A method and system for identifying a low-angle structural plane of a pumped storage upper reservoir

Abstract

The present application relates to a kind of draw and store upper reservoir gentle dip structure face identification method and system, wherein the method includes the following steps: selecting several surveyed draw and store power station upper reservoir area as sample area, obtains the exploration sample data of sample area;Sample area is gridded into three-dimensional voxel grid, and the exploration sample data is preprocessed, the exploration feature of each voxel in three-dimensional voxel grid is extracted, while adding whether the sample label of belonging to gentle dip structure face, into training sample set;Recognition model based on machine learning model is constructed, and the recognition model is iteratively trained using training sample set, and the voxel gentle dip structure face probability recognition model is obtained;The exploration data of the upper reservoir area to be identified of draw and store power station is obtained, and the voxel gentle dip structure face probability recognition model is used to obtain the prediction probability whether each voxel in the region belongs to gentle dip structure face, and then the gentle dip structure face in target area is identified.

Description

Technical Field

[0001] This invention relates to a method and system for identifying gently dipping structural surfaces in pumped storage reservoirs, belonging to the field of surveying automation technology. Background Technology

[0002] Gentle-dip structural surfaces are key potential geological defects in pumped-storage power station construction. Their spatial distribution, connectivity, and material composition directly determine the stability of the rock mass. Gentle-dip structural surfaces typically refer to geological discontinuities with a dip angle ≤30°. Their formation is controlled by multiple factors and can be classified into three typical types according to their genesis: primary structural surfaces; tectonic structural surfaces; and secondary structural surfaces. The core of their engineering identification lies in their spatial distribution characteristics: regional extension of sedimentary layers, retention of slickensides / steps and other tectonic evidence by gentle-dip faults, and unloading fractures opening towards the free face (opening degree 1–50mm). Currently, the following problems exist in the survey: 1. Surface geological mapping, as the most basic and intuitive means, also has significant limitations in the survey of gentle-dip structural surfaces. Due to various factors such as the degree of topographic dissection, the intensity of erosion, and subsequent deposition, surface outcrops are often scattered, isolated, and fragmented. This makes it difficult to obtain complete and continuous spatial distribution information of gently dipping structural surfaces within a region or project area during surveying. Furthermore, the cover layers such as Quaternary loose deposits, colluvial deposits, vegetation, and artificial fill obscure the gently dipping structural surfaces in the underlying bedrock, making it difficult to reliably determine key elements such as the dip extension range, pinch-out location, and intersection relationship with other structural surfaces due to the lack of outcrops. Simultaneously, due to the limited number of outcrops, many potentially hazardous deep gently dipping structural surfaces are easily overlooked in preliminary surveys. 2. Drilling is a direct means of obtaining deep underground geological information, but its inherent techniques also have limitations when exploring gently dipping structural surfaces. The mechanical vibration, drill string friction, and flushing fluid scouring during drilling can easily cause irreversible physical disturbances to the core sample, especially when traversing weak and fractured gently dipping structural surfaces. Structural materials (such as fault gouge, mylonite, and strongly weathered zones) are often crushed, eroded, worn away, or lost, resulting in core samples appearing as "rock cakes" or as sandy or muddy fragments. This disturbance not only destroys key information such as the material composition, thickness, and structural characteristics of the structural surface itself, but also completely alters its original state and mechanical properties (such as shear strength), causing subsequent laboratory tests and engineering evaluations to deviate from the true basis.

[0003] There is an urgent need for a scheme to identify gently tilted structural surfaces that can integrate multi-source data for accurate and automatic identification. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention proposes a method and system for identifying the gently tilted structural surface of a pumped storage reservoir.

