Modeling Method, Device, Equipment and Storage Medium for Continuous Variables in Meandering Rivers

Abstract

An embodiment of the present invention provides a method, device, equipment and storage medium for continuous variable modeling of meandering rivers. The method includes: establishing a training image of continuous variables of meandering rivers according to simulation grids; scanning the training image of continuous variables of meandering rivers through a data template to obtain a plurality of training patterns; for each empty grid node in the simulation grids, scanning the grid node through the data template, and under the constraint of soft data trend, obtaining the training pattern of the grid node, determining the training pattern corresponding to the training pattern of the grid node among the plurality of training patterns of the meandering river training image, and referring the determined training pattern to the grid node to complete the simulation of the grid node. This solution can characterize complex spatial structures and geometric forms, and can better reflect the geological sedimentation law of meandering rivers; it is realized under the constraint of soft data trend, which is conducive to overcoming the disadvantage of poor channel continuity of multi-point geostatistics without trend constraint.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas exploration, and particularly relates to a modeling method, device, equipment and storage medium for continuous variables in meandering rivers. Background Art

[0002] With the continuous improvement of oil and gas exploration and development levels and the continuous expansion of research fields, the spatial structure and geometric shape of exploration target geological bodies are becoming increasingly complex, and the requirements for geological body modeling are also getting higher and higher. Taking meandering rivers as an example, meandering rivers have a high degree of planar curvature, complex internal structures, continuous lateral migration of river channels, formation of point bars on the concave banks of river channels, and development of lateral accretion beds within point bars during flood events or diagenesis, dividing point bars into multiple crescent-shaped lateral accretion bodies, resulting in complex physical property spatial distributions and remaining oil distribution patterns and great modeling difficulties.

[0003] Traditional two-point geostatistical modeling methods are based on variograms and can only express the correlation between two points, making it difficult to characterize complex spatial structures and geometric shapes (such as meandering river channels). A new pixel-based multi-point geostatistical stochastic simulation method has been proposed in the prior art. Multi-point geostatistics uses a "training image" to replace the variogram to express the correlation between multiple points. The training image is a digital image that can express the actual reservoir structure, geometric shape, and distribution pattern, and it can reflect prior geological concepts and sedimentation patterns. The conditional probability distribution function is obtained by scanning the training image with given data events. Subsequently, foreign researchers have proposed many new multi-point algorithms or improved algorithms, such as the Snesim, Simpat, Growth-sim algorithms, etc. However, most of these algorithms can only be used for discrete variable simulations and are not suitable for continuous variables.

[0004] In addition, during the process of oil and gas geological research, deterministic trend surfaces or volume data can be obtained through seismic attributes or inversion. This data can be used for reservoir modeling, greatly improving the certainty of modeling, which is a common idea in conventional modeling methods, but it is less used in multi-point geological modeling.

[0005] Therefore, there is still a lack of a method in the prior art that can characterize complex spatial structures and geometric shapes and is suitable for multi-point geological modeling of continuous variables. Summary of the Invention

[0006] The embodiments of the present invention provide a modeling method for continuous variables in meandering rivers to solve the technical problems in the prior art of multi-point geological modeling of continuous variables, such as the inability to characterize complex spatial structures and geometric shapes and poor continuity. The method includes:

[0007] Establishing a simulation grid based on well logging data;

[0008] Establishing a continuous meandering river training image based on geological understanding and sedimentation laws;

[0009] Set multiple data filters. The data filters are a series of weight values related to a data template. Scan the training image through the data template to obtain multiple training patterns of the training image. Each training pattern is the sum of the score values of multiple data filters each time the data template scans.

[0010] For each empty grid node in the simulation grid, scan the grid node through the data template, and under the constraint of soft data trend, obtain the training pattern of the grid node. Determine the training pattern corresponding to the training pattern of the grid node among the multiple training patterns of the training image, and reference the determined training pattern to the grid node to complete the simulation of the grid node.

