Upper reservoir water level prediction method based on guide vane opening degree-active power-steel pipe pressure
By constructing a water level prediction model for the upper reservoir based on guide vane opening, active power, and steel pipe pressure, and using an encoder and decoder for feature extraction and prediction, the problem of inaccurate water level prediction caused by traditional manual experience judgment is solved, and higher prediction accuracy is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-05
- Publication Date
- 2026-03-10
AI Technical Summary
Traditional methods of human experience in predicting water levels in the upper reservoir of hydropower stations are subjective and result in low accuracy.
By constructing a water level prediction model for the upper reservoir based on guide vane opening, active power, and steel pipe pressure, an encoder is used for feature extraction and attention weighting, combined with a decoder for upper reservoir water level prediction. The predicted values are then used for model training, and the model parameters are optimized to improve prediction accuracy.
The accuracy of upper reservoir water level prediction has been improved by strengthening the feature weights at key time points and optimizing model parameters through deep feature extraction and mapping transformation to obtain accurate upper reservoir water level prediction results.
Smart Images
Figure CN121638564A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for predicting upper reservoir water level based on guide vane opening degree, active power, and steel pipe pressure. Background Technology
[0002] Currently, in order to ensure the safe and stable operation of hydropower stations, accurate prediction of the upper reservoir water level is of paramount importance.
[0003] In traditional techniques, the prediction of the upper reservoir water level of a hydropower station is generally based on manual experience. However, this method of manual experience is subjective and prone to prediction errors, resulting in low accuracy in the prediction of the upper reservoir water level. Summary of the Invention
[0004] Therefore, it is necessary to provide a method, device, computer equipment, computer-readable storage medium, and computer program product for predicting upper reservoir water level based on guide vane opening degree, active power, and steel pipe pressure, which can improve the accuracy of upper reservoir water level prediction.
[0005] Firstly, this application provides a method for predicting the upper reservoir water level based on guide vane opening degree, active power, and steel pipe pressure. The method includes:
[0006] Based on the operating data of the hydropower station's generating units, the characteristic data of the generating units are determined; the operating data includes guide vane opening data, active power data, and steel pipe pressure data.
[0007] Based on the feature data, a reservoir water level prediction model is constructed; the reservoir water level prediction model includes an encoder and a decoder.
[0008] The encoder of the upper reservoir water level prediction model performs feature extraction and attention weighting on the feature data to obtain enhanced feature data.
[0009] The enhanced feature data is processed by the decoder of the upper reservoir water level prediction model to predict the upper reservoir water level, thereby obtaining the predicted upper reservoir water level value of the hydropower station.
[0010] The upper reservoir water level prediction model is trained using the predicted upper reservoir water level to obtain the target upper reservoir water level prediction model.
[0011] The current operating data of the hydropower station is input into the target upper reservoir water level prediction model for upper reservoir water level prediction processing to obtain the upper reservoir water level prediction result of the hydropower station.
[0012] In one embodiment, the encoder of the upper reservoir water level prediction model performs feature extraction and attention weighting processing on the feature data to obtain enhanced feature data, including:
[0013] The encoder of the upper reservoir water level prediction model is used to perform feature extraction processing on the feature data to obtain time-series feature data.
[0014] The time-series feature data is subjected to attention-weighted processing to obtain the enhanced feature data.
[0015] In one embodiment, the attention-weighted processing of the temporal feature data to obtain the enhanced feature data includes:
[0016] Based on the aforementioned temporal feature data, attention weights are determined;
[0017] Based on the attention weights, the target time point feature data in the time series feature data is subjected to attention weighting processing to obtain the enhanced feature data.
[0018] In one embodiment, the step of performing upper reservoir water level prediction processing on the enhanced feature data through the decoder of the upper reservoir water level prediction model to obtain the upper reservoir water level prediction value of the hydropower station includes:
[0019] The enhanced feature data is processed by the decoder of the upper reservoir water level prediction model to obtain the enhanced feature data after feature mapping.
[0020] The enhanced feature data after feature mapping is processed to predict the upper reservoir water level, thereby obtaining the predicted upper reservoir water level value.
[0021] In one embodiment, before training the upper reservoir water level prediction model using the upper reservoir water level prediction value to obtain the target upper reservoir water level prediction model, the method further includes:
[0022] Obtain historical reservoir water level data;
[0023] The historical reservoir water level data is processed by a sliding window mean to obtain the target upper reservoir water level data;
[0024] The step of training the upper reservoir water level prediction model using the predicted upper reservoir water level to obtain the target upper reservoir water level prediction model includes:
[0025] Based on the predicted upper reservoir water level and the target upper reservoir water level data, the upper reservoir water level prediction model is trained to obtain the target upper reservoir water level prediction model.
[0026] In one embodiment, determining the characteristic data of the generating units based on the operating data of the hydropower station units includes:
[0027] The running data is subjected to outlier removal processing to obtain outlier-removed running data;
[0028] The feature data is determined based on the operational data after outlier removal.
[0029] Secondly, this application also provides a device for predicting the upper reservoir water level based on guide vane opening degree, active power, and steel pipe pressure. The device includes:
[0030] The data determination module is used to determine the characteristic data of the generating units based on the operating data of the hydropower station's generating units; the operating data includes guide vane opening data, active power data, and steel pipe pressure data.
[0031] The model building module is used to build a reservoir water level prediction model based on the feature data; the reservoir water level prediction model includes an encoder and a decoder.
[0032] The feature extraction module is used to perform feature extraction and attention weighting processing on the feature data through the encoder of the upper reservoir water level prediction model to obtain enhanced feature data.
[0033] The water level prediction module is used to perform upper reservoir water level prediction processing on the enhanced feature data through the decoder of the upper reservoir water level prediction model to obtain the upper reservoir water level prediction value of the hydropower station.
