Organic matter content prediction method, device, equipment and storage medium

Abstract

The invention discloses an organic matter content prediction method, device and equipment and a storage medium, and relates to the field of geosciences and applied sciences. The method comprises the following steps: acquiring seismic data and logging data, generating a large-offset incident angle gather and a small-offset incident angle gather based on the seismic data, matching the logging data with the seismic data through well-seismic calibration and time-depth conversion to obtain an initial low-frequency model, and calculating a low-frequency model according to the initial low-frequency model; then, based on seismic data, the initial low-frequency model, the large-offset incident angle gather and the small-offset incident angle gather, an approximate equation is adopted for inversion, an inversion result is obtained, according to the inversion result, anisotropic parameters are obtained through calculation, then an organic matter content prediction model is trained according to logging data and the anisotropic parameters, and an organic matter content prediction model is obtained. And obtaining a trained organic matter content prediction model. According to the method, the anisotropic parameters in the seismic data are utilized to accurately predict the content of the organic matters in the shale reservoir.

Description

Technical Field

[0001] The embodiments of the present application relate to the fields of earth science and applied science, and particularly to a method, device, equipment and storage medium for predicting organic matter content. Background Art

[0002] At the present stage, in addition to the exploration of conventional oil and gas reservoirs, the exploration of unconventional oil and gas reservoirs is also the focus of oil and gas exploration.

[0003] In the related art, pre-stack seismic AVO (Amplitude Versus Offset) inversion is used to obtain underground elastic and anisotropic parameters, and based on the underground elastic and anisotropic parameters, reservoir lithology, physical properties and oil and gas bearing properties are further predicted.

[0004] However, due to the large number of parameters to be inverted in the vertically transverse isotropic layer, the problems of high ill-posedness and low accuracy occur when using the related art for inversion. Summary of the Invention

[0005] The embodiments of the present application provide a method, device, equipment and storage medium for predicting organic matter content. The technical solution is as follows:

[0006] According to one aspect of the embodiments of the present application, a method for predicting organic matter content is provided, and the method includes:

[0007] Obtain seismic data and logging data, where the seismic data is the data collected after seismic waves propagate in underground media, and the logging data is the data measured from a drilling well;

[0008] Based on the seismic data, generate a large-offset incident angle gather and a small-offset incident angle gather;

[0009] Match the logging data and the seismic data through well-seismic calibration and time-depth conversion to obtain the initial low-frequency model, which is used to describe the wave velocity distribution and density distribution of the underground media, and the initial low-frequency model includes a velocity distribution model and a density distribution model;

[0010] Based on the seismic data, the initial low-frequency model, the large-offset incident angle gather and the small-offset incident angle gather, perform inversion using an approximate equation to obtain an inversion result, where the inversion result includes an anisotropic inversion result and an isotropic inversion result;

[0011] Based on the anisotropic inversion result and the isotropic inversion result, calculate anisotropic parameters, where the anisotropic parameters are used to indicate the degree of anisotropy;

[0012] Based on the logging data and the anisotropic parameters, train an organic matter content prediction model to obtain a trained organic matter content prediction model; wherein, the trained organic matter content prediction model is used to predict the organic matter content of the shale reservoir.

[0013] According to one aspect of the embodiments of the present application, there is provided an organic matter content prediction device, the device includes:

[0014] An acquisition module, configured to acquire seismic data and logging data, where the seismic data is data collected after seismic waves propagate in underground media, and the logging data is data measured from a drilling well;

[0015] A generation module, configured to generate a large-offset incidence angle gather and a small-offset incidence angle gather based on the seismic data;

[0016] A matching module, configured to match the logging data and the seismic data through well-seismic calibration and time-depth conversion to obtain the initial low-frequency model, where the initial low-frequency model is used to describe the wave velocity distribution and density distribution of the underground media, and the initial low-frequency model includes a velocity distribution model and a density distribution model;

[0017] An inversion module, configured to perform inversion on the seismic data, the initial low-frequency model, the large-offset incidence angle gather, and the small-offset incidence angle gather by using an approximate equation to obtain an inversion result, where the inversion result includes an anisotropic inversion result and an isotropic inversion result;

[0018] A calculation module, configured to calculate anisotropic parameters based on the anisotropic inversion result and the isotropic inversion result, where the anisotropic parameters are used to indicate the degree of anisotropy;

[0019] A training module, configured to train an organic matter content prediction model based on the logging data and the anisotropic parameters to obtain a trained organic matter content prediction model; wherein, the trained organic matter content prediction model is used to predict the organic matter content of the shale reservoir.

[0020] According to one aspect of the embodiments of the present application, there is provided a computer device, the computer device includes a processor and a memory, and a computer program is stored in the memory, and the computer program is loaded and executed by the processor to implement the above method.

[0021] According to one aspect of the embodiments of the present application, there is provided a computer-readable storage medium, and a computer program is stored in the computer-readable storage medium, and the computer program is loaded and executed by a processor to implement the above method.

[0022] The technical solution provided by the embodiment of the present application at least includes the following beneficial effects:

[0023] By acquiring seismic data and logging data, based on the seismic data, a large-offset incidence angle gather and a small-offset incidence angle gather are generated. The logging data and the seismic data are matched through well-seismic calibration and time-depth conversion to obtain an initial low-frequency model. Then, based on the seismic data, the initial low-frequency model, the large-offset incidence angle gather, and the small-offset incidence angle gather, an approximate equation is used for inversion to obtain an inversion result. According to this inversion result, anisotropic parameters are calculated. Then, based on the logging data and the anisotropic parameters, the organic matter content prediction model is trained to obtain a trained organic matter content prediction model, realizing the accurate prediction of the organic matter content in the shale reservoir by using the anisotropic parameters in the seismic data. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a schematic diagram of the implementation environment of the solution provided by an embodiment of the present application;

[0025] Figure 2 is a flowchart of the organic matter content prediction method provided by an embodiment of the present application;

[0026] Figure 3 is a flowchart of the method provided by an embodiment of the present application;

[0027] Figure 4 is a comparison diagram of the anisotropic inversion results provided by an embodiment of the present application;

[0028] Figure 5 is a comparison diagram of the isotropic inversion results provided by an embodiment of the present application;

[0029] Figure 6 is a schematic diagram of the anisotropic parameter estimation results provided by an embodiment of the present application;

[0030] Figure 7 is a schematic diagram of the organic matter content estimation results provided by an embodiment of the present application;

[0031] Figure 8 is a block diagram of the organic matter content prediction device provided by an embodiment of the present application;

[0032] Figure 9 is a block diagram of the computer device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] To make the objectives, technical solutions, and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the accompanying drawings.

