A method and system for short-term prediction of power generation of small hydropower station
Silt deposit features are obtained through sonar detection and combined with hybrid neural network models to track the changes in sediment deposition in real time, solving the problem that sediment deposition effects are ignored in the prediction of power generation of small hydropower stations, improving the accuracy and reliability of prediction, ensuring prediction The continuous accuracy of the results provides support for the efficient operation and power dispatch of small hydropower stations.
Patent Information
- Application Number
- CN202411178622.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-08-27
AI Technical Summary
The existing power generation power prediction method for small hydropower stations ignores the impact of sediment deposition on reservoir capacity and power generation capacity, resulting in the prediction results being inaccurate and reliable enough.
The characteristics of sediment accumulation distribution are obtained through sonar detection technology, combined with hybrid neural network models, the dynamic changes of sediment deposits are tracked in real time, and the power generation power prediction model is adaptively adjusted to generate more accurate short-term power prediction results.
It significantly improves the accuracy and reliability of short-term prediction of power generation power in small hydropower stations, ensures the continuous accuracy of prediction results under dynamic changes in the sediment deposition environment, and supports efficient operation and power scheduling optimization.
Smart Images

Figure CN118713075B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of prediction of power generation capacity of small hydropower stations, and in particular to a method and system for short-term prediction of power generation capacity of small hydropower stations. Background Art
[0002] As a clean energy, small hydropower plays an important role in increasing energy supply and promoting the transformation and optimization of energy structure. However, small hydropower stations have poor regulation capabilities and strong output fluctuations. Large-scale grid-connected use will have a certain impact on the local power grid. Therefore, the accuracy of short-term prediction results of small hydropower station power generation has become a key guarantee for helping small hydropower stations to reasonably formulate power generation plans, ensure the stable operation of the power grid and improve the utilization rate of water resources.
[0003] Most of the existing methods for predicting the power generation of small hydropower stations are analyzed and studied based on the meteorological conditions, spatial distribution and historical cumulative rainfall in the hydropower station basin. Although they can meet the prediction needs of the power generation of small hydropower stations to a certain extent, they all ignore the impact of sediment deposition at the bottom of the reservoir during the operation of the hydropower station on the power generation capacity. The sensitivity of small hydropower stations to sediment deposition is much higher than that of large hydropower stations. Even a relatively small amount of sediment deposition may have a significant impact on the storage capacity and power generation capacity of small hydropower stations. The greater the amount of sediment deposition, the more the effective storage capacity of the reservoir is reduced, and the more obvious the decrease in power generation will be. The sediment deposition method also has an important impact on the storage capacity and power generation of small hydropower stations. For example, a large amount of sediment accumulation in a certain area of the reservoir basin will significantly reduce the storage capacity of the area, causing a sudden drop in the local water level, resulting in an increase in head loss, making the power generation output drop more significant. Therefore, it is urgent to provide a method for predicting the power generation of small hydropower stations that can comprehensively consider the impact of sediment deposition and deposition mode on the reservoir capacity, so as to improve the accuracy and reliability of the short-term prediction results of the power generation of small hydropower stations. Summary of the invention
[0004] The purpose of the present invention is to provide a short-term prediction method for the power generation of a small hydropower station. By combining the amount of sediment deposition and the deposition mode to track and analyze the dynamic changes of sediment deposition in real time, the power generation prediction model can adapt to the dynamic changes of the sediment deposition environment, effectively improve the accuracy and reliability of the short-term prediction of the power generation of the small hydropower station, and thus provide important support for the efficient operation of the small hydropower station and the optimization of power dispatching.
[0005] In order to achieve the above objectives, it is necessary to provide a short-term prediction method and system for the power generation of a small hydropower station in response to the above technical problems.
[0006] In a first aspect, an embodiment of the present invention provides a method for short-term prediction of power generation of a small hydropower station, the method comprising the following steps:
[0007] Acquire sonar detection data of different reservoir basins of the target hydropower station reservoir, and obtain corresponding regional sediment accumulation distribution characteristics based on the sonar detection data; the regional sediment accumulation distribution characteristics include regional sediment deposition thickness distribution, regional sediment deposition volume distribution and regional sediment deposition morphological parameter distribution;
[0008] According to the sediment accumulation distribution characteristics of the region, reservoir capacity impact analysis is performed to obtain multiple sedimentation degree classification areas, and sedimentary terrain volume difference analysis is performed on each sedimentation degree classification area to obtain the corresponding reservoir capacity change amount of the classification area;
[0009] Determine whether the storage capacity change of each graded area meets the preset storage capacity reduction condition. If so, take the corresponding sedimentation degree graded area as the key analysis area, and obtain the local head loss value of the key analysis area by simulating the water flow movement of the sediment deposition reservoir;
[0010] Generate a modified water level storage capacity curve of the target hydropower station reservoir according to the storage capacity changes of all graded areas and all local head loss values, and obtain a power generation control standard according to the modified water level storage capacity curve;
[0011] Power generation control is performed according to the power generation control standard, and power generation power is predicted based on the real-time collected sediment deposition monitoring data of different reservoir basins of the target hydropower station reservoir and the hybrid neural network model to obtain the corresponding power generation power prediction value.
[0012] Furthermore, the step of acquiring sonar detection data of different reservoir basin areas of the target hydropower station reservoir and obtaining the corresponding regional sediment accumulation distribution characteristics according to the sonar detection data includes:
[0013] A full coverage scan of the target hydropower station reservoir is performed by a multi-beam sonar detector to obtain underwater topographic information of different reservoir basin areas;
[0014] The time domain reflection coefficient analysis and spectrum analysis are performed on the underwater topography information of each reservoir basin area to obtain the corresponding regional sediment layer interface information;
[0015] Calculate the corresponding regional sediment thickness distribution according to the regional sediment layer interface information and the preset underwater terrain model;
[0016] According to the regional sediment deposition thickness distribution, the corresponding regional sediment deposition volume distribution is obtained based on the triangular prism volume formula;
[0017] A three-dimensional surface of a sediment accumulation body is constructed by using a triangular mesh model, and the interior of the sediment accumulation body is filled by a voxelization method to obtain a corresponding regional sediment accumulation three-dimensional model;
[0018] A morphological parameter analysis is performed based on the three-dimensional model of regional sediment accumulation to obtain a corresponding regional sediment deposition morphological parameter distribution; the regional sediment deposition morphological parameter distribution includes a slope distribution and a curvature distribution.
[0019] Furthermore, the step of performing reservoir capacity impact analysis according to the sediment accumulation distribution characteristics of the region to obtain multiple sedimentation degree classification areas includes:
[0020] According to the distribution of regional sediment deposition morphological parameters in each reservoir basin area, a corresponding spatial variation function model is selected, and according to the spatial variation function model, the regional sediment deposition thickness distribution and the regional sediment deposition volume distribution, a corresponding regional sediment deposition thickness distribution map and a regional sediment deposition volume distribution map are obtained by the Kriging interpolation method;
[0021] Performing elevation difference analysis on the regional sediment thickness distribution map and the original reservoir basin regional topographic feature map to obtain a corresponding regional topographic change map;
[0022] According to the terrain change maps of all regions, the area proportions of regions with different degrees of change are counted, and the threshold value of reservoir volume impact change is obtained according to the area proportions of regions with different degrees of change;
[0023] Obtaining the sediment deposition impact area according to the regional terrain change map and the reservoir volume impact change threshold;
[0024] The sedimentation degree of all sediment deposition affected areas is graded by fuzzy clustering algorithm, and multiple sedimentation degree graded areas are obtained.
[0025] Furthermore, the step of performing sedimentary terrain volume difference analysis on each sedimentary degree graded area to obtain the corresponding storage capacity change amount of the graded area includes:
[0026] The zoning statistics tool and grid calculator are used to analyze each sedimentation degree classification area to obtain the corresponding sedimentation area and average sedimentation thickness;
[0027] According to the sedimentation area and average sedimentation thickness of each sedimentation degree classification area, the optimal boundary value of each sedimentation degree classification area is obtained;
[0028] According to the optimal demarcation value, a distribution map of sediment deposition differences is generated by using a grid reclassification tool;
[0029] According to the sediment deposition degree difference distribution map, the terrain volume difference before and after sediment deposition is analyzed for each sediment degree classification area, and the reservoir capacity change of the corresponding classification area is obtained.
[0030] Furthermore, the step of obtaining the local head loss value of the key analysis area by simulating the water flow movement condition of the sediment deposition reservoir includes:
[0031] Acquire a regional terrain mask of the key analysis area, and adjust terrain data in a pre-constructed two-dimensional hydrodynamic model according to the regional terrain mask;
[0032] According to the adjusted two-dimensional hydrodynamic model, the water flow movement condition of the sediment deposition reservoir is simulated and analyzed to obtain corresponding hydrodynamic simulation results; the hydrodynamic simulation results include regional flow velocity change data and water level change data;
[0033] According to the hydrodynamic simulation results, water level data of each key analysis area is obtained, and a water level change distribution map is generated by Kriging interpolation method according to the water level data of each key analysis area;
[0034] According to the water level change distribution diagram, the water level change situation of each key analysis area is obtained, and according to the water level change situation of each key analysis area, the corresponding local head loss value is calculated based on the Bernoulli equation.