[0034] The model training module is used to train the upper reservoir water level prediction model using the upper reservoir water level prediction value to obtain the target upper reservoir water level prediction model.
[0035] The data input module is used to input the current operating data of the hydropower station into the target upper reservoir water level prediction model for upper reservoir water level prediction processing, and to obtain the upper reservoir water level prediction result of the hydropower station.
[0036] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:
[0037] Based on the operating data of the hydropower station's generating units, the characteristic data of the generating units are determined; the operating data includes guide vane opening data, active power data, and steel pipe pressure data.
[0038] Based on the feature data, a reservoir water level prediction model is constructed; the reservoir water level prediction model includes an encoder and a decoder.
[0039] The encoder of the upper reservoir water level prediction model performs feature extraction and attention weighting on the feature data to obtain enhanced feature data.
[0040] The enhanced feature data is processed by the decoder of the upper reservoir water level prediction model to predict the upper reservoir water level, thereby obtaining the predicted upper reservoir water level value of the hydropower station.
[0041] The upper reservoir water level prediction model is trained using the predicted upper reservoir water level to obtain the target upper reservoir water level prediction model.
[0042] The current operating data of the hydropower station is input into the target upper reservoir water level prediction model for upper reservoir water level prediction processing to obtain the upper reservoir water level prediction result of the hydropower station.
[0043] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:
[0044] Based on the operating data of the hydropower station's generating units, the characteristic data of the generating units are determined; the operating data includes guide vane opening data, active power data, and steel pipe pressure data.
[0045] Based on the feature data, a reservoir water level prediction model is constructed; the reservoir water level prediction model includes an encoder and a decoder.
[0046] The encoder of the upper reservoir water level prediction model performs feature extraction and attention weighting on the feature data to obtain enhanced feature data.
[0047] The enhanced feature data is processed by the decoder of the upper reservoir water level prediction model to predict the upper reservoir water level, thereby obtaining the predicted upper reservoir water level value of the hydropower station.
[0048] The upper reservoir water level prediction model is trained using the predicted upper reservoir water level to obtain the target upper reservoir water level prediction model.
[0049] The current operating data of the hydropower station is input into the target upper reservoir water level prediction model for upper reservoir water level prediction processing to obtain the upper reservoir water level prediction result of the hydropower station.
[0050] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:
[0051] Based on the operating data of the hydropower station's generating units, the characteristic data of the generating units are determined; the operating data includes guide vane opening data, active power data, and steel pipe pressure data.
[0052] Based on the feature data, a reservoir water level prediction model is constructed; the reservoir water level prediction model includes an encoder and a decoder.
[0053] The encoder of the upper reservoir water level prediction model performs feature extraction and attention weighting on the feature data to obtain enhanced feature data.
[0054] The enhanced feature data is processed by the decoder of the upper reservoir water level prediction model to predict the upper reservoir water level, thereby obtaining the predicted upper reservoir water level value of the hydropower station.
[0055] The upper reservoir water level prediction model is trained using the predicted upper reservoir water level to obtain the target upper reservoir water level prediction model.
[0056] The current operating data of the hydropower station is input into the target upper reservoir water level prediction model for upper reservoir water level prediction processing to obtain the upper reservoir water level prediction result of the hydropower station.
[0057] The aforementioned method, apparatus, computer equipment, computer-readable storage medium, and computer program product for predicting upper reservoir water level based on guide vane opening, active power, and steel pipe pressure determine the characteristic data of the hydropower station's generating units based on their operating data. The operating data includes guide vane opening data, active power data, and steel pipe pressure data. An upper reservoir water level prediction model is constructed based on this characteristic data. The upper reservoir water level prediction model includes an encoder and a decoder. The encoder of the upper reservoir water level prediction model performs feature extraction and attention weighting processing on the characteristic data to obtain enhanced feature data. The decoder of the upper reservoir water level prediction model performs upper reservoir water level prediction processing on the enhanced feature data to obtain the predicted upper reservoir water level value of the hydropower station. The predicted upper reservoir water level value is used to train the upper reservoir water level prediction model to obtain a target upper reservoir water level prediction model. The current operating data of the hydropower station is input into the target upper reservoir water level prediction model for upper reservoir water level prediction processing to obtain the predicted upper reservoir water level result of the hydropower station. This scheme determines characteristic data based on the guide vane opening, active power, and steel pipe pressure data of the hydropower station units, which is beneficial for obtaining multi-dimensional operational characteristics related to the upper reservoir water level changes. Constructing an upper reservoir water level prediction model including an encoder and decoder facilitates deep feature extraction and mapping transformation of the characteristic data. Encoder-based feature extraction processing helps extract temporal features and variation patterns from the characteristic data. Attention-weighted processing of the characteristic data by the encoder yields enhanced feature data, which strengthens the feature weights at key time points and highlights feature information that significantly impacts upper reservoir water level prediction. Decoder-based upper reservoir water level prediction processing maps the enhanced feature data to predicted upper reservoir water level values. Training the upper reservoir water level prediction model using the predicted values yields a target upper reservoir water level prediction model, which optimizes model parameters to improve prediction capabilities. Inputting current operational data into the target upper reservoir water level prediction model for prediction processing helps obtain accurate upper reservoir water level prediction results, thus improving the accuracy of upper reservoir water level prediction. Attached Figure Description
[0058] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0059] Figure 1 This is a flowchart illustrating a method for predicting upper reservoir water level based on guide vane opening, active power, and steel pipe pressure in one embodiment.
[0060] Figure 2 This is a flowchart illustrating the steps for determining enhanced feature data in one embodiment;
[0061] Figure 3 This is a structural block diagram of an upper reservoir water level prediction device based on guide vane opening degree, active power, and steel pipe pressure in one embodiment.
