River channel simulation and deep learning methods for constructing river sand strata frameworks

By simulating the spatial morphology of channel sand using sine and power functions, and combining deep confidence networks and forward modeling, the problem of inaccurate stratigraphic framework construction for channel sand was solved, resulting in a more reasonable low-frequency model and improving the accuracy of lithological prediction and physical property analysis for channel sand.

CN117420599BActive Publication Date: 2026-11-17CHINA PETROLEUM & CHEMICAL CORP +1
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202210811325.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-11
Publication Date
2026-11-17
Estimated Expiration
2042-07-11

AI Technical Summary

Technical Problem

Existing technologies cannot effectively construct stratigraphic frameworks that conform to the spatial distribution of river sand bodies, resulting in inaccurate prediction of river sandstone properties and characterization of hydrocarbon content.

Method used

The top and bottom surfaces of the channel sand spatial morphology are simulated using sine and power functions. Combined with depth confidence networks and forward modeling, a stratigraphic framework that conforms to the channel morphology is formed.

Benefits of technology

The constructed stratigraphic framework better matches the spatial distribution morphology of channel sand, enabling the generation of more reasonable low-frequency models and improving the accuracy of channel sand lithology prediction and physical property analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117420599B_ABST
    Figure CN117420599B_ABST
Patent Text Reader

Abstract

This invention provides a river channel simulation and deep learning method for constructing a stratigraphic framework for channel sands. The method includes: Step 1, statistically analyzing the channel parameters of the Guantao Formation in the target area to clarify the spatial distribution morphology of the channel sands; Step 2, simulating the envelope surface of the channel sand spatial morphology in batches based on sine and power functions; Step 3, determining three-dimensional samples of the channel sands based on forward modeling; Step 4, obtaining the top and bottom stratigraphic levels of the channel sands based on a deep confidence network; Step 5, forming a stratigraphic framework that conforms to the channel morphology based on the stratigraphic contact relationship, the standard layer, and the envelope surface of the channel sand body. The stratigraphic framework constructed by this method for constructing a stratigraphic framework for channel sands more closely matches the spatial distribution morphology of the channel sands, resulting in a more reasonable low-frequency model, which is more conducive to subsequent prediction of channel sand lithology, physical property analysis, and characterization of hydrocarbon content.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of exploration geophysics, and in particular to a method for river channel simulation and deep learning for constructing a framework for river sand formations. Background Technology

[0002] Channel sand reservoirs are an important type of reservoir for increasing reserves and production in major oilfields, accounting for a significant proportion of total reserves. Predicting channel sand reservoirs is a crucial factor limiting the accuracy of well placement in this type of reservoir. Various geophysical methods are used in channel sand exploration. Amplitude-based properties can clearly characterize the planar distribution of channel sand bodies; sparse pulse inversion can predict the spatial distribution of channel sand bodies, thereby obtaining the top and bottom surfaces and thickness parameters; stochastic inversion can predict thin sand bodies, compensating for the insufficient vertical resolution of sparse pulse inversion.

[0003] However, years of exploration and development practice have revealed that some channel sand bodies characterized by strong amplitudes in the amplitude attributes were not actually encountered in actual drilling. Furthermore, the channel morphology depicted by sparse pulse inversion and stochastic inversion does not align with geological understanding. Analysis indicates that the stratigraphic framework of the channel sand bodies does not conform to their spatial distribution, particularly the stochastic inversion results, which are overly dependent on the stratigraphic framework. Currently, no commercial software can construct a stratigraphic model that accurately reflects the morphology of channel sand bodies, thus failing to provide a reasonable low-frequency model to support subsequent prediction of channel sandstone properties, physical property analysis, and characterization of hydrocarbon content.

