Hydropower station unit overcurrent prediction method based on LSTM deep learning

Abstract

The invention discloses a hydropower station set overcurrent prediction method based on LSTM deep learning, and the method achieves the accurate prediction of the total water diversion flow of a generator set through the original observation data of a reservoir hydropower station through the construction and training of an LSTM deep learning model for a time sequence variable regression problem. Comprising the steps of giving output and water level original observation data of a certain reservoir hydropower station in a long-time sequence, giving an actual measurement or simulation flow process of a confidence section of the corresponding reservoir hydropower station in the long-time sequence, carrying out balance analysis from the confidence section to the reservoir hydropower station, and constructing an LSTM deep learning network for predicting a single-feature long-time sequence through a multi-feature long-time sequence. Determining a network layer structure attribute and a connection mode, and carrying out LSTM deep learning model training and model verification; real-time and rapid decision support can be provided for dispatching operation personnel, dispatching operation of hydropower station units is further optimized, and power generation benefits of a hydropower station are improved while water volume calculation precision of a power station reservoir is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of hydropower generation, and in particular relates to a method for predicting overcurrent of a hydropower station unit based on LSTM deep learning. Background Art

[0002] For a long time, the LSTM model has been widely used in various disciplines and production practices in various industries for its excellent time series processing capabilities. At the same time, in the field of reservoir scheduling, especially in the operation and management of reservoirs and hydropower stations, measurement technology means and calculation methods have been developed to accurately obtain key data such as flow and water level. However, there are still errors in the measurement or calculation of the flow of hydropower station units, which brings challenges to the power generation scheduling of hydropower stations and even the optimization of reservoir operation. The Three Gorges and Gezhouba reservoirs occupy a pivotal position in China's hydropower construction and Yangtze River management. These two reservoirs play a key role in water resources management in the Yangtze River Basin, preventing and controlling floods, improving navigation conditions, protecting the coastal ecological environment, and promoting the coordinated development of the regional economy. The Three Gorges Reservoir is located about 40 kilometers upstream of the Gezhouba Reservoir. It only takes about 30 minutes for the Three Gorges outflow flood to evolve to the upstream of the Gezhouba Reservoir. Therefore, the flattening of the flood process is not obvious, and the Three Gorges outflow has a decisive influence on the inflow of the Gezhouba Reservoir. At the same time, the section from the Three Gorges to the Gezhouba Dam is relatively short, and there are no large tributaries flowing into it, resulting in a very small flow rate in the Three Gorges-Gezhouba Dam section. Therefore, in theory, the Three Gorges outflow and the Gezhouba inflow should be roughly equal, and the lag time should not exceed one hour. However, actual data show that there is an obvious imbalance in water volume between the Three Gorges and the Gezhouba Reservoir. A large number of examples and studies have shown that this phenomenon is mainly caused by inaccurate calculation of the water volume of the Gezhouba Dam. The Gezhouba Reservoir Hydropower Station underwent capacity expansion and transformation of the generator sets from 2006 to 2022. After the transformation, the efficiency of the units was improved, which was manifested in the same output and head conditions. The referenced power generation flow rate decreased, which resulted in the calculated value of the overall power generation referenced flow rate of Gezhouba being larger than the actual value, resulting in the outflow of Gezhouba being always greater than the outflow of the Three Gorges. The inflow of Gezhouba was obtained by inverse calculation based on the outflow and the measured water level changes. Therefore, the inflow of Gezhouba was systematically larger than the outflow of the Three Gorges for a long time. And this difference will fluctuate with the change of flow, and the relevant laws are difficult to observe.

