Density attribute earthquake prediction method and device, medium and equipment
Through rock physics modeling and convolutional neural network deep learning methods, the problem of identifying effective reservoirs in conventional elastic parameters under medium and deep conditions is solved, density attribute prediction under the condition of few wells on the sea is realized, signal-to-noise ratio of the prediction results is improved, and technical support is provided for exploration.
Patent Information
- Application Number
- CN202510101039.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-22
- Publication Date
- 2025-05-06
AI Technical Summary
Under medium and deep conditions, it is difficult for the prior art to accurately identify effective reservoirs through conventional elastic parameters, and due to the small well data in the offshore exploration stage, deep learning methods are limited in the application of density prediction.
A large number of pseudo-well curves are generated through petrophysical modeling, the label data set is expanded, and deep learning training is performed using convolutional neural networks. Through transfer learning and diffusion filtering processing, the signal-to-noise ratio of density attribute prediction is improved.
Under the condition of offshore small wells, deep learning density attribute prediction based on sample expansion of rock physics model is realized, which improves the signal-to-noise ratio of the prediction results, and provides effective technical support for the comprehensive evaluation of targets and well site deployment in the exploration stage.
Smart Images
Figure CN119937008A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of oil and gas exploration, and in particular, relates to a density attribute earthquake prediction method. Background Art
[0002] Through the analysis of a large amount of actual data, it is found that under medium-deep conditions, when the P-wave impedance and P-S wave velocity ratio cannot distinguish effective reservoirs, density is often an effective parameter for distinguishing lithology, physical properties and even hydrocarbon content. The conventional inversion method based on the convolution model calculates elastic parameters, and the reflection coefficient is calculated by the Zoeppritz simplified equation. Density inversion requires pre-stack data with a large angle of incidence to make the results stable. However, when the actual data has a large angle of incidence, the long-distance tensile distortion will be more serious, and the data is almost unusable.
[0003] Using deep learning methods to predict density is a development direction. Deep learning is a branch of artificial intelligence. It is essentially a neural network that can simulate the human brain for analysis and learning. Deep learning methods require a large amount of labeled data. Previous studies have shown that increasing training samples can greatly improve the prediction accuracy of neural networks. Big data is the basis of deep learning. However, there are often few well data in the offshore exploration stage. There is not enough labeled data for neural network training using conventional methods. For deep learning density prediction under offshore conditions with few wells, there is often a lack of a large amount of labeled data, and the application of deep learning methods is limited. In order to expand the training samples, conventional methods often use rock physics modeling to simulate different reservoir segment parameters to obtain a large number of samples, and then perform post-stack deep feedforward neural network attribute training. This method is mainly used for multi-attribute analysis of post-stack data because the neural network belongs to a fully connected model and each neuron is directly connected to the lower-level neuron. The amount of calculation is large. The pre-stack data contains rich information about the change of amplitude with offset, which can effectively reflect the change of reservoir parameters. After filtering and sampling through the convolution layer and the pooling layer, the convolutional neural network model has fewer sampling points and is a local connection model with higher computational efficiency. It can be applied to the density attribute prediction of prestack data. However, the results obtained by this method often have a low signal-to-noise ratio and poor presentation of the prediction results.
[0004] At present, the industry as a whole has little research on the direct use of pre-stack data for density attribute prediction and practical application, and there is no mature geophysical method that can predict density attributes in areas with few wells during the exploration stage. Summary of the invention
[0005] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention provides a density attribute seismic prediction method, which aims to solve the problem that conventional elastic parameters in the prior art cannot accurately identify effective reservoirs under medium-deep conditions. The final reservoir prediction results provide technical support for target comprehensive evaluation and well location deployment.
[0006] To achieve the above object, the present invention adopts the following technical solutions:
[0007] In a first aspect, the present invention provides a density attribute earthquake prediction method, comprising the following steps:
[0008] For the wells that have been drilled in the study area, the P-wave velocity, S-wave velocity, density, porosity, mineral content and water saturation curves of the target layer in the study area are collected, and the bulk modulus and density of the rock skeleton in the study area are adjusted to forward-model the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock, so that the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock are well consistent with those of the target layer in the study area.