[0005] The technical solution of the present invention is as follows: On the one hand, this invention proposes a method for identifying gently dipping structural surfaces in pumped-storage upper reservoirs, comprising the following steps: Several surveyed upper reservoir areas of pumped storage power stations were selected as sample areas, and exploration sample data of the sample areas were obtained, including surface mapping data, drilling data and physical exploration data. The sample area is meshed into a three-dimensional voxel mesh, and the exploration sample data is preprocessed. Based on the preprocessed data, the exploration features of each voxel in the three-dimensional voxel mesh are extracted. At the same time, a sample label is added to each voxel to indicate whether it belongs to a gently dipping structural surface. The exploration features and corresponding sample labels of each voxel are put into the training sample set. A recognition model based on a machine learning model is constructed, and the recognition model is iteratively trained using a training sample set to obtain a trained probabilistic recognition model of voxel gently tilted structural surfaces. The exploration data of the upper reservoir area of ​​the pumped storage power station to be identified is obtained, gridded, and the exploration features of each voxel in the area are extracted. The data are then input into the voxel gently dipping structural surface probability identification model to obtain the predicted probability of whether each voxel in the area belongs to the gently dipping structural surface. Based on the predicted probability of whether each voxel belongs to the gently dipping structural surface, the gently dipping structural surface in the target area is identified. In a preferred embodiment, the exploration sample data includes: The surface mapping data includes the attitude data of outcrops in the measured area and the marking data of gently dipping structural surfaces marked by artificial means. The marking data of gently dipping structural surfaces is binary data indicating whether the corresponding outcrops belong to gently dipping structural surfaces as marked by the explorers. Drilling data includes core sketches, water injection tests, and drill bit parameters. The physical exploration data includes borehole television measurement data.

[0006] In a preferred embodiment, the step of preprocessing the exploration sample data includes: Implementing coordinate system one specifically includes: Surface mapping data is bound to the geodetic coordinate system via a drone system; Based on the drilling data and the geodetic coordinates of the borehole opening, the core image data and the drilling parameters of the drill bit are bound to the geodetic coordinate system according to the relationship between depth and the geodetic coordinates of the borehole opening. Seismic CT data is converted to the geodetic coordinate system using rigid body transformation based on the coordinates of the survey line.

[0007] In a preferred embodiment, the step of extracting the exploration features of each voxel in the three-dimensional voxel grid based on the preprocessed data specifically includes: Exploration features include mapping features, drilling features, and geophysical features, among which: For mapping features, acquire the attitude data of known surface voxels and the artificially marked gently dipping structural surface marker data corresponding to all outcrops in the region, and calculate the attitude data and gently dipping structural surface marker data of unknown surface voxels. Use the attitude data and gently dipping structural surface marker data as the mapping features of the corresponding voxels. For drilling features, the core type of the corresponding drilling point is identified by the core image data and used as the core image feature. The drilling speed, torque, and pump pressure in the drilling parameter data of the drill bit are extracted, and the drilling mutation quantification index calculated by the drilling speed and torque is used as the drilling feature of the corresponding drilling point. The core image feature and the drilling feature of the same drilling point are stitched together to form the drilling feature of the corresponding drilling point. For any voxel, the drilling feature of the voxel is determined by the drilling feature of the nearest drilling point to the voxel. For geophysical features, the in-hole structural surface quantification features of known drilling points are extracted from borehole television measurement data and stitched together to form the geophysical features of that drilling point. For any voxel, the geophysical features of that voxel are determined by weighted fusion of geophysical features from multiple neighboring drilling points.

[0008] In a preferred embodiment, the step of calculating the attitude data of unknown surface voxels and the marker data of gently dipping structural surfaces includes: Obtaining attitude data includes dip angle and dip direction. The dip angle and dip direction of known surface voxels are converted into unit normal vectors. Kriging interpolation is performed using the unit normal vectors of known surface voxels to obtain the unit normal vectors of each unknown surface voxel. Then, the dip angle and dip direction of the unknown surface voxels are calculated based on the unit normal vectors of each unknown surface voxel. Artificially labeled gently dipping structural surfaces with known surface voxels corresponding to all outcrops in the region are obtained as (0,1) labels. Gaussian logistic regression interpolation is performed on the region based on the (0,1) labels of each known surface voxel to obtain the probability value of each unknown surface voxel belonging to a gently dipping structural surface.

[0009] In a preferred embodiment, the machine learning model employs a 3D U-Net network, and its loss function is set as a composite loss function, specifically as follows: ; ; ; in, For composite loss function, The Dice loss function, The weighted cross-entropy loss function is... and These are the preset Dice weighting coefficients and cross-entropy weighting coefficients, respectively. For sample labels, This represents the predicted probability of whether a voxel belongs to a gently dipping structural surface, as predicted by the model. The preset smoothing constant, The preset positive class weight.