[0062] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0063] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0064] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0065] In one exemplary embodiment, such as Figure 1 As shown, a method for predicting upper reservoir water level based on guide vane opening, active power, and steel pipe pressure is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc.; the server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. In this embodiment, the method includes the following steps:
[0066] Step S101: Determine the characteristic data of the generating units based on the operating data of the hydropower station units; the operating data includes guide vane opening data, active power data, and steel pipe pressure data;
[0067] Step S102: Construct an upper reservoir water level prediction model based on the feature data; the upper reservoir water level prediction model includes an encoder and a decoder.
[0068] Step S103: The encoder of the upper reservoir water level prediction model performs feature extraction and attention weighting on the feature data to obtain enhanced feature data.
[0069] Step S104: The enhanced feature data is processed by the decoder of the upper reservoir water level prediction model to obtain the upper reservoir water level prediction value of the hydropower station.
[0070] Step S105: Use the predicted upper reservoir water level to train the upper reservoir water level prediction model to obtain the target upper reservoir water level prediction model.
[0071] Step S106: Input the current operating data of the hydropower station into the target upper reservoir water level prediction model for upper reservoir water level prediction processing to obtain the upper reservoir water level prediction result of the hydropower station.
[0072] Among them, the guide vane opening data refers to the real-time sampling data of the opening degree of the guide vanes of the turbine unit, which reflects the operating conditions of the unit.
[0073] Among them, active power data can refer to real-time sampling data of the active power output of the hydro-generator unit, which reflects the power generation load status of the unit.
[0074] Among them, the steel pipe pressure data can refer to the real-time sampling data of the water flow pressure in the water intake steel pipe connected to the unit, which reflects the energy state of the water flow.
[0075] The unit's operating data can refer to real-time high-frequency sampling data including guide vane opening data, active power data, and steel pipe pressure data. For example, the guide vane opening, active power, and steel pipe pressure of four units can be collected to form 12-dimensional variable input data.
[0076] Feature data can refer to data extracted from the unit's operating data for model input. This feature data undergoes preprocessing, including outlier removal, to ensure data stability. For example, data from a 300-second time window can be used as the model's input features.
[0077] Among them, the upper reservoir water level prediction model can refer to a deep learning prediction model based on the fusion of Temporal Convolutional Network (TCN) and Attention mechanism. This model includes two parts: an encoder and a decoder.
[0078] The encoder can refer to a multi-layered stacked temporal convolutional network structure, such as a 4-layer stacked temporal convolutional network with channels of 64, 128, 256, and 512 respectively. Gradients are stabilized through residual connections and normalization, and an attention mechanism is added at the end.
[0079] The decoder can refer to a structure that maps the high-dimensional features output by the encoder to the target output dimension through a multi-layer perceptron (MLP). For example, the high-dimensional features output by the encoder can be mapped to a dimension that is 3 times the batch size and 1 times the batch size, so as to achieve a second-level prediction of the water level in the next 3 seconds.
[0080] Feature extraction processing can refer to the process of extracting temporal features from feature data through the temporal convolutional network of the encoder. This process extracts temporal patterns through residual connections and normalization.
[0081] Among them, attention-weighted processing can refer to weighting the extracted features through an attention mechanism to strengthen the feature weights at key time points, such as strengthening the feature weights at water level change points.
[0082] Among them, enhanced feature data can refer to high-dimensional feature data obtained after feature extraction and attention weighting by the encoder. This data contains temporal features and weighted information of key time points.
[0083] Among them, the upper reservoir water level prediction processing can refer to the process of processing the enhanced feature data through a decoder to map the high-dimensional features into the upper reservoir water level prediction value.
[0084] The predicted upper reservoir water level can refer to the upper reservoir water level value at a future moment output by the model, such as the upper reservoir water level in the next 3 seconds, including the upper reservoir water level for each second.
[0085] Among them, model training processing can refer to the process of using continuous high-frequency sampling data as a training set and optimizing model parameters through backpropagation. For example, 27 days of continuous high-frequency sampling data can be used as a training set for training.
[0086] Among them, the target upper reservoir water level prediction model can refer to the upper reservoir water level prediction model (neural network model) that has been optimized by parameter optimization after model training.
[0087] The current operating data can refer to the guide vane opening data, active power data, and steel pipe pressure data of the hydropower station at the current moment. This data is fed into the sliding data buffer stack, for example, it can be fed into the 300-second sliding data buffer stack.
[0088] The upper reservoir water level prediction result can refer to the upper reservoir water level value at a future moment output by the target upper reservoir water level prediction model, such as the upper reservoir water level in the next 3 seconds, including the upper reservoir water level in every second.
[0089] Optionally, when the terminal determines the characteristic data of the generating units based on the operating data of the hydropower station's units, it collects guide vane opening data, active power data, and steel pipe pressure data from the four generating units of the hydropower station as operating data, forming 12-dimensional variable characteristic data. Simultaneously, it performs sliding window mean processing on the upper reservoir water level as the target variable to remove high-frequency noise and focus on trend changes, and performs outlier removal processing on the characteristic data to ensure data stability. When constructing the upper reservoir water level prediction model based on the characteristic data, a four-layer stacked temporal convolutional network structure is used in the encoder of the upper reservoir water level prediction model. The number of channels was set to 64, 128, 256, and 512 respectively. Gradients were stabilized through residual connections and normalization. An attention mechanism was added at the end of the encoder. In the decoder of the upper reservoir water level prediction model, a multilayer perceptron structure was used to map the high-dimensional features output by the encoder to a target output dimension multiplied by 3 and 1 (batch size multiplied by 1), achieving second-level prediction of the water level in the next 3 seconds. When performing feature extraction and attention-weighted processing on the feature data, temporal features were extracted from the feature data through the temporal convolutional network of the encoder. Temporal patterns were extracted through residual connections and normalization, and then attention was added. The intention mechanism strengthens the feature weights for key time points, resulting in enhanced feature data that includes temporal features and weighted information from key time points. When processing this enhanced feature data for upper reservoir water level prediction, the high-dimensional features of the enhanced feature data are mapped to predicted upper reservoir water levels using a multilayer perceptron in the decoder. When training the upper reservoir water level prediction model using these predicted values, a loss function coupled with first-order differences is used as the loss function. Continuous high-frequency sampled data is used as the training set, and the input size is the batch size multiplied by 300 multiplied by 12, where 300 represents 300 seconds. The window, 12, represents 12-dimensional feature data. The model parameters of the upper reservoir water level prediction model are optimized through backpropagation to obtain the target upper reservoir water level prediction model. When the current operating data is input into the target upper reservoir water level prediction model for upper reservoir water level prediction processing, the current operating data of the hydropower station is collected through remote input / output modules or network communication. The current operating data is then fed into a 300-second sliding data buffer stack. Each refresh triggers preprocessing and inference. The preprocessed current operating data is then input into the target upper reservoir water level prediction model for upper reservoir water level prediction processing to obtain the upper reservoir water level prediction result of the hydropower station.