[0009] By changing the shale content, porosity and water saturation parameters of the reservoir section, and using the bulk modulus and density parameters of the rock skeleton in the study area, forward modeling is continued to generate a large number of P-wave velocity, S-wave velocity and density curves of different pseudo-wells, and synthetic pre-stack seismic gathers are generated under the guidance of the Zoeppritz equation to expand the training data;
[0010] A large number of P-wave velocity, S-wave velocity and density curves of different pseudo-wells generated by forward simulation and synthetic pre-stack seismic gathers are used to train the convolutional neural network to obtain a trained convolutional neural network model.
[0011] The trained convolutional neural network model is applied to actual wells and actual data gathers. The trained convolutional neural network model is corrected using actual wells and actual data gathers through transfer learning. The corrected convolutional neural network model is used to apply the actual data gathers to predict density attributes and obtain density attribute prediction results.
[0012] As a preferred embodiment, the aforementioned “according to the wells drilled in the study area, the P-wave velocity, S-wave velocity, density, porosity, mineral content and water saturation curves of the target layer in the study area are collected, and the bulk modulus and density of the rock skeleton in the study area are adjusted, and the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock are forward modeled, so that the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock forward modeled have good consistency with the P-wave velocity, S-wave velocity and density curves of the target layer in the study area” is specifically:
[0013] The Gassmann equation is used to calculate the bulk modulus and shear modulus of fluid-saturated rock:
[0014]
[0015] G=G d
[0016] Where K is the bulk modulus of the fluid-saturated rock; K d is the bulk modulus of dry rock; K m is the bulk modulus of the rock skeleton; K f is the bulk modulus of the pore fluid; is the porosity of the fluid-saturated rock; G is the shear modulus of the fluid-saturated rock; G d is the shear modulus of dry rock;
[0017] The bulk modulus of the rock skeleton is calculated using the VRH average formula:
[0018] K M =(K V +K R ) / twenty two)
[0019] K V =V qua K qua +V shale K shale (3)
[0020]
[0021] Where V qua is the volume fraction of quartz in the rock skeleton; V shale is the volume fraction of clay in the rock skeleton; K qua is the bulk modulus of quartz in the rock skeleton; K shale is the bulk modulus of quartz in the rock framework;
[0022] The bulk modulus of the pore fluid is calculated by the WOOD formula:
[0023]
[0024] In the formula, K oil , K water and K gas are the bulk moduli of oil, water and gas respectively; V oil 、V water and V gas are the percentages of oil, water and gas respectively;
[0025] Based on the calculated bulk modulus and shear modulus of fluid-saturated rock, the following equations are used to forward model the P-wave velocity, S-wave velocity and density curves of fluid-saturated rock:
[0026]
[0027] ρ d =ρ qua V qua +ρ shale V shale +... (9)
[0028] ρ f =ρ w S w +ρ g (1-S w ) (10)
[0029] Where V p V is the longitudinal wave velocity of the fluid-saturated rock; s is the shear wave velocity of the fluid-saturated rock; ρ sat is the density of the fluid-saturated rock; ρ d is the density of dry rock, which mainly depends on the volume fraction of quartz, clay and other diagenetic matrix in the rock skeleton, and the density of quartz, clay and other diagenetic matrix in the rock matrix; ρ f is the density of the mixed fluid; ρ g is the density of gas; ρ w is the density of water; S w is the water saturation.
[0030] As a preferred method: before generating synthetic prestack seismic gathers under the guidance of the Zoeppritz equation, it is necessary to select Ricker wavelets according to the dominant frequency of the seismic data of the target layer, and generate synthetic prestack seismic gathers by combining Ricker wavelets with the Zoeppritz equation.
[0031] Preferably, the convolutional neural network training comprises the following steps:
[0032] Extract synthetic pre-stack seismic gather data, i.e., the variation of amplitude with offset and the variation of amplitude with time as input images for the convolutional neural network model;
[0033] Use convolutional layers to extract different features from the input image;
[0034] Then use the pooling layer to downsample, reduce the number of samples, use iterations, reduce parameters, and convert the input data to feature dimensions;
[0035] Finally, the data in the feature dimension is converted to the sample label dimension through the fully connected layer and then output to obtain the trained convolutional neural network model.
[0036] As a preference: the following steps are also included:
[0037] The obtained density attribute prediction results are subjected to diffusion filtering and amplitude-preserving denoising to improve the signal-to-noise ratio of the density attribute prediction results.