[0003] At present, the Three Gorges Cascade Control Center, some universities and laboratories have conducted a lot of research on this issue. It is generally believed that due to its high design and construction standards and strict operation management, and no major transformation of the power station units has been carried out since the completion of the water storage, the outflow measurement of the Three Gorges Reservoir is relatively accurate, which can be used as a relatively accurate reference flow for the mainstream of the Yangtze River in the region; therefore, it is necessary to propose a hydropower station unit overcurrent prediction method based on LSTM deep learning to solve the above problem. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method for predicting overcurrent of hydropower station units based on LSTM deep learning, aiming to solve the problem that there are large errors in the measurement and calculation of overcurrent of hydropower station units in the prior art, and to use the LSTM deep learning model to accurately predict the overcurrent of hydropower stations.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: A method for predicting overcurrent of a hydropower station unit based on LSTM deep learning, comprising the following steps: S1, establish an observation data set of factors affecting the flow of hydropower stations, including the measured water level sequence upstream of the hydropower station, the measured water level sequence downstream of the hydropower station, the total output sequence of the hydropower station and the output sequence of each unit; S2, determine any confidence section upstream and downstream of the reservoir and hydropower station, obtain or calculate the confidence section flow, and conduct the evolution and water balance analysis of the confidence section flow to the upstream and downstream of the reservoir; S3, establishing a hydropower station flow data set; S4, perform dimension segmentation and feature segmentation of the influencing factor observation data set and the flow data set, and divide the training set and the validation set; S5, build a deep learning model based on LSTM, and set the attributes of the input layer, LSTM layer, Drop layer, fully connected layer and regression layer according to the observation data set and the flow data set; S6, conduct LSTM deep learning model training; S7, adjust the number of LSTM hidden units and the neuron drop rate parameters according to the training results, and return to S6 until the training results meet the requirements; S8, perform LSTM deep learning model verification and unit overcurrent prediction.

[0006] Preferably, in the data set in step S1, the output sequence of each unit is accurate to the individual output of each unit, and the number of units determines the number of features of the observation data set; the number of units is assumed to be n.

[0007] Preferably, the confidence sections described in step S2 only include river sections whose flow or water level observation data are considered to be accurate and whose errors are less than a threshold.

[0008] Preferably, in step S2, the evolution and water balance analysis is divided into two categories: One type is that the confidence section is located downstream of the hydropower station under study, and the other type is that the confidence section is located upstream of the hydropower station under study. For the first type of problem, the flow process of the confidence section is inverted to the reservoir outflow control section; for the second type of problem, the flow process of the confidence section is evolved to the reservoir inflow control section, and the reservoir outflow is calculated based on the actual upstream water level process measured in the corresponding period of the reservoir using the water level storage capacity curve and the water balance principle.

[0009] Preferably, in step S2, the evolution method adopts Muskingum method, and the Muskingum model is based on the following assumptions: The velocity of the river current is uniform, the channel shape is constant, and the lateral flux in the channel is negligible; Based on the water balance equation: ; (1) Among them, In, Qut, and Wr are the inflow, outflow, and trough storage of the river section, respectively. The following table shows the initial and final states of the time period; River channel storage equation: ; (2) Among them, Q' is the indicated storage flow, K represents the slope of the storage-flow relationship curve, and x is the flow specific gravity coefficient; The Muskingum confluence equation is calculated by formula (1) and formula (2): ; (3) ; (4) ; (5) ; (6) in .

[0010] Preferably, the dimension segmentation method described in step S4 is: The observed data set is divided into c samples with equal time step a and corresponding to the same number of features b in each time step, that is, the input data shape dimension is determined to be [a, b, c].

[0011] Preferably, the training set and validation set described in step S4 are divided in a ratio of 70% training and 30% validation. Both the training set and the validation set contain data of the reservoir and hydropower station under different operating conditions of high water level and low water level, and do not include the situation when the reservoir is abandoned; if there is water abandonment, the reservoir gate discharge flow data is obtained and eliminated from the flow data set.

