Reservoir dam discharge capacity calibration system and method

Abstract

The invention discloses a method and a system for calibrating the discharge capacity of a reservoir dam, and particularly relates to the technical field of dam monitoring. The method comprises the following steps: firstly, constructing a CFD model of a reservoir drainage system, and optimizing the calculation precision of a key region by adopting an adaptive grid technology; then, respectively calculating an error sensitivity coefficient and a dynamic grid adjustment coefficient by utilizing a machine learning algorithm, automatically detecting a numerical diffusion influence region, and judging whether optimization adjustment is triggered or not; in a high-error region, adaptive grid encryption, dynamic time step adjustment and turbulence model optimization are adopted, so that errors are effectively reduced, and the accuracy of discharge capacity calibration is improved; in addition, a discharge capacity curve is generated through multiple rounds of dynamic adjustment, the discharge capacity under the future working condition is predicted in combination with a deep learning model, the calculation efficiency is optimized, and the CFD calculation cost is reduced; finally, a high-precision discharge capacity curve and a flow velocity distribution diagram are output, reliable data support is provided for reservoir scheduling, and the accuracy and safety of flood control scheduling are improved.

Description

Technical Field

[0001] The invention relates to the technical field of dam monitoring, and in particular to a system and method for grading the discharge capacity of a reservoir dam. Background Art

[0002] The discharge capacity of a reservoir dam is directly related to reservoir scheduling, flood control safety and downstream ecological environment. Traditional discharge capacity calibration methods mainly rely on field measurements, physical model tests and empirical formula calculations, but these methods have problems such as long test cycles, high costs and limited applicability. In recent years, the development of computational fluid dynamics (CFD) technology has provided an efficient and accurate numerical simulation method for discharge capacity calibration, which can be used to predict discharge characteristics under different water levels, gate openings and complex working conditions.

[0003] The prior art has the following deficiencies: In the process of CFD-based discharge capacity calibration, numerical diffusion is a problem that is not easy to detect but may seriously affect the calculation accuracy. Numerical diffusion is an artificial viscosity effect caused by discretization errors, which causes the velocity gradient and free surface fluctuations in the flow field to be over-smoothed, thereby underestimating the kinetic energy and turbulence characteristics of high-speed discharge. In the calibration of reservoir discharge capacity, numerical diffusion may lead to the following serious consequences: 1) Underestimation of the actual discharge capacity of the spillway and spillway, making the reservoir scheduling plan too conservative and affecting flood control efficiency; 2) Misjudgment of the hydraulic jump position of the energy dissipation pool, resulting in failure of the optimization design of the energy dissipation structure, causing downstream riverbed scour or structural damage. Summary of the invention

[0004] The purpose of the present invention is to provide a system and method for calibrating the discharge capacity of a reservoir dam to solve the shortcomings of the background technology.

[0005] In order to achieve the above object, the present invention provides the following technical solution: a method for calibrating the discharge capacity of a reservoir dam, comprising: Obtain the geometric parameters of the reservoir dam discharge facilities, including the structural dimensions of the spillway, spillway tunnel, energy dissipation pool and bottom holes, build a CFD model of the reservoir dam discharge system, and use adaptive grids to optimize the calculation accuracy of key areas; Based on the machine learning algorithm, the error sensitivity coefficient and dynamic grid adjustment coefficient in each key area are calculated respectively, and the area affected by numerical diffusion is detected by the calculated error sensitivity coefficient to determine whether to trigger optimization adjustment; For high error areas, dynamic grid adjustment coefficients are used to perform adaptive grid encryption, dynamic time step adjustment, and turbulence model optimization to reduce errors and improve calculation accuracy. Perform multiple rounds of dynamic adjustment under different working conditions to generate a discharge capacity curve, optimize the calculation efficiency in combination with a deep learning prediction model, and output the discharge capacity curve and the flow velocity distribution map to support the optimization of reservoir operation.

