Reservoir water level prediction method
By meshing the reservoir and training the grid water level short-term prediction model, combining historical data and change speed sequence, the problem of low accuracy of reservoir water level prediction in the existing technology is solved, and more accurate water level prediction is achieved.
Patent Information
- Application Number
- CN202411155696.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-22
- Publication Date
- 2024-12-20
- Estimated Expiration
- 2044-08-22
AI Technical Summary
The prior art has low accuracy in reservoir water level prediction, which is mainly due to the failure to comprehensively collect and consider multi-dimensional data, resulting in inaccurate prediction results.
By meshing the water surface structure model of the reservoir, a short-term prediction model of the grid water level is constructed and trained, and prediction is made based on historical water level data and change speed sequences, and the accuracy of reservoir water level prediction is improved through the correction of the surrounding grid.
It improves the accuracy and resolution of reservoir water level prediction, can capture local changes and dynamic trends more accurately, and provides more accurate short-term water level prediction.
Smart Images
Figure CN118674127B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a reservoir water level prediction method, which is applicable to the technical field of reservoir water level prediction. Background Art
[0002] Reservoir water level forecasting is an important part of water resources management and flood prevention and disaster reduction. Accurate water level forecasting can help managers formulate reasonable reservoir scheduling strategies, optimize water resources allocation, improve flood prevention capabilities, and reduce losses caused by natural disasters.
[0003] However, the existing technology usually estimates the water level of the reservoir based on the rainfall, evaporation, leakage, reservoir scheduling operation, upstream water inflow, downstream drainage, etc. in the reservoir area. But in fact, it is difficult to collect comprehensive multi-faceted data at the same time and comprehensively consider various factors. Therefore, the existing technology usually only considers the main factors: rainfall, upstream water inflow and downstream drainage, resulting in the problem of low reservoir water level prediction accuracy in the existing technology. Summary of the invention
[0004] The technical problem to be solved by the present invention is: in view of the above-mentioned problems, a reservoir water level prediction method is provided to solve the problem of low reservoir water level prediction accuracy in the prior art.
[0005] The technical solution adopted by the present invention is: a reservoir water level prediction method, characterized in that it includes:
[0006] Based on the exploration structure of the reservoir, a water surface structure model of the reservoir is constructed and the water surface structure model is meshed;
[0007] Based on the historical water level data sequence of each grid and the water level data after a preset time length of the historical water level data sequence, a grid water level short-term prediction model of each grid is trained;
[0008] Obtain the water level data sequence of the prediction time length and the period before the current time of each grid, and combine the grid water level short-term prediction model of each grid to obtain the predicted water level after the prediction time length of each grid;
[0009] Based on the predicted water levels of the grids surrounding each grid, the predicted water level of the grid is corrected to obtain the corrected water level of each grid;
[0010] Based on the corrected water level of each grid, the predicted water level of the reservoir is determined.
[0011] The grid water level short-term prediction model of each grid is trained based on the historical water level data sequence of each grid and the water level data of the historical water level data sequence after a preset time length, including:
[0012] Based on the historical water level data sequence of each grid, a water level change speed sequence is constructed;
[0013] Based on the water level change rate sequence, determine the water level change coefficient;
[0014] Based on the historical water level data sequence of each grid, the water level variation coefficient and the water level data after a preset time length, and the preset time length, a grid water level short-term prediction model of each grid is trained.
[0015] The water level change speed sequence is constructed based on the historical water level data sequence of each grid, including:
[0016] Subtract the water level data at time t from the water level data at time t-1 in the historical water level data sequence to obtain the water level difference at time t. Divide the water level difference at time t by the interval time to obtain the water level change rate at time t. The water level change rates at multiple times constitute a water level change rate sequence.
[0017] The step of determining the water level variation coefficient based on the water level variation speed sequence includes:
[0018] Among them, ε is the water level change coefficient, c is the denominator coefficient, a T is the water level change rate at the Tth moment in the water level change rate sequence, T is the latest moment in the sequence, and a t is the water level change rate at the tth moment in the water level change rate sequence.