[0035] Furthermore, the power generation control standard includes an inflow control standard and a unit start-stop water level standard;
[0036] The steps of generating a modified water level storage capacity curve of the target hydropower station reservoir according to the storage capacity changes of all graded areas and all local head loss values, and obtaining a power generation control standard according to the modified water level storage capacity curve include:
[0037] According to the storage capacity changes of the graded areas of all the graded areas of sedimentation degree, a total storage capacity change is obtained, and according to the total storage capacity change, a current reservoir capacity of the target hydropower station reservoir is obtained;
[0038] Acquire the historical water level and storage capacity time series data of the target hydropower station reservoir, and obtain the water level and storage capacity curve through polynomial regression fitting according to the historical water level and storage capacity time series data, the current reservoir capacity and the corresponding current reservoir water level;
[0039] According to the local head loss values of all key analysis areas, an average local head loss value is obtained, and the reservoir water level in the water level and storage capacity curve is corrected by the average local head loss value to obtain a corrected water level and storage capacity curve;
[0040] According to the modified water level storage capacity curve, an inflow flow optimization data set is generated by a Monte Carlo simulation method; the inflow flow optimization data set includes a plurality of groups of inflow flows and outflow flows corresponding to different water levels;
[0041] A multi-objective reservoir operation optimization model is established with the goal of maximizing power generation and minimizing water abandonment. According to the inflow optimization data set, a genetic algorithm is used to solve the multi-objective reservoir operation optimization model to obtain the inflow control standard; the inflow control standard includes inflow control thresholds corresponding to different water levels;
[0042] Through the fuzzy comprehensive evaluation method, a unit start-stop evaluation index system is established, and based on the unit start-stop evaluation index system, a comprehensive analysis is carried out based on the Delphi method and the hierarchical analysis method to obtain the unit start-stop water level standard; the unit start-stop evaluation index system includes a first-level indicator of power generation efficiency, a first-level indicator of flood discharge capacity, a first-level indicator of reservoir safety, and a number of second-level indicators corresponding to each first-level indicator.
[0043] Furthermore, the sediment deposition monitoring data includes sediment deposition thickness distribution data, sediment deposition volume distribution data and sediment accumulation morphology image data;
[0044] The step of performing power generation prediction based on the real-time collected sediment deposition monitoring data of different reservoir basins of the target hydropower station reservoir and the hybrid neural network model to obtain the corresponding power generation prediction value comprises:
[0045] According to the preset ARIMA model, the sediment thickness distribution data and sediment volume distribution data of each reservoir basin area are analyzed respectively to obtain the corresponding sediment thickness change rate and sediment volume growth rate;
[0046] According to the preset convolutional neural network, feature extraction is performed on the sediment accumulation morphology image data of each reservoir basin area to obtain the corresponding sediment accumulation morphology evolution characteristics;
[0047] The sediment thickness distribution data, sediment volume distribution data, sediment thickness change rate, sediment volume growth rate and sediment accumulation morphology evolution characteristics of all reservoir basin areas are input into the hybrid neural network model to predict the power generation and obtain the power generation prediction value.
[0048] Furthermore, the hybrid neural network model is obtained by weighted integration of a long short-term memory network model and a support vector regression model; the steps of constructing the hybrid neural network model include:
[0049] Acquire the historical hydropower station characteristic parameter time series data, the historical power generation time series data, and the historical sediment deposition time series data of different reservoir basin areas of the target hydropower station reservoir; the historical sediment deposition time series data includes the historical sediment thickness distribution time series data, the historical sediment volume distribution time series data and the historical accumulation morphology image time series data;
[0050] The historical sediment deposition time series data of each reservoir basin area are analyzed respectively to obtain the corresponding historical sediment evolution time series data and historical sediment influencing factor time series data; the historical sediment evolution time series data include the historical sediment thickness change rate time series data, the historical sediment volume growth rate time series data and the historical accumulation morphology evolution characteristic time series data; the historical sediment influencing factor time series data include the sediment thickness change coefficient time series data, the volume growth coefficient time series data and the local accumulation morphology coefficient time series data;
[0051] The random forest regression model is trained according to the preset hydrological parameter characteristics, the historical power generation time series data, and the historical sedimentation influencing factor time series data of each reservoir basin area, and the weight value of each sedimentation influencing factor is obtained according to the trained random forest regression model;
[0052] Comprehensive analysis is performed based on the historical sedimentation influencing factor time series data of each reservoir basin area to obtain the comprehensive sedimentation influencing factor time series data, and the head loss correction coefficient is obtained based on the weight value of each sedimentation influencing factor;
[0053] According to the comprehensive sedimentation influencing factor time series data and the head loss correction coefficient, the historical power generation time series data is corrected to obtain the historical power generation correction time series data;
[0054] Constructing a model training data set based on the historical power generation correction time series data, the historical hydropower station characteristic parameter time series data, and the historical sediment thickness distribution time series data, the historical sediment volume distribution time series data, and the historical sediment evolution time series data of different reservoir basin areas;
[0055] According to the model training data set, integrated learning training is performed on the preset long short-term memory network model and the preset support vector regression model to obtain the hybrid neural network model.
[0056] Furthermore, the method further comprises:
[0057] According to the preset update cycle, the latest deposition influence factor time series data and the corresponding latest deposition influence factor weight are obtained, and the latest power generation power correction time series data are generated according to the latest deposition influence factor time series data and the corresponding latest deposition influence factor weight, and an adaptive update training data set is constructed according to the latest power generation power correction time series data to train and update the hybrid neural network model.
[0058] In a second aspect, an embodiment of the present invention provides a short-term prediction system for power generation of a small hydropower station, the system comprising:
[0059] A sediment accumulation analysis module is used to obtain sonar detection data of different reservoir basins of a target hydropower station reservoir, and obtain corresponding regional sediment accumulation distribution characteristics based on the sonar detection data; the regional sediment accumulation distribution characteristics include regional sediment deposition thickness distribution, regional sediment deposition volume distribution and regional sediment deposition morphological parameter distribution;
[0060] A reservoir capacity impact analysis module is used to perform reservoir capacity impact analysis based on the sediment accumulation distribution characteristics of the region, obtain multiple sedimentation degree classification areas, and perform sedimentary terrain volume difference analysis on each sedimentation degree classification area to obtain the corresponding reservoir capacity change amount of the classification area;
[0061] The head loss analysis module is used to determine whether the storage capacity change of each graded area meets the preset storage capacity reduction condition. If so, the corresponding sedimentation degree graded area is used as the key analysis area, and the local head loss value of the key analysis area is obtained by simulating the water flow movement of the sediment deposition reservoir;
[0062] A power generation control analysis module, used to generate a modified water level storage capacity curve of the target hydropower station reservoir according to the storage capacity changes of all graded areas and all local head loss values, and to obtain a power generation control standard according to the modified water level storage capacity curve analysis;
[0063] The power generation prediction module is used to perform power generation control according to the power generation control standard, and to predict the power generation based on the real-time collected sediment deposition monitoring data of different reservoir basins of the target hydropower station reservoir and the hybrid neural network model to obtain the corresponding power generation prediction value.
[0064] The present application provides a method and system for short-term prediction of power generation of a small hydropower station. The method can obtain sonar detection data of different reservoir basins of a target hydropower station reservoir, and obtain regional sediment accumulation distribution characteristics corresponding to regional sediment deposition thickness distribution, regional sediment deposition volume distribution and regional sediment deposition morphological parameter distribution based on the sonar detection data. According to the regional sediment deposition distribution characteristics, a reservoir capacity impact analysis is performed to obtain multiple sedimentation degree classification areas, and a sedimentation terrain volume difference analysis is performed on each sedimentation degree classification area to obtain the corresponding classification area reservoir capacity change, and it is judged whether the reservoir capacity change of each classification area meets the preset When the reservoir capacity reduction condition is met, the corresponding sedimentation degree classification area is used as the key analysis area, and the local head loss value of the key analysis area is obtained by simulating the water flow movement of the sediment deposition reservoir, and the modified water level storage capacity curve of the target hydropower station reservoir is generated according to the storage capacity change of all classification areas and all local head loss values, and the power generation control standard is obtained according to the analysis of the modified water level storage capacity curve, and the power generation control is performed according to the power generation control standard, and the power generation power is predicted according to the real-time collected sediment deposition monitoring data of different reservoir basins of the target hydropower station reservoir and the hybrid neural network model, and the corresponding power generation power prediction value is obtained. Compared with the prior art, the short-term prediction method of the power generation power of a small hydropower station, by combining the sediment deposition amount and the deposition mode to track and analyze the dynamic changes of sediment deposition in real time, enables the power generation prediction model to adapt to the dynamic changes of the sediment deposition environment, effectively improves the accuracy and reliability of the short-term prediction of the power generation power of the small hydropower station, and then provides important support for the efficient operation of the small hydropower station and the optimization of power dispatching, and has high practical value and promotion and application prospects. BRIEF DESCRIPTION OF THE DRAWINGS
[0065] Figure 1 It is a schematic flow chart of a method for short-term prediction of power generation of a small hydropower station in an embodiment of the present invention;
[0066] Figure 2 2 is another schematic flow chart of a method for short-term prediction of power generation of a small hydropower station according to an embodiment of the present invention;
[0067] Figure 3 2 is a schematic diagram of the structure of a short-term prediction system for power generation of a small hydropower station according to an embodiment of the present invention;
[0068] Figure 4 It is another structural schematic diagram of the short-term prediction system for power generation of a small hydropower station in an embodiment of the present invention. DETAILED DESCRIPTION
[0069] In order to make the purpose, technical scheme and beneficial effects of the present application clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the embodiments described below are part of the embodiments of the present invention and are only used to illustrate the present invention, but are not used to limit the scope of the present invention. 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.
