A water conservancy digital twin real-time modeling data processing method and system

Abstract

The application provides a kind of water conservancy digital twin real-time modeling data processing method and system, it is related to data processing technical field, the method includes: according to the optimized space reference surface, extraction analog quantity monitoring signal and state quantity monitoring signal, extract wavelet energy entropy and wavelet singularity through wavelet transform, and construct equipment state feature set in combination with state quantity monitoring signal;According to the basic probability distribution model of equipment state feature set, the evidence base element is obtained;Based on the distance measurement between evidence base element, the reliability weight is calculated, the evidence base element is weighted and fused based on the reliability weight, and the fusion confidence distribution result is obtained;According to the fusion confidence distribution result, drive water conservancy digital twin to carry out real-time update, realize the synchronization and real-time modeling of water conservancy project state.The application can effectively improve the geometric shape fitting effect of water conservancy digital twin under complex terrain conditions and the reliability of equipment state diagnosis.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for real-time modeling data processing of water conservancy digital twins. Background Technology

[0002] In the process of digital twin modeling of water conservancy projects, constructing a spatial reference surface that can accurately reflect the complex terrain features is an important foundation for realizing the expression of the engineering structure morphology and state simulation. Existing technologies typically preprocess multi-source heterogeneous monitoring data to form a standardized spatial sampling point matrix, and then construct a continuous reference surface based on global interpolation or surface fitting methods. However, in actual water conservancy projects, such as in areas with steep slopes, complex geological structures at dam shoulders, and densely distributed spillway dam sections, the surface morphology often exhibits significant local curvature variations and non-uniform geometric distribution. Taking a large-scale water conservancy project as an example, the slope of the banks on both sides of the dam site varies greatly, and the dam shoulders... There is a significant geometric transition at the boundary with the bank slope, and the connection between the dam surface and the flood discharge facilities involves multi-scale structural connections. In such scenarios, conventional global interpolation methods (such as inverse distance weighted or ordinary kriging) often struggle to fully account for the detailed features of areas with drastic local morphological changes in order to maintain the overall smoothness of the surface. This may lead to a certain deviation between the constructed spatial reference surface and the physical entity at slope turning points or structural connection points. This deviation may gradually become apparent in subsequent block partitioning, sampling point mapping, and parameter inversion processes, thereby affecting the expression of the digital twin in terms of engineering geometry and structural response. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method and system for real-time modeling data processing of water conservancy digital twins, which can effectively improve the geometric shape fitting effect and equipment status diagnosis reliability of water conservancy digital twins under complex terrain conditions.

[0004] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a method for real-time modeling data processing of water conservancy digital twins, the method comprising: Step 1: Acquire multi-source heterogeneous data, preprocess the multi-source heterogeneous data to obtain a standardized spatial sampling point matrix; Step 2: Construct a spatial reference surface based on the standardized spatial sampling point matrix; extract geometric primitive regions on the spatial reference surface and subdivide the geometric primitive regions to obtain multiple subdivided blocks; Step 3: Based on multiple segmented molecular blocks, map each sampling point in the standardized spatial sampling point matrix to the corresponding segmented molecular block, and obtain the morphological correction coefficient based on the block characteristics of each segmented molecular block; use the morphological correction coefficient to iteratively optimize the parameters of the spatial reference surface to obtain the optimized spatial reference surface. Step 4: Based on the optimized spatial reference surface, extract analog monitoring signals and state monitoring signals, extract wavelet energy entropy and wavelet singularity through wavelet transform, and construct the equipment state feature set by combining the state monitoring signals; Step 5: Construct a basic probability allocation model based on the equipment status feature set to obtain evidence primitives; calculate the credibility weights based on the distance metric between evidence primitives; and perform weighted fusion of evidence primitives based on the credibility weights to obtain the fusion credibility allocation result. Step 6: Based on the fusion confidence allocation results, drive the water conservancy digital twin to update in real time, so as to realize the synchronization and real-time modeling of the water conservancy project status.

[0005] Secondly, a real-time modeling data processing system for water conservancy digital twins includes: The data preprocessing module is used to acquire multi-source heterogeneous data, preprocess the multi-source heterogeneous data, and obtain a standardized spatial sampling point matrix; The subdivision module is used to construct a spatial reference surface based on a standardized spatial sampling point matrix; extract geometric primitive regions on the spatial reference surface, and subdivide the geometric primitive regions to obtain multiple subdivided blocks; The optimization module is used to map each sampling point in the standardized spatial sampling point matrix to the corresponding segmented molecular block based on multiple segmented molecular blocks, obtain the morphological correction coefficient based on the block characteristics of each segmented molecular block, and use the morphological correction coefficient to iteratively optimize the parameters of the spatial reference surface to obtain the optimized spatial reference surface. The extraction module is used to extract analog monitoring signals and state monitoring signals based on the optimized spatial reference surface, extract wavelet energy entropy and wavelet singularity through wavelet transform, and construct a set of equipment state features by combining the state monitoring signals. The calculation module is used to construct a basic probability allocation model based on the device state feature set to obtain evidence primitives; calculate the credibility weight based on the distance metric between evidence primitives; and perform weighted fusion of evidence primitives based on the credibility weight to obtain the fusion credibility allocation result. The update module is used to drive the water conservancy digital twin to update in real time based on the fusion confidence allocation results, so as to realize the synchronization and real-time modeling of the water conservancy project status.

[0006] Thirdly, a computing device, comprising: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0007] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0008] The above-described solution of the present invention has at least the following beneficial effects: By calculating the weighted coordinate deviation and central moment parameter between the sampling points and the regional center point, the subdivision boundary is dynamically adjusted to obtain subdivided blocks that fit the local geometric features. Based on the local sampling point set of each subdivided block, the normal deviation and tangential deviation are calculated to determine the morphological correction coefficient and iteratively optimize the spatial reference surface. This effectively solves the problem of losing details of drastically changing local areas in order to maintain global smoothness. It is suitable for complex scenarios such as steep slopes, complex geological structures on dam shoulders, and densely distributed floodgate sections. The constructed spatial reference surface can ensure both overall smoothness and fit the local geometric shape. By quantifying the differences between evidence through evidence distance, the credibility weight is assigned based on the degree of difference, making the fusion process more reasonable and improving the reliability of the fusion credibility allocation result. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of a real-time modeling data processing method for water conservancy digital twins provided by an embodiment of the present invention.

[0010] Figure 2 This is a schematic diagram of a real-time modeling data processing system for water conservancy digital twins provided by an embodiment of the present invention. Detailed Implementation

[0011] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0012] like Figure 1 As shown in the figure, an embodiment of the present invention proposes a real-time modeling data processing method for water conservancy digital twins, the method comprising the following steps: Step 1: Acquire multi-source heterogeneous data, preprocess the multi-source heterogeneous data to obtain a standardized spatial sampling point matrix; Step 2: Construct a spatial reference surface based on the standardized spatial sampling point matrix; extract geometric primitive regions on the spatial reference surface and subdivide the geometric primitive regions to obtain multiple subdivided blocks; Step 3: Based on multiple segmented molecular blocks, map each sampling point in the standardized spatial sampling point matrix to the corresponding segmented molecular block, and obtain the morphological correction coefficient based on the block characteristics of each segmented molecular block; use the morphological correction coefficient to iteratively optimize the parameters of the spatial reference surface to obtain the optimized spatial reference surface. Step 4: Based on the optimized spatial reference surface, extract analog monitoring signals and state monitoring signals, extract wavelet energy entropy and wavelet singularity through wavelet transform, and construct the equipment state feature set by combining the state monitoring signals; Step 5: Construct a basic probability allocation model based on the equipment status feature set to obtain evidence primitives; calculate the credibility weights based on the distance metric between evidence primitives; and perform weighted fusion of evidence primitives based on the credibility weights to obtain the fusion credibility allocation result. Step 6: Based on the fusion confidence allocation results, drive the water conservancy digital twin to update in real time, so as to realize the synchronization and real-time modeling of the water conservancy project status.

[0013] In this embodiment of the invention, by calculating the weighted coordinate deviation and central moment parameter between the sampling point and the regional center point, the subdivision boundary is dynamically adjusted to obtain subdivided blocks that fit the local geometric features. Based on the local sampling point set of each subdivided block, the normal deviation and tangential deviation are calculated to determine the morphological correction coefficient and iteratively optimize the spatial reference surface. This effectively solves the problem of losing details of drastically changing local areas in order to maintain global smoothness. It is applicable to complex scenarios such as steep slopes, complex geological structures on dam shoulders, and densely distributed floodgate sections, so that the constructed spatial reference surface can ensure both overall smoothness and fit the local geometric shape. By quantifying the differences between evidence through evidence distance, the credibility weight is assigned based on the degree of difference, making the fusion process more reasonable and improving the reliability of the fusion credibility allocation result.

