion model training device provided in the third embodiment of the present invention;

[0049] Figure 3 This is a schematic diagram of the structure of a wave impedance inversion device provided in Embodiment 4 of the present invention;

[0050] Figure 4 The wave impedance inversion model training and wave impedance inversion flow chart provided in the fifth embodiment of the present invention;

[0051] Figure 5 The time domain post-stack seismic profile provided in the fifth embodiment of the present invention;

[0052] Figure 6 This is a waveform diagram of the broadband Ricker wavelet provided in the fifth embodiment of the present invention;

[0053] Figure 7 A spectrum comparison diagram of the broadband Ricker wavelet provided in the fifth embodiment of the present invention and a conventional Ricker wavelet with the same main frequency;

[0054] Figure 8 This is a schematic cross-sectional diagram of the time domain well logging interpolation impedance model provided in the fifth embodiment of the present invention;

[0055] Fig. 9 This is a cross-sectional schematic diagram of the time domain reflection coefficient model provided in the fifth embodiment of the present invention;

[0056] Fig.10 This is a schematic diagram of a time-domain noise-free broadband high-resolution seismic profile provided in Embodiment 5 of the present invention;

[0057] Fig.11 A schematic diagram of a preset generative adversarial network structure provided in Embodiment 5 of the present invention;

[0058] Fig.12 A flowchart of the training of a generative adversarial neural network structure provided in Embodiment 5 of the present invention;

[0059] Fig.13 A schematic diagram of an inverse impedance profile obtained by using the method of the present invention provided in Embodiment 5 of the present invention;

[0060] Fig.14 This is a schematic diagram of the inverse impedance profile obtained by a conventional method provided in the fifth embodiment of the present invention;

[0061] Fig.15 This is a schematic diagram of the structure of an electronic device provided in Embodiment 8 of the present invention;

[0062] Reference numerals:

[0063] Figure 2 In: 201- broadband wavelet library construction module, 202- logging interpolation impedance model construction module, 203- broadband high-resolution data generation module, 204- training sample set construction module, 205- generative adversarial network construction module, 206- objective function determination module, 207- model training module;

[0064] Figure 3 Middle: 301-normalization processing module, 302-wave impedance inversion module;

[0065] Fig.15 In it: 1500 - electronic device, 1501 - processor, 1502 - communication bus, 1503 - user interface, 1504 - communication interface, 1505 - memory. DETAILED DESCRIPTION

[0066] The present invention will be further described below in conjunction with the embodiments shown in the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention, and it should be understood that the present invention is not limited to the exemplary embodiments described here.

[0067] It should be noted that the relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

[0068] It should also be understood that any component, data or structure mentioned in the embodiments of the present invention can generally be understood as one or more, unless explicitly limited or otherwise indicated in the context.

[0069] In addition, the term "and / or" in the present disclosure is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present disclosure generally indicates that the associated objects before and after are in an "or" relationship.

[0070] It should also be understood that the description of the various embodiments of the present invention focuses on the differences between the various embodiments, and the same or similar aspects thereof can be referenced to each other, and for the sake of brevity, they will not be described one by one.

[0071] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0072] The embodiments of the present invention may not discuss in detail the technologies, methods, and devices known to ordinary technicians in the relevant fields, but where appropriate, the technologies, methods, and devices should be considered as part of the specification.

[0073] Embodiments of the present invention can be applied to electronic devices such as terminal devices, computer systems, servers, etc., which can operate with many other general or special computing system environments or configurations. Examples of well-known terminal devices, computing systems, environments and / or configurations suitable for use with electronic devices such as terminal devices, computer systems, servers, etc. include, but are not limited to: personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network personal computers, small computer systems, large computer systems, and distributed cloud computing technology environments including any of the above systems, etc.

[0074] Electronic devices such as terminal devices, computer systems, servers, etc. can be described in the general context of computer system executable instructions (such as program modules) executed by computer systems. Generally, program modules can include routines, programs, object programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in a distributed cloud computing environment, where tasks are performed by remote processing devices linked through a communication network. In a distributed cloud computing environment, program modules can be located on local or remote computing system storage media including storage devices.

[0075] Embodiment 1

[0076] In the first embodiment of the present invention, Figure 1 As shown, a wave impedance inversion model training method is provided, and the method specifically includes the following steps:

[0077] Step S101: constructing a broadband wavelet library through broadband Ricker wavelets according to post-stack seismic data and logging data of each well.

[0078] Step S102: acquiring the target layer logging data from the logging data of each well, and establishing a logging interpolation impedance model according to the target layer logging data.

[0079] Step S103: constructing a noise-free broadband high-resolution data set and a noisy broadband high-resolution data set according to the well logging interpolation impedance model and the broadband wavelet library.

[0080] Step S104: constructing a training sample set for training a wave impedance inversion model according to the noise-free broadband high-resolution data set, the noisy broadband high-resolution data set and the logging interpolation impedance model.