[0070] The short-term prediction method for the power generation of a small hydropower station provided by the present invention can be understood as a method proposed based on the current situation that the prediction analysis of the power generation of a small hydropower station ignores the influence of sediment deposition, resulting in poor prediction results. It uses sonar detection technology to accurately capture the spatial distribution characteristics of reservoir sediment deposition, and on this basis, evaluates the influence of sediment deposition on reservoir capacity and water flow characteristics. A hybrid neural network is introduced to track and analyze the dynamic changes of sediment deposition in real time based on the combination of sediment deposition amount and deposition mode, and adaptively predicts the power generation of a small hydropower station based on the influence of the dynamic changes of sediment deposition on power generation. The method can significantly improve the accuracy and timeliness of short-term power generation prediction, while ensuring the continuous accuracy of the prediction results under the conditions of dynamic changes in the sediment deposition environment. The following embodiments will explain the short-term prediction method for the power generation of a small hydropower station of the present invention in detail.
[0071] In one embodiment, Figure 1 As shown, a method for short-term prediction of power generation of a small hydropower station is provided, comprising the following steps:
[0072] S11. Acquire sonar detection data of different reservoir basin areas of the target hydropower station reservoir, and obtain corresponding regional sediment accumulation distribution characteristics based on the sonar detection data; the regional sediment accumulation distribution characteristics include regional sediment deposition thickness distribution, regional sediment deposition volume distribution and regional sediment deposition morphological parameter distribution; wherein, different reservoir basin areas can be understood as dividing the reservoir basin of the entire target hydropower station reservoir into scanning areas based on the single detection range of the actual sonar detection instrument. The specific division method is not specifically limited here, and it is only necessary that the sum of the different reservoir basin areas can achieve full coverage of the entire reservoir basin range.
[0073] Specifically, the step of acquiring sonar detection data of different reservoir basin areas of the target hydropower station reservoir and obtaining the corresponding regional sediment accumulation distribution characteristics according to the sonar detection data includes:
[0074] A full-coverage scan of the target hydropower station reservoir is performed by a multi-beam sonar detector to obtain underwater terrain information of different reservoir basin areas; wherein the underwater terrain information can be obtained directly based on the detection and scanning results of the multi-beam sonar detector, or on the basis of the detection and scanning results obtained by the multi-beam sonar detector, a single-beam sonar detector is used to perform water depth correction to obtain a more accurate underwater terrain three-dimensional model. For example, a multi-beam sonar detector is first used to scan at a fan angle of 20 degrees to obtain underwater terrain point cloud data with a resolution of 0.1 meters, and then the multi-beam scanning data is corrected according to the water depth correction curve obtained by fitting the historical single-beam sounding data and the newly collected single-beam sounding points by the least squares method, so as to generate a three-dimensional underwater terrain model with an accuracy of centimeters.
[0075] The underwater terrain information of each reservoir basin area is subjected to time domain reflection coefficient analysis and spectrum analysis respectively to obtain the corresponding regional sediment layer interface information; the regional sediment layer interface information can be understood as the interface position information between the sediment layer and the bottom of the reservoir basin; in practical applications, the underwater terrain information (sonar raw data) can be subjected to time domain reflection coefficient analysis after selecting a reflection coefficient threshold to preliminarily identify the sediment layer interface, and at the same time, the energy change characteristics of the preset frequency band range (15-25kHz) can be identified based on the spectrum analysis to assist in judging the preliminarily identified sediment layer interface, so as to obtain more accurate regional sediment layer interface information. The corresponding detailed implementation method can be realized by referring to the relevant existing technology and will not be described in detail here.
[0076] According to the regional sediment layer interface information and the preset underwater terrain model, the corresponding regional sediment deposition thickness distribution is calculated; wherein, the preset underwater terrain model can be selected according to the actual application scenario, or it can be a three-dimensional underwater terrain model obtained by combining the aforementioned multi-beam sonar detector with the single-beam sonar detector, which is not specifically limited here; correspondingly, the regional sediment deposition thickness distribution can be understood as the sediment deposition thickness of each regional grid obtained by superimposing and analyzing the preset underwater terrain model and the obtained regional sediment layer interface information, and the specific calculation method can also be implemented by referring to the relevant existing technology.
[0077] According to the regional sediment deposition thickness distribution, the corresponding regional sediment deposition volume distribution is obtained based on the triangular prism volume formula; wherein the regional sediment deposition volume distribution can be obtained by multiplying the sediment deposition thickness of each regional grid obtained by the above calculation by the area and summing the results.
[0078] A triangular mesh model is used to construct the three-dimensional surface of a silt accumulation body, and the interior of the silt accumulation body is filled by a voxelization method to obtain a corresponding three-dimensional model of regional silt accumulation; wherein, the triangular mesh model (Delaunay) can be understood as an algorithm model for three-dimensional reconstruction of regional silt accumulation bodies; in practical applications, a three-dimensional surface is first constructed based on a preset grid resolution (such as 0.2 meters) using a Delaunay triangular mesh algorithm, and then the interior is filled using an octree voxelization method based on a preset voxel size (for example, 0.1 meters) to obtain a complete three-dimensional model of the silt accumulation body.
[0079] A morphological parameter analysis is performed based on the regional sediment accumulation three-dimensional model to obtain a corresponding regional sediment deposition morphological parameter distribution; the regional sediment deposition morphological parameter distribution includes a slope distribution and a curvature distribution; wherein the slope distribution can be obtained by calculating the normal vector of each triangular facet in the regional sediment accumulation three-dimensional model, and the corresponding curvature distribution can be obtained by calculating the principal curvature using the curvature tensor; in practical applications, the corresponding sediment accumulation area can be divided into different sediment deposition types based on the regional sediment deposition morphological parameter distribution.
[0080] S12. Perform reservoir capacity impact analysis based on the sediment accumulation distribution characteristics of the region to obtain multiple sedimentation degree classification areas, and perform sedimentation terrain volume difference analysis on each sedimentation degree classification area to obtain the corresponding reservoir capacity change of the classification area; wherein, the number of classifications of the sedimentation degree classification areas can be selected according to actual application needs, and can also be obtained based on the clustering analysis of sediment accumulation distribution characteristics in the actual area, for example, it can be divided into 5 different levels such as extremely high sedimentation area, high sedimentation area, medium sedimentation area, low sedimentation area and extremely low sedimentation area to distinguish different sediment accumulation degrees.
[0081] Specifically, the step of performing reservoir capacity impact analysis according to the sediment accumulation distribution characteristics of the region to obtain multiple sedimentation degree classification areas includes:
[0082] According to the distribution of regional sediment deposition morphological parameters in each reservoir basin area, the corresponding spatial variation function model is selected, and according to the spatial variation function model, the regional sediment deposition thickness distribution and the regional sediment deposition volume distribution, the corresponding regional sediment deposition thickness distribution map and regional sediment deposition volume distribution map are obtained by the Kriging interpolation method; wherein, the regional sediment deposition thickness distribution map can be understood as generating sediment deposition thickness grid data by performing Kriging interpolation on the regional sediment deposition thickness based on the preset grid resolution size according to the spatial variation function model, and based on the preset contour line interval, using The sediment thickness distribution map is drawn using the CONREC contour tracing algorithm. For example, in practical applications, based on the regional sediment deposition morphological parameter distribution data, the spatial variation function model is selected by comparing the goodness of fit of the spherical, exponential and Gaussian models. The model is then used for ordinary Kriging interpolation to generate 10 m × 10 m resolution sediment thickness grid data. Assuming that there are 100,000 grid cells in total, the CONREC contour tracing algorithm is used to draw the sediment thickness distribution map. Correspondingly, the regional sediment deposition volume distribution map can also be understood as a sediment deposition volume distribution map obtained by spatially interpolating the sediment deposition volume using the block kriging interpolation method in order to reduce the impact of small-scale variations while improving the accuracy of large-scale sediment deposition volume estimation, and drawing the obtained sediment deposition volume grid data using the three-dimensional voxel rendering method; for example, in practical applications, the block kriging interpolation method can be used to divide the study area into regular blocks of 50 meters × 50 meters, each containing 25 grid units, to reduce the impact of small-scale variations; the interpolated sediment deposition volume grid data is visualized using the three-dimensional voxel rendering method, and each voxel represents a deposition volume of 1000 cubic meters.
[0083] The elevation difference analysis is performed on the regional sediment deposition thickness distribution map and the original reservoir basin area terrain characteristic map to obtain the corresponding regional terrain change map; wherein, the regional terrain change map can be understood as a terrain elevation difference distribution map before and after sediment deposition obtained by superimposing the regional sediment deposition thickness distribution map with the original reservoir basin area terrain characteristic map using raster algebraic operations.