[0012] Preferably, in the LSTM deep learning model described in step S5, the principle and structure of the LSTM long short-term memory network layer are as follows: LSTM network introduces a new cell state Specializes in linear cyclic information transmission and nonlinearly outputs information to the external state of the hidden layer At the same time, the "gate" structure can be used to remove or add "cell state" information to retain important content and remove unimportant content. The three "gates" are input gate , Forget Gate and output gate ; The calculation process is: First, use the external state of the previous moment and the current input , calculate three gates and candidate states ; Combined with forget gate and input gate To update the memory unit ; Combined output gate , passing the information of the internal state to the external state ; The specific calculation process is shown in formulas (8) to (15): For the input variable set , first use the nonlinear activation function tanh to transform its input: ; (7) Given the input vector, the calculation is as follows: ； (8) In the formula, Represents the weight matrix of the input variables, representing the weight of different input features on the overall prediction. The bias vector of the input variable indicates the additional degrees of freedom that need to be applied to the input variable after the weight is assigned by the weight matrix. It is a model hyperparameter together with the weight matrix. Forget Gate Control the internal state of the previous moment The amount of information that needs to be forgotten, yes The combined variables input before time, also called hidden layer states, yes Type activation function, which is calculated as follows: ； (9) In the formula, represents the weight matrix of the input variable of the forget gate, The weight matrix representing the hidden state of the forget gate, represents the forget gate bias vector; Input Gate Control the candidate status at the current moment The amount of information that needs to be saved is calculated as follows: ； (10) In the formula, represents the weight matrix of the input gate input variables, represents the input gate hidden state weight matrix, represents the input gate bias vector; Cell state First, calculate the candidate value of the cell state , as follows: ； (11) ; (12) In the formula, represents the cell state input variable weight matrix, represents the cell state hidden state weight matrix, represents the cell state deviation vector; Output Gate Control the internal state of the current moment Need to output to external state The amount of information is calculated as follows: ； (13) In the formula, represents the output gate input variable weight matrix, represents the output gate hidden state weight matrix, represents the output gate bias vector; current The hidden layer state at time The calculation of is as follows: ； (14) In the formula, represents the S-type activation function; The final output is: ； (15) In the formula, represents the output variable weight matrix, The bias vector representing the final output.

[0013] Preferably, the training optimization process described in step S7 measures the model training effect by detecting the model MSE, RMSE, MAE and loss Loss, adjusts the hyperparameters of the learning rate, batch size, number of LSTM hidden units and neuron drop probability, and gradually optimizes the training effect.

[0014] Preferably, the verification and prediction process described in step S8 is performed by loading the trained LSTM deep learning model, inputting data that conforms to the shape dimension size, the input including the observed or predicted water level and output in a certain period of time, and outputting the predicted total water diversion and power generation flow of the hydropower station in the corresponding period of time.

[0015] The beneficial effects of the present invention are as follows: 1. The present invention can accurately and quickly predict the corresponding unit overcurrent according to the water level, output measured or forecast data of the hydropower station when there are confidence sections and flow processes upstream and downstream of the hydropower station. It can not only be used for real-time data calculation of hydropower station operation, but also for flow simulation in different situations in scientific research and production practice. Using data from different periods for deep learning model training can form a specialized model suitable for different water inflow periods, and more targetedly predict the power generation reference flow under specific working conditions such as flood season and water storage period. At the same time, after the training of this LSTM deep learning model is completed, the memory occupancy is small and the prediction function call is convenient and fast. It can not only provide data support for scientific research, forecasting and scheduling and other businesses, but also provide real-time and fast decision support for scheduling and operation personnel, further optimize the scheduling and operation of hydropower station units, and improve the calculation accuracy of the water volume of the power station reservoir while improving the power generation efficiency of the hydropower station.