[0006] Preferably, the calculation method of the error sensitivity coefficient is as follows: Define the error characteristic variables and construct the data matrix X: ; m is the number of grid cells, and n is the number of error influence variables; Since the value ranges of different variables are different, standardization processing is required to make their mean 0 and variance 1: Calculate the covariance matrix C of the standardized data matrix X': ; where: C is an n×n-dimensional covariance matrix, and T is the matrix transpose; Perform eigenvalue decomposition on the covariance matrix C to obtain the eigenvalues and the corresponding eigenvectors , and the expression is: ; The eigenvalue : represents the variance of the i-th principal component, and the eigenvector represents the direction of the i-th principal component; Select the first k principal components with a cumulative contribution rate reaching 95% for calculation, and calculate the principal component scores: ; where: V is a matrix composed of the selected k eigenvectors, and Z is the data matrix after dimensionality reduction, containing the principal component scores; Define the error sensitivity coefficient as ESC, and the expression is: ; where: is the eigenvalue of the j-th principal component, is the weight coefficient corresponding to the eigenvector.

[0007] Preferably, set the error influence region threshold TESC. If ESC≥TESC, it is considered that the numerical diffusion error in the region is large, and it is marked as a high-error region; if ESC<TESC, it is considered that the numerical diffusion error in the region is small, and it is marked as a low-error region, and record the index set of the grid regions that need to be optimized , and calculate the proportion of the high-error region : ; N represents the total number of grid cells in the CFD calculation model. If the calculated proportion of the high-error region is greater than or equal to the set high-error region proportion threshold, trigger the optimization adjustment.

[0008] Preferably, the method for obtaining the dynamic grid adjustment coefficient is as follows: Construct the data set Y, including the local grid scale, turbulent kinetic energy, free surface fluctuation, and calculation error; Determine the optimal number of clusters w, using the elbow method: ; In the formula, represents the data points belonging to the q-th cluster, is the sum of squared errors within the cluster, represents the set of grid points in the p-th cluster, is the center point of the pth cluster; Randomly select w initial cluster centers and calculate the Euclidean distance from each grid point to all cluster centers , and assign it to the nearest cluster : ; e represents the number of the e-th grid unit in the CFD calculation model, N represents the number of all grid units in the CFD calculation model, and for each cluster , update the mean center: ; If the cluster center no longer changes or the maximum number of iterations is reached, the iteration is terminated; Assign a mesh refinement factor to each cluster : Cluster center error Normalize: ;in Represents the error characteristic component of the pth cluster, h is the number of clusters; combined with the weights of each cluster , calculate the dynamic grid adjustment coefficient DMRC: DMRC .

[0009] Preferably, the adaptive grid encryption controls grid optimization through a dynamic grid adjustment coefficient DMRC, and the optimized grid size calculation formula is: ;in: is the optimized grid cell size, is the initial grid cell size, is the dynamic grid adjustment coefficient of the e-th grid cell.

[0010] Preferably, the dynamic time step adjustment is performed by DMRC calculation to optimize the time step, and the optimized time step calculation formula is: ;in: is the optimized time step, is the original time step, is the maximum DMRC value in the high error region.

[0011] Preferably, the deep learning prediction model is based on a long short-term memory network and includes: The LSTM model is trained using historical discharge capacity data, with input features including water level, gate opening, discharge flow, and flow velocity distribution; The LSTM layer extracts time series features and learns the changing trend of the discharge capacity curve; Calculate the mean square error. If the error exceeds the set threshold, optimize the CFD calculation to improve the prediction accuracy. Update LSTM training data based on CFD calculation results.

[0012] The present invention also provides a reservoir dam discharge capacity rating system, comprising a parameter acquisition module, a data calculation module, a grid adjustment module and a calculation efficiency optimization module; Parameter acquisition module: obtains the geometric parameters of the reservoir dam discharge facilities, including the structural dimensions of the spillway, spillway tunnel, energy dissipation pool and bottom hole, builds the CFD model of the reservoir dam discharge system, and uses adaptive grids to optimize the calculation accuracy of key areas; Data calculation module: Based on the machine learning algorithm, the error sensitivity coefficient and dynamic grid adjustment coefficient in each key area are calculated respectively, and the area affected by numerical diffusion is detected through the calculated error sensitivity coefficient to determine whether to trigger optimization adjustment; Grid adjustment module: For high error areas, dynamic grid adjustment coefficients are used to perform adaptive grid encryption, dynamic time step adjustment, and turbulence model optimization to reduce errors and improve calculation accuracy; Computational efficiency optimization module: Perform multiple rounds of dynamic adjustments under different working conditions to generate a discharge capacity curve, combine deep learning prediction models to optimize computational efficiency, and output discharge capacity curves and flow velocity distribution maps to support reservoir scheduling optimization.