[0034] Geophysical exploration, abbreviated as geophysical prospecting, refers to the exploration of geological conditions such as formation lithology and geological structure by studying and observing the changes in various geophysical fields. Since different rock formations in the earth's crust often vary in density, elasticity, conductivity, magnetism, radioactivity, and thermal conductivity, these differences will cause local changes in the corresponding geophysical fields. By measuring the distribution and variation characteristics of these physical fields and analyzing them in combination with known geological data, the purpose of inferring geological properties can be achieved. This method has both exploration and testing functions. Compared with drilling, it has the advantages of light equipment, low cost, high efficiency, and wide working space.

[0035] Seismic inversion technology is a geophysical method used to understand the underground geological structure. By analyzing the way seismic waves propagate and reflect underground, it is possible to determine the properties, distribution, and shape of underground rock formations. Seismic inversion is a process of solving for the spatial structure and physical properties of underground rock formations using seismic data observed on the surface, constrained by known geological laws and well logging data.

[0036] AVO technology is used to study the variation characteristics of seismic reflection amplitude with the distance between the shot point and the receiver, i.e., the offset (or incident angle), to explore the variation of the reflection coefficient response with the offset (or incident angle), and then determine the lithological characteristics and physical property parameters of the overlying and underlying media of the reflection interface.

[0037] A shale reservoir refers to the shale layer in underground rocks, which contains storable and recoverable natural gas and oil, and has the characteristics of "self-generation and self-storage", poor physical properties, extremely low permeability, and strong anisotropy.

[0038] Shale oil and gas sweet spots refer to the formation conditions that can achieve economic feasibility and high output when extracting shale oil and shale gas from shale formations. Shale oil and gas sweet spots include geological sweet spots and engineering sweet spots. In the process of identifying and predicting geological sweet spots, the key is to accurately identify the physical property parameters of underground rocks such as porosity, organic matter content, hydrocarbon-bearing property, and reservoir thickness. Engineering sweet spots are parameters related to fracturing development, such as in-situ stress, fractures, brittleness, and pressure. Organic matter is the main hosting matrix of shale gas and also the material basis for determining shale gas production. Generally, the higher the remaining organic matter content in mature shale, the better the gas-bearing property of the formation system. Predicting the organic matter content of shale gas is of great significance. Conventional seismic prediction of organic matter content relies on high-precision density inversion, and the problem of strong non-uniqueness is particularly prominent when facing strongly anisotropic shale formations in density seismic inversion.

[0039] The Vertical Transverse Isotropy (VTI) property means that in a homogeneous vertically transversely isotropic medium, seismic waves exhibit transverse isotropy when propagating horizontally and anisotropy when propagating in other directions. Seismic responses and acoustic logging responses also have such characteristics. Seismic anisotropy has a huge impact on seismic propagation velocity and amplitude and plays a crucial role in quantitative seismic interpretation of shale oil and gas reservoirs.

[0040] Please refer to Figure 1 , which shows a schematic diagram of the implementation environment of the solution provided by an embodiment of the present application. This solution implementation environment can be implemented as a training and use system for the organic matter content prediction model. This solution implementation environment may include: a model training device 10 and a model use device 20.

[0041] The model training device 10 can be an electronic device such as a mobile phone, a tablet computer, a laptop computer, a desktop computer, a server, etc., or some other electronic device with strong computing power. The model training device 10 is used to train the organic matter content prediction model.

[0042] In the embodiment of the present application, the organic matter content prediction model is a machine learning model obtained through training and is used to predict the organic matter content of underground shale reservoirs. For example, seismic data and logging data are acquired, and after seismic inversion, an inversion result is obtained. According to the inversion result, anisotropy parameters are calculated, and the predicted organic matter content of the shale reservoir is calculated through the organic matter content prediction model. The model training device 10 can train this product in a machine learning manner to enable it to have the ability to predict the organic matter content of shale reservoirs. The specific model training method can refer to the following embodiments.

[0043] The trained organic matter content model can be deployed in the model use device 20 for use. The model use device 20 can be a terminal device such as a mobile phone, a tablet computer, a laptop computer, a desktop computer, etc., or a server. When product quality prediction is required, the model use device 20 can implement the above product quality prediction function through the trained product quality prediction model.

[0044] The model training device 10 and the model use device 20 can be two independently existing devices or the same device.

[0045] The method provided by the embodiments of the present application, the execution subject of each step can be a computer device, which refers to an electronic device with data calculation, processing, and storage functions. The computer device can be a terminal device such as a mobile phone, a tablet computer, a laptop computer, a desktop computer, etc., or a server. Among them, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. For example, the electronic device can be like Figure 1 the model training device 10 in

[0046] Please refer to Figure 2 , which shows a flowchart of an organic matter content prediction method provided by an embodiment of the present application. The method may include at least one of the following steps 210 to 260:

[0047] Step 210, obtain seismic data and logging data. Seismic data is the data collected after seismic waves propagate in underground media, and logging data is the data measured from a drilling well.

[0048] Seismic data is the seismic waves artificially excited on the earth's surface in geophysical exploration. During the propagation of seismic waves from top to bottom, reflection and refraction phenomena will occur due to the inconsistency of the formation media and lithology, and then the seismic waves reflected from the formation are received by a seismic exploration geophone on the earth's surface, and the seismic wave data collected during this process. Seismic data includes recorded data such as the reflected seismic waves, the arrival time and intensity of the seismic waves.

[0049] In some embodiments, the seismic data includes at least one seismic trace. A seismic trace, also called a seismic record trace, is the basic unit for recording the propagation and reflection information of seismic waves in geophysical exploration. Each seismic trace represents the record of seismic waves within a certain period of time, usually including time and seismic wave amplitude information. Seismic traces are the basic units of seismic data and are used to analyze and interpret the properties and structures of underground rock formations.

[0050] Logging data refers to various physical information of rock formations in a drilling well predicted by logging methods, that is, geophysical parameters such as the conductivity, acoustic properties, radioactivity, and electrochemical properties of rock formations are measured along the drilling profile by special downhole instruments. Logging methods include but are not limited to the following methods: geophysical logging, exploration logging, development logging, acoustic logging, etc. Which logging method is specifically used to measure the logging data is determined by relevant technical personnel, and the present application does not make any limitations in this regard.