[0038] As a preferred embodiment, the “perform a diffusion filtering and amplitude-preserving denoising process on the obtained density attribute prediction result to improve the signal-to-noise ratio of the density attribute prediction result” is specifically:
[0039] First, the density attribute prediction results are high-pass filtered to remove the low-frequency density attributes;
[0040] Then, the density attribute with low frequency removed is subjected to the denoising process of diffusion filtering and amplitude preservation;
[0041] Finally, the processed density attribute is added to the density attribute with low frequency removed to obtain the final density prediction result.
[0042] In a second aspect, the present invention provides a density attribute earthquake prediction device, comprising:
[0043] The first processing unit is used to collect the P-wave velocity, S-wave velocity, density, porosity, mineral content and water saturation curves of the target layer in the study area for the wells that have been drilled in the study area, and adjust the bulk modulus and density of the rock skeleton in the study area to forward-model the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock, so that the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock are well consistent with the P-wave velocity, S-wave velocity and density curves of the target layer in the study area;
[0044] The second processing unit is used to change the mud content, porosity and water saturation parameters of the reservoir section, and use the bulk modulus and density parameters of the rock skeleton in the study area to continue forward simulation to generate a large number of P-wave velocity, S-wave velocity and density curves of different pseudo-wells, and generate synthetic pre-stack seismic gathers under the guidance of the Zoeppritz equation to expand the training data;
[0045] The third processing unit is used to train the convolutional neural network using the P-wave velocity, S-wave velocity and density curves of a large number of different pseudo-wells and the synthetic pre-stack seismic gathers generated by the forward simulation to obtain a trained convolutional neural network model;
[0046] The fourth processing unit is used to apply the trained convolutional neural network model to actual wells and actual track gathers, modify the trained convolutional neural network model using the actual wells and actual track gathers through transfer learning, and use the modified convolutional neural network model to apply the actual track gathers to predict density attributes, so as to obtain density attribute prediction results.
[0047] In a third aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program is executed by a processor to control the device where the processor is located to implement the steps of the earthquake prediction method based on density attributes described in the first aspect of the present invention.
[0048] In a fourth aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the earthquake prediction method based on density attributes described in the first aspect of the present invention.
[0049] The present invention adopts the above technical solution, which has the following advantages:
[0050] 1. The present invention generates a large number of pseudo-well curves through rock physics modeling, uses the pseudo-well curve forward simulation to expand the label data set, and further trains the forward simulation results through the convolutional neural network deep learning method. Next, after transfer learning of the model algorithm, the actual pre-stack gathers are used for density prediction, and finally the density attribute prediction results are subjected to diffusion filtering to improve the signal-to-noise ratio of the prediction results.
[0051] 2. The present invention can carry out deep learning density attribute prediction based on rock physics model sample expansion under offshore conditions with few wells; at the same time, through diffusion filtering and amplitude preservation denoising processing, the problem of low signal-to-noise ratio of density attribute prediction results of pre-stack data can be solved, and the signal-to-noise ratio of prediction results can be effectively improved; in the case of amplitude preservation processing of seismic data, the density prediction results are consistent with the verification wells, and the density attributes are reliable.
[0052] 3. The present invention can provide effective technical support for comprehensive evaluation of targets and well site deployment in areas with few wells during the exploration phase through fine characterization. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Throughout the accompanying drawings, the same reference numerals are used to represent the same components. In the accompanying drawings:
[0054] Figure 1 A flow chart of a density attribute earthquake prediction method provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0055] In order to make the purpose, technical scheme and advantages of the present invention clearer, the specific embodiments of the present invention are further described below in conjunction with the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.
[0056] The seismic prediction method of density attribute provided by the present invention includes: forward modeling the P-wave velocity, S-wave velocity and density curve of fluid-saturated rock; using the bulk modulus and density parameters of the rock skeleton in the study area, continuing forward simulation to generate a large number of P-wave velocity, S-wave velocity and density curves of different pseudo-wells, and generating synthetic pre-stack seismic gathers under the guidance of the Zoeppritz equation; using the P-wave velocity, S-wave velocity and density curves of different pseudo-wells and the synthetic pre-stack seismic gathers to train a convolutional neural network to obtain a trained convolutional neural network model; applying the convolutional neural network model to actual wells and gathers, correcting the trained convolutional neural network model, and using the corrected convolutional neural network model to apply the actual gathers to predict density attributes. The present invention predicts density attributes by a deep learning method based on rock physics model sample expansion, solving the problem that conventional elastic parameters in the prior art cannot identify effective reservoirs under medium-deep conditions.