[0013] In the above technical solution, the technical effects and advantages provided by the present invention are: The present invention optimizes the calculation accuracy of key areas through adaptive grid optimization, uses principal component analysis to calculate the error sensitivity coefficient, identifies the area affected by numerical diffusion, and combines K-Means clustering to calculate the dynamic grid adjustment coefficient, automatically optimizing the grid division and calculation parameters. In high-error areas, the grid density, time step and turbulence model are dynamically adjusted to reduce calculation errors and improve the accuracy of discharge capacity determination. In addition, the discharge capacity curve is predicted using a long short-term memory network to optimize CFD calculation efficiency, reduce calculation costs, and output high-precision discharge capacity curves and flow velocity distribution maps to provide data support for reservoir scheduling. On the whole, the present invention improves the accuracy of discharge capacity assessment, calculation stability and scheduling optimization level, and has wide application value in flood control safety, water resources scheduling and other aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0015] Figure 1 The figure is a flow chart of the method of the present invention.

[0016] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Example 1, please refer to Figure 1 As shown, a method for calibrating the discharge capacity of a reservoir dam described in this embodiment includes: Obtain the geometric parameters of the reservoir dam discharge facilities, including the structural dimensions of the spillway, spillway tunnel, energy dissipation pool and bottom holes, build a CFD model of the reservoir dam discharge system, and use adaptive grids to optimize the calculation accuracy of key areas; Based on the machine learning algorithm, the error sensitivity coefficient and dynamic grid adjustment coefficient in each key area are calculated respectively, and the area affected by numerical diffusion is detected by the calculated error sensitivity coefficient to determine whether to trigger optimization adjustment; For high error areas, dynamic grid adjustment coefficients are used to perform adaptive grid encryption, dynamic time step adjustment, and turbulence model optimization to reduce errors and improve calculation accuracy. Perform multiple rounds of dynamic adjustments under different operating conditions to generate a discharge capacity curve, combine deep learning prediction models to optimize computing efficiency, and output the discharge capacity curve and flow velocity distribution map to support reservoir scheduling optimization.

[0019] The steps of constructing the CFD model of the reservoir dam discharge system and optimizing the key areas of the adaptive mesh include: Step 1: Obtain geometric parameters of spillway facilities: Obtain key parameters such as structural dimensions, boundary shapes, slopes, and cross-sectional forms of spillway, spillway tunnel, energy dissipation pool, and bottom holes. Refer to dam construction drawings, topographic survey data, and 3D laser scanning data to improve geometric accuracy. Statistical data on different water levels, gate opening angles, flow rates, and other parameters to clarify calculation boundary conditions. Combine historical water flow observation data to optimize model input parameters.

[0020] Step 2: Construct a CFD model of the reservoir dam spillway system: Use modeling software such as CAD, SolidWorks, and ANSYS Space Claim to create a three-dimensional geometric model based on the structural data of the spillway facility. Simplify the details of non-critical areas (such as small auxiliary structures) to improve calculation efficiency. Determine the calculation scope, including the spillway facility and its surrounding waters, and ensure that the boundary is away from the critical flow area to reduce the impact of reflection. Set the inflow boundary (reservoir area), outflow boundary (downstream river channel), and atmospheric boundary (free surface) to ensure calculation stability. Use tools such as ICEM CFD, Gambit, and Ansys Meshing for preliminary meshing. Use unstructured mesh to improve adaptability in complex geometric areas.

[0021] Step 3: Determine the key optimization areas (grid densification areas) including: High gradient change areas: spillway inlet, bottom hole inlet, around the gate, water velocity and pressure mutation areas. Turbulence and backflow areas: spillway bends, energy dissipation pools, drop sections, and areas with complex flow patterns. Free water surface areas: top of overflow weir, spillway surface, and areas with obvious water-air interaction. Negative pressure and cavitation risk areas: high-speed spillway jet area, bottom of energy dissipation pool, and areas where cavitation may occur.

