Three-dimensional geological implicit modeling method based on global constraint features, medium and equipment

Abstract

The invention discloses a three-dimensional geological implicit modeling method based on global constraint features, a medium and equipment, and relates to the technical field of geological modelling, the three-dimensional geological implicit modeling method based on global constraint features mainly comprises the following steps: carrying out vertical dimension sampling and horizontal dimension sampling on original geological data; a profile data set and a drilling data set are obtained, global feature extraction is carried out on the profile data set and the drilling data set to obtain global vertical features and global horizontal features, global feature fusion and coordinate matching are carried out on the global vertical features and the global horizontal features to obtain a machine learning data set, and a Stacking machine learning model is trained to obtain a trained Stacking machine learning model; and predicting a research area by using the trained Stacking machine learning model to obtain a three-dimensional geologic model. According to the three-dimensional geological implicit modeling method based on the global constraint features, the medium and the equipment provided by the invention, the global expression of the underground space in the three-dimensional geological modeling process can be enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of geological modeling, and more specifically to a three-dimensional geological implicit modeling method, medium and equipment based on global constraint features. Background Art

[0002] Since underground space is essentially three-dimensional, the three-dimensional geological model can be used to visualize the abstract geological structure and vividly and intuitively show the spatial distribution and relationship of geological units. This is of great significance for studying underground space and geological structure, ensuring the scientificity and feasibility of geological space planning, and reducing development and exploration risks.

[0003] With the development of artificial intelligence, 3D geological modeling has gradually shifted from explicit modeling methods based on manual interaction to implicit modeling methods based on machine learning. However, the current mainstream 3D geological modeling methods based on machine learning only consider local features such as the location characteristics of discrete points and geophysical characteristics, and have certain deficiencies in the overall morphological expression of underground space. There is also the problem that it is difficult to infer the global cognition of geological targets based on a small amount of sampling data. Therefore, there is an urgent need for 3D geological modeling methods based on global feature constraints. Summary of the invention

[0004] The purpose of the present invention is to provide a three-dimensional geological implicit modeling method, medium and equipment based on global constraint features, which can enhance the global expression of underground space in the three-dimensional geological modeling process.

[0005] The present invention provides a three-dimensional geological implicit modeling method based on global constraint features, comprising the following steps: S1: acquiring original geological data, sampling the original geological data in vertical dimensions and horizontal dimensions, and obtaining a profile data set and a borehole data set; S2: extracting global features based on the profile data set and the borehole data set, and obtaining global vertical features and global horizontal features; S3: performing global feature fusion and coordinate matching based on the global vertical features and the global horizontal features, and obtaining a machine learning data set; dividing the machine learning data set according to a preset ratio, and obtaining a training set and a test set; S4: constructing a Stacking machine learning model, and training the Stacking machine learning model using the training set and the test set, and obtaining a trained Stacking machine learning model; S5: predicting a study area using the trained Stacking machine learning model, and obtaining a three-dimensional geological model.

[0006] Furthermore, step S2 of the above-mentioned three-dimensional geological implicit modeling method based on global constraint features specifically includes: S21: obtaining global vertical features using a convolutional neural network based on the profile data set; S22: performing data balancing, implicit function interpolation, normal vector calculation and Kriging interpolation on the borehole data set to obtain global horizontal features.

[0007] Furthermore, step S21 of the above-mentioned three-dimensional geological implicit modeling method based on global constraint features specifically includes: S211: cropping and encoding the profile data set according to a preset size to obtain a two-dimensional profile image data set; S212: extracting features using a convolutional neural network based on the two-dimensional profile image data set to obtain global vertical features.

[0008] Furthermore, step S22 of the above-mentioned three-dimensional geological implicit modeling method based on global constraint features specifically includes: S221: using synthetic minority class oversampling technology and edited nearest neighbor method to balance the drilling data set to obtain a balanced data set; S222: based on the balanced data set, using radial basis function interpolation method to obtain geological equivalent interface; S223: based on the geological equivalent interface, using K nearest neighbor estimation method and singular value decomposition method to obtain horizontal trend characteristics of geological equivalent interface; S224: based on the horizontal trend characteristics of the geological equivalent interface, using Kriging method to interpolate the normal vectors of discrete points of the geological equivalent interface to obtain global horizontal characteristics.

[0009] Furthermore, step S3 of the above-mentioned three-dimensional geological implicit modeling method based on global constraint features specifically includes: S31: fusing the global vertical features and the local horizontal features to obtain global features; S32: matching the global features with the spatial position coordinates to obtain a machine learning data set; S33: dividing the machine learning data set according to a preset ratio to obtain a training set and a test set.

[0010] Furthermore, the Stacking machine learning model of the above-mentioned three-dimensional geological implicit modeling method based on global constraint features includes a base classifier and a meta-classifier; the base classifier includes a support vector machine module, a random forest module, a K nearest neighbor estimation module, a gradient boosting decision tree module and an extreme gradient boosting tree module; the meta-classifier includes a logistic regression module.