[0019] The grid water level short-term prediction model includes:
[0020] Among them, L TF is the predicted water level of the grid water level short-term prediction model, W1 is the first outer layer weight coefficient, W2 is the second outer layer weight coefficient, W3 is the third outer layer weight coefficient, L t is the water level data at time t in the water level data sequence, ω t is the water level data L at time t t The weight coefficient, b t is the water level data L at time t t The bias coefficient, T F For future time, T C is the current time, ω T is the weight coefficient of time length, b T is the bias coefficient of time length, ε is the water level variation coefficient, ω ε is the weight coefficient of the water level variation coefficient ε, b ε is the bias coefficient of the water level variation coefficient ε, σ is the Sigmoid function, and T is the latest time in the sequence.
[0021] The method of obtaining the prediction time length and the water level data sequence of each grid within a period of time before the current time, combining the grid water level short-term prediction model of each grid, and obtaining the predicted water level of each grid after the prediction time length includes:
[0022] Based on the water level data sequence of each grid in a period of time before the current time, a water level change speed sequence is constructed;
[0023] Based on the water level change rate sequence, determine the water level change coefficient;
[0024] The water level data sequence, water level variation coefficient and prediction time length of each grid in a period of time before the current time are input into the corresponding grid water level short-term prediction model to obtain the predicted water level of each grid after the prediction time length.
[0025] The method of correcting the predicted water level of each grid based on the predicted water levels of the grids surrounding each grid to obtain the corrected water level of each grid includes:
[0026] Centered on each grid area;
[0027] Calculate the deviation coefficient between the predicted water level of the grid area within the central neighborhood and the predicted water level of the central grid area;
[0028] According to the deviation coefficient, the predicted water level in the central grid area is corrected to obtain the corrected water level in the grid area.
[0029] The calculation of the deviation coefficient between the predicted water level of the grid area within the central neighborhood and the predicted water level of the central grid area includes:
[0030] Among them, θ is the deviation coefficient, e is the natural constant, L o Predict water level for the central grid area, L i is the predicted water level of the i-th grid area within the central neighborhood, N is the number of grid areas within the central neighborhood, and i is a positive integer.
[0031] The method of correcting the predicted water level of the central grid area according to the deviation coefficient to obtain the corrected water level of the grid area includes:
[0032] Among them, L r is the corrected water level in the central grid area, θ is the deviation coefficient, L o is the predicted water level in the central grid area, L i is the predicted water level of the i-th grid area within the central neighborhood, N is the number of grid areas within the central neighborhood, and i is a positive integer.
[0033] The method of determining the predicted water level of the reservoir based on the corrected water level of each grid comprises:
[0034] Based on the corrected water level of each grid, calculate the corrected water level mean of each horizontal row grid in the water surface structure model;
[0035] Calculate the first predicted water level based on the corrected water level mean of each horizontal row grid;
[0036] Based on the corrected water level of each grid, calculate the corrected water level mean of each longitudinal column grid in the water surface structure model;
[0037] Calculate the second predicted water level based on the corrected water level mean of each vertical column grid;
[0038] The first predicted water level and the second predicted water level are weighted to obtain the predicted water level of the reservoir.
[0039] The calculating the first predicted water level based on the corrected water level mean of each horizontal row grid includes:
[0040] Among them, L F is the first predicted water level, L j is the corrected mean water level of the jth horizontal row grid, j is a positive integer, and M is the number of corrected mean water levels of the horizontal row grids.
[0041] The calculating the second predicted water level based on the corrected water level mean of each longitudinal column grid includes:
[0042] Among them, L F is the second predicted water level, L j is the corrected mean water level of the jth vertical column grid, j is a positive integer, and M is the number of corrected mean water levels of the vertical column grids.
[0043] A reservoir water level prediction device, characterized by comprising:
[0044] A meshing module is used to construct a water surface structure model of the reservoir based on the exploration structure of the reservoir, and to mesh the water surface structure model;
[0045] A model training module, used to train a grid water level short-term prediction model for each grid based on the historical water level data sequence of each grid and the water level data after a preset time length of the historical water level data sequence;
[0046] The grid water level prediction module is used to obtain the prediction time length and the water level data sequence of each grid within a period of time before the current time, and combine the grid water level short-term prediction model of each grid to obtain the predicted water level after the prediction time length of each grid;
[0047] A grid water level correction module is used to correct the predicted water level of the grid based on the predicted water levels of the grids surrounding each grid to obtain the corrected water level of each grid;
[0048] The reservoir water level determination module is used to determine the predicted water level of the reservoir based on the corrected water level of each grid.