[0010] In a preferred embodiment, the step of identifying gently tilted structural surfaces in the target region based on the predicted probability of whether each voxel belongs to a gently tilted structural surface specifically includes: Suspected voxels are selected by setting a threshold and predicting whether each voxel belongs to a gently tilted structure surface. The selected suspected voxels are then filled with holes and isolated voxels are removed to form at least one continuous three-dimensional suspected structure surface. For each three-dimensional suspected structural surface, starting from any voxel, adjacent voxels are gradually merged based on the similarity of normal vectors to form a single structural surface. For any single structural surface, extract the coordinates of all voxels, and fit the best-fit plane of the single structural surface based on the coordinates of all voxels. The dip angle of the best-fit plane is calculated using the unit normal vector of the best-fit plane, and the dip angle of the best-fit plane is used to determine whether the corresponding individual structural surface is a gently dipping structural surface.

[0011] On the other hand, the present invention also proposes a system for identifying the gently sloping structural surface of a pumped storage upper reservoir, comprising: The raw data acquisition module is used to select several surveyed upper reservoir areas of pumped storage power stations as sample areas and acquire exploration sample data of the sample areas, including surface mapping data, drilling data and physical exploration data. The data processing module is used to mesh the sample area into a three-dimensional voxel grid and preprocess the exploration sample data. Based on the preprocessed data, the exploration features of each voxel in the three-dimensional voxel grid are extracted. At the same time, a sample label is added to each voxel to indicate whether it belongs to a gently dipping structural surface. The exploration features and corresponding sample labels of each voxel are put into the training sample set. The model training module is used to build a recognition model based on a machine learning model. The recognition model is iteratively trained using a training sample set to obtain a trained voxel gently tilted surface probabilistic recognition model. The identification application module is used to acquire exploration data of the upper reservoir area of ​​the pumped storage power station to be identified, perform gridding and extract the exploration features of each voxel in the area, input them into the voxel gentle dip structure surface probability identification model, obtain the predicted probability of whether each voxel in the area belongs to the gentle dip structure surface, and identify the gentle dip structure surface in the target area based on the predicted probability of whether each voxel belongs to the gentle dip structure surface.

[0012] In another aspect, the present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method as described in any embodiment of the present invention.

[0013] In another aspect, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any embodiment of the present invention.

[0014] Additional aspects and advantages of the invention will be set forth in the following description, and some of them will be obvious from the description, or may be learned by practicing the invention. Furthermore, various aspects and advantages of the invention may be realized and obtained by means of method steps and combinations particularly pointed out in the appended claims. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the method flow of Embodiment 1 of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0018] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0019] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0020] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0021] Example 1: See Figure 1This embodiment provides a method for identifying gently dipping structural surfaces in pumped storage reservoirs, including the following steps: S100. Select several pumped storage power station reservoir areas that have undergone detailed surveys as representative sample areas, and collect geological exploration sample data from these sample areas, specifically including high-precision surface mapping data, core and stratigraphic data obtained from drilling, and geophysical data obtained from various physical exploration methods.

[0022] In a preferred embodiment of this invention, in step S100, the exploration sample data includes: The surface mapping data includes the attitude data of outcrops in the measured area and the marking data of gently dipping structural surfaces marked by artificial means. The marking data of gently dipping structural surfaces is binary data indicating whether the corresponding outcrops belong to gently dipping structural surfaces as marked by the explorers. Drilling data includes core sketches, water injection tests, and drill bit parameters. The physical exploration data includes borehole television measurement data.

[0023] S200. The sample area is meshed into a three-dimensional voxel grid, and the exploration sample data is preprocessed. Based on the preprocessed exploration data, representative geological exploration features are extracted from the three-dimensional voxel grid one voxel at a time. These features can reflect relevant information about rock mass structure and geological structure. At the same time, for each voxel, based on its spatial attributes and geological expert interpretation, it is determined whether it belongs to a gently dipping structural surface and is assigned a corresponding sample label. Finally, the exploration features corresponding to each voxel and the labeled sample labels are integrated into a complete training sample and stored in the training sample set.

[0024] In a preferred embodiment of this invention, step S200, which involves meshing the sample region into a three-dimensional voxel mesh, specifically involves: S210. Identify anchor points of interest in the sample area, which may be boreholes that have exposed gently dipping structural surfaces, geophysical anomaly areas (such as low resistivity anomaly areas, high resistivity anomaly areas), outcrops / traces of surface structural surfaces, fracture traces discovered by borehole television, etc.

