A method and system for operating a pumped storage unit

Abstract

The application discloses a kind of pumped storage unit operation control method and system, wherein the method includes screening new energy unit output data and future weather data, obtains new energy output prediction result according to the multi-source prediction dataset constructed by screening result;Determine each data subset of load data under different dimensions, input each data subset into corresponding prediction model, obtain medium and long term load prediction result;Load data and future weather data are input into short-term load prediction model to obtain short-term load prediction result;Determine optimal reservoir storage target curve based on new energy output prediction result and medium and long term load prediction result, and determine optimal unit output result based on new energy output prediction result and short-term load prediction result;The operation control strategy generated by optimal reservoir storage target curve and optimal unit output result is executed.The method of the application realizes the operation optimization control of pumped storage unit.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pumped storage technology, in particular to a pumped storage unit operation control method and system. BACKGROUND

[0002] With the global emphasis on environmental protection and sustainable development, the installed capacity and power generation of new energy such as solar energy and wind energy are increasing. However, new energy has the characteristics of uncertainty, which brings great challenges to the stable operation of the power system. On this basis, pumped storage units, as an important energy storage facility, have multiple functions such as peak shaving, frequency modulation, pressure regulation and backup in the power system.

[0003] However, the traditional pumped storage unit operation control method is often based on artificial experience or a single model, without considering the uncertainty and complexity of new energy output and load, resulting in low operation efficiency.

[0004] Therefore, how to reasonably control the operation of pumped storage units to adapt to the complex changing environment of the power system after the access of new energy has become a technical problem to be solved by those skilled in the art. SUMMARY

[0005] The present application provides a pumped storage unit operation control method and system to solve the technical problem of how to reasonably control the operation of pumped storage units.

[0006] In order to solve the above technical problems, the present application provides a pumped storage unit operation control method, comprising:

[0007] Respectively acquiring load data of a target region matched with a target pumped storage unit, new energy unit output data of the target region and future meteorological data of the target region;

[0008] Filtering the new energy unit output data and the future meteorological data based on the mutual information method, predicting the target new energy unit output data of the target region according to a multi-source prediction data set constructed from the filtering results, and obtaining a new energy output prediction result;

[0009] Determining each data subset of the load data in different dimensions, inputting each data subset into a prediction model of the corresponding dimension for prediction, and fusing each prediction result to obtain a medium and long term load prediction result of the target region, wherein the data subset at least includes seasonal dimension data, linear trend dimension data and random data;

[0010] Inputting the load data and the future meteorological data into a pre-constructed short-term load prediction model to obtain a short-term load prediction result of the target region;

[0011] determine an optimal reservoir storage target curve of the target pumped storage unit based on the new energy output prediction result and the medium and long term load prediction result, and determine an optimal unit output result of the target pumped storage unit based on the new energy output prediction result and the short term load prediction result;

[0012] execute a pumped storage unit operation control strategy generated based on the optimal reservoir storage target curve and the optimal unit output result.

[0013] As one of the preferred schemes, the mutual information method is used to filter the new energy unit output data and the future weather data, and a multi-source prediction data set is constructed based on the filtering results, including:

[0014] The new energy unit output data is taken as a target variable, and the future weather data is taken as a feature variable to construct a feature matrix;

[0015] The mutual information value between different weather data and the new energy unit output data is calculated according to a mutual information calculation method to obtain a mutual information value array;

[0016] The mutual information value array is traversed, and weather data features with mutual information values greater than a preset mutual information threshold value are filtered out;

[0017] The filtered future weather data features and the new energy unit output data are spliced to obtain the multi-source prediction data.

[0018] As one of the preferred schemes, the target new energy unit output data of the target region is predicted to obtain a new energy output prediction result, including:

[0019] The pre-acquired historical multi-source prediction data is divided in time sequence to obtain a training data set, a validation data set and a test data set respectively;

[0020] An initial new energy output prediction model based on LSTM is constructed;

[0021] The training data set is input into the initial new energy output prediction model for training, in the training process, a particle swarm optimization algorithm and the validation data set are used to optimize the parameters of the initial new energy output prediction model, and the accuracy of the initial new energy output prediction model is evaluated according to the test data set to obtain a trained new energy output prediction model;

[0022] In the actual prediction process, the acquired real-time multi-source prediction data is input into the new energy output prediction model for prediction to obtain the new energy output prediction result.

[0023] As one of the preferred solutions, the determining of each data subset of the load data in different dimensions comprises:

[0024] The load data is converted into time series data with time as the index and load value as the data column;

[0025] The time series data is decomposed based on the STL decomposition method to obtain seasonal dimension data, trend dimension data and random data.

[0026] As one of the preferred solutions, the inputting of each data subset into the corresponding dimension prediction model for prediction and the fusion of each prediction result to obtain the medium and long-term load prediction result of the target area comprises:

[0027] The seasonal dimension data is predicted based on a seasonal autoregressive integrated moving average model to obtain a seasonal prediction result;

[0028] The trend dimension data is predicted according to a linear regression model to obtain a trend prediction result;

[0029] The random data is predicted based on an autoregressive integrated moving average model to obtain a random prediction result;

[0030] The seasonal prediction result, the trend prediction result and the random prediction result are fused to obtain a medium and long-term load prediction result.