[0051] Step 220, based on the seismic data, generate a large-offset incident angle gather and a small-offset incident angle gather.

[0052] It should be noted that the "angle of incidence" mentioned in this application refers to the incident angle when seismic waves intersect with the medium boundary underground or on the Earth's surface, that is, the intersection angle between the seismic waves and the interface.

[0053] The angle-of-incidence gather, also known as the seismic angle gather, is used to describe the response of underground rock formations to seismic waves at different angles of incidence, and includes a set of seismic traces, where each trace represents seismic wave data recorded at a different angle of incidence.

[0054] The offset is the horizontal distance between the seismic energy source and the seismic monitoring instrument. Different offsets will cause seismic waves to propagate along different paths and angles, resulting in different reflection and refraction phenomena underground. Smaller offsets usually reflect information about the shallow subsurface, while larger offsets can provide information about the deeper subsurface structure.

[0055] The large-offset angle-of-incidence gather is an angle-of-incidence gather with a large offset, and is used to interpret the properties and distribution of deep underground rock formations.

[0056] The small-offset angle-of-incidence gather is an angle-of-incidence gather with a small offset, and is used to interpret the properties and distribution of shallow underground rock formations.

[0057] In some embodiments, preprocess the seismic data to obtain the preprocessed seismic data; perform a stacked section on the preprocessed seismic data to generate a seismic wave gather; perform angle gather conversion and angle-by-angle stacking on the seismic wave gather to obtain a seismic wave angle gather; divide the seismic wave angle gather based on a first value to generate a large-offset angle-of-incidence gather and a small-offset angle-of-incidence gather, where the first value is used to distinguish the offset corresponding to the large-offset angle-of-incidence gather from the offset corresponding to the small-offset angle-of-incidence gather.

[0058] A seismic section is a commonly used data representation method in the fields of geophysical exploration and seismology, and is used to display the underground geological structure and the propagation of seismic waves. A seismic section is a longitudinal display of seismic data, usually a series of seismic traces arranged together according to time or spatial coordinates.

[0059] Stacked section is a data processing technique in geophysical exploration, which is used to improve the signal-to-noise ratio (SNR) of seismic data and enhance the imaging of underground structures, and is completed by adding the data of multiple seismic traces or through complex stacking methods.

[0060] A seismic wave gather refers to a set of seismic traces or seismic record traces, that is, a seismic wave gather includes at least one seismic trace or seismic record trace.

[0061] In some embodiments, the seismic wave trace gather can be converted into a seismic wave angle gather by a ray theory-based method, or can be converted into a seismic wave angle gather by a wave theory-based method. Of course, it can also be other angle gather conversion methods, which are not limited in this application.

[0062] In some embodiments, angle-domain stacking refers to stacking seismic data within a first incident angle range, and the first incident angle range is used to indicate the incident angle range for angle-domain stacking, which is set by relevant personnel and is not limited in this application.

[0063] In some embodiments, the first value can be a threshold set according to experiments, or can be a threshold for dividing the large-offset incident angle gather and the small-offset incident angle gather set in other ways, which is not limited in this application.

[0064] In some embodiments, the seismic traces in the large-offset incident angle gather are sorted according to the incident angle from large to small or from small to large.

[0065] In some embodiments, the seismic traces in the small-offset incident angle gather are sorted according to the incident angle from large to small or from small to large.

[0066] By the above method, high-quality seismic data is provided for subsequent anisotropic inversion and isotropic inversion, improving the accuracy of the inversion results.

[0067] In some embodiments, the preprocessing includes denoising, filtering, and NMO correction.

[0068] Denoising is a process of filtering out or reducing noise components from seismic data. The methods of denoising include but are not limited to the following: wavelet denoising, band-pass filtering, and equalization filtering.

[0069] Filtering is used to adjust the spectral characteristics of the data, highlighting the required signal components while removing noise or unwanted frequency components.

[0070] NMO correction refers to the process of correcting the static time shift on the seismic record traces to ensure that the data is aligned in time or depth.

[0071] Exemplarily, as Figure 3 shown, which shows a schematic diagram of a seismic wave angle gather provided by an embodiment of this application.

[0072] By the above method, the accuracy and usability of the seismic data are ensured, reducing the impact of poor-quality data on the subsequent inversion results.

[0073] Step 230: Match the logging data and seismic data through well-seismic calibration and time-depth conversion to obtain an initial low-frequency model. The initial low-frequency model is used to describe the wave velocity distribution and density distribution of the underground medium, and includes a velocity distribution model and a density distribution model.

[0074] Well-seismic calibration is the process of using logging data in underground wells to correct the time scale of seismic record traces. Time-depth conversion is the process of converting the arrival time in seismic data into the corresponding depth value.

[0075] In some embodiments, the initial low-frequency model is the starting model used for inverting the underground structure in geophysical exploration, including a velocity distribution model and a density distribution model. Among them, the velocity distribution model describes the P-wave and S-wave velocity distributions of seismic waves at different depths underground, showing the approximate velocity changes when seismic waves propagate in the underground structure. The density distribution model describes the density changes of the underground medium at different depths.

[0076] Through the above method, the initial low-frequency model provides a preliminary estimate for initial modeling and analysis in seismic inversion, providing a starting point for seismic inversion.

[0077] Step 240: Based on the seismic data, the initial low-frequency model, the large-offset incidence angle gather, and the small-offset incidence angle gather, perform inversion using an approximate equation to obtain an inversion result. The inversion result includes an anisotropic inversion result and an isotropic inversion result.

[0078] The approximate equation refers to the basic physical equation in seismic inversion, which is used to simulate and interpret seismic data and estimate parameters of the underground structure, such as velocity, density, lithology, etc. It can include but is not limited to the following equations: Rüger approximate equation, Zhang approximate equation, etc.

[0079] In some embodiments, the approximate equation includes an anisotropic approximate equation and an isotropic approximate equation; based on the initial low-frequency model, generate initial inversion parameters, which include S-wave velocity, P-wave velocity, and medium density; determine the well-side trace seismic data from the seismic data, extract the wavelet from the well-side trace seismic data to obtain a mixed-phase wavelet; based on the large-offset incidence angle gather, the mixed-phase wavelet, the initial inversion parameters, and the anisotropic approximate equation, perform anisotropic inversion to obtain an anisotropic inversion result; based on the small-offset incidence angle gather, the mixed-phase wavelet, the initial inversion parameters, and the isotropic approximate equation, perform isotropic inversion to obtain an isotropic inversion result.