[0025] S211. Based on the determined anchor point of interest, determine the region of interest. If the anchor point is a single point, the region of interest can be a three-dimensional region centered on the anchor point (e.g., a three-dimensional region with a radius of 10m and the borehole as the center; or a region centered on the outcrop, extending 5m outward in the X and Y directions and 5m downward in depth). If the anchor point is a geophysical anomaly area, the region of interest can directly adopt the vector boundary of the geophysical anomaly area.

[0026] S212. Define non-uniform voxel partitioning rules, set up multi-level regions, and each region uses a different voxel size. The closer to the center of the region of interest, the smaller the voxel size; the farther away from the center of the region of interest, the larger the voxel size. For example: The region within the region of interest, and whose spatial distance from the center of the region of interest is less than or equal to 3m, is defined as the core region. The voxel size within this region is set to 0.5m × 0.5m × 0.25m. The region within the region of interest, and whose spatial distance from the center of the region of interest is greater than 3m, is defined as the transition region. The voxel size in this region is set to 1m×1m×0.5m. The region that is not within the region of interest is defined as the outer region, and the voxel size in this region is set to 2m×2m×1m.

[0027] S213. Using a 3D meshing tool, the sample area is meshed according to the voxel division rules defined in step S212. First, a global mesh is generated based on the voxel size of the outer area. Then, the mesh in the area of ​​interest is replaced with smaller voxels according to the corresponding rules. The voxel boundaries between the three levels of areas are treated with a gradient transition to avoid mesh gaps.

[0028] By following the steps above, the grid density is aligned with the more valuable areas, significantly reducing the workload of regional gridding while ensuring the recognition accuracy of the core areas.

[0029] In a preferred embodiment of this example, step S200, the step of preprocessing the exploration sample data, includes: S220. Implement coordinate system one, specifically including: S221. Bind the surface mapping data to the geodetic coordinate system using an unmanned aerial vehicle system; S222. Based on the drilling data and the geodetic coordinates of the borehole opening, bind the core image data and the drilling parameter data of the drill bit to the geodetic coordinate system according to the relationship between depth and geodetic coordinates of the borehole opening. S223. Using the borehole opening geodetic coordinates as a reference, and combining the borehole depth, borehole dip / azimuth angle, and borehole circumferential orientation, map the two-dimensional image features of the borehole wall (such as structural surfaces and cracks) to the three-dimensional geodetic coordinate system.

[0030] In a preferred embodiment of this example, step S200, which involves extracting the exploration features of each voxel in the three-dimensional voxel grid based on the preprocessed data, specifically comprises: Exploration features include mapping features, drilling features, and geophysical features, among which: S231. For mapping features, acquire the attitude data of known surface voxels and the artificially marked gently dipping structural surface marker data corresponding to all outcrop points in the region, and deduce the attitude data and gently dipping structural surface marker data of unknown surface voxels, using the attitude data and gently dipping structural surface marker data as the mapping features of the corresponding voxels; wherein, the steps of deduce the attitude data and gently dipping structural surface marker data of unknown surface voxels include: Obtaining attitude data includes dip angle and dip direction. The dip angle and dip direction of known surface voxels are converted into unit normal vectors. Kriging interpolation is performed using the unit normal vectors of known surface voxels to obtain the unit normal vectors of each unknown surface voxel. Then, the dip angle and dip direction of the unknown surface voxels are calculated based on the unit normal vectors of each unknown surface voxel. Artificially labeled gently dipping structural surfaces with known surface voxels corresponding to all outcrops in the region are obtained as (0,1) labels. Gaussian logistic regression interpolation is performed on the region based on the (0,1) labels of each known surface voxel to obtain the probability value of each unknown surface voxel belonging to a gently dipping structural surface.

[0031] S232. For drilling features, the core type of the corresponding drilling point is identified by the core image data and used as the core image feature. The drilling speed, torque, and pump pressure in the drilling parameter data of the drill bit are extracted, and the drilling mutation quantification index calculated by the drilling speed and torque is used as the drilling feature of the corresponding drilling point. The core image feature and the drilling feature of the same drilling point are stitched together to form the drilling feature of the corresponding drilling point. For any voxel, the drilling feature of the voxel is determined by the drilling feature of the drilling point closest to the voxel. S233. For geophysical features, the in-hole structural surface quantification features of known drilling points are extracted from borehole television measurement data and stitched together to form the geophysical features of the drilling point. For any voxel, the geophysical features of the voxel are determined by weighted fusion of geophysical features from multiple neighboring drilling points.

[0032] S300. Construct a recognition model based on a machine learning model, and use the training sample set to iteratively train the recognition model to obtain a trained voxel gently tilting structure surface probability recognition model.

[0033] In a preferred embodiment of this invention, the machine learning model employs a 3D U-Net network, and its loss function is set as a composite loss function, specifically: ; ; ; in, For composite loss function, The Dice loss function, The weighted cross-entropy loss function is... and These are the preset Dice weighting coefficients and cross-entropy weighting coefficients, respectively. For sample labels, This represents the predicted probability of whether a voxel belongs to a gently dipping structural surface, as predicted by the model. The preset smoothing constant, The preset positive class weight.

[0034] S400. Obtain detailed exploration data of the upper reservoir area of ​​the pumped storage power station to be identified. Preprocess and mesh the raw data to divide the entire area into a three-dimensional voxel network as in step S200. Then, extract the corresponding exploration features for each voxel unit and input the exploration feature data of each voxel into a pre-trained voxel gently dipping structure surface probability identification model to obtain the predicted probability of whether each voxel in the area belongs to a gently dipping structure surface. Finally, based on the predicted probability results of all voxel units, identify the specific location and distribution range of the gently dipping structure surfaces in the target area. In a preferred embodiment of this example, the step of identifying gently tilted structural surfaces in the target region based on the predicted probability of whether each voxel belongs to a gently tilted structural surface specifically includes: Suspected voxels are selected by setting a threshold and predicting whether each voxel belongs to a gently tilted structure surface. The selected suspected voxels are then filled with holes and isolated voxels are removed to form at least one continuous three-dimensional suspected structure surface. For each three-dimensional suspected structural surface, starting from any voxel, adjacent voxels are gradually merged based on the similarity of normal vectors to form a single structural surface. For any single structural surface, extract the coordinates of all voxels, and fit the best-fit plane of the single structural surface based on the coordinates of all voxels. The dip angle of the best-fit plane is calculated using the unit normal vector of the best-fit plane, and the dip angle of the best-fit plane is used to determine whether the corresponding individual structural surface is a gently dipping structural surface.

[0035] Example 2: This embodiment proposes a system for identifying gently sloping structural surfaces in pumped storage upper reservoirs, including: The raw data acquisition module is used to select several surveyed upper reservoir areas of pumped storage power stations as sample areas and acquire exploration sample data of the sample areas, including surface mapping data, drilling data and physical exploration data; this module is used to implement the function of step S100 in Embodiment 1, which will not be described in detail here. The data processing module is used to mesh the sample area into a three-dimensional voxel mesh, preprocess the exploration sample data, extract the exploration features of each voxel in the three-dimensional voxel mesh based on the preprocessed data, add a sample label to each voxel indicating whether it belongs to a gently dipping structural surface, and put the exploration features and corresponding sample labels of each voxel into the training sample set. This module is used to implement the function of step S200 in Embodiment 1, which will not be described in detail here. The model training module is used to construct a recognition model based on a machine learning model. Iterative training of the recognition model is performed using a training sample set to obtain a trained probabilistic recognition model of a voxel gently tilted structure surface. This module is used to implement the function of step S300 in Example 1, which will not be described in detail here. The identification application module is used to acquire exploration data of the upper reservoir area of ​​the pumped storage power station to be identified, perform gridding and extract the exploration features of each voxel in the area, input them into the voxel gently dipping structure surface probability identification model, obtain the predicted probability of whether each voxel in the area belongs to the gently dipping structure surface, and identify the gently dipping structure surface in the target area based on the predicted probability of whether each voxel belongs to the gently dipping structure surface; this module is used to implement the function of step S400 in Embodiment 1, which will not be described in detail here.

[0036] Example 3: This embodiment proposes an electronic 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, it implements the method described in any embodiment of the present invention.

[0037] Example 4: This embodiment proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in any embodiment of the present invention.