[0011] Furthermore, step S5 of the above-mentioned three-dimensional geological implicit modeling method based on global constraint features specifically includes: S51: determining the modeling range boundary according to the study area, and establishing a bounding box according to the modeling range boundary; S52: performing regular grid division on the bounding box according to a preset accuracy to obtain a three-dimensional grid simulation model; S53: according to the global characteristics of the study area, using the three-dimensional grid simulation model to obtain a three-dimensional geological model geological feature data set; S54: according to the three-dimensional geological model geological feature data set, using the trained Stacking machine learning model to obtain a three-dimensional geological model.

[0012] Furthermore, the above three-dimensional geological implicit modeling method based on global constraint features also includes: using information entropy to perform uncertainty analysis on the three-dimensional geological model.

[0013] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned three-dimensional geological implicit modeling method based on global constraint features are implemented.

[0014] The present invention also provides a computer device, including a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned three-dimensional geological implicit modeling method based on global constraint features when executing the program.

[0015] The implementation of the three-dimensional geological implicit modeling method, medium and equipment based on global constraint features provided by the present invention has the following beneficial effects: The present invention starts from a global perspective and extracts feature vectors and normal vectors from the vertical and horizontal dimensions respectively, forming a global consideration of the multi-dimensional underground space, that is, by extracting multi-dimensional global features of the global stratification features of the vertical dimension of geological data and the global trend features of the horizontal dimension as geological rules as constraints; by adding self-attention to improve the Stacking ensemble learning model, an ensemble learning model integrating multiple rules is proposed, and the self-attention improved Stacking ensemble learning model is used for geological prediction, and finally a three-dimensional geological model is generated, which fully integrates the geological features, solves the problem that traditional ensemble learning fails to extract effective features when processing multi-dimensional data, obtains a more accurate and robust ensemble learning model, provides a complete three-dimensional geological implicit modeling process, enhances the global expression of underground space in the three-dimensional geological modeling process, better fits the underground space structure as a whole, and provides a new idea for effectively extracting and integrating global features in the three-dimensional geological modeling process, and provides a reliable foundation for geological research and engineering applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1 It is a flow chart of a three-dimensional geological implicit modeling method based on global constraint features provided by the present invention; Figure 2 It is a schematic diagram of the steps of three-dimensional geological implicit modeling based on global feature constraints provided by the present invention; Figure 3 It is a schematic diagram of profile feature extraction provided by the present invention; Figure 4 It is a flow chart of the formation normal vector calculation method provided by the present invention; Figure 5 is a flow chart of the normal vector field calculation method provided by the present invention; Figure 6 It is a schematic diagram of multi-dimensional global feature space coordinate matching provided by the present invention; Figure 7 It is a flow chart of the integrated learning model integrating multiple rules provided by the present invention; Figure 8 It is a structural block diagram of the computer device provided by the present invention. DETAILED DESCRIPTION

[0017] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0018] Figure 1 A schematic diagram of a 3D geological implicit modeling method based on global constraint features of this embodiment is shown. In this embodiment, the 3D geological implicit modeling method based on global constraint features includes the following steps: S1: acquiring original geological data, and performing vertical dimension sampling and horizontal dimension sampling on the original geological data to obtain a profile data set and a drilling data set; Specifically, the original geological data includes a priori geological models and original drilling data; It should be noted that, in the stage of processing the original data in step S1, global feature extraction is performed from different dimensions, and corresponding data sets are constructed according to the original geological data: a profile data set is constructed in the vertical dimension, and a borehole data set is constructed in the horizontal dimension; the profile data set is the two-dimensional image data of the geological profile, and the borehole data set is the discrete point data after the original borehole is resampled; Specifically, in step S1, the process of constructing the profile data set includes: collecting three-dimensional profiles from two specific directions, north-south and east-west, using the prior geological model obtained through prior knowledge of stratigraphic connectivity; selecting a mutually perpendicular north-south three-dimensional profile s1 and an east-west three-dimensional profile s2 for each intersecting grid unit as the profile data matched by the grid unit; converting the three-dimensional profile into a two-dimensional profile image, retaining the acquisition number, performing image scale matching, generating a two-dimensional profile image data set, and obtaining a profile data set; Specifically, in step S1, the process of constructing the borehole data set includes: calculating the x, y, and z coordinate values ​​of each geological boundary point according to the point information and stratum thickness information of the original borehole data; resampling the borehole data to fill the blank strata between the stratum boundary points; adding corresponding attribute values ​​to the geological boundary discrete point data to form a borehole data set; S2: extracting global features according to the profile data set and the drilling data set to obtain global vertical features and global horizontal features; Specifically, Figure 2 The flowchart of the three-dimensional geological implicit modeling method based on global feature constraints in this embodiment is shown; step S2 specifically includes: S21: obtaining global vertical features using a convolutional neural network according to the profile data set; It should be noted that, in step S21, the global feature extraction of the profile data set in the vertical direction is performed using a convolutional neural network, which mainly includes: performing a convolution operation on the profile data set to obtain the global hierarchical features of the profile in the vertical direction, that is, the profile image features, thereby obtaining the global vertical features; Specifically, step S21 includes: S211: cutting and encoding the cross-sectional data set according to a preset size to obtain a two-dimensional cross-sectional image data set; Specifically, in step S211, the geological profile images in the profile data set are uniformly cropped to a preset size and encoded to form a two-dimensional profile image data set; S212: extracting features using a convolutional neural network according to the two-dimensional profile image data set to obtain global vertical features; It should be noted that, in step S212, a convolutional neural network (CNN) is used to extract the features of the geological profile in the two-dimensional profile image data set to obtain the global layered features of the profile in the vertical direction; the convolutional neural network uses a Laplace convolution kernel to extract the stratum boundary features; Specifically, the convolutional neural network includes a Laplacian convolution kernel; Specifically, the Laplace filter is used to identify the stratigraphic relationship between soil sediments. It can be estimated by the following formula:

[0019] In the formula represents the horizontal distance, Represents vertical depth, pixel intensity Represents the soil type, whose index is an integer; for the discrete domain, the Laplacian filter used is simplified as follows:

[0020] It should be noted that the non-zero convolution value obtained by Laplace filtering indicates the edge position between different strata boundaries; a single filter is convolved with the training image to construct a feature map corresponding to a given grid scale, and then a feature pooling layer is used to filter out unimportant strata relationships, such as Figure 3 Schematic diagram of profile feature extraction in the embodiment; finally, the extracted features are output and flattened through a fully connected layer to obtain a one-dimensional feature vector, that is, a feature vector of the target geological profile, thereby obtaining a global vertical feature; S22: performing data balancing, implicit function interpolation, normal vector calculation and Kriging interpolation on the drilling data set to obtain global level features; It should be noted that in step S22, global horizontal feature extraction is performed based on the borehole data set, and the normal vector field is constructed by implicit function interpolation to complete the global horizontal trend feature extraction in the horizontal dimension. The main steps are: data balancing of the borehole data set, and then using KNN and SVD decomposition algorithms to calculate the normal vector information of discrete points, and finally using Kriging interpolation to calculate the normal vector field of the study area, that is, the horizontal trend field of the geological body, to obtain the global trend feature, that is, the global horizontal feature; the process of extracting the global trend feature of the horizontal dimension using the borehole data set is as follows: Figure 4 As shown; Specifically, step S22 includes: S221: performing data balancing on the drilling data set using a synthetic minority class oversampling technique and an edited nearest neighbor method to obtain a balanced data set; Specifically, in step S221, the drilling data set is balanced using the synthetic minority over-sampling technique (SMOTE) and the edited nearest neighbors (ENN) method to form a balanced data set; S222: obtaining a geological equivalent interface using a radial basis function interpolation method according to the balanced data set; Specifically, in step S222, according to the balanced data set, the geological equivalent interface is fitted to extract the horizontal trend characteristic information of the geological equivalent interface; the radial basis function (RBF) interpolation method is used to perform implicit function interpolation on the unevenly distributed discrete points of the same equivalent layer, and the geological grid units are evenly filled in the space to form a geological equivalent interface; the algorithm form is s( x ), as follows:

[0021] in, is a set of different interpolation centers of the grid to be filled, is the weight coefficient, is a type of radial basis function, is a first-degree polynomial; S223: According to the geological equivalent interface, using a K nearest neighbor estimation method and a singular value decomposition method, obtaining horizontal trend characteristics of the geological equivalent interface; Specifically, in step S223, the K-Nearest Neighbors (KNN) estimation method and the singular value decomposition (SVD) method are used to calculate the normal vectors of the grid discrete points of the geological equivalent interface, obtain the directional characteristic information of all grid discrete points of the geological equivalent interface, and obtain the horizontal trend characteristics of the geological equivalent interface. The specific flow chart is as follows: Figure 5 As shown; K nearest neighbor estimation (KNN) is performed on the geological target points to determine a field whose distance The measurement formula is as follows:

[0022] in , All belong to n dimensional real vector space , It is expressed as Euclidean distance, which is also the distance measurement algorithm used in this embodiment; select a suitable K value, search for the K geological sample point sets closest to the test sample obtained from the geological training sample, and use the least squares method to perform plane fitting based on the geological sample point set with the same geological interface obtained by the K-nearest neighbor algorithm to calculate the normal vector of the test sample, i.e., the target point. The fitting plane equation formula is as follows:

[0023] Among them, a, b, c are plane parameters, that is, eigenvectors, and then e is obtained by covariance SVD matrix decomposition as follows:

[0024] The obtained plane parameters a, b, c are the characteristic vectors, that is, the plane normal vectors, which represent the point normal vectors of the discrete point here. By analogy, the discrete point normal vectors of each geological equivalent interface are finally obtained, and then the horizontal trend characteristics of the entire geological interface are obtained. S224: According to the horizontal trend characteristics of the geological equivalent interface, the normal vectors of the discrete points of the geological equivalent interface are interpolated by using the Kriging method to obtain the global horizontal characteristics; Specifically, in step S224, three-dimensional Kriging interpolation is performed on the normal vector of the isosurface, normal vector features are assigned to the discrete points of the entire geological body grid, and a horizontal trend field of the geological body is constructed, thereby constraining the horizontal trend of the entire three-dimensional geological body. The steps include: First, determine the interpolation set N, that is, the three-dimensional grid center point data set, and determine the three-directional variation range of the ellipsoid. By fitting the variation function between the known data points To complete the three-dimensional variogram fitting model calculation, as follows:

[0025] in is the axial anisotropy weight function, , , is the variation function of the three axes, which determines the three-directional variation range of the search ellipsoid. The calculation formula is as follows:

[0026] in, is a vector The number of point value pairs separated by and A certain position in space and distance from it A certain position in space Two regionalized variables; Secondly, a three-dimensional geological space search ellipsoid is established and the geological sample points are searched. The center coordinates of the search ellipsoid are determined by taking the center of the three-dimensional grid as the interpolation point. , and set the radius of the ellipsoid in three axes , and ; These radii are consistent with the three axes of the variogram and cannot exceed the range of each axis , and ; The sample points within the search ellipsoid are denoted as , the conditions are:

[0027] It should be noted that this conditional formula represents the sample point The coordinates of the sample point are inside or on the boundary of the search ellipsoid. If this condition is met, it means that the sample point falls within the range of the search ellipsoid. Then, the coefficient vector of the point to be interpolated is calculated using the ordinary Kriging interpolation weight coefficient equation. By using the selected sample points that meet the conditions and the coefficient vector of the point to be interpolated, the interpolation result of the point to be interpolated is calculated as follows:

[0028] in, represents the unsampled position in the center of the 3D geological grid, express Geological sampling points, represents the weight of ordinary kriging, express u The estimated value at 0; Finally, all the points to be interpolated are traversed and the interpolation results are output; thus, each grid of the entire 3D grid model is given a normal feature, forming a normal vector field, completing the construction of the horizontal trend field of the geological body, and obtaining the global horizontal feature, which provides the global trend feature of the stratum in the horizontal direction for 3D geological modeling; S3: performing global feature fusion and coordinate matching according to the global vertical features and the global horizontal features to obtain a machine learning data set; dividing the machine learning data set according to a preset ratio to obtain a training set and a test set; Specifically, step S3 includes: S31: fusing the global vertical feature and the local horizontal feature to obtain a global feature; Specifically, in step S31, Figure 6 As shown in the figure, the global vertical features and the local horizontal features are fused to obtain the global features. The feature fusion form is as follows:<x, y, z, s1, s2, nx, ny, nz> , where x, y, z represent spatial position coordinates; s1 and s2 are two one-dimensional vectors representing the vertical characteristic information of the cross section; nx, ny, nz are the normal information representing the horizontal trend characteristics; S32: Matching the global features with the spatial position coordinates to obtain a machine learning data set; Specifically, in step S32, the extracted drilling point data are added with corresponding global features to form a geological feature data set; first, the spatial point coordinates of each grid of the three-dimensional grid model are obtained by calculating the proton coordinates of the center of the grid unit; secondly, vertical or horizontal feature information is given to each grid, and the spatial point coordinates are matched with the global vertical features and the global horizontal features, so that each grid center point contains spatial point coordinates and multi-dimensional global feature information, and its feature vector form is as follows:

[0029] in Represents the coordinates of the center point of the grid; Two one-dimensional vectors represent the vertical feature information of the cross section intersecting the grid; To represent the normal information of the horizontal features of the grid; finally, the feature vector is normalized, and the maximum and minimum value normalization method is used to process the data to obtain a machine learning data set; S33: Divide the machine learning data set according to a preset ratio to obtain a training set and a test set; Specifically, the above preset ratio is 8:2; S4: construct a Stacking machine learning model, and train the Stacking machine learning model using the training set and the test set to obtain a trained Stacking machine learning model; Specifically, the Stacking machine learning model includes a base classifier and a meta classifier; the base classifier includes a support vector machine module, a random forest module, a K nearest neighbor estimation module, a gradient boosting decision tree module and an extreme gradient boosting tree module; the meta classifier includes a logistic regression module; It should be noted that the Stacking machine learning model includes the base classifiers of the first layer and the meta classifiers of the second layer; the base classifiers of the first layer include the Support Vector Machine (SVM) module, the Random Forest (RF) module, the K-nearest neighborhood (KNN) module, the Gradient Boosting Decision Tree (GBDT) module and the eXtreme Gradient Boosting (XGBoost) module; the meta classifiers of the second layer include the Logistic Regression (LR) module; Specifically, in step S4, the Stacking machine learning model is trained using the training set, the outputs of the five base classifiers are used as the training data of the meta-classifier of the second layer of Stacking, the logistic regression module of the second layer is trained, and the model accuracy is evaluated using the test set to obtain a trained Stacking machine learning model; Specifically, Stacking integrates the machine learning process such as Figure 7 As shown, the process is as follows: First, given a training dataset ,in Represents the feature vector, including spatial coordinate features, vertical vector features and horizontal vector features. represents the corresponding target variable; Secondly, train the first layer of the Stacking machine learning model, SVM, RF, KNN, GBDT, and XGBoost, five base classifiers, using the training data set To train each base classifier , where t ranges from 1 to , by indexing each sample in the training dataset Iterate on it to build a new prediction dataset ,in The range is from 1 to , , is the base classifier For samples The training characteristics of is the target variable; Then, the probability features output by the first layer are input into the Attention self-attention mechanism for vector weighting to extract effective features. It is input into the self-attention layer as a new feature vector Perform attention weighting to construct a weighted data set ; Finally, the vector weighted data is used as the training data for the second layer of logistic regression to train the second layer, that is, using the new weighted data set To train the meta-classifier , and use the test set to evaluate the model accuracy to obtain the trained Stacking machine learning model; S5: using the trained Stacking machine learning model to predict the study area and obtain a three-dimensional geological model; Specifically, step S5 includes: S51: determining a modeling range boundary according to the research area, and establishing a bounding box according to the modeling range boundary; It should be noted that the bounding box can accommodate the entire modeling area; S52: performing regular grid division on the bounding box according to a preset accuracy to obtain a three-dimensional grid simulation model; S53: according to the global characteristics of the study area, using the three-dimensional grid simulation model, obtain a three-dimensional geological model geological characteristic data set; Specifically, in step S53, each grid of the three-dimensional grid simulation model is regarded as a geological body unit, and the voxel center of each grid is calculated to obtain the spatial position information of the discrete point; the global characteristics of the study area are matched to the voxel center point of each grid, and the geological body to be predicted is converted into new three-dimensional discrete point data to obtain the geological characteristic data set of the three-dimensional geological model; S54: obtaining a three-dimensional geological model according to the geological feature data set of the three-dimensional geological model and using the trained Stacking machine learning model; Specifically, in step S54, the three-dimensional discrete point data of the geological feature data set of the three-dimensional geological model is predicted using the trained Stacking machine learning model to obtain the attribute value of each discrete point and form a three-dimensional geological implicit field model; the discrete point attribute value is assigned to each grid to realize the implicit field visualization of the modeling area, that is, to obtain the prediction result; the geological attributes of the prediction result are extracted as the label of each geological unit; based on these labels and the visualized prediction results, a three-dimensional geological model is obtained.

[0030] It should be noted that the above three-dimensional geological implicit modeling method based on global constraint characteristics may also include: using information entropy to perform uncertainty analysis on the three-dimensional geological model; Specifically, using information entropy to perform uncertainty analysis on a three-dimensional geological model includes the following steps: First, the accuracy of the Stacking ensemble learning model is evaluated by precision, recall, F1 score and ROC curve, and then the uncertainty of the 3D geological model is analyzed by information entropy, as shown in the formula:

[0031] in, is the information entropy, represents the number of geological categories, Indicates that the unit is predicted to be The probability of the class.

[0032] In some embodiments, the above-mentioned three-dimensional geological implicit modeling method based on global constraint features can also be implemented in the following manner.

[0033] Please refer to Figure 2 This embodiment provides a three-dimensional geological implicit modeling method based on global feature constraints, which specifically includes the following steps: Step S01: Raw data processing stage: according to the global feature extraction of different dimensions, the corresponding data set is constructed from the geological raw data, the profile data set is constructed in the vertical dimension, and the drilling data set is constructed in the horizontal dimension; Step S02: Global feature extraction stage: Using the profile data set constructed in step S01, extract global hierarchical features through CNN to complete global feature extraction in the vertical dimension; using the borehole data set constructed in step S01, construct a normal vector field through implicit function interpolation to complete global horizontal trend feature extraction in the horizontal dimension; Step S03: Global feature fusion stage: perform data fusion on the global vertical features and global horizontal features extracted in S02, match them with the spatial position coordinates to form a machine learning data set, and divide it into a training set and a test set; Step S04: Stacking model training phase: Use the training set data in step S03 to train the Stacking machine learning model, and use the test set to evaluate the model accuracy; Step S05: 3D geological model establishment and evaluation phase: Use the trained Staking machine learning model to predict the entire 3D geological model, complete the 3D geological implicit modeling, and complete the visualization through the 3D grid model, and finally perform uncertainty analysis.

[0034] Specifically, in step S01, the profile data set and the drilling data set are constructed as follows: First, the prior geological model obtained through the prior knowledge of stratigraphic connectivity is used to collect 3D profiles from two specific directions, the north-south direction and the east-west direction. Then, each cross grid selects mutually perpendicular 3D profiles s1 (north-south direction) and s2 (east-west direction) as the profile data matched by the grid unit. Finally, the 3D profile is converted into a 2D profile image, the acquisition number is retained, and image scale matching is performed to generate a 2D profile image data set. Based on the original borehole data point information and stratum thickness information, the x, y, and z coordinate values ​​of each geological boundary point are calculated; then the borehole data is resampled to fill the blank strata between the stratum boundary points; finally, the corresponding attribute values ​​are added to the geological boundary discrete point data to form a borehole data set.