[0090] In the above-mentioned method for predicting the upper reservoir water level based on guide vane opening, active power, and steel pipe pressure, the characteristic data of the generating units are determined according to the operating data of the hydropower station units. The operating data includes guide vane opening data, active power data, and steel pipe pressure data. Based on the characteristic data, an upper reservoir water level prediction model is constructed. The upper reservoir water level prediction model includes an encoder and a decoder. Through the encoder of the upper reservoir water level prediction model, feature extraction and attention weighting processing are performed on the characteristic data to obtain enhanced feature data. Through the decoder of the upper reservoir water level prediction model, upper reservoir water level prediction processing is performed on the enhanced feature data to obtain the upper reservoir water level prediction value of the hydropower station. The upper reservoir water level prediction model is trained using the upper reservoir water level prediction value to obtain the target upper reservoir water level prediction model. The current operating data of the hydropower station is input into the target upper reservoir water level prediction model for upper reservoir water level prediction processing to obtain the upper reservoir water level prediction result of the hydropower station. This scheme determines characteristic data based on the guide vane opening, active power, and steel pipe pressure data of the hydropower station units, which is beneficial for obtaining multi-dimensional operational characteristics related to the upper reservoir water level changes. Constructing an upper reservoir water level prediction model including an encoder and decoder facilitates deep feature extraction and mapping transformation of the characteristic data. Encoder-based feature extraction processing helps extract temporal features and variation patterns from the characteristic data. Attention-weighted processing of the characteristic data by the encoder yields enhanced feature data, which strengthens the feature weights at key time points and highlights feature information that significantly impacts upper reservoir water level prediction. Decoder-based upper reservoir water level prediction processing maps the enhanced feature data to predicted upper reservoir water level values. Training the upper reservoir water level prediction model using the predicted values yields a target upper reservoir water level prediction model, which optimizes model parameters to improve prediction capabilities. Inputting current operational data into the target upper reservoir water level prediction model for prediction processing helps obtain accurate upper reservoir water level prediction results, thus improving the accuracy of upper reservoir water level prediction.
[0091] In one exemplary embodiment, reference is made to Figure 2 The encoder of the upper reservoir water level prediction model performs feature extraction and attention weighting on the feature data to obtain enhanced feature data, including:
[0092] Step S201: The encoder of the upper reservoir water level prediction model is used to extract features from the feature data to obtain time-series feature data.
[0093] Step S202: Attention weighting is applied to the time-series feature data to obtain enhanced feature data.
[0094] Among them, time-series feature data can refer to feature data containing time-series patterns obtained after feature extraction processing of feature data by an encoder. This time-series feature data extracts the time-series change patterns and time dependencies in the feature data.
[0095] Optionally, when the terminal performs feature extraction processing on the feature data through the encoder of the upper reservoir water level prediction model, the feature data is input into the temporal convolutional network of the encoder for feature extraction processing. The temporal pattern of the feature data is extracted through the residual connection and normalization processing of the temporal convolutional network to obtain temporal feature data. When performing attention weighting processing on the temporal feature data, the temporal feature data is input into the attention mechanism for attention weighting processing. The feature weight of key time points in the temporal feature data is strengthened through the attention mechanism to obtain enhanced feature data.
[0096] For example, when the terminal performs feature extraction processing on the feature data through the encoder of the upper reservoir water level prediction model, the feature data is first input into the encoder's four-layer stacked temporal convolutional network for feature extraction. The first temporal convolutional network performs the first feature extraction, the second temporal convolutional network performs the second feature extraction, the third temporal convolutional network performs the third feature extraction, and the fourth temporal convolutional network performs the fourth feature extraction. In each temporal convolutional network, the temporal pattern of the feature data is extracted through residual connections and normalization processing to obtain temporal feature data. When performing attention weighting processing on the temporal feature data, the temporal feature data is input into the attention mechanism for attention weighting processing. The attention mechanism calculates the attention weight value of each time point in the temporal feature data, and the features of each time point in the temporal feature data are weighted according to the attention weight value to strengthen the feature weight of key time points in the temporal feature data. Key time points may include abrupt changes in the upper reservoir water level, resulting in enhanced feature data containing weighted features.
[0097] The technical solution provided in this embodiment obtains time-series feature data by performing feature extraction processing on the feature data, which is beneficial for extracting the time-series patterns and time dependencies in the feature data; and obtains enhanced feature data by performing attention-weighted processing on the time-series feature data, which is beneficial for strengthening the feature weights of key time points and highlighting the time point features that have an important impact on the prediction of the upper reservoir water level, thereby improving the accuracy of the prediction of the upper reservoir water level.
[0098] In an exemplary embodiment, attention-weighted processing is performed on temporal feature data to obtain enhanced feature data, including: determining attention weights based on the temporal feature data; and performing attention-weighted processing on target time point feature data in the temporal feature data according to the attention weights to obtain enhanced feature data.