[0049] A storage medium stores a computer program that can be executed by a processor, characterized in that the steps of the reservoir water level prediction method are implemented when the computer program is executed.
[0050] A reservoir water level prediction device comprises a memory and a processor, wherein the memory stores a computer program executable by the processor, and is characterized in that the steps of the reservoir water level prediction method are implemented when the computer program is executed.
[0051] The beneficial effects of the present invention are as follows: the present invention grids the water surface structure model of the reservoir so that the reservoir is divided into multiple grid areas, establishes and trains a grid water level short-term prediction model in each grid area, predicts the water level of each grid area through the trained grid water level short-term prediction model, and corrects the predicted water level of the grid according to the predicted water level of the grids surrounding each grid, thereby improving the water level prediction accuracy of each grid area, and then predicts the reservoir water level according to the corrected water level of each grid area, thereby improving the prediction accuracy of the reservoir water level.
[0052] The present invention can realize independent analysis of different areas of the reservoir by gridding the water surface structure model of the reservoir, thereby improving the resolution of the overall water level prediction. This method can capture local changes that may be ignored by traditional overall models, thereby improving the accuracy of prediction.
[0053] The present invention dynamically reflects the water level change trend of each grid area in the reservoir by constructing a water level change speed sequence. This dynamic tracking can more accurately predict the water level change in the short term in the future. The present invention establishes a grid water level short-term prediction model based on the water level data sequence and the water level change coefficient, so that the grid water level short-term prediction model takes into account the current water level situation and understands the water level change situation, thereby providing a more accurate prediction.
[0054] The present invention calculates the difference between the predicted water level of the central grid area and the predicted water level of the surrounding grid areas to obtain a deviation coefficient. When the deviation coefficient is larger, the water level correction amplitude of the central grid area is larger, so that the same area has the same level of water level.
[0055] The present invention divides a reservoir into a plurality of grid areas, wherein there are a plurality of rows of grid areas in the horizontal direction and a plurality of columns of grid areas in the vertical direction. The horizontal corrected water level mean of each row of grid areas and the vertical corrected water level mean of each column of grid areas are calculated. The first predicted water level is calculated through each horizontal corrected water level mean, and the second predicted water level is calculated through each vertical corrected water level mean. The first predicted water level and the second predicted water level are combined to obtain a predicted reservoir water level, thereby improving the prediction accuracy of the reservoir water level. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] Figure 1 This is a flow chart of Example 1. DETAILED DESCRIPTION
[0057] Example 1: Figure 1 As shown, this embodiment is a reservoir water level prediction method, which specifically includes the following steps:
[0058] S1. Based on the exploration structure of the reservoir, a water surface structure model of the reservoir is constructed and the water surface structure model is meshed.
[0059] This embodiment can realize independent analysis of different areas of the reservoir by gridding the reservoir structure model, thereby improving the resolution of the overall water level prediction. This method can capture local changes that may be ignored by traditional overall models, thereby improving the accuracy of prediction.
[0060] S2. Based on the historical water level data sequence of each grid and the water level data after a preset time length of the historical water level data sequence, a grid water level short-term prediction model of each grid is trained.
[0061] S21, obtaining the historical water level data sequence of each grid and the water level data after a preset time length of the historical water level data sequence, and constructing a water level change speed sequence based on the historical water level data sequence of each grid.
[0062] Subtract the water level data at the tth moment in the historical water level data sequence from the water level data at the t-1th moment to obtain the water level difference at the tth moment. Divide the water level difference at the tth moment by the interval time to obtain the water level change rate at the tth moment. The water level change rates at multiple moments form a water level change rate sequence, where t is the number of the moment.
[0063] S22, determining a water level variation coefficient based on the water level variation speed sequence;
[0064] Among them, ε is the water level change coefficient, c is the denominator coefficient, a T is the water level change rate at the Tth moment in the water level change rate sequence, T is the latest moment in the sequence, and a t is the water level change rate at the tth moment in the water level change rate sequence, and t is the moment number.