[0016] 2. The method for predicting the overcurrent of a hydropower station unit provided by the present invention can reduce the errors in the calculation of the water volume of a reservoir hydropower station to a certain extent. Some of these errors are due to systematic deviations caused by maintenance of hydropower station unit equipment, changes in operating conditions, etc., and some are due to the error transmission caused by insufficient representativeness of observation data such as water level or observation errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the flow chart of the present invention; Figure 2 A schematic diagram of a deep learning network architecture in an embodiment of the present invention; Figure 3 This is a generalized diagram of the LSTM model structure in an embodiment of the present invention; Figure 4 Schematic diagram of the prediction effect based on the LSTM model in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] Embodiment 1: like Figure 1 As shown, a method for predicting overcurrent of a hydropower station unit based on LSTM deep learning includes the following steps: S1, establish an observation data set of factors affecting the flow of hydropower stations, including the measured water level sequence upstream of the hydropower station, the measured water level sequence downstream of the hydropower station, the total output sequence of the hydropower station and the output sequence of each unit; S2, determine any confidence section upstream and downstream of the reservoir and hydropower station, obtain or calculate the confidence section flow, and conduct the evolution and water balance analysis of the confidence section flow to the upstream and downstream of the reservoir; S3, establishing a hydropower station flow data set; S4, perform dimension segmentation and feature segmentation of the influencing factor observation data set and the flow data set, and divide the training set and the validation set; S5, build a deep learning model based on LSTM, and set the attributes of the input layer, LSTM layer, Drop layer, fully connected layer and regression layer according to the observation data set and the flow data set; S6, conduct LSTM deep learning model training; S7, adjust the number of LSTM hidden units and the neuron drop rate parameters according to the training results, and return to S6 until the training results meet the requirements; S8, perform LSTM deep learning model verification and unit overcurrent prediction.

[0019] Preferably, in the data set in step S1, the output sequence of each unit is accurate to the output of each unit alone, and the number of units determines the number of features of the observation data set; the number of units is assumed to be n, and taking the Gezhouba Power Station as an example, there are 21 generator sets and 1 power supply unit, then n=22.

[0020] Preferably, the confidence sections described in step S2 only include river sections whose flow or water level observation data are considered to be accurate and whose errors are less than a threshold value; for example, a section a located upstream of the reservoir under study, whose flow measurement process adopts advanced technical means, and whose flow process is considered to be accurate and reliable after evaluation, then the measured flow process of this section is the confidence section flow; or a large-scale water conservancy project is built upstream of the reservoir under study, and its downstream flow is precisely controlled, then the outflow flow of the upstream reservoir and the outflow representative station are the confidence section flow and the confidence section.

[0021] Preferably, in step S2, the evolution and water balance analysis is divided into two categories: One type is that the confidence section is located downstream of the hydropower station under study, and the other type is that the confidence section is located upstream of the hydropower station under study. For the first type of problem, the flow process of the confidence section is inverted to the reservoir outflow control section; for the second type of problem, the flow process of the confidence section is evolved to the reservoir inflow control section, and the reservoir outflow is calculated based on the actual upstream water level process measured in the corresponding period of the reservoir using the water level storage capacity curve and the water balance principle.

[0022] Preferably, in step S2, the evolution method adopts the Muskingum method. Other commonly used software or model methods may also be used under the condition of having other parameters and computing capabilities. Taking the Muskingum method as an example: The Muskingum model is based on the following assumptions: The velocity of the river current is uniform, the channel shape is constant, and the lateral flux in the channel is negligible; Based on the water balance equation: ; (1) Among them, In, Qut, and Wr are the inflow, outflow, and trough storage of the river section, respectively. The following table shows the initial and final states of the time period; River channel storage equation: ; (2) Among them, Q' is the indicated storage flow, K represents the slope of the storage-flow relationship curve, and x is the flow specific gravity coefficient; The Muskingum confluence equation is calculated by formula (1) and formula (2): ; (3) ; (4) ; (5) ; (6) in .

[0023] Preferably, the dimension segmentation method described in step S4 is: The observation data set is divided into c samples with the same time step a and the same number of features b in each time step, that is, the input data shape dimension is determined to be [a, b, c]; taking Gezhouba as an example, assuming that the current observation data set is 1000 time steps, the corresponding features in each step include upstream water level, downstream water level, total power station output, and output of each unit, then the number of features is 3+n=25. The data set is divided into 10 samples, then the time step of each sample is 100, and the input data dimension format is [100, 25, 10].

[0024] Preferably, the training set and validation set described in step S4 are divided in a ratio of 70% training and 30% validation. Both the training set and the validation set contain data of the reservoir and hydropower station under different operating conditions of high water level and low water level, and do not include the situation when the reservoir is abandoned; if there is water abandonment, the reservoir gate discharge flow data is obtained and eliminated from the flow data set.