[0004] Chinese patent application CN202010022741.5 discloses a method, device, and readable storage medium for channel sand modeling. The channel sand modeling method includes the following steps: obtaining basic parameters of the reservoir to be modeled; generating several first channel sand body models and corresponding training condition data based on the obtained basic parameters; establishing a generator network; establishing a discriminator network; training the discriminator network and the generator network to obtain a generator; and inputting well data into the generator to obtain a sand body model of the reservoir to be modeled. The beneficial effects of this invention are: by generating a large number of first channel sand body models and corresponding training condition data through a goal-based simulation method, and then training the generator network and discriminator network using these first channel sand body models and training condition data, the generator network, after extensive training, obtains a generator capable of generating realistic channel sand body models. Finally, inputting well data from the study area into the generator yields a sand body model of the reservoir to be modeled in the study area.

[0005] Chinese patent application CN201410844281.9 discloses a method for identifying channel sandstone reservoirs, comprising: obtaining seismic response characteristics of channel sand bodies through forward modeling of the reservoir, and determining seismic identification markers for channel sand bodies in different geological bodies; extracting corresponding seismic attributes based on the seismic identification markers of channel sand bodies to determine channel boundaries; determining the depositional stages of channel sands using stratigraphic slicing technology under an isochronous stratigraphic framework; and determining the thickness of channel sand bodies based on the statistical fitting relationship between the thickness of sand bodies in actual wells and seismic attributes. Based on this method, in-depth seismic attribute extraction and analysis are conducted, utilizing known well reservoir and oil-gas layer calibration results, drilling statistical analysis, and combined with model forward modeling results to establish a channel sand body identification and description technology. This method can accurately identify and describe ultra-deep, thin-layered, and narrow-channel channel sandstone reservoirs, with minimal impact from seismic resolution.

[0006] Chinese patent application CN201510490826.5 discloses a method for fine identification of channel sand bodies. This invention first determines the seismic response characteristics corresponding to different sand body depositional patterns to identify channel features. Then, through desanding tests, it determines the phase relationship of sand layers on a time profile, and identifies the envelope of the channel sand bodies based on the seismic response characteristics. Stratigraphic slices are taken at equal time intervals from top to bottom to obtain the superposition relationship of channels at different stages. Cluster analysis of five basic attributes and other unconventional attributes is conducted to determine the boundaries of the channel sand bodies. Fine stratigraphic correlation can determine the internal spatial superposition relationship of the sand bodies, accurately determining the oil-bearing area and reserves of individual sand bodies, and determining the deployment range of the development well network. Through the above process, this invention solves technical problems such as complex reservoir planar distribution in multi-source convergence areas, strong sand body heterogeneity, and insufficient seismic data resolution to predict thin reservoirs, greatly improving drilling success rates and possessing broad market application prospects.

[0007] The existing technologies described above are significantly different from our invention and have failed to solve the technical problems we want to address. Therefore, we have invented a new method for river channel simulation and deep learning based on the construction of a grid for river sandy strata. Summary of the Invention

[0008] The purpose of this invention is to provide a channel simulation and deep learning method for constructing a channel sandstone stratigraphic framework, which can obtain a more reasonable low-frequency model and is more conducive to subsequent channel sandstone property prediction, physical property analysis and hydrocarbon characterization.

[0009] The objective of this invention can be achieved through the following technical measures: a river channel simulation and deep learning method based on a river channel sandy stratum framework, comprising:

[0010] Step 1: Statistically analyze the river parameters of the Guantao Formation in the target area to clarify the spatial distribution morphology of river sand in the area;

[0011] Step 2: Based on the sine function and the power function, simulate the envelope surface of the spatial morphology of riverbed sand in batches;

[0012] Step 3: Based on forward modeling, determine the three-dimensional samples of riverbed sand;

[0013] Step 4: Based on the deep belief network, obtain the top and bottom layers of the channel sand.

[0014] Step 5: Based on the stratigraphic contact relationship, and using the standard layer and the envelope of the channel sand body, form a stratigraphic framework that conforms to the channel morphology.

[0015] The objective of this invention can also be achieved through the following technical measures:

[0016] In step 1, for the target area Guantao Formation, parameters such as river channel direction, river width, river bend length, and curvature index are statistically analyzed to clarify the spatial distribution morphology of river sand in the target area.