[0014] In a preferred embodiment of the present invention, step 1 involves acquiring multi-source heterogeneous data, preprocessing the multi-source heterogeneous data to obtain a standardized spatial sampling array, wherein the multi-source heterogeneous data includes three-dimensional spatial scanning data and time-series data on water level, flow rate, gate opening, and rainfall for monitoring operating conditions. Specifically, this includes acquiring multi-source heterogeneous data of the water conservancy project area through a three-dimensional laser scanner, water level sensor, flow rate sensor, gate opening sensor, and rainfall sensor. The multi-source heterogeneous data specifically includes three-dimensional spatial scanning data and time-series data on water level, flow rate, gate opening, and rainfall for monitoring operating conditions. The three-dimensional spatial scanning data is collected by a three-dimensional laser scanner deployed in key areas of the water conservancy project, covering all key structural areas of the water conservancy project, including the dam body, slopes, bank slopes, and flood discharge facilities. During the data collection process, each area is comprehensively scanned according to a preset scanning interval, ranging from 0.05 meters to 0.2 meters. This ensures that the scanned data can completely cover all geometric structures of the project, especially areas with drastic local morphological changes, such as slope turning points, the junction of dam shoulders and bank slopes, and the connection between the dam body and flood discharge facilities, to avoid scanning blind spots. Water level data is collected from real-time water level information of key locations such as reservoirs and rivers in the water conservancy project through water level sensors. Flow data is collected from real-time water flow information of rivers and flood discharge channels through flow sensors. Gate opening data is collected from real-time opening information of each flood discharge gate and intake gate through gate opening sensors. Rainfall data is collected from real-time rainfall information in the project area through rainfall sensors. All time-series data are continuously collected according to a preset collection frequency, ranging from 1 minute to 5 minutes.

[0015] After data acquisition, the acquired multi-source heterogeneous data undergoes preprocessing, which mainly includes four stages: data denoising, data completion, data standardization, and spatial registration. For data denoising, anomalies and noise points caused by scanning interference in the 3D spatial scanning data are identified. A preset deviation threshold range of 0.01 to 0.03 meters is used. Points whose coordinate deviations from adjacent scanning points exceed this threshold are identified as noise points and removed. Valid scanning points with deviations within the threshold range are retained to ensure the accuracy of the 3D spatial scanning data. For random noise in the time-series data of water level, flow rate, gate opening, and rainfall, the values ​​of 3 to 5 adjacent acquisition points are averaged. The calculated average value replaces the abnormally fluctuating values, thus denoising the time-series data and preventing noise data from interfering with subsequent processing.

[0016] Data completion processing addresses missing scan points in 3D spatial scan data due to occlusion, scanning angle limitations, or other reasons. This involves finding valid scan points surrounding the missing point, calculating the distance between the missing point and these valid scan points, selecting the 3 to 5 closest valid scan points, and averaging the coordinates of these valid scan points as the coordinates of the missing point. The average coordinates are calculated as follows: the x-coordinate of the missing point equals the sum of the x-coordinates of the selected valid scan points divided by the number of valid scan points; the y-coordinate of the missing point equals the sum of the y-coordinates of the selected valid scan points divided by the number of valid scan points; and the vertical coordinate of the missing point equals the sum of the vertical coordinates of the selected valid scan points. Divide by the number of valid scan points to complete the missing points; for missing values ​​in time series data, calculate the missing value by linear interpolation based on the valid time series data adjacent to the missing value. That is, select the valid data before and after the missing value, calculate the difference between the two valid data, multiply the difference by the proportion of the position of the missing value to the interval between the two valid data, and add the value of the previous valid data to calculate the missing value and complete it to ensure the continuity of time series data. The method for calculating the proportion of the position of the missing value is the number of acquisition intervals between the missing value and the previous valid data divided by the total number of acquisition intervals between the two valid data.

[0017] Data standardization involves converting the coordinates of the 3D spatial scan data into a pre-defined engineering coordinate system. The pre-defined engineering coordinate system uses the 2000 geodetic coordinate system to ensure that the coordinates of all scan points are under the same reference. For time-series data of water level, flow rate, gate opening, and rainfall, normalization is performed according to their respective measurement ranges, converting all time-series data values ​​to a pre-defined range of 0 to 1. The normalization calculation method is as follows: the normalized value of each time-series data point equals the actual value of that time-series data point minus the minimum value of that type of time-series data, then divided by the maximum and minimum values ​​of that type of time-series data. The difference; spatial registration processing involves spatially associating the preprocessed 3D spatial scanning data with the temporal data. Based on the spatial location of the time-series data acquisition points, the corresponding time-series data is associated with the corresponding coordinate points of the 3D spatial scanning data, forming comprehensive data containing spatial coordinates and temporal condition information. Then, these comprehensive data are gridded and sampled according to a preset sampling density, which ranges from 0.1 meters to 0.3 meters. The sampling method involves uniformly selecting sampling points in the comprehensive data according to the preset sampling density to ensure uniform distribution of sampling points, ultimately forming a standardized spatial sampling point array.

[0018] This embodiment addresses the preprocessing of multi-source heterogeneous data by comprehensively denoising, completing, standardizing, and spatially registering the data, ensuring the accuracy, integrity, and consistency of the standardized spatial sampling point array; it also achieves effective correlation between spatial data and time-series operational data.

[0019] In a preferred embodiment of the present invention, step 2, constructing a spatial reference surface based on a standardized spatial sampling point matrix; extracting geometric primitive regions on the spatial reference surface and subdividing the geometric primitive regions to obtain multiple subdivided blocks, may include: Step 201: Based on the standardized spatial sampling point matrix, construct a spatial reference surface through spatial interpolation and mesh reconstruction; identify key structural regions based on the geometric features of the spatial reference surface, and extract the key structural regions as geometric primitive regions. Specifically, this includes: constructing the spatial reference surface based on the standardized spatial sampling point matrix through spatial interpolation and mesh reconstruction; during spatial interpolation, select all sampling points in the standardized spatial sampling point matrix as interpolation nodes, use the Kriging interpolation method, and set the interpolation weights from 0.01 to 0.99. Based on the coordinates of each interpolation node and the associated working condition data, calculate the coordinates and working condition data at any position between the interpolation nodes. The calculation method is that the coordinates and working condition data at any position are equal to the coordinates of all interpolation nodes. The interpolated data and the working condition data are multiplied by their respective interpolation weights. The interpolation weights are calculated based on the distance between the interpolation node and the arbitrary position. The closer the distance, the greater the weight, ensuring that the interpolated surface can connect all sampling points continuously and smoothly. During the grid reconstruction process, the interpolated continuous data is divided according to the preset grid size, which ranges from 0.2 meters to 0.5 meters, forming a regular spatial grid. Each grid corresponds to a grid cell. The value of each grid cell is calculated by averaging the values ​​of all sampling points in that cell. That is, the value of each grid cell is equal to the sum of the values ​​of all sampling points in that cell divided by the number of sampling points. In this way, a spatial reference surface that can initially reflect the geometric shape and working condition distribution of the water conservancy project is constructed.

[0020] After constructing the spatial reference surface, key structural regions are identified based on its geometric features and extracted as geometric primitive regions. During geometric feature identification, the focus is on analyzing the curvature changes, slope changes, and structural connections of the spatial reference surface. The preset curvature threshold ranges from 0.02 to 0.05 meters, and the preset slope threshold ranges from 30 to 60 degrees. The curvature and slope values ​​of each grid unit on the spatial reference surface are calculated. The curvature value is calculated using the coordinates of the four vertices of the grid unit, and the slope value is calculated using the ratio of the height difference to the horizontal distance of the grid unit. Grid units with curvature and slope values ​​exceeding the corresponding preset thresholds are excluded from the calculation. Candidate units for key structural regions were identified, mainly concentrated in areas with dramatic local morphological changes, such as slope turning points, the junction of dam abutments and bank slopes, and the connection between the dam body and flood discharge facilities. Simultaneously, screening was conducted based on the actual structural distribution of the water conservancy project, specifically including key structural areas such as the connection between the dam body and bank slopes, the connection between the flood discharge gate and the dam body, slope transition sections, the transition area between the dam crest and dam body, and the areas surrounding the inlet and outlet of the water conveyance tunnel. Candidate units were screened against these actual structural distribution locations, eliminating misjudged units that were not within the actual key structural regions due to data errors. The continuous regions formed by the screened candidate units were then designated as key structural regions, and these key structural regions were extracted one by one as geometric primitive regions.

[0021] Step 202: Based on the distribution of the standardized spatial sampling point matrix within the geometric primitive region, calculate the weighted coordinate deviation between each sampling point and the center point of the region. Determine the central moment parameter reflecting the shape characteristics of the region based on the distribution of the weighted coordinate deviation. Specifically, this includes: for each geometric primitive region, determining the center point of the region, calculating the average coordinate of all standardized spatial sampling points within the geometric primitive region, and using this average coordinate as the coordinate of the center point of the geometric primitive region. That is, the x-coordinate of the center point is equal to the sum of the x-coordinates of all sampling points in the region divided by the number of sampling points; the y-coordinate of the center point is equal to the sum of the y-coordinates of all sampling points in the region divided by the number of sampling points; and the y-coordinate of the center point is equal to the sum of the y-coordinates of all sampling points in the region divided by the number of sampling points.