[0057] The density attribute earthquake prediction method and device provided by the embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0058] Embodiment 1:
[0059] See also Figure 1 , this embodiment provides a density attribute earthquake prediction method, comprising the following steps:
[0060] S100. For the wells that have been drilled in the study area, collect the P-wave velocity, S-wave velocity, density, porosity, mineral content and water saturation curves of the target layer in the study area, adjust the bulk modulus and density of the rock skeleton in the study area, and forward model the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock, so that the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock are well consistent with the P-wave velocity, S-wave velocity and density curves of the target layer in the study area, specifically:
[0061] S101. Using Gassman n The bulk modulus and shear modulus of fluid-saturated rock are calculated using the following equations:
[0062]
[0063] G=G d
[0064] Where K is the bulk modulus of the fluid-saturated rock; K d is the bulk modulus of dry rock, obtained by laboratory measurement; K m is the bulk modulus of the rock skeleton; K f is the bulk modulus of the pore fluid; is the porosity of the fluid-saturated rock; G is the shear modulus of the fluid-saturated rock; G d is the shear modulus of dry rock, obtained by laboratory measurement;
[0065] S102. Calculate the bulk modulus of the rock skeleton using the VRH average formula (i.e., the method of averaging the upper and lower boundaries of Voigt and Reuss):
[0066] K M =(K V +K R ) / twenty two)
[0067] K V =V qua K qua +V shale K shale (3)
[0068]
[0069] Where V qua is the volume fraction of quartz in the rock skeleton; V shale is the volume fraction of clay in the rock skeleton; K qua is the bulk modulus of quartz in the rock skeleton; K shale is the bulk modulus of quartz in the rock framework;
[0070] S103. Calculate the bulk modulus of the pore fluid using the WOOD formula:
[0071]
[0072] In the formula, K oil , K water and K gas are the bulk moduli of oil, water and gas, respectively, obtained through empirical parameters; V oil 、V water and V gas are the percentages of oil, water and gas, respectively, obtained from the water saturation logging curve;
[0073] S104. Based on the calculated bulk modulus and shear modulus of the fluid-saturated rock, the following equations are used to forward model the longitudinal wave velocity, transverse wave velocity and density curve of the fluid-saturated rock:
[0074]
[0075]
[0076] ρ d =ρ qua V qua +ρ shale V shale +... (9)
[0077] ρ f =ρ w S w +ρ g (1-S w ) (10)
[0078] Where V p V is the longitudinal wave velocity of the fluid-saturated rock; s is the shear wave velocity of the fluid-saturated rock; ρ sat is the density of the fluid-saturated rock; ρ d is the density of dry rock, which mainly depends on the volume fraction of quartz, clay and other diagenetic matrix in the rock skeleton, and the density of quartz, clay and other diagenetic matrix in the rock matrix; ρ f is the density of the mixed fluid; ρ g is the density of gas; ρ w is the density of water; S w is the water saturation.
[0079] S200. By changing the mud content, porosity and water saturation parameters of the reservoir section, and using the bulk modulus and density parameters of the rock skeleton in the study area, forward simulation is continued to generate a large number of P-wave velocity, S-wave velocity and density curves of different pseudo-wells, and synthetic pre-stack seismic gathers are generated under the guidance of the Zoeppritz equation to expand the training data.
[0080] In the above embodiment, preferably, before generating synthetic prestack seismic gathers under the guidance of the Zoeppritz equation, it is necessary to select Ricker wavelets according to the main frequency of the seismic data of the target layer segment, and generate synthetic prestack seismic gathers by combining Ricker wavelets with the Zoeppritz equation.
[0081] The Zoeppritz equation is the reflection coefficient equation commonly used in industry.