[0038] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0039] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0040] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0041] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0042] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for identifying gently dipping structural surfaces in pumped-storage upper reservoirs, characterized in that, Includes the following steps: Several surveyed upper reservoir areas of pumped storage power stations were selected as sample areas, and exploration sample data of the sample areas were obtained, including surface mapping data, drilling data and physical exploration data. The sample area is meshed into a three-dimensional voxel mesh, and the exploration sample data is preprocessed. Based on the preprocessed data, the exploration features of each voxel in the three-dimensional voxel mesh are extracted. At the same time, a sample label is added to each voxel to indicate whether it belongs to a gently dipping structural surface. The exploration features and corresponding sample labels of each voxel are put into the training sample set. A recognition model based on a machine learning model is constructed, and the recognition model is iteratively trained using a training sample set to obtain a trained probabilistic recognition model of voxel gently tilted structural surfaces. The exploration data of the upper reservoir area of ​​the pumped storage power station to be identified is obtained, gridded and the exploration features of each voxel in the area are extracted and input into the voxel gently dipped structure surface probability identification model. The predicted probability of whether each voxel in the area belongs to the gently dipped structure surface is obtained, and the gently dipped structure surface in the target area is identified based on the predicted probability of whether each voxel belongs to the gently dipped structure surface. Specifically, the step of extracting the exploration features of each voxel in the three-dimensional voxel grid based on the preprocessed data is as follows: Exploration features include mapping features, drilling features, and geophysical features, among which: For mapping features, acquire the attitude data of known surface voxels and the artificially marked gently dipping structural surface marker data corresponding to all outcrops in the region, and calculate the attitude data and gently dipping structural surface marker data of unknown surface voxels. Use the attitude data and gently dipping structural surface marker data as the mapping features of the corresponding voxels. For drilling features, the core type of the corresponding drilling point is identified by the core image data and used as the core image feature. The drilling speed, torque, and pump pressure in the drilling parameter data of the drill bit are extracted, and the drilling mutation quantification index calculated by the drilling speed and torque is used as the drilling feature of the corresponding drilling point. The core image feature and the drilling feature of the same drilling point are stitched together to form the drilling feature of the corresponding drilling point. For any voxel, the drilling feature of the voxel is determined by the drilling feature of the nearest drilling point to the voxel. For geophysical features, the in-hole structural surface quantification features of known drilling points are extracted from borehole television measurement data and stitched together to form the geophysical features of that drilling point. For any voxel, the geophysical features of that voxel are determined by weighted fusion of geophysical features from multiple neighboring drilling points.

2. The method for identifying the gently sloping structural surface of a pumped-storage upper reservoir according to claim 1, characterized in that, In the exploration sample data: The surface mapping data includes the attitude data of outcrops in the measured area and the marking data of gently dipping structural surfaces marked by artificial means. The marking data of gently dipping structural surfaces is binary data indicating whether the corresponding outcrops belong to gently dipping structural surfaces as marked by the explorers. Drilling data includes core sketches, water injection tests, and drill bit parameters. The physical exploration data includes borehole television measurement data.

3. The method for identifying the gently sloping structural surface of a pumped storage reservoir according to claim 2, characterized in that, The steps for preprocessing the exploration sample data include: Implementing coordinate system one specifically includes: Surface mapping data is bound to the geodetic coordinate system via a drone system; Based on the drilling data and the geodetic coordinates of the borehole opening, the core image data and the drilling parameters of the drill bit are bound to the geodetic coordinate system according to the relationship between depth and the geodetic coordinates of the borehole opening. The borehole television measurement data is mapped to the geodetic coordinate system according to the geodetic coordinate system of the borehole opening.

4. The method for identifying the gently sloping structural surface of a pumped-storage upper reservoir according to claim 1, characterized in that, The steps for calculating the attitude data of unknown surface voxels and the marker data of gently dipping structural surfaces include: Obtaining attitude data includes dip angle and dip direction. The dip angle and dip direction of known surface voxels are converted into unit normal vectors. Kriging interpolation is performed using the unit normal vectors of known surface voxels to obtain the unit normal vectors of each unknown surface voxel. Then, the dip angle and dip direction of the unknown surface voxels are calculated based on the unit normal vectors of each unknown surface voxel. Artificially labeled gently dipping structural surfaces with known surface voxels corresponding to all outcrops in the region are obtained as (0,1) labels. Gaussian logistic regression interpolation is performed on the region based on the (0,1) labels of each known surface voxel to obtain the probability value of each unknown surface voxel belonging to a gently dipping structural surface.