[0022] The weighted coordinate deviation between each sampling point and the center point of the geometric primitive region is calculated. This calculation includes determining the coordinate differences between each sampling point and the center point in the horizontal, vertical, and axial directions. Specifically, the horizontal coordinate difference is obtained by subtracting the horizontal coordinate of the center point from the horizontal coordinate of the sampling point; the vertical coordinate difference is obtained by subtracting the vertical coordinate of the center point from the vertical coordinate of the sampling point; and the axial coordinate difference is obtained by subtracting the vertical coordinate of the center point from the vertical coordinate of the sampling point. Furthermore, different weights are assigned to each sampling point based on its distance from the center point: sampling points closer to the center point have smaller weights, and sampling points farther from the center point have larger weights. The preset base weight is 1.0. The weight values ​​are determined by the inverse relationship of distance, that is, the weight of each sampling point is equal to the preset base weight divided by the distance from the sampling point to the center point. The distance from the sampling point to the center point is calculated by taking the square root of the sum of the squares of the differences in the three coordinate directions. The weight values ​​range from 0.1 to 10. The difference in each coordinate direction is multiplied by the weight of the sampling point to obtain the weighted coordinate deviation of each sampling point in the three coordinate directions. That is, the weighted deviation of the horizontal coordinate is equal to the difference in the horizontal coordinate multiplied by the weight of the sampling point, the weighted deviation of the vertical coordinate is equal to the difference in the vertical coordinate multiplied by the weight of the sampling point, and the weighted deviation of the vertical coordinate is equal to the difference in the vertical coordinate multiplied by the weight of the sampling point.

[0023] The calculation of the central moment parameter is mainly based on the distribution of weighted coordinate deviations. The average value of the weighted coordinate deviations of all sampling points in the three coordinate directions is calculated. The average value in each coordinate direction is equal to the sum of the weighted coordinate deviations of all sampling points in that direction divided by the number of sampling points. Then, the difference between the weighted coordinate deviation of each sampling point and the average value in the corresponding direction is calculated. These differences are squared and then multiplied by their respective sampling point weights. Finally, all the results are added together, and the sum is the central moment parameter of the geometric primitive region. The value range of the central moment parameter is from 0.001 to 100.0.

[0024] Step 203: Generate normalized distribution weights for each sampling point based on the central moment parameter. Adaptively adjust the boundary of the geometric primitive region using these normalized distribution weights to determine the partitioning boundary. Adaptively partition the geometric primitive region along the partitioning boundary to obtain multiple partitioned blocks. Specifically, this includes: generating normalized distribution weights for each sampling point based on the central moment parameter obtained in step 202; during the generation of normalized distribution weights, first calculate the absolute value of the weighted coordinate deviation of each sampling point, then add the absolute values ​​of the weighted coordinate deviations of all sampling points to obtain the total deviation. Divide the absolute value of the weighted coordinate deviation of each sampling point by the total deviation to obtain the initial distribution weight of each sampling point. Then, normalize all the initial distribution weights so that the sum of the normalized distribution weights of all sampling points equals one. That is, the normalized distribution weight of each sampling point is equal to the initial distribution weight of that sampling point divided by the sum of all initial distribution weights. The normalized distribution weight ranges from 0.001 to 0.999, reflecting the importance of the sampling point's distribution within the geometric primitive region.

[0025] The boundary of the geometric primitive region is adaptively adjusted using normalized distribution weights to determine the partitioning boundary. During the adjustment process, the initial boundary of the geometric primitive region is obtained, which is the minimum bounding boundary of the region. The initial boundary is adjusted according to the normalized distribution weight of each sampling point. The preset weight threshold ranges from 0.05 to 0.1. For sampling points with a normalized distribution weight greater than the preset weight threshold, it indicates that the geometric shape at the location of the sampling point has a significant change, and the boundary needs to be adjusted towards that sampling point. The adjustment distance is equal to the normalized distribution weight of the sampling point multiplied by a preset adjustment coefficient, where the preset adjustment coefficient ranges from [value missing]. The boundary is set to 0.1 to 0.3 meters to ensure that the boundary covers the key areas surrounding the sampling point. For sampling points whose normalized distribution weight is less than or equal to the preset weight threshold, it indicates that the geometric shape of the sampling point is relatively gentle. The boundary can be appropriately moved away from the sampling point, and the adjustment distance is also equal to the normalized distribution weight of the sampling point multiplied by the preset adjustment coefficient to avoid overly fine subdivision that would reduce the efficiency of subsequent processing. During the adjustment process, the rationality of the boundary is continuously checked to ensure that the adjusted boundary can completely surround all sampling points within the geometric primitive area and accurately reflect the geometric shape characteristics of the area, thus finally determining a reasonable subdivision boundary.

[0026] The geometric primitive region is adaptively partitioned along a defined partitioning boundary to obtain multiple partitioned sub-blocks. During the partitioning process, the geometric primitive region is divided into multiple continuous sub-regions based on the distribution of the partitioning boundary, and each sub-region is a partitioned sub-block. The principle followed during the partitioning process is that the sampling points within each partitioned sub-block are relatively evenly distributed, and the geometry of each partitioned sub-block is relatively regular. At the same time, it is ensured that the partitioned sub-blocks can accurately reflect the local geometric features of the region. The preset sampling point density threshold ranges from 5 to 10 points per square meter. For regions with a sampling point density greater than the preset density threshold and drastic geometric changes, the number of partitioned sub-blocks is larger and the size is smaller, with the sub-block size ranging from 0.1 meters to 0.2 meters, which can accurately capture local detailed features. For regions with a sampling point density less than or equal to the preset density threshold and gentle geometric changes, the number of partitioned sub-blocks is smaller and the size is larger, with the sub-block size ranging from 0.3 meters to 0.5 meters. Through this adaptive partitioning method, multiple partitioned sub-blocks are obtained.

[0027] In this embodiment, during the construction of the spatial reference surface, the initial surface is constructed by combining the characteristics of the complex geometric structure of the water conservancy project with spatial interpolation and grid reconstruction. Then, the key structural regions are accurately identified as geometric primitive regions, which can focus on the key parts of the project and avoid ignoring local details in the global surface construction.

[0028] In a preferred embodiment of the present invention, step 3 involves mapping each sampling point in the standardized spatial sampling point matrix to the corresponding segmented segment based on multiple segmented segments, obtaining morphological correction coefficients based on the characteristics of each segmented segment, and iteratively optimizing the parameters of the spatial reference surface using the morphological correction coefficients to obtain the optimized spatial reference surface, which may include: Step 301: Based on the boundaries of each segmented sub-block, allocate the sampling points in the standardized spatial sampling point matrix to the corresponding segmented sub-blocks to obtain the local sampling point set within each segmented sub-block; for each segmented sub-block, extract the minimum bounding boundary of the local sampling point set, select initial support line pairs on the minimum bounding boundary, rotate the support lines sequentially and calculate the boundary width in each support line direction, and dynamically adjust the segmentation boundary of the segmented sub-blocks according to the change in boundary width to obtain the optimized segmented sub-block boundary. Specifically, this includes: allocating each sampling point in the standardized spatial sampling point matrix obtained in Step 1 to the corresponding segmented sub-blocks; during the allocation process, determining the coordinates of each sampling point one by one. If the coordinates of a sampling point are within the boundary of a certain segmented region, then the sampling point is assigned to that segmented region. In this way, all sampling points are assigned to their respective segmented regions, resulting in a local sampling point set within each segmented region. (where j is the subdivision block number and n is the number of sampling points within the subdivision block). Each local sampling point set contains only the sampling points within the subdivision block, ensuring that the local sampling point set can accurately reflect the geometric features of the corresponding subdivision block.

[0029] Then, for each segmented region, the minimum bounding boundary of the local sampling point set within that segmented region is extracted. During the extraction of the minimum bounding boundary, the local sampling point set is calculated. Maximum value of the horizontal axis and minimum value The maximum value of the y-axis and minimum value The maximum value of the vertical coordinate and minimum value ,by Starting from the bottom left corner, Assuming the upper right corner is the endpoint, construct a minimal cuboid that can completely enclose the local sampling point set. The surface of this cuboid is the minimum outer boundary of the segmented block.

[0030] To accurately verify the fit between the minimum bounding boundary and the local sampling point set, a convex polygon intersection area calculation algorithm is added. The specific calculation process is as follows: the minimum bounding boundary of the subdivided block is projected onto the horizontal projection plane ( Projecting onto a fixed value, we obtain a convex polygon A, and simultaneously set the local sampling points within the segmented sub-block. First, project the polygons onto the same horizontal projection plane and remove points whose coordinates coincide after projection to obtain convex polygon B. Second, use the Sutherland-Hodgman algorithm to calculate the intersection area of ​​convex polygons A and B. Using convex polygon A as the polygon to be clipped and convex polygon B as the clipping window, sequentially select each edge of convex polygon B as a clipping edge. Initialize the output vertex list to contain all vertices of convex polygon A. (m is the number of vertices in the convex polygon A). For each clipping edge, iterate through the vertices in the output vertex list, and determine whether each vertex (x, y) is inside the clipping edge. Keep the inside vertices. If the current vertex is outside and the previous vertex is inside, calculate the intersection of the line segment containing the current vertex and the previous vertex with the clipping edge. The intersection point is added to the output vertex list. If the current vertex is on the inside and the previous vertex is on the outside, the intersection point is also calculated and added to the output vertex list, while the outside vertex is discarded. After traversing all the clipping edges of the convex polygon B, the polygon formed by the output vertex list is the intersection polygon of A and B. The area of ​​this intersection polygon is calculated using the Shoelace formula. (where t is the number of vertices of the intersecting polygon,) Let be the coordinates of the t-th vertex of the intersecting polygon. Obtain the intersection area of ​​the convex polygons; calculate the intersection area and the area of ​​convex polygon A. ratio ,like If the value is less than 0.95, it indicates that the minimum bounding boundary does not fit the local sampling point set well enough. Return to re-extract the minimum bounding boundary until... 0.95 ensures that the minimum outer boundary can accurately enclose the local sampling point set.