[0031] As one of the preferred solutions, the inputting of the load data and the future meteorological data into a pre-constructed short-term load prediction model to obtain the short-term load prediction result of the target area comprises:

[0032] Feature extraction is performed on the load data and the future meteorological data to obtain a training data set;

[0033] An initial short-term load prediction model based on LSTM is constructed;

[0034] The training data set is input into the initial short-term load prediction model for training, and in the training process, the parameters of the initial short-term load prediction model are optimized through a back propagation algorithm to obtain a trained short-term load prediction model;

[0035] In the actual prediction process, the real-time multi-source prediction data and real-time load data after feature extraction are input into the initial short-term load prediction model for prediction to obtain the short-term load prediction result.

[0036] As one preferred embodiment, determining the optimal reservoir storage target curve for the target pumped storage unit based on the new energy output forecast results and the medium- and long-term load forecast results includes:

[0037] The predicted output of the new energy source and the predicted medium- and long-term load are input into a pre-constructed medium- and long-term optimization objective function for solution, resulting in the optimal reservoir storage target curve for the target pumped storage unit; wherein, the medium- and long-term optimization objective function is expressed as:

[0038]

[0039] in, For the first The predicted output of new energy sources over a given time period. For the first Medium- to long-term load forecasts for a given period. For pumped storage units in Power generation in the first time period, It is the first Pumping costs for a given time period It is the first Maintenance costs over a specific period of time. , , These are the corresponding weighting coefficients. It represents the total number of time periods in a medium- to long-term plan.

[0040] As one preferred embodiment, determining the optimal unit output of the target pumped storage unit based on the new energy output forecast results and the short-term load forecast results includes:

[0041] The new energy output forecast results and the short-term load forecast results are input into a pre-constructed short-term optimization objective function for solution, to obtain the optimal unit output result of the target pumped storage unit; wherein, the short-term optimization objective function is expressed as:

[0042]

[0043] The constraints of the short-run optimization objective function are expressed as follows:

[0044]

[0045] in, For the first Power generation of pumped storage units during a given time period For the first The pumping power of the pumped storage unit during a certain period of time This indicates that the pumped storage unit was in the first an output of a period, a new energy output prediction value of a period, a new energy output prediction value of a period, a short-term load prediction value of a period. a short-term load prediction value of a period.

[0046] As one of the preferred solutions, after executing the pumped storage unit operation control strategy generated based on the optimal reservoir storage target curve and the optimal unit output result, the method further comprises:

[0047] real-time acquisition of new energy unit output data, load data and future weather data of a target pumped storage unit affected area;

[0048] updating the new energy output prediction result, the medium and long term load prediction result and the load prediction result according to the real-time acquired data;

[0049] obtaining an optimization result according to the updated prediction result and the optimal reservoir storage target curve and the optimal unit output result;

[0050] dynamically adjusting the pumped storage unit control strategy according to the optimization result.

[0051] Another embodiment of the application provides a pumped storage unit operation control system, comprising:

[0052] an acquisition module for acquiring load data of a target area matched with a target pumped storage unit, new energy unit output data of the target area and future weather data of the target area respectively;

[0053] an output prediction module for filtering the new energy unit output data and the future weather data based on a mutual information method, predicting target new energy unit output data of the target area based on a multi-source prediction data set constructed from the filtering result, and obtaining a new energy output prediction result;

[0054] a first load prediction module for determining each data subset of the load data under different dimensions, inputting each data subset and the future weather data into a prediction model of a corresponding dimension for prediction, and fusing each prediction result to obtain a medium and long term load prediction result of the target area, wherein the data subset at least includes seasonal dimension data, linear trend dimension data and random data;

[0055] a second load prediction module for inputting the load data and the future weather data into a pre-constructed short-term load prediction model to obtain a short-term load prediction result of the target area;

[0056] an analysis module configured to determine an optimal reservoir storage target curve of the target pumped storage unit based on the new energy output prediction result and the medium and long term load prediction result, and determine an optimal unit output result of the target pumped storage unit based on the new energy output prediction result and the short term load prediction result;

[0057] an execution module configured to execute a pumped storage unit operation control strategy generated based on the optimal reservoir storage target curve and the optimal unit output result.

[0058] Compared with the prior art, the embodiments of the present application have at least one of the following advantages:

[0059] (1) The present application determines the optimal reservoir storage target curve based on the new energy output prediction result and the medium and long term load prediction result, can consider the change trend of new energy power generation and load demand from a long-term perspective, and reasonably plans the reservoir storage capacity. In the new energy large generation period, more water can be stored, and more water can be discharged for power generation when the load peak and new energy output are insufficient, so as to realize the optimal utilization of water resources, improve the energy storage efficiency and regulation capacity of pumped storage power stations, and ensure the stable power supply of the power system.

[0060] (2) The scheme of the present application enables the pumped storage unit to better match the new energy unit and the power load, enhances the ability of the power system to cope with new energy uncertainty and load changes, improves the flexibility and adaptability of the power system, helps to build a new type of power system mainly based on new energy, and promotes the green and low-carbon transformation of energy. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 is a flowchart of the operation control method of the pumped storage unit in one of the embodiments of the present application;

[0062] Figure 2 is a schematic diagram of the operation control system of the pumped storage unit in one of the embodiments of the present application. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments of the present application. The purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0064] In the description of the present application, the terms "first", "second", "third" and the like are used only for descriptive purposes and are not to be construed as indicating or implying relative importance or a specific order of the indicated technical features. Thus, features defined with "first", "second", "third" and the like can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specified.