[0084] According to the terrain change maps of all regions, the area proportions of regions with different degrees of change are counted, and the threshold value of reservoir volume impact change is obtained according to the area proportions of regions with different degrees of change; among them, the threshold value of reservoir volume impact change can be understood as the elevation change threshold used to determine whether regional sediment deposition has an impact on the effective volume of the library; the specific determination method is to count the area proportions of different degrees of change in the terrain change maps of all regions, so as to quantify the degree of change of the reservoir terrain by sediment deposition, and at the same time use the Otsu adaptive threshold segmentation method to automatically determine the threshold value affecting the effective volume of the reservoir according to the distribution characteristics of sediment deposition thickness; for example, based on the statistics of all regional terrain change maps, the area proportions of different degrees of change are 30% of the total area for 0-0.5 meter change areas, 25% for 0.5-1 meter change areas, etc., then the Otsu adaptive threshold segmentation method can be used to process the sediment deposition thickness raster data, and the threshold value affecting the effective volume of the reservoir is automatically determined to be 1.2 meters.
[0085] According to the regional terrain change map and the reservoir volume impact change threshold, the sediment deposition influence area is obtained; that is, the area where the terrain elevation difference change value is greater than the reservoir volume impact change threshold is marked as the area where sediment deposition has a significant impact on the effective volume of the reservoir, that is, the sediment deposition influence area that needs to be focused on and analyzed.
[0086] The fuzzy clustering algorithm is used to classify the sedimentation degree of all sedimentation affected areas, and multiple sedimentation degree classification areas are obtained; among them, the division result of the sedimentation degree classification area can be understood as the clustering result of sedimentation areas of different degrees of high, medium and low obtained by classifying and iterating all sedimentation affected areas according to the preset number of iterations after determining the optimal number of clusters based on the silhouette coefficient method.
[0087] In this embodiment, the change in storage capacity of the graded area can be understood as the situation that each area with different sediment deposition degrees will cause the reservoir storage capacity to decrease; specifically, the step of performing sedimentary terrain volume difference analysis on each sedimentation degree graded area to obtain the corresponding change in storage capacity of the graded area includes:
[0088] The zonal statistics tool and the raster calculator are used to analyze each sedimentation degree classification area to obtain the corresponding sedimentation area area and average sedimentation thickness; specifically, if the sedimentation degree classification areas obtained by the above cluster analysis have five different levels, namely, extremely high sedimentation area, high sedimentation area, medium sedimentation area, low sedimentation area and extremely low sedimentation area, the zonal statistics tool can be used to calculate the area of each sedimentation degree classification area. The corresponding results show that the area of the extremely high sedimentation area is 2.5 square kilometers, the high sedimentation area is 7.8 square kilometers, the medium sedimentation area is 15.3 square kilometers, the low sedimentation area is 22.1 square kilometers, and the extremely low sedimentation area is 31.4 square kilometers; at the same time, the raster calculator is used to calculate the average sedimentation thickness of each area, and the extremely high sedimentation area is 3.2 meters, the high sedimentation area is 2.5 meters, the medium sedimentation area is 1.8 meters, the low sedimentation area is 1.2 meters, and the extremely low sedimentation area is 0.6 meters.
[0089] According to the sedimentation area and average sedimentation thickness of each sedimentation degree classification area, the optimal boundary value of each sedimentation degree classification area is obtained; wherein, the optimal boundary value can be understood as the boundary value between two adjacent sedimentation degree classification areas calculated by the Jenks optimization method. For example, for the five sedimentation degree classification areas in the above example, the corresponding four boundary values obtained by the Jenks optimization method are 0.8 meters, 1.5 meters, 2.2 meters and 2.9 meters.
[0090] According to the optimal dividing value, a sediment deposition degree difference distribution map is generated through a grid reclassification tool; wherein, the grid reclassification tool can use an ArcGIS tool. The specific method of generating a sediment deposition degree difference distribution map for identifying the sediment deposition degree differences in different regions based on the optimal dividing value can be implemented by referring to the relevant existing technology and will not be described in detail here.
[0091] According to the sediment deposition degree difference distribution map, the terrain volume difference analysis before and after sediment deposition is carried out on each sedimentation degree classification area to obtain the corresponding reservoir capacity change of the classification area; wherein, the terrain volume difference analysis before and after sediment deposition can be understood as the volume difference between the original terrain and the terrain after deposition obtained based on the sediment deposition degree difference distribution map using the Cut Fill tool. The corresponding analysis can obtain the volume change of the sedimentation area in the different sedimentation degree classification areas. For example, based on the above example analysis, the volume change of the extremely high sedimentation area is 80 million cubic meters, the high sedimentation area is 195 million cubic meters, the medium sedimentation area is 275.4 million cubic meters, the low sedimentation area is 265.2 million cubic meters, and the extremely low sedimentation area is 188.4 million cubic meters.
[0092] S13. Determine whether the storage capacity change of each graded area meets the preset storage capacity reduction condition. If so, take the corresponding sedimentation degree graded area as the key analysis area, and obtain the local head loss value of the key analysis area by simulating the water flow movement condition of the silt deposition reservoir; wherein, the preset storage capacity reduction condition can be set according to the actual application requirements. In this embodiment, it is preferably set to that the regional storage capacity change is greater than the preset storage capacity reduction threshold; that is, the area where the storage capacity change of the graded area is greater than the storage capacity reduction threshold is taken as the key analysis area for subsequent analysis of its impact on the local head loss.
[0093] Specifically, the step of simulating the water flow movement of the sediment deposition reservoir to obtain the local head loss value of the key analysis area includes:
[0094] A regional terrain mask of the key analysis area is obtained, and the terrain data in the pre-constructed two-dimensional hydrodynamic model is adjusted according to the regional terrain mask; wherein the regional terrain mask can be understood as a regional terrain mask obtained by masking the sediment deposition in the key analysis area. The specific generation method can be implemented by existing technology and will not be described in detail here.
[0095] According to the adjusted two-dimensional hydrodynamic model, the water flow movement of the sediment deposition reservoir is simulated and analyzed to obtain the corresponding hydrodynamic simulation results; the hydrodynamic simulation results include regional flow velocity change data and water level change data; wherein, the two-dimensional hydrodynamic model can be constructed by MIKE21FM software, and the model is used to solve the shallow water equations based on the finite volume method, and the water flow movement of the reservoir after sediment deposition is simulated to obtain the required hydrodynamic simulation results; in the actual simulation process, the upstream boundary can be set as the flow process line (the flow rate is a change process of 500-1000 cubic meters / second), the downstream boundary can be set as the water level process line (the fluctuation process of 145-150 meters), the initial condition is the static water level (145 meters), and the reservoir area is divided into triangular unstructured grids (for example, the grid size is 10 meters × 10 meters), and the hydrodynamic simulation is performed according to the preset time step (1 hour) and the preset simulation period (30 days), for example, the flow velocity in the key analysis area increases by 20-30%, and the water level rises by 0.5-1.5 meters.
[0096] According to the hydrodynamic simulation results, the water level data of each key analysis area is obtained, and based on the water level data of each key analysis area, a water level change distribution map is generated through the Kriging interpolation method.
[0097] According to the water level change distribution map, the water level change of each key analysis area is obtained, and according to the water level change of each key analysis area, the corresponding local head loss value is calculated based on the Bernoulli equation; wherein, the water level change can be understood as the water level change in the sediment accumulation area obtained by calculating the water level difference before and after sediment deposition through a grid calculator based on the water level change distribution map generated by the ordinary Kriging interpolation method mentioned above; the local head loss calculated based on the Bernoulli equation can be understood as the calculation of the energy loss between the upstream and downstream sections of the sediment accumulation area while considering friction loss and local loss, and the average head loss value of each sub-area can also be calculated through grid statistics to generate a local head loss distribution map with a preset height resolution, which will not be described in detail here.
[0098] S14. Generate a modified water level and storage capacity curve of the target hydropower station reservoir according to the storage capacity changes of all graded areas and all local head loss values, and obtain a power generation control standard based on the analysis of the modified water level and storage capacity curve; wherein the modified water level and storage capacity curve can be understood as a water level and storage capacity relationship curve that is more in line with the actual characteristics of the target hydropower station reservoir based on the storage capacity changes caused by sediment deposition and the local head loss; correspondingly, the power generation control standard can be understood as a strategy standard for adjusting the power generation dispatch of the target hydropower station according to the modified water level and storage capacity curve. Preferably, the power generation control standard includes an inflow control standard and a unit start and stop water level standard.
[0099] Specifically, the steps of generating a modified water level storage capacity curve of the target hydropower station reservoir according to the storage capacity changes of all graded areas and all local head loss values, and obtaining a power generation control standard according to the modified water level storage capacity curve analysis include:
[0100] According to the changes in the storage capacity of all graded areas of graded sedimentation degree, the total storage capacity change is obtained, and according to the total storage capacity change, the current reservoir capacity of the target hydropower station reservoir is obtained; wherein, the total storage capacity change can be understood as the total storage capacity affected by silt deposition obtained by weighted aggregation of the changes in the storage capacity of all graded areas of graded sedimentation degree, and the ideal storage capacity of the target hydropower station reservoir minus the total storage capacity change can be obtained to obtain the current reservoir capacity that actually affects the power generation power.
[0101] The historical water level and storage capacity time series data of the target hydropower station reservoir are obtained, and the water level and storage capacity curve is obtained through polynomial regression fitting according to the historical water level and storage capacity time series data, the current reservoir capacity and the corresponding current reservoir water level; wherein, the polynomial order used for the polynomial regression fitting can select the highest order with the highest R² value and no overfitting based on the goodness of fit R² value.