[0082] S300. Perform convolutional neural network training using a large number of P-wave velocity, S-wave velocity and density curves of different pseudo-wells and synthetic pre-stack seismic gathers generated by forward modeling in step S200 to obtain a trained convolutional neural network model;
[0083] Convolutional neural networks (CNN) are a type of feedforward neural network with deep structure and convolution operation, and are one of the representative algorithms of deep learning. The convolutional neural network model mainly includes the following five layers: input layer, convolution layer, pooling layer, fully connected layer and output layer. Among them, convolutional neural network training includes the following steps:
[0084] S301. Extracting synthetic pre-stack seismic gather data, namely, the change of amplitude with offset and the change of amplitude with time as input images of the convolutional neural network model;
[0085] S302. Using the convolution layer to perform operations, extract different features from the input image;
[0086] S303. Use the pooling layer to downsample, reduce the number of samples, use iteration, reduce parameters, and convert the input data to feature dimensions;
[0087] S304. Finally, the data of the feature dimension is converted into the sample label dimension through the fully connected layer and then output to obtain the trained convolutional neural network model.
[0088] Compared with other types of neural network models, the convolutional neural network model is a locally connected model. After filtering and sampling the data through the convolution layer and the pooling layer, the number of sampling points is fewer and the computational efficiency is higher.
[0089] S400. Apply the trained convolutional neural network model to actual wells and actual trace gathers, modify the trained convolutional neural network model using the actual wells and actual trace gathers through transfer learning, use the modified convolutional neural network model to apply the actual trace gathers to perform density attribute prediction, and obtain density attribute prediction results.
[0090] Transfer learning refers to using a trained model to train another model. This trained model has been trained on a very large data set. The process of transfer learning is essentially the process of continuing to update the fitting of the original nonlinear relationship with new sample data.
[0091] By transfer learning, the convolutional neural network model trained in step S300 is corrected using actual wells and actual track gathers, which can reduce the uncertainty of the convolutional neural network obtained only by forward modeling track gathers and forward logging curves, especially reducing the impact of the difference between the wavelet and the actual seismic data, the energy difference between the synthetic track gather and the actual seismic data on the convolutional neural network model.
[0092] S500. Performing a diffusion filtering and amplitude-preserving denoising process on the density attribute prediction result obtained in step S400 to improve the signal-to-noise ratio of the density attribute prediction result, specifically:
[0093] S501. First, high-pass filter the density attribute prediction result in step S400 to remove low-frequency density attributes, for example, remove density attributes below 6 Hz;
[0094] S502. Then, the density attribute with low frequency removed is subjected to a diffusion filtering and amplitude-preserving denoising process;
[0095] S503. Finally, the processed density attribute is added to the density attribute after removing the low frequency to obtain the final density prediction result.
[0096] After testing, under the condition of amplitude preservation processing of seismic data, the density attributes of the participating well locations of this method are consistent with those of the drilled wells, and the density attributes of the verification well locations are well consistent, which verifies the effectiveness of the method.
[0097] Embodiment 2:
[0098] The above-mentioned embodiment 1 provides an earthquake prediction method with density attributes, and correspondingly, this embodiment provides an earthquake prediction device with density attributes. The earthquake prediction device provided in this embodiment can implement the method of embodiment 1, and the earthquake prediction device can be implemented by software, hardware, or a combination of software and hardware. For example, the earthquake prediction device may include integrated or separate functional modules or functional units to execute the corresponding steps in each method of embodiment 1. Since the earthquake prediction device of this embodiment is basically similar to the method embodiment, the description process of this embodiment is relatively simple, and the relevant parts can refer to the partial description of embodiment 1. The earthquake prediction device of this embodiment is only schematic.
[0099] The density attribute earthquake prediction device provided in this embodiment includes:
[0100] The first processing unit is used to collect the P-wave velocity, S-wave velocity, density, porosity, mineral content and water saturation curves of the target layer in the study area for the wells that have been drilled in the study area, and adjust the bulk modulus and density of the rock skeleton in the study area to forward-model the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock, so that the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock are well consistent with the P-wave velocity, S-wave velocity and density curves of the target layer in the study area;
[0101] The second processing unit is used to change the mud content, porosity and water saturation parameters of the reservoir section, and use the bulk modulus and density parameters of the rock skeleton in the study area to continue forward simulation to generate a large number of P-wave velocity, S-wave velocity and density curves of different pseudo-wells, and generate synthetic pre-stack seismic gathers under the guidance of the Zoeppritz equation to expand the training data;
[0102] The third processing unit is used to train the convolutional neural network using the P-wave velocity, S-wave velocity and density curves of a large number of different pseudo-wells and the synthetic pre-stack seismic gathers generated by the forward simulation to obtain a trained convolutional neural network model;
[0103] The fourth processing unit is used to apply the trained convolutional neural network model to actual wells and actual track gathers, modify the trained convolutional neural network model using the actual wells and actual track gathers through transfer learning, and use the modified convolutional neural network model to apply the actual track gathers to predict density attributes, so as to obtain density attribute prediction results.