[0080] In some embodiments, the anisotropic approximate equation is the Zhang approximate equation.

[0081] In some embodiments, the isotropic approximate equation is the Rüger approximate equation.

[0082] Wellside seismic data is obtained by placing seismic sensors near the wellhead and measuring the propagation of seismic waves in the vertical direction along the wellbore.

[0083] A mixed-phase wavelet is a wavelet used to simulate the seismic waveform when the seismic source is excited, and has certain amplitude and phase information of the seismic wave.

[0084] In the above manner, starting data is provided for anisotropic inversion and isotropic inversion.

[0085] In some embodiments, based on the initial inversion parameters and the anisotropic approximation equation, an anisotropic forward model is constructed; seismic simulation is performed based on the anisotropic forward model and the mixed-phase wavelet to generate seismic synthetic data; according to the large-offset incident angle gather and the seismic synthetic data, the anisotropic error is determined, and the anisotropic error is used to measure the difference between the large-offset incident angle gather and the seismic synthetic data; when the anisotropic error is greater than the first threshold, the parameters of the anisotropic forward model are adjusted according to the anisotropic error, and the step of performing seismic simulation again based on the anisotropic forward model and the mixed-phase wavelet to generate seismic synthetic data is executed; wherein, the first threshold is used to control the process of training the anisotropic forward model; when the anisotropic error is less than or equal to the first threshold, the training of the anisotropic forward model is stopped, and the trained anisotropic forward model is obtained; according to the parameters of the trained anisotropic forward model, the anisotropic inversion result is obtained.

[0086] In some embodiments, based on the anisotropic approximation equation, a convolutional model is used to establish an anisotropic forward model, and other methods can also be used to establish the anisotropic forward model, which is not limited in this application.

[0087] Seismic synthetic data is seismic data generated by simulation or synthesis methods, and is used to compare with the actually collected seismic data to adjust the parameters of the anisotropic forward model.

[0088] In some embodiments, a regularization method under the Bayesian framework can be used for anisotropic inversion, and inversion methods based on the least squares method, the gradient descent method, etc. can also be used for inversion, which is not limited in this application.

[0089] In some embodiments, the parameters of the anisotropic forward model after the training is completed are determined as the anisotropic inversion result.

[0090] In some embodiments, the anisotropic inversion result may include the intercept term, gradient term, and anisotropic velocity term of the amplitude change with offset, and may also include other parameters, which is not limited in this application.

[0091] The derivation of the inversion process using the regularization method under the Bayesian framework is introduced below. For the detailed derivation process, please refer to the following text.

[0092] In some embodiments, based on the inversion initial parameters and the isotropic approximation equation, an isotropic forward model is constructed; seismic simulation is performed based on the isotropic forward model and the mixed-phase wavelet to generate seismic synthetic data; according to the small-offset incident angle gather and the seismic synthetic data, the isotropic error is determined, and the isotropic error is used to measure the difference between the small-offset incident angle gather and the seismic synthetic data; when the isotropic error is greater than the second threshold, the parameters of the isotropic forward model are adjusted according to the isotropic error, and the step of performing seismic simulation again based on the isotropic forward model and the mixed-phase wavelet to generate seismic synthetic data is executed; wherein, the second threshold is used to control the training process of the isotropic forward model; when the isotropic error is less than or equal to the second threshold, the training of the isotropic forward model is stopped to obtain the trained isotropic forward model; according to the parameters of the trained isotropic forward model, the isotropic inversion result is obtained.

[0093] In some embodiments, based on the isotropic approximation equation, a convolutional model is used to establish an isotropic forward model, or other methods can also be used to establish the isotropic forward model, and the present application does not limit this.

[0094] In some embodiments, the isotropic inversion can be performed using the regularization method under the Bayesian framework, or other inversion methods such as the least squares method and the gradient descent method can also be used for inversion, and the present application does not limit this.

[0095] In some embodiments, the parameters of the anisotropic forward model after the training is completed are determined as the anisotropic inversion result.

[0096] In some embodiments, the anisotropic inversion result may include the intercept term, gradient term, and anisotropic velocity term of the amplitude change with offset, or may also include other parameters, and the present application does not limit this.

[0097] The derivation of the inversion process using the regularization method under the Bayesian framework is introduced below. For the detailed derivation process, please refer to the following text.

[0098] By the above method, accurate anisotropic inversion results and isotropic inversion results are obtained, preparing for the subsequent training of an accurate organic matter prediction model.

[0099] Step 250, based on the anisotropic inversion result and the isotropic inversion result, calculate the anisotropic parameter, and the anisotropic parameter is used to indicate the degree of anisotropy.

[0100] The anisotropic parameter is used to describe the properties of anisotropic rocks and underground structures.

[0101] In some embodiments, anisotropic parameters are calculated based on the gradient term in the anisotropic inversion result and the shear wave modulus of the isotropic inversion result. For example, assuming that when performing anisotropic inversion, the approximate equation used is the Zhang approximate equation; assuming that when performing anisotropic inversion, the approximate equation used is the Rüger approximate equation, that is,

[0102] The Zhang approximate equation is:

[0103]

[0104] The Rüger approximate equation is:

[0105]

[0106] where the AVO intercept term is the gradient term is the anisotropic velocity term the vertical P-wave velocity is the P-wave impedance is the shear wave modulus is the P-wave velocity is the vertical S-wave velocity is the medium density is ρ; Δ represents the difference between the upper and lower interfaces, represents the average value of the P-wave velocities of the upper and lower interfaces, represents the average value of the S-wave velocities of the upper and lower interfaces, is a constant, θ is the incident angle, and the anisotropic parameter σ is:

[0107]

[0108] where represents the predicted value of the shear wave modulus, and the anisotropic parameter can be calculated according to the above formula.

[0109] By the above method, anisotropic parameters are obtained, which can better predict the organic matter content.

[0110] Step 260: Based on the logging data and the anisotropic parameters, train the organic matter content prediction model to obtain the trained organic matter content prediction model; wherein, the trained organic matter content prediction model is used to predict the organic matter content of the shale reservoir.