[0065] In the description of the present application, it should be noted that, unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connecting" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be the communication inside two elements. The terms "vertical", "horizontal", "left", "right", "up", "down" and similar expressions used herein are for illustrative purposes only, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be construed as limiting the present application. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those skilled in the art, the specific meanings of the above terms in the present application can be understood in specific cases.

[0066] In the description of the present application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present application are the same as those commonly understood by those skilled in the art. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. For those skilled in the art, the specific meanings of the above terms in the present application can be understood in specific cases.

[0067] An embodiment of the present application provides a method for operating a pumped storage unit, specifically, please refer to Figure 1 , Figure 1 The flowchart shows the method for operating a pumped storage unit in one embodiment of the present application, which includes steps S1-S6:

[0068] S1: respectively acquiring load data of a target region, new energy unit output data of the target region and future weather data of the target region matched with a target pumped storage unit;

[0069] The load data can be directly obtained from real-time load data and historical load data of the target region through a data interface of the power system, such as an energy management system (EMS) or a distribution management system (DMS), which can provide load data information of different voltage levels and different regions, covering the power consumption of various users such as industrial, commercial and residential users.

[0070] For wind power, photovoltaic and other new energy power stations, each is equipped with its own monitoring system. For example, the supervisory control and data acquisition system (SCADA) of a wind power plant can monitor the operating status, wind speed, wind direction and power generation data of each wind turbine in real time; the monitoring system of a photovoltaic power station can monitor the power generation, light intensity and temperature of the photovoltaic panel. By establishing a communication connection with the monitoring system of these new energy power stations, real-time output data and historical output data of the new energy units can be directly obtained.

[0071] The future weather data can be obtained by establishing cooperation with local meteorological departments or professional meteorological service agencies through their data interface.

[0072] S2: filtering the new energy unit output data and the future weather data based on the mutual information method, predicting the target new energy unit output data of the target region based on a multi-source prediction data set constructed from the filtering results, and obtaining a new energy output prediction result;

[0073] The new energy unit output data reflects the actual power generation of wind power generation, photovoltaic power generation and other new energy equipment at a specific time point, and is the target object for subsequent prediction. The future weather data includes various meteorological elements such as temperature, humidity, wind speed, wind direction and light intensity, which have a direct or indirect impact on the output of the new energy unit.

[0074] Preferably, in an embodiment of the present application, the new energy unit output data and the future weather data are filtered based on the mutual information method, and a multi-source prediction data set is constructed from the filtering results, including:

[0075] The new energy unit output data is taken as a target variable, and the future weather data is taken as a feature variable to construct a feature matrix;

[0076] The mutual information value between different weather data and the new energy unit output data is calculated according to the mutual information calculation method to obtain a mutual information value array;

[0077] The mutual information value array is traversed, and the weather data features with a mutual information value greater than a preset mutual information threshold value are selected;

[0078] The selected future weather data features are spliced with the new energy unit output data to obtain a multi-source prediction data.

[0079] wherein the mutual information is a concept in information theory, used to measure the correlation between two random variables. The present application uses the mutual information calculation method to measure the degree of correlation between different meteorological data (feature variables) and new energy unit output data (target variables). For each kind of meteorological data, the mutual information value between it and the new energy unit output data is calculated. Assuming that there are K kinds of meteorological data, finally a mutual information value array with a length of K is obtained, and each element in the array represents the mutual information value between a kind of meteorological data and the new energy unit output data.

[0080] A mutual information threshold is preset, which is determined according to experience, experiment or data analysis, and is used to judge whether the correlation between the meteorological data and the new energy unit output data is strong enough. Traverse the mutual information value array, and check whether each mutual information value is greater than the preset mutual information threshold one by one. Screen out the meteorological data features whose mutual information values are greater than the threshold. These screened out meteorological data have strong correlation with the new energy unit output data.

[0081] The screened out future meteorological data features are spliced with the new energy unit output data. In an embodiment of the present application, if the wind speed and the light intensity are screened out as the meteorological data features, then the two kinds of meteorological data are arranged in time sequence corresponding to the new energy unit output data to form a new data set, i.e. multi-source prediction data.

[0082] Preferably, in an embodiment of the present application, the target new energy unit output data of the target region is predicted to obtain a new energy output prediction result, which includes:

[0083] The pre-acquired historical multi-source prediction data is divided in time sequence to obtain a training data set, a verification data set and a test data set respectively;

[0084] An initial new energy output prediction model based on LSTM is constructed;

[0085] The training data set is input into the initial new energy output prediction model for training. In the training process, the particle swarm optimization algorithm and the verification data set are used to optimize the parameters of the initial new energy output prediction model, and the accuracy of the initial new energy output prediction model is evaluated according to the test data set, so as to obtain a trained new energy output prediction model;

[0086] In the actual prediction process, the acquired real-time multi-source prediction data is input into the new energy output prediction model for prediction to obtain a new energy output prediction result.

[0087] A certain period of historical multi-source prediction data is obtained in advance, which is based on the above method and contains key meteorological data related to new energy unit output and corresponding output data. The historical multi-source prediction data is divided in chronological order.

