A reservoir lithology prediction method, device, electronic device and storage medium
Through remote sensing satellite data and convolutional neural network technology, combined with geological statistics and seismic convolution forward performance, the problem of energy consistency of seismic amplitude in the loess plateau area is solved, the accuracy and quantitativeness of reservoir lithologic prediction are improved, and the oil and gas are satisfied Requirements for geological research.
Patent Information
- Application Number
- CN202110138106.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-02-01
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2041-02-01
AI Technical Summary
In the loess plateau area, the problem of energy consistency of seismic amplitude leads to poor lateral prediction and insufficient quality controllability of existing reservoir lithologic prediction technologies, which cannot meet the needs of oil and gas geological research.
Combining remote sensing satellite data, convolutional neural network method and geological statistical modeling technology, reservoir and elastic data are obtained through geological statistical simulation, and seismic convolution forward evolution method is used to convert it into simulated seismic data, and the actual measured data of the well point is as Training objectives, training neural network models for reservoir lithology prediction.
It improves the lithology identification accuracy, overcomes the shortcomings of a single measured seismic data, significantly improves the reservoir lithology prediction accuracy and quantitativeity under the surface conditions of the loess plateau, and meets the requirements of oil and gas geological research.
Smart Images

Figure CN114841049B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of geophysical exploration, and relates to the field of lithology prediction in seismic reservoirs. More specifically, it relates to a reservoir lithology prediction method, prediction device, electronic equipment and storage medium for the problem of loess plateau surface seismic amplitude energy consistency. Background Art
[0002] Due to the lateral changes in the distribution of loess layers on the surface of the Loess Plateau, there are significant differences in the artificial seismic source excitation environment, shallow seismic wave absorption and attenuation, and detector receiving conditions at different plane positions, resulting in a serious problem of lateral energy consistency in seismic data.
[0003] At present, shallow fine velocity modeling and attenuation compensation are used to solve this problem and perform spatial correction of seismic energy. The main technologies include velocity analysis and attenuation parameter estimation based on tomography, near-surface velocity modeling based on refraction wave method, and attenuation depth migration (Q-PSDM) that iteratively corrects the absorption attenuation coefficient during imaging. Although there are differences in the technical means of implementation, they are all limited to the field of seismic data signal processing. For example, the velocity analysis and attenuation parameters based on tomography technology have high requirements for the seismic observation system. The illumination intensity and acquisition density have seriously restricted the application effect of this method in the actual loess tableland area. The near-surface modeling technology based on refraction waves has problems such as high acquisition cost and low lateral density. It is difficult to finely depict the lateral drastic changes of the loess layer in the loess tableland area. The attenuation depth migration technology (Q-PSDM) that iteratively corrects the absorption attenuation coefficient during the imaging process is mainly used in marine seismic exploration because the technology itself has high requirements for the seismic observation system and high acquisition costs. At the same time, there is a serious incompatibility problem between the pre-stack time migration technology system commonly used in the loess tableland area and the technical requirements of the high-cost pre-stack depth migration technology that must be performed first.
[0004] For the reservoir lithology prediction problem based on seismic amplitude interpretation, which is downstream of seismic data processing, the commonly used technologies have prominent problems such as long technical links, poor lateral predictability, and insufficient quality controllability, and cannot meet the requirements of oil and gas geological research under the surface conditions of the Loess Plateau.
[0005] Therefore, there is an urgent need in this field for a reservoir lithology prediction method for the problem of surface seismic amplitude energy consistency in the loess plateau. Summary of the invention
[0006] The present invention is the first to use remote sensing satellite surface data, convolutional neural network methods and geostatistical modeling technology to predict reservoir lithology under the condition that there is a consistency problem in the amplitude energy of loess plateau surface earthquakes, thereby overcoming the shortcomings of single measured seismic data and improving the accuracy of lithology identification.
[0007] According to one aspect of the present invention, a reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem is provided, comprising:
[0008] Through geostatistical simulation, reservoir and elastic data are obtained;
[0009] Using a seismic convolution forward modeling method, the elastic data is converted into simulated seismic data;
[0010] The neural network model is trained using the measured seismic data, satellite remote sensing data, and simulated seismic data at the well point as input and the measured lithology at the well point as the training target;
[0011] The neural network model is used to predict reservoir lithology.