[0017] In step 2, it can be seen from the river channel deposition pattern diagram and the current river channel morphology that the river channel morphology, i.e. the envelope, can be dissected into a cross section similar to a sine curve and a longitudinal section similar to a power function.

[0018] In step 2, multiple sinusoidal function accumulation signals and power functions are used to fit the spatial envelope of the river channel, i.e., the top and bottom surfaces of the riverbed sand, for both the cross-section and longitudinal section of the river channel; the specific formulas are as follows:

[0019]

[0020] Formula (1) is the fitting formula for the cross-section of the river channel, which is composed of multiple sine functions to simulate the planar distribution of the river channel. The x-axis is the direction of the river channel, and the y-axis is the direction of the river channel width. A i B represents the maximum variation of the sine function along the y-axis. i The x-axis represents the displacement of the sine function along the x-axis; Formula (2) is the fitting formula for the bottom of the longitudinal section of the river channel, which is an arc surface, the z-axis is the direction of the river channel depth, and the origin is taken at the middle position of the river surface; Formula (3) is the range of values ​​for many parameters, where D is the deepest point of the river channel, C is the maximum width of the river channel, and L is the length of the river channel.

[0021] In step 2, based on the simulated spatial distribution patterns of various river sands, the simulated river patterns are rotated in planar orientation within the river direction range to obtain more river patterns with different orientations.

[0022] In step 3, a forward model is constructed based on various river morphologies with different geological parameters such as different river widths, different river lengths, and different river orientations.

[0023] In step 3, based on the forward model of channel sand containing various morphologies, three-dimensional seismic images of channel sand are obtained using forward modeling technology.

[0024] In step 3, the three-dimensional seismic image of the river channel sand is used as the sample feature value and a label is set. Both the top and bottom surfaces of the river channel belong to the river channel envelope, and the envelope is labeled as 1, while other areas are labeled as 0.

[0025] In step 4, using forward samples as input, a deep belief network is used to continuously train the hyperparameters to obtain a training model with smaller errors.

[0026] In step 4, the trained model is applied to actual 3D seismic data to predict the envelope of the river channels contained in the actual seismic data.

[0027] In step 4, if the predicted envelope is not complete, the river channel simulation formula from step 2 is used to repair it.

[0028] In step 5, the target section standard layer is loaded, and based on the stratigraphic contact relationship, a stratigraphic framework conforming to the channel morphology is formed based on the standard layer and the envelope surface of the channel sand body.

[0029] This invention presents a river channel simulation and deep learning method for constructing a stratigraphic framework for channel sands. Starting from geological understandings such as river direction, width, bend length, and curvature index, it uses sine and power functions to simulate the top and bottom surfaces of the spatial morphology of channel sands in batches. Based on forward modeling and a deep confidence network, the top and bottom stratigraphic levels of the channel sands are obtained. According to stratigraphic contact relationships, a stratigraphic framework conforming to the channel morphology is constructed. Compared with traditional methods, the stratigraphic framework constructed by this method better reflects the spatial distribution of channel sands, resulting in a more reasonable low-frequency model, which is more conducive to subsequent prediction of channel sand lithology, physical property analysis, and characterization of hydrocarbon potential. Attached Figure Description

[0030] Figure 1 This is a flowchart of a specific embodiment of the river channel simulation and deep learning method for constructing a grid of river sand strata according to the present invention;

[0031] Figure 2 This is a schematic diagram of key geological parameters of Neogene channel sands in an oilfield, as statistically analyzed in a specific embodiment of the present invention.

[0032] Figure 3 This is a cross-sectional and longitudinal anatomical diagram of a river channel in a specific embodiment of the present invention;

[0033] Figure 4 This is a three-dimensional display of a simulated river channel in a specific embodiment of the present invention;

[0034] Figure 5 This is a top view of the envelope of multiple rivers in a certain work area predicted in a specific embodiment of the present invention. Detailed Implementation

[0035] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0036] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, and / or combinations thereof.