[0099] Among them, attention weight can refer to the weight value calculated based on time series feature data through the attention mechanism to characterize the importance of each time point. This attention weight is used to differentiate the weights of features at different time points in the time series feature data.
[0100] Among them, the target time point feature data can refer to the feature data of important time points in the time series feature data. The target time point feature data includes the feature data of key time points, which may include the abrupt change in the water level of the upper reservoir.
[0101] Optionally, when the terminal determines the attention weight based on the temporal feature data, it inputs the temporal feature data into the attention mechanism and calculates the attention weight of each time point in the temporal feature data through the attention mechanism; when performing attention weighting processing on the target time point feature data in the temporal feature data according to the attention weight, it determines the target time point feature data in the temporal feature data based on the attention weight, and performs weighting processing on the target time point feature data according to the attention weight to obtain the enhanced feature data.
[0102] For example, when a terminal determines attention weights based on temporal feature data, the temporal feature data is input into the attention mechanism. The query matrix, key matrix, and value matrix of the attention mechanism are used to transform the temporal feature data. The correlation between the query matrix and the key matrix is calculated to obtain an attention score. The attention score is then normalized to obtain the attention weight of each time point in the temporal feature data. The larger the value of the attention weight, the higher the importance of the corresponding time point. When performing attention weighting on the target time point feature data in the temporal feature data according to the attention weight, the time point in the temporal feature data whose attention weight is greater than a preset weight threshold is determined as the target time point. The temporal feature data corresponding to the target time point is obtained as the target time point feature data. According to the attention weight, the target time point feature data is weighted and calculated with the corresponding attention weight to enhance the expressive power of the target time point feature data in the overall features, resulting in enhanced feature data containing the weighted target time point feature data.
[0103] The technical solution provided in this embodiment determines the attention weight based on time-series feature data, which helps to quantify the importance of each time point in the time-series feature data. By performing attention-weighted processing on the target time point feature data in the time-series feature data according to the attention weight, enhanced feature data is obtained, which helps to strengthen the expressive ability of the target time point feature data in the overall features, highlight the target time point feature data that has an important impact on the prediction of the upper reservoir water level, thereby helping to improve the accuracy of the upper reservoir water level prediction.
[0104] In an exemplary embodiment, the enhanced feature data is processed by the decoder of the upper reservoir water level prediction model to obtain the upper reservoir water level prediction value of the hydropower station. This includes: performing feature mapping processing on the enhanced feature data by the decoder of the upper reservoir water level prediction model to obtain the feature-mapped enhanced feature data; and performing upper reservoir water level prediction processing on the feature-mapped enhanced feature data to obtain the upper reservoir water level prediction value.
[0105] Feature mapping processing can refer to the process of transforming and mapping the dimensions of the enhanced feature data through a decoder. This feature mapping process gradually maps and reduces the dimensionality of the high-dimensional features of the enhanced feature data.
[0106] Among them, the enhanced feature data after feature mapping can refer to the dimensionality-reduced feature data obtained after feature mapping processing by the decoder. The dimension of the enhanced feature data after feature mapping matches the target output dimension of the upper reservoir water level prediction value.
[0107] Optionally, when the terminal performs feature mapping processing on the enhanced feature data through the decoder of the upper reservoir water level prediction model, it inputs the enhanced feature data into the multilayer perceptron of the decoder for feature mapping processing. The multilayer perceptron performs dimensional transformation and mapping transformation on the enhanced feature data, mapping the high-dimensional features of the enhanced feature data to feature data of the target output dimension, thus obtaining the enhanced feature data after feature mapping. When performing upper reservoir water level prediction processing on the enhanced feature data after feature mapping, the upper reservoir water level prediction value is generated based on the enhanced feature data after feature mapping.
[0108] For example, when the terminal performs feature mapping processing on the enhanced feature data through the decoder of the upper reservoir water level prediction model, the enhanced feature data is first input into the multilayer perceptron of the decoder for feature mapping processing. The first layer of the multilayer perceptron performs the first dimensional transformation on the enhanced feature data, and the second layer of the multilayer perceptron performs the second dimensional transformation on the data after the first dimensional transformation. Through each layer of the multilayer perceptron, the high-dimensional feature mapping of the enhanced feature data is gradually reduced in dimensionality, and the high-dimensional feature mapping of the enhanced feature data is transformed into feature data with a target output dimension that is 3 times the batch size and 1 times the batch size. Here, 3 represents the next 3 seconds, and 1 represents the upper reservoir water level prediction value for each second. When performing upper reservoir water level prediction processing on the feature-mapped enhanced feature data, the upper reservoir water level prediction value is generated based on the target output dimension of the feature-mapped enhanced feature data. The upper reservoir water level prediction value includes the upper reservoir water level for the next 3 seconds, with the upper reservoir water level for each second as an independent prediction value.
[0109] The technical solution provided in this embodiment, by performing feature mapping processing on the enhanced feature data, obtains enhanced feature data after feature mapping. This is beneficial for mapping the high-dimensional features of the enhanced feature data into feature data that matches the target output dimension of the upper reservoir water level prediction value. By performing upper reservoir water level prediction processing on the enhanced feature data after feature mapping, the upper reservoir water level prediction value is obtained. This is beneficial for generating accurate upper reservoir water level prediction values based on the mapped feature data, thereby improving the accuracy of upper reservoir water level prediction.
[0110] In an exemplary embodiment, before training the upper reservoir water level prediction model using the upper reservoir water level prediction value to obtain the target upper reservoir water level prediction model, the method further includes: acquiring historical reservoir water level data; performing sliding window averaging on the historical reservoir water level data to obtain the target upper reservoir water level data; and training the upper reservoir water level prediction model using the upper reservoir water level prediction value to obtain the target upper reservoir water level prediction model, which includes: training the upper reservoir water level prediction model based on the upper reservoir water level prediction value and the target upper reservoir water level data to obtain the target upper reservoir water level prediction model.