[0012] Furthermore, the geostatistical simulation includes: by giving the lithology ratio of each lithology within the formation, the regular distribution of elastic parameters and the spatial variation function of lithology and elasticity, respectively using sequential indicator simulation and sequential Gaussian simulation technology to achieve spatial simulation of lithology and elasticity, and obtain multiple sets of lithology parameter models and elasticity parameter models.
[0013] Further, converting the elastic data into simulated seismic data comprises:
[0014] The elastic data is converted into a reflection coefficient that varies with offset using the Aki-Richard equation;
[0015] The reflection coefficients are converted to seismic reflections using convolution equations and seismic wavelets.
[0016] Further, the Aki-Richard equation is as follows:
[0017]
[0018] Where R is the reflection coefficient, θ = (θ1 + θ2) / 2 is the incident angle, θ1 and θ2 are the incident angle and transmission angle respectively, and p is the ray parameter. Δρ=ρ2-ρ1, ρ=(ρ2+ρ1) / 2, ρ1 is the longitudinal wave velocity, shear wave velocity and density above the interface, ρ2 is the longitudinal wave velocity, shear wave velocity and density below the interface.
[0019] Furthermore, the convolution equation is as follows:
[0020] S(t, θ)=R(t, θ)*W(t, θ) (2)
[0021] Where t is the earthquake travel time, W(t, θ) is the earthquake wavelet, S(t, θ) is the synthetic seismic record, and * is the convolution operator.
[0022] Furthermore, the objective function of the neural network model is as follows:
[0023]
[0024] Among them, W and b are weight coefficients and bias terms respectively, x and y are input data and training targets, P is the dimension of the data, and T is the matrix transpose operator.
[0025] Furthermore, the Adam method is used to achieve the convergence of the objective function.
[0026] According to another aspect of the present invention, a reservoir lithology prediction device for the problem of surface seismic amplitude energy consistency in loess plateau is provided, comprising:
[0027] The simulation module obtains reservoir and elastic data through geostatistical simulation;
[0028] A forward modeling module, which converts the elastic data into simulated seismic data using a seismic convolution forward modeling method;
[0029] The training module uses the measured seismic data, satellite remote sensing data, and simulated seismic data at the well point as inputs and the measured lithology at the well point as training targets to train the neural network model;
[0030] The prediction module uses the neural network model to predict reservoir lithology.
[0031] According to another aspect of the present invention, there is provided an electronic device, the electronic device comprising:
[0032] A memory storing executable instructions;
[0033] A processor runs the executable instructions in the memory to implement the reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem.
[0034] According to another aspect of the present invention, there is provided a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem is implemented.
[0035] The solution of the present invention uses remote sensing satellite data with high lateral density and intuitively reflecting the thickness of the loess layer as input data to reflect the influence of the loess tableland surface on the spatial variation of seismic amplitude energy, establishes the correlation between the abnormal energy distribution in the actual seismic data and the loess tableland surface, and describes the noise energy of non-geological factors in the seismic data.
[0036] The solution of the present invention obtains the lateral changes of stratum elastic parameters caused by geological factors by utilizing geostatistical simulation technology, and converts the geological-elastic parameters into seismic amplitude data in combination with seismic convolution forward modeling technology to present the geological trend of lateral earthquake energy changes.
[0037] The solution of the present invention utilizes a convolutional neural network method, takes measured seismic data at well points, satellite remote sensing data, and geostatistical simulated seismic data as input, takes measured lithology at well points as training targets, trains a neural network model, and ultimately achieves lithology prediction between wells. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] The above and other objects, features and advantages of the present disclosure will become more apparent through a more detailed description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, wherein like reference numerals generally represent like components in the exemplary embodiments of the present disclosure.
[0039] Figure 1 The present invention is a flow chart of a reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem.
[0040] Figure 2 1 is an example of probability distribution of elastic parameters of different lithologies according to an embodiment of the present invention.
[0041] Figure 3 It is a forward seismic profile diagram according to an embodiment of the present invention.
[0042] Figure 4 The lithology comparison between the learning target (bottom) and the training result (top) of 22 well points according to an embodiment of the present invention.
[0043] Figure 5 The figure shows how the objective function error of the sample well convolutional neural network model changes with the number of iterations according to an embodiment of the present invention.
[0044] Figure 6 (a) is the lateral distribution diagram of seismic amplitude energy, (b) is the satellite remote sensing data diagram, (c) is the predicted sandstone thickness trend diagram, and (d) is the geological sedimentation trend recognition diagram. DETAILED DESCRIPTION
[0045] The preferred embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.