[0035] Specifically, in step S02, the specific process of extracting the global hierarchical features of the vertical dimension using the profile data set is as follows: First, the geological profile is encoded, and then the stratigraphic relationship between soil sediments is identified using a Laplace filter. It can be estimated by the following formula: (1) In the formula represents the horizontal distance, Represents vertical depth, pixel intensity Represents soil type, its index is an integer; For the discrete domain, the Laplace filter used is simplified as follows: (2) The non-zero convolution value obtained by Laplace filtering indicates the edge position between different strata boundaries; a single filter is convolved with the training image to construct a feature map corresponding to a given grid scale, and then a feature pooling layer is used to filter out unimportant strata relationships, such as Figure 3 It is a schematic diagram of profile feature extraction of the embodiment; finally, the extracted features are output and flattened through a fully connected layer to obtain a one-dimensional feature vector, that is, the target geological profile feature vector.

[0036] Specifically, in step S02, the process of extracting the global trend feature of the horizontal dimension using the drilling data set is as follows: Figure 4 , specifically: Step 1: Data balancing: Use SMOTE-ENN to process the unbalanced data set and obtain a balanced data set; Step 2, extracting the horizontal trend characteristic information of the geological equivalent interface: In this embodiment, the RBF interpolation method is used to form the geological equivalent interface. Secondly, the K nearest neighbor combined with singular value decomposition algorithm (KNN-SVD) is used to estimate the normal vector of the geological equivalent interface grid point to extract the horizontal trend characteristics of the geological equivalent interface, such as Figure 5 As shown; First, the geological isovalue interface is fitted, and the RBF interpolation algorithm is used to perform implicit function interpolation on the unevenly distributed discrete points of the same isovalue layer, and the geological grid units are evenly filled in the space. The algorithm form is s( x ), as follows: (3) in is a set of different interpolation centers of the grid to be filled, is the weight coefficient, is a type of radial basis function, is a first-degree polynomial; Then the normal vector of the discrete points is calculated using the KNN-SVD algorithm, and the directional characteristic information of all grid discrete points on the geological equivalent interface is obtained; K nearest neighbor estimation (KNN) is performed on the geological target points to determine an area whose distance The measurement formula is as follows: (4) in , All belong to n dimensional real vector space , It is expressed as Euclidean distance, which is also the distance metric algorithm used in this study; Select a suitable K value, search for the K geological sample point sets closest to the test sample from the geological training samples, and use the least squares method to perform plane fitting to calculate the normal vector of the test sample (target point) based on the geological sample point set with the same geological interface obtained by the K-nearest neighbor algorithm. The fitting plane equation formula is as follows: (5) Among them, a, b, and c are plane parameters, i.e., eigenvectors; then, e is obtained by covariance SVD matrix decomposition as follows: (6) The obtained plane parameters a, b, c are the characteristic vectors, that is, the plane normal vectors, which represent the point normal vectors of the discrete point here. By analogy, the normal vectors of the discrete points of each geological equivalent interface are finally obtained, and then the horizontal trend characteristics of the entire geological interface are obtained. Step 3, constructing the horizontal trend field of the geological body: performing three-dimensional Kriging interpolation on the normal vectors of the isosurfaces, assigning normal vector characteristics to the discrete points of the entire geological body grid, and constructing the horizontal trend field of the geological body, thereby realizing the constraint on the horizontal trend of the entire three-dimensional geological body; First, determine the interpolation set N, that is, the three-dimensional grid center point data set, and determine the three-directional variation range of the ellipsoid. By fitting the variation function between the known data points To complete the three-dimensional variogram fitting model calculation, as follows: (7) in is the axial anisotropy weight function, , , is the variation function of the three axes, which determines the three-directional variation range of the search ellipsoid. The calculation formula is as follows: (8) in is a vector The number of point value pairs separated by and A certain position in space and distance from it A certain position in space Two regionalized variables; Secondly, a three-dimensional geological space search ellipsoid is established and the geological sample points are searched. The center coordinates of the search ellipsoid are determined by taking the center of the three-dimensional grid as the interpolation point. , and set the radius of the ellipsoid in three axes , and ; These radii are consistent with the three axes of the variogram and cannot exceed the range of each axis , and ; The sample points within the search ellipsoid are denoted as , the conditions are: (9) This conditional formula represents the sample point The coordinates of the sample point are inside or on the boundary of the search ellipsoid. If this condition is met, it means that the sample point falls within the range of the search ellipsoid. Then, the coefficient vector of the point to be interpolated is calculated using the ordinary Kriging interpolation weight coefficient equation. By using the selected sample points that meet the conditions and the coefficient vector of the point to be interpolated, the interpolation result of the point to be interpolated is calculated as follows: (10) in u 0 means the center of the 3D geological grid is not sampled. express Geological sampling points, represents the weight of ordinary kriging, express u The estimated value at 0; Finally, all the points to be interpolated are traversed and the interpolation results are output; thus, each grid of the entire three-dimensional grid model is given a normal feature, and the horizontal trend field of the geological body is constructed, which provides the global trend characteristics of the strata in the horizontal direction for the three-dimensional geological modeling research of this embodiment.