[0031] Initial support line pairs are selected on the minimum outer boundary. During the selection process, a straight line is selected on each of the two opposite faces of the minimum outer boundary as a support line. The two support lines are parallel to each other and can cover the main geometric extension direction of the segmented block. The support lines are then rotated sequentially by a preset rotation angle. The range of values ​​is During the rotation, the direction of the support line is gradually adjusted according to the preset rotation angle. After each rotation, the boundary width in the current support line direction is calculated. The boundary width is calculated by finding the maximum value of all sampling points in the segmented block in the current support line direction. and minimum value Boundary width The partitioning boundaries of the subdivided sub-blocks are dynamically adjusted based on changes in boundary width, with a preset upper limit for boundary width. The value range is from 0.5m to 1.0m, with a preset lower limit for the boundary width. The value range is from 0.1m to 0.2m. During the adjustment process, if the boundary width in a certain support line direction... This indicates that the size of the subdivided block in this direction is too large, and it is necessary to add a dividing line in this direction to further divide the subdivided block into two smaller sub-blocks, reducing the boundary width to... Within the range; if the boundary width in the direction of a certain support line This indicates that the size of the segmented molecular blocks in this direction is too small. Adjacent segmented molecular blocks can be appropriately merged to increase the boundary width. Within the specified range, ensure that the boundary width of each segmented molecular block is within a preset reasonable range; through this dynamic adjustment method, the optimized boundary of the segmented molecular block is obtained.

[0032] Step 302: Based on the optimized boundary of the segmented sub-blocks, redetermine the local sampling point set within each segmented sub-block. Calculate the normal and tangential deviations between each sampling point within each segmented sub-block and the predicted surface shape of the current spatial reference surface on the segmented sub-block. Determine the normal shape correction coefficient of the segmented sub-block based on the numerical distribution of the normal deviation of each sampling point, and determine the tangential shape correction coefficient of the segmented sub-block based on the numerical distribution of the tangential deviation of each sampling point. Specifically, this includes: based on the optimized boundary of the segmented sub-blocks obtained in Step 301, redetermine the local sampling point set within each segmented sub-block. During the redetermining process, it is again determined whether the coordinates of each sampling point are within the optimized boundary of the segmented sub-block. Sampling points within the range are retained in the local sampling point set of the segmented sub-block, while sampling points outside the range are removed, ensuring that the redetermined local sampling point set perfectly matches the optimized boundary of the segmented sub-blocks. In the calculation of normal deviation, the predicted surface shape of the current spatial reference surface on the segmented block is determined. Based on the coordinates of the center point of the segmented block and the coordinates of the surrounding sampling points, the normal vector of the predicted surface at the sampling point is determined. The direction of the normal vector is perpendicular to the predicted surface. The normal vector is calculated by constructing a plane using the coordinates of three adjacent sampling points around the sampling point. The normal vector of this plane is the normal vector of the predicted surface at the sampling point. Then, the difference between the actual coordinates of the sampling point and the coordinates of the predicted surface at that position is calculated. This difference is projected onto the direction of the normal vector. The projected value is the normal deviation of the sampling point. The projection calculation method is to divide the dot product of the difference vector and the normal vector by the magnitude of the normal vector. The normal deviation ranges from -0.1 meters to 0.1 meters. If the projected value is positive, it means that the sampling point is located outside the predicted surface; if it is negative, it means that the sampling point is located inside the predicted surface.

[0033] In the calculation of tangential deviation, the tangential plane of the predicted surface at the sampling point is first determined. The tangential plane is perpendicular to the normal vector and includes the tangent direction of the sampling point on the predicted surface. The difference between the actual coordinates of the sampling point and the coordinates of the predicted surface at that position is calculated. This difference is projected onto the tangential plane, and the projected value is the tangential deviation of the sampling point. The projection calculation method is to subtract the projection vector in the direction of the normal vector from the difference vector. The magnitude of this vector is the tangential deviation, which ranges from 0 to 0.1 meters. Based on the numerical distribution of the normal deviations of each sampling point, the normal shape correction coefficient of the segmented block is determined. The normal deviations of all sampling points within the segmented block are counted, and the absolute values ​​of all normal deviations are calculated. Then, the absolute values ​​of all normal deviations are added together to obtain the sum of normal deviations. The absolute value of the normal deviation of each sampling point is divided by the sum of normal deviations to obtain the normal deviation weight of each sampling point. The normal deviation weight ranges from 0.001 to 0.999. The normal deviation weight of each sampling point is then calculated. The deviation is multiplied by the corresponding normal deviation weight, and all results are summed to obtain the normal shape correction coefficient of the segmented block. The normal shape correction coefficient ranges from -0.1 to 0.1 and reflects the overall deviation of the sampling points in the normal direction within the segmented block. The tangential shape correction coefficient of the segmented block is determined based on the numerical distribution of the tangential deviation of each sampling point. The calculation process is the same as that of the normal shape correction coefficient. The tangential deviation of all sampling points in the segmented block is counted, and the absolute value of all tangential deviations is calculated. The absolute values ​​of all tangential deviations are summed to obtain the total tangential deviation. The absolute value of the tangential deviation of each sampling point is divided by the total tangential deviation to obtain the tangential deviation weight of each sampling point. The tangential deviation weight ranges from 0.001 to 0.999. The tangential deviation of each sampling point is then multiplied by the corresponding tangential deviation weight, and all results are summed to obtain the tangential shape correction coefficient of the segmented block. The tangential shape correction coefficient ranges from 0 to 0.1.

[0034] Step 303: Combine the normal shape correction coefficient and the tangential shape correction coefficient to obtain the shape correction coefficient of the segmented sub-block; use the shape correction coefficient of each segmented sub-block as an optimization variable to iteratively adjust the control points of the spatial reference surface until the fitting error between the overall surface shape and each local sampling point set meets the convergence condition, and use the spatial reference surface at the end of the iteration as the optimized spatial reference surface. Specifically, this includes: combining the normal shape correction coefficient and the tangential shape correction coefficient of each segmented sub-block obtained in step 302 to obtain the shape correction coefficient of that segmented sub-block. During the combination process, the normal shape correction coefficient and the tangential shape correction coefficient are weighted and summed according to a preset ratio. The normal weight ranges from 0.5 to 0.7, and the preset tangential weight ranges from 0.3 to 0.5. The sum of the preset normal weight and the preset tangential weight is 1.0. That is, the shape correction coefficient is equal to the normal shape correction coefficient multiplied by the preset normal weight plus the tangential shape correction coefficient multiplied by the preset tangential weight. The preset normal weight and the preset tangential weight are determined according to the structural characteristics of the water conservancy project. The preset normal weight is 0.7 and the preset tangential weight is 0.3 in the dam area and 0.5 and the preset tangential weight is 0.5 in the slope area. The shape correction coefficient ranges from -0.1 to 0.1 to ensure that the combined shape correction coefficient can fully reflect the overall shape deviation of the segmented sub-block.

[0035] The shape correction coefficient of each segmented sub-block is used as an optimization variable to iteratively adjust the control points of the spatial reference surface. The control points of the spatial reference surface are key nodes selected when constructing the surface. Each control point corresponds to a coordinate value. During the iterative adjustment process, the segmented sub-block corresponding to each control point is determined, and the shape correction coefficient of the segmented sub-block is used as the basis for adjustment. The preset adjustment ratio range is 0.1 to 0.3. If the shape correction coefficient is positive, it means that the sampling points in the segmented sub-block are generally deviated from the outside of the prediction surface, and the corresponding control points need to be adjusted outward. The adjustment range is equal to the shape correction coefficient multiplied by the preset adjustment ratio. That is, the larger the shape correction coefficient, the larger the adjustment range. If the shape correction coefficient is negative, it means that the sampling points in the segmented sub-block are generally deviated from the inside of the prediction surface, and the corresponding control points need to be adjusted inward. The adjustment range is also equal to the absolute value of the shape correction coefficient multiplied by the preset adjustment ratio.

[0036] After each adjustment, the fitting error between the overall surface shape and each local sampling point set is calculated. The fitting error is calculated by dividing the actual coordinates of all sampling points within each subdivided block into the predicted coordinates of the adjusted surface at that location, squaring each difference, and then summing all the squared differences to obtain the total fitting error. The total fitting error ranges from 0 to 0.01 square meters, reflecting the degree of fit between the adjusted surface and the actual sampling points. The calculated total fitting error is then compared with a preset convergence condition, which is that the total fitting error is less than or equal to... A preset error threshold is set, ranging from 0.005 meters to 0.01 meters. This threshold is determined based on the accuracy requirements of hydraulic engineering modeling and can meet the accuracy requirements of the digital twin for geometric representation. If the total fitting error is greater than the preset error threshold, it indicates that the surface fitting accuracy has not met the requirements, and iterative adjustments to the control points are needed, repeating the above adjustment and error calculation process. If the total fitting error is less than or equal to the preset error threshold, it indicates that the surface fitting accuracy has met the requirements, the iteration process terminates, and the spatial reference surface at the end of the iteration is used as the optimized spatial reference surface.

[0037] In this embodiment, the calculation of the morphology correction coefficient and the iterative optimization process of the spatial reference surface, by accurately calculating the normal and tangential morphology correction coefficients of each segmented sub-block, and adjusting the surface control points in a targeted manner, gradually reduces the fitting error between the surface and the actual sampling points, and effectively corrects the deviation of the initial spatial reference surface.