[0081] Step S105: constructing a generative adversarial network for wave impedance inversion according to the data structure characteristics of the training sample set and the post-stack seismic data.

[0082] Step S106: Determine the objective function of the wave impedance inversion according to the preset wave impedance inversion requirement.

[0083] Step S107: training the generative adversarial network according to the training sample set and the objective function to obtain a trained wave impedance inversion model.

[0084] Optionally, in step S101, the broadband wavelet library includes broadband wavelets of each well and comprehensive broadband wavelets.

[0085] Based on the post-stack seismic data and the logging data of each well, a broadband wavelet library is constructed through broadband Ricker wavelets, including:

[0086] Based on the post-stack seismic data and the logging data of each well, each well is calibrated by broadband Ricker wavelet to determine the broadband wavelet of each well. Based on the broadband wavelet of each well, a comprehensive broadband wavelet is determined to construct a broadband wavelet library.

[0087] Specifically, in one implementation, based on the broadband sub-waves of each well, the determined comprehensive broadband sub-wave may include:

[0088] According to the broadband wavelets of each well, the average value of the broadband wavelets is calculated and taken as the comprehensive broadband wavelet.

[0089] More specifically, in step S101, the Ricker wavelet R(t) is known, and the broadband Ricker wavelet Y(t) is synthesized by Ricker wavelets of different widths and is defined as:

[0090]

[0091]

[0092] Among them, q and p are the upper and lower limits of the integration of the parameter f0 in the Ricker wavelet R(t).

[0093] In step S101, a broadband wavelet library is constructed by using broadband Ricker wavelets, which are good approximations of impulse functions, with narrow main lobes, small side lobe amplitudes, and simple waveforms. Under the same main lobe width, broadband Ricker wavelets have higher actual resolution.

[0094] It should be understood that the post-stack seismic data are the post-stack seismic data in the study area, and the logging data of each well are the logging data of each well in the study area.

[0095] Optionally, in step S102, a well logging interpolation impedance model is established according to the well logging data of the target layer, including:

[0096] Perform horizon interpretation based on post-stack seismic data to obtain seismic horizons;

[0097] A stratigraphic framework is constructed according to the seismic horizon, and the logging data of the target layer is interpolated along the stratigraphic framework using the Kriging method to obtain the logging interpolation impedance model at the preset sampling time interval.

[0098] It should be noted that the present invention does not impose any special limitation on the preset sampling time interval, and technicians can set it according to actual needs.

[0099] In one implementation, the preset sampling time interval may be set to 1 ms, that is, a well logging interpolation impedance model with 1 ms sampling is obtained.

[0100] In addition, the target layer section mentioned in step S102 refers to the layer section where the tight sandstone reservoir is located.

[0101] Specifically, it is known that the impedance value at the i-th wellbore is z(x i ), set the weight λ of Kriging interpolation i , then the impedance at the point to be interpolated is:

[0102]

[0103] The above formula is the expression of the well logging interpolation impedance model, where z * (x0) is the impedance at the point to be interpolated and n is the total number of wells in the study area.

[0104] Optionally, in step S103, a noise-free broadband high-resolution data set and a noisy broadband high-resolution data set are constructed according to the well logging interpolation impedance model and the broadband wavelet library, including:

[0105] Generate a reflection coefficient model based on the logging interpolation impedance model, and convolve each wavelet in the broadband wavelet library with the reflection coefficient model to generate noise-free broadband high-resolution data corresponding to each wavelet, so as to construct a noise-free broadband high-resolution data set;

[0106] Adding preset Gaussian noise to the noise-free broadband high-resolution data to generate noisy broadband high-resolution data corresponding to each wavelet, so as to construct a noisy broadband high-resolution data set;

[0107] The preset Gaussian noise includes Gaussian noise having the same signal-to-noise ratio as the post-stack seismic data.

[0108] In step S103, the reflection coefficient model is convolved with the broadband wavelet of each well and the comprehensive broadband wavelet to obtain noise-free broadband high-resolution data, which expands the training samples and improves the diversity of samples. Gaussian noise with the same signal-to-noise ratio as the original post-stack seismic data is added to the noise-free broadband high-resolution data to obtain noisy broadband high-resolution data, which is subsequently used as input data for model training, thereby improving the robustness and generalization ability of the generative adversarial network.

[0109] Gaussian noise is a type of noise generated by adding normally distributed random values ​​with mean zero and standard deviation 0 to the input data. It is a continuous probability distribution defined by its probability density function (PDF):

[0110]

[0111] Wherein, x is a random variable, σ is the standard deviation, u is the mean, and in the present invention, u is 0.

[0112] Optionally, in step S104, a training sample set is constructed according to the noise-free broadband high-resolution data set, the noisy broadband high-resolution data set and the well logging interpolation impedance model, including:

[0113] Each piece of broadband high-resolution data in the noise-free broadband high-resolution data set and the noisy broadband high-resolution data set is normalized to obtain multiple input data, and the logging interpolation impedance model is used as label data to construct a training sample set.