[0088] An initial new energy output prediction model is constructed based on a long short-term memory network (LSTM). LSTM is a special recurrent neural network (RNN) that can effectively handle long-term dependencies in time series data, making it suitable for predicting new energy unit output, which changes over time.

[0089] The basic structure of the LSTM model is determined, including the number of neurons in the input layer, hidden layer and output layer, as well as the number of network layers and other parameters. For example, the number of input layer neurons can be determined according to the number of multi-source prediction data features, the hidden layer can be set to multiple LSTM units, and the number of output layer neurons is 1, which is used to output the predicted new energy unit output value.

[0090] The training data set is input into the initial new energy output prediction model for training. During the training process, the model will continuously adjust its parameters based on the input data to minimize the error between the predicted value and the actual value.

[0091] The particle swarm optimization algorithm (PSO) is used to optimize the parameters of the initial new energy output prediction model. PSO is a swarm intelligence-based optimization algorithm that searches for the optimal solution by particles in the solution space. In model training, the PSO algorithm can help find a set of optimal model parameters to make the model's prediction performance better.

[0092] During the training process, the validation data set is used to evaluate the model. The validation data set does not participate in the training of the model, but is used to check the generalization ability of the model during the training process. If the error of the model on the validation data set starts to increase, it may indicate that overfitting has occurred, in which case the model parameters need to be adjusted or some measures to prevent overfitting need to be taken, such as adding a regularization term.

[0093] The accuracy of the optimized initial new energy output prediction model is evaluated based on the test data set. By inputting the data in the test data set into the model, the predicted results are obtained and compared with the actual new energy unit output data, and various evaluation indicators such as mean square error (MSE) and mean absolute error (MAE) are calculated to determine the prediction accuracy of the model. After multiple adjustments and optimizations, a trained new energy output prediction model is obtained, which has high accuracy and generalization ability in predicting new energy unit output.

[0094] In the actual prediction process, real-time multi-source prediction data is obtained, which is also screened and constructed according to the mutual information method described above, and contains key meteorological data and corresponding new energy unit output data at the prediction time (if there is real-time output data).

[0095] The obtained real-time multi-source prediction data is input into the trained new energy output prediction model for prediction. The model processes the real-time data according to the learned patterns and rules, and finally outputs the predicted new energy output result. This result can provide important reference for power system dispatching, energy management, etc., help to reasonably arrange power resources, and improve the stability and reliability of the power system.

[0096] S3: Determine each data subset of the load data in different dimensions, input each data subset into the prediction model of the corresponding dimension for prediction, and fuse each prediction result to obtain the medium and long term load prediction result of the target area, wherein the data subset at least includes seasonal dimension data, linear trend dimension data and random data;

[0097] Preferably, in an embodiment of the present application, determining each data subset of the load data in different dimensions comprises:

[0098] The load data is converted into time series data with the time of the load data as the index and the load value as the data column.

[0099] The time series data is decomposed and processed based on the STL decomposition method to obtain seasonal dimension data, trend dimension data and random data.

[0100] The STL decomposition method (Seasonal-Trend decomposition using Loess) is a commonly used time series decomposition method, which is based on the local weighted regression (Loess) algorithm. The converted time series data is decomposed and processed. The seasonal dimension data reflects the repeated change pattern of the load data in a fixed period. For example, in power load, the daily peak and valley of electricity consumption shows a certain periodicity, and this periodic change is reflected in the seasonal dimension data. Through the STL decomposition method, the periodic change characteristics can be accurately extracted.

[0101] The trend dimension data represents the overall change trend of the load data in a long time. For example, with the development of economy, population growth or technological progress, the power load may show a gradually rising or falling trend, and this part of the trend information is contained in the trend dimension data.

[0102] The random data is the unpredictable fluctuation part left after removing the seasonality and trend. This part of data may be affected by some sudden factors, such as temporary electricity demand surge under extreme weather, abnormal electricity behavior of individual large industrial users, etc. The load fluctuation caused by these random factors is reflected in the random data. Through STL decomposition, the load data is completely decomposed into three different dimensional data subsets, providing a basis for more accurate prediction later.

[0103] Preferably, in an embodiment of the present application, each data subset is input into the prediction model of the corresponding dimension for prediction, and the respective prediction results are fused to obtain the medium and long term load prediction result of the target area, including:

[0104] The seasonality dimension data is predicted based on a seasonal autoregressive integrated moving average model to obtain a seasonal prediction result;

[0105] The trend dimension data is predicted according to a linear regression model to obtain a trend prediction result;

[0106] The random data is predicted based on an autoregressive integrated moving average model to obtain a random prediction result;

[0107] The seasonal prediction result, the trend prediction result and the random prediction result are fused to obtain the medium and long term load prediction result.

[0108] The SARIMA model is a prediction model specially used for processing time series data with seasonal characteristics. The extracted seasonality dimension data is input into the SARIMA model. The model learns the change law of the load in different seasons and different time periods according to the historical seasonality data pattern. For example, for the electricity peak period in summer and winter, and the electricity difference between weekdays and weekends, etc. Seasonal characteristics, the SARIMA model can predict the load change in the corresponding time period in the future through learning from historical data, so as to obtain the seasonal prediction result.

[0109] The linear regression model is a simple and effective prediction model, which is suitable for data with linear trend. The trend dimension data reflects the long-term change trend of the load, which can usually be approximated by a linear relationship. The trend dimension data is input into the linear regression model. The model will fit a best linear regression straight line according to the historical trend data, and predict the future trend change through the extension of the straight line. For example, if the power load has shown a trend of increasing by a certain percentage every year in the past few years, the linear regression model can predict the load growth in the next few years according to this growth trend to obtain the trend prediction result.

[0110] ARIMA model is mainly used for processing non-stationary time series data. By differencing the data to make it stationary, and then using autoregressive and moving average methods to model and predict. Although random data is unpredictable, there are some statistical rules in a certain extent.

[0111] The random data is input into the ARIMA model. The model analyzes the autocorrelation and partial autocorrelation characteristics of the random data, determines the appropriate model parameters, and predicts the future fluctuations of the random data. Although it is impossible to accurately predict each random fluctuation, the ARIMA model can give a prediction value within a certain probability range, and obtain a random prediction result.

[0112] The seasonal prediction result, trend prediction result and random prediction result obtained in the foregoing are fused. A common fusion method is simple weighted summation, and each prediction result is assigned a weight according to its reliability and importance. For example, if the seasonal factor has a greater impact on the load, a higher weight can be assigned to the seasonal prediction result; if the trend prediction is relatively stable and reliable, the weight can also be appropriately increased.

[0113] The three prediction results are combined into a final medium and long term load prediction result through weighted summation. This result comprehensively considers the changes of load data in seasonal, trend and random dimensions, and can more comprehensively and accurately reflect the future medium and long term load change trend compared with single model or single dimension prediction method, and provides more reliable decision basis for planning, scheduling and resource allocation of power system.

[0114] S4: inputting the load data and future meteorological data into the pre-constructed short-term load prediction model to obtain a short-term load prediction result of the target area;

[0115] Preferably, in an embodiment of the present application, the load data and future meteorological data are input into the pre-constructed short-term load prediction model to obtain a short-term load prediction result of the target area, comprising:

[0116] Feature extraction is performed on the load data and future meteorological data to obtain a training data set;

[0117] An initial short-term load prediction model based on LSTM is constructed;

[0118] The training data set is input into the initial short-term load prediction model for training. In the training process, the parameters of the initial short-term load prediction model are optimized through a back propagation algorithm to obtain a trained short-term load prediction model;

[0119] In the actual prediction process, the real-time multi-source prediction data and real-time load data after feature extraction are input into the initial short-term load prediction model for prediction to obtain a short-term load prediction result.

[0120] Specifically, the extracted load data features and future meteorological data features are integrated, organized according to certain formats and rules, and formed into a training data set. The samples in the training data set are sequentially input into the initial short-term load prediction model, the data enters from the input layer, is processed by the LSTM layer, the LSTM unit processes and memorizes the long-term dependency relationship in the data through the gating mechanism (input gate, forget gate, output gate), extracts features and models the input data, and finally obtains a prediction result through the output layer.

[0121] The prediction result of the model is compared with the actual load data label, and a loss function is used to calculate the difference between the prediction result and the true value. Common loss functions include mean square error, mean absolute error, and mean absolute percentage error. According to the result of the loss function, the gradient of each parameter is calculated through the back propagation algorithm. The gradient represents the degree of influence of the parameter on the loss function. The parameters of the model are updated in the opposite direction of the gradient, so that the value of the loss function gradually decreases. This process continuously adjusts the weights and biases in the LSTM network to optimize the performance of the model. The process of forward propagation, loss calculation, and back propagation is repeated for multiple iterations to train the model. As the training progresses, the parameters of the model are continuously optimized, the value of the loss function gradually decreases, and the fitting degree of the model to the training data continuously improves until the preset stopping condition is met, such as reaching the maximum number of training rounds, the loss function value converging to a certain threshold, etc. A trained short-term load prediction model is obtained.

[0122] Real-time load data and multi-source prediction data (including real-time meteorological data, etc.) are collected and processed according to the feature extraction method to extract the corresponding features. The real-time data after feature extraction is input into the trained short-term load prediction model. The model processes and analyzes the input data based on the learned relationship between load data and meteorological data, etc. Through the memory and prediction ability of the LSTM network, the short-term load prediction result of the target area is output.

[0123] S5: determining an optimal reservoir water storage target curve of the target pumped storage unit based on the new energy output prediction result and the medium and long-term load prediction result, and determining an optimal unit output result of the target pumped storage unit based on the new energy output prediction result and the short-term load prediction result;

[0124] Preferably, in an embodiment of the present application, the determination of the optimal reservoir water storage target curve of the target pumped storage unit based on the new energy output prediction result and the medium and long-term load prediction result comprises:

[0125] The predicted output of new energy sources and the predicted medium- and long-term load are input into a pre-constructed medium- and long-term optimization objective function for solution, resulting in the target curve for the optimal reservoir storage capacity of the target pumped storage unit; wherein, the medium- and long-term optimization objective function is expressed as:

[0126]

[0127] in, For the first The predicted output of new energy sources over a given time period. For the first Medium- to long-term load forecasts for a given period. For pumped storage units in Power generation in the first time period, It is the first Pumping costs for a given time period It is the first Maintenance costs over a specific period of time. , , These are the corresponding weighting coefficients. It represents the total number of time periods in a medium- to long-term plan.

[0128] Specifically, the previously obtained new energy output forecast results and medium- and long-term load forecast results are substituted into the medium- and long-term optimization objective function. These data reflect the power generation of new energy sources and the changing trend of power load in the future.

[0129] The function is solved using a suitable optimization algorithm. Common optimization algorithms include linear programming algorithms (such as the simplex method) and nonlinear programming algorithms (such as gradient descent and quasi-Newton methods). Based on the power generation principle of pumped storage units and the relationship between reservoir water volume and power generation, the reservoir water volume corresponding to each time period is further determined. For example, given the unit's power generation efficiency and the reservoir head, the required reservoir water volume change to generate the corresponding power generation can be calculated using physical models or empirical formulas, thus obtaining the optimal reservoir water volume target curve. This curve describes the optimal value of reservoir water volume that should be achieved in each time period within the medium- to long-term planning timeframe to achieve the economical and stable operation of the power system.

[0130] Preferably, in one embodiment of the present invention, determining the optimal unit output of the target pumped storage unit based on the new energy output forecast results and short-term load forecast results includes:

[0131] The predicted output of new energy sources and the predicted short-term load are input into a pre-constructed short-term optimization objective function for solving, to obtain the optimal unit output of the target pumped storage unit; wherein, the short-term optimization objective function is expressed as:

[0132]

[0133] The constraints of the short-term optimization objective function are represented as:

[0134]

[0135] wherein, is the power generation of the pumped storage unit in the i-th time period, is the pumping power of the pumped storage unit in the i-th time period, represents the output of the pumped storage unit in the i-th time period, represents the new energy output prediction value in the i-th time period, represents the short-term load prediction value in the i-th time period. Specifically, the new energy output prediction result and the short-term load prediction result are substituted into the short-term optimization objective function and the constraints. These real-time or near-term prediction data can reflect the new energy generation and load demand of the power system in the current and short term. By solving the objective function to find a set of optimal pumped storage unit power generation and pumping power, the optimal unit output result of the target pumped storage unit in each time period is determined. According to these results, the pumped storage unit can be directly controlled in the short term to generate and pump water, so as to realize the real-time balance and stable operation of the power system, for example, to increase the unit power generation when the new energy output is insufficient and the load is high; to increase the pumping power for energy storage when the new energy is generated in large quantities and the load is low. S6: Execute the pumped storage unit operation control strategy generated based on the optimal reservoir water storage target curve and the optimal unit output result. Based on the determined optimal reservoir water storage target curve and optimal unit output result, the operation control strategy of the pumped storage unit is generated. This strategy will specify in detail how the pumped storage unit should adjust its power generation and pumping capacity in each time period to follow the optimal reservoir water storage target curve and achieve the optimal unit output. For example, in a certain time period, according to the optimal unit output result, if the unit needs to generate power to meet the load demand, the strategy will clearly specify the size of the power output by the unit; at the same time, combined with the optimal reservoir water storage target curve, if the reservoir water storage capacity is higher than the target value at this time, the strategy may arrange appropriate pumping operation to store the excess water for subsequent power generation or power balance adjustment.

[0136]

[0137]

[0138]

[0139] ​​​​​The control system of the pumped storage power station will receive this operation control strategy and control the unit in real time according to the requirements of the strategy. The automation system of the power station will adjust the guide vane opening of the unit (for power generation conditions) or the speed of the water pump (for pumping conditions) according to the strategy instructions to accurately control the power generation and pumping capacity of the unit. At the same time, the relevant monitoring system will monitor the operating state of the unit in real time, including the power output of the unit, the change of the reservoir water level, the operating parameters of the equipment, etc., to ensure that the unit operates stably according to the strategy requirements.

[0140] Preferably, in one embodiment of the present application, after executing the pumped storage unit operation control strategy generated based on the optimal reservoir storage target curve and the optimal unit output result, it further comprises:

[0141] Real-time acquisition of new energy unit output data, load data and future weather data in the target pumped storage unit influence area;

[0142] Updating the new energy output prediction result, the medium and long-term load prediction result and the load prediction result according to the real-time acquired data;

[0143] According to the updated prediction result, the optimal reservoir storage target curve and the optimal unit output result are obtained.

[0144] Specifically, the updated new energy output prediction result and the medium and long-term load prediction result are input into the medium and long-term optimization objective function to be solved again. Since the prediction result has changed, the optimal reservoir storage target curve may also change. For example, if the updated new energy output prediction result shows that the new energy power generation capacity will increase significantly in the future, in order to better absorb new energy power, it may be necessary to adjust the reservoir storage capacity and increase the pumping energy storage during the period of large new energy power generation, so as to obtain the optimized optimal reservoir storage target curve.

[0145] The updated new energy output prediction result and the short-term load prediction result are substituted into the short-term optimization objective function and the constraint condition to be solved again. The changes of the power system reflected by the real-time data may make the original optimal unit output result no longer optimal, for example, the real-time load suddenly increases, but the new energy output does not change accordingly, which requires the power generation and pumping capacity of the unit to be adjusted again to meet the balance of power supply and demand and achieve the optimal operating state, so as to obtain the optimized optimal unit output result. Through continuous real-time data acquisition, prediction result updating and optimization calculation, the operation control strategy of the pumped storage unit can always adapt to the dynamic changes of the power system, and the stability, reliability and economy of the power system can be improved.