[0111] The organic matter content prediction model is an artificial intelligence model constructed based on a support vector machine, and is used to predict the organic matter content of the underground shale reservoir according to the anisotropic parameters.

[0112] In some embodiments, the organic matter content in the logging data is determined as the sample label, and the anisotropic parameter is determined as the sample data; the sample data is input into the organic matter content prediction model to obtain the predicted organic matter content; according to the sample label and the predicted organic matter content, the model loss is obtained, and the model loss is used to measure the difference between the sample label and the predicted organic matter content; when the model loss is greater than the second value, the parameters of the organic matter content prediction model are adjusted according to the model loss, and the process starts again from the step of inputting the sample data into the organic matter content prediction model to obtain the predicted organic matter content. The second value is the threshold for stopping the training of the organic matter content prediction model; when the model loss is less than or equal to the second value, the training is stopped to obtain the trained organic matter content prediction model.

[0113] In some embodiments, a linear kernel can be selected as the kernel function of the support vector machine, or a polynomial kernel, a Gaussian kernel, or other kernel functions can be selected. The present application does not make any limitations in this regard.

[0114] In some embodiments, methods such as gradient descent, random search, and grid search can be used to adjust the parameters of the organic matter content prediction model according to the total loss. The present application does not make any limitations in this regard.

[0115] Through the above method, by predicting through the trained organic matter content model, a more accurate organic matter content can be obtained.

[0116] Please refer to Figure 4 , which shows a comparison diagram of anisotropic inversion results provided by an embodiment of the present application. From left to right, the inversion parameters are the AVO intercept term, the gradient term, and the anisotropic velocity, respectively. The black dashed line is the model data, the black solid line is the actual logging data, and the gray solid line is the inversion result.

[0117] Please refer to Figure 5 , which shows a comparison diagram of isotropic inversion results provided by an embodiment of the present application. From left to right, the inversion parameters are the longitudinal wave impedance, the shear wave modulus, and the longitudinal wave velocity, respectively. The black dashed line is the model data, the black solid line is the actual logging data, and the gray solid line is the inversion result.

[0118] Please refer to Figure 6 , which shows a schematic diagram of the estimated anisotropic parameter results provided by an embodiment of the present application. The black solid line is the calculation result of the actual logging data, and the gray solid line is the inversion result.

[0119] Please refer to Figure 7 , which shows a schematic diagram of the estimated organic matter content results provided by an embodiment of the present application. The black solid line is the calculation result of the actual logging data, and the gray solid line is the inversion result.

[0120] In summary, for the technical solution provided in the embodiments of the present application, by acquiring seismic data and logging data, based on the seismic data, a large-offset incidence angle gather and a small-offset incidence angle gather are generated. The logging data and the seismic data are matched through well-seismic calibration and time-depth conversion to obtain an initial low-frequency model. Then, based on the seismic data, the initial low-frequency model, the large-offset incidence angle gather, and the small-offset incidence angle gather, an inversion is performed using an approximate equation to obtain an inversion result. According to this inversion result, anisotropic parameters are calculated. Then, based on the logging data and the anisotropic parameters, a prediction model for organic matter content is trained to obtain a trained prediction model for organic matter content, realizing the accurate prediction of the organic matter content in shale reservoirs using the anisotropic parameters in seismic data.

[0121] The derivation process of Bayesian inversion is briefly introduced below.

[0122] The derivation process of Bayesian inversion for anisotropy is as follows:

[0123] (1) Establish an anisotropic approximate equation for the PP reflection coefficient based on the Zhang approximate equation:

[0124]

[0125] Among them, the AVO intercept term is The gradient term is The anisotropic velocity term The vertical shear wave velocity is The medium density is ρ; Δ represents the difference between the upper and lower interfaces, represents the average value of the P-wave velocities of the upper and lower interfaces, represents the average value of the S-wave velocities of the upper and lower interfaces, is a constant, the incident angle is θ, and the anisotropic parameter is

[0126] (2) Perform inversion using the regularization method under the Bayesian framework. The advantage of Bayesian theory is that for data that cannot be measured repeatedly in large quantities, the inference process can be constrained by prior information, making the posterior distribution reliable.

[0127] The observed data d collected obs cannot have a certain repeatability and contains a certain error, which makes it inaccurate to estimate the subsurface model m under the condition of d obs . The same Bayesian equation can be established:

[0128]

[0129] Among them, p(m|d obs ) is under d obsThe probability distribution of the subsurface model m under certain conditions, where p(m) is the prior distribution satisfied by the assumed subsurface medium, and different selections of the prior distribution are often based on human choices. p(d obs |m) is the probability distribution of obtaining the observed data under this model, that is, the likelihood function p(d obs ) is the distribution of the observed data, which should not depend on the model and is often related to the noise distribution.

[0130] Assume that the seismic data noise follows a Gaussian distribution and is independent (correlation coefficient is 0), and the noise n is distributed as:

[0131]

[0132] where p 0 is a constant, is the inverse matrix of the covariance matrix of the noise, and p(n) is the likelihood function.

[0133] The general relationship between seismic data and the subsurface medium model in AVO linear inversion:

[0134] d obs = G·m + n

[0135] where G is the linear forward operator. n = d obs - G·m follows a Gaussian distribution and conforms to the linear principle, and the distribution law of the likelihood function:

[0136]

[0137] It is usually considered that the marginal distribution probability of the observed data d obs is a constant. Putting p(m) into the exponential term, assume p(m) = e -μQ(m) , where Q(m) represents the regularization constraint term obtained from the prior model, then there is:

[0138]

[0139] According to the MAP principle, the objective function J of the inversion can be established:

[0140]

[0141] Taking the derivative of this function gives:

[0142]

[0143] In this way, the Bayesian inversion solution is established, where Q(m) is a term closely related to the selection of the prior distribution. It can also be seen from the above formula that the essence of Bayesian inversion is also to constrain both the error and the model simultaneously. The term is the weight of the error, the μ term is the weight of the model, and d is the obtained seismic data.

[0144] (3) According to the anisotropic approximation equation in (1), a forward model is established using the convolution model. Since the forward model is a function that varies with time (i.e., in the depth direction), assuming that this variation is not very large within a sampling interval and can be considered a continuous function of time, the assumed gradient change can be regarded as the derivative of the natural logarithm, and the difference can be regarded as the derivative of the time function.