[0114] Specifically, considering that seismic data has positive and negative amplitude values, the network convolution training process is affected by the convolution kernel operation and activation function. Negative parameters may cause node death and inaccurate training results. In order to facilitate neural network operations, referring to the application characteristics of deep learning, the seismic data is preprocessed to normalize the positive and negative amplitude values ​​to the [0,1] interval. The normalization formula is:

[0115]

[0116] Among them, S norm Represent the seismic data before and after normalization, S max , S min are the maximum and minimum amplitude values ​​of the seismic data, respectively.

[0117] According to the above normalization formula, the noise-free and noisy broadband high-resolution data are normalized respectively as input data, and the logging interpolation impedance model is used as label data to construct the training sample set of the impedance inversion neural network.

[0118] Optionally, in step S105, a generative adversarial network for wave impedance inversion is constructed, including constructing a generative adversarial network suitable for high-resolution impedance inversion of thin reservoirs.

[0119] See also Fig.11 The constructed generative adversarial network structure includes two parts: the generator and the discriminator, and both the generator and the discriminator are composed of convolutional neural networks. G(z) represents the estimated impedance generated by the generator, z(t) represents the real impedance, and s(t) represents the post-stack seismic record. The generator consists of a five-layer network, including two convolutional layers (Convolutional, referred to as CONV), two sub-sampling layers (Sub-Sampling) and one fully connected layer (Fully-Connect); the discriminator consists of a three-layer network, including one fully connected layer and two deconvolution layers (Transposed Convolution, referred to as CONV_tran). The network first trains the discriminator, which is essentially a binary classifier. It marks the input seismic impedance training sample as 1, that is, the real impedance, and marks the output value of the initial generator as 0, that is, the generated impedance. The generator is a convolutional neural network that only contains the forward propagation algorithm, and its network parameter updates are updated by the discriminator.

[0120] For the generator, the preprocessed post-stack seismic data and impedance data are extracted through the convolution layer and the downsampling layer. This process mainly obtains high-dimensional data features through convolution kernel operations. The convolution layer calculation process is:

[0121]

[0122] in, is a two-dimensional convolution operator; σ is an activation function; W is a network weight parameter, initialized to a small random number; b is a network bias value, initialized to 0; H is the high-dimensional data output after the convolution kernel operation. Through the analysis of the characteristics of seismic data and the verification of experimental results, the activation function in the present invention is preferably a ReLU function to improve the nonlinear fitting ability of the model.

[0123] σ(x)=max(0,x)

[0124] Where x is the input feature value. The network weight parameters and bias values ​​are continuously updated as the network is continuously trained. The dimension of the output data of the convolutional layer is high, which affects the training speed of the network and the accuracy of the training results. To this end, the downsampling process of the image is used to reduce the dimension of the high-dimensional feature data, and retain the feature data that carries position information and has strong correlation. The feature data is transmitted to the fully connected layer, and the error is judged by the loss function.

[0125] Preferably, the loss function of the generative adversarial network in the present invention adopts JS divergence, and the formula is:

[0126]

[0127] Among them, P g (z) represents the probability distribution of the estimated impedance generated by the generator, P r (z) represents the probability distribution of the true impedance, and KL represents the KL divergence, and the formula is:

[0128]

[0129] For the discriminator, the network parameters are updated according to the loss function. When the generated sample has a large deviation from the real sample, the loss function can provide a large gradient. The network updates the network parameters along the gradient descent direction and passes the updated parameters to the discriminator and the generator. With continuous training, the discriminator's ability to identify the authenticity of the impedance is continuously enhanced, and finally the optimal discriminator is achieved.

[0130] The training process of the generative adversarial network is cross-training, and the discriminator and generator are continuously trained and optimized at the same time. The generator does not contain a back-propagation algorithm, and its network parameter update depends on the error back-propagation algorithm of the discriminator. After each parameter update of the generator, an estimated impedance is generated from a new set of post-stack seismic records. The generator is similar to an inverse convolutional neural network. After the impedance is input to the generator, it is converted into network data features through a fully connected layer. The specific form can be compared to the data features obtained after the impedance data in the discriminator passes through the convolution layer. At this time, the data features obtained by the generator from the post-stack seismic records are reconstructed through the deconvolution layer, and the reconstructed estimated impedance G(z) is obtained after passing through two layers of deconvolution layers.

[0131] The specific formula of the deconvolution layer is:

[0132] z=(d-1)×m+k

[0133] Among them, z is the generated impedance; d is the eigenvalue after the full connection layer transformation; m is the sliding step size, preferably, in the present invention, m is 2; k represents the convolution kernel size, preferably, in the present invention, k is 5.

[0134] Optionally, in step S106, the preset wave impedance inversion requirement includes the thin reservoir high-resolution impedance inversion requirement. After the generative adversarial network framework is built, in order to achieve high-resolution impedance inversion, the impedance generated by the generator needs to be able to continuously approach the real impedance of the post-stack seismic record:

[0135] J = min(G(z)-z) 2

[0136] Where G(z) represents the estimated impedance produced by the generator.