[0111] Among them, historical reservoir water level data can refer to the historical data of the upper reservoir water level collected by the hydropower station during historical periods. This historical reservoir water level data is used as the target variable for model training.
[0112] Among them, the sliding window mean processing can refer to the method of using a sliding window to perform moving average processing on historical reservoir water level data. This sliding window mean processing can remove high-frequency noise from historical reservoir water level data and focus on the trend changes of historical reservoir water level data.
[0113] The target upper reservoir water level data can refer to the smoothed upper reservoir water level data obtained after sliding window averaging, and this target upper reservoir water level data serves as the target ground truth data for model training.
[0114] Optionally, before training the upper reservoir water level prediction model using the predicted upper reservoir water level, the terminal obtains historical reservoir water level data of the hydropower station, which serves as the target variable for model training. When performing moving window averaging on the historical reservoir water level data, a moving average is applied to remove high-frequency noise from the historical data and focus on the trend changes, thus obtaining the target upper reservoir water level data. When training the upper reservoir water level prediction model based on the predicted upper reservoir water level and the target upper reservoir water level data, the loss value between the predicted upper reservoir water level and the target upper reservoir water level is calculated. The model parameters of the upper reservoir water level prediction model are optimized based on the loss value using a backpropagation algorithm to obtain the target upper reservoir water level prediction model.
[0115] For example, before using the predicted upper reservoir water level to train the upper reservoir water level prediction model, the terminal acquires historical reservoir water level data collected by the hydropower station within a historical period. This historical water level data serves as the target variable for model training, and the historical period can include multiple consecutive days. When performing a moving window averaging on the historical water level data, the window length is set, which can be set to multiple sampling points. The moving average of the historical water level data is calculated within the moving window, removing high-frequency noise and focusing on the historical water level data. The trend changes in water level data are used to obtain smoothed target upper reservoir water level data. When training the upper reservoir water level prediction model based on the predicted upper reservoir water level and the target upper reservoir water level data, the predicted upper reservoir water level and the target upper reservoir water level data are input into the loss function. The loss function calculates the loss value between the predicted upper reservoir water level and the target upper reservoir water level data. The loss function can be a loss function coupled with first-order difference. Through the backpropagation algorithm, the gradient of each layer in the upper reservoir water level prediction model is calculated based on the loss value. The model parameters of the upper reservoir water level prediction model are optimized based on the gradients. After multiple rounds of iterative training, the target upper reservoir water level prediction model with optimized model parameters is obtained.
[0116] The technical solution provided in this embodiment obtains target upper reservoir water level data by performing sliding window averaging on historical reservoir water level data. This helps to remove high-frequency noise from historical reservoir water level data and focus on the trend changes of historical reservoir water level data, thus obtaining stable and reliable training target data. By training the upper reservoir water level prediction model based on the predicted upper reservoir water level value and the target upper reservoir water level data, it is beneficial to optimize the model parameters based on the difference between the predicted upper reservoir water level value and the target upper reservoir water level data, improve the model's predictive ability, and thus improve the accuracy of upper reservoir water level prediction.
[0117] In an exemplary embodiment, determining the characteristic data of the generating units based on the operating data of the hydropower station units includes: performing outlier removal processing on the operating data to obtain outlier-removed operating data; and determining the characteristic data based on the outlier-removed operating data.
[0118] Outlier removal processing refers to the process of detecting and removing abnormal values in operational data. This outlier removal process ensures the stability of operational data by identifying and removing values that deviate from the normal range in the operational data.
[0119] The outlier-removed running data can refer to stable and reliable running data obtained after outlier removal processing, which does not contain outlier values that deviate from the normal range.
[0120] Optionally, when the terminal determines the characteristic data of the generating units based on the operating data of the hydropower station's generating units, it performs outlier removal processing on the operating data. The outlier removal processing identifies and removes abnormal values that deviate from the normal range in the operating data, resulting in outlier-removed operating data. When determining characteristic data based on the outlier-removed operating data, the terminal extracts the guide vane opening data, active power data, and steel pipe pressure data of the generating units based on the outlier-removed operating data, and uses the guide vane opening data, active power data, and steel pipe pressure data of the generating units as characteristic data.
[0121] For example, when the terminal determines the characteristic data of a hydropower station's generating units based on their operating data, it performs outlier removal processing on the operating data. An outlier removal method based on standard deviation is used to detect outliers, calculating the mean and standard deviation of the operating data to determine the outlier threshold range. This threshold range can be set as the mean plus or minus a certain number of standard deviations. Values in the operating data that exceed the outlier threshold range are identified as outliers and removed, resulting in outlier-removed operating data. When determining characteristic data based on this outlier-removed operating data, the terminal extracts guide vane opening data, active power data, and steel pipe pressure data from multiple generating units of the hydropower station. The hydropower station may contain four generating units, each containing guide vane opening data, active power data, and steel pipe pressure data. These four units' guide vane opening data, active power data, and steel pipe pressure data are used as characteristic data, which includes 12-dimensional variables.
[0122] The technical solution provided in this embodiment obtains outlier-removed operating data by performing outlier removal processing on the operating data. This helps to identify and remove abnormal values that deviate from the normal range in the operating data, ensuring the stability and reliability of the operating data. By determining feature data based on the outlier-removed operating data, it is beneficial to extract the unit's feature data based on stable and reliable operating data, improve the quality of the feature data, and thus improve the accuracy of upper reservoir water level prediction.
[0123] The following application example illustrates the upper reservoir water level prediction method based on guide vane opening, active power, and steel pipe pressure provided in this application. This application example demonstrates the application of this method to a terminal.