[0038] In a preferred embodiment of the present invention, step 4, extracting analog monitoring signals and state monitoring signals based on the optimized spatial reference surface, extracting wavelet energy entropy and wavelet singularity through wavelet transform, and constructing a device state feature set by combining the state monitoring signals, may include: Step 401: Based on the optimized spatial reference surface, feature response nodes are deployed at key structural locations on the surface. Analog and state monitoring signals corresponding to each feature response node are extracted from multi-source heterogeneous data. The extracted analog monitoring signals are then subjected to multi-level wavelet decomposition to obtain a wavelet coefficient sequence. Specifically, based on the optimized spatial reference surface obtained in Step 303 and considering the key structural characteristics of the hydraulic engineering project, feature response nodes are deployed at key structural locations on the surface. These key structural locations mainly include the dam-slope connection section, the connection between the spillway and the dam body, slope transition sections, the transition area between the dam crest and the dam body, and the areas around the inlet and outlet of the water conveyance tunnel, where geometrical changes are significant. For parts that are severely affected and have a significant impact on engineering safety, at least one characteristic response node should be installed at each critical structural location. After installation, analog quantity monitoring signals and state quantity monitoring signals corresponding to each characteristic response node should be extracted from the multi-source heterogeneous data preprocessed in step 1. The analog quantity monitoring signals include continuously changing monitoring data such as displacement, settlement, stress, strain, water pressure, and flow velocity at each characteristic response node. The state quantity monitoring signals include discrete state data such as equipment operating status and structural integrity status corresponding to each characteristic response node, such as normal equipment operating status, abnormal equipment operating status, structural integrity status, minor structural damage status, and severe structural damage status.

[0039] After extracting the analog monitoring signal, multi-level wavelet decomposition is performed on the analog monitoring signal corresponding to each characteristic response node to obtain the wavelet coefficient sequence. Specifically, the decomposition process involves selecting a suitable wavelet basis function. Considering the continuity and fluctuation characteristics of the analog monitoring signal in water conservancy projects, the db4 wavelet is selected as the basis function. The specific expression of the db4 wavelet is as follows: ,in The wavelet basis function is a Gaussian function, which can effectively capture abrupt changes and detailed information in the monitoring signal. The number of wavelet decomposition levels is determined based on the sampling frequency and data length of the analog monitoring signal. Generally, 5 to 8 levels are selected. If the sampling frequency is 100Hz and the data length is 1024 points, 6 levels are selected to ensure that the high-frequency detail components and low-frequency trend components in the monitoring signal are fully extracted. During the decomposition process, the analog monitoring signal is used as the original signal f(t), and the first level of wavelet decomposition is performed to obtain the low-frequency components of the first level. and high frequency components The decomposition formula is ,in The scaling function for the db4 wavelet. The wavelet function is the db4 wavelet, and * denotes convolution operation; then the low-frequency components of the first layer... Perform a second-level wavelet decomposition to obtain the low-frequency components of the second level. and high frequency components The decomposition formula is This process continues until the preset number of decomposition levels is completed. After each decomposition level, corresponding wavelet coefficients are obtained. The wavelet coefficients corresponding to the low-frequency components are approximation coefficients, and the wavelet coefficients corresponding to the high-frequency components are detail coefficients. The wavelet coefficients of all decomposition levels are arranged in ascending order of decomposition scale. (n is the number of decomposition levels), forming the wavelet coefficient sequence corresponding to the analog monitoring signal.

[0040] Step 402: Calculate the wavelet energy entropy based on the wavelet coefficient sequence; combine the wavelet coefficient sequences of each decomposition scale into a wavelet coefficient matrix according to the scale dimension and time dimension; perform singular value decomposition on the wavelet coefficient matrix to obtain a singular value sequence; determine the wavelet singularity based on the distribution of the singular value sequence; use the wavelet energy entropy and wavelet singularity as analog quantity features and fuse them with the state quantity monitoring signal to obtain the equipment state feature set. Specifically, this includes: calculating the energy of the wavelet coefficients at each decomposition scale. The wavelet energy at each decomposition scale is equal to the sum of the squares of all wavelet coefficients at that scale, which is to square each wavelet coefficient at that scale separately, and then... The wavelet energy at each scale is obtained by summing the results of all squaring operations. The total wavelet energy at all decomposition scales is calculated by summing the wavelet energy at each decomposition scale. The proportion of wavelet energy at each decomposition scale to the total energy is calculated by dividing the wavelet energy at each decomposition scale by the total wavelet energy. The energy proportion at each scale is multiplied by the logarithm of that proportion, and all the results are summed. Finally, the sum is negative, and the result is the wavelet energy entropy. The wavelet energy entropy reflects the energy distribution characteristics of the analog monitoring signal, and thus reflects the stability of the equipment's operating status.

[0041] The wavelet coefficient sequences at each decomposition scale are combined into a wavelet coefficient matrix according to the scale dimension and the time dimension. Specifically, the decomposition scale is used as the rows of the matrix, and the time series of wavelet coefficients at each scale is used as the columns. Each decomposition scale corresponds to one row of the matrix, and each wavelet coefficient at that scale corresponds to one element in that row. The wavelet coefficient sequences from all decomposition scales are sequentially filled into the matrix to form the wavelet coefficient matrix. Singular value decomposition is then performed on the wavelet coefficient matrix to obtain singular value sequences. The decomposition process involves calculating the transpose of the wavelet coefficient matrix and then multiplying the wavelet coefficient matrix by its transpose to obtain a single singular value sequence. The matrix is ​​first constructed, and then the eigenvalues ​​of the matrix are calculated. All eigenvalues ​​are arranged in descending order, and the square root of each eigenvalue is taken. The result is the singular value. All singular values ​​are arranged in order to form a singular value sequence. The wavelet singularity is determined based on the distribution of the singular value sequence. The specific determination process is as follows: the sum of the first three largest singular values ​​in the singular value sequence is calculated, and then the sum of the entire singular value sequence is calculated. The sum of the first three largest singular values ​​is divided by the sum of the singular value sequence. The ratio obtained is the wavelet singularity. The wavelet singularity can reflect the abrupt change characteristics in the analog monitoring signal, corresponding to the abnormal changes in the operating status of the equipment.

[0042] Wavelet energy entropy and wavelet singularity are used as analog features and fused with state monitoring signals to obtain a set of equipment state features. The fusion process is as follows: for each feature response node, the two analog features corresponding to the node, wavelet energy entropy and wavelet singularity, are associated with the state monitoring signal corresponding to the node to ensure that each feature item corresponds to the monitoring data of the same feature response node. The associated features of all feature response nodes are summarized, duplicate feature items are removed, and the analog features of different feature response nodes and state monitoring signals are integrated to form a unified set of equipment state features.

[0043] This embodiment, by deploying characteristic response nodes at key locations on the optimized spatial reference surface, accurately extracts analog and state monitoring signals, and combines wavelet transform to extract wavelet energy entropy and wavelet singularity, thus achieving comprehensive and accurate extraction of equipment state characteristics. It can fully capture the detailed changes in the operation and structural state of water conservancy engineering equipment.

[0044] In a preferred embodiment of the present invention, step 5, constructing a basic probability allocation model based on the device state feature set to obtain evidence primitives; calculating confidence weights based on the distance metric between evidence primitives; and performing weighted fusion of the evidence primitives based on the confidence weights to obtain a fused confidence allocation result, may include: Step 501: Based on the equipment state feature set, construct a basic probability allocation model. Define an identification framework within the basic probability allocation model and treat each feature item in the equipment state feature set as a proposition under the identification framework. Assign a basic probability value to each proposition. Combine each proposition with its corresponding basic probability value to form an evidence primitive. Specifically, this includes: constructing a basic probability allocation model based on the equipment state feature set obtained in step 402. The construction, training, and implementation process of this model is as follows: During the construction process, considering the actual operating and structural states of the water conservancy project equipment, clarify that the core function of the model is to assign probabilities to equipment state features, thereby quantifying the credibility of the equipment state corresponding to each feature item. Determine the model's input and output. The input is all feature items in the equipment state feature set, with each feature item serving as the model's input vector. The output is the basic probability value corresponding to each feature item, ranging from 0 to 1, and the sum of all output basic probability values ​​is one.

[0045] The allocation rules for the model are determined. The allocation rules are pre-defined based on the operational experience of water conservancy projects over the past five years and historical monitoring data. The correlation thresholds between feature items and equipment status are clarified. When the value of a feature item is within the preset normal range, the support level for the corresponding propositions of normal equipment operation and structural integrity is increased. When the value of a feature item exceeds the preset normal range but is within the minor anomaly threshold, the support level for the corresponding propositions of minor equipment anomaly and minor structural damage is increased. When the value of a feature item exceeds the minor anomaly threshold, the support level for the corresponding propositions of severe equipment anomaly and severe structural damage is increased. This ensures that the basic probability assignment can accurately reflect the support level of the feature items for the equipment status.