[0025] Embodiment 2: like Figure 2 As shown, in the LSTM deep learning model described in step S5, the principle and structure of the LSTM long short-term memory network layer are as follows: LSTM network introduces a new cell state Specializes in linear cyclic information transmission and nonlinearly outputs information to the external state of the hidden layer At the same time, the "gate" structure can be used to remove or add "cell state" information to retain important content and remove unimportant content. The three "gates" are input gate , Forget Gate and output gate .

[0026] like Figure 3 As shown, the LSTM network calculation process is: First, use the external state of the previous moment and the current input , calculate three gates and candidate states ; Combined with forget gate and input gate To update the memory unit ; Combined output gate , passing the information of the internal state to the external state ; The specific calculation process is shown in formulas (8) to (15): For the input variable set , first use the nonlinear activation function tanh to transform its input: ; (7) Given the input vector, the calculation is as follows: ； (8) In the formula, Represents the weight matrix of the input variables, representing the weight of different input features on the overall prediction. The bias vector of the input variable indicates the additional degrees of freedom that need to be applied to the input variable after the weight is assigned by the weight matrix. It is a model hyperparameter together with the weight matrix. Forget Gate Control the internal state of the previous moment The amount of information that needs to be forgotten, yes The combined variables input before time, also called hidden layer states, yes Type activation function, which is calculated as follows: ； (9) In the formula, represents the weight matrix of the input variable of the forget gate, The weight matrix representing the hidden state of the forget gate, represents the forget gate bias vector; Input Gate Control the candidate status at the current moment The amount of information that needs to be saved is calculated as follows: ； (10) In the formula, represents the weight matrix of the input gate input variables, represents the input gate hidden state weight matrix, represents the input gate bias vector; Cell state First, calculate the candidate value of the cell state , as follows: ； (11) ; (12) In the formula, represents the cell state input variable weight matrix, represents the cell state hidden state weight matrix, represents the cell state deviation vector; Output Gate Control the internal state of the current moment Need to output to external state The amount of information is calculated as follows: ； (13) In the formula, represents the output gate input variable weight matrix, represents the output gate hidden state weight matrix, represents the output gate bias vector; current The hidden layer state at time The calculation of is as follows: ； (14) In the formula, represents the S-type activation function; The final output is: ； (15) In the formula, represents the output variable weight matrix, The bias vector representing the final output.

[0027] The specific symbols are explained in Table 1 below: Table 1: Explanation of symbols in the formula;

[0028] Preferably, the training optimization process described in step S7 measures the model training effect by detecting the model MSE, RMSE, MAE and loss Loss, adjusts the hyperparameters of the learning rate, batch size, number of LSTM hidden units and neuron drop probability, and gradually optimizes the training effect.

[0029] Preferably, the verification and prediction process described in step S8 is performed by loading the trained LSTM deep learning model, inputting data that conforms to the shape dimension size, the input including the observed or predicted water level and output in a certain period of time, and outputting the predicted total water diversion and power generation flow of the hydropower station in the corresponding period of time.

[0030] Embodiment three: Furthermore, the attributes of each layer of the LSTM deep learning model described in step S5 are set as shown in Table 1 below: Table 1: Deep learning network analysis property settings:

[0031] Taking Gezhouba as an example, the input feature dimension of the input layer is set to 25, the minimum length of minlength is set to 350, and no normalization is performed; the number of hidden units of the LSTM layer is set to 128, the output mode is set to sequence to output the hidden state of each time step, the StateActivationFunction state unit activation function is set to tanh to facilitate processing sequence data, the GateActivationFunction gate activation function is set to sigmoid, the InputWeightsLearnRateFactor input weight learning rate factor is set to 1 to use the global learning rate, and the InputWeightsL2Factor controls the input weight The L2 regularization factor is set to 1, the RecurrentWeightsLearnRateFactor controls the learning rate factor of the recursive weights and is set to 1, the BiasLearnRateFactor controls the learning rate factor of the bias and is set to 1, and the InputWeightsInitializer initialization weight strategy is set to glorot to help accelerate convergence; the DropoutProbability random neuron drop probability of the Drop layer is set to 0.3 to adapt to smaller network data sets; the number of NumUnits neurons in the FC fully connected layer is set to 1, because only one output feature, the total overcurrent, needs to be predicted, and the rest, such as the learning rate factor and regularization factor that control the weights and biases, are the same as the LSTM layer; the output dimension of the regression layer OutputSize is set to 1.