[0037] In step S03, the vertical features and horizontal features extracted in step S02 are fused to form an integrated learning data set such as Figure 6 , the specific steps are: First, the spatial point coordinates of each grid of the three-dimensional grid model are obtained by calculating the proton coordinates of the grid unit center; secondly, vertical or horizontal feature information is given to each grid, and the spatial point coordinates are matched with the global vertical features and global horizontal features, so that each grid center point contains spatial point coordinates and multi-dimensional global feature information, and its feature vector form is as follows: (11) in Represents the coordinates of the center point of the grid; Two one-dimensional vectors represent the vertical feature information of the cross section intersecting the grid; To represent the normal information of the horizontal features of the grid; The feature vectors were normalized, the maximum and minimum value normalization method was used to process the data, and then the ensemble learning data set was randomly divided into a training data set and a test data set in a ratio of 8:2.

[0038] In step S04, stacking integrated machine learning is performed, such as Figure 7 As shown, the specific process is as follows: First, given a training data set ,in Represents the feature vector, including spatial coordinate features, vertical vector features and horizontal vector features. represents the corresponding target variable; Then use the training dataset To train each base classifier (t ranges from 1 to ), by indexing each sample in the training dataset ( The range is from 1 to ) to construct a new prediction dataset ,in , is the base classifier For samples The training characteristics of is the target variable; Will It is input into the self-attention layer as a new feature vector Perform attention weighting to construct a weighted data set ; Finally, using the new weighted dataset To train the meta-classifier .

[0039] In step S05, the specific process of three-dimensional model prediction is as follows: Select the three-dimensional grid model of the study area, calculate the center of each grid voxel, and obtain the spatial position information of discrete points; Match the global features to the center point of each grid voxel to form a geological feature data set of a three-dimensional geological model; Use the trained Stacking model to predict the geological feature data set, obtain the attribute value of each discrete point, and construct a three-dimensional geological implicit field model; Assign discrete point attribute values ​​to each grid to visualize the implicit field in the modeling area.

[0040] Model evaluation: First, the accuracy of the Stacking ensemble learning model is evaluated through precision, recall, F1 score and ROC curve. Secondly, the uncertainty of the 3D geological model is analyzed through information entropy. The description is as follows: (12) in represents the number of geological categories, Indicates that the unit is predicted to be The probability of the class.

[0041] This embodiment provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned three-dimensional geological implicit modeling method based on global constraint features are implemented. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (Flash Memory), a hard disk (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory.

[0042] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the above-mentioned three-dimensional geological implicit modeling method based on global constraint features are implemented.

[0043] like Figure 8As shown, the computer device may include: at least one processor 121, such as a CPU (Central Processing Unit), at least one communication interface 123, a memory 124, and at least one communication bus 122. Among them, the communication bus 122 is used to realize the connection and communication between these components. Among them, the communication interface 123 may include a display screen (Display), a keyboard (Keyboard), and the optional communication interface 123 may also include a standard wired interface and a wireless interface. The memory 124 may be a high-speed RAM memory (Random Access Memory, volatile random access memory), or a non-volatile memory (non-volatile memory), such as at least one disk memory. The memory 124 may also be at least one storage device located away from the aforementioned processor 121. Among them, the memory 124 stores an application program, and the processor 121 calls the program code stored in the memory 124 to perform any of the above method steps. Among them, the communication bus 122 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The communication bus 122 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8Only one line is used to represent it, but it does not mean that there is only one bus or one type of bus. Among them, the memory 124 may include a volatile memory (English: volatile memory), such as a random access memory (English: random-access memory, abbreviated: RAM); the memory may also include a non-volatile memory (English: non-volatile memory), such as a flash memory (English: flash memory), a hard disk drive (English: hard disk drive, abbreviated: HDD) or a solid-state drive (English: solid-state drive, abbreviated: SSD); the memory 124 may also include a combination of the above-mentioned types of memory. Among them, the processor 121 may be a central processing unit (English: central processing unit, abbreviated: CPU), a network processor (English: network processor, abbreviated: NP) or a combination of CPU and NP. Among them, the processor 121 may further include a hardware chip. The above-mentioned hardware chip may be an application-specific integrated circuit (English: application-specific integrated circuit, abbreviated: ASIC), a programmable logic device (English: programmable logic device, abbreviated: PLD) or a combination thereof. The PLD can be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof. Optionally, the memory 124 is also used to store program instructions. The processor 121 can call the program instructions to implement the three-dimensional geological implicit modeling method based on global constraint features as in this embodiment.