[0124] A deep learning prediction model based on the fusion of Temporal Convolutional Network and Attention mechanism is used to predict the input of the unit guide vane opening, unit active power and unit steel pipe pressure to obtain the upper reservoir water level in the next 3 seconds.
[0125] Key data selection and preprocessing:
[0126] (1) Data selection: The input features are determined to be the guide vane opening, active power and steel pipe pressure of the four units, a total of 12 variables, and the target variable is the upper reservoir water level.
[0127] (2) Preprocessing: The target variable, namely the upper reservoir water level, is processed by sliding window mean to remove high-frequency noise and focus on trend changes; at the same time, outlier removal is performed on the input features to ensure data stability.
[0128] Model structure and loss function design:
[0129] (1) Model structure construction:
[0130] The encoder employs a 4-layer stacked temporal convolutional network with channels of 64, 128, 256, and 512 respectively. Gradients are stabilized through residual connections and normalization. Finally, an attention mechanism is added to enhance the feature weights at key time points, which may include water level mutation points.
[0131] Decoder: By using a multilayer perceptron to map the high-dimensional features output by the encoder to the target output dimension, it achieves second-level prediction of the water level in the next 3 seconds.
[0132] (2) Loss function design: The Huber loss coupled with first-order difference is adopted (a robust regression loss function that combines the advantages of mean square error and mean absolute error). The Huber loss ensures robustness to large errors and accuracy to small errors, while the first-order difference loss constrains the trend of the predicted sequence and avoids physically unreasonable violent oscillations.
[0133] Model training and deployment:
[0134] (1) Training process: Use continuous high-frequency sampled data as the training set and optimize the model parameters through backpropagation.
[0135] (2) Deployment plan:
[0136] Data access: Real-time data is collected via remote I / O modules or network communication, based on industrial protocols, and fed into a sliding data buffer stack. Each refresh triggers preprocessing and inference.
[0137] Inference Deployment: Supports deployment on servers or edge devices. Servers may include CPUs (Central Processing Units) or GPUs (Graphics Processing Units) to ensure that inference time meets real-time requirements.
[0138] Model validation and optimization:
[0139] (1) Validation method: Use 3 days of unseen data as the validation set to evaluate the model’s ability to predict water level trends, with a focus on validating the response performance at abrupt change points.
[0140] (2) Optimization mechanism: The prediction results are post-processed by the outlier removal mechanism based on standard deviation to avoid abnormal output caused by sudden fluctuations; the model parameters are iteratively adjusted in combination with the verification results to improve robustness.
[0141] Model inference:
[0142] After the model training is completed, the guide vane opening, active power and steel pipe pressure of the unit are input into the trained model, which is a deep learning prediction model based on the fusion of temporal convolutional network and attention mechanism. The model performs feature extraction, attention mechanism processing and decoding processing on the input data to predict the upper reservoir water level in the next 3 seconds, including the upper reservoir water level in every second.
[0143] The technical solution provided in this application example determines feature data based on the guide vane opening data, active power data, and steel pipe pressure data of the hydropower station units, which is beneficial for obtaining multi-dimensional operational features related to the upper reservoir water level changes. Constructing an upper reservoir water level prediction model including an encoder and decoder facilitates deep feature extraction and mapping transformation of the feature data. Feature extraction processing of the feature data through the encoder helps extract temporal features and variation patterns. Attention-weighted processing of the feature data through the encoder yields enhanced feature data, which strengthens the feature weights at key time points and highlights feature information that significantly impacts upper reservoir water level prediction. Upper reservoir water level prediction processing of the enhanced feature data through the decoder facilitates mapping the enhanced feature data to predicted upper reservoir water level values. Training the upper reservoir water level prediction model using the predicted upper reservoir water level values yields a target upper reservoir water level prediction model, which helps optimize model parameters to improve prediction capabilities. Inputting current operational data into the target upper reservoir water level prediction model for upper reservoir water level prediction processing helps obtain accurate upper reservoir water level prediction results, thereby improving the accuracy of upper reservoir water level prediction.
[0144] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0145] Based on the same inventive concept, this application also provides an upper reservoir water level prediction device based on guide vane opening degree, active power, and steel pipe pressure for implementing the above-mentioned upper reservoir water level prediction method based on guide vane opening degree, active power, and steel pipe pressure. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more upper reservoir water level prediction device embodiments based on guide vane opening degree, active power, and steel pipe pressure provided below can be found in the limitations of the upper reservoir water level prediction method based on guide vane opening degree, active power, and steel pipe pressure described above, and will not be repeated here.
[0146] In one exemplary embodiment, such as Figure 3 As shown, a reservoir water level prediction device based on guide vane opening degree, active power, and steel pipe pressure is provided. This reservoir water level prediction device 300 may include:
[0147] The data determination module 301 is used to determine the characteristic data of the generating units based on the operating data of the hydropower station units; the operating data includes guide vane opening data, active power data, and steel pipe pressure data;
[0148] The model building module 302 is used to build an upper reservoir water level prediction model based on feature data; the upper reservoir water level prediction model includes an encoder and a decoder.
[0149] The feature extraction module 303 is used to perform feature extraction and attention weighting on the feature data through the encoder of the upper reservoir water level prediction model to obtain enhanced feature data.
[0150] The water level prediction module 304 is used to perform upper reservoir water level prediction processing on the enhanced feature data through the decoder of the upper reservoir water level prediction model to obtain the upper reservoir water level prediction value of the hydropower station.
[0151] The model training module 305 is used to train the upper reservoir water level prediction model using the upper reservoir water level prediction value to obtain the target upper reservoir water level prediction model.
[0152] The data input module 306 is used to input the current operating data of the hydropower station into the target upper reservoir water level prediction model for upper reservoir water level prediction processing, and obtain the upper reservoir water level prediction result of the hydropower station.