[0046] During model training, historical monitoring data of water conservancy projects over the past five years were selected as the training dataset. This dataset includes feature data corresponding to the status of various equipment and labeled basic probability values. The training data was preprocessed to remove outliers and missing data, and the feature values ​​were normalized to map to 0 to 1, preventing feature values ​​of different magnitudes from affecting the model's assignment accuracy. The preprocessed training data was then input into the model to initialize its assignment parameters. Gradient descent was used to iteratively optimize the model parameters. During iteration, the mean squared error between the model's output basic probability values ​​and the labeled values ​​was used as the loss function. The loss function was calculated as the reciprocal of the number of training data points multiplied by the first labeled value. The basic probability assignment is calculated by adding the squared difference between the first labeled value and the corresponding model output value to the squared difference between the second labeled value and the corresponding model output value, and so on until the squared difference between the e-th labeled value and the corresponding model output value. Here, the labeled value is the basic probability assignment of the labeled value, the model output value is the basic probability assignment of the model output value, and e is the number of training data. In each iteration, the model parameters are adjusted to reduce the loss function value until the loss function value is less than the preset threshold of 0.001, at which point the iteration stops and the model training is completed. During the model implementation process, the trained model is deployed to the data processing system, receives the input of the device state feature set, and outputs the basic probability assignment corresponding to each feature item through the allocation rules inside the model and the trained parameters, thereby realizing the basic probability allocation function.

[0047] In the basic probability assignment model, an identification framework is defined. This framework is a set of all possible equipment status propositions, specifically including propositions such as normal equipment operation, minor equipment anomaly, severe equipment anomaly, structural integrity, minor structural damage, and severe structural damage, covering all possible equipment and structural states. Each feature item in the equipment status feature set is treated as a proposition within the identification framework, with each feature item corresponding to a unique proposition, ensuring that each feature item can find a corresponding proposition within the framework. A basic probability is assigned to each proposition. Specifically, this assignment process involves determining the degree of support each feature item provides for its corresponding proposition based on historical monitoring data from the past five years. The higher the probability, the larger the basic probability assignment; the lower the support level, the smaller the basic probability assignment. The specific range of the basic probability assignment is as follows: the basic probability assignment range for the propositions of normal equipment operation and structural integrity is 0.1 to 0.4; the basic probability assignment range for the propositions of minor equipment abnormality and minor structural damage is 0.05 to 0.25; and the basic probability assignment range for the propositions of severe equipment abnormality and severe structural damage is 0.01 to 0.15. When assigning the value, the number of times the proposition corresponding to the feature item appears in the historical data is counted, and then the total number of times the feature item appears is counted. The initial probability assignment is obtained by dividing the number of times the proposition corresponding to the feature item appears by the total number of times the feature item appears.

[0048] The initial probability assignments are corrected according to preset correction rules. For feature items that significantly impact equipment status (i.e., stress values ​​and displacement values), the initial probability assignment is increased by 0.05 to 0.1, remaining within the range of the basic probability assignment for the corresponding proposition after correction. For feature items that have a smaller impact on equipment status (i.e., flow velocity and water pressure values), the initial probability assignment is decreased by 0.02 to 0.05, also remaining within the range of the basic probability assignment for the corresponding proposition after correction, ensuring the rationality of the basic probability assignments. The basic probability assignment ranges from 0 to 1, and the sum of the basic probability assignments for all propositions is one; that is, the sum of the basic probability assignments for all six propositions equals one. The basic probability assignment for each proposition is the value corresponding to that proposition. Each proposition and its corresponding basic probability assignment are combined to form evidence primitives. Each evidence primitive contains one proposition and its corresponding basic probability assignment, specifically a combination of the proposition and its corresponding basic probability assignment. All evidence primitives together constitute the evidence set.

[0049] Step 502: Based on the evidence primitives, calculate the evidence distance between every two evidence primitives, and use the evidence distance as a measure of difference between evidence primitives; calculate the credibility coefficient of each evidence primitive based on the evidence distance between each evidence primitive and the remaining evidence primitives. Specifically, this includes: based on all the evidence primitives obtained in Step 501, calculate the evidence distance between every two evidence primitives, and use the evidence distance as a measure of difference between evidence primitives; the calculation process is as follows: select a method for calculating the evidence distance, and combine the characteristics of the evidence primitives, use cosine distance as the calculation method for the evidence distance. This calculation method is preset as the standard calculation method for evidence distance, and is suitable for measuring the difference in basic probability assignments of multiple propositions; for any two evidence primitives, the first evidence primitive contains basic probability assignments corresponding to six propositions, and the second evidence primitive also contains basic probability assignments corresponding to six propositions. Extract the basic probability assignments corresponding to the two evidence primitives, and use the two basic probability assignments as two vectors respectively. That is, the first vector is the basic probability assignments corresponding to the six propositions in the first evidence primitive arranged in sequence, and the second vector is the basic probability assignments corresponding to the six propositions in the second evidence primitive arranged in sequence, and then calculate... To calculate the dot product of two vectors, multiply corresponding elements of each vector and then sum the results. Specifically, multiply the first element of the first vector by the first element of the second vector, add the results of multiplying the first and second elements of the first and second vectors, and so on until the sixth element of the first vector is multiplied by the sixth element of the second vector. To calculate the magnitude of each vector, square each element of the vector, sum the squares, and then take the square root of the sum. Specifically, calculate the magnitude of the first vector... The first vector is the square root of the sum of the squares of all its elements, and the second vector is the square root of the sum of the squares of all its elements. The cosine of the two vectors is obtained by dividing the dot product of the two vectors by the product of their magnitudes. Subtracting this cosine from one gives the evidence distance between the two evidence units. Specifically, the evidence distance is calculated by subtracting (the dot product of the two vectors divided by the product of their magnitudes) from one. The larger the evidence distance, the greater the difference between the two evidence units; the smaller the evidence distance, the higher the consistency between the two evidence units. The evidence distance ranges from 0 to 1.

[0050] Then, based on the evidence distance between each evidence element and the remaining evidence elements, the credibility coefficient of each evidence element is calculated. The credibility coefficient ranges from 0 to 1, and the sum of the credibility coefficients of all evidence elements is one. Specifically, for each evidence element (the evidence element number is the corresponding sequence number), the evidence distance between that evidence element and all other evidence elements is calculated. All calculated evidence distances are summed to obtain the total evidence distance for that evidence element (the total evidence distance is the sum of the evidence distances between that element and every other element). The total evidence distance ranges from 0 to (the total number of evidence elements minus one). The average of the total evidence distances of all evidence elements is calculated, and the total evidence distance of all evidence elements is calculated. The evidence distances are summed and then divided by the total number of evidence units to obtain the average total evidence distance (the average total evidence distance is the sum of the total evidence distances of all units divided by the total number of units). The average total evidence distance ranges from zero to (the total number of evidence units minus one). The average total evidence distance is divided by the total evidence distance of the evidence unit to obtain the initial credibility coefficient of that evidence unit. The initial credibility coefficient ranges from 0 to positive infinity. All initial credibility coefficients are normalized by dividing the initial credibility coefficient of each evidence unit by the sum of all initial credibility coefficients to obtain the final credibility coefficient. The specific calculation method is to divide the initial credibility coefficient of each unit by the sum of the initial credibility coefficients of all units.

[0051] Step 503 involves using the credibility coefficient as a weighting coefficient to perform a weighted average of the basic probability assignments for each evidence element, and then performing a fusion calculation on the weighted average basic probability assignments to obtain the fusion credibility allocation result. Specifically, this includes: using the credibility coefficient obtained in step 502 as a weighting coefficient to perform a weighted average of the basic probability assignments for each evidence element. The weighting coefficient, i.e., the credibility coefficient, ranges from 0 to 1, and the sum of all weighting coefficients is 1. The weighted averaging process specifically involves: for each proposition, collecting all evidence elements containing that proposition, and multiplying the basic probability assignment of that proposition in each evidence element by the value of that proposition. Based on the credibility coefficient corresponding to each primitive, the weighted basic probability assignment for each evidence primitive to the proposition is obtained. Specifically, the basic probability assignment for the proposition in the k-th evidence primitive is multiplied by the credibility coefficient of the k-th evidence primitive, and the value of the weighted basic probability assignment ranges from 0 to 1. The weighted basic probability assignments of all evidence primitives containing the proposition are added together and then divided by the sum of the credibility coefficients of all evidence primitives containing the proposition to obtain the weighted average basic probability assignment for the proposition, which ranges from 0 to 1. For all propositions in the identification framework, the corresponding weighted average basic probability assignment is calculated in the above way.

[0052] The weighted average basic probability assignments are fused to obtain the fused reliability allocation result. Specifically, the fusion calculation process involves selecting the DS evidence fusion rule as the fusion method. This method is pre-set as a standard method for multi-evidence fusion, effectively handling conflicts between multiple pieces of evidence and improving the reliability of the fusion result. The core rule is to achieve multi-evidence fusion by calculating orthogonal sums, eliminating interference from conflicting evidence. During the fusion process, the orthogonal sum of the weighted average basic probability assignments of all evidence primitives is calculated. First, the sum of the products of the weighted average basic probability assignments of all propositions is calculated to obtain the fusion constant. Specifically, this constant is calculated by subtracting the weighted average of all different propositions from 1. The fusion reliability value is calculated by multiplying the weighted average basic probability of each proposition by the reciprocal of the fusion constant, where the fusion constant ranges from 0 to 1. Then, the fusion reliability value for each proposition is obtained by multiplying its weighted average basic probability by the reciprocal of the fusion constant, and then dividing by the sum of the weighted average basic probability of all propositions multiplied by the reciprocal of the fusion constant. The fusion reliability value ranges from 0 to 1, and the sum of all proposition fusion reliability values ​​is 1. After calculation, all propositions and their corresponding fusion reliability values ​​are organized to form the fusion reliability allocation result.