[0065] S23, training a grid water level short-term prediction model for each grid based on the historical water level data sequence of each grid, the water level variation coefficient, the water level data after a preset time length, and the preset time length;
[0066] Among them, L TF is the predicted water level of the grid water level short-term prediction model, W1 is the first outer layer weight coefficient, W2 is the second outer layer weight coefficient, W3 is the third outer layer weight coefficient, L t is the water level data at time t in the water level data sequence, ω t is the water level data L at time t t The weight coefficient, b t is the water level data L at time t t The bias coefficient, T F For future time, T C is the current time, ω T is the weight coefficient of time length, b T is the bias coefficient of time length, ε is the water level variation coefficient, ω ε is the weight coefficient of the water level variation coefficient ε, b ε is the bias coefficient of the water level variation coefficient ε, σ is the Sigmoid function, T is the latest moment in the sequence, and t is the moment number.
[0067] The grid water level short-term prediction model in this embodiment takes into account the water level data sequence, prediction time length and water level variation coefficient, and assigns different weights and biases to the water level data sequence, prediction time length and water level variation coefficient, respectively, so that the model accurately assigns weights and biases to each variable, and then uses W1, W2 and W3 to integrate the results of each part, making the prediction results more comprehensive and reliable.
[0068] In this example, each weight and bias can be trained using the gradient descent method. During training, the following are used: historical water level data sequence, prediction time length and water level change coefficient constitute the training sample, and the future time T F The corresponding water level (the water level corresponding to the preset time length after the historical water level data sequence) is used as a label to form a training sample label group. Multiple groups of sample label groups are used to train the weights and biases in the grid water level short-term prediction model until the predicted water level is the same or similar to the label.
[0069] This embodiment dynamically reflects the water level change trend of each grid area in the reservoir by constructing a water level change speed sequence. This dynamic tracking can more accurately predict water level changes in the future within a short period of time.
[0070] This embodiment establishes a grid water level short-term prediction model based on the water level data sequence and the water level variation coefficient, so that the grid water level short-term prediction model takes into account the current water level situation and understands the water level variation situation, thereby providing a more accurate prediction.
[0071] This embodiment uses the mean of the water level change rate sequence to reflect the overall water level change situation, and combines the water level change rate at the latest moment so that the water level change coefficient can reflect the current water level change situation in a timely manner.
[0072] S3. Obtain the prediction time length and the water level data sequence of each grid in a period of time before the current time, and combine the grid water level short-term prediction model of each grid to obtain the predicted water level after the prediction time length of each grid.
[0073] S31. Construct a water level change speed sequence based on the water level data sequence of each grid within a period of time before the current time.
[0074] Subtract the water level data at the tth moment in the water level data sequence from the water level data at the t-1th moment to obtain the water level difference at the tth moment. Divide the water level difference at the tth moment by the interval time to obtain the water level change rate at the tth moment. The water level change rates at multiple moments form a water level change rate sequence, where t is the number of the moment.
[0075] S32, determining a water level variation coefficient based on a water level variation speed sequence;
[0076] Among them, ε is the water level change coefficient, c is the denominator coefficient, a T is the water level change rate at the Tth moment in the water level change rate sequence, T is the latest moment in the sequence, and a t is the water level change rate at the tth moment in the water level change rate sequence, and t is the moment number.
[0077] S33, inputting the water level data sequence, water level variation coefficient and prediction time length of each grid in a period of time before the current time into the trained grid water level short-term prediction model of the corresponding grid, and obtaining the predicted water level of each grid after the prediction time length.
[0078] In this embodiment, the grid water level short-term prediction model must rely on the current water level data sequence to perform short-term predictions of 1 hour and 2 hours. If a long-term prediction is performed, the water level data sequence and the water level variation coefficient will change greatly in the future, and accurate predictions cannot be performed. In this example, the water level data sequence can be measured by a buoy-type water level gauge.
[0079] S4. Based on the predicted water levels of the grids surrounding each grid, the predicted water level of the grid is corrected to obtain the corrected water level of each grid.
[0080] S41, taking each grid area as the center.
[0081] S42, calculating the deviation coefficient between the predicted water level of the grid area within the central neighborhood and the predicted water level of the central grid area;
[0082] Among them, θ is the deviation coefficient, e is the natural constant, L o Predict water level for the central grid area, L i is the predicted water level of the i-th grid area within the central neighborhood, N is the number of grid areas within the central neighborhood, and i is a positive integer.