[0044] A three-dimensional geological implicit modeling device based on global feature constraints can be run in computing devices such as desktop computers, notebooks, PDAs and cloud servers. The three-dimensional geological implicit modeling device based on global feature constraints can include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the example is only an example of a three-dimensional geological implicit modeling device based on global feature constraints, and does not constitute a limitation on a three-dimensional geological implicit modeling device based on global feature constraints. It can include more or less components than the example, or combine certain components, or different components. For example, the three-dimensional geological implicit modeling device based on global feature constraints can also include input and output devices, network access devices, buses, etc. The so-called processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, field programmable gate arrays or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the three-dimensional geological implicit modeling device based on global feature constraints, and uses various interfaces and lines to connect the various parts of the three-dimensional geological implicit modeling device based on global feature constraints. The memory can be used to store the computer program and / or module, and the processor implements various functions of the three-dimensional geological implicit modeling device based on global feature constraints by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card, a secure digital card, a flash memory card, at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0045] According to the second aspect of the present application, the present invention also provides a three-dimensional geological implicit modeling device based on global feature constraints. The three-dimensional geological implicit modeling device based on global feature constraints includes a memory, a processor, and a three-dimensional geological implicit modeling program based on global feature constraints stored in the memory and executable on the processor. When the three-dimensional geological implicit modeling program based on global feature constraints is executed by the processor, the steps of the three-dimensional geological implicit modeling method based on global feature constraints described in any of the aforementioned embodiments are implemented.

[0046] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

[0047] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0048] The above description is only an embodiment of the present application and is not intended to limit the present invention. For those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

[0049] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A three-dimensional geological implicit modeling method based on global constraint features, characterized in that: The following steps are involved: S1: acquiring original geological data, and performing vertical dimension sampling and horizontal dimension sampling on the original geological data to obtain a profile data set and a drilling data set; S2: extracting global features according to the profile data set and the drilling data set to obtain global vertical features and global horizontal features; S3: performing global feature fusion and coordinate matching according to the global vertical features and the global horizontal features to obtain a machine learning data set; dividing the machine learning data set according to a preset ratio to obtain a training set and a test set; S4: construct a Stacking machine learning model, and train the Stacking machine learning model using the training set and the test set to obtain a trained Stacking machine learning model; S5: Use the trained Stacking machine learning model to predict the study area and obtain a three-dimensional geological model.

2. The three-dimensional geological implicit modeling method based on global constraint features according to claim 1 is characterized in that: Step S2 specifically includes: S21: obtaining global vertical features using a convolutional neural network according to the profile data set; S22: performing data balancing, implicit function interpolation, normal vector calculation and Kriging interpolation on the drilling data set to obtain global level features.

3. The three-dimensional geological implicit modeling method based on global constraint features according to claim 2 is characterized in that: Step S21 specifically includes: S211: cutting and encoding the cross-sectional data set according to a preset size to obtain a two-dimensional cross-sectional image data set; S212: According to the two-dimensional profile image data set, a convolutional neural network is used to extract features to obtain global vertical features.

4. The three-dimensional geological implicit modeling method based on global constraint features according to claim 2 is characterized in that: Step S22 specifically includes: S221: performing data balancing on the drilling data set using a synthetic minority class oversampling technique and an edited nearest neighbor method to obtain a balanced data set; S222: obtaining a geological equivalent interface using a radial basis function interpolation method according to the balanced data set; S223: According to the geological equivalent interface, using a K nearest neighbor estimation method and a singular value decomposition method, obtaining horizontal trend characteristics of the geological equivalent interface; S224: According to the horizontal trend characteristics of the geological equivalent interface, the Kriging method is used to interpolate the normal vectors of the discrete points of the geological equivalent interface to obtain global horizontal characteristics.

5. The three-dimensional geological implicit modeling method based on global constraint features according to claim 1 is characterized in that: Step S3 specifically includes: S31: fusing the global vertical feature and the local horizontal feature to obtain a global feature; S32: Matching the global features with the spatial position coordinates to obtain a machine learning data set; S33: Divide the machine learning data set according to a preset ratio to obtain a training set and a test set.

6. The three-dimensional geological implicit modeling method based on global constraint features according to claim 1 is characterized in that: The Stacking machine learning model includes a base classifier and a meta classifier; the base classifier includes a support vector machine module, a random forest module, a K nearest neighbor estimation module, a gradient boosting decision tree module and an extreme gradient boosting tree module; the meta classifier includes a logistic regression module.

7. The three-dimensional geological implicit modeling method based on global constraint features according to claim 1 is characterized in that: Step S5 specifically includes: S51: determining a modeling range boundary according to the research area, and establishing a bounding box according to the modeling range boundary; S52: performing regular grid division on the bounding box according to a preset accuracy to obtain a three-dimensional grid simulation model; S53: according to the global characteristics of the study area, using the three-dimensional grid simulation model, obtain a three-dimensional geological model geological characteristic data set; S54: According to the geological feature data set of the three-dimensional geological model, using the trained Stacking machine learning model, a three-dimensional geological model is obtained.

8. The three-dimensional geological implicit modeling method based on global constraint features according to claim 1 is characterized in that: The three-dimensional geological implicit modeling method based on global constraint features also includes: using information entropy to perform uncertainty analysis on the three-dimensional geological model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the three-dimensional geological implicit modeling method based on global constraint features as described in any one of claims 1 to 8 are implemented.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the three-dimensional geological implicit modeling method based on global constraint features as described in any one of claims 1-8 are implemented.

Images