[0046] At present, the surface seismic amplitude energy consistency processing technology of the Loess Plateau, which uses seismic data signal processing as a means, has a low improvement effect due to the limitations of its own application conditions. The papers consulted are mainly based on structural imaging and qualitative interpretation of seismic amplitude, which cannot meet the technical requirements of quantitative seismic amplitude energy.
[0047] The present invention is the first to use remote sensing satellite surface data, convolutional neural network methods and geostatistical modeling technology to predict reservoir lithology under the condition that there is a consistency problem in the amplitude energy of loess plateau surface earthquakes. By adding noise components (satellite remote sensing data) reflecting non-geological factors and geostatistical seismic simulation data reflecting geological factors to the input data, the insufficient use of actual seismic data containing both non-geological (loess plateau surface) and geological factors is overcome, and the powerful sample learning and classification capabilities of convolutional neural networks are used to improve the accuracy of lithology prediction.
[0048] like Figure 1 As shown, the present invention provides a reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem, comprising:
[0049] Through geostatistical simulation, reservoir and elastic data are obtained;
[0050] Using a seismic convolution forward modeling method, the elastic data is converted into simulated seismic data;
[0051] The neural network model is trained using the measured seismic data, satellite remote sensing data, and simulated seismic data at the well point as input and the measured lithology at the well point as the training target;
[0052] The neural network model is used to predict reservoir lithology.
[0053] Geostatistical simulation, that is, by giving the lithology ratio of each lithology in the formation, the regular distribution of elastic parameters, and the spatial variogram of lithology and elasticity, sequential indicator simulation and sequential Gaussian simulation technology are used to achieve spatial simulation of lithology and elasticity, and obtain multiple sets of lithology and elasticity parameter models. In actual work, it is necessary to use the lithology and elasticity data measured at the well point to estimate the geostatistical parameters (lithology ratio, elastic parameter probability distribution, variogram type and range).
[0054] Further, converting the elastic data into simulated seismic data comprises:
[0055] The elastic data is converted into a reflection coefficient that varies with offset using the Aki-Richard equation;
[0056] The reflection coefficients are converted to seismic reflections using convolution equations and seismic wavelets.
[0057] Further, the Aki-Richard equation is as follows:
[0058]
[0059] Where R is the reflection coefficient, θ = (θ1 + θ2) / 2 is the incident angle, θ1 and θ2 are the incident angle and transmission angle respectively, and p is the ray parameter. Δρ=ρ2-ρ1, ρ=(ρ2+ρ1) / 2, ρ1 is the longitudinal wave velocity, shear wave velocity and density above the interface, ρ2 is the longitudinal wave velocity, shear wave velocity and density below the interface.
[0060] Furthermore, the convolution equation is as follows:
[0061] S(t, θ)=R(t, θ)*W(t, θ) (2)
[0062] Where t is the earthquake travel time, W(t, θ) is the earthquake wavelet, S(t, θ) is the synthetic seismic record, and * is the convolution operator.
[0063] Furthermore, the objective function of the neural network model is as follows:
[0064]
[0065] Among them, W and b are weight coefficients and bias terms respectively, x and y are input data and training targets, P is the dimension of the data, and T is the matrix transpose operator.
[0066] Furthermore, the Adam method is used to achieve the convergence of the objective function.
[0067] The present invention effectively combines remote sensing data, geological modeling-seismic forward modeling data and measured data by using machine learning methods, and implements lithology prediction technology through MATLAB programming, providing favorable guarantees for lithology identification on the surface of the loess plateau. The effect of the invention is proved by actual data, and good results have been achieved in the application of reservoir lithology prediction under loess plateau surface conditions.
[0068] Table 1 is a comparison of the data prediction results provided by the method of the present invention and the seismic signal processing method.
[0069] Table 1 Comparison of compliance rates
[0070] Seismic processing method Seismic data prediction agreement rate The method of the present invention predicts the compliance rate Coal seam (3 meters above) 57% 90% Lithology (above 8 meters) 62% 85% Sandstone (over 8 meters) 64% 87%
[0071] This invention uses the convolutional neural network method for the first time, combining satellite remote sensing data reflecting the lateral changes in the thickness of the loess plateau, forward modeling data reflecting the geological change trend obtained based on geostatistical simulation and seismic forward modeling, and actual collected seismic data, to achieve the initial lithology prediction work under the condition that the loess plateau surface causes the consistency problem of seismic energy, effectively overcome the shortcomings of the method that relies solely on seismic data signal processing means, and improve the accuracy and quantification of seismic lithology prediction. This method has achieved remarkable results in the actual application of the typical loess plateau surface conditions where the seismic energy has consistency problems.