[0083] S43, correcting the predicted water level of the central grid area according to the deviation coefficient to obtain a corrected water level of the grid area;
[0084] Among them, L r is the corrected water level in the central grid area, θ is the deviation coefficient, L o is the predicted water level in the central grid area, L i is the predicted water level of the i-th grid area within the central neighborhood, N is the number of grid areas within the central neighborhood, and i is a positive integer.
[0085] In this embodiment, the reservoir is divided into grid areas, and there are multiple grid areas around each grid area. According to the prediction situation of each grid area, the predicted water level at the center is corrected to obtain the corrected water level.
[0086] This embodiment will calculate the difference between the predicted water level at the center and the predicted water level in the surrounding grid area to obtain a deviation coefficient. The larger the difference, the larger the deviation coefficient. When the deviation coefficient is larger, the larger the correction amplitude of the water level at the center is, so that the same area has the same level of water level.
[0087] S5. Determine the predicted water level of the reservoir based on the corrected water level of each grid.
[0088] S51. Based on the corrected water level of each grid, calculate the corrected water level mean of each horizontal row grid in the water surface structure model.
[0089] S52, calculating a first predicted water level based on the corrected water level mean of each horizontal row grid;
[0090] S53, based on the corrected water level of each grid, calculating the corrected water level mean of each longitudinal column grid in the water surface structure model;
[0091] S54, calculating a second predicted water level based on the corrected water level mean of each longitudinal column grid;
[0092] S55. Perform weighted processing on the first predicted water level and the second predicted water level to obtain the predicted water level of the reservoir.
[0093] In this embodiment, the calculation formulas for the first predicted water level and the second predicted water level are as follows:
[0094] Among them, L F is the first predicted water level or the second predicted water level, L j is the jth transverse corrected water level mean (corresponding to the first predicted water level) or the jth longitudinal corrected water level mean (corresponding to the second predicted water level), j is a positive integer, and M is the number of transverse corrected water level means or longitudinal corrected water level means.
[0095] This embodiment divides the reservoir into multiple grid areas, with multiple rows of grid areas in the horizontal direction and multiple columns of grid areas in the vertical direction. The horizontal corrected water level mean of each row of grid areas and the vertical corrected water level mean of each column of grid areas are calculated. The first predicted water level is calculated through the horizontal corrected water level means, and the second predicted water level is calculated through the vertical corrected water level means. The predicted reservoir water level is obtained by combining the first predicted water level and the second predicted water level, thereby improving the prediction accuracy of the reservoir water level.
[0096] Embodiment 2: This embodiment is a reservoir water level prediction device, including a grid division module, a model training module, a grid water level prediction module, a grid water level correction module and a reservoir water level determination module.
[0097] In this example, the grid division module is used to construct a water surface structure model of the reservoir based on the exploration structure of the reservoir, and to grid the water surface structure model; the model training module is used to train the grid water level short-term prediction model of each grid based on the historical water level data sequence of each grid and the water level data after a preset time length of the historical water level data sequence; the grid water level prediction module is used to obtain the prediction time length and the water level data sequence within a period of time before the current time of each grid, and combine the grid water level short-term prediction model of each grid to obtain the predicted water level after the prediction time length of each grid.
[0098] In this embodiment, the grid water level correction module is used to correct the predicted water level of the grid based on the predicted water levels of the grids surrounding each grid to obtain the corrected water level of each grid; the reservoir water level determination module is used to determine the predicted water level of the reservoir based on the corrected water level of each grid.
[0099] Embodiment 3: This embodiment is a storage medium on which a computer program that can be executed by a processor is stored. When the computer program is executed, the steps of the reservoir water level prediction method in Embodiment 1 are implemented.
[0100] Embodiment 4: This embodiment is a reservoir water level prediction device, which has a memory and a processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed, the steps of the reservoir water level prediction method in Embodiment 1 are implemented.