[0137] In order to make the value of J in the formula as close to 0 as possible, according to the requirements of high-resolution impedance inversion of thin reservoirs, in one implementation, the objective function V of the generative adversarial network is set to:

[0138] min G max D V[D,G]=E x [lgD(z)]+E′ w [lg(1-D(G(z′)))]

[0139] Among them, D(z) represents the optimal discriminator, and D(G(z′)) represents the identification result of the optimal discriminator on the estimated impedance generated by the generator. The generative adversarial network mainly learns the probability distribution characteristics of the data, and the objective function is used to measure the degree of consistency between the generated impedance and the real impedance. Analyzing the objective function, for a fixed generator G, the discriminator D needs to continuously improve its identification ability, that is, maximize the cross entropy between the two, so as to accurately distinguish the difference between the generated samples and the real samples; for a fixed discriminator D, it is necessary to continuously improve the accuracy of the samples generated by the generator G, so that D is difficult to distinguish the difference between the two. There is a constant confrontation and game between the two during the learning process.

[0140] Optionally, in step S107, the generative adversarial network is trained according to the training sample set and the objective function, including: inputting the training sample set into the constructed generative adversarial network structure, starting network training, continuously updating network parameters, and obtaining a network model suitable for high-resolution impedance inversion of thin reservoirs.

[0141] The specific implementation method is as follows:

[0142] The training sample set is input into the constructed generative adversarial network structure to start network training. The entire generative adversarial network training process is carried out alternately by the generator and the discriminator. Each training will alternately input the well logging interpolation impedance model in the training set and the estimated impedance generated by the generator into the discriminator network, extract the data features through the convolution layer, and then reduce the dimension of the feature data through the downsampling layer to prevent overfitting, and finally output 32 feature atlases. With the feature atlas as the identification condition, the feature of the well logging interpolation impedance model is marked as 1 (true impedance), and the generated impedance is judged as the middle value between 0 (pseudo impedance) and 1 (true impedance) through the loss function error. If the judgment result is greater than 0.5, it is considered that the generated impedance is the true impedance corresponding to the post-stack seismic data.

[0143] During the training of the entire network, the estimated impedance output by the generator and the real impedance are continuously matched in the discriminator. The parameters of the generator and the discriminator are updated according to the objective function. Through the game between the generator and the discriminator, the estimated impedance is continuously converged to the real impedance. When the generative adversarial network reaches the Nash equilibrium, that is, the moment when the discriminator cannot distinguish the estimated impedance from the real impedance, it is considered that the result impedance generated by the generator in the last iteration is the real impedance in the corresponding post-stack seismic data.

[0144] The wave impedance inversion model training method provided in this embodiment does not need to collect a large amount of observation data and high-resolution data pairs in advance during training. It only needs to build a broadband sub-wave library based on post-stack seismic data and logging data of each well, and then obtain a noise-free broadband high-resolution data set, a noisy broadband high-resolution data set and a logging interpolation impedance model. Then, a training sample set can be constructed to train the generative adversarial network and obtain a wave impedance inversion model with higher accuracy.

[0145] Embodiment 2

[0146] In a second embodiment of the present invention, a wave impedance inversion method is provided, and the method specifically includes:

[0147] Normalizing the post-stack seismic data, and inputting the normalized post-stack seismic data into a preset wave impedance inversion model to perform wave impedance inversion, and obtaining wave impedance inversion results;

[0148] The preset wave impedance inversion model is obtained according to the wave impedance inversion model training method in the first embodiment.

[0149] In this embodiment, based on the trained network model, high-resolution impedance inversion is performed on actual post-stack seismic data, and normalization preprocessing is also performed on the seismic data to adjust the value range to the [0,1] interval to prevent the negative amplitude of the seismic record from affecting the operation of the convolution kernel and the activation function. The preprocessed seismic record is input into the previously trained neural network model to obtain the corresponding impedance inversion result.

[0150] The wave impedance inversion method provided in this embodiment performs high-resolution impedance inversion of thin reservoirs according to the wave impedance inversion model obtained by the wave impedance inversion model training method provided in Example 1, which can effectively improve the inversion accuracy of thin reservoirs.

[0151] Embodiment 3

[0152] In a third embodiment of the present invention, a wave impedance inversion model training device is provided, such as Figure 2 As shown, the device comprises:

[0153] The broadband wavelet library construction module 201 is used to construct a broadband wavelet library through broadband Ricker wavelets according to the post-stack seismic data and the logging data of each well; wherein the broadband wavelet library includes broadband wavelets of each well and comprehensive broadband wavelets;

[0154] The well logging interpolation impedance model building module 202 is used to obtain the well logging data of the target layer from the well logging data of each well, and to establish the well logging interpolation impedance model according to the well logging data of the target layer;

[0155] The broadband high-resolution data generation module 203 is used to construct a noise-free broadband high-resolution data set and a noisy broadband high-resolution data set according to the well logging interpolation impedance model and the broadband wavelet library;

[0156] A training sample set construction module 204 is used to construct a training sample set for training a wave impedance inversion model according to a noise-free broadband high-resolution data set, a noisy broadband high-resolution data set, and a well logging interpolation impedance model;

[0157] The generative adversarial network construction module 205 is used to construct a generative adversarial network for wave impedance inversion according to the data structure characteristics of the training sample set and the post-stack seismic data.