[0072] To facilitate understanding of the solutions and effects of the embodiments of the present invention, a specific application example is given below. Those skilled in the art should understand that the example is only for facilitating understanding of the present invention, and any specific details thereof are not intended to limit the present invention in any way.
[0073] Example 1
[0074] This embodiment provides a reservoir lithology prediction method for the problem of surface seismic amplitude energy consistency in the loess plateau.
[0075] First, a geostatistical simulation is performed.
[0076] Geostatistical simulation, that is, by giving the lithology ratio of each lithology in the formation, the regular distribution of elastic parameters, and the spatial variogram of lithology and elasticity, sequential indicator simulation and sequential Gaussian simulation techniques are used to achieve spatial simulation of lithology and elasticity, and obtain multiple sets of lithology and elasticity parameter models. In actual work, it is necessary to use the lithology and elasticity data measured at the well point to estimate the geostatistical parameters (lithology ratio, elastic parameter probability distribution, variogram type and range). In actual implementation, there are certain requirements for the quantity and representativeness of well sample data, and it is necessary to ensure that the statistical parameters have statistical effects. In addition, data such as lithofacies plane maps established by geologists based on geological laws and well data can be used for trend constraints in geostatistical simulation.
[0077] Figure 2 The following are examples of probability distribution of elastic parameters for different lithologies. In the figure, D represents coal, C represents mudstone, B represents water sand, and A represents gas sandstone. The top part is the probability distribution of P-wave impedance and the bottom part is the probability distribution of P-wave velocity ratio.
[0078] Next, we conduct seismic convolution forward modeling. Since the simulated lithology data, elastic data and seismic data belong to different physical quantities, we need to use the seismic convolution forward modeling method to convert the elastic data into seismic data.
[0079] First, the Aki-Richard equation is used to convert the elastic parameters obtained from geostatistical simulation into reflection coefficients that vary with offset. Equation (1) is the mathematical form of the Aki-Richard equation:
[0080]
[0081] Where R is the reflection coefficient, θ = (θ1 + θ2) / 2 is the incident angle, θ1 and θ2 are the incident angle and transmission angle respectively, and p is the ray parameter. Δρ=ρ2-ρ1, ρ=(ρ2+ρ1) / 2, ρ1 is the longitudinal wave velocity, shear wave velocity and density above the interface, ρ2 is the longitudinal wave velocity, shear wave velocity and density below the interface.
[0082] Secondly, the reflection coefficient is converted into seismic reflection using the convolution equation and seismic wavelet. Equation (2) is the mathematical form of the convolution equation:
[0083] S(t, θ)=R(t, θ)*W(t, θ) (2)
[0084] Where t is the earthquake travel time, W(t, θ) is the seismic wavelet, S(t, θ) is the synthetic seismic record, and * is the convolution operator. Since the geostatistical simulation obtains multiple sets of results, the forward seismic data obtained are also multiple sets. Figure 3 This is the seismic profile obtained by forward modeling.
[0085] Next, convolutional neural network training is performed, and finally the convolutional neural network is used for lithology identification.
[0086] In this embodiment, a deep feed-forward neural network (DFNN) is used to implement it. By adding convolutional layers and pooling layers, a deeper network can be implemented to achieve deep learning. The specific network settings (number of convolutional layers and pooling layers) need to be adjusted according to the actual data and research difficulty. First, the descriptive ability of the neural network becomes stronger with the increase of convolutional layers, but there are problems such as long calculation time and overfitting. The default setting is 4 convolutional layers. If the result does not converge, the number of convolutional layers can be increased. If the verification data has an overfitting problem, the number of pooling layers can be increased to reduce the impact of overfitting.
[0087] The training of the neural network is mainly to obtain the optimal weight coefficient W and bias term b. The mathematical form of the objective function (Formula 3) is as follows:
[0088]
[0089] Wherein, W and b are weight coefficient and bias term respectively, and x and y are input data and training target. In the present invention, the Adam method is used to achieve the convergence of the objective function. In practical application, it is divided into two major links: training and application of convolutional neural network model.