[0102] According to the local head loss values of all key analysis areas, the local head loss average value is obtained, and the reservoir water level in the water level capacity curve is corrected by the local head loss average value to obtain a corrected water level capacity curve; wherein, the corrected water level capacity curve can be understood as a reservoir characteristic curve equation obtained by further correcting the above-fitted water level capacity curve based on the characteristic that the local head loss value will affect the reservoir water level and using the local head loss value as a water level correction factor, that is, the water level capacity curve obtained by subtracting the local head loss average value from the reservoir water level in the above water level capacity curve.
[0103] According to the modified water level and storage capacity curve, an optimized inflow data set is generated through the Monte Carlo simulation method; the optimized inflow data set includes multiple groups of inflows and outflows corresponding to different water levels. The actual acquisition method can be understood as setting the log-normal distribution obeyed by the inflow according to the statistical characteristics of historical reservoir hydrological data, and performing a preset number of Monte Carlo simulations based on the inflow distribution and the modified water level and storage capacity curve to generate a certain number of water level-inflow-outflow data.
[0104] With the goal of maximizing power generation and minimizing water abandonment, a multi-objective reservoir scheduling optimization model is established, and based on the inflow optimization data set, a genetic algorithm is used to solve the multi-objective reservoir scheduling optimization model to obtain the inflow control standard; the inflow control standard includes inflow control thresholds corresponding to different water levels; wherein, the multi-objective reservoir scheduling optimization model only needs to simultaneously meet the two goals of maximizing power generation and minimizing water abandonment, and the specific model expression can be constructed according to actual application requirements, which is not specifically limited here.
[0105] Through the fuzzy comprehensive evaluation method, the unit start-stop evaluation index system is established, and according to the unit start-stop evaluation index system, the Delphi method and the hierarchical analysis method are used to conduct a comprehensive analysis to obtain the unit start-stop water level standard; the unit start-stop evaluation index system includes the first-level index of power generation efficiency, the first-level index of flood discharge capacity, the first-level index of reservoir safety, and several second-level indicators corresponding to each first-level index; among them, the second-level indicators corresponding to the first-level index of power generation efficiency include head utilization and unit efficiency, etc.; the second-level indicators corresponding to the first-level index of flood discharge capacity include the number of flood discharge gates and flood discharge flow, etc.; the second-level indicators corresponding to the first-level index of reservoir safety include dam body stability, flood control storage capacity, etc. In practical applications, multiple unit start-up water level standard schemes are pre-set and the unit start-stop evaluation index system is constructed. After calculating the values of each second-level evaluation index with reference to relevant existing technologies, the weight values of each second-level evaluation index are determined by the Delphi method, and then the hierarchical analysis method is used to calculate the comprehensive scores of each first-level index. When the first-level index comprehensive score matrix meets the consistency judgment condition, the score values of each unit start-up water level standard scheme are calculated, and the required unit start-stop water level standard is obtained accordingly. It should be noted that the analysis process of the Delphi method and the hierarchical analysis method involved in obtaining the start-stop water level standard of the unit in this embodiment can be implemented by referring to the relevant existing technologies, and will not be described in detail here. The power generation control standard obtained by this method is more in line with the scheduling and operation scenario of a hydropower station with silt deposition, and can effectively ensure the continuous and stable operation of a small hydropower station.
[0106] S15. Execute power generation control according to the power generation control standard, and perform power generation prediction based on the real-time collected sedimentation monitoring data of different reservoir basin areas of the target hydropower station reservoir and the hybrid neural network model to obtain the corresponding power generation prediction value; wherein the real-time collected sedimentation monitoring data of different reservoir basin areas preferably includes sedimentation thickness distribution data, sedimentation volume distribution data and sedimentation accumulation morphology image data, so as to achieve accurate prediction of power generation under the condition of sensing the dynamic changes of sedimentation. Specifically, the step of performing power generation prediction based on the real-time collected sedimentation monitoring data of different reservoir basin areas of the target hydropower station reservoir and the hybrid neural network model to obtain the corresponding power generation prediction value includes:
[0107] According to the preset ARIMA model, the sediment thickness distribution data and sediment volume distribution data of each reservoir basin area are analyzed respectively to obtain the corresponding sediment thickness change rate and sediment volume growth rate; among them, the sediment thickness change rate can be understood as combining the currently collected sediment thickness distribution data with the historical sediment thickness distribution data obtained at multiple previous collection times into time series data, and then calculating the sediment thickness time series change based on the sediment thickness distribution spatial model constructed by the Kriging three-dimensional interpolation algorithm, and then calculating the sediment thickness time series change based on the preset ARIMA (Autoregressive Integrated Moving Average The sediment thickness growth rate is obtained by analyzing the calculated sediment thickness time series changes using the ARIMA model, for example, the daily average sediment thickness change rate is obtained. Correspondingly, the sediment volume growth rate can also be understood as combining the currently collected sediment volume distribution data with the historical sediment volume distribution data obtained at multiple previous collection moments into time series data, calculating the sediment volume time series changes based on the sediment volume distribution spatial model constructed using the Kriging three-dimensional interpolation algorithm, and then analyzing the calculated sediment volume time series changes based on the preset ARIMA model to obtain the sediment volume growth rate, for example, the daily average sediment volume growth rate is obtained.
[0108] According to the preset convolutional neural network, feature extraction is performed on the sediment accumulation morphology image data of each reservoir basin area respectively to obtain the corresponding sediment accumulation morphology evolution characteristics; wherein, the sediment accumulation morphology evolution characteristics can be understood as the rapid evolution characteristics of sediment accumulation morphology in different regions obtained by combining the currently collected sediment accumulation morphology image data with the historical sediment accumulation morphology image data obtained at multiple previous collection moments into time series data, and then inputting the preset convolutional neural network to extract feature parameter changes. The preset convolutional neural network in this embodiment can be selected according to actual application requirements, such as using a network structure with 5 convolutional layers and 3 fully connected layers, which is not specifically limited here.
[0109] The sediment thickness distribution data, sediment volume distribution data, sediment thickness change rate, sediment volume growth rate and sediment accumulation morphology evolution characteristics of all reservoir basin areas are input into the hybrid neural network model to predict power generation, and obtain the predicted power generation value; wherein, the hybrid neural network model can be understood as a network model that can be used for short-term prediction of power generation of small hydropower stations by integrating and training multiple neural network models with sediment thickness distribution, sediment volume distribution, sediment thickness change rate, sediment volume growth rate and sediment accumulation morphology evolution characteristics as model input variables, combined with the characteristic parameters of the hydropower station (such as head, flow, etc.) and historical power generation data; this embodiment preferably sets the hybrid neural network model to be obtained by weighted integration of the long short-term memory network model and the support vector regression model. Specifically, the construction steps of the hybrid neural network model include:
[0110] The historical hydropower station characteristic parameter time series data, the historical power generation time series data, and the historical sediment deposition time series data of different reservoir basin areas of the target hydropower station reservoir are obtained; wherein, the historical hydropower station characteristic parameter time series data can be understood as the historical time series data including various sub-characteristic parameters such as head and flow, which are not specifically limited here; the historical sediment deposition time series data include historical sediment thickness distribution time series data, historical sediment volume distribution time series data and historical accumulation morphology image time series data; it should be noted that the thickness distribution data and volume distribution data corresponding to each historical period in the historical sediment thickness distribution time series data and the historical sediment volume distribution time series data can be obtained by referring to the description of the acquisition process of the sediment deposition distribution characteristics of the aforementioned region, and the accumulation morphology image data corresponding to each historical period in the corresponding historical accumulation morphology image time series data can also be obtained by referring to the aforementioned underwater terrain three-dimensional model acquisition method, which will not be described in detail here.
[0111] The historical sediment deposition time series data of each reservoir basin area are analyzed respectively to obtain the corresponding historical sediment evolution time series data and historical sediment influencing factor time series data; wherein, the historical sediment evolution time series data include the historical sediment thickness change rate time series data, the historical sediment volume growth rate time series data and the historical accumulation morphology evolution characteristic time series data, which can be obtained by the acquisition process of the sediment thickness change rate, the sediment volume growth rate and the sediment accumulation morphology evolution characteristics described above, which will not be repeated here; the historical sediment influencing factor time series data include the sediment thickness change coefficient time series data, the volume growth coefficient time series data and the local accumulation morphology coefficient time series data, wherein, each The sedimentation thickness variation coefficient data at each historical moment can be obtained by first calculating the sedimentation thickness standard deviation and the sedimentation thickness average value of the corresponding historical sedimentation thickness distribution sub-time series data (time series data before the historical moment), and then calculating the ratio of the sedimentation thickness standard deviation to the sedimentation thickness average value; the volume growth coefficient data at each historical moment can be obtained by exponential fitting based on the corresponding historical sedimentation volume distribution sub-time series data (time series data before the historical moment); the local accumulation morphology coefficient data at each historical moment can be obtained by extracting texture features based on the corresponding historical accumulation morphology image data (time series data before the historical moment) using the gray level co-occurrence matrix method.