[0053] This embodiment constructs a basic probability allocation model, transforms equipment state features into evidence primitives, calculates credibility weights by combining evidence distance, and performs weighted fusion of evidence primitives. This effectively solves the conflict problem between multi-source evidence and improves the reliability and accuracy of the fusion credibility allocation results.

[0054] In a preferred embodiment of the present invention, step 6, driving the water conservancy digital twin to perform real-time updates based on the fusion confidence allocation result, to achieve synchronization and real-time modeling of the water conservancy project status, may include: Step 601: Based on the fusion reliability allocation result, analyze the reliability value of each state proposition in the fusion reliability allocation result, compare the reliability value of each state proposition with the preset reliability threshold, and select state propositions with reliability values ​​exceeding the reliability threshold as active state propositions. Specifically, this includes: based on the fusion reliability allocation result obtained in step 503, analyzing the reliability value of each state proposition and determining its reliability level. During the analysis process, each state proposition and its corresponding fusion reliability value are extracted one by one and organized according to a one-to-one correspondence. The reliability value of each state proposition is compared with a preset reliability threshold. The preset reliability threshold is determined based on the accuracy requirements of water conservancy project safety monitoring and operational experience, with a specific value range of 0.65 to 0.85. 0.7 is usually selected as the standard reliability threshold, which can effectively distinguish between valid and invalid state propositions. Among them, those with a reliability value greater than 0.7 are valid state propositions, and those less than or equal to 0.7 are invalid state propositions. During the comparison, the reliability value of each state proposition is compared with the reliability threshold one by one. Those with a reliability value exceeding the threshold are selected as active state propositions; those that do not exceed the threshold are removed and do not participate in subsequent parameter adjustments. After the selection is completed, all active state propositions are summarized to form an active state proposition set.

[0055] Step 602: Based on the activation state proposition, find the corresponding water conservancy digital twin parameter adjustment item from the preset state-parameter mapping relationship for each activation state proposition. Combine all the found parameter adjustment items to form a parameter adjustment instruction set. Specifically, this includes: based on the activation state proposition obtained in step 601, calling the preset state-parameter mapping relationship. The specific construction process of this mapping relationship is as follows: sort out the core parameter system of the water conservancy digital twin, clarify the key parameters in the digital twin that are related to the actual state of the project, mainly including coordinate parameters, morphological parameters, and operational parameters. Among them, coordinate parameters cover the spatial coordinates of key structures such as dam body, floodgate, and water conveyance tunnel; morphological parameters cover morphological features such as structural curvature, slope, and cross-sectional dimensions; and operational parameters cover operational indicators such as equipment operating power, speed, and pressure. Clarify the value range, accuracy requirements, and correlation logic of each parameter with the actual state of the project. Summarize all possible equipment and structural state propositions of the water conservancy project, namely, normal equipment operation, minor equipment abnormality, serious equipment abnormality, structural integrity, and structural minor abnormality. The system addresses six key issues: minor damage, severe structural damage, and more. It clarifies the actual engineering manifestations of each issue. For example, minor structural damage corresponds to local dam displacement exceeding the normal range but not exceeding the safety threshold, and slight curvature deviation. The system establishes a correlation between these status issues and parameter adjustment items. Based on the past five years of operational data from water conservancy projects and structural design standards, it clarifies the parameter adjustment requirements for each status issue. For instance, minor equipment anomalies require adjusting operating parameters to within ±3% of the rated range, while minor structural damage requires adjusting coordinate parameters to within ±3 cm and curvature to within ±0.005. The system verifies and optimizes this correlation by selecting engineering operation cases from the past three years and substituting the status issues from these cases into the initially constructed correlation. This verifies the rationality and effectiveness of the parameter adjustment items. If the deviation between the adjusted digital twin and the actual engineering status exceeds 3%, the adjustment range in the correlation is revised until the deviation meets the requirements. Finally, a standardized mapping relationship between status and parameters is established, clarifying the logical correspondence between each status issue and its corresponding parameter adjustment item and adjustment range. This mapping relationship includes the correspondence between all possible state propositions and their corresponding digital twin parameter adjustment items. Each state proposition corresponds to one or more parameter adjustment items. These parameters can be adjusted to ensure that the digital twin is consistent with the actual state of the project. The adjustment range for coordinate parameters is ±5 cm of the actual coordinates of the project. Among the morphological parameters, the adjustment range for curvature is ±0.01 of the actual curvature, the adjustment range for slope is ±0.5 degrees of the actual slope, and the adjustment range for operating parameters is ±5% of the rated operating parameters of the equipment. From the preset state-parameter mapping relationship, the parameter adjustment item corresponding to each active state proposition is searched one by one. The search process uses the active state proposition as the search keyword and accurately matches it in the mapping relationship to ensure that each active state proposition can find a corresponding parameter adjustment item. If no item is found, the corresponding parameter adjustment item is temporarily determined according to the emergency adjustment rules, combined with the parameter standards and structural design requirements of the water conservancy project equipment. After the search is completed, all parameter adjustment items are summarized, duplicates are removed, and they are sorted by digital twin structure and parameter type to form a parameter adjustment instruction set.

[0056] Step 603: The parameter adjustment instruction set is sent to the water conservancy digital twin. The water conservancy digital twin updates its internal state parameters according to the parameter adjustment instruction set, completing the synchronization and real-time modeling of the water conservancy project status. Specifically, this includes: sending the parameter adjustment instruction set obtained in step 602 to the water conservancy digital twin. The sending process is executed according to a preset dedicated communication specification. This dedicated communication specification is constructed in conjunction with industrial data communication and control network technology standards, adapting to the instruction transmission requirements of the water conservancy digital twin. Specific content includes communication interface specifications, using a dedicated wired communication interface for the digital twin, with the interface model matching the interface of the digital twin control unit, supporting high-speed data transmission and strong anti-interference capabilities, adapting to the complex environment of the water conservancy project site; transmission method specifications, explicitly using wired transmission, prioritizing the industrial Ethernet transmission protocol to ensure that instruction transmission is not affected by factors such as on-site electromagnetic interference and environmental humidity; transmission parameters. The specifications include a preset transmission rate of 1000 megabytes per second, a transmission delay controlled within 50 milliseconds, and a command transmission error rate not exceeding 10 to the power of -6. Data verification specifications employ Cyclic Redundancy Check (CRC) to perform real-time verification of the transmitted parameter adjustment command set. If verification passes, transmission is complete; if verification fails, transmission is immediately re-encoded until verification passes. Transmission security specifications encrypt command data using a dedicated encryption algorithm to prevent tampering or theft, ensuring secure command transmission. Based on these specifications, commands are transmitted to the control unit via a wired connection through a dedicated communication interface of the digital twin, ensuring stable and accurate command transmission and preventing loss or errors. After transmission, the control unit receives the command set, parses each parameter adjustment item, and clarifies the parameter name, adjustment direction, and requirements. The parsing process follows a standardized procedure to ensure that the parsing results are completely consistent with the command requirements.

[0057] The water conservancy digital twin updates its internal state parameters according to the parameter adjustment instruction set. The update process follows a preset real-time update procedure, adjusting the corresponding parameters one by one as required by the instructions: after the coordinate parameters are adjusted, the geometric shape of the digital twin is consistent with the actual engineering, with an error of no more than 5 cm; after the shape parameters are adjusted, the curvature error is no more than 0.01 and the slope error is no more than 0.5 degrees; after the operating parameters are adjusted, the actual operating state of the engineering equipment is simulated, with an error of no more than 3% of the rated parameters. During the update process, the adjustment effect is monitored in real time at a frequency of 10 times / second to ensure that each parameter adjustment meets the standards. If the standards are not met, they are readjusted in time until the requirements are met. After the parameter adjustment is completed, the internal state parameters of the water conservancy digital twin are synchronized with the actual operation and structural state of the engineering, completing the synchronization of the engineering state and real-time modeling.

[0058] This embodiment filters activation state propositions by parsing the fusion confidence allocation results, finds the corresponding parameter adjustment items, and issues adjustment instructions, thereby realizing the real-time updating of the water conservancy digital twin. This ensures that the state of the digital twin can be synchronized with the actual operation and structural state of the water conservancy project, improving the accuracy and timeliness of real-time modeling of the digital twin.

[0059] like Figure 2 As shown, embodiments of the present invention also provide a real-time modeling data processing system for water conservancy digital twins, comprising: The data preprocessing module is used to acquire multi-source heterogeneous data, preprocess the multi-source heterogeneous data, and obtain a standardized spatial sampling point matrix; The subdivision module is used to construct a spatial reference surface based on a standardized spatial sampling point matrix; extract geometric primitive regions on the spatial reference surface, and subdivide the geometric primitive regions to obtain multiple subdivided blocks; The optimization module is used to map each sampling point in the standardized spatial sampling point matrix to the corresponding segmented molecular block based on multiple segmented molecular blocks, obtain the morphological correction coefficient based on the block characteristics of each segmented molecular block, and use the morphological correction coefficient to iteratively optimize the parameters of the spatial reference surface to obtain the optimized spatial reference surface. The extraction module is used to extract analog monitoring signals and state monitoring signals based on the optimized spatial reference surface, extract wavelet energy entropy and wavelet singularity through wavelet transform, and construct a set of equipment state features by combining the state monitoring signals. The calculation module is used to construct a basic probability allocation model based on the device state feature set to obtain evidence primitives; calculate the credibility weight based on the distance metric between evidence primitives; and perform weighted fusion of evidence primitives based on the credibility weight to obtain the fusion credibility allocation result. The update module is used to drive the water conservancy digital twin to update in real time based on the fusion confidence allocation results, so as to realize the synchronization and real-time modeling of the water conservancy project status.