[0011] An embodiment of the present invention also provides a meandering river continuous variable modeling device to solve the technical problems in the prior art that multi-point geological modeling of continuous variables cannot represent complex spatial structures and geometric forms and has poor continuity. The device includes:

[0012] A simulation grid establishment module for establishing a simulation grid according to well logging data;

[0013] A training image establishment module for establishing a meandering river continuous variable training image according to geological understanding and sedimentation rules;

[0014] A training image scanning module for setting multiple data filters. The data filters are a series of weight values related to a data template. Scan the training image through the data template to obtain multiple training patterns of the training image. Each training pattern is the sum of the score values of multiple data filters each time the data template scans.

[0015] A grid node simulation module for, for each empty grid node in the simulation grid, scanning the grid node through the data template, and under the constraint of soft data trend, obtaining the training pattern of the grid node. Determining the training pattern corresponding to the training pattern of the grid node among the multiple training patterns of the training image, and referencing the determined training pattern to the grid node to complete the simulation of the grid node.

[0016] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned meandering river continuous variable modeling methods to solve the technical problems in the prior art that multi-point geological modeling of continuous variables cannot represent complex spatial structures and geometric forms and has poor continuity.

[0017] An embodiment of the present invention also provides a computer-readable storage medium storing a computer program for executing the above-mentioned meandering river continuous variable modeling method, so as to solve the technical problems in the prior art that the multi-point geological modeling of continuous variables cannot represent complex spatial structures and geometric forms and has poor continuity.

[0018] In the embodiment of the present invention, a simulation grid is established based on well logging data, that is, well point data is sampled into the grid as hard data of the simulation grid. Then, for the empty grid nodes in the simulation grid, a data template is formed by setting multiple data filters. Multiple training patterns are obtained by scanning the meandering river continuous variable training image with the data template, and then the empty grid node is scanned with the data template. Under the constraint of the soft data trend, the training pattern of the empty grid node is obtained. Finally, the training pattern corresponding to the training pattern of the empty grid node among the multiple training patterns is referenced to the empty grid node, that is, the simulation of the grid node is completed. The next grid node is simulated according to a random path until all empty grid nodes are simulated. Since the simulation grid is established based on well logging data, and then the empty grid nodes in the simulation grid are simulated by scanning the training image with the data template to obtain training patterns; since the training image considers not only the distance of spatial variables but also the positional relationship, it is beneficial to establish a multi-point geological statistical model of continuous variables under the guidance of geological patterns. Compared with traditional geostatistical algorithms, it can represent complex spatial structures and geometric forms and better reflect the geological laws of meandering rivers; at the same time, during the process of simulating nodes, it is realized under the constraint of the soft data trend, which is beneficial to overcoming the disadvantage of poor channel continuity of multi-point geostatistics without trend constraint. Description of the Drawings

[0019] The drawings described herein are used to provide a further understanding of the present invention, form a part of this application, and do not limit the present invention. In the drawings:

[0020] Figure 1 is a flowchart of a meandering river continuous variable modeling method provided by an embodiment of the present invention;

[0021] Figure 2 is a schematic diagram of original well logging data provided by an embodiment of the present invention;

[0022] Figure 3 is a schematic diagram of a meandering river porosity training image provided by an embodiment of the present invention;

[0023] Figure 4 is a schematic diagram of the process of obtaining filter scores by a two-dimensional filter of well logging data provided by an embodiment of the present invention;

[0024] Figure 5It is a schematic diagram of a two - dimensional filter two - step division method provided by an embodiment of the present invention;

[0025] Figure 6 It is a schematic diagram of a multi - point simulation result without trend constraint provided by an embodiment of the present invention;

[0026] Figure 7 It is a schematic diagram of a soft data - constrained body provided by an embodiment of the present invention;

[0027] Figure 8 It is a schematic diagram of a multi - point simulation result under trend constraint provided by an embodiment of the present invention;