[0037] The present invention provides a river channel simulation and deep learning method for constructing a stratigraphic framework for river sand. Starting from geological knowledge such as river direction, river width, river bend length, and curvature index, it uses sine and power functions to simulate the top and bottom surfaces of the spatial morphology of river sand in batches. Based on forward modeling and deep confidence networks, the top and bottom stratigraphic levels of river sand are obtained. According to the stratigraphic contact relationship, a stratigraphic framework that conforms to the river channel morphology is constructed.

[0038] The following are several specific embodiments of the application of the present invention.

[0039] Example 1

[0040] In a specific embodiment 1 of the present invention, the river channel simulation and deep learning method for constructing a grid of river sand strata includes the following steps:

[0041] Step 1: Statistically analyze parameters such as river channel direction, river width, river bend length, and curvature index of the Guantao Formation in the target area to clarify the spatial distribution pattern of river sand in the area.

[0042] For the Guantao Formation in the target area, parameters such as river channel direction, width, bend length, and curvature index were statistically analyzed to clarify the spatial distribution of sand in the river channel. This lays the theoretical and data foundation for subsequent batch simulations of river channel morphology.

[0043] Step 2: Based on sine and power functions, simulate the envelope (top and bottom layers) of the spatial morphology of channel sand in batches.

[0044] As can be seen from the river channel depositional pattern diagram and the current river channel morphology, the river channel morphology (envelope) can be dissected into a cross section similar to a sine curve and a longitudinal section similar to a power function.

[0045] For both the cross-section and longitudinal section of the river channel, multiple sinusoidal function summations and power functions are used to fit the spatial envelope of the river channel (i.e., the top and bottom surfaces of the channel sand). The specific formulas are as follows:

[0046]

[0047] Formula (1) is the fitting formula for the cross-section of the river channel, which is a simulation of the planar distribution of the river channel by accumulating multiple sine functions. The x-axis is the direction of the river channel, and the y-axis is the direction of the river channel width. Formula (2) is the fitting formula for the bottom of the longitudinal section of the river channel, which is an arc surface. The z-axis is the direction of the river channel depth, and the origin is taken at the middle position of the river surface. Formula (3) is the range of values ​​for many parameters, where D is the deepest point of the river channel, C is the maximum width of the river channel, and L is the length of the river channel.

[0048] Based on the simulated spatial distribution of various riverbed sand morphologies, within the range of riverbed orientation, the simulated various riverbed morphologies are rotated in planar orientation to obtain more riverbed morphologies with different orientations.

[0049] Step 3: Based on forward modeling, determine the three-dimensional samples of riverbed sand;

[0050] A forward model is constructed based on simulating various river morphologies with different geological parameters such as river width, river length, and river direction.

[0051] Based on a forward model of channel sand containing various morphologies, three-dimensional seismic images of channel sand are obtained using forward modeling technology.

[0052] The 3D seismic image of the river channel sand was used as the sample feature value and labeled. Both the top and bottom surfaces of the river channel belong to the river channel envelope, and the envelope was labeled as 1, while other areas were labeled as 0.

[0053] Step 4: Based on the deep belief network, obtain the top and bottom layers of the channel sand.

[0054] Using forward-modeling samples as input, a deep belief network is used to continuously train hyperparameters, resulting in a training model with smaller errors.

[0055] The trained model is applied to actual 3D seismic data to predict the envelope of the river channels contained in the actual seismic data.

[0056] If the predicted envelope is not complete, it can be repaired using the river channel simulation formula in step 2.

[0057] Step 5: Based on the stratigraphic contact relationship, and using the standard layer and the envelope surface (top and bottom layers) of the channel sand body, form a stratigraphic framework that conforms to the channel morphology.

[0058] The target stratigraphic segment is loaded with a standard layer. Based on the stratigraphic contact relationship, and using the standard layer and the envelope of the channel sand body, a stratigraphic framework conforming to the channel morphology is formed.

[0059] Example 2

[0060] In a specific embodiment 2 of the present invention, such as Figure 1 As shown, Figure 1 This is a flowchart of the river channel simulation and deep learning method for constructing a grid of sandy soil layers in the river channel, as described in this invention.

[0061] In step 1, for the target area, Guantao Formation, parameters such as river channel direction, width, bend length, and curvature index are statistically analyzed to clarify the spatial distribution morphology of river sand in the target area. This lays the theoretical and data foundation for subsequent batch simulations of river channel morphology. Figure 2 As shown, Figure 2 This involves statistically analyzing key geological parameters of Neogene channel sands in a certain oilfield, including the range of values ​​for parameters such as channel direction, width, bend length, channel thickness, curvature index, strata, and seismic activity. By statistically analyzing these parameters, the spatial distribution morphology of channel sands in this area is clarified, laying a theoretical and data foundation for subsequent batch simulations of channel morphology.

[0062] In step 201, it can be seen from the channel deposition model diagram and the current channel morphology that the channel morphology (envelope) can be dissected into a cross-section similar to a sine curve and a longitudinal section similar to a power function. For example... Figure 3 As shown, Figure 3 It is an anatomical diagram of the cross-section and longitudinal section of the river channel. With the center point of the river surface as the origin, the x-axis is along the direction of the river channel, the y-axis is along the width of the river channel, and the z-axis is along the depth of the river channel. The cross-section of the river channel is located on the plane formed by the x-axis and y-axis, and its shape is similar to a sine function. The longitudinal section of the river channel is located on the plane formed by the y-axis and z-axis, and the bottom shape of the longitudinal section is similar to a power function.

[0063] In step 201, multiple sinusoidal function accumulation signals and power functions are used to fit the spatial envelope of the river channel (i.e., the top and bottom surfaces of the channel sand) for both the cross-section and longitudinal section. The specific formulas are as follows:

[0064]

[0065] Formula (1) is the fitting formula for the cross-section of the river channel, which simulates the planar distribution of the river channel by accumulating multiple sine functions. The x-axis represents the direction of the river channel, and the y-axis represents the width of the river channel. Formula (2) is the fitting formula for the bottom of the longitudinal section of the river channel, which is an arc surface. The z-axis represents the depth of the river channel, and the origin is taken at the middle position of the river surface. Formula (3) is the range of values ​​for many parameters, where D is the deepest point of the river channel, C is the maximum width of the river channel, and L is the length of the river channel. Figure 4 As shown, Figure 4It is a 3D display of a simulated river channel, which includes both the top and bottom surfaces of the channel, and the channel shape is simulated quite clearly.

[0066] In step 202, based on the simulated spatial distribution patterns of various riverbed sands, the simulated riverbed patterns are rotated in planar orientation within the riverbed direction range to obtain more riverbed patterns with different orientations.

[0067] In step 301, a forward model is constructed based on various river morphologies with different geological parameters such as different river widths, different river lengths, and different river directions.

[0068] In step 302, based on the forward model of channel sand containing various morphologies, a three-dimensional seismic image of the channel sand is obtained using forward modeling technology.

[0069] In step 303, the three-dimensional seismic image of the river channel sand is used as the sample feature value and a label is set. Both the top and bottom surfaces of the river channel belong to the river channel envelope. The envelope is labeled as 1, and other areas are labeled as 0.

[0070] In step 401, using forward samples as input, a deep belief network is used to continuously train the hyperparameters to obtain a training model with smaller errors.

[0071] In step 402, the trained model is applied to actual 3D seismic data to predict the envelope of the river channels contained in the actual seismic data. For example... Figure 5 As shown, Figure 5 It is a top view of the envelope of multiple rivers in a certain work area, and the river morphology is relatively complete.

[0072] In step 403, if the predicted envelope is not complete in shape, it can be repaired using the river channel simulation formula from step 2.