[0146] Another embodiment of the present application provides a pumped storage unit operation control system, specifically, please refer to Figure 2 ,Figure 2 A schematic diagram of an operation control system of a pumped storage unit in one embodiment of the present application is shown, comprising:

[0147] The acquisition module 11 is configured to acquire load data of a target region, new energy unit output data of the target region and future weather data of the target region matched with a target pumped storage unit respectively;

[0148] The output prediction module 12 is configured to filter the new energy unit output data and the future weather data based on a mutual information method, and predict target new energy unit output data of the target region based on a multi-source prediction data set constructed from the filtering results to obtain new energy output prediction results.

[0149] The first load prediction module 13 is configured to determine each data subset of the load data in different dimensions, input each data subset and the future weather data into a prediction model of a corresponding dimension for prediction, and fuse each prediction result to obtain medium and long term load prediction results of the target region, wherein the data subset at least includes seasonal dimension data, linear trend dimension data and random data.

[0150] The second load prediction module 14 is configured to input the load data, the new energy unit output data and the future weather data into a pre-constructed short-term load prediction model to obtain short-term load prediction results of the target region.

[0151] The analysis module 15 is configured to determine an optimal reservoir storage target curve of the target pumped storage unit based on the new energy output prediction results and the medium and long term load prediction results, and determine an optimal unit output result of the target pumped storage unit based on the new energy output prediction results and the short-term load prediction results.

[0152] The execution module 16 is configured to execute a pumped storage unit operation control strategy generated based on the optimal reservoir storage target curve and the optimal unit output result.

[0153] Compared with the prior art, the beneficial effects of the embodiments of the present application are at least one of the following:

[0154] (1) The present application determines the optimal reservoir storage target curve based on the new energy output prediction results and the medium and long term load prediction results, which can consider the change trend of new energy generation and load demand from a long-term perspective, and reasonably plan the reservoir storage capacity. In the new energy generation period, more water can be stored, and in the load peak period and new energy output shortage, more water can be discharged for power generation, realizing the optimized utilization of water resources, improving the energy storage efficiency and regulation capacity of the pumped storage power station, and ensuring the stable power supply of the power system.

[0155] (2) The scheme of the application can better match the pumped storage unit with new energy units and power load, enhance the ability of the power system to cope with new energy uncertainty and load change, improve the flexibility and adaptability of the power system, help to build a new type of power system with new energy as the main body, and promote the green and low-carbon transformation of energy.

[0156] The above-described embodiments only express several embodiments of the present application, and the description is more specific and detailed, but it should not be understood as limiting the scope of the patent of the present application. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, several modifications and improvements can be made, which belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for operating and controlling a pumped storage unit, characterized in that, include: The load data of the target area that matches the target pumped storage unit, the output data of the new energy unit in the target area, and the future meteorological data of the target area are obtained respectively. The process involves filtering the power output data of the new energy generating units and the future meteorological data based on mutual information, and constructing a multi-source prediction dataset based on the filtering results. Specifically, this includes: constructing a feature matrix using the power output data of the new energy generating units as the target variable and the future meteorological data as the feature variable; calculating the mutual information values ​​between different meteorological data and the power output data of the new energy generating units using a mutual information calculation method, obtaining a mutual information value array; traversing the mutual information value array and filtering out meteorological data features whose mutual information values ​​are greater than a preset mutual information threshold; and concatenating the filtered future meteorological data features with the power output data of the new energy generating units to obtain the multi-source prediction data. The output data of the target new energy units in the target area are predicted to obtain the new energy output prediction results. The load data is determined into various data subsets under different dimensions. Each data subset is input into the prediction model of the corresponding dimension for prediction. The prediction results are then fused to obtain the medium- and long-term load prediction results for the target area. The data subsets include at least seasonal dimension data, linear trend dimension data, and random data. The load data and the future meteorological data are input into a pre-constructed short-term load forecasting model to obtain the short-term load forecasting results for the target area. Based on the new energy output forecast results and the medium- and long-term load forecast results, the optimal reservoir water storage target curve for the target pumped storage unit is determined. Based on the new energy output forecast results and the short-term load forecast results, the optimal unit output result of the target pumped storage unit is determined; Execute the pumped storage unit operation control strategy generated based on the optimal reservoir water storage target curve and the optimal unit output result; The process of determining the optimal unit output of the target pumped storage unit based on the new energy output forecast results and the short-term load forecast results includes: The new energy output forecast results and the short-term load forecast results are input into a pre-constructed short-term optimization objective function for solution, to obtain the optimal unit output result of the target pumped storage unit; wherein, the short-term optimization objective function is expressed as: The constraints of the short-run optimization objective function are expressed as follows: in, For the first Power generation of pumped storage units during a given time period For the first The pumping power of the pumped storage unit during a certain period of time. This indicates that the pumped storage unit was in the first The effort exerted during each period, Indicates the first Forecast values ​​of new energy power output for each time period Indicates the first Short-term load forecasts for each time period.