[0028] Figure 9 It is a schematic diagram of a two - point geostatistical simulation result provided by an embodiment of the present invention;

[0029] Figure 10 It is a flowchart of implementing the above - mentioned meandering river continuous variable modeling method provided by an embodiment of the present invention;

[0030] Figure 11 It is a structural block diagram of a computer device provided by an embodiment of the present invention;

[0031] Figure 12 It is a structural block diagram of a meandering river continuous variable modeling device provided by an embodiment of the present invention. Detailed implementation manners

[0032] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with the implementation manners and the accompanying drawings. Herein, the illustrative implementation manners of the present invention and their descriptions are used to explain the present invention, but do not limit the present invention.

[0033] In an embodiment of the present invention, a meandering river continuous variable modeling method is provided. As Figure 1 shown, the method includes:

[0034] Step 102: Establish a simulation grid according to well logging data;

[0035] Step 104: Establish a meandering river continuous variable training image according to geological understanding and sedimentation laws;

[0036] Step 106: Set multiple data filters. The data filter is a series of weight values related to a data template. Scan the meandering river training image through the data template to obtain multiple training patterns of the meandering river training image. Among them, each training pattern is the sum of the score values of multiple data filters each time the data template scans.

[0037] Step 108: For each empty grid node in the simulated grid, scan the grid node through the data template. Under the constraint of the soft data trend, obtain the training pattern of the grid node. Determine the training pattern corresponding to the training pattern of the grid node among the multiple training patterns of the meandering river training image, and reference the determined training pattern to the grid node to complete the simulation of the grid node.

[0038] As can be seen from Figure 1 the process shown, in the embodiment of the present invention, a simulated grid is established based on well logging data, that is, well point data is sampled into the grid as the hard data of the simulated grid. Then, for the empty grid nodes in the simulated grid, a data template is formed by setting multiple data filters, and multiple training patterns are obtained by scanning the meandering river training image with the data template. Then, the empty grid node is scanned through the data template. Under the constraint of the soft data trend, the training pattern of the empty grid node is obtained. Finally, the training pattern corresponding to the training pattern of the empty grid node among the multiple training patterns is referenced to the empty grid node, that is, the simulation of the grid node is completed, and the next grid node is simulated according to a random path until the simulation of all empty grid nodes is completed. Since the simulated grid is established based on well logging data, and then the empty grid nodes in the simulated grid are simulated by scanning the training image with the data template; since the training image considers not only the distance of spatial variables but also the positional relationship, it is beneficial to establish a multiple-point geostatistical model of continuous variables under the guidance of geological patterns. Compared with traditional geostatistical algorithms, it can represent complex spatial structures and geometric forms and can better reflect the geological laws of meandering rivers; at the same time, during the process of simulating nodes, it is realized under the constraint of the soft data trend, which is beneficial to overcoming the disadvantage of poor channel continuity of multi-point geostatistics without trend constraint.

[0039] Specifically, as Figure 10 shown, before implementing the meandering river continuous variable modeling, modeling data needs to be prepared, that is, the measured well data can be marked on the nearest grid node to establish a simulated grid. For example, according to the size of the work area, a suitable grid step size is selected to establish a simulated grid, and the well point data is sampled into the grid through grid coarsening as the hard data of the simulated grid, as Figure 2 shown.

[0040] Specifically, as Figure 10As shown, after establishing the simulation grid, a continuous meandering river training image can be established based on the simulation grid. The present application does not specifically limit the process of establishing the continuous meandering river training image, and it can be implemented by existing methods. For example, a continuous meandering river training image can be established through outcrop modeling, modern sediment investigation, fine geological research in dense well pattern areas, core observation and analysis, seismic information mining, etc. The continuous meandering river training image established by the method of converting modern sediment images is as Figure 3 shown.

[0041] Specifically, when implementing, after establishing the continuous meandering river training image, in order to enable the modeling of continuous variables of the meandering river to reflect the geological laws of the meandering river, in this embodiment, during the process of simulating grid nodes, a data template using multiple data filters is proposed to scan the meandering river training image and grid nodes. For example, multiple data filters are set. The multiple data filters are a series of weight values related to the data template, and the size of the data filter is the same as the size of the data template, which is convenient for applying to the training mode centered on u 0 . Each node of the data filter corresponds to an offset vector h 0 relative to the center u i of the data template, that is, the data filter can be defined as:

[0042] {f(h i ); i = 1…J}

[0043] T J = {u 0 ; h i , i = 1…J}

[0044] where u 0 is the central node of the data template, f(h i ) is the data filter, i is the grid variable, J is the maximum number of grids of the data filter, T J is the search template, h i = (x, y, z) i is the offset vector, and x, y, z are grid coordinates, integers;

[0045] Furthermore, by scanning the meandering river training image with the data template, multiple training modes of the meandering river training image are obtained, where each training mode is the sum of the score values of multiple data filters each time the data template scans;

[0046] Specifically, the data template can define the training mode by scanning the training image, and each training mode is the sum of the score values of multiple data filters each time the data template scans, as Figure 4As shown, during the scanning process, the data filter extracts the conditional data events of the training image. Then, the sum of the products of the weight values of the data filter and the point values of the conditional data events of the training image is the score value of the data filter. Specifically, the score value S of the data filter can be calculated by the following formula T (u):

[0047]

[0048] where S T (u) is the filter score, and pat(u + h i ) is the pattern grid node value; J = n x ×n y ×n z is the maximum grid data, and n x , n y , n z are the number of grids in three directions.

[0049] Since a single data filter is not sufficient to capture the information carried by the training pattern, a column of multiple data filters needs to be set to obtain various information contained in the training pattern. These data filters form a vector composed of the sum of the scores of each training pattern As Figure 5 shown,[[]]END]]

[0050]

[0051] where k = 1, …, F, and k is the number of filters;

[0052] In this way, the dimension of the training pattern is reduced from J = n x ×n y ×n z to F. For continuous training images, the F filters can be directly used to form the training pattern.

[0053] Specifically, in each direction of the data template, any combination of an average filter, a gradient filter, and a curvature filter can be set.

[0054] Specifically, in the X / Y / Z three directions, the data filter grid settings should be consistent with the data template. Assume that the number of nodes in the X direction of the data template is n i (n i is odd),

[0055] m i =(n i -1) / 2

[0056] α i =-m i , …, m

[0057] Average filter:

[0058] Gradient filter:

[0059] Curvature filter:

[0060] m i is half of the number of nodes in a certain direction of the data template, and α i is the distance from the center point of the template; through permutation and combination, for two-dimensional data, there are 2 or 3 combinations of data filters, that is, 6 filter selections, and for three-dimensional data, there are 9 filter selections.

[0061] In specific implementation, for each empty grid node in the simulation grid, the process of the data template scanning this grid node to obtain the training mode of this grid node is similar to the process of the data template scanning the training image to obtain the training mode, which will not be elaborated again. Furthermore, the training mode corresponding to the training mode of this grid node can be referenced into this grid node, that is, the data of the training mode corresponding to the training mode of this grid node is filled into this grid node as known data points to complete the simulation of this grid node, and then the next grid node is simulated according to a random path until all nodes are simulated.

[0062] In specific implementation, in order to improve the matching of the training mode, in this embodiment, determining the training mode corresponding to the training mode of this grid node among multiple training modes of the meandering river training image includes:

[0063] By dividing the score value space of the training mode, classifying multiple training modes of the meandering river training image, and the training modes of the same class form a mode prototype, obtaining multiple mode prototypes, where the mode prototype is the average of all training modes within the mode prototype;

[0064] Determining the mode prototype corresponding to the training mode of this grid node;

[0065] Determining any one training mode in the determined mode prototype as the training mode corresponding to the training mode of this grid node.

[0066] Specifically, similar training modes have similar F·K scores. Therefore, as Figure 10 shown, the training modes can be classified by dividing the score value space, and similar training modes are classified into one class, that is, the mode prototype. The mode prototype can be defined as the average of all training modes falling into a certain training mode class. The mode prototype and the filter data template have the same size.

[0067] For continuous training images, the prototype value of a prototype model can be calculated through the data template T J to calculate.

[0068]

[0069] prot(h i ) is the prototype value of a certain classification, and h i is the i-th offset in the data template T J . c is the number of training images in the prototype model classification, and u j (i = 1,..., c) is the central grid of a specific training pattern.

[0070] Specifically, for example, a two-step method can be used for training pattern classification. As Figure 5 shown, the maximum CPU efficiency can be obtained.

[0071] First, use a fast classification algorithm to divide all training patterns into some coarser pattern sets, called parent categories (i.e., pattern prototypes). Each parent category represents its own pattern prototype. A parent category may include many training patterns and needs to be further refined. A similar classification method can be used to further divide the parent category into multiple subcategories. Each subcategory can have its own pattern prototype.

[0072] In specific implementation, when determining the pattern prototype corresponding to the training pattern of the grid node, a minimum distance algorithm can be used. For example, calculate the distance between the training pattern of the grid node and each pattern prototype; determine the pattern prototype with the minimum distance as the pattern prototype corresponding to the training pattern of the grid node.

[0073] Specifically, the pattern prototype corresponding to the training pattern of the grid node can be calculated through the following minimum distance algorithm formula:

[0074]

[0075] L(dev, prot) is the distance function; dev(u 0 +h i ) is the node value of the data template in the area to be simulated at the position of u 0 +h i ; prot(u 0 ) is the score value in a certain classification prototype model; W is the weight value of each grid; J is the number of grids of the data template; u 0 is the central grid of the pattern prototype; h i is the offset of u from the central grid. w mThe determination of values is divided into four types: m = 1, original hard data; m = 2, simulated data; m = 3, reference data; m = 4, soft data; w m are the weight values of the four types; N m is the number of grids of this type in the data event; W 1 +W 2 +W 3 +W 4 = 1, and W 1 > W 2 > W 3 ,W 4

[0076] Specifically, after classifying and creating a pattern prototype (parent category, subcategory) in the training mode, the simulation implementation can begin. Access the simulation grid G along a random path. At each empty grid node u, the data template T and a data filter template of the same size are used to extract the conditional data event dev(u), obtain the training mode of this grid node, and then find the pattern prototype that is closest to the training mode of this grid node through the minimum distance algorithm. If this pattern prototype still has subcategories, then randomly select a training mode from the subcategories and reference it to the grid node of the simulation grid G. The referenced part is frozen as hard data, and then the simulation of the next grid node is carried out.

[0077] Specifically, by setting a series of multiple linear data filters and classifying the training mode within the range of the linear data filters, and then performing the training mode matching method, the purpose of reducing the dimension can be achieved.

[0078] During specific implementation, in the process of obtaining the training mode of this grid node by scanning this grid node through the data template, in order to improve the continuity of the simulation grid node and further improve the continuity of the meandering river continuous variable modeling, this embodiment proposes to obtain the training mode of this grid node under the soft data trend constraint. Specifically, as Figure 10 shown, when using the data template to scan this grid node to extract the conditional data event, use the data template to scan the soft data to extract the conditional data event, and then use the conditional data event extracted from the soft data to replace the unmarked grids within the range of the data template until all the unmarked grids within the data template range centered on u 0 are filled, then obtain the training mode of this grid node, and then reference the training mode corresponding to the training mode of this grid node to this grid node, that is, fill the data of the training mode corresponding to the training mode of this grid node into this grid node as known data points to complete the simulation of this grid node, and then simulate the next grid node according to the random path until all nodes are simulated.

[0079] Specifically, if the grid node simulation process is not completed under the soft data trend constraint, the obtained multi-point simulation result without trend constraint is as follows Figure 6 shown. The soft data is taken as an example as shown in Figure 7 shown. The grid node simulation process is completed under the soft data trend constraint, and the obtained multi-point simulation result under trend constraint is as follows Figure 8 shown. By comparison, it can be seen that without the soft data trend constraint, relying solely on the geological patterns contained in the training image for pattern recognition and matching has multiple solutions, resulting in poor continuity of meandering channels and even locally inconsistent geological results. Under the soft data trend constraint, the smoothness of the simulation results is improved, the channel morphology is natural, the continuity is good, and it well conforms to geological understanding.

[0080] Specifically, the soft data volume can be a spatial trend volume.

[0081] Specifically, for each empty grid node in the simulation grid, by using a data template to scan the training image and the grid node, and then filling the data of the training pattern of the training image corresponding to the training pattern of the grid node into the grid node as known data points to complete the simulation of the grid node, and then simulating the next grid node according to a random path until all nodes are simulated. The grid node simulation process is somewhat similar to selecting pictures from a pile of similar pictures to complete a jigsaw puzzle. The existing Snesim method stores all training patterns in a search tree, while the above-mentioned meandering river continuous variable modeling method only stores the central grid points in memory. Therefore, the requirement for random access memory can be greatly reduced.

[0082] Specifically, the meandering river porosity model established by using the above-mentioned meandering river continuous variable modeling method is as follows Figure 8 shown. The meandering river porosity model established by using the existing two-point geostatistical method is as follows Figure 9 shown. By comparison, it can be seen that the traditional two-point geostatistical method, based on the variogram, can only consider the correlation between two spatial points and is difficult to accurately represent complex spatial structures and reproduce the geometric shapes of complex targets. The above-mentioned meandering river continuous variable modeling method uses the training image instead of the variogram. Its advantages are mainly reflected in that it not only considers the distance of spatial variables but also considers their positional relationships, can express complex geometric bodies, can well show the positional relationships and morphological characteristics of meandering river point bars, crevasse splays, channels and lateral accretion bodies, and characterizes the changes in porosity of each microfacies unit.

[0083] In this embodiment, a computer device is provided, as shown in Figure 11As shown, it includes a memory 1102, a processor 1104, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned meandering river continuous variable modeling methods.

[0084] Specifically, the computer device can be a computer terminal, a server, or a similar computing device.

[0085] In this embodiment, a computer-readable storage medium is provided, which stores a computer program for executing any of the above-mentioned meandering river continuous variable modeling methods.

[0086] Specifically, the computer-readable storage medium includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer-readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassette tapes, magnetic disk storage, or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device. As defined herein, computer-readable storage media do not include transitory media such as modulated data signals and carrier waves.

[0087] Based on the same inventive concept, an apparatus for meandering river continuous variable modeling is also provided in an embodiment of the present invention, as described in the following embodiments. Since the principle of the apparatus for meandering river continuous variable modeling to solve problems is similar to that of the meandering river continuous variable modeling method, the implementation of the apparatus for meandering river continuous variable modeling can refer to the implementation of the meandering river continuous variable modeling method, and the repeated parts will not be described again. Hereinafter, the term "unit" or "module" can be a combination of software and / or hardware that can implement a predetermined function. Although the apparatuses described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0088] Figure 12 is a structural block diagram of the apparatus for meandering river continuous variable modeling according to an embodiment of the present invention, as Figure 12 shown, the apparatus includes:

[0089] An analog grid establishment module 1202, configured to establish an analog grid according to well logging data;

[0090] A training image building module 1204, configured to build a meandering river continuous variable training image according to the simulation grid;

[0091] A training image scanning module 1206, configured to set a plurality of data filters, where the data filter is a series of weight values related to a data template, and scan the training image through the data template to obtain a plurality of training patterns of the training image, where each training pattern is the sum of the score values of the plurality of data filters each time the data template scans;

[0092] A grid node simulation module 1208, configured to, for each empty grid node in the simulation grid, scan the grid node through the data template, and under the constraint of the soft data trend, obtain the training pattern of the grid node, determine the training pattern corresponding to the training pattern of the grid node among the plurality of training patterns of the training image, and reference the determined training pattern to the grid node to complete the simulation of the grid node.

[0093] In one embodiment, the training image scanning module is further configured to, when using the data template to scan the grid node to extract conditional data events, use the data template to scan the soft data to extract conditional data events, and replace the unmarked grids within the range of the data template with the conditional data events extracted from the soft data to obtain the training pattern of the grid node.

[0094] In one embodiment, the grid node simulation module includes:

[0095] A classification unit, configured to classify the plurality of training patterns of the meandering river training image by dividing the score value space of the training patterns, where the training patterns of the same class form a pattern prototype, and obtain a plurality of pattern prototypes, where the pattern prototype is the average of all training patterns within the pattern prototype;

[0096] A pattern prototype determination unit, configured to determine the pattern prototype corresponding to the training pattern of the grid node;

[0097] A training pattern determination unit, configured to determine any one of the training patterns in the determined pattern prototype as the training pattern corresponding to the training pattern of the grid node.

[0098] In one embodiment, the pattern prototype determination unit is specifically configured to calculate the distance between the training pattern of the grid node and each pattern prototype; and determine the pattern prototype with the smallest distance as the pattern prototype corresponding to the training pattern of the grid node.

[0099] In one embodiment, the training image scanning module is further configured to set any combination of an average filter, a gradient filter, and a curvature filter in each direction of the data template.

[0100] The embodiments of the present invention achieve the following technical effects: A simulation grid is established based on well logging data, that is, well point data is sampled into the grid as hard data of the simulation grid. Then, for the empty grid nodes in the simulation grid, a data template is formed by setting multiple data filters, and multiple training patterns of the meandering river training image are scanned by the data template. Then, the empty grid node is scanned by the data template, and under the constraint of the soft data trend, the training pattern of the empty grid node is obtained. Finally, the training pattern corresponding to the training pattern of the empty grid node among multiple training patterns is referenced to the empty grid node, that is, the simulation of the grid node is completed, and the next grid node is simulated according to a random path until the simulation of all empty grid nodes is completed. Since the simulation grid is established based on well logging data, and then the empty grid nodes in the simulation grid are simulated by scanning the training image with the data template to obtain the training pattern; since the training image considers not only the distance of spatial variables but also the positional relationship, it is beneficial to establish a multiple-point geostatistical model of continuous variables under the guidance of geological patterns. Compared with traditional geostatistical algorithms, it can represent complex spatial structures and geometric forms and can better reflect the geological laws of meandering rivers; at the same time, during the process of simulating nodes, it is realized under the constraint of the soft data trend, which is beneficial to overcoming the disadvantage of poor channel continuity of multi-point geostatistics without trend constraint.

[0101] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the embodiments of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device, so that they can be stored in a storage device and executed by the computing device. And in some cases, the steps shown or described can be executed in a different order from here, or they can be separately made into individual integrated circuit modules, or multiple modules or steps among them can be made into a single integrated circuit module to be implemented. In this way, the embodiments of the present invention are not limited to any specific combination of hardware and software.

[0102] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the embodiments of the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for modeling continuous variables in meandering rivers, characterized in that, it includes: establishing a simulation grid according to well logging data; establishing a training image of continuous variables in meandering rivers according to geological understanding and sedimentation rules; setting multiple data filters, where the data filter is a series of weight values related to a data template, scanning the training image through the data template to obtain multiple training patterns of the training image, and each training pattern is the sum of the score values of multiple data filters during each scan of the data template; for each empty grid node in the simulation grid, scanning the grid node through the data template, obtaining the training pattern of the grid node under the constraint of soft data trend, determining the training pattern corresponding to the training pattern of the grid node among the multiple training patterns of the training image, and referring the determined training pattern to the grid node to complete the simulation of the grid node; scanning the grid node through the data template to obtain the training pattern of the grid node under the constraint of soft data trend, including: when using the data template to scan the grid node to extract conditional data events, using the data template to scan soft data to extract conditional data events, and replacing the unmarked grids within the range of the data template with the conditional data events extracted from the soft data to obtain the training pattern of the grid node; determining the training pattern corresponding to the training pattern of the grid node among the multiple training patterns of the training image, including: classifying the multiple training patterns of the training image by dividing the score value space of the training pattern, and the training patterns of the same class form a pattern prototype, obtaining multiple pattern prototypes, where the pattern prototype is the average of all training patterns within the pattern prototype; determining the pattern prototype corresponding to the training pattern of the grid node; and determining any one training pattern in the determined pattern prototype as the training pattern corresponding to the training pattern of the grid node.

2. The method for modeling continuous variables in meandering rivers according to claim 1, characterized in that, determining the pattern prototype corresponding to the training pattern of the grid node includes: calculating the distance between the training pattern of the grid node and each pattern prototype; determining the pattern prototype with the minimum distance as the pattern prototype corresponding to the training pattern of the grid node.

3. The method for modeling continuous variables in meandering rivers according to any one of claims 1 to 2, characterized in that, setting multiple data filters includes: setting any combination of an average filter, a gradient filter, and a curvature filter in each direction of the data template.

4. A device for modeling continuous variables in meandering rivers, characterized in that, it includes: a simulation grid establishment module for establishing a simulation grid according to well logging data; a training image establishment module for establishing a training image of continuous variables in meandering rivers according to geological understanding and sedimentation rules; A training image scanning module, configured to set a plurality of data filters, where the data filters are a series of weight values related to a data template, and to scan the training image through the data template to obtain a plurality of training patterns of the training image, where each training pattern is the sum of the score values of the plurality of data filters each time the data template scans; A grid node simulation module, configured to, for each empty grid node in the simulation grid, scan the grid node through the data template, and under the constraint of the soft data trend, obtain the training pattern of the grid node, determine the training pattern corresponding to the training pattern of the grid node among the plurality of training patterns of the training image, and reference the determined training pattern to the grid node to complete the simulation of the grid node; The training image scanning module is further configured to, when using the data template to scan the grid node to extract conditional data events, use the data template to scan the soft data to extract conditional data events, and replace the unmarked grids within the range of the data template with the conditional data events extracted from the soft data to obtain the training pattern of the grid node; The grid node simulation module includes: a classification unit, configured to classify the plurality of training patterns of the training image by dividing the score value space of the training patterns, where the training patterns of the same class form a pattern prototype, and obtain a plurality of pattern prototypes, where the pattern prototype is the average of all training patterns within the pattern prototype; a pattern prototype determination unit, configured to determine the pattern prototype corresponding to the training pattern of the grid node; and a training pattern determination unit, configured to determine any one of the training patterns in the determined pattern prototype as the training pattern corresponding to the training pattern of the grid node.

5. The meandering river continuous variable modeling device according to claim 4, wherein, The pattern prototype determination unit is specifically configured to calculate the distance between the training pattern of the grid node and each pattern prototype; and determine the pattern prototype with the minimum distance as the pattern prototype corresponding to the training pattern of the grid node.

6. The meandering river continuous variable modeling device according to any one of claims 4 to 5, wherein, The training image scanning module is further configured to set any combination of an average filter, a gradient filter, and a curvature filter in each direction of the data template.

7. A computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein, When the processor executes the computer program, it implements the meandering river continuous variable modeling method according to any one of claims 1 to 3.

8. A computer-readable storage medium, wherein, The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the meandering river continuous variable modeling method according to any one of claims 1 to 3.

Images