[0145] Assume that the number of angle samples is M, and the angle is denoted as θ 1 ,…,θ M , and the number of time samples is N. Written in matrix form, where M and N are positive integers, it can be expressed as:

[0146] K·D·F=R

[0147] By given the coefficient matrix K, the difference matrix D, the data logarithm matrix F, and the reflection coefficient matrix R, and then multiplying by the wavelet convolution matrix S. It can be written in the forward form, thus establishing the same forward formula as discussed above, where d is the obtained seismic data:

[0148] S·K·D·F=d

[0149] The establishment process of each matrix is as follows. Coefficient matrix K:

[0150]

[0151] Among them,

[0152]

[0153]

[0154] A(θ k ,t 1 ) represents the coefficient of parameter A at the first time sampling point when the angle is θ k .

[0155] A(θ k ,t 1 ), B(θ k ,t 1 ), C(θ k ,t 1 ) are the same by analogy.

[0156] Difference equation D:

[0157]

[0158] Wavelet convolution matrix S:

[0159]

[0160] Among them, S(θ k ) represents the wavelet convolution matrix at the angle of θ k , where k = 1, 2, …, M.

[0161] Model data matrix F:

[0162] F = [ln A(t 1 )…ln A(t N ), ln B(t 1 )…ln B(t N ), ln C(t 1 )…ln C(t N )] T

[0163] It can be deduced that G in the inversion objective function is:

[0164] G = S·K·D

[0165] Substituting G back into the inversion objective function, the inversion result can be obtained.

[0166] The derivation process of isotropic Bayesian inversion is as follows:

[0167] (1) Establish an isotropic approximation equation of the PP reflection coefficient based on the Rüger approximation equation:

[0168]

[0169] Among them, the vertical P-wave velocity is the P-wave impedance is the shear modulus is the P-wave velocity is the vertical S-wave velocity is the medium density is ρ; Δ represents the difference between the upper and lower interfaces, represents the average value of the P-wave velocities of the upper and lower interfaces, represents the average value of the S-wave velocities of the upper and lower interfaces, is a constant, and θ is the incident angle.

[0170] The following steps are the same as (2) and (3) in the derivation process of anisotropic Bayesian inversion, so they will not be elaborated here.

[0171] The following is the device embodiment of the present application, which can be used to execute the method embodiment of the present application. For the details not disclosed in the device embodiment of the present application, please refer to the method embodiment of the present application.

[0172] Please refer to Figure 8, which shows a block diagram of an organic matter content prediction device provided by an embodiment of the present application. The device has the functions of implementing the above method examples, and the functions can be implemented by hardware or by hardware executing corresponding software. The device can be the computer device introduced above or can be set in the computer device. As Figure 8 shown, the device 800 may include an acquisition module 810, a generation module 820, a matching module 830, an inversion module 840, a calculation module 850, and a training module 860.

[0173] The acquisition module 810 is configured to acquire seismic data and logging data. The seismic data is data collected after seismic waves propagate in underground media, and the logging data is data measured from a drilling well.

[0174] The generation module 820 is configured to generate a large-offset incident angle gather and a small-offset incident angle gather based on the seismic data.

[0175] The matching module 830 is configured to match the logging data and the seismic data through well-seismic calibration and time-depth conversion to obtain the initial low-frequency model. The initial low-frequency model is used to describe the wave velocity distribution and density distribution of the underground media, and the initial low-frequency model includes a velocity distribution model and a density distribution model.

[0176] The inversion module 840 is configured to perform inversion on the seismic data, the initial low-frequency model, the large-offset incident angle gather, and the small-offset incident angle gather by using an approximate equation to obtain an inversion result. The inversion result includes an anisotropic inversion result and an isotropic inversion result.

[0177] The calculation module 850 is configured to calculate anisotropic parameters based on the anisotropic inversion result and the isotropic inversion result. The anisotropic parameters are used to indicate the degree of anisotropy.

[0178] The training module 860 is configured to train an organic matter content prediction model based on the logging data and the anisotropic parameters to obtain a trained organic matter content prediction model. The trained organic matter content prediction model is used to predict the organic matter content of a shale reservoir.

[0179] In some embodiments, the approximate equation includes an anisotropic approximate equation and an isotropic approximate equation. The inversion module includes an initial sub-module, a wavelet sub-module, an anisotropic inversion sub-module, and an isotropic inversion sub-module ( Figure 8 not shown in the figure).

[0180] The initial sub-module is configured to generate inversion initial parameters based on the initial low-frequency model. The inversion initial parameters include shear wave velocity, longitudinal wave velocity, and medium density.

[0181] A wavelet sub-module, configured to determine near-wellbore seismic data from the seismic data, perform wavelet extraction on the near-wellbore seismic data, and obtain a mixed-phase wavelet.

[0182] An anisotropic inversion sub-module, configured to perform anisotropic inversion based on the large-offset incidence angle gather, the mixed-phase wavelet, the inversion initial parameters, and the anisotropic approximation equation, and obtain the anisotropic inversion result.

[0183] An isotropic inversion sub-module, configured to perform isotropic inversion based on the small-offset incidence angle gather, the mixed-phase wavelet, the inversion initial parameters, and the isotropic approximation equation, and obtain the isotropic inversion result.

[0184] In some embodiments, the anisotropic inversion sub-module is configured to: construct an anisotropic forward model based on the inversion initial parameters and the anisotropic approximation equation; perform seismic simulation based on the anisotropic forward model and the mixed-phase wavelet to generate seismic synthetic data; determine an anisotropic error according to the large-offset incidence angle gather and the seismic synthetic data, where the anisotropic error is used to measure the difference between the large-offset incidence angle gather and the seismic synthetic data; when the anisotropic error is greater than a first threshold, adjust the parameters of the anisotropic forward model according to the anisotropic error, and start executing the step of performing seismic simulation based on the anisotropic forward model and the mixed-phase wavelet to generate seismic synthetic data again; where the first threshold is used to control the training process of the anisotropic forward model; when the anisotropic error is less than or equal to the first threshold, stop the training of the anisotropic forward model to obtain a trained anisotropic forward model; and obtain the anisotropic inversion result according to the parameters of the trained anisotropic forward model.

[0185] In some embodiments, the isotropic inversion sub-module is configured to: construct an isotropic forward model based on the inversion initial parameters and the isotropic approximation equation; perform seismic simulation based on the isotropic forward model and the mixed-phase wavelet to generate seismic synthetic data; determine an isotropic error according to the small-offset incident angle gather and the seismic synthetic data, where the isotropic error is used to measure the difference between the small-offset incident angle gather and the seismic synthetic data; when the isotropic error is greater than a second threshold, adjust the parameters of the isotropic forward model according to the isotropic error, and start executing again from the step of performing seismic simulation based on the isotropic forward model and the mixed-phase wavelet to generate seismic synthetic data; wherein, the second threshold is used to control the training process of the isotropic forward model; when the isotropic error is less than or equal to the second threshold, stop the training of the isotropic forward model to obtain the trained isotropic forward model; obtain the isotropic inversion result according to the parameters of the trained isotropic forward model.

[0186] In some embodiments, the generating module 820 is configured to: preprocess the seismic data to obtain preprocessed seismic data; perform stacking profiles on the preprocessed seismic data to generate seismic wave gathers; perform angle gather conversion and angle-by-angle stacking on the seismic wave gathers to obtain seismic wave angle gathers; divide the seismic wave angle gathers based on a first value to generate the large-offset incident angle gather and the small-offset incident angle gather, where the first value is used to distinguish the offset corresponding to the large-offset incident angle gather and the offset corresponding to the small-offset incident angle gather.

[0187] In some embodiments, the preprocessing includes denoising, filtering, and NMO correction.

[0188] In some embodiments, the training module 860 is configured to: determine the organic matter content in the logging data as a sample label, and determine the anisotropic parameter as sample data; input the sample data into the organic matter content prediction model to obtain a predicted organic matter content; obtain a model loss according to the sample label and the predicted organic matter content, where the model loss is used to measure the difference between the sample label and the predicted organic matter content; when the model loss is greater than a second value, adjust the parameters of the organic matter content prediction model according to the model loss, and start executing again from the step of inputting the sample data into the organic matter content prediction model to obtain a predicted organic matter content, where the second value is the threshold for stopping the training of the organic matter content prediction model; when the model loss is less than or equal to the second value, stop the training to obtain the trained organic matter content prediction model.

[0189] In summary, for the technical solution provided in the embodiment of the present application, by acquiring seismic data and logging data, based on the seismic data, a large-offset incidence angle gather and a small-offset incidence angle gather are generated. The logging data and the seismic data are matched through well-seismic calibration and time-depth conversion to obtain an initial low-frequency model. Then, based on the seismic data, the initial low-frequency model, the large-offset incidence angle gather, and the small-offset incidence angle gather, an inversion is performed using an approximate equation to obtain an inversion result. According to this inversion result, anisotropic parameters are calculated. Then, based on the logging data and the anisotropic parameters, the organic matter content prediction model is trained to obtain a trained organic matter content prediction model, realizing the accurate prediction of the organic matter content in the shale reservoir using the anisotropic parameters in the seismic data.

[0190] It should be noted that for the device provided in the above embodiment, when implementing its functions, only the division of the above functional modules is used as an example. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device provided in the above embodiment and the method embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0191] Please refer to Figure 9 , which shows a structural block diagram of a computer device 900 provided in an embodiment of the present application. The computer device 900 may be Figure 1 the computer device 10 in the shown implementation environment, and is used to implement the organic matter content prediction method provided in the above embodiment. Specifically:

[0192] Generally, the computer device 900 includes: a processor 910 and a memory 920.

[0193] The processor 910 may include one or more processing cores, such as a quad-core processor, an octa-core processor, etc. The processor 910 may be implemented in at least one of the following hardware forms: Digital Signal Processing (DSP), Field Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 910 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the wake state, also known as the Central Processing Unit (CPU); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 910 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 910 may further include an AI processor, which is used to process computational operations related to machine learning.

[0194] The memory 920 may include one or more computer-readable storage media, and the computer-readable storage media may be non-transitory. The memory 920 may further include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash storage devices. In some embodiments, the non-transitory computer-readable storage media in the memory 920 are used to store computer programs, and the computer programs are configured to be executed by one or more processors to implement the above-mentioned organic matter content prediction method.

[0195] Those skilled in the art can understand that Figure 9 the structure shown in does not constitute a limitation on the computer device 900, and it may include more or fewer components than shown in the figure, or combine certain components, or adopt a different component layout.

[0196] In an exemplary embodiment, a computer-readable storage medium is further provided. A computer program is stored in the storage medium. When the computer program is executed by a processor, the above-mentioned organic matter content prediction method is implemented. Optionally, the computer-readable storage medium may include: Read-Only Memory (ROM for short), Random Access Memory (RAM for short), Solid State Drives (SSD for short), or optical discs, etc. Among them, the random access memory may include Resistance Random Access Memory (ReRAM for short) and Dynamic Random Access Memory (DRAM for short).

[0197] In an exemplary embodiment, a computer program product is further provided. The computer program product includes a computer program, and the computer program is stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium, and the processor executes the computer program, so that the computer device executes the above-mentioned organic matter content prediction method.

[0198] As used herein, "a plurality of" means two or more. "And / or" describes the association relationship of associated objects and indicates that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally represents an "or" relationship between the associated objects before and after.

[0199] As used herein, "greater than or equal to" may mean greater than or equal to or greater than, and "less than or equal to" may mean less than or equal to or less than.

[0200] In addition, the step numbers described herein only exemplarily show a possible execution sequence among the steps. In some other embodiments, the above steps may not be executed in the order of the numbers. For example, two steps with different numbers may be executed simultaneously, or two steps with different numbers may be executed in the reverse order of the illustration. The embodiments of the present application do not limit this.

[0201] The above are only exemplary embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting organic matter content, characterized in that: The method comprises: Acquire seismic data and well logging data, wherein the seismic data is data collected after seismic waves propagate in underground media, and the well logging data is data measured from drilled wells; Based on the seismic data, generating a large offset incident angle gather and a small offset incident angle gather; Matching the logging data and the seismic data through well-seismic calibration and time-depth conversion to obtain the initial low-frequency model, which is used to describe the wave velocity distribution and density distribution of the underground medium. The initial low-frequency model includes a velocity distribution model and a density distribution model; Based on the seismic data, the initial low-frequency model, the large offset incident angle gathers and the small offset incident angle gathers, an approximate equation is used to perform inversion to obtain an inversion result, wherein the inversion result includes an anisotropic inversion result and an isotropic inversion result; Based on the anisotropic inversion result and the isotropic inversion result, an anisotropic parameter is calculated, and the anisotropic parameter is used to indicate the degree of anisotropy; Based on the logging data and the anisotropy parameter, the organic matter content prediction model is trained to obtain a trained organic matter content prediction model; wherein the trained organic matter content prediction model is used to predict the organic matter content of the shale reservoir.

