Hydropower cluster output prediction method based on lightgbm-transformer-lstm

Abstract

The present application relates to a hydropower cluster output prediction method based on LightGBM-Transformer-LSTM, comprising the following steps: collecting historical power generation data and historical meteorological data of the target hydropower cluster and preprocessing; determining the time delay feature through the historical power generation data, and constructing the basic feature data; constructing an output prediction model, including three base models, namely the LightGBM base model, the Transformer base model and the LSTM base model; constructing three kinds of input feature data corresponding to the three base models respectively, training the output prediction model; and predicting the power generation data of the target hydropower cluster at the future time through the trained output prediction model.

Description

TECHNICAL FIELD

[0001] The present application relates to a LightGBM-Transformer-LSTM-based hydropower cluster output prediction method, belonging to the technical field of hydropower output prediction. BACKGROUND

[0002] Hydropower generation, as an important part of renewable energy, is widely used worldwide, especially in power system dispatching and energy management. However, hydropower generation has strong seasonality and uncertainty, influenced by precipitation, temperature, watershed conditions and other factors. Therefore, accurate hydropower generation prediction is of great significance for the stable operation of power systems, energy optimization scheduling, unit start-stop arrangement, etc.

[0003] Existing hydropower generation prediction methods can be roughly divided into two categories: traditional statistical models and modern machine learning models. Traditional statistical methods, such as autoregressive integrated moving average models (ARIMA), rely on the regularity of historical data and use time series modeling to predict future power generation. However, these methods have weak ability to handle high-frequency noise and nonlinear relationships, and cannot effectively capture the complex dynamic changes in hydropower generation data. For example, under the influence of factors such as precipitation and temperature, hydropower generation may experience sharp fluctuations, and models such as ARIMA often perform poorly in such situations. SUMMARY

[0004] To solve the problems existing in the prior art, the present application proposes a LightGBM-Transformer-LSTM-based hydropower cluster output prediction method.

[0005] The technical solution of the present application is as follows:

[0006] On the one hand, the present application proposes a LightGBM-Transformer-LSTM-based hydropower cluster output prediction method, including the following steps:

[0007] Collecting historical power generation data and historical weather data of the target hydropower cluster and performing preprocessing;

[0008] Determining the time delay feature with the greatest correlation between the power generation data at a given time and the time-delayed power generation data based on the historical power generation data, and constructing basic feature data based on the preprocessed historical power generation data, historical weather data and time delay feature;

[0009] The output prediction model is constructed, including three base models, which are a LightGBM base model, a Transformer base model and an LSTM base model; three kinds of input feature data corresponding to the three base models are respectively constructed based on basic feature data, and the three kinds of input feature data are respectively input into the three base models; in the training process of the output prediction model, the prediction results of the base models are fused through a weighted integration method, and the output prediction model is iteratively trained according to the loss of the fused prediction results and the real power generation data, and the weight given to each base model in the weighted integration process is adjusted;

[0010] The trained output prediction model is used to predict the power generation data of the target hydropower cluster at a future time.

[0011] As a preferred embodiment, the method for determining the time delay feature with the greatest correlation between the power generation data at a given time and the time-delayed power generation data based on historical power generation data is specifically:

[0012] The Pearson correlation coefficient between the power generation data at different time delays and the power generation data at a given time is calculated by using the Pearson correlation calculation method, and the time delay with the maximum Pearson correlation coefficient is taken as the time delay feature.

[0013] As a preferred embodiment, the method for constructing three kinds of input feature data corresponding to the three base models based on basic feature data is specifically:

[0014] For the LightGBM base model, environmental data of the hydropower cluster is introduced based on the basic feature data to form the first input feature data.

[0015] For the Transformer base model, meteorological sequence data within a first given time window is introduced based on the basic feature data to form the second input feature data.

[0016] For the LSTM base model, power generation sequence data and meteorological sequence data within a second given time window are introduced based on the basic feature data to form the third input feature.

[0017] The first given time window is greater than twice the second given time window.

[0018] As a preferred embodiment, in the output prediction model, the fused prediction result is specifically:

[0019]

[0020] wherein, is the fused prediction result, is the prediction result of the i-th base model, w i is the weight parameter of the i-th base model,m is the number of base models;

[0021] The loss function of the output prediction model training process is set as:

[0022]

[0023] wherein, is the loss function, n is the number of samples, y i represents the true value of the i th sample, represents the prediction value of the i th sample by the j th base model.

[0024] The weight gradient of each base model is obtained by taking the partial derivative of the calculated loss with respect to the weight of each base model, and the weight of the corresponding base model is updated according to the weight gradient of each base model.

[0025] In another aspect, the present application also provides a hydropower cluster output prediction system based on LightGBM-Transformer-LSTM, comprising:

[0026] A data acquisition module is configured to acquire historical power generation data and historical meteorological data of a target hydropower cluster and perform preprocessing.

[0027] A basic feature construction module is configured to determine a time delay feature with the largest correlation between power generation data at a given time and time-delayed power generation data based on historical power generation data, and construct basic feature data based on the preprocessed historical power generation data, historical meteorological data and time delay features.

[0028] An integrated model training module is configured to construct an output prediction model including three base models, namely a LightGBM base model, a Transformer base model and an LSTM base model. Based on the basic feature data, three kinds of input feature data corresponding to the three base models are constructed, and the three kinds of input feature data are input into the three base models. In the training process of the output prediction model, the prediction results of each base model are fused through a weighted integration method, and the output prediction model is iteratively trained according to the loss of the fused prediction results and the true power generation data, and the weights assigned to each base model in the weighted integration process are adjusted.

[0029] A hydropower cluster output prediction module is configured to predict the power generation data of the target hydropower cluster at a future time through the trained output prediction model.

[0030] As a preferred embodiment, the method of determining a time delay feature with the largest correlation between power generation data at a given time and time-delayed power generation data based on historical power generation data is specifically:

[0031] The Pearson correlation coefficient of the power generation data at different time delays and the power generation data at a given time is calculated by using the Pearson correlation calculation method, and the time delay with the maximum Pearson correlation coefficient is taken as the time delay feature.

[0032] As a preferred embodiment, the method of constructing three kinds of input feature data corresponding to the three base models based on the basic feature data is specifically:

[0033] For the LightGBM base model, on the basis of the basic feature data, environmental data of the hydropower cluster is introduced to form the first input feature data.

[0034] For the Transformer base model, on the basis of the basic feature data, meteorological sequence data within a first given time window is introduced to form the second input feature data.

[0035] For the LSTM base model, on the basis of the basic feature data, power generation sequence data and meteorological sequence data within a second given time window are introduced to form the third input feature.

[0036] Wherein, the first given time window is greater than twice the second given time window.

[0037] As a preferred embodiment, the fused prediction result in the output prediction model is specifically:

[0038]

[0039] Wherein, is the fused prediction result, is the prediction result of the i-th base model, w i is the weight parameter of the i-th base model, m is the number of base models;

[0040] The loss function of the output prediction model training process is set as:

[0041]

[0042] Wherein, is the loss function, n is the sample number, y i represents the true value of the i-th sample, represents the prediction value of the j-th base model for the i-th sample.

[0043] The weight gradient of each base model is obtained by taking the partial derivative of the calculated loss with respect to the weight of each base model, and the weight of the corresponding base model is updated according to the weight gradient of each base model.

[0044] In still another aspect, the present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for predicting output of a hydropower cluster based on LightGBM-Transformer-LSTM according to any one of the embodiments of the present application when executing the program.

[0045] In still another aspect, the present application also provides a computer readable storage medium, having stored thereon a computer program, which, when executed by a processor, implements the method for predicting output of a hydropower cluster based on LightGBM-Transformer-LSTM according to any one of the embodiments of the present application.

[0046] Additional aspects and advantages of the present application will be set forth in the description that follows, and in part will be obvious from that description, or can be learned by practice of the present application. Moreover, the various aspects and advantages of the present application can be realized and obtained by means of the methods and combinations particularly pointed out in the appended claims. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 Flow chart of the method according to the first embodiment of the present application;

[0048] Figure 2 Schematic diagram of the workflow according to the first embodiment of the present application. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0050] It should be understood that the step numbers used herein are only for the convenience of description, and are not limited to the execution sequence of the steps.

[0051] It should be understood that the terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, unless otherwise clearly indicated by the context, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0052] The terms "comprise" and "include" indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not exclude one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0053] The term "and / or," refers to any combination of one or more of the associated listed items, and all possible combinations, and includes these combinations.

[0054] Embodiment one:

[0055] With the rapid development of machine learning techniques, more and more nonlinear modeling methods such as Support Vector Machine (SVM), Random Forest (RF), Gradient Boosting Tree (GBT) are applied to the prediction of hydropower generation. These methods can handle complex nonlinear relationships, especially when there are high interactions and complex patterns between data features, showing better modeling ability. However, although machine learning models can better fit nonlinear data, there are still some problems: first, many machine learning models rely on a large amount of training data, and in the case of insufficient sample size or data imbalance, their prediction effect may not be stable; second, machine learning methods are usually "black box" models, lacking data interpretability, making it difficult to provide detailed decision-making basis for power dispatch departments.

[0056] With the rise of deep learning, Long Short-Term Memory (LSTM) as a classic time series modeling method has gradually been applied to the prediction of hydropower generation. LSTM can handle long-term dependencies in time series data through its memory cells, so it has obvious advantages in capturing seasonal changes, periodic fluctuations, and other aspects of hydropower generation. However, LSTM may face computational efficiency problems when processing long time series, and is also susceptible to data noise, leading to a decrease in prediction accuracy.

[0057] The Transformer model, originally used in the field of natural language processing, has attracted widespread attention for its efficiency in handling long sequence data. Unlike LSTM, Transformer captures long-term dependencies through a self-attention mechanism (Self-Attention), showing high parallelism and computational efficiency in modeling complex time series. However, although Transformer has advantages in handling global dependencies of time series, its modeling ability for short-term dependencies and local dynamic changes is relatively weak.

[0058] In view of the limitations of the prior art, the present embodiment provides a hydropower cluster output prediction method based on LightGBM-Transformer-LSTM, which combines the global modeling ability of LightGBM, the long-term dependency capturing ability of Transformer, and the short-term dependency modeling ability of LSTM, comprehensively improving the accuracy and robustness of hydropower generation prediction. By weighting the prediction results of multiple base models, the present invention can handle global characteristics, long and short-term dependencies, and nonlinear dynamic changes simultaneously, thus providing a new solution for accurate prediction of hydropower generation.

[0059] With reference to the drawings Figure 1 and Figure 2 The method provided by the embodiment comprises the following steps:

[0060] S100, collect historical power generation data and historical meteorological data of a target hydropower cluster and perform preprocessing.

[0061] In step S100, for the historical power generation data, the missing values in the data due to sensor failure or recording error are filled in by using a spline interpolation method, so as to ensure data integrity and continuity; the historical meteorological data are obtained by the European Center for Medium-Range Weather Forecasts (ECMWF), including relative humidity, specific rainwater content, specific snow water content and air temperature; the historical power generation data and the meteorological data are normalized, and the selected historical power generation data and meteorological data are normalized, so as to ensure that the input feature data are in the same scale range and avoid the influence of different dimensions on model training. The normalization formula is as follows:

[0062]

[0063] Wherein, x is the original data, x min and x max are the minimum value and the maximum value of the data set respectively.

[0064] The normalized data not only include the historical power generation data of the target power station, but also fuse the related meteorological data (such as rainfall, temperature, humidity, etc.) of the target power station, so as to comprehensively reflect the influence of environmental factors on power generation. Through this step, we can comprehensively consider the historical power generation characteristics and environmental changes in the input data, and provide more rich input information for the model.

[0065] S200, determine the time delay feature with the largest correlation between the power generation data at a given time and the power generation data after time delay through the historical power generation data; the target is to find the best position of the data delay in the time dimension, so as to maximize the capture of the time dependence of the power generation change.

[0066] Then, the basic feature data are constructed based on the preprocessed historical power generation data, historical meteorological data and time delay feature.

[0067] S300, construct an output prediction model, including three base models, namely a LightGBM base model, a Transformer base model and an LSTM base model; wherein:

[0068] The LightGBM base model learns the relationship between the input features and the power generation of the target hydropower station by training a decision tree model, and is mainly used to capture the global characteristics and nonlinear features in the data.

[0069] The Transformer-based model captures long-term dependencies in the target power plant generation data through a multi-head self-attention mechanism, generating a prediction result with global characteristics. The present application captures long-term dependencies in the target power plant generation data through a Transformer model. The Transformer adopts a multi-head self-attention mechanism, which can consider the relationships between different positions in the sequence when processing the input sequence, capturing long-distance dependency characteristics. In this way, the model can not only focus on the generation at the current time point, but also combine the influencing factors at the distant time point, thereby generating a prediction result with global information. The advantage of the Transformer is that it can be more efficient than traditional recurrent neural networks (RNN) and LSTM when processing long time series data.

[0070] The LSTM-based model models the time series data through the forget gate, input gate and output gate, and the memory unit retains important historical information and discards irrelevant information, which can efficiently capture the short-term dependencies of the data. The local prediction result generated by the LSTM model can accurately reflect the short-term fluctuations of the target power plant, providing accurate local information for the final prediction.

[0071] Then, based on the characteristics of the three base models, new features are added to the shared features, and three kinds of input feature data corresponding to the three base models are constructed. The three kinds of input feature data are input into the three base models, and the prediction results of the base models are fused through a weighted integration method during the training process of the output prediction model. The output prediction model is iteratively trained according to the loss between the fused prediction result and the real generation data, and the weights assigned to each base model in the weighted integration process are adjusted.

[0072] S400, predicting the generation data of the target hydropower cluster at the future time through the trained output prediction model.

[0073] As a preferred embodiment of the present embodiment, in step S200, the method of determining the time delay feature with the greatest correlation between the given time generation data and the time delayed generation data based on historical generation data is specifically:

[0074] The Pearson correlation coefficient between the generation data at different time delays and the generation data at the given time is calculated using the Pearson correlation calculation method, and the time delay with the largest Pearson correlation coefficient is taken as the time delay feature. The Pearson correlation calculation formula is specifically:

[0075]

[0076] where r XYis the Pearson correlation coefficient, cov(X, Y) is the covariance of the power generation data at a given time and the power generation data under time delay Y, and X and Y are the standard deviations of the power generation data at a given time and the power generation data under time delay Y, respectively.

[0077] As a preferred embodiment of the present embodiment, in step S300, the method of constructing three kinds of input feature data corresponding to the three base models based on the basic feature data is specifically:

[0078] For the LightGBM base model, considering that LightGBM focuses on global characteristics and nonlinear relationships between features, on the basis of the basic feature data, environmental data such as reservoir capacity and altitude are introduced to form the first input feature data.

[0079] For the Transformer base model, as a model for processing long-time dependencies, on the basis of the basic feature data, historical data in a longer time range can be focused on, especially when dealing with seasonal and periodic fluctuations, meteorological sequence data in a first given time window is added to form the second input feature data; the meteorological sequence data in the first given time window is, for example, daily meteorological data in the past 30 days.

[0080] For the LSTM base model, it pays more attention to the time sequence of time series data, so its input features also include historical sequence data such as power generation and meteorological data within a time window on the basis of the basic feature data, power generation sequence data and meteorological sequence data within a second given time window are introduced to form the third input feature, the meteorological sequence data within the second given time window is, for example, daily power generation and meteorological data in the past 10 days.

[0081] As a preferred embodiment of the present embodiment, in the output prediction model,

[0082] For each base model, the loss function is:

[0083]

[0084] wherein, is the loss function of the i-th base model, y i is the true value of the i-th sample, is the predicted value of the i-th sample by the j-th base model, and n is the number of samples.

[0085] The fused prediction result is specifically:

[0086]

[0087] wherein, is the prediction result of the fusion, is the prediction result of the i-th base model, w i is the weight parameter of the i-th base model, and m is the number of base models;

[0088] The loss function of the entire output prediction model training process is set as:

[0089]

[0090] wherein, is the loss function, n is the number of samples, y i represents the true value of the i-th sample, represents the prediction value of the j-th base model for the i-th sample;

[0091] In the processing prediction model, the weight wi of each base model is determined by optimizing the objective function of the entire integrated model, that is, minimizing the loss. The optimization of the weight is carried out by the method of gradient descent. The gradient of the loss function with respect to each weight, that is, the contribution of each weight to the total loss function, is calculated by the back propagation algorithm. The formula for calculating the weight gradient of each base model is:

[0092]

[0093] wherein, is the gradient of the loss function with respect to the weight of the j-th base model,

[0094] is the weighted fusion prediction result.

[0095] After calculating the gradient of each base model, the weight is updated using the gradient descent method. The formula for updating the weight is as follows:

[0096]

[0097] wherein, η is the learning rate, which controls the step size of each update.

[0098] Embodiment Two:

[0099] The embodiment proposes a hydropower cluster output prediction system based on LightGBM-Transformer-LSTM, which comprises:

[0100] A data acquisition module is used to acquire historical power generation data and historical meteorological data of the target hydropower cluster and to preprocess them. This module is used to realize the function of step S100 in embodiment one, and will not be described here again;

[0101] The basic feature construction module is configured to determine a time delay feature with which power generation data at a given time is most relevant to time delayed power generation data based on historical power generation data, and construct basic feature data based on the preprocessed historical power generation data, historical weather data and the time delay feature. The module is configured to implement the function of step S200 in Embodiment I, and thus will not be described herein again.

[0102] The integrated model training module is configured to construct an output prediction model including three base models, i.e., a LightGBM base model, a Transformer base model and an LSTM base model. Three kinds of input feature data corresponding to the three base models are constructed based on the basic feature data, and the three kinds of input feature data are respectively input into the three base models. In the training process of the output prediction model, the prediction results of the base models are fused by a weighted integration method, the output prediction model is iteratively trained according to the loss of the fused prediction results and the real power generation data, and the weights assigned to the base models in the weighted integration process are adjusted. The module is configured to implement the function of step S300 in Embodiment I, and thus will not be described herein again.

[0103] The hydropower cluster output prediction module is configured to predict the power generation data of a target hydropower cluster at a future time by using the trained output prediction model. The module is configured to implement the function of step S400 in Embodiment I, and thus will not be described herein again.

[0104] As a preferred embodiment of the present embodiment, the method of determining a time delay feature with which power generation data at a given time is most relevant to time delayed power generation data based on historical power generation data is specifically as follows:

[0105] The Pearson correlation coefficient of power generation data at different time delays and power generation data at a given time is calculated by using the Pearson correlation calculation method, and the time delay with the largest Pearson correlation coefficient is taken as the time delay feature.

[0106] As a preferred embodiment of the present embodiment, the method of constructing three kinds of input feature data corresponding to the three base models based on the basic feature data is specifically as follows:

[0107] For the LightGBM base model, environmental data in which the hydropower cluster is located is introduced based on the basic feature data to form first input feature data.

[0108] For the Transformer base model, weather sequence data within a first given time window is introduced based on the basic feature data to form second input feature data.

[0109] For the LSTM base model, power sequence data and weather sequence data within a second given time window are introduced based on the basic feature data to form third input feature.

[0110] The first given time window is greater than twice the second given time window.

[0111] As a preferred embodiment of the present embodiment, in the output prediction model, the fused prediction result is specifically:

[0112]

[0113] wherein, is the fused prediction result, is the prediction result of the i-th base model, w i

[0114] is the weight parameter of the i-th base model, m is the number of base models;

[0115] The loss function of the output prediction model training process is set as:

[0116]

[0117] wherein, is the loss function, n is the number of samples, y i represents the true value of the i-th sample, represents the prediction value of the j-th base model for the i-th sample;

[0118] According to the calculated loss, the weight of each base model is derived to obtain the weight gradient of each base model, and the weight of the corresponding base model is updated according to the weight gradient of each base model.

[0119] Embodiment three:

[0120] The present embodiment provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the program to realize the hydropower cluster output prediction method based on LightGBM-Transformer-LSTM according to any embodiment of the present application.

[0121] Embodiment four:

[0122] The present embodiment provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to realize the hydropower cluster output prediction method based on LightGBM-Transformer-LSTM according to any embodiment of the present application.

[0123] In the embodiments of the present application, "at least one" means one or more, and "multiple" means two or more. The "and / or" describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can represent the cases of A alone, A and B together, and B alone. Wherein A and B can be singular or plural. The character " / " generally represents an "or" relationship between the front and rear associated objects. "At least one of the following" and the like means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0124] Those of ordinary skill in the art can realize that the units and algorithm steps described in the embodiments disclosed herein can be realized in electronic hardware, computer software, and a combination of electronic hardware and computer software. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0125] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0126] In several embodiments provided by the present application, any function realized in the form of a software function unit and sold or used as an independent product can be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that make contributions to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (Read-Only Memory; hereinafter referred to as: ROM), a random access memory (Random Access Memory; hereinafter referred to as: RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0127] The above only describes the embodiments of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation based on the content of the present application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for predicting the output of hydropower clusters based on LightGBM-Transformer-LSTM, characterized in that, Includes the following steps: Collect and preprocess historical power generation data and historical meteorological data of the target hydropower cluster. By using historical power generation data, determine the time delay feature that has the greatest correlation between power generation data at a given moment and power generation data after the time delay; Basic feature data are constructed based on preprocessed historical power generation data, historical meteorological data, and time delay characteristics; A power output prediction model is constructed, comprising three base models: LightGBM, Transformer, and LSTM. Based on basic feature data, three types of input feature data are constructed for each of the three base models. These three types of input feature data are then input into the three base models. During the training of the power output prediction model, the prediction results of each base model are fused using a weighted ensemble method. The power output prediction model is iteratively trained based on the loss between the fused prediction results and the actual power generation data, and the weights assigned to each base model during the weighted ensemble process are adjusted. The power generation data of the target hydropower cluster at future moments are predicted using a trained power output prediction model. Specifically, the method for constructing three types of input feature data corresponding to the three base models based on basic feature data is as follows: For the LightGBM base model, environmental data related to the location of the hydropower cluster is introduced on the basis of basic feature data to form the first input feature data; For the Transformer base model, meteorological sequence data within a first given time window is introduced into the basic feature data to form the second input feature data. For the LSTM base model, in addition to the basic feature data, power generation sequence data and meteorological sequence data within a second given time window are introduced to form the third input feature. The first given time window is greater than twice the size of the second given time window.

2. The method for predicting the output of a hydropower cluster based on LightGBM-Transformer-LSTM according to claim 1, characterized in that, The method for determining the time delay feature that has the greatest correlation between power generation data at a given moment and power generation data after a time delay, using historical power generation data, is as follows: The Pearson correlation method is used to calculate the Pearson correlation coefficient between power generation data and power generation data at a given time under different time delays, and the time delay with the largest Pearson correlation coefficient is taken as the time delay feature.

3. The hydropower cluster output prediction method based on LightGBM-Transformer-LSTM according to claim 1, characterized in that, In the power output prediction model, the fused prediction result is specifically as follows: ； in, To integrate the prediction results, Let i be the prediction result of the i-th base model. Let be the weight parameters of the i-th base model. The number of base models; The loss function for the power output prediction model training process is set as follows: ； in, Let n be the loss function and n be the number of samples. This represents the true value of the i-th sample. This represents the prediction value of the j-th base model for the i-th sample; The partial derivative of the calculated loss with respect to the weights of each base model is used to obtain the weight gradient of each base model. The weights of the corresponding base model are then updated based on the weight gradient of each base model.

4. A hydropower cluster output prediction system based on LightGBM-Transformer-LSTM, characterized in that, include: The data acquisition module is used to collect historical power generation data and historical meteorological data of the target hydropower cluster and perform preprocessing. The basic feature construction module is used to determine the time delay feature that has the greatest correlation between the power generation data at a given moment and the power generation data after the time delay, based on historical power generation data; Basic feature data are constructed based on preprocessed historical power generation data, historical meteorological data, and time delay characteristics; An integrated model training module is used to construct a power output prediction model, which includes three base models: LightGBM, Transformer, and LSTM. Based on basic feature data, three types of input feature data are constructed for each of the three base models. These three types of input feature data are then input into the three base models. During the training of the power output prediction model, the prediction results of each base model are fused using a weighted ensemble method. The power output prediction model is iteratively trained based on the loss between the fused prediction results and the actual power generation data, and the weights assigned to each base model during the weighted ensemble process are adjusted. The hydropower cluster output prediction module is used to predict the power generation data of the target hydropower cluster at future times using a trained output prediction model. Specifically, the method for constructing three types of input feature data corresponding to the three base models based on basic feature data is as follows: For the LightGBM base model, environmental data related to the location of the hydropower cluster is introduced on the basis of basