Multi-point geostatistical stratigraphic model construction method coupling data driving and geological constraint

Abstract

The invention discloses a multi-point geostatistical stratigraphic model construction method coupling data driving and geological constraint, and relates to the field of stratigraphic modeling and geostatistical analysis, and the method comprises the steps: obtaining and preprocessing original drilling data, and generating a modeling data set; the invention discloses a lithofacies simulation method based on fusion data driving and geological constraint. The lithofacies simulation method comprises the following steps: constructing a three-dimensional training image by adopting nearest neighbor resampling and indicating Kriging interpolation; carrying out geological constraint correction by utilizing a vertical proportion curve and a transition probability function, and carrying out multipoint geostatistics stochastic simulation by taking drilling data as hard condition data, combining with the training image and adopting an SNESIM algorithm to generate a three-dimensional stratum model; and finally, carrying out coordinate and depth back calculation to realize correspondence between a model result and an actual geographic coordinate system. By introducing a coupling mechanism of geological constraint and data driving, the geological rationality and simulation precision of the model are improved, manual intervention and random errors are reduced, and a three-dimensional space structure of a complex stratum can be efficiently and reliably reconstructed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of stratum modeling and geostatistical analysis, in particular to a method for constructing a multiple-point geostatistics stratum model coupled with data driving and geological constraints. BACKGROUND

[0002] Three-dimensional stratum modeling is a key technology for constructing an underground structure model, and is widely used in the fields of geology, hydrology and engineering. At present, the main methods include deterministic modeling and stochastic modeling. Deterministic modeling relies on a large amount of accurate data, and generates a model through manual interpretation and interpolation, but it is easy to produce an over-smooth result when the data is sparse, and it is highly dependent on expert experience, low in efficiency and strong in subjectivity. In order to overcome these limitations, the multiple-point geostatistics method is introduced, which can better reproduce the connectivity and heterogeneity of the stratum by training images to depict the spatial structure and distribution of the geological body, and can effectively reconstruct the complex geological structure under the condition of limited data.

[0003] However, the existing training image construction method and the multiple-point geostatistics modeling process still have the following limitations: (1) the traditional training image has its own limitations. Manual drawing and target simulation rely on expert experience and are highly subjective; the method based on seismic or prototype model is subject to the difficulty and high cost of data acquisition; the process simulation method is highly scientific, but has high computational complexity and is difficult to be directly applied to the work area with limited data. (2) The multiple-point geostatistics modeling method lacks generalizability and landing. The existing method often focuses on the algorithm itself, and lacks systematic integration of the whole process such as data preprocessing, spatial coordinate normalization, hard data mapping, search tree construction, simulation and post-processing, which is difficult to be directly applied to actual engineering modeling.

[0004] In summary, in view of the complex scientific problem of improving the reliability of three-dimensional training images under the constraints of low precision and low data volume, in order to realize the accurate characterization of the three-dimensional soil-groundwater system in the regional complex geological environment, it is urgent to propose a multiple-point geostatistics three-dimensional stratum model construction method which is driven by limited geological data and integrates geological constraint information, to realize a high-precision, objective and engineering usable modeling process of stratum spatial structure. This has important significance for improving the geological reasonableness, simulation accuracy and practical application value of the stratum model. SUMMARY

[0005] In order to solve the above-mentioned problems, the present application provides a method for constructing a multiple-point geostatistics stratum model coupled with data driving and geological constraints, which optimizes the acquisition method of three-dimensional training images, overcomes the related limitations of traditional training images, realizes the whole process automation of stratum modeling from original data preprocessing to model output, can effectively reproduce the spatial heterogeneity and connectivity of complex stratum, and realizes the high-precision, objective and geological reasonableness reconstruction of three-dimensional stratum structure under the condition of limited geological data.

[0006] To achieve the above object, the application provides a method for constructing a multi-point geostatistical stratum model by coupling data driving and geological constraints, comprising the following steps: S1: obtaining original drilling data of a modeling area and performing pretreatment to generate a modeling data set; S2: based on the modeling data set, adopting a facies simulation method that fuses data driving and geological constraints to construct a three-dimensional training image; S3: taking the modeling data set as hard condition data and based on the three-dimensional training image constructed in S2, using an SNESIM algorithm to perform multi-point geostatistical random simulation to generate a three-dimensional stratum model; S4: performing post-processing on the three-dimensional stratum model to realize coordinate and depth back calculation.

[0007] Preferably, S1 specifically comprises: S11: extracting and organizing original drilling stratum data in the modeling area, wherein the original drilling stratum data comprises drilling number, latitude and longitude coordinates, depth and lithology layering information; S12: jointly calculating the original drilling latitude and longitude coordinates and the boundary point coordinates of the modeling area to obtain relative coordinates of each drilling in the coordinate system of the modeling area; S13: performing normalization processing on the depth information of the drilling data to unify the depth scale; S14: organizing the drilling number, relative coordinates, unified depth value and lithology layering information into a structured spreadsheet data set to obtain the modeling data set.

[0008] Preferably, S2 specifically comprises: S21: based on the modeling data set, using a nearest neighbor interpolation method to resample all drilling data according to a unified vertical resolution to generate a standardized spatial point set, wherein the spatial point set comprises X, Y and Z direction coordinate values and corresponding lithology categories; S22: regarding each type of facies as a categorical variable and converting it into a binary indicator variable; for each type of facies, independently performing indicator kriging interpolation to establish a spatial variation function model of the occurrence probability of each facies in a three-dimensional space to generate a three-dimensional probability body; S23: based on geological constraint parameters, performing constraint correction on the three-dimensional probability body; the geological constraint parameters comprise a vertical proportion curve and a transition probability function; S24: based on the corrected three-dimensional probability body, constructing a three-dimensional training image that fuses geological knowledge and data driving features.

[0009] Preferably, the vertical proportion curve is expressed as: ; wherein, the proportion of facies classes at vertical positions the number of samples of classes at depths the total number of samples at the depth layer.

[0010] Preferably, the transition probability function is expressed as: ; i , j h

[0011] Preferably, S3 specifically comprises: S31: taking the modeling dataset as hard condition data, mapping the nearest network grid points on the simulation rule grid, selecting a custom data template to scan the three-dimensional training image generated by S2, and establishing a multi-point conditional probability search tree; S32: sequentially selecting the nodes to be estimated by adopting the sequential simulation path, randomly extracting the simulation values according to the conditional probability distribution of the search tree and updating the conditional dataset; S33: repeating S32 until all nodes to be estimated are simulated; S34: outputting the simulation results as a three-dimensional stratigraphic model.

[0012] Preferably, the establishment of the multi-point conditional probability search tree in S31 specifically comprises: S311: constructing equidistant three-dimensional regular grids in the modeling area, the grid spacing is determined according to the drilling distribution density and the modeling resolution, mapping the drilling data in the normalized modeling dataset to the nearest grid node, and marking the mapped node as a hard data point, and marking the unmapped node as a node to be estimated; S312: customizing a multi-point data template according to the spatial scale characteristics of the geological body and the simulation accuracy requirement; S313: taking the custom template as a sliding window to perform full-space scanning on the three-dimensional training image constructed in S2, for each spatial configuration within the template, counting the frequency of each spatial configuration within the template in the training image, and forming a pair of conditional probabilities of multi-point spatial configuration and corresponding lithology class; S314: organizing the scanned spatial configurations in a tree-like manner according to the spatial position sequence of the nodes within the template, and constructing a multi-point conditional probability search tree, wherein the root node is an empty template, the child nodes correspond to the sampling points within the template in turn, and the leaf nodes store the corresponding lithology class conditional probability distribution.

[0013] Preferably, S4 specifically comprises:​​​​​​​​ S41: The relative coordinates of each node in the three-dimensional stratum model are inversely calculated into actual latitude and longitude coordinates according to the boundary point information of the original coordinate system of the modeling area; S42: The unified depth value is restored according to the surface elevation and depth reference of the original borehole to obtain a depth value corresponding to the actual stratum.

[0014] Therefore, the present application adopts the above-mentioned multi-point geostatistical stratum model construction method coupled with data driving and geological constraints. The user only needs to arrange the borehole data into the standard format required by the algorithm and input the modeling area information, and the whole process from data preprocessing, training image construction to three-dimensional stratum model generation can be automatically completed by the related script. The present application effectively simplifies the user's artificial definition and input process of the stratum structure by introducing geological constraint parameters and a multi-point spatial statistical model, and significantly reduces the subjective intervention and randomness in the model establishment process. The present application uses the geological constraint mechanism combining the vertical scale curve and the transition probability function to mine the statistical distribution characteristics of the stratum in the three-dimensional space from the limited original borehole data, realizes the spatial correlation constraint and reasonable assignment of the geological structure, and effectively overcomes the problems of difficult to maintain the continuity of the stratum structure and serious smoothing of the results in the traditional method. The present application improves the operation efficiency and spatial reconstruction accuracy of the model while ensuring the geological rationality, and provides an efficient, reliable and generalizable technical approach for three-dimensional stratum modeling in a complex geological environment.

[0015] The technical solutions of the present application will be further described in detail below with the aid of the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 FIG. 1 is a flowchart of the multi-point geostatistical stratum model construction method coupled with data driving and geological constraints in the present application; Figure 2 FIG. 2 is a schematic diagram of the borehole layering structure in the embodiment of the present application; Figure 3 FIG. 3 is a schematic diagram of the borehole attribute in the embodiment of the present application; Figure 4 FIG. 4 is a schematic diagram of the three-dimensional training image result in the embodiment of the present application; wherein, (a) is a three-dimensional training image result profile, and (b) is a three-dimensional training image result stereogram; Figure 5 FIG. 5 is a schematic diagram of the three-dimensional stratum structure result in the embodiment of the present application; wherein, (a) is a three-dimensional stratum structure result profile, and (b) is a three-dimensional stratum structure result stereogram. DETAILED DESCRIPTION

[0017] The following detailed description of embodiments of the application provided in the accompanying drawings is not intended to limit the scope of the application claimed, but merely represents selected embodiments of the application. Based upon the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work are within the scope of protection of the present application.

[0018] Unless otherwise defined, technical terms or scientific terms used in the present application shall have the common meaning understood by one of ordinary skill in the art to which the present application pertains.

[0019] The terms such as "comprising" or "including" or the like used in the present application mean that the elements before the term encompass the elements listed after the term, and do not exclude the possibility of also encompassing other elements. The terms "in", "out", "up", "down", etc. indicate the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application, and when the absolute position of the described object changes, the relative positional relationship may also change accordingly. In the present application, unless otherwise specified and limited, the term "attached" and the like should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or it can be integrated; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0020] Embodiments A method for constructing a multiple-point geostatistical stratigraphic model coupled with data driving and geological constraints, as shown in Figure 1 includes the following steps: S1: obtaining the original drilling data of the modeling area and preprocessing to generate the modeling data set; the modeling data set is a normalized modeling data set containing relative coordinates and unified depth information; S1 specifically includes: S11: extracting and organizing the original drilling stratigraphic data in the modeling area, the original drilling stratigraphic data including drilling number, latitude and longitude coordinates, depth and lithology layering information; S12: jointly calculating the original drilling latitude and longitude coordinates and the boundary point coordinates of the modeling area to obtain the relative coordinates of each drilling in the modeling area coordinate system; S12 jointly calculates the original drilling latitude and longitude coordinates and the boundary point coordinates of the modeling area, specifically including: S121: using a two-dimensional affine transformation model to establish a conversion model. The mathematical expression is: Where: (L, B) are latitude and longitude coordinates, (X, Y) are relative coordinates in the modeling area coordinate system, and (a, b, c, d, e, f) are transformation coefficients to be determined. The six coefficients together determine the different scaling and shear deformations in the two coordinate axis directions. S122: Based on the coordinates of the regional boundary points, the transformation parameters are solved. Using their known correspondence in the latitude and longitude coordinate system and the modeling coordinate system, eight equations can be formed. The transformation coefficient set (a, b, c, d, e, f) of the above linear model is solved by the least squares method. S123: In actual calculations, due to the small size of the study area and the small deformation, the model can be simplified by translation, using the reference point C as the origin and adopting an incremental approach: , The simplified transformation model is as follows: ,in:( , (x, y) represents the difference between latitude and longitude coordinates and the origin, and (x, y) represents the relative coordinates in the modeling region coordinate system. , , , The simplified transformation coefficients can be solved using the above eight equations. S124: Enables batch conversion of borehole coordinates for any borehole point. Its modeling coordinates are obtained through: , ; S13: Normalize the depth information of the borehole data to unify the depth scale; S14: Organize the borehole numbers, relative coordinates, uniform depth values, and lithological stratification information into a structured spreadsheet dataset to obtain a standardized modeling dataset.

[0021] S2: Based on a standardized modeling dataset, a lithofacies simulation method that integrates data-driven and geologically constrained approaches is used to construct three-dimensional training images; S2 specifically includes: S21: Based on the standardized modeling dataset, the nearest neighbor interpolation method is used to resample all borehole data at a uniform vertical resolution to generate a standardized spatial point set, which includes X, Y, and Z coordinate values ​​and corresponding lithology categories. S22: Treat all types of lithofacies as categorical variables and transform them into binary indicator variables; for each type of lithofacies, independently perform indicator kriging interpolation to establish a spatial variation function model of the probability of occurrence of each lithofacies in three-dimensional space, and generate a three-dimensional probability volume; S23: Based on geological constraint parameters, the three-dimensional probabilistic volume is constrained and corrected to reflect the macroscopic sequence structure and microscopic adjacency relationship of the strata; the geological constraint parameters include the vertical scale curve and the transition probability function; Let denote the proportion of facies class at vertical position , define the vertical proportion curve as , where is the sample number of class at depth , and is the total number of samples at this depth layer, the vertical proportion curve controls the macro-scale sequence trend of the stratum; Let the lithology class be i , j , the step size be h , and the transition probability be defined as , which is used to quantify the contact relationship between facies and control the smooth transition of small-scale adjacency relationship; According to the constraint results obtained by the vertical proportion curve and the transition probability function, the three-dimensional probability body is corrected to form a macro-scale stratum structure model with geological rationality.

[0022] S24: Based on the corrected three-dimensional probability body, a three-dimensional training image integrating geological knowledge and data-driven features is constructed to provide training samples for SNESIM modeling.

[0023] S3: Taking the normalized modeling dataset as hard condition data and based on the three-dimensional training image constructed in S2, a three-dimensional stratum model is generated by using the SNESIM algorithm for multiple-point geostatistical random simulation; S3 specifically includes: S31: Taking the normalized modeling dataset as hard condition data, mapping it on the nearest network grid points of the simulation regular grid, selecting a custom data template to scan the three-dimensional training image generated in S2, and establishing a multiple-point conditional probability search tree; The multiple-point conditional probability search tree is established in S31, specifically including: S311: A three-dimensional regular grid with equal intervals is constructed in the modeling area, the grid interval is determined according to the drilling distribution density and the modeling resolution, the drilling data in the normalized modeling dataset is mapped to the nearest grid node, and the mapped node is marked as a hard data point, and the unmapped node is marked as a to-be-estimated node; S312: According to the spatial scale characteristics of the geological body and the simulation accuracy requirement, a multiple-point data template is customized; the template shape can be a cube, an ellipsoid or an irregular polygon, and the template contains a plurality of sampling nodes to capture the spatial structure characteristics of multiple points; ​S313: Full-space scanning is performed on the three-dimensional training image constructed in S2 by using the self-defined template as a sliding window. For each spatial configuration within the template, the frequency of each spatial configuration within the template appearing in the training image is counted to form a pair of multi-point spatial configuration and corresponding lithology category conditional probability; S314: The spatial configurations obtained by scanning are tree-organized according to the spatial position sequence of the nodes within the template to construct a multi-point conditional probability search tree, wherein the root node is an empty template, the child nodes correspond to the sampling points within the template in turn, and the leaf nodes store the corresponding lithology category conditional probability distribution.

[0024] S32: Sequential simulation paths are adopted to sequentially select the nodes to be estimated, and simulation values are randomly extracted according to the conditional probability distribution of the search tree and the conditional data set is updated; S33: S32 is repeated until the simulation of all the nodes to be estimated is completed. S34: The simulation results are output as a three-dimensional stratigraphic model to realize the prediction of the spatial distribution of the stratum based on multi-point geostatistics.

[0025] S4: The three-dimensional stratigraphic model is post-processed to realize the inverse calculation of coordinates and depth.

[0026] Specifically, it includes: S41: The relative coordinates of each node in the three-dimensional stratigraphic model are inversely calculated into the actual latitude and longitude coordinates according to the boundary point information of the original coordinate system of the modeling area; S42: The unified depth value is restored according to the surface elevation and depth reference of the original borehole to obtain the depth value corresponding to the actual stratum.

[0027] Embodiment 1 By using the method provided in the present application, the spatial coordinates and stratum drillability data of 18 boreholes in a certain area are modeled, including the following steps: 1. The original borehole data in the modeling area is arranged, and the borehole stratification structure is shown in Figure 2 The electronic table containing the borehole number, latitude and longitude coordinates, borehole depth and lithology stratification information is established. The borehole data is imported into the data processing module, the borehole coordinates are projected and converted, and the modeling area coordinate system is established to represent all borehole positions in a unified relative coordinate system.

[0028] 2. In order to avoid the deviation caused by uneven sampling interval of the original borehole data, the nearest neighbor interpolation method is used for vertical resampling of the borehole data. A unified vertical resolution of 1m is set, and the resampling range is-40m to 6m. The resampling process maintains the discrete properties of the lithofacies code (HFU value), ensures that the interpolation result is within the original depth interval range, and generates a standardized space point set containing x, y, z coordinates and lithofacies category hfu value. The generated borehole attribute diagram is shown in Figure 3shown.

[0029] 3. Construct a three-dimensional regular grid model based on the standardized point set. Set the cell size of X and Y directions as 100 m, the number of grid cells as X = 75, Y = 44, Z = 47, and the origin coordinates as (0, 0, -40). After completion, a regular spatial modeling framework is obtained, providing a unified calculation domain for three-dimensional stratigraphic simulation.

[0030] 4. Treat each type of lithofacies as a categorical variable and convert it into a binary indicator variable. Perform indicator Kriging interpolation independently for each type of lithofacies to estimate the probability of its occurrence at each point in three-dimensional space and generate the corresponding three-dimensional probability volume. This step realizes the conversion from sparse borehole data to continuous probability field, quantifying the spatial uncertainty of the stratigraphic distribution.

[0031] 5. Perform geological constraint correction on the probability volume generated by indicator Kriging. Introduce vertical proportion curve (VPC) and transition probability function (TPF) for collaborative verification and constraint. The vertical proportion curve is used to reflect the overall proportion trend of each type of lithofacies in the vertical direction, and the transition probability is used to describe the contact relationship and sequence transition characteristics of lithofacies in the vertical direction. After integrating both into the correction process, a three-dimensional training image with statistical consistency and geological reasonableness is generated.

[0032] 6. According to the corrected three-dimensional probability volume, the spatial distribution proportions of each lithofacies are calculated to determine the four types of lithofacies units: HFU-0, HFU-1, HFU-2, and HFU-3, with overall proportions of 0.20, 0.19, 0.41, and 0.20, respectively. Use the visualization module to generate a three-dimensional training image, as shown in Figure 4 .

[0033] 7. Based on the generated three-dimensional training image, use the SNESIM algorithm for multiple-point geostatistical random simulation. Set the simulation origin as (-40, 0, 0), the simulation grid cell size as Δx = 100 m, Δy = 100 m, Δz = 1 m, the number of directional grid cells as X = 75, Y = 44, Z = 47, the number of multiple grid layers as 3, the service coefficient as 0.2, and the minimum repetition number as 5. According to the characteristics of the riverine facies sedimentary body in the study area, set the search radius as hx = 55 m, hy = 35 m, hz = 18 m, and the principal axis angles a1, a2, a3 as 0°.

[0034] 8. Use the borehole data in step 2 as hard conditioning data and the training image generated in step 5 as a statistical sample to sequentially select the nodes to be estimated using the sequential simulation path. Use the multiple-point conditional probability search tree to retrieve the corresponding conditional probability distribution, perform random sampling and assignment, and update the conditional data set until all nodes are simulated. To ensure statistical stability, the maximum simulation implementation number is set to 80. The simulation results are shown in Figure 5 .

[0035] 9. After the simulation is completed, the generated three-dimensional stratum model is subjected to geological analysis and result verification. The simulation result shows that the stratum model is basically consistent with the actual drilling exposure data in spatial distribution, and from shallow to deep, there are low-permeability layer, medium-permeability layer, high-permeability layer and bottom impermeable layer, and the transition between each stratum unit is smooth and the connectivity is good, which is consistent with the geological characteristics of the study area.

[0036] 10. The finally generated three-dimensional stratum model is subjected to post-processing. By combining the boundary coordinates and surface elevation information of the modeling area, the relative coordinates are converted into actual geographical longitude and latitude coordinates, and the unified depth is restored to the real stratum depth.

[0037] Therefore, the application adopts the above-mentioned multi-point geostatistical stratum model construction method coupled with data driving and geological constraints, realizes the full-process automatic modeling from data preprocessing, training image construction to stratum model generation and post-processing under the condition of limited geological data. On the premise of maintaining geological rationality, the method takes into account the modeling accuracy and calculation efficiency, significantly reduces the manual interpretation and subjective intervention. By fusing the geological constraint parameters and multi-point spatial statistical characteristics, the spatial connectivity and structural continuity of the stratum model are effectively improved, which provides an efficient, stable and engineering application solution for the three-dimensional stratum fine modeling under complex geological environment.

[0038] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application but not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that: it can still modify or equivalently replace the technical solutions of the present application, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for building a multiple-point geostatistics stratigraphic model coupling data drive and geologic constraint, characterized in that, The method comprises the following steps: S1: obtaining original drilling data of a modeling area and preprocessing to generate a modeling data set; S2: based on the modeling data set, adopting a facies simulation method of fusing data driving and geological constraints to construct a three-dimensional training image; S3: taking the modeling data set as hard condition data, and based on the three-dimensional training image constructed in S2, using SNESIM algorithm to perform multiple-point geostatistical random simulation to generate a three-dimensional stratigraphic model; S4: post-processing the three-dimensional stratigraphic model to realize coordinate and depth inverse calculation.

2. The method of claim 1, wherein, S1 specifically comprises: S11: extracting and organizing original drilling stratigraphic data in the modeling area, wherein the original drilling stratigraphic data comprises drilling number, latitude and longitude coordinates, depth and lithology layering information; S12: jointly calculating the original drilling latitude and longitude coordinates and the boundary point coordinates of the modeling area to obtain the relative coordinates of each drilling in the modeling area coordinate system; S13: normalizing the depth information of the drilling data to unify the depth scale; S14: organizing the drilling number, relative coordinates, unified depth value and lithology layering information into a structured spreadsheet data set to obtain the modeling data set.

3. The method of claim 2, wherein the method further comprises: S2 specifically comprises: S21: based on the modeling data set, using nearest neighbor interpolation method to resample all drilling data at a unified vertical resolution to generate a standardized spatial point set, wherein the spatial point set comprises X, Y and Z direction coordinate values and corresponding lithology categories; S22: regarding each type of facies as a categorical variable and converting it into a binary indicator variable; for each type of facies, independently performing indicator kriging interpolation to establish a spatial variation function model of the occurrence probability of each facies in the three-dimensional space to generate a three-dimensional probability body; S23: based on geological constraint parameters, modifying the three-dimensional probability body; the geological constraint parameters comprise a vertical proportion curve and a transition probability function; S24: based on the modified three-dimensional probability body, constructing a three-dimensional training image fusing geological knowledge and data driven features.

4. The method of claim 3, wherein the method further comprises: The vertical proportion curve is expressed as: ； in, In vertical position lithofacies category proportion, For depth Category The number of samples, This represents the total number of samples at this depth level.

5. The method of claim 4, wherein the method further comprises: The transition probability function is expressed as: ； wherein, i , j is a lithology class, h is a step size, is a lithology class at a spatial location z.

6. The method for constructing a data-driven and geologically-constrained multiple-point geostatistical facies model according to claim 1, wherein, S3 specifically comprises: S31: taking the modeling data set as hard condition data, mapping on the nearest network grid point of the simulation rule grid, selecting a self-defined data template to scan the three-dimensional training image generated in S2 to establish a multiple-point conditional probability search tree; S32: using a sequential simulation path to sequentially select a to-be-estimated node, randomly extracting a simulation value according to the conditional probability distribution of the search tree and updating the conditional data set; S33: repeating S32 until all to-be-estimated nodes are simulated; S34: outputting the simulation result as a three-dimensional stratigraphic model.

7. The method for constructing a data-driven and geologically-constrained multiple-point geostatistical facies model according to claim 6, wherein, In S31, the multiple-point conditional probability search tree is established, specifically comprising: S311: constructing equidistant three-dimensional regular grids in the modeling area, wherein the grid spacing is determined according to the drilling distribution density and the modeling resolution, mapping the drilling data in the modeling data set to the nearest grid node, marking the mapped node as a hard data point, and marking the unmapped node as a to-be-estimated node; S312: self-defining a multiple-point data template according to the spatial scale characteristics of the geological body and the simulation accuracy requirement; S313: Full-space scanning is performed on the three-dimensional training image constructed in S2 using the self-defined template as a sliding window. For each spatial configuration within the template, the frequency of each spatial configuration within the template appearing in the training image is counted to form a pair of multi-point spatial configuration and corresponding lithology category conditional probability; S314: The spatial configurations obtained by scanning are tree-organized according to the spatial position sequence of the nodes within the template to construct a multi-point conditional probability search tree, wherein the root node is an empty template, the child nodes correspond to the sampling points within the template in turn, and the leaf nodes store the corresponding lithology category conditional probability distribution.

8. The method for constructing a data-driven and geologically-constrained multiple-point geostatistical stratigraphic model according to claim 1, wherein, S4 specifically comprises: S41: The relative coordinates of each node in the three-dimensional stratigraphic model are inversely calculated into actual latitude and longitude coordinates according to the boundary point information of the original coordinate system of the modeling area; S42: The unified depth value is restored according to the surface elevation and depth reference of the original borehole to obtain a depth value corresponding to the actual stratum.

Images