[0032] like Figure 4 As shown, taking the Gezhouba Power Station as an example, the prediction effect of the hydropower station overcurrent prediction method based on the LSTM deep learning provided by the present invention is demonstrated. It can be seen from the figure that the predicted value curve and the actual value curve have a high degree of overlap. Due to the capacity expansion and transformation of the Gezhouba Reservoir unit and possible changes in equipment operating conditions, the calculation results of the water diversion and power generation flow of the Gezhouba unit appear systematically larger in the flood season, and fluctuating errors appear in the non-flood season. The current solution to this problem in the Three Gorges cascade regulation is manual correction, and the water diversion and power generation flow is scaled by introducing empirical coefficients. This method provides another way to solve such problems. Based on big data and machine learning ideas, a black box model is established to consider all kinds of influencing factors that affect the accuracy of water volume calculation in hydropower stations, and the external side only pays attention to the input and output of the LSTM deep learning model, where the input is the most original unprocessed observation data, and the output is a reliable and accurate hydropower station water diversion and power generation flow.

Claims

1. A method for predicting overcurrent of hydropower station units based on LSTM deep learning, characterized in that: The following steps are involved: S1, establish an observation data set of factors affecting the flow of hydropower stations, including the measured water level sequence upstream of the hydropower station, the measured water level sequence downstream of the hydropower station, the total output sequence of the hydropower station and the output sequence of each unit; S2, determine any confidence section upstream and downstream of the reservoir and hydropower station, obtain or calculate the confidence section flow, and conduct the evolution and water balance analysis of the confidence section flow to the upstream and downstream of the reservoir; S3, establishing a hydropower station flow data set; S4, perform dimension segmentation and feature segmentation of the influencing factor observation data set and the flow data set, and divide the training set and the validation set; S5, build a deep learning model based on LSTM, and set the attributes of the input layer, LSTM layer, Drop layer, fully connected layer and regression layer according to the observation data set and the flow data set; S6, conduct LSTM deep learning model training; S7, adjust the number of LSTM hidden units and the neuron drop rate parameters according to the training results, and return to S6 until the training results meet the requirements; S8, perform LSTM deep learning model verification and unit overcurrent prediction.

2. According to a method for predicting overcurrent of a hydropower station unit based on LSTM deep learning according to claim 1, it is characterized in that: In the data set in step S1, the output sequence of each unit is accurate to the individual output of each unit. The number of units determines the number of features in the observation data set; the number of units is assumed to be n.

3. According to a method for predicting overcurrent of a hydropower station unit based on LSTM deep learning according to claim 1, it is characterized in that: The confidence sections described in step S2 only include river sections whose flow or water level observation data are considered to be accurate and whose errors are less than a threshold.

4. A method for predicting overcurrent of a hydropower station unit based on LSTM deep learning according to claim 1, characterized in that: In step S2, the evolution and water balance analysis is divided into two categories: One type is that the confidence section is located downstream of the hydropower station under study, and the other type is that the confidence section is located upstream of the hydropower station under study; For the first type of problem, the flow process of the confidence section is inverted to the reservoir outflow control section; For the second type of problem, the confidence section flow process is evolved to the reservoir inflow control section. Based on the measured upstream water level process of the reservoir in the corresponding period, the reservoir outflow is calculated using the water level storage capacity curve and the water balance principle.

5. A method for predicting overcurrent of a hydropower station unit based on LSTM deep learning according to claim 1, characterized in that: In step S2, the evolution method adopts the Muskingum method, and the Muskingum model is based on the following assumptions: The velocity of the river current is uniform, the channel shape is constant, and the lateral flux in the channel is negligible; Based on the water balance equation: ；（1） Among them, In, Qut, and Wr are the inflow, outflow, and trough storage of the river section, respectively. The following table shows the initial and final states of the time period; River channel storage equation: ；（2） Among them, Q' is the indicated storage flow, K represents the slope of the storage-flow relationship curve, and x is the flow specific gravity coefficient; The Muskingum confluence equation is calculated by formula (1) and formula (2): ；（3） ；（4） ；（5） ；（6） in .