[0153] In an exemplary embodiment, the feature extraction module 303 is further configured to perform feature extraction processing on the feature data through the encoder of the upper reservoir water level prediction model to obtain time-series feature data; and to perform attention weighting processing on the time-series feature data to obtain enhanced feature data.
[0154] In an exemplary embodiment, the feature extraction module 303 is further configured to determine attention weights based on temporal feature data; and to perform attention-weighted processing on the target time point feature data in the temporal feature data according to the attention weights to obtain enhanced feature data.
[0155] In an exemplary embodiment, the water level prediction module 304 is further configured to perform feature mapping processing on the enhanced feature data through the decoder of the upper reservoir water level prediction model to obtain the enhanced feature data after feature mapping; and to perform upper reservoir water level prediction processing on the enhanced feature data after feature mapping to obtain the upper reservoir water level prediction value.
[0156] In an exemplary embodiment, the device 300 further includes: a mean processing module, used to acquire historical reservoir water level data; perform sliding window mean processing on the historical reservoir water level data to obtain target upper reservoir water level data; and a model training module 305, used to perform model training processing on the upper reservoir water level prediction model based on the upper reservoir water level prediction value and the target upper reservoir water level data to obtain the target upper reservoir water level prediction model.
[0157] In an exemplary embodiment, the data determination module 301 is further configured to perform outlier removal processing on the running data to obtain outlier-removed running data; and determine feature data based on the outlier-removed running data.
[0158] The modules in the aforementioned upper reservoir water level prediction device based on guide vane opening, active power, and steel pipe pressure can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.
[0159] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a method for predicting the upper reservoir water level based on guide vane opening degree, active power, and steel pipe pressure. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.
[0160] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0161] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.
[0162] In one exemplary embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above-described method embodiments.
[0163] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0164] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0165] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are 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.
[0166] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for predicting the upper reservoir water level based on the guide vane opening degree, active power and steel pipe pressure, characterized in that, The method comprises: According to the operation data of the unit of the hydropower station, the characteristic data of the unit is determined; the operation data comprises guide vane opening data, active power data and steel pipe pressure data; According to the characteristic data, an upper reservoir water level prediction model is constructed; the upper reservoir water level prediction model comprises an encoder and a decoder; Through the encoder of the upper reservoir water level prediction model, the characteristic data is subjected to feature extraction processing and attention weighting processing to obtain enhanced characteristic data; Through the decoder of the upper reservoir water level prediction model, the enhanced characteristic data is subjected to upper reservoir water level prediction processing to obtain the upper reservoir water level prediction value of the hydropower station; The upper reservoir water level prediction model is subjected to model training processing by using the upper reservoir water level prediction value to obtain a target upper reservoir water level prediction model; The current operation data of the hydropower station is input into the target upper reservoir water level prediction model for upper reservoir water level prediction processing to obtain the upper reservoir water level prediction result of the hydropower station.
2. The method of claim 1, wherein, The method comprises: Through the encoder of the upper reservoir water level prediction model, the characteristic data is subjected to feature extraction processing and attention weighting processing to obtain enhanced characteristic data, comprising: Through the encoder of the upper reservoir water level prediction model, the characteristic data is subjected to feature extraction processing to obtain time sequence characteristic data; 3. The method of claim 2, wherein, The time sequence characteristic data is subjected to attention weighting processing to obtain the enhanced characteristic data. The method comprises: Based on the time sequence characteristic data, the attention weight is determined; 4. The method of claim 1, wherein, According to the attention weight, the target time point characteristic data in the time sequence characteristic data is subjected to attention weighting processing to obtain the enhanced characteristic data. The method comprises: Through the decoder of the upper reservoir water level prediction model, the enhanced characteristic data is subjected to feature mapping processing to obtain feature-mapped enhanced characteristic data; 5. The method of claim 1, wherein, The feature-mapped enhanced characteristic data is subjected to upper reservoir water level prediction processing to obtain the upper reservoir water level prediction value. Before the upper reservoir water level prediction model is subjected to model training processing by using the upper reservoir water level prediction value to obtain a target upper reservoir water level prediction model, the method further comprises: Obtaining historical upper reservoir water level data; The historical upper reservoir water level data is subjected to sliding window mean value processing to obtain target upper reservoir water level data; The method comprises:
6. The method according to any one of claims 1 to 5, characterized in that, According to the upper reservoir water level prediction value and the target upper reservoir water level data, the upper reservoir water level prediction model is subjected to model training processing to obtain the target upper reservoir water level prediction model. The method comprises: According to the operation data of the unit of the hydropower station, the characteristic data of the unit is determined, comprising:
7. A device for predicting the upper reservoir water level based on the guide vane opening, active power, and steel pipe pressure, characterized by, The operation data is subjected to outlier rejection processing to obtain outlier-rejected operation data; According to the outlier-rejected operation data, the characteristic data is determined. The device comprises: The data determination module is configured to determine characteristic data of a unit of the hydropower station according to operation data of the unit, wherein the operation data comprises wicket gate opening data, active power data and steel pipe pressure data. The model construction module is configured to construct an upper reservoir water level prediction model according to the characteristic data, wherein the upper reservoir water level prediction model comprises an encoder and a decoder. The feature extraction module is configured to perform feature extraction processing and attention weighting processing on the characteristic data through the encoder of the upper reservoir water level prediction model to obtain enhanced characteristic data. The water level prediction module is configured to perform upper reservoir water level prediction processing on the enhanced characteristic data through the decoder of the upper reservoir water level prediction model to obtain an upper reservoir water level prediction value of the hydropower station. The model training module is configured to perform model training processing on the upper reservoir water level prediction model by using the upper reservoir water level prediction value to obtain a target upper reservoir water level prediction model. The data input module is configured to input current operation data of the hydropower station into the target upper reservoir water level prediction model for upper reservoir water level prediction processing to obtain an upper reservoir water level prediction result of the hydropower station. 8.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-7. The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.
9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 6.