[0158] The objective function determination module 206 is used to determine the objective function of the wave impedance inversion according to the preset wave impedance inversion requirements.

[0159] The model training module 207 is used to train the generative adversarial network according to the training sample set and the objective function to obtain the wave impedance inversion model.

[0160] Optionally, the broadband wavelet library construction module 201 constructs a broadband wavelet library through broadband Ricker wavelets according to the post-stack seismic data and the logging data of each well, including:

[0161] Based on the post-stack seismic data and the logging data of each well, each well is calibrated by broadband Ricker wavelet to determine the broadband wavelet of each well. Based on the broadband wavelet of each well, a comprehensive broadband wavelet is determined to construct a broadband wavelet library.

[0162] Optionally, the well logging interpolation impedance model building module 202 builds a well logging interpolation impedance model according to the well logging data of the target layer, including:

[0163] Perform horizon interpretation based on post-stack seismic data to obtain seismic horizons;

[0164] A stratigraphic framework is constructed according to the seismic horizon, and the logging data of the target layer is interpolated along the stratigraphic framework using the Kriging method to obtain a logging interpolation impedance model with a preset sampling time interval.

[0165] Optionally, the broadband high-resolution data generation module 203 constructs a noise-free broadband high-resolution data set and a noisy broadband high-resolution data set according to the well logging interpolation impedance model and the broadband wavelet library, including:

[0166] Generate a reflection coefficient model based on the logging interpolation impedance model, and convolve each wavelet in the broadband wavelet library with the reflection coefficient model to generate noise-free broadband high-resolution data corresponding to each wavelet, so as to construct a noise-free broadband high-resolution data set;

[0167] Adding preset Gaussian noise to the noise-free broadband high-resolution data to generate noisy broadband high-resolution data corresponding to each wavelet, so as to construct a noisy broadband high-resolution data set;

[0168] The preset Gaussian noise includes Gaussian noise having the same signal-to-noise ratio as the post-stack seismic data.

[0169] Optionally, the training sample set construction module 204 constructs a training sample set according to the noise-free broadband high-resolution data set, the noisy broadband high-resolution data set and the well logging interpolation impedance model, including:

[0170] Each piece of broadband high-resolution data in the noise-free broadband high-resolution data set and the noisy broadband high-resolution data set is normalized to obtain multiple input data, and the logging interpolation impedance model is used as label data to construct a training sample set.

[0171] The wave impedance inversion model training device provided in this embodiment does not need to collect a large amount of observation data and high-resolution data pairs in advance during training. It only needs to build a broadband sub-wave library based on post-stack seismic data and logging data of each well, and then obtain a noise-free broadband high-resolution data set, a noisy broadband high-resolution data set and a logging interpolation impedance model. A training sample set can be constructed to train the generative adversarial network and obtain a wave impedance inversion model with higher accuracy.

[0172] Embodiment 4

[0173] In a fourth embodiment of the present invention, a wave impedance inversion device is provided, such as Figure 3 As shown, the device comprises:

[0174] A normalization processing module 301 is used to perform normalization processing on post-stack seismic data;

[0175] The wave impedance inversion module 302 is used to input the normalized post-stack seismic data into a preset wave impedance inversion model to perform wave impedance inversion and obtain a wave impedance inversion result; wherein the preset wave impedance inversion model is obtained according to the wave impedance inversion model training method in the first embodiment.

[0176] The wave impedance inversion method provided in this embodiment performs high-resolution impedance inversion of thin reservoirs according to the wave impedance inversion model obtained by the wave impedance inversion model training method provided in Example 1, which can effectively improve the inversion accuracy of thin reservoirs.

[0177] Embodiment 5

[0178] In the fifth embodiment of the present invention, taking the tight sandstone reservoir in the Sichuan Basin in southwest China as an example, the wave impedance inversion model training method and the wave impedance inversion method in the above embodiments of the present invention are used to train a wave impedance inversion model, and high-resolution impedance inversion is performed on the tight sandstone reservoir in the Sichuan Basin in southwest China based on the wave impedance inversion model. The specific process is as follows: Figure 4 shown.

[0179] Figure 5 The embodiment shows a post-stack seismic profile with a time domain range of 1.5s to 2.0s and a duration of 0.5s. The tight sandstone reservoir is located above the strong wave trough of 1.8s to 1.5s, showing a thin-layered seismic response characteristic. Therefore, the method provided in the above embodiment of the present invention is used to perform high-resolution impedance inversion of the thin reservoir.

[0180] In step (1), based on the post-stack seismic data of the study area and the logging data of each well in the study area, the well seismic calibration is performed on each well in the study area based on the broadband Ricker wavelet, the broadband wavelet corresponding to each well is determined, and the comprehensive broadband wavelet suitable for the study area is obtained to form a broadband wavelet library.

[0181] The broadband Ricker wavelet waveform is as follows Figure 6 As shown, its main lobe is relatively narrow, the side lobe amplitude is small, the waveform is simple, and it has a higher resolution. Figure 7 This is a spectrum comparison diagram of the broadband Ricker wavelet and the conventional Ricker wavelet with the same main frequency. It can be seen that the high-frequency band of the broadband Ricker wavelet is wider than that of the conventional Ricker wavelet.

[0182] In step (2), firstly, the actual post-stack seismic data is used to interpret the horizon and build a stratigraphic framework; then the Kriging method is used to interpolate the well logging data of the target layer along the stratigraphic framework to obtain a well logging interpolation impedance model with 1 ms sampling, such as Figure 8 shown.

[0183] In step (3), the corresponding reflection coefficient model is generated according to the well logging interpolation impedance model. In order to expand the training samples and improve the diversity of samples, the reflection coefficient model is convolved with each well sub-wavelet and the comprehensive sub-wavelet to obtain noise-free broadband high-resolution data. In order to improve the robustness and generalization ability of the adversarial network, the signal-to-noise ratio of the original post-stack seismic data is evaluated, and Gaussian noise with the corresponding signal-to-noise ratio is added to the noise-free broadband high-resolution data to obtain noisy broadband high-resolution data.

[0184] Fig. 9 for Figure 8 The target layer reflection coefficient model generated by the logging interpolation model is shown in the figure. Fig.10 The noise-free broadband high-resolution data is obtained by convolving the reflection coefficient model with the comprehensive broadband wavelet extracted in step (1). The resolution of this data is relatively high. Figure 5 Actual seismic data shown are high.

[0185] In step (4), since seismic data has positive and negative amplitude values, the network convolution training process is affected by the convolution kernel operation and activation function. Negative parameters may cause node death and inaccurate training results. In order to facilitate neural network operations, referring to the application characteristics of deep learning, the noise-free and noisy broadband high-resolution seismic data are preprocessed to normalize the positive and negative amplitude values ​​to within the interval. The normalized broadband high-resolution seismic data is used as input data, and the well logging interpolation impedance model is used as label data to construct a training sample set for the impedance inversion neural network.

[0186] In step (5), according to the data structure characteristics of the training sample set and the actual post-stack seismic data, a generative adversarial network structure suitable for high-resolution impedance inversion of thin reservoirs is constructed, including a generator and a discriminator. The structure is as follows: Fig.11 As shown in the figure, the generator consists of a five-layer network, including two convolutional layers (CONV), two sub-sampling layers (Sub-Sampling) and a fully connected layer (Fully-Connect); the discriminator consists of a three-layer network, including a fully connected layer and two deconvolution layers (CONV_tran).

[0187] In step (6), according to the requirements of high-resolution impedance inversion of thin reservoirs, an inversion objective function is set, which is used to measure the degree of consistency between the generated impedance and the true impedance. Through continuous training of the network and continuous adjustment of parameters, the impedance generated by the generator can continuously approach the true impedance of the post-stack seismic record.

[0188] In step (7), the training sample set is input into the constructed generative adversarial network structure to start network training. The entire generative adversarial network training process is carried out alternately by the generator and the discriminator. The estimated impedance output by the generator and the real impedance are continuously matched in the discriminator. Through the game between the generator and the discriminator, the network parameters are continuously updated so that the estimated impedance continues to converge to the real impedance. The process is as follows: Fig.12 shown.

[0189] When the generative adversarial network reaches the Nash equilibrium, that is, the moment when the discriminator cannot distinguish the estimated impedance from the true impedance, the result impedance generated by the generator in the last iteration is considered to be the true impedance in the corresponding post-stack seismic record. At this time, a network model suitable for high-resolution impedance inversion of thin reservoirs is obtained.

[0190] In step (8), based on the trained network model, Figure 5 The actual post-stack seismic data shown in the figure are used for high-resolution impedance inversion. The seismic data are also normalized and preprocessed to adjust the value range to within the interval to avoid the negative amplitude of the seismic record affecting the operation of the convolution kernel and the activation function.

[0191] The pre-processed seismic records are input into the previously trained neural network model to obtain the corresponding impedance inversion results, such as Fig.13 shown.

[0192] Fig.14 This is the conventional post-stack impedance inversion result. Fig.13 and Fig.14 It can be seen that the high-resolution impedance inversion method for thin reservoirs based on generative adversarial networks proposed in the present invention has a higher resolution of inversion results than the conventional post-stack impedance inversion method, can better distinguish tight sandstone thin reservoirs in details, and is more consistent with actual drilling.