[0022] Adaptive Meshing Technology (AMR) is used to optimize mesh density: local mesh encryption is used to refine meshes in key areas to improve flow field analysis capabilities. Dynamic Mesh is used to optimize meshes in real time for free water surfaces and turbulent eddy regions. Boundary layer mesh refinement is used to improve near-wall flow calculation accuracy and optimize energy loss calculations.

[0023] Optimize time step and calculation stability: Use adaptive time step control to optimize calculation accuracy and stability. Select a suitable turbulence model (such as RANS-LES hybrid model) to ensure the accuracy of turbulence analysis.

[0024] The PCA algorithm is used to calculate the error sensitivity coefficients in key areas of the reservoir dam discharge system (such as spillways, spillway tunnels, energy dissipation pools, and bottom holes) to identify the variables that mainly affect the error, reduce data dimensions, and improve calculation efficiency.

[0025] The error sensitivity coefficient is calculated by defining the error characteristic variables, including: Grid scale: grid unit size, which affects the calculation accuracy; Flow rate gradient: , affecting the turbulence accuracy; Pressure gradient: , affecting the energy loss calculation; Time step: calculate stability related parameters; Turbulence model error: differences in calculation results of different turbulence models; Free surface fluctuation: numerical error at the water-air interface.

[0026] Construct the data matrix X: ; m is the number of grid cells, and n is the number of error-influencing variables; since the value ranges of different variables are different, standardization processing is required to make their mean 0 and variance 1: Calculate the covariance matrix C of the standardized data matrix X′: ; where: C is an n×n-dimensional covariance matrix, and T is the matrix transpose; Perform eigenvalue decomposition on the covariance matrix C to obtain the eigenvalues and the corresponding eigenvectors , and the expression is: ; the eigenvalue : represents the variance (contribution rate) of the i-th principal component, and the eigenvector represents the direction of the i-th principal component; usually, the first k principal components with a cumulative contribution rate reaching 95% are selected for calculation, and the principal component scores (dimensionality reduction transformation) are calculated: ; where: V is a matrix composed of the selected k eigenvectors (dimensionality reduction transformation matrix), and Z is the data matrix after dimensionality reduction, containing the principal component scores. Define the error sensitivity coefficient as ESC, and the error sensitivity coefficient represents the influence degree of the key area on the overall error, and the expression is: ; where: is the eigenvalue of the j-th principal component, is the weight coefficient corresponding to the eigenvector.

[0027] Set the error influence area threshold TESC (empirical value, such as 0.05 - 0.1). If ESC≥TESC, it is considered that the numerical diffusion error in the area is large, and it is marked as a high-error area; if ESC<TESC, it is considered that the numerical diffusion error in the area is small, and it is marked as a low-error area, and record the set of grid area indices that need to be optimized .

[0028] Calculate the proportion of high-error areas : ; N represents the total number of grid cells in the CFD calculation model. If the calculated proportion of high-error areas is greater than or equal to the set high-error area proportion threshold, trigger optimization adjustment. Perform K-Means clustering on the high-error areas to automatically identify the grid optimization strategy.

[0029] Calculate the dynamic grid adjustment coefficient of the key area of the reservoir dam spillway system through the K-Means clustering algorithm, automatically identify the high-error areas, and allocate appropriate grid densities to improve the CFD calculation accuracy and efficiency.

[0030] Define feature variables and construct data set Y: Local grid scale: the grid size of the current area, Turbulent kinetic energy: an indicator of local turbulence intensity, reflecting the complexity of flow; Free water surface fluctuations: the rate of change of water surface height at the water-air interface, which affects the numerical diffusion error; Computational Error: A metric that evaluates the local error based on historical CFD calculations.

[0031] To determine the optimal number of clusters w, the elbow method can be used: ; In the formula, is the intra-cluster error sum of squares, represents the data point belonging to the qth cluster, represents the set of grid points of the p-th cluster, is the center point of the pth cluster. When the downward trend slows down, the w value is selected as the optimal number of clusters.