[0112] According to the preset hydrological parameter characteristics, the historical power generation time series data, and the historical deposition influencing factor time series data of each reservoir basin area, the random forest regression model is trained, and the weight values of each deposition influencing factor are obtained according to the trained random forest regression model; wherein the preset hydrological parameter characteristics may include flow, flow velocity, water level, sediment content, water quality, flood season, ice age, groundwater and riverbed geology, etc., which are not specifically limited here; the process of training the random forest regression model in this embodiment can be understood as using the preset hydrological parameter characteristics and the deposition influencing factor data of each reservoir basin area as the model input, and using the power generation data as the model output to train and verify the random forest regression model using a 5-fold cross validation method, and optimizing and updating the model parameters by optimizing the model hyperparameters using a grid search method; correspondingly, the weight values of each deposition influencing factor can be understood as being based on the trained random forest regression model, first using the feature importance analysis method to obtain the weights of each deposition influencing factor in each area, and then averaging the same deposition influencing factor weights of all areas to obtain the weights of each deposition influencing factor relative to the entire reservoir basin.
[0113] A comprehensive analysis is conducted based on the historical sedimentation influence factor time series data of each reservoir basin area to obtain the comprehensive sedimentation influence factor time series data, and the head loss correction coefficient is obtained based on the weight value of each sedimentation influence factor; among them, the comprehensive sedimentation influence factor time series data can be understood as first calculating the average value of the same historical sedimentation influence factor time series data of all reservoir basin areas to obtain the historical sedimentation influence factor average time series data of each sedimentation influence factor, and then performing weighted averaging of the average time series data of each historical sedimentation influence factor according to the corresponding sedimentation influence factor weight to obtain the time series data. Correspondingly, the head loss correction coefficient can be understood as the head loss calculation correction coefficient obtained by multiplying the weight values of each sedimentation influence factor based on the Bernoulli equation theory.
[0114] According to the comprehensive sedimentation influencing factor time series data and the head loss correction coefficient, the historical power generation time series data is corrected to obtain the historical power generation correction time series data; wherein the power correction data at a certain historical moment can be expressed as:
[0115]
[0116] In the formula, and They represent the power generation data and comprehensive sedimentation influencing factor data at time t respectively; It represents the correction factor of head loss; Indicates the corrected power generation data at time t. This embodiment introduces the comprehensive sedimentation influence factor and the head loss correction coefficient to (Q represents the power generation flow, H represents the working head, and η is the generator efficiency) The calculated historical electric power data is corrected and then used for subsequent power generation prediction model training, which can provide reliable data support for the hybrid neural network model to accurately capture the impact of dynamic changes in sediment deposition on power generation, thereby ensuring the effectiveness of adaptive prediction of small hydropower station power generation.
[0117] A model training data set is constructed based on the historical power generation correction time series data, the historical hydropower station characteristic parameter time series data, and the historical sediment thickness distribution time series data, historical sediment volume distribution time series data, and historical sediment evolution time series data of different reservoir basin areas; it should be noted that the construction process of the model training data set may include outlier processing, missing value processing, and normalization processing of various types of data, which will not be described in detail here.
[0118] According to the model training data set, the preset long short-term memory network model and the preset support vector regression model are integrated and trained to obtain the hybrid neural network model; wherein, the preset long short-term memory network model can select an LSTM model including 2 LSTM layers (100 neurons per layer) and 1 fully connected layer (50 neurons); correspondingly, the preset support vector regression model can select a model with a kernel function of RBF; based on the model training data set, the preset long short-term memory network model and the preset support vector regression model are iteratively trained respectively, and the two trained models are weighted integrated according to the preset fusion weights to obtain the required hybrid neural network model. It should be noted that the training method of the preset long short-term memory network model and the preset support vector regression model can be implemented with reference to the existing technology, which will not be repeated here.
[0119] The hybrid neural network model obtained through the above-mentioned method steps can track and analyze the dynamic changes of sediment deposition in real time based on a combination of sediment deposition amount and deposition mode, and effectively capture the impact of the dynamic changes of sediment deposition on power generation. It can make up for the prediction defects caused by the neglect of the impact of sediment deposition in the existing small hydropower station power generation prediction, and significantly improve the accuracy and reliability of short-term power generation prediction. At the same time, considering that the sediment deposition and storage capacity and water level of the actual power station reservoir are dynamically changing, in order to ensure the high efficiency of the prediction analysis and the continuous accuracy of the prediction results under the conditions of dynamic changes in the sediment deposition environment, this embodiment preferably updates the online training of the hybrid neural network model according to a preset period. Specifically, if Figure 2 As shown, the method also includes:
[0120] S16. According to a preset update cycle, the latest deposition influence factor time series data and the corresponding latest deposition influence factor weight are obtained, and the latest power generation power correction time series data is generated according to the latest deposition influence factor time series data and the corresponding latest deposition influence factor weight, and an adaptive update training data set is constructed according to the latest power generation power correction time series data to train and update the hybrid neural network model.
[0121] In this embodiment, the preset update cycle can be selected according to the actual application requirements, such as once every 6 hours or once every 24 hours, etc., and no specific limitation is made this time; the corresponding latest deposition influence factor time series data can be understood as the latest time series data obtained by adding the deposition influence factor data at the update time obtained based on the aforementioned method to the historical deposition influence factor time series data of the previous update cycle; correspondingly, the latest deposition influence factor weight can be understood as the weight value of each deposition influence factor obtained by the aforementioned deposition influence factor weight acquisition method based on the latest deposition influence factor time series data obtained. It should be noted that the method for obtaining the latest power generation correction time series data can also refer to the relevant description of the aforementioned historical power generation correction time series data, which will not be repeated here. The training update of the corresponding hybrid neural network model can be understood as an online iterative optimization using the stochastic gradient descent method and the mean square error as the objective function to adapt to the impact of dynamic changes in sediment deposition on power generation.
[0122] The hybrid neural network model trained, constructed, and updated online dynamically by the above method can directly predict and analyze the power generation at the next moment based on the sediment deposition thickness distribution data, sediment deposition volume distribution data, sediment deposition thickness change rate, sediment volume growth rate, and sediment accumulation morphology evolution characteristics of all reservoir basin areas obtained in real time online in combination with the characteristic parameters of the hydropower station, and obtain the predicted value of power generation considering the sediment deposition situation; at the same time, it can also be used in combination with the preset hydropower station characteristic parameter prediction model (for example, using the ARIMA model to predict and analyze the reservoir water level change trend in the future) and the sediment deposition dynamic change prediction model (for example, using a one-dimensional sediment transport mathematical model to simulate and predict the sediment deposition thickness distribution and sediment deposition volume distribution in the future while considering the suspended load and bed load transport), that is, based on the prediction results of the hydropower station characteristic parameter prediction model, the sediment deposition dynamic change prediction model, and the hybrid neural network model, the basic analysis data of each time to be predicted are updated in turn, so as to realize the rolling prediction of the target hydropower station power generation for a period of time in the future.
[0123] The embodiment of the present application obtains sonar detection data of different reservoir basin areas of the target hydropower station reservoir, and obtains regional sediment accumulation distribution characteristics including regional sediment deposition thickness distribution, regional sediment deposition volume distribution and regional sediment deposition morphological parameter distribution according to the sonar detection data, performs reservoir capacity impact analysis according to the regional sediment deposition distribution characteristics to obtain multiple sedimentation degree classification areas, performs sedimentation terrain volume difference analysis on each sedimentation degree classification area, obtains the corresponding classification area reservoir capacity change, and judges whether the storage capacity change of each classification area meets the preset storage capacity reduction condition. If it meets the condition, the corresponding sedimentation degree classification area is used as the key analysis area, and after the local head loss value of the key analysis area is obtained by simulating the water flow movement condition of the sediment deposition reservoir, a corrected water level storage capacity curve of the target hydropower station reservoir is generated according to the storage capacity change of all classification areas and all local head loss values, and the power generation control standard is obtained according to the corrected water level storage capacity curve analysis. The invention also performs power generation control according to the power generation control standard, and predicts the power generation according to the real-time collected sediment deposition monitoring data of different reservoir basins of the target hydropower station reservoir and the hybrid neural network model, obtains the corresponding power generation prediction value, and provides a scheme for online updating and adjusting the hybrid neural network model, which effectively solves the application defect that the existing small hydropower station power generation prediction analysis ignores the influence of sediment deposition and leads to poor prediction effect. It can not only track and analyze the dynamic changes of sediment deposition in real time by combining sediment deposition amount and deposition mode, so that the power generation prediction model can adapt to the dynamic changes of sediment deposition environment, effectively improve the accuracy and reliability of short-term prediction of power generation of small hydropower station, but also ensure the high efficiency of prediction analysis and the continuous accuracy of prediction results under the conditions of dynamic changes of sediment deposition environment, thereby providing important support for the efficient operation of small hydropower stations and power dispatch optimization, and has high practical value and promotion and application prospects.
[0124] It should be noted that although the steps in the above flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders.