6. A method for predicting overcurrent of a hydropower station unit based on LSTM deep learning according to claim 1, characterized in that: The dimension segmentation method described in step S4 is: The observed data set is divided into c samples with equal time step a and corresponding to the same number of features b in each time step, that is, the input data shape dimension is determined to be [a, b, c].

7. A method for predicting overcurrent of a hydropower station unit based on LSTM deep learning according to claim 1, characterized in that: The training set and validation set described in step S4 are divided into 70% training and 30% validation ratios. Both the training set and the validation set contain data of the reservoir and hydropower station under different operating conditions of high water level and low water level, and do not include the situation when the reservoir is abandoned. If there is water abandonment, the reservoir gate discharge flow data is obtained and eliminated from the flow data set.

8. A method for predicting overcurrent of a hydropower station unit based on LSTM deep learning according to claim 1, characterized in that: In the LSTM deep learning model described in step S5, the principle and structure of the LSTM long short-term memory network layer are as follows: LSTM network introduces a new cell state Specializes in linear cyclic information transmission and nonlinearly outputs information to the external state of the hidden layer At the same time, the "gate" structure can be used to remove or add "cell state" information to retain important content and remove unimportant content. The three "gates" are input gate , Forget Gate and output gate ; The calculation process is: First, use the external state of the previous moment and the current input , calculate three gates and candidate states ; Combined with forget gate and input gate To update the memory unit ; Combined output gate , passing the information of the internal state to the external state ; The specific calculation process is shown in formulas (8) to (15): For the input variable set , first use the nonlinear activation function tanh to transform its input: ；（7） Given the input vector, the calculation is as follows: ； （8） In the formula, Represents the weight matrix of the input variables, representing the weight of different input features on the overall prediction. The bias vector of the input variable indicates the additional degrees of freedom that need to be applied to the input variable after the weight is assigned by the weight matrix. It is a model hyperparameter together with the weight matrix. Forget Gate Control the internal state of the previous moment The amount of information that needs to be forgotten, yes The combined variables input before time, also called hidden layer states, yes Type activation function, which is calculated as follows: ； （9） In the formula, represents the weight matrix of the input variable of the forget gate, The weight matrix representing the hidden state of the forget gate, represents the forget gate bias vector; Input Gate Control the candidate status at the current moment The amount of information that needs to be saved is calculated as follows: ； （10） In the formula, represents the weight matrix of the input gate input variables, represents the input gate hidden state weight matrix, represents the input gate bias vector; Cell state First, calculate the candidate value of the cell state , as follows: ； （11） ； （12） In the formula, represents the cell state input variable weight matrix, represents the cell state hidden state weight matrix, represents the cell state deviation vector; Output Gate Control the internal state of the current moment Need to output to external state The amount of information is calculated as follows: ； （13） In the formula, represents the output gate input variable weight matrix, represents the output gate hidden state weight matrix, represents the output gate bias vector; current The hidden layer state at time The calculation of is as follows: ； （14） In the formula, represents the S-type activation function; The final output is: ； （15） In the formula, represents the output variable weight matrix, The bias vector representing the final output.

9. A method for predicting overcurrent of a hydropower station unit based on LSTM deep learning according to claim 1, characterized in that: The training optimization process described in step S7 measures the model training effect by detecting the model MSE, RMSE, MAE and loss Loss, adjusts the hyperparameters of the learning rate, batch size, number of LSTM hidden units and neuron drop probability, and gradually optimizes the training effect.

10. A method for predicting overcurrent of a hydropower station unit based on LSTM deep learning according to claim 1, characterized in that: The verification and prediction process described in step S8 loads the trained LSTM deep learning model, inputs data that conform to the shape dimension size, and the input includes the observed or predicted water level and output for a certain period of time, and outputs the predicted total water diversion and power generation flow of the hydropower station for the corresponding period of time.

Images