Claims
1. A reservoir water level prediction method, characterized in that: include: Based on the exploration structure of the reservoir, a water surface structure model of the reservoir is constructed and the water surface structure model is meshed; Based on the historical water level data sequence of each grid and the water level data after a preset time length of the historical water level data sequence, a grid water level short-term prediction model of each grid is trained; Obtain the water level data sequence of the prediction time length and the period before the current time of each grid, and combine the grid water level short-term prediction model of each grid to obtain the predicted water level after the prediction time length of each grid; Based on the predicted water levels of the grids surrounding each grid, the predicted water level of the grid is corrected to obtain the corrected water level of each grid; Determine the predicted water level of the reservoir based on the corrected water level of each grid; The grid water level short-term prediction model of each grid is trained based on the historical water level data sequence of each grid and the water level data of the historical water level data sequence after a preset time length, including: Based on the historical water level data series of each grid, a water level change speed series is constructed; Based on the water level change rate sequence, determine the water level change coefficient; Based on the historical water level data sequence of each grid, the water level variation coefficient and the water level data after a preset time length, and the preset time length, a grid water level short-term prediction model of each grid is trained; The step of determining the water level variation coefficient based on the water level variation speed sequence includes: Among them, ε is the water level change coefficient, c is the denominator coefficient, a T is the water level change rate at the Tth moment in the water level change rate sequence, T is the latest moment in the sequence, and a t is the water level change rate at the tth moment in the water level change rate sequence; The grid water level short-term prediction model includes: Among them, L TF is the predicted water level of the grid water level short-term prediction model, W1 is the first outer layer weight coefficient, W2 is the second outer layer weight coefficient, W3 is the third outer layer weight coefficient, L t is the water level data at time t in the water level data sequence, ω t is the water level data L at time t t The weight coefficient, b t is the water level data L at time t t The bias coefficient, T F For future time, T C is the current time, ω T is the weight coefficient of time length, b T is the bias coefficient of time length, ε is the water level variation coefficient, ω ε is the weight coefficient of the water level variation coefficient ε, b ε is the bias coefficient of the water level variation coefficient ε, σ is the Sigmoid function, and T is the latest time in the sequence; The method of correcting the predicted water level of each grid based on the predicted water levels of the grids surrounding each grid to obtain the corrected water level of each grid includes: Centered on each grid area; Calculate the deviation coefficient between the predicted water level of the grid area within the central neighborhood and the predicted water level of the central grid area; According to the deviation coefficient, the predicted water level of the central grid area is corrected to obtain the corrected water level of the grid area; The method of determining the predicted water level of the reservoir based on the corrected water level of each grid comprises: Based on the corrected water level of each grid, calculate the corrected water level mean of each horizontal row grid in the water surface structure model; Calculate the first predicted water level based on the corrected water level mean of each horizontal row grid; Based on the corrected water level of each grid, calculate the corrected water level mean of each longitudinal column grid in the water surface structure model; Calculate the second predicted water level based on the corrected water level mean of each vertical column grid; The first predicted water level and the second predicted water level are weighted to obtain the predicted water level of the reservoir.
2. The reservoir water level prediction method according to claim 1, characterized in that: The water level change speed sequence is constructed based on the historical water level data sequence of each grid, including: Subtract the water level data at time t from the water level data at time t-1 in the historical water level data sequence to obtain the water level difference at time t. Divide the water level difference at time t by the interval time to obtain the water level change rate at time t. The water level change rates at multiple times constitute a water level change rate sequence.
3. The reservoir water level prediction method according to claim 1, characterized in that: The method of obtaining the prediction time length and the water level data sequence of each grid within a period of time before the current time, combining the grid water level short-term prediction model of each grid, and obtaining the predicted water level of each grid after the prediction time length includes: Based on the water level data sequence of each grid in a period of time before the current time, a water level change speed sequence is constructed; Based on the water level change rate sequence, determine the water level change coefficient; The water level data sequence, water level variation coefficient and prediction time length of each grid in a period of time before the current time are input into the corresponding grid water level short-term prediction model to obtain the predicted water level of each grid after the prediction time length.
4. The reservoir water level prediction method according to claim 1, characterized in that: The calculation of the deviation coefficient between the predicted water level of the grid area within the central neighborhood and the predicted water level of the central grid area includes: Among them, θ is the deviation coefficient, e is the natural constant, L o Predict water level for the central grid area, L i is the predicted water level of the i-th grid area within the central neighborhood, N is the number of grid areas within the central neighborhood, and i is a positive integer.
5. The reservoir water level prediction method according to claim 1, characterized in that: The method of correcting the predicted water level of the central grid area according to the deviation coefficient to obtain the corrected water level of the grid area includes: Among them, L r is the corrected water level in the central grid area, θ is the deviation coefficient, L o is the predicted water level in the central grid area, L i is the predicted water level of the i-th grid area within the central neighborhood, N is the number of grid areas within the central neighborhood, and i is a positive integer.