[0032] Perform K-Means iterative calculations, specifically: randomly select w initial cluster centers, calculate the Euclidean distance from each grid point to all cluster centers , and assign it to the nearest cluster : ; e represents the number of the e-th grid unit in the CFD calculation model, N represents the number of all grid units in the CFD calculation model, and for each cluster , update the mean center: ; If the cluster center no longer changes or the maximum number of iterations is reached, the iteration is terminated.

[0033] Assign a mesh refinement factor to each cluster : Cluster center error Normalize: ;in represents the error characteristic component of the pth cluster, and h is the number of clusters.

[0034] Combine the weights of each cluster (Based on velocity gradient, turbulence energy and other characteristics), calculate the dynamic grid adjustment coefficient DMRC: DMRC The larger the DMRC value, the higher the mesh density required in this area.

[0035] The obtained dynamic grid adjustment coefficient is compared with the preset threshold. If the dynamic grid adjustment coefficient is greater than or equal to the preset threshold, adaptive grid encryption, dynamic time step adjustment and turbulence model optimization are performed on the high error area, including: Calculate the adaptive grid size according to DMRC: ;in: is the optimized grid cell size, is the initial grid cell size, is the dynamic grid adjustment coefficient of the e-th grid unit. The larger the DMRC value, the greater the error impact and the higher the grid density.

[0036] Mesh refinement is performed in high error areas to improve local calculation accuracy. Adaptive mesh refinement is used to dynamically optimize mesh division during the calculation process.

[0037] Dynamic time step adjustment includes: using DMRC to calculate the new time step : ;in: is the optimized time step, is the original time step, The maximum DMRC value in the high error area. Reduce the time step in the high error area to improve the time discretization accuracy and reduce the error. Use adaptive time step control to automatically adjust the calculation stability.

[0038] Turbulence model optimization includes: For high error areas: Use LES (Large Eddy Simulation) or RANS-LES hybrid model to improve calculation accuracy. Use more sophisticated turbulent wall functions to improve near-wall flow simulation capabilities. In areas where the dynamic grid adjustment coefficient is greater than or equal to the pre-set threshold, increase the resolution of the turbulence model and improve numerical stability. In free water areas, optimize the VOF (Volume of Fluid) free surface capture algorithm to reduce errors. Use WENO (Weighted Essentially Non-Oscillation Format) to reduce numerical diffusion errors in high gradient areas.

[0039] Under different working conditions, the LSTM (long short-term memory network) deep learning model is used to predict the discharge capacity curve, reduce the number of CFD calculations, improve calculation efficiency, and optimize reservoir scheduling in combination with flow velocity distribution data.

[0040] Collect and organize historical discharge capacity data of the reservoir discharge system, including: Water level (H): changes in reservoir water level under different working conditions. Gate opening (G): percentage of gate opening (0%~100%). Discharge flow (Q): amount of water passing through the discharge facility per unit time. Velocity distribution (V(x,y)): velocity field in key areas such as spillways and energy dissipation pools. Historical CFD calculation results: discharge capacity data under different grid densities and time steps.

[0041] Since different parameters have different numerical ranges, they need to be standardized so that the data have the same scale and improve the training accuracy and stability of the model.

[0042] Constructing LSTM deep learning model: Input layer: Receive historical discharge capacity data, including water level, gate opening, flow rate and other features. LSTM layer: Extract time series features and learn the change trend of discharge capacity curve. Fully connected layer: Used to output future discharge capacity curve. Activation function: Use ReLU for nonlinear mapping to improve prediction accuracy.

[0043] The historical data is used to train the model so that it can learn the changing rules of discharge capacity. The validation data is used to evaluate the prediction ability of the model and adjust the model parameters to improve the generalization ability.

[0044] Predict the discharge capacity under future working conditions, input new water level, gate opening, flow rate and other data, and the model predicts the future discharge capacity curve. Use the mean square error to evaluate the error between the discharge capacity curve predicted by the model and the actual CFD calculation result. Set the error threshold TMSE (such as 0.05~0.1). If the prediction error exceeds the threshold, the CFD calculation needs to be optimized.

[0045] Combined with CFD calculation to optimize the discharge capacity curve: Optimize mesh density: refine the mesh in high error areas to improve local calculation accuracy.

[0046] Optimize time step: adjust time discretization to improve calculation stability.