[0073] In step 5, the target section standard layer is loaded, and based on the stratigraphic contact relationship, a stratigraphic framework conforming to the channel morphology is formed based on the standard layer and the envelope surface of the channel sand body.

[0074] Example 3

[0075] In a specific embodiment 3 of the present invention, the river channel simulation and deep learning method for constructing a grid of river sand strata includes the following steps:

[0076] Step 1: Statistically analyze parameters such as river channel direction, river width, river bend length, and curvature index of the Guantao Formation in the target area to clarify the spatial distribution pattern of river sand in the area.

[0077] Step 2: Based on sine and power functions, simulate the envelope (top and bottom layers) of the spatial morphology of channel sand in batches.

[0078] As can be seen from the river channel depositional pattern diagram and the current river channel morphology, the river channel morphology (envelope) can be dissected into a cross section similar to a sine curve and a longitudinal section similar to a power function.

[0079] For both the cross-section and longitudinal section of the river channel, multiple sinusoidal function summations and power functions are used to fit the spatial envelope of the river channel (i.e., the top and bottom surfaces of the channel sand). The specific formulas are as follows:

[0080]

[0081] Formula (1) is the fitting formula for the cross-section of the river channel, which is a simulation of the planar distribution of the river channel by accumulating multiple sine functions. The x-axis is the direction of the river channel, and the y-axis is the direction of the river channel width. Formula (2) is the fitting formula for the bottom of the longitudinal section of the river channel, which is an arc surface. The z-axis is the direction of the river channel depth, and the origin is taken at the middle position of the river surface. Formula (3) is the range of values ​​for many parameters, where D is the deepest point of the river channel, C is the maximum width of the river channel, and L is the length of the river channel.

[0082] Based on the simulated spatial distribution of various riverbed sand morphologies, within the range of riverbed orientation, the simulated various riverbed morphologies are rotated in planar orientation to obtain more riverbed morphologies with different orientations.

[0083] Step 3: Based on forward modeling, determine the three-dimensional samples of riverbed sand;

[0084] A forward model is constructed based on simulating various river morphologies with different geological parameters such as river width, river length, and river direction.

[0085] Based on a forward model of channel sand containing various morphologies, three-dimensional seismic images of channel sand are obtained using forward modeling technology.

[0086] The 3D seismic image of the river channel sand was used as the sample feature value and labeled. Both the top and bottom surfaces of the river channel belong to the river channel envelope, and the envelope was labeled as 1, while other areas were labeled as 0.

[0087] Step 4: Based on the deep belief network, obtain the top and bottom layers of the channel sand.

[0088] Using forward-modeling samples as input, a deep belief network is used to continuously train hyperparameters (including the number of hidden layers, the number of nodes in each layer, the number of training iterations, the learning rate, etc.) to obtain a training model with smaller errors.

[0089] Based on the experts' interpretation of the results, sample data was added to revise and improve the training model.

[0090] The trained model is applied to actual 3D seismic data to predict the envelope of the river channels contained in the actual seismic data.

[0091] Step 5: Based on the stratigraphic contact relationship, and using the standard layer and the envelope surface (top and bottom layers) of the channel sand body, form a stratigraphic framework that conforms to the channel morphology.

[0092] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0093] Except for the technical features described in the specification, all other technologies are known to those skilled in the art.