[0060] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0061] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0062] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0063] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for real-time modeling data processing of water conservancy digital twins, characterized in that, The method includes: Step 1: Acquire multi-source heterogeneous data, preprocess the multi-source heterogeneous data to obtain a standardized spatial sampling point matrix; Step 2: Construct a spatial reference surface based on the standardized spatial sampling point matrix; extract geometric primitive regions on the spatial reference surface and subdivide the geometric primitive regions to obtain multiple subdivided blocks. The subdivision of the geometric primitive regions to obtain multiple subdivided blocks includes: calculating the weighted coordinate deviation between each sampling point and the center point of the region based on the distribution of the standardized spatial sampling point matrix within the geometric primitive regions; determining the central moment parameter reflecting the shape characteristics of the region based on the distribution of the weighted coordinate deviation; generating normalized distribution weights for each sampling point based on the central moment parameter; adaptively adjusting the boundary of the geometric primitive regions using the normalized distribution weights to determine the subdivision boundary; and adaptively subdividing the geometric primitive regions along the subdivision boundary to obtain multiple subdivided blocks. Step 3: Based on multiple segmented sub-blocks, map each sampling point in the standardized spatial sampling point matrix to the corresponding segmented sub-block, and obtain the shape correction coefficient based on the block characteristics of each segmented sub-block; use the shape correction coefficient to iteratively optimize the parameters of the spatial reference surface to obtain the optimized spatial reference surface. The shape correction coefficient obtained based on the block characteristics of each segmented sub-block includes: calculating the normal and tangential deviations between each sampling point in each segmented sub-block and the predicted surface shape of the current spatial reference surface on the segmented sub-block; determining the normal shape correction coefficient of the segmented sub-block based on the numerical distribution of the normal deviation of each sampling point; and determining the tangential shape correction coefficient of the segmented sub-block based on the numerical distribution of the tangential deviation of each sampling point. Step 4: Based on the optimized spatial reference surface, extract analog quantity monitoring signals and state quantity monitoring signals, extract wavelet energy entropy and wavelet singularity through wavelet transform, and construct equipment state feature set by combining state quantity monitoring signals; Step 5: Construct a basic probability allocation model based on the equipment status feature set to obtain evidence primitives; calculate the credibility weights based on the distance metric between evidence primitives; and perform weighted fusion of evidence primitives based on the credibility weights to obtain the fusion credibility allocation result. Step 6: Based on the fusion confidence allocation results, drive the water conservancy digital twin to update in real time, so as to realize the synchronization and real-time modeling of the water conservancy project status.

2. The method for real-time modeling data processing of water conservancy digital twins according to claim 1, characterized in that, The multi-source heterogeneous data includes three-dimensional spatial scanning data as well as water level, flow rate, gate opening, and rainfall time series data used to monitor operating conditions.

3. The method for real-time modeling data processing of water conservancy digital twins according to claim 2, characterized in that, Based on the standardized spatial sampling point matrix, a spatial reference surface is constructed; geometric primitive regions are extracted from the spatial reference surface, including: Based on the standardized spatial sampling point matrix, a spatial reference surface is constructed through spatial interpolation and grid reconstruction; key structural regions are identified based on the geometric features of the spatial reference surface, and these key structural regions are extracted as geometric primitive regions.

4. The method for real-time modeling data processing of water conservancy digital twins according to claim 3, characterized in that, Step 3 includes: Based on the boundaries of each segmented sub-block, the sampling points in the standardized spatial sampling point matrix are assigned to the corresponding segmented sub-blocks to obtain the local sampling point set within each segmented sub-block. For each segmented sub-block, the minimum outer boundary of the local sampling point set is extracted. Initial support line pairs are selected on the minimum outer boundary, and the support lines are rotated sequentially and the boundary width in each support line direction is calculated. The segmentation boundary of the segmented sub-block is dynamically adjusted according to the change in the boundary width to obtain the optimized segmented sub-block boundary. The local sampling point set within each segmented molecular block is redefined based on the optimized segmented molecular block boundary. The normal shape correction coefficient and the tangential shape correction coefficient are combined to obtain the shape correction coefficient of the segmented sub-blocks. The shape correction coefficient of each segmented sub-block is used as the optimization variable to iteratively adjust the control points of the spatial reference surface until the fitting error between the overall surface shape and each local sampling point set meets the convergence condition. The spatial reference surface at the end of the iteration is used as the optimized spatial reference surface.

5. The method for real-time modeling data processing of water conservancy digital twins according to claim 4, characterized in that, Based on the optimized spatial reference surface, analog monitoring signals and state monitoring signals are extracted. Wavelet energy entropy and wavelet singularity are extracted through wavelet transform. Combined with the state monitoring signals, a device state feature set is constructed, including: Based on the optimized spatial reference surface, characteristic response nodes are set up at key structural locations on the surface. Analog and state monitoring signals corresponding to each characteristic response node are extracted from multi-source heterogeneous data. Multi-level wavelet decomposition is performed on the extracted analog monitoring signals to obtain wavelet coefficient sequences. Based on the wavelet coefficient sequence, the wavelet energy entropy is calculated; the wavelet coefficient sequences of each decomposition scale are combined into a wavelet coefficient matrix according to the scale dimension and the time dimension, and the wavelet coefficient matrix is ​​subjected to singular value decomposition to obtain a singular value sequence. The wavelet singularity is determined according to the distribution of the singular value sequence; the wavelet energy entropy and wavelet singularity are used as analog quantity features and fused with the state quantity monitoring signal to obtain the equipment state feature set.

6. The method for real-time modeling data processing of water conservancy digital twins according to claim 5, characterized in that, A basic probability allocation model is constructed based on the equipment state feature set to obtain evidence primitives; Credibility weights are calculated based on distance metrics between evidence primitives. These credibility weights are then used to weight and fuse the evidence primitives to obtain the fused credibility allocation result, including: Based on the equipment state feature set, a basic probability allocation model is constructed. In the basic probability allocation model, an identification framework is defined and each feature item in the equipment state feature set is taken as a proposition under the identification framework. A basic probability value is assigned to each proposition and the basic probability value corresponding to each proposition is combined to form an evidence primitive. Based on the evidence units, calculate the evidence distance between every two evidence units and use the evidence distance as a measure of the difference between the evidence units; calculate the credibility coefficient of each evidence unit based on the evidence distance between each evidence unit and the remaining evidence units. The credibility coefficient is used as a weighting coefficient to calculate the weighted average of the basic probability assignments of each evidence primitive. The weighted average basic probability assignments are then fused to obtain the fused credibility allocation result.

7. The method for real-time modeling data processing of water conservancy digital twins according to claim 6, characterized in that, Based on the fusion confidence allocation results, the water conservancy digital twin is driven to update in real time, realizing the synchronization and real-time modeling of the water conservancy project status, including: Based on the fusion reliability assignment results, the reliability value of each state proposition in the fusion reliability assignment results is analyzed, and the reliability value of each state proposition is compared with the preset reliability threshold. State propositions with reliability values ​​exceeding the reliability threshold are selected as active state propositions. Based on the activation state proposition, the corresponding water conservancy digital twin parameter adjustment item is found from the preset state and parameter mapping relationship, and all the found parameter adjustment items are combined to form a parameter adjustment instruction set. The parameter adjustment instruction set is sent to the water conservancy digital twin, and the water conservancy digital twin updates its internal status parameters according to the parameter adjustment instruction set, thus completing the synchronization and real-time modeling of the water conservancy project status.

8. A real-time modeling data processing system for water conservancy digital twins, wherein the system implements the method as described in any one of claims 1 to 7, characterized in that, include: The data preprocessing module is used to acquire multi-source heterogeneous data, preprocess the multi-source heterogeneous data, and obtain a standardized spatial sampling point matrix; The subdivision module is used to construct a spatial reference surface based on a standardized spatial sampling point matrix; extract geometric primitive regions on the spatial reference surface, and subdivide the geometric primitive regions to obtain multiple subdivided blocks; The optimization module is used to map each sampling point in the standardized spatial sampling point matrix to the corresponding segmented molecular block based on multiple segmented molecular blocks, and to obtain the morphological correction coefficient based on the block characteristics of each segmented molecular block. The parameters of the spatial reference surface are iteratively optimized using morphological correction coefficients to obtain the optimized spatial reference surface; The extraction module is used to extract analog monitoring signals and state monitoring signals based on the optimized spatial reference surface, extract wavelet energy entropy and wavelet singularity through wavelet transform, and construct a device state feature set by combining the state monitoring signals. The calculation module is used to construct a basic probability allocation model based on the device state feature set to obtain evidence primitives; calculate the credibility weights based on the distance metric between evidence primitives; and perform weighted fusion of evidence primitives based on the credibility weights to obtain the fusion credibility allocation result. The update module is used to drive the water conservancy digital twin to update in real time based on the fusion confidence allocation results, so as to realize the synchronization and real-time modeling of the water conservancy project status.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images