[0047] Optimize turbulence model: Improve turbulence calculation and reduce the impact of errors.

[0048] Rerun the CFD calculation to obtain the optimized discharge capacity curve and update the LSTM training data to make the model prediction more accurate.

[0049] Compare the LSTM predicted discharge capacity curve with the CFD calculated discharge capacity curve to analyze the prediction accuracy. When the error requirements are met, directly use the LSTM prediction results to reduce the number of CFD calculations and improve calculation efficiency.

[0050] Combine CFD calculations with the velocity distribution predicted by deep learning to analyze the flow characteristics of the discharge facility. Determine key areas such as water flow scour and turbulent areas to provide reference for reservoir operation.

[0051] Example 2, please refer to Figure 2 As shown, a reservoir dam discharge capacity rating system described in this embodiment includes a parameter acquisition module, a data calculation module, a grid adjustment module and a calculation efficiency optimization module; Parameter acquisition module: obtains the geometric parameters of the reservoir dam discharge facilities, including the structural dimensions of the spillway, spillway tunnel, energy dissipation pool and bottom hole, builds the CFD model of the reservoir dam discharge system, and uses adaptive grids to optimize the calculation accuracy of key areas; Data calculation module: Based on the machine learning algorithm, the error sensitivity coefficient and dynamic grid adjustment coefficient in each key area are calculated respectively, and the area affected by numerical diffusion is detected through the calculated error sensitivity coefficient to determine whether to trigger optimization adjustment; Grid adjustment module: For high error areas, dynamic grid adjustment coefficients are used to perform adaptive grid encryption, dynamic time step adjustment, and turbulence model optimization to reduce errors and improve calculation accuracy; Computational efficiency optimization module: Perform multiple rounds of dynamic adjustments under different working conditions to generate a discharge capacity curve, combine deep learning prediction models to optimize computational efficiency, and output discharge capacity curves and flow velocity distribution maps to support reservoir scheduling optimization.

[0052] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0053] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for grading the discharge capacity of a reservoir dam, characterized in that: include: Obtain the geometric parameters of the reservoir dam discharge facilities, including the structural dimensions of the spillway, spillway tunnel, energy dissipation pool and bottom holes, build a CFD model of the reservoir dam discharge system, and use adaptive grids to optimize the calculation accuracy of key areas; Based on the machine learning algorithm, the error sensitivity coefficient and dynamic grid adjustment coefficient in each key area are calculated respectively, and the area affected by numerical diffusion is detected by the calculated error sensitivity coefficient to determine whether to trigger optimization adjustment; For high error areas, dynamic grid adjustment coefficients are used to perform adaptive grid encryption, dynamic time step adjustment, and turbulence model optimization to reduce errors and improve calculation accuracy. Perform multiple rounds of dynamic adjustments under different operating conditions to generate a discharge capacity curve, combine deep learning prediction models to optimize computing efficiency, and output the discharge capacity curve and flow velocity distribution map to support reservoir scheduling optimization.

2. A method for grading the discharge capacity of a reservoir dam according to claim 1, characterized in that: The key areas include high gradient change areas, turbulence and backflow areas, free water surface evolution areas and local negative pressure areas.

3. A method for grading the discharge capacity of a reservoir dam according to claim 2, characterized in that: The calculation method of the error sensitivity coefficient is: define the error characteristic variable and construct the data matrix X: ; m is the number of grid cells, n is the number of error-affecting variables; Since different variables have different value ranges, they need to be standardized so that their mean is 0 and their variance is 1: Calculate the covariance matrix C of the standardized data matrix X′: ; Where: C is the n×n dimensional covariance matrix, T is the matrix transpose; Perform eigenvalue decomposition on the covariance matrix C and obtain the eigenvalue and the corresponding eigenvector , the expression is: ; Eigenvalue : represents the variance of the i-th principal component, eigenvector Indicates the direction of the i-th principal component; select the first k principal components whose cumulative contribution rate reaches 95% for calculation, and calculate the principal component score: ; Where: V is a matrix composed of the selected k eigenvectors, Z is the reduced-dimensional data matrix, including the principal component scores; the error sensitivity coefficient is defined as ESC, and the expression is: ;in: is the eigenvalue of the jth principal component, is the weight coefficient corresponding to the eigenvector.