6. The reservoir water level prediction method according to claim 1, characterized in that: The calculating the first predicted water level based on the corrected water level mean of each horizontal row grid includes: Among them, L F is the first predicted water level, L j is the corrected mean water level of the jth horizontal row grid, j is a positive integer, and M is the number of corrected mean water levels of the horizontal row grids.
7. The reservoir water level prediction method according to claim 1, characterized in that: The calculating the second predicted water level based on the corrected water level mean of each longitudinal column grid includes: Among them, L F is the second predicted water level, L j is the corrected mean water level of the jth vertical column grid, j is a positive integer, and M is the number of corrected mean water levels of the vertical column grids.
8. A reservoir water level prediction device, characterized in that: include: A meshing module is used to construct a water surface structure model of the reservoir based on the exploration structure of the reservoir, and to mesh the water surface structure model; A model training module, used to train a grid water level short-term prediction model for each grid based on the historical water level data sequence of each grid and the water level data after a preset time length of the historical water level data sequence; The grid water level prediction module is used to obtain the prediction time length and the water level data sequence of each grid within a period of time before the current time, and combine the grid water level short-term prediction model of each grid to obtain the predicted water level after the prediction time length of each grid; A grid water level correction module is used to correct the predicted water level of the grid based on the predicted water levels of the grids surrounding each grid to obtain the corrected water level of each grid; A reservoir water level determination module is used to determine the predicted water level of the reservoir based on the corrected water level of each grid; The grid water level short-term prediction model of each grid is trained based on the historical water level data sequence of each grid and the water level data of the historical water level data sequence after a preset time length, including: Based on the historical water level data series of each grid, a water level change speed series is constructed; Based on the water level change rate sequence, determine the water level change coefficient; Based on the historical water level data sequence of each grid, the water level variation coefficient and the water level data after a preset time length, and the preset time length, a grid water level short-term prediction model of each grid is trained; The step of determining the water level variation coefficient based on the water level variation speed sequence includes: Among them, ε is the water level change coefficient, c is the denominator coefficient, a T is the water level change rate at the Tth moment in the water level change rate sequence, T is the latest moment in the sequence, and a t is the water level change rate at the tth moment in the water level change rate sequence; The grid water level short-term prediction model includes: Among them, L TF is the predicted water level of the grid water level short-term prediction model, W1 is the first outer layer weight coefficient, W2 is the second outer layer weight coefficient, W3 is the third outer layer weight coefficient, L t is the water level data at time t in the water level data sequence, ω t is the water level data L at time t t The weight coefficient, b t is the water level data L at time t t The bias coefficient, T F For future time, T C is the current time, ω T is the weight coefficient of time length, b T is the bias coefficient of time length, ε is the water level variation coefficient, ω ε is the weight coefficient of the water level variation coefficient ε, b ε is the bias coefficient of the water level variation coefficient ε, σ is the Sigmoid function, and T is the latest time in the sequence; The method of correcting the predicted water level of each grid based on the predicted water levels of the grids surrounding each grid to obtain the corrected water level of each grid includes: Centered on each grid area; Calculate the deviation coefficient between the predicted water level of the grid area within the central neighborhood and the predicted water level of the central grid area; According to the deviation coefficient, the predicted water level of the central grid area is corrected to obtain the corrected water level of the grid area; The method of determining the predicted water level of the reservoir based on the corrected water level of each grid comprises: Based on the corrected water level of each grid, calculate the corrected water level mean of each horizontal row grid in the water surface structure model; Calculate the first predicted water level based on the corrected water level mean of each horizontal row grid; Based on the corrected water level of each grid, calculate the corrected water level mean of each longitudinal column grid in the water surface structure model; Calculate the second predicted water level based on the corrected water level mean of each vertical column grid; The first predicted water level and the second predicted water level are weighted to obtain the predicted water level of the reservoir.
9. A storage medium having stored thereon a computer program executable by a processor, characterized in that: When the computer program is executed, the steps of the reservoir water level prediction method according to any one of claims 1 to 7 are implemented.
10. A reservoir water level prediction device, comprising a memory and a processor, wherein the memory stores a computer program executable by the processor, characterized in that: When the computer program is executed, the steps of the reservoir water level prediction method according to any one of claims 1 to 7 are implemented.