[0104] Embodiment 3:
[0105] This embodiment provides a processing device for implementing the earthquake prediction method of density attributes provided in this embodiment 1. The processing device can be a processing device for a client, such as a mobile phone, a laptop computer, a tablet computer, a desktop computer, etc., to execute the method of embodiment 1.
[0106] The processing device includes a processor, a memory, a communication interface and a bus, and the processor, the memory and the communication interface are connected through the bus to complete mutual communication. The memory stores a computer program that can be run on the processor, and the processor executes the method provided in this embodiment 1 when running the computer program.
[0107] Preferably, the memory may be a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory.
[0108] Preferably, the processor may be a central processing unit (CPU), a digital signal processor (DSP) or other general-purpose processors of various types, which are not limited here.
[0109] Embodiment 4:
[0110] The earthquake prediction method of density attributes of the present embodiment 1 can be specifically implemented as a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing the method described in the present embodiment 1.
[0111] Computer readable storage media can be tangible devices that hold and store instructions used by instruction execution devices. Computer readable storage media can be, for example, but not limited to, electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any combination thereof.
[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and they should all be included in the scope of the claims and specification of the present application. In particular, as long as there is no structural conflict, the various technical features mentioned in the various embodiments can be combined in any way. The present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions that fall within the scope of the claims.
Claims
1. A density attribute earthquake prediction method, characterized in that: The following steps are involved: For the wells that have been drilled in the study area, the P-wave velocity, S-wave velocity, density, porosity, mineral content and water saturation curves of the target layer in the study area are collected, and the bulk modulus and density of the rock skeleton in the study area are adjusted to forward-model the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock, so that the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock are well consistent with those of the target layer in the study area. By changing the shale content, porosity and water saturation parameters of the reservoir section, and using the bulk modulus and density parameters of the rock skeleton in the study area, forward modeling is continued to generate a large number of P-wave velocity, S-wave velocity and density curves of different pseudo-wells, and synthetic pre-stack seismic gathers are generated under the guidance of the Zoeppritz equation to expand the training data; A large number of P-wave velocity, S-wave velocity and density curves of different pseudo-wells generated by forward simulation and synthetic pre-stack seismic gathers are used to train the convolutional neural network to obtain a trained convolutional neural network model. The trained convolutional neural network model is applied to actual wells and actual data gathers. The trained convolutional neural network model is corrected using actual wells and actual data gathers through transfer learning. The corrected convolutional neural network model is used to apply the actual data gathers to predict density attributes and obtain density attribute prediction results.
2. The earthquake prediction method according to claim 1, characterized in that: The "according to the wells drilled in the study area, the P-wave velocity, S-wave velocity, density, porosity, mineral content and water saturation curves of the target layer in the study area are collected, and the bulk modulus and density of the rock skeleton in the study area are adjusted, and the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock are forward modeled, so that the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock forward modeled have good consistency with the P-wave velocity, S-wave velocity and density curves of the target layer in the study area" is specifically: The Gassmann equation is used to calculate the bulk modulus and shear modulus of fluid-saturated rock: G=G d Where K is the bulk modulus of the fluid-saturated rock; K d is the bulk modulus of dry rock; K m is the bulk modulus of the rock skeleton; K f is the bulk modulus of the pore fluid; is the porosity of the fluid-saturated rock; G is the shear modulus of the fluid-saturated rock; G d is the shear modulus of dry rock; The bulk modulus of the rock skeleton is calculated using the VRH average formula: K M =(K V +K R ) / 2 (2) K V =V qua K qua +V shale K shale (3) Where V qua is the volume fraction of quartz in the rock skeleton; V shale is the volume fraction of clay in the rock skeleton; K qua is the bulk modulus of quartz in the rock skeleton; K shale is the bulk modulus of quartz in the rock framework; The bulk modulus of the pore fluid is calculated by the WOOD formula: In the formula, K oil , K water and K gas are the bulk moduli of oil, water and gas respectively; V oil 、V water and V gas are the percentages of oil, water and gas respectively; Based on the calculated bulk modulus and shear modulus of fluid-saturated rock, the following equations are used to forward model the P-wave velocity, S-wave velocity and density curves of fluid-saturated rock: r d =ρ qua V qua +r shale V shale +... (9) r f =ρ w S w +r g (1-S w ) (10) Where V p V is the longitudinal wave velocity of the fluid-saturated rock; s is the shear wave velocity of the fluid-saturated rock; ρ sat is the density of the fluid-saturated rock; ρ d is the density of dry rock, which mainly depends on the volume fraction of quartz, clay and other diagenetic matrix in the rock skeleton, and the density of quartz, clay and other diagenetic matrix in the rock matrix; ρ f is the density of the mixed fluid; ρ g is the density of gas; ρ w is the density of water; S w is the water saturation.