[0193] In view of the thin thickness of tight sandstone reservoirs, this embodiment uses broadband Ricker wavelet for well seismic calibration, establishes a broadband wavelet library suitable for the study area, and convolves with the reflection coefficient of the logging interpolation impedance model to obtain broadband high-resolution data. This is used as a training sample set to carry out training of a generative adversarial network, and a thin reservoir high-resolution impedance inversion method based on a generative adversarial network is formed, thereby improving the inversion accuracy of thin reservoirs and providing strong data support for oil and gas geophysical exploration of tight sandstone thin reservoirs in the study area.

[0194] Embodiment 6

[0195] In the sixth embodiment of the present invention, a computer program product is further provided. The computer program product includes a computer program or instructions. When the computer program or instructions are executed by a processor, all or part of the steps of the method described in the above embodiments are implemented.

[0196] Further, the computer program product may include one or more computer executable components configured to execute the embodiments when the program is run; the computer program product may also include a computer program tangibly contained on a readable medium, the computer program including program code for executing any method in the embodiments of the present invention. In such an embodiment, the computer program may be downloaded and installed from a network through a communication part, and / or installed from a removable medium.

[0197] Embodiment 7

[0198] In the seventh embodiment of the present invention, a computer-readable storage medium is further provided. The computer program stored in the computer-readable storage medium, when executed by one or more processors, implements all or part of the steps of the method described in the above embodiments.

[0199] Each functional unit in each embodiment of the present invention may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. If the functions are implemented in the form of software functional units and sold or used as independent products, they may be stored in a computer-readable storage medium.

[0200] Computer readable storage media can use any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (Random Access Memory, referred to as RAM), a read-only memory (Read-Only Memory, referred to as ROM), an erasable programmable read-only memory (Electrically Erasable Programmable read only memory, referred to as EPROM), an optical fiber, a portable compact disc read-only memory (Compact Disc Read-Only Memory, referred to as CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0201] Embodiment 8

[0202] In the eighth embodiment of the present invention, an electronic device 1500 is also provided, which may be a mobile phone, a computer, or a tablet computer. Fig.15 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention is shown in FIG. Fig.15As shown, the electronic device 1500 includes: at least one processor 1501, at least one communication bus 1502, a user interface 1503, at least one external communication interface 1504, and a memory 1505. Among them, the communication bus 1502 is configured to realize the connection and communication between these components. Among them, the user interface 1503 may include a display screen, and the external communication interface 1504 may include a standard wired interface and a wireless interface. A computer program is stored in the memory 1505, and the memory 1505 and one or more processors 1501 are connected to each other in communication. When the computer program is executed by one or more processors, the processor 1501 is configured to execute the computer program stored in the memory to implement all or part of the steps of the method in the above embodiment.

[0203] The processor can be an application specific integrated circuit (ASIC), a digital signal processor (DSP), a programmable logic device (PLD), a field programmable gate array (FPGA), a controller, a microcontroller, a microprocessor or other electronic components to execute all or part of the steps of the method described in the above embodiment, which will not be repeated in this embodiment.

[0204] The memory can be implemented by any type of volatile or non-volatile storage device or a combination of them, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0205] The wave impedance inversion model training method, wave impedance inversion method and electronic device provided by the present invention do not need to collect a large amount of observation data and high-resolution data pairs in advance during training. It is only necessary to build a broadband wavelet library based on post-stack seismic data and logging data of each well, and then obtain a noise-free broadband high-resolution data set, a noisy broadband high-resolution data set and a logging interpolation impedance model, so as to construct a training sample set to train a generative adversarial network and obtain a wave impedance inversion model with higher accuracy; high-resolution impedance inversion of thin reservoirs is performed according to the wave impedance inversion model obtained through training, which can effectively improve the inversion accuracy of thin reservoirs.

[0206] The basic principles of the present invention are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present disclosure. In addition, the specific details disclosed above are only for the purpose of illustration and facilitation of understanding, and are not limiting. The above details do not limit the present invention to being implemented by adopting the above specific details.

[0207] The block diagrams of the devices, apparatuses, equipment, and systems involved in the present invention are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems may be connected, arranged, and configured in any manner. Words such as "including," "comprising," "having," and the like are open-ended words, meaning "including but not limited to," and may be used interchangeably therewith.

[0208] It should also be noted that in the apparatus, device and method disclosed herein, each component or each step can be decomposed and / or reassembled, and such decomposition and / or reassembly should be regarded as an equivalent solution of the present invention.