2. The operation control method for a pumped storage unit as described in claim 1, characterized in that, The process of predicting the output data of the target new energy units in the target area to obtain the new energy output prediction result includes: The historical multi-source prediction data obtained in advance is divided into training dataset, validation dataset and test dataset according to the time sequence. Construct an initial renewable energy output prediction model based on LSTM; The training dataset is input into the initial new energy power output prediction model for training. During the training process, the parameters of the initial new energy power output prediction model are optimized using the particle swarm optimization algorithm and the validation dataset. The accuracy of the initial new energy power output prediction model is evaluated based on the test dataset to obtain a well-trained new energy power output prediction model. In the actual prediction process, the acquired real-time multi-source prediction data is input into the new energy output prediction model for prediction, and the new energy output prediction result is obtained.

3. The operation control method for a pumped storage unit as described in claim 1, characterized in that, Determining the subsets of the load data under different dimensions includes: Using the time of the load data as an index and the load value as a data column, the load data is converted into time series data; The time series data will be decomposed based on the STL decomposition method to obtain seasonal dimension data, trend dimension data, and random data.

4. The operation control method for a pumped storage unit as described in claim 1, characterized in that, The step of inputting each subset of data into the prediction model of the corresponding dimension for prediction, and fusing the prediction results to obtain the medium- and long-term load prediction results for the target area includes: The seasonal dimension data are predicted based on the seasonal autoregressive integral moving average model to obtain seasonal prediction results; The trend dimension data are predicted based on the linear regression model to obtain the trend prediction results; The random data is predicted based on the autoregressive integral moving average model to obtain random prediction results; The seasonal forecast results, the trend forecast results, and the stochastic forecast results are fused to obtain the medium- and long-term load forecast results.

5. The operation control method for a pumped storage unit as described in claim 1, characterized in that, The step of inputting the load data and the future meteorological data into a pre-constructed short-term load forecasting model to obtain the short-term load forecasting results for the target area includes: Feature extraction is performed on the load data and the future weather data to obtain a training dataset; Construct an initial short-term load forecasting model based on LSTM; The training dataset is input into the initial short-term load forecasting model for training. During the training process, the parameters of the initial short-term load forecasting model are optimized through the backpropagation algorithm to obtain a trained short-term load forecasting model. In the actual forecasting process, the real-time multi-source forecasting data and real-time load data after feature extraction are input into the initial short-term load forecasting model for forecasting to obtain the short-term load forecasting result.

6. The operation control method for a pumped storage unit as described in claim 1, characterized in that, After executing the pumped storage unit operation control strategy generated based on the optimal reservoir water storage target curve and the optimal unit output result, the following is also included: Real-time acquisition of power output data, load data, and future weather data of new energy units in the area affected by the target pumped storage unit; The new energy output forecast results, the medium- and long-term load forecast results, and the load forecast results are updated based on the real-time acquired data. Based on the updated prediction results, the optimal reservoir storage target curve and the optimal unit output results are used to obtain the optimization results; Based on the optimization results, the control strategy of the pumped storage unit is dynamically adjusted.

7. An operation control system for a pumped storage unit, characterized in that, include: The acquisition module is used to acquire load data of the target area that matches the target pumped storage unit, output data of the new energy unit in the target area, and future meteorological data of the target area, respectively. The power output prediction module is used to filter the power output data of the new energy generating units and the future meteorological data based on the mutual information method, and construct a multi-source prediction dataset based on the filtering results. Specifically, it includes: constructing a feature matrix with the power output data of the new energy generating units as the target variable and the future meteorological data as the feature variable; calculating the mutual information values ​​between different meteorological data and the power output data of the new energy generating units according to the mutual information calculation method, obtaining a mutual information value array; traversing the mutual information value array and filtering out meteorological data features whose mutual information values ​​are greater than a preset mutual information threshold; concatenating the filtered future meteorological data features with the power output data of the new energy generating units to obtain the multi-source prediction data; and predicting the target power output data of the target new energy generating units in the target area to obtain the new energy power output prediction result. The first load forecasting module is used to determine the various data subsets of the load data under different dimensions, input each data subset and the future meteorological data into the corresponding dimension forecasting model for forecasting, and fuse the various forecasting results to obtain the medium- and long-term load forecasting results of the target area. The data subsets include at least seasonal dimension data, linear trend dimension data and random data. The second load forecasting module is used to input the load data and the future meteorological data into a pre-constructed short-term load forecasting model to obtain the short-term load forecasting results for the target area. The analysis module is used to determine the optimal reservoir water storage target curve for the target pumped storage unit based on the new energy output forecast results and the medium- and long-term load forecast results. Based on the new energy output forecast results and the short-term load forecast results, the optimal unit output result of the target pumped storage unit is determined; The execution module is used to execute the pumped storage unit operation control strategy generated based on the optimal reservoir water storage target curve and the optimal unit output result; The process of determining the optimal unit output of the target pumped storage unit based on the new energy output forecast results and the short-term load forecast results includes: The new energy output forecast results and the short-term load forecast results are input into a pre-constructed short-term optimization objective function for solution, to obtain the optimal unit output result of the target pumped storage unit; wherein, the short-term optimization objective function is expressed as: The constraints of the short-run optimization objective function are expressed as follows: in, For the first Power generation of pumped storage units during a given time period For the first The pumping power of the pumped storage unit during a certain period of time. This indicates that the pumped storage unit was in the first The effort exerted during each period, Indicates the first Forecast values ​​of new energy power output for each time period Indicates the first Short-term load forecasts for each time period.

Images