[0090] Figure 4 Comparison of learning objectives and training results for 22 well points. In the left figure, light gray is mudstone, dark gray is sandstone, and the second dark gray is coal seam. The comparison results can be more clearly seen through the color-coded map. In the right figure, gray is mudstone, red is sandstone, and blue is coal seam. Figure 4 The left and right pictures in the middle are exactly the same pictures, the difference is that the left picture is a grayscale picture and the right picture is a color picture.
[0091] By comparison, it can be seen that the method of the present invention has a strong learning ability for rock layers with a certain thickness, ensuring the convergence of the model.
[0092] Example 2
[0093] The experiment of this embodiment is aimed at the 150km2 loess plateau coverage area, using 22 wells as training samples and 13 wells as verification wells to conduct effect testing. Figure 5 The number of iterations and errors of model training. The red figure shows the convergence of all 22 sample wells as samples for model training, and the black figure shows the convergence of 20 wells after randomly deducting 2 wells for sample training. It can be seen from the figure that the objective function of the convolutional neural network model converges faster using the Adam method.
[0094] The trained convolutional neural network model was used to predict the lithology of 14 post-test wells.
[0095] Table 2-13 Comparison of prediction (left) and actual measurement (right) of post-test wells
[0096]
[0097] Table 2 is a comparison table of the predictions (left) and actual measurements (right) of 13 post-test wells. In the table, black represents coal, light gray represents mudstone, and dark gray represents sandstone. As can be seen from the figure, although there are certain errors in the prediction of individual wells (lithology prediction results and lithology development locations), the overall prediction of the target layer lithology in the loess plateau area can be well achieved, verifying the robustness of the algorithm.
[0098] Example 3
[0099] Figure 6The plane characteristics (c) of the reservoir lithology prediction results under the surface conditions of the loess plateau are compared with the lateral energy distribution of seismic amplitude (a), satellite remote sensing data (b) and the understanding of geological sedimentary trends. As can be seen from the figure, there is a strong correlation between the distribution trend of seismic amplitude energy (a) and satellite remote sensing data (b), that is, the imaging northwest-southeast distribution characteristics. At the same time, the understanding of geological sedimentary trends (d) shows a northeast-southwest feature, and the distribution trend of predicted sandstone thickness is relatively consistent with it, showing the basic structure of river channel distribution. Due to the descriptive role of satellite remote sensing data for the amplitude energy of non-geological factors and the implementation of geostatistical simulation and seismic forward modeling, the prediction risk caused by the lateral consistency of amplitude energy is effectively compensated, and the uncertainty of lithology prediction is reduced.
[0100] This method has achieved remarkable results in practical application in typical loess plateau surface conditions where there is a consistency problem in seismic energy.
[0101] Example 4
[0102] This embodiment provides a reservoir lithology prediction device for the loess plateau surface seismic amplitude energy consistency problem, comprising:
[0103] The simulation module obtains reservoir and elastic data through geostatistical simulation;
[0104] A forward modeling module, which converts the elastic data into simulated seismic data using a seismic convolution forward modeling method;
[0105] The training module uses the measured seismic data, satellite remote sensing data, and simulated seismic data at the well point as inputs and the measured lithology at the well point as training targets to train the neural network model;
[0106] The prediction module uses the neural network model to predict reservoir lithology.
[0107] The simulation module, forward modeling module, training module and prediction module are electrically connected in sequence. The simulation module sends the obtained and elastic data to the forward modeling module to convert the elastic data into simulated seismic data. The training module receives the measured seismic data, satellite remote sensing data and simulated seismic data at the well point, and trains the neural network model with the measured lithology of the well point as the training target. The prediction module uses the trained neural network model to predict the reservoir lithology.
[0108] Example 5
[0109] This embodiment provides an electronic device, which includes: a memory storing executable instructions; a processor running the executable instructions in the memory to implement the reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem.
[0110] An electronic device according to an embodiment of the present disclosure includes a memory and a processor.
[0111] The memory is used to store non-temporary computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.
[0112] The processor may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In one embodiment of the present disclosure, the processor is used to run the computer-readable instructions stored in the memory.
[0113] Those skilled in the art should be able to understand that in order to solve the technical problem of how to obtain a good user experience, the present embodiment may also include well-known structures such as a communication bus and an interface, and these well-known structures should also be included in the protection scope of the present disclosure.
[0114] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.
[0115] Example 6
[0116] This embodiment provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem is implemented.
[0117] According to the computer-readable storage medium of the embodiment of the present disclosure, non-transitory computer-readable instructions are stored thereon. When the non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the above-mentioned methods of each embodiment of the present disclosure are executed.
[0118] The above-mentioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or mobile hard disk), media with built-in rewritable non-volatile memory (e.g., memory card) and media with built-in ROM (e.g., ROM box).
[0119] Those skilled in the art should understand that the purpose of the above description of the embodiments of the present invention is only to exemplarily illustrate the beneficial effects of the embodiments of the present invention, and is not intended to limit the embodiments of the present invention to any given examples.
[0120] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.
Claims
1. A reservoir lithology prediction method for the problem of surface seismic amplitude energy consistency in loess plateau, characterized in that: include: Through geostatistical simulation, the elasticity data of the reservoir is obtained; Using a seismic convolution forward modeling method, the elastic data is converted into simulated seismic data; The neural network model is trained using the measured seismic data, satellite remote sensing data, and simulated seismic data at the well point as input and the measured lithology at the well point as the training target; Using the neural network model to predict reservoir lithology; The geostatistical simulation includes: by giving the lithology ratio of each lithology in the stratum, the regular distribution of elastic parameters and the spatial variation function of lithology and elasticity, using sequential indicator simulation and sequential Gaussian simulation technology respectively, to achieve spatial simulation of lithology and elasticity, and obtain multiple groups of lithology parameter models and elasticity parameter models; Wherein, converting the elastic data into simulated seismic data comprises: The elastic data is converted into a reflection coefficient that varies with offset using the Aki-Richard equation; Converting the reflection coefficient into a seismic reflection using a convolution equation and a seismic wavelet; Wherein, the Aki-Richard equation is as follows: Where R is the reflection coefficient, θ = (θ1 + θ2) / 2 is the incident angle, θ1 and θ2 are the incident angle and transmission angle respectively, and p is the ray parameter. Δρ=ρ2-ρ1, ρ=(ρ2+ρ1) / 2, ρ1 is the longitudinal wave velocity, shear wave velocity and density above the interface, ρ2 is the longitudinal wave velocity, shear wave velocity and density below the interface.
2. The reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem according to claim 1 is characterized by: The convolution equation is as follows: S(t,θ)=R(t,θ)*W(t,θ) (2) Where t is the earthquake travel time, W(t,θ) is the earthquake wavelet, S(t,θ) is the synthetic earthquake record, and * is the convolution operator.
3. The reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem according to claim 1 is characterized by: The objective function of the neural network model is as follows: Among them, W and b are weight coefficients and bias terms respectively, x and y are input data and training targets, P is the dimension of the data, and T is the matrix transpose operator.
4. The reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem according to claim 3 is characterized by: The Adam method is used to achieve convergence of the objective function.
5. A reservoir lithology prediction device for the problem of consistency of loess plateau surface seismic amplitude energy, characterized in that: include: The simulation module obtains reservoir and elastic data through geostatistical simulation; A forward modeling module, which converts the elastic data into simulated seismic data using a seismic convolution forward modeling method; The training module uses the measured seismic data, satellite remote sensing data, and simulated seismic data at the well point as inputs and the measured lithology at the well point as training targets to train the neural network model; A prediction module, using the neural network model to predict reservoir lithology; The geostatistical simulation includes: by giving the lithology ratio of each lithology in the stratum, the regular distribution of elastic parameters and the spatial variation function of lithology and elasticity, using sequential indicator simulation and sequential Gaussian simulation technology respectively, to achieve spatial simulation of lithology and elasticity, and obtain multiple groups of lithology parameter models and elasticity parameter models; Wherein, converting the elastic data into simulated seismic data comprises: The elastic data is converted into a reflection coefficient that varies with offset using the Aki-Richard equation; Converting the reflection coefficient into a seismic reflection using a convolution equation and a seismic wavelet; Wherein, the Aki-Richard equation is as follows: Where R is the reflection coefficient, θ = (θ1 + θ2) / 2 is the incident angle, θ1 and θ2 are the incident angle and transmission angle respectively, and p is the ray parameter. Δρ=ρ2-ρ1, ρ=(ρ2+ρ1) / 2, ρ1 is the longitudinal wave velocity, shear wave velocity and density above the interface, ρ2 is the longitudinal wave velocity, shear wave velocity and density below the interface.
6. An electronic device, characterized in that: The electronic device comprises: A memory storing executable instructions; A processor, wherein the processor runs the executable instructions in the memory to implement the reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem according to any one of claims 1-4.
7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements a reservoir lithology prediction method for the loess plateau surface seismic amplitude energy consistency problem described in any one of claims 1-4.