[0125] In one embodiment, Figure 3 As shown, a short-term prediction system for power generation of a small hydropower station is provided, the system comprising:
[0126] The sediment accumulation analysis module 1 is used to obtain the sonar detection data of different reservoir basin areas of the target hydropower station reservoir, and obtain the corresponding regional sediment accumulation distribution characteristics according to the sonar detection data; the regional sediment accumulation distribution characteristics include regional sediment deposition thickness distribution, regional sediment deposition volume distribution and regional sediment deposition morphological parameter distribution;
[0127] The reservoir capacity impact analysis module 2 is used to perform reservoir capacity impact analysis according to the sediment accumulation distribution characteristics of the region, obtain multiple sedimentation degree classification areas, and perform sedimentary terrain volume difference analysis on each sedimentation degree classification area to obtain the corresponding reservoir capacity change amount of the classification area;
[0128] The head loss analysis module 3 is used to determine whether the storage capacity change of each graded area meets the preset storage capacity reduction condition. If so, the corresponding sedimentation degree graded area is used as the key analysis area, and the local head loss value of the key analysis area is obtained by simulating the water flow movement of the sediment deposition reservoir;
[0129] The power generation control analysis module 4 is used to generate a modified water level storage capacity curve of the target hydropower station reservoir according to the storage capacity changes of all graded areas and all local head loss values, and to obtain a power generation control standard according to the modified water level storage capacity curve analysis;
[0130] The power generation prediction module 5 is used to perform power generation control according to the power generation control standard, and to predict the power generation according to the real-time collected sediment deposition monitoring data of different reservoir basins of the target hydropower station reservoir and the hybrid neural network model to obtain the corresponding power generation prediction value.
[0131] In one embodiment, Figure 4 As shown, a short-term prediction system for power generation of a small hydropower station is provided, and the system also includes:
[0132] The model dynamic update module 6 is used to obtain the latest deposition influence factor time series data and the corresponding latest deposition influence factor weight according to a preset update cycle, and generate the latest power generation power correction time series data according to the latest deposition influence factor time series data and the corresponding latest deposition influence factor weight, and construct an adaptive update training data set according to the latest power generation power correction time series data to train and update the hybrid neural network model.
[0133] For the specific definition of the short-term prediction system for the power generation of a small hydropower station, please refer to the definition of the short-term prediction method for the power generation of a small hydropower station mentioned above, and the corresponding technical effects can also be obtained equivalently, which will not be repeated here. Each module in the above-mentioned short-term prediction system for the power generation of a small hydropower station can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0134] In summary, the embodiment of the present invention provides a method and system for short-term prediction of power generation of a small hydropower station. The method for short-term prediction of power generation of a small hydropower station realizes the acquisition of sonar detection data of different reservoir basin areas of the target hydropower station reservoir, and obtains the regional sediment accumulation distribution characteristics corresponding to the regional sediment deposition thickness distribution, the regional sediment deposition volume distribution and the regional sediment deposition morphological parameter distribution according to the sonar detection data, performs reservoir capacity impact analysis according to the regional sediment deposition distribution characteristics to obtain multiple sedimentation degree classification areas, and performs sedimentation terrain volume difference analysis on each sedimentation degree classification area to obtain the corresponding reservoir capacity change of the classification area, and judges whether the reservoir capacity change of each classification area meets the preset reservoir capacity reduction condition respectively. If it meets the condition, the corresponding sedimentation degree classification area is used as the key analysis area, and after the local head loss value of the key analysis area is obtained by simulating the water flow movement condition of the sedimentation reservoir, a corrected water level and storage capacity curve of the target hydropower station reservoir is generated according to the reservoir capacity change of all classification areas and all local head loss values, and the corrected water level and storage capacity curve is generated according to the corrected water level and storage capacity curve of all classification areas. The invention discloses a technical scheme for obtaining power generation control standards by analyzing the position and storage capacity curve, and executing power generation control according to the power generation control standards, and predicting power generation according to the sediment deposition monitoring data of different reservoir basins of the target hydropower station reservoir collected in real time and the hybrid neural network model, so as to obtain the corresponding power generation prediction value. The method uses sonar detection technology to accurately capture the spatial distribution characteristics of reservoir sediment deposition, and on this basis evaluates the influence of sediment deposition on storage capacity and water flow characteristics, introduces a hybrid neural network to track and analyze the dynamic changes of sediment deposition in real time based on the combination of sediment deposition amount and deposition mode, and adaptively predicts the power generation of small hydropower stations based on the influence of the dynamic changes of sediment deposition on power generation. The method significantly improves the accuracy and reliability of short-term prediction of power generation of small hydropower stations, and ensures the high efficiency of prediction and analysis under the conditions of dynamic changes in sediment deposition environment and the continuous accuracy of prediction results, thereby providing important support for the efficient operation of small hydropower stations and power dispatch optimization, and has high practical value and promotion and application prospects.
[0135] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0136] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.
Claims
1. A short-term prediction method for power generation of a small hydropower station, characterized in that: The method comprises the following steps: Acquire sonar detection data of different reservoir basins of the target hydropower station reservoir, and obtain corresponding regional sediment accumulation distribution characteristics based on the sonar detection data; the regional sediment accumulation distribution characteristics include regional sediment deposition thickness distribution, regional sediment deposition volume distribution and regional sediment deposition morphological parameter distribution; According to the sediment accumulation distribution characteristics of the region, reservoir capacity impact analysis is performed to obtain multiple sedimentation degree classification areas, and sedimentary terrain volume difference analysis is performed on each sedimentation degree classification area to obtain the corresponding reservoir capacity change amount of the classification area; Determine whether the storage capacity change of each graded area meets the preset storage capacity reduction condition. If so, take the corresponding sedimentation degree graded area as the key analysis area, and obtain the local head loss value of the key analysis area by simulating the water flow movement of the sediment deposition reservoir; Generate a modified water level storage capacity curve of the target hydropower station reservoir according to the storage capacity changes of all graded areas and all local head loss values, and obtain a power generation control standard according to the modified water level storage capacity curve; Execute power generation control according to the power generation control standard, and perform power generation prediction based on the real-time collected sediment deposition monitoring data of different reservoir basins of the target hydropower station reservoir and the hybrid neural network model to obtain the corresponding power generation prediction value; the sediment deposition monitoring data includes sediment deposition thickness distribution data, sediment deposition volume distribution data and sediment accumulation morphology image data; The step of predicting the power generation according to the real-time collected sediment deposition monitoring data of different reservoir basins of the target hydropower station reservoir and the hybrid neural network model to obtain the corresponding power generation prediction value comprises: According to the preset ARIMA model, the sediment thickness distribution data and sediment volume distribution data of each reservoir basin area are analyzed respectively to obtain the corresponding sediment thickness change rate and sediment volume growth rate; According to the preset convolutional neural network, feature extraction is performed on the sediment accumulation morphology image data of each reservoir basin area to obtain the corresponding sediment accumulation morphology evolution characteristics; The sediment thickness distribution data, sediment volume distribution data, sediment thickness change rate, sediment volume growth rate and sediment accumulation morphology evolution characteristics of all reservoir basin areas are input into the hybrid neural network model to predict the power generation to obtain the predicted power generation value; the hybrid neural network model is a network model for short-term prediction of the power generation of small hydropower stations, which is obtained by integrating the long short-term memory network model and the support vector regression model with the sediment thickness distribution, sediment volume distribution, sediment thickness change rate, sediment volume growth rate and sediment accumulation morphology evolution characteristics as model input variables and combining the characteristic parameters of the hydropower station and the historical power generation data to obtain an integrated training network model.
2. The method for short-term prediction of power generation of a small hydropower station according to claim 1, characterized in that: The step of obtaining sonar detection data of different reservoir basin areas of the target hydropower station reservoir and obtaining the corresponding regional sediment accumulation distribution characteristics according to the sonar detection data includes: A full coverage scan of the target hydropower station reservoir is performed by a multi-beam sonar detector to obtain underwater topographic information of different reservoir basin areas; The time domain reflection coefficient analysis and spectrum analysis are performed on the underwater topography information of each reservoir basin area to obtain the corresponding regional sediment layer interface information; Calculate the corresponding regional sediment thickness distribution according to the regional sediment layer interface information and the preset underwater terrain model; According to the regional sediment deposition thickness distribution, the corresponding regional sediment deposition volume distribution is obtained based on the triangular prism volume formula; A three-dimensional surface of a sediment accumulation body is constructed by using a triangular mesh model, and the interior of the sediment accumulation body is filled by a voxelization method to obtain a corresponding regional sediment accumulation three-dimensional model; A morphological parameter analysis is performed based on the three-dimensional model of regional sediment accumulation to obtain a corresponding regional sediment deposition morphological parameter distribution; the regional sediment deposition morphological parameter distribution includes a slope distribution and a curvature distribution.
3. The method for short-term prediction of power generation of a small hydropower station according to claim 1, characterized in that: The step of performing reservoir capacity impact analysis according to the sediment accumulation distribution characteristics of the region to obtain multiple sedimentation degree classification areas comprises: According to the distribution of regional sediment deposition morphological parameters in each reservoir basin area, a corresponding spatial variation function model is selected, and according to the spatial variation function model, the regional sediment deposition thickness distribution and the regional sediment deposition volume distribution, a corresponding regional sediment deposition thickness distribution map and a regional sediment deposition volume distribution map are obtained by the Kriging interpolation method; Performing elevation difference analysis on the regional sediment thickness distribution map and the original reservoir basin regional topographic feature map to obtain a corresponding regional topographic change map; According to the terrain change maps of all regions, the area proportions of regions with different degrees of change are counted, and the threshold value of reservoir volume impact change is obtained according to the area proportions of regions with different degrees of change; Obtaining the sediment deposition impact area according to the regional terrain change map and the reservoir volume impact change threshold; The sedimentation degree of all sediment deposition affected areas is graded by fuzzy clustering algorithm, and multiple sedimentation degree graded areas are obtained.
4. The method for short-term prediction of power generation of a small hydropower station according to claim 1, characterized in that: The step of analyzing the difference in sedimentary terrain volume for each sedimentary degree classification area to obtain the corresponding storage capacity change amount of the classification area includes: The zoning statistics tool and grid calculator are used to analyze each sedimentation degree classification area to obtain the corresponding sedimentation area and average sedimentation thickness; According to the sedimentation area and average sedimentation thickness of each sedimentation degree classification area, the optimal boundary value of each sedimentation degree classification area is obtained; According to the optimal demarcation value, a distribution map of sediment deposition differences is generated by using a grid reclassification tool; According to the sediment deposition degree difference distribution map, the terrain volume difference before and after sediment deposition is analyzed for each sediment degree classification area, and the reservoir capacity change of the corresponding classification area is obtained.
5. The method for short-term prediction of power generation of a small hydropower station according to claim 1, characterized in that: The step of obtaining the local head loss value of the key analysis area by simulating the water flow movement condition of the sediment deposition reservoir comprises: Acquire a regional terrain mask of the key analysis area, and adjust terrain data in a pre-constructed two-dimensional hydrodynamic model according to the regional terrain mask; According to the adjusted two-dimensional hydrodynamic model, the water flow movement condition of the sediment deposition reservoir is simulated and analyzed to obtain corresponding hydrodynamic simulation results; the hydrodynamic simulation results include regional flow velocity change data and water level change data; According to the hydrodynamic simulation results, water level data of each key analysis area is obtained, and a water level change distribution map is generated by Kriging interpolation method according to the water level data of each key analysis area; According to the water level change distribution diagram, the water level change situation of each key analysis area is obtained, and according to the water level change situation of each key analysis area, the corresponding local head loss value is calculated based on the Bernoulli equation.
6. The method for short-term prediction of power generation of a small hydropower station according to claim 1, characterized in that: The power generation control standards include the inflow control standards and the unit start-stop water level standards; The steps of generating a modified water level storage capacity curve of the target hydropower station reservoir according to the storage capacity changes of all graded areas and all local head loss values, and obtaining a power generation control standard according to the modified water level storage capacity curve analysis include: According to the storage capacity changes of the graded areas of all the graded areas of sedimentation degree, a total storage capacity change is obtained, and according to the total storage capacity change, a current reservoir capacity of the target hydropower station reservoir is obtained; Acquire the historical water level and storage capacity time series data of the target hydropower station reservoir, and obtain the water level and storage capacity curve through polynomial regression fitting according to the historical water level and storage capacity time series data, the current reservoir capacity and the corresponding current reservoir water level; According to the local head loss values of all key analysis areas, an average local head loss value is obtained, and the reservoir water level in the water level storage capacity curve is corrected by the local head loss average value to obtain a corrected water level storage capacity curve; According to the modified water level storage capacity curve, an inflow flow optimization data set is generated by a Monte Carlo simulation method; the inflow flow optimization data set includes a plurality of groups of inflow flows and outflow flows corresponding to different water levels; A multi-objective reservoir operation optimization model is established with the goal of maximizing power generation and minimizing water abandonment. According to the inflow optimization data set, a genetic algorithm is used to solve the multi-objective reservoir operation optimization model to obtain the inflow control standard; the inflow control standard includes inflow control thresholds corresponding to different water levels; Through the fuzzy comprehensive evaluation method, a unit start-stop evaluation index system is established, and based on the unit start-stop evaluation index system, a comprehensive analysis is carried out based on the Delphi method and the hierarchical analysis method to obtain the unit start-stop water level standard; the unit start-stop evaluation index system includes a first-level indicator of power generation efficiency, a first-level indicator of flood discharge capacity, a first-level indicator of reservoir safety, and a number of second-level indicators corresponding to each first-level indicator.
7. The method for short-term prediction of power generation of a small hydropower station according to claim 1, characterized in that: The hybrid neural network model is obtained by weighted integration of a long short-term memory network model and a support vector regression model; the construction steps of the hybrid neural network model include: Acquire the historical hydropower station characteristic parameter time series data, the historical power generation time series data, and the historical sediment deposition time series data of different reservoir basin areas of the target hydropower station reservoir; the historical sediment deposition time series data includes the historical sediment thickness distribution time series data, the historical sediment volume distribution time series data and the historical accumulation morphology image time series data; The historical sediment deposition time series data of each reservoir basin area are analyzed respectively to obtain the corresponding historical sediment evolution time series data and historical sediment influencing factor time series data; the historical sediment evolution time series data include the historical sediment thickness change rate time series data, the historical sediment volume growth rate time series data and the historical accumulation morphology evolution characteristic time series data; the historical sediment influencing factor time series data include the sediment thickness change coefficient time series data, the volume growth coefficient time series data and the local accumulation morphology coefficient time series data; The random forest regression model is trained according to the preset hydrological parameter characteristics, the historical power generation time series data, and the historical sedimentation influencing factor time series data of each reservoir basin area, and the weight value of each sedimentation influencing factor is obtained according to the trained random forest regression model; Comprehensive analysis is performed based on the historical sedimentation influencing factor time series data of each reservoir basin area to obtain the comprehensive sedimentation influencing factor time series data, and the head loss correction coefficient is obtained based on the weight value of each sedimentation influencing factor; According to the comprehensive sedimentation influencing factor time series data and the head loss correction coefficient, the historical power generation time series data is corrected to obtain the historical power generation correction time series data; Constructing a model training data set based on the historical power generation correction time series data, the historical hydropower station characteristic parameter time series data, and the historical sediment thickness distribution time series data, the historical sediment volume distribution time series data, and the historical sediment evolution time series data of different reservoir basin areas; According to the model training data set, integrated learning training is performed on the preset long short-term memory network model and the preset support vector regression model to obtain the hybrid neural network model.
8. The method for short-term prediction of power generation of a small hydropower station according to claim 1, characterized in that: The method further comprises: According to the preset update cycle, the latest deposition influence factor time series data and the corresponding latest deposition influence factor weight are obtained, and the latest power generation power correction time series data are generated according to the latest deposition influence factor time series data and the corresponding latest deposition influence factor weight, and according to the latest power generation power correction time series data, an adaptive update training data set is constructed to train and update the hybrid neural network model.
9. A short-term prediction system for power generation of a small hydropower station, characterized in that: The system comprises: A sediment accumulation analysis module is used to obtain sonar detection data of different reservoir basins of a target hydropower station reservoir, and obtain corresponding regional sediment accumulation distribution characteristics based on the sonar detection data; the regional sediment accumulation distribution characteristics include regional sediment deposition thickness distribution, regional sediment deposition volume distribution and regional sediment deposition morphological parameter distribution; A reservoir capacity impact analysis module is used to perform reservoir capacity impact analysis based on the sediment accumulation distribution characteristics of the region, obtain multiple sedimentation degree classification areas, and perform sedimentary terrain volume difference analysis on each sedimentation degree classification area to obtain the corresponding reservoir capacity change amount of the classification area; The head loss analysis module is used to determine whether the storage capacity change of each graded area meets the preset storage capacity reduction condition. If so, the corresponding sedimentation degree graded area is used as the key analysis area, and the local head loss value of the key analysis area is obtained by simulating the water flow movement of the sediment deposition reservoir; A power generation control analysis module, used to generate a modified water level storage capacity curve of the target hydropower station reservoir according to the storage capacity changes of all graded areas and all local head loss values, and to obtain a power generation control standard according to the modified water level storage capacity curve analysis; A power generation prediction module is used to perform power generation control according to the power generation control standard, and to predict power generation according to the real-time collected sediment deposition monitoring data of different reservoir basins of the target hydropower station reservoir and the hybrid neural network model to obtain a corresponding power generation prediction value; the sediment deposition monitoring data includes sediment deposition thickness distribution data, sediment deposition volume distribution data and sediment accumulation morphology image data; The method of predicting power generation based on the real-time collected sedimentation monitoring data of different reservoir basins of the target hydropower station reservoir and the hybrid neural network model to obtain the corresponding power generation prediction value includes: According to the preset ARIMA model, the sediment thickness distribution data and sediment volume distribution data of each reservoir basin area are analyzed respectively to obtain the corresponding sediment thickness change rate and sediment volume growth rate; According to the preset convolutional neural network, feature extraction is performed on the sediment accumulation morphology image data of each reservoir basin area to obtain the corresponding sediment accumulation morphology evolution characteristics; The sediment thickness distribution data, sediment volume distribution data, sediment thickness change rate, sediment volume growth rate and sediment accumulation morphology evolution characteristics of all reservoir basin areas are input into the hybrid neural network model to predict the power generation to obtain the predicted power generation value; the hybrid neural network model is a network model for short-term prediction of the power generation of small hydropower stations, which is obtained by integrating the long short-term memory network model and the support vector regression model with the sediment thickness distribution, sediment volume distribution, sediment thickness change rate, sediment volume growth rate and sediment accumulation morphology evolution characteristics as model input variables and combining the characteristic parameters of the hydropower station and the historical power generation data to obtain an integrated training network model.