[0209] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present invention. Therefore, the present invention is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0210] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present invention to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A wave impedance inversion model training method, characterized in that: The method comprises: According to the post-stack seismic data and the logging data of each well, a broadband wavelet library is constructed by using broadband Ricker wavelets; wherein the broadband wavelet library includes broadband wavelets of each well and comprehensive broadband wavelets; Acquire target layer logging data from the logging data of each well, and establish a logging interpolation impedance model based on the target layer logging data; Constructing a noise-free broadband high-resolution data set and a noisy broadband high-resolution data set according to the well logging interpolation impedance model and the broadband wavelet library; Constructing a training sample set for training a wave impedance inversion model according to the noise-free broadband high-resolution data set, the noisy broadband high-resolution data set and the well logging interpolation impedance model; Constructing a generative adversarial network for wave impedance inversion according to the data structure characteristics of the training sample set and the post-stack seismic data; According to the preset wave impedance inversion requirements, determine the objective function of wave impedance inversion; The generative adversarial network is trained according to the training sample set and the objective function to obtain a trained wave impedance inversion model.

2. The wave impedance inversion model training method according to claim 1, characterized in that: The method of constructing a broadband wavelet library by broadband Ricker wavelet based on post-stack seismic data and well logging data of each well includes: Based on the post-stack seismic data and the logging data of each well, each well is calibrated by broadband Ricker wavelet to determine the broadband wavelet of each well, and based on the broadband wavelet of each well, a comprehensive broadband wavelet is determined to construct the broadband wavelet library.

3. The wave impedance inversion model training method according to claim 1, characterized in that: The step of establishing a well logging interpolation impedance model according to the well logging data of the target layer comprises: Performing horizon interpretation according to the post-stack seismic data to obtain seismic horizons; A stratigraphic framework is constructed according to the seismic horizon, and the well logging data of the target layer section is interpolated along the stratigraphic framework using the Kriging method to obtain a well logging interpolation impedance model at a preset sampling time interval.

4. The wave impedance inversion model training method according to claim 1, characterized in that: The step of constructing a noise-free broadband high-resolution data set and a noisy broadband high-resolution data set according to the well logging interpolation impedance model and the broadband wavelet library comprises: Generate a reflection coefficient model according to the logging interpolation impedance model, and convolve each wavelet in the broadband wavelet library with the reflection coefficient model to generate noise-free broadband high-resolution data corresponding to each wavelet, so as to construct the noise-free broadband high-resolution data set; Adding preset Gaussian noise to the noise-free broadband high-resolution data to generate noisy broadband high-resolution data corresponding to each wavelet, so as to construct the noisy broadband high-resolution data set; The preset Gaussian noise includes Gaussian noise having the same signal-to-noise ratio as the post-stack seismic data.

5. The wave impedance inversion model training method according to claim 1, characterized in that: The step of constructing a training sample set according to the noise-free broadband high-resolution data set, the noisy broadband high-resolution data set and the well logging interpolation impedance model comprises: Each piece of broadband high-resolution data in the noise-free broadband high-resolution data set and the noisy broadband high-resolution data set is normalized to obtain multiple input data, and the well logging interpolation impedance model is used as label data to construct a training sample set.

6. A wave impedance inversion method, characterized in that: The method comprises: Normalizing the post-stack seismic data, and inputting the normalized post-stack seismic data into a preset wave impedance inversion model to perform wave impedance inversion, and obtaining wave impedance inversion results; Wherein, the preset wave impedance inversion model is obtained according to the method described in any one of claims 1 to 5.

7. A wave impedance inversion model training device, characterized in that: The device comprises: A broadband wavelet library construction module is used to construct a broadband wavelet library through broadband Ricker wavelets according to post-stack seismic data and well logging data of each well; wherein the broadband wavelet library includes broadband wavelets of each well and comprehensive broadband wavelets; A well logging interpolation impedance model building module is used to obtain target layer logging data from the well logging data of each well, and to build a well logging interpolation impedance model based on the target layer logging data; A broadband high-resolution data generation module, used to construct a noise-free broadband high-resolution data set and a noisy broadband high-resolution data set according to the well logging interpolation impedance model and the broadband wavelet library; A training sample set construction module, used to construct a training sample set for training a wave impedance inversion model according to the noise-free broadband high-resolution data set, the noisy broadband high-resolution data set and the well logging interpolation impedance model; A generative adversarial network construction module, used to construct a generative adversarial network for wave impedance inversion according to the data structure characteristics of the training sample set and the post-stack seismic data; An objective function determination module is used to determine the objective function of wave impedance inversion according to preset wave impedance inversion requirements; The model training module is used to train the generative adversarial network according to the training sample set and the objective function to obtain a trained wave impedance inversion model.

8. A wave impedance inversion device, characterized in that: The device comprises: Normalization processing module, used for normalization processing of post-stack seismic data; The wave impedance inversion module is used to input the normalized post-stack seismic data into a preset wave impedance inversion model to perform wave impedance inversion and obtain a wave impedance inversion result; wherein the preset wave impedance inversion model is obtained according to the method described in any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that: The computer program stored in the computer-readable storage medium, when executed by one or more processors, implements the steps of the method according to any one of claims 1 to 6.

10. An electronic device, characterized in that: The method comprises a memory and one or more processors, wherein a computer program is stored in the memory, and when the computer program is executed by the one or more processors, the steps of the method according to any one of claims 1 to 6 are performed.

Images