3. The earthquake prediction method according to claim 2, characterized in that: Before generating synthetic prestack seismic gathers under the guidance of the Zoeppritz equation, it is necessary to select Ricker wavelets according to the dominant frequency of the seismic data in the target layer, and generate synthetic prestack seismic gathers by combining Ricker wavelets with the Zoeppritz equation.
4. The earthquake prediction method according to claim 3, characterized in that: The convolutional neural network training includes the following steps: Extract synthetic pre-stack seismic gather data, i.e., the variation of amplitude with offset and the variation of amplitude with time as input images for the convolutional neural network model; Use convolutional layers to extract different features from the input image; Then use the pooling layer to downsample, reduce the number of samples, use iterations, reduce parameters, and convert the input data to feature dimensions; Finally, the data in the feature dimension is converted to the sample label dimension through the fully connected layer and then output to obtain the trained convolutional neural network model.
5. The earthquake prediction method according to claim 4, characterized in that: The following steps are also included: The obtained density attribute prediction results are subjected to diffusion filtering and amplitude-preserving denoising to improve the signal-to-noise ratio of the density attribute prediction results.
6. The earthquake prediction method according to claim 5, characterized in that: The aforementioned “performing a diffusion filtering and amplitude-preserving denoising process on the obtained density attribute prediction result to improve the signal-to-noise ratio of the density attribute prediction result” is specifically: First, the density attribute prediction results are high-pass filtered to remove the low-frequency density attributes; Then, the density attribute with low frequency removed is subjected to the denoising process of diffusion filtering and amplitude preservation; Finally, the processed density attribute is added to the density attribute with low frequency removed to obtain the final density prediction result.
7. A density attribute earthquake prediction device, characterized in that: include: The first processing unit is used to collect the P-wave velocity, S-wave velocity, density, porosity, mineral content and water saturation curves of the target layer in the study area for the wells that have been drilled in the study area, and adjust the bulk modulus and density of the rock skeleton in the study area to forward-model the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock, so that the P-wave velocity, S-wave velocity and density curves of the saturated fluid rock are well consistent with the P-wave velocity, S-wave velocity and density curves of the target layer in the study area; The second processing unit is used to change the mud content, porosity and water saturation parameters of the reservoir section, and use the bulk modulus and density parameters of the rock skeleton in the study area to continue forward simulation to generate a large number of P-wave velocity, S-wave velocity and density curves of different pseudo-wells, and generate synthetic pre-stack seismic gathers under the guidance of the Zoeppritz equation to expand the training data; The third processing unit is used to train the convolutional neural network using the P-wave velocity, S-wave velocity and density curves of a large number of different pseudo-wells and the synthetic pre-stack seismic gathers generated by the forward simulation to obtain a trained convolutional neural network model; The fourth processing unit is used to apply the trained convolutional neural network model to actual wells and actual track gathers, modify the trained convolutional neural network model using the actual wells and actual track gathers through transfer learning, and use the modified convolutional neural network model to apply the actual track gathers to predict density attributes, so as to obtain density attribute prediction results.
8. A computer-readable storage medium, characterized in that: A computer program is stored, and the computer program is executed by a processor to control the device where the processor is located to implement the steps of the earthquake prediction method based on density attributes as described in any one of claims 1 to 6.
9. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the earthquake prediction method of density attributes as claimed in any one of claims 1 to 6 when executing the computer program.