4. A method for grading the discharge capacity of a reservoir dam according to claim 3, characterized in that: Set the threshold TESC for the error influence area. If ESC ≥ TESC, it is considered that the numerical diffusion error in the area is large, and the area is marked as a high-error area; if ESC < TESC, it is considered that the numerical diffusion error in the area is small, and the area is marked as a low-error area, and record the index set of the grid areas that need to be optimized , calculate the proportion of the high-error area : ; N represents the number of all grid cells in the CFD calculation model. If the calculated proportion of the high-error area is greater than or equal to the set threshold of the proportion of the high-error area, an optimization adjustment is triggered.

5. A method for grading the discharge capacity of a reservoir dam according to claim 4, characterized in that: The method for obtaining the dynamic grid adjustment coefficient is as follows: constructing a data set Y, including local grid scale, turbulent kinetic energy, free water surface fluctuation and calculation error; determining the optimal clustering number w, using the elbow method: ; In the formula, represents the data point belonging to the qth cluster, is the intra-cluster error sum of squares, represents the set of grid points of the p-th cluster, is the center point of the pth cluster; Randomly select w initial cluster centers and calculate the Euclidean distance from each grid point to all cluster centers , and assign it to the nearest cluster : ; e represents the number of the e-th grid unit in the CFD calculation model, N represents the number of all grid units in the CFD calculation model, and for each cluster , update the mean center: ; If the cluster center no longer changes or the maximum number of iterations is reached, the iteration is terminated; Assign a mesh refinement factor to each cluster : Cluster center error Normalize: ;in Represents the error characteristic component of the pth cluster, h is the number of clusters; combined with the weights of each cluster , calculate the dynamic grid adjustment coefficient DMRC: DMRC .

6. A method for grading the discharge capacity of a reservoir dam according to claim 5, characterized in that: The adaptive grid encryption controls the grid optimization through the dynamic grid adjustment coefficient DMRC, and the optimized grid size calculation formula is: ;in: is the optimized grid cell size, is the initial grid cell size, is the dynamic grid adjustment coefficient of the e-th grid cell.

7. A method for grading the discharge capacity of a reservoir dam according to claim 6, characterized in that: The dynamic time step adjustment optimizes the time step by DMRC calculation, and the optimized time step calculation formula is: ;in: is the optimized time step, is the original time step, is the maximum DMRC value in the high error region.

8. A method for grading the discharge capacity of a reservoir dam according to claim 7, characterized in that: The deep learning prediction model is based on a long short-term memory network and includes: The LSTM model is trained using historical discharge capacity data, with input features including water level, gate opening, discharge flow, and flow velocity distribution; The LSTM layer extracts time series features and learns the changing trend of the discharge capacity curve; Calculate the mean square error. If the error exceeds the set threshold, optimize the CFD calculation to improve the prediction accuracy. Update LSTM training data based on CFD calculation results.

9. A reservoir dam discharge capacity rating system, used to implement a reservoir dam discharge capacity rating method according to any one of claims 1 to 8, characterized in that: It includes parameter acquisition module, data calculation module, grid adjustment module and calculation efficiency optimization module; Parameter acquisition module: obtains the geometric parameters of the reservoir dam discharge facilities, including the structural dimensions of the spillway, spillway tunnel, energy dissipation pool and bottom hole, builds the CFD model of the reservoir dam discharge system, and uses adaptive grids to optimize the calculation accuracy of key areas; Data calculation module: Based on the machine learning algorithm, the error sensitivity coefficient and dynamic grid adjustment coefficient in each key area are calculated respectively, and the area affected by numerical diffusion is detected through the calculated error sensitivity coefficient to determine whether to trigger optimization adjustment; Grid adjustment module: For high error areas, dynamic grid adjustment coefficients are used to perform adaptive grid encryption, dynamic time step adjustment, and turbulence model optimization to reduce errors and improve calculation accuracy; Computational efficiency optimization module: Perform multiple rounds of dynamic adjustments under different working conditions to generate a discharge capacity curve, combine deep learning prediction models to optimize computational efficiency, and output discharge capacity curves and flow velocity distribution maps to support reservoir scheduling optimization.

Images