Claims

1. A river channel simulation and deep learning method for a river channel sand formation framework, characterized in that, The river channel simulation and deep learning methods for constructing a lattice framework for river sand formations include: Step 1: Statistically analyze the river parameters of the Guantao Formation in the target area to clarify the spatial distribution morphology of river sand in the area; Step 2: Based on the sine function and the power function, simulate the envelope of the spatial distribution morphology of channel sand in batches. This includes: As can be seen from the channel deposition model diagram and the current channel morphology, the channel morphology, i.e. the envelope, can be dissected into a cross-section similar to a sine curve and a longitudinal section similar to a power function. For the cross-section and longitudinal section of the river channel, multiple sinusoidal function accumulation signals and power functions are used to fit the spatial envelope of the river channel, i.e., the top and bottom surfaces of the river channel sand; the specific formulas are as follows: ; Equation (1) is the fitting formula of river cross section, which is simulated by multiple sinusoidal functions to accumulate the river planar distribution pattern, x axis is the river trend, y axis is the river width direction, represents the maximum change amplitude of sinusoidal function along y axis, represents the displacement amount of sinusoidal function along x axis; Equation (2) is the fitting formula of river longitudinal section bottom, which is an arc surface, z axis is the river depth direction, and the origin is taken at the middle position of river surface; Equation (3) is the value range of numerous parameters, wherein D is the deepest part of the river, C is the maximum width of the river, and L is the length of the river; Step 3: Based on forward modeling, determine the three-dimensional samples of riverbed sand; Step 4: Based on the deep belief network, obtain the top and bottom layers of the channel sand. Step 5: Based on the stratigraphic contact relationship, and using the standard layer and the envelope of the channel sand body, form a stratigraphic framework that conforms to the channel morphology.

2. The river simulation and deep learning method for river sand formation framework building according to claim 1, characterized in that, In step 1, for the target area Guantao Formation, the parameters such as river direction, river width, river length, and curvature index are statistically analyzed to clarify the spatial distribution pattern of river sand in the target area.

3. The method for river channel simulation and deep learning based on the framework construction of river sand strata as described in claim 1, characterized in that, In step 2, based on the simulated spatial distribution patterns of various river sands, the simulated river patterns are rotated in planar orientation within the river direction range to obtain more river patterns with different orientations.

4. The method for river channel simulation and deep learning based on the framework construction of river sandy strata according to claim 1, characterized in that, In step 3, a forward model is constructed based on various river morphologies with different geological parameters such as different river widths, different river lengths, and different river orientations.

5. The method for river channel simulation and deep learning based on the framework construction of river sandy strata according to claim 4, characterized in that, In step 3, based on the forward model of channel sand containing various morphologies, three-dimensional seismic images of channel sand are obtained using forward modeling technology.

6. The method for river channel simulation and deep learning based on the framework construction of river sand strata as described in claim 5, characterized in that, In step 3, the three-dimensional seismic image of the river channel sand is used as the sample feature value and a label is set. Both the top and bottom surfaces of the river channel belong to the river channel envelope, and the envelope is labeled as 1, while other areas are labeled as 0.

7. The method for river channel simulation and deep learning based on the framework construction of river sandy strata according to claim 1, characterized in that, In step 4, using forward samples as input, a deep belief network is used to continuously train the hyperparameters to obtain a training model with smaller errors.

8. The method for river channel simulation and deep learning based on the framework construction of river sandy strata as described in claim 7, characterized in that, In step 4, the trained model is applied to actual 3D seismic data to predict the envelope of the river channels contained in the actual seismic data.

9. The method for river channel simulation and deep learning based on the framework construction of river sandy strata as described in claim 8, characterized in that, In step 4, if the predicted envelope is not complete, the river channel simulation formula from step 2 is used to repair it.

10. The method for river channel simulation and deep learning based on the framework construction of river sandy strata according to claim 1, characterized in that, In step 5, the target section standard layer is loaded, and based on the stratigraphic contact relationship, a stratigraphic framework conforming to the channel morphology is formed based on the standard layer and the envelope surface of the channel sand body.

Citation Information

Patent Citations

  • Identification methods for channel sandstone reservoirs

    CN104502969B

  • A method for fine identification of river channel sand bodies

    CN105372703B

  • Riverway sand modeling method and device and readable storage medium

    CN111209674A

  • Quantitative identification and prediction method of favorable reservoir of tuffaceous sandstone

    CN106842301A

  • Shallow water delta finger-shaped sand dam reservoir configuration identification method and device, medium and equipment

    CN113534288A