2. The method according to claim 1, characterized in that The approximate equation includes an anisotropic approximate equation and an isotropic approximate equation; The inversion is performed based on the seismic data, the initial low-frequency model, the large offset incident angle gathers and the small offset incident angle gathers using an approximate equation to obtain an inversion result, including: Based on the initial low-frequency model, generating initial inversion parameters, the initial inversion parameters including shear wave velocity, longitudinal wave velocity and medium density; Determine the well bypass seismic data from the seismic data, extract wavelets from the well bypass seismic data, and obtain mixed phase wavelets; Based on the large offset incident angle gather, the mixed phase wavelet, the inversion initial parameters and the anisotropic approximate equation, anisotropic inversion is performed to obtain the anisotropic inversion result; Based on the small offset incident angle gather, the mixed phase wavelet, the inversion initial parameters and the isotropic approximate equation, isotropic inversion is performed to obtain the isotropic inversion result.

3. The method according to claim 2, characterized in that The anisotropic inversion is performed based on the large offset incident angle gather, the mixed phase wavelet, the inversion initial parameter and the anisotropic approximate equation to obtain the anisotropic inversion result, including: Based on the inversion initial parameters and the anisotropic approximate equation, constructing an anisotropic forward model; Performing seismic simulation based on the anisotropic forward model and the mixed phase wavelet to generate seismic synthetic data; Determining an anisotropic error according to the large offset incident angle gather and the seismic synthetic data, wherein the anisotropic error is used to measure the difference between the large offset incident angle gather and the seismic synthetic data; When the anisotropic error is greater than a first threshold, adjusting the parameters of the anisotropic forward model according to the anisotropic error, and starting again from the step of performing seismic simulation based on the anisotropic forward model and the mixed phase wavelet to generate seismic synthetic data; wherein the first threshold is used to control the process of training the anisotropic forward model; When the anisotropic error is less than or equal to the first threshold, stopping the anisotropic forward model training to obtain a trained anisotropic forward model; The anisotropic inversion result is obtained according to the parameters of the trained anisotropic forward model.

4. The method according to claim 2, characterized in that: The isotropic inversion is performed based on the small offset incident angle gather, the mixed phase wavelet, the inversion initial parameters and the isotropic approximate equation to obtain the isotropic inversion result, including: Based on the inversion initial parameters and the isotropic approximate equation, constructing an isotropic forward model; Performing seismic simulation based on the isotropic forward model and the mixed phase wavelet to generate seismic synthetic data; Determine an isotropic error according to the small offset incident angle gather and the seismic synthetic data, wherein the isotropic error is used to measure the difference between the small offset incident angle gather and the seismic synthetic data; When the isotropic error is greater than a second threshold, adjusting the parameters of the isotropic forward model according to the isotropic error, and starting again from the step of performing seismic simulation based on the isotropic forward model and the mixed phase wavelet to generate seismic synthetic data; wherein the second threshold is used to control the process of training the isotropic forward model; When the isotropic error is less than or equal to the second threshold, stopping the isotropic forward model training to obtain a trained isotropic forward model; The isotropic inversion result is obtained according to the parameters of the trained isotropic forward model.

5. The method according to claim 1, characterized in that The generating of large offset incident angle gathers and small offset incident angle gathers based on the seismic data comprises: Preprocessing the seismic data to obtain preprocessed seismic data; Stacking sections of the preprocessed seismic data to generate seismic wave gathers; Performing angle gather conversion and angle stacking on the seismic wave gathers to obtain seismic wave angle gathers; Based on a first value, the seismic wave angle gather is divided to generate the large offset incident angle gather and the small offset incident angle gather, wherein the first value is used to distinguish the offset corresponding to the large offset incident angle gather and the offset corresponding to the small offset incident angle gather.

6. The method according to claim 5, characterized in that The preprocessing includes denoising, filtering and dynamic correction.

7. The method according to claim 1, characterized in that The organic matter content prediction model is trained based on the well logging data and the anisotropy parameter to obtain the trained organic matter content prediction model, including: Determine the organic matter content in the well logging data as a sample label, and determine the anisotropy parameter as sample data; Inputting the sample data into the organic matter content prediction model to obtain predicted organic matter content; According to the sample label and the predicted organic matter content, a model loss is obtained, wherein the model loss is used to measure the difference between the sample label and the predicted organic matter content; When the model loss is greater than a second value, adjusting the parameters of the organic matter content prediction model according to the model loss, and starting again from the step of inputting the sample data into the organic matter content prediction model to obtain the predicted organic matter content, wherein the second value is a threshold for stopping training of the organic matter content prediction model; When the model loss is less than or equal to the second value, the training is stopped to obtain the trained organic matter content prediction model.

8. An organic matter content prediction device, characterized in that: The device comprises: An acquisition module, used to acquire seismic data and well logging data, wherein the seismic data is data collected after seismic waves propagate in underground media, and the well logging data is data measured from drilling wells; A generating module, used for generating a large offset incident angle gather and a small offset incident angle gather based on the seismic data; A matching module, used for matching the well logging data and the seismic data through well-seismic calibration and time-depth conversion to obtain the initial low-frequency model, wherein the initial low-frequency model is used to describe the wave velocity distribution and density distribution of the underground medium, and the initial low-frequency model includes a velocity distribution model and a density distribution model; An inversion module is used for performing inversion on the seismic data, the initial low-frequency model, the large offset incident angle gather and the small offset incident angle gather using an approximate equation to obtain an inversion result, wherein the inversion result includes an anisotropic inversion result and an isotropic inversion result; A calculation module, used for calculating anisotropic parameters based on the anisotropic inversion result and the isotropic inversion result, wherein the anisotropic parameters are used to indicate the degree of anisotropy; A training module is used to train an organic matter content prediction model based on the logging data and the anisotropy parameter to obtain a trained organic matter content prediction model; wherein the trained organic matter content prediction model is used to predict the organic matter content of a shale reservoir.

9. A computer device, characterized in that: The computer device comprises a processor and a memory, wherein a computer program is stored in the memory, and the processor executes the computer program to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and the computer program is used to be executed by a processor to implement the method according to any one of claims 1 to 7.

Images