5. The method for identifying the gently dipping structural surface of a pumped-storage upper reservoir according to claim 1, characterized in that, The machine learning model uses a 3D U-Net network, and its loss function is a composite loss function, specifically: ； ； ； in, For composite loss function, The Dice loss function, The weighted cross-entropy loss function is... and These are the preset Dice weighting coefficients and cross-entropy weighting coefficients, respectively. For sample labels, This represents the predicted probability of whether a voxel belongs to a gently dipping structural surface, as predicted by the model. The preset smoothing constant, The preset positive class weight.

6. The method for identifying the gently dipping structural surface of a pumped storage reservoir according to claim 1, characterized in that, The specific steps for identifying gently tilted structural surfaces in the target region based on the predicted probability of whether each voxel belongs to a gently tilted structural surface are as follows: Suspected voxels are selected by setting a threshold and predicting whether each voxel belongs to a gently tilted structure surface. The selected suspected voxels are then filled with holes and isolated voxels are removed to form at least one continuous three-dimensional suspected structure surface. For each three-dimensional suspected structural surface, starting from any voxel, adjacent voxels are gradually merged based on the similarity of normal vectors to form a single structural surface. For any single structural surface, extract the coordinates of all voxels, and fit the best-fit plane of the single structural surface based on the coordinates of all voxels. The dip angle of the best-fit plane is calculated using the unit normal vector of the best-fit plane, and the dip angle of the best-fit plane is used to determine whether the corresponding individual structural surface is a gently dipping structural surface.

7. A system for identifying gently dipping structural surfaces in pumped-storage upper reservoirs, characterized in that, include: The raw data acquisition module is used to select several surveyed upper reservoir areas of pumped storage power stations as sample areas and acquire exploration sample data of the sample areas, including surface mapping data, drilling data and physical exploration data. The data processing module is used to mesh the sample area into a three-dimensional voxel grid and preprocess the exploration sample data. Based on the preprocessed data, the exploration features of each voxel in the three-dimensional voxel grid are extracted. At the same time, a sample label is added to each voxel to indicate whether it belongs to a gently dipping structural surface. The exploration features and corresponding sample labels of each voxel are put into the training sample set. The model training module is used to build a recognition model based on a machine learning model. The recognition model is iteratively trained using a training sample set to obtain a trained voxel gently tilted surface probabilistic recognition model. The identification application module is used to acquire exploration data of the upper reservoir area of ​​the pumped storage power station to be identified, perform gridding and extract the exploration features of each voxel in the area, input them into the voxel gently dipped structure surface probability identification model, obtain the predicted probability of whether each voxel in the area belongs to the gently dipped structure surface, and identify the gently dipped structure surface in the target area based on the predicted probability of whether each voxel belongs to the gently dipped structure surface. Specifically, the step of extracting the exploration features of each voxel in the three-dimensional voxel grid based on the preprocessed data is as follows: Exploration features include mapping features, drilling features, and geophysical features, among which: For mapping features, acquire the attitude data of known surface voxels and the artificially marked gently dipping structural surface marker data corresponding to all outcrops in the region, and calculate the attitude data and gently dipping structural surface marker data of unknown surface voxels. Use the attitude data and gently dipping structural surface marker data as the mapping features of the corresponding voxels. For drilling features, the core type of the corresponding drilling point is identified by the core image data and used as the core image feature. The drilling speed, torque, and pump pressure in the drilling parameter data of the drill bit are extracted, and the drilling mutation quantification index calculated by the drilling speed and torque is used as the drilling feature of the corresponding drilling point. The core image feature and the drilling feature of the same drilling point are stitched together to form the drilling feature of the corresponding drilling point. For any voxel, the drilling feature of the voxel is determined by the drilling feature of the nearest drilling point to the voxel. For geophysical features, the in-hole structural surface quantification features of known drilling points are extracted from borehole television measurement data and stitched together to form the geophysical features of that drilling point. For any voxel, the geophysical features of that voxel are determined by weighted fusion of geophysical features from multiple neighboring drilling points.

8. An electronic 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 program, it implements the method for identifying the gently tilted structural surface of a pumped storage reservoir as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the method for identifying the gently tilted structural surface of a pumped storage reservoir as described in any one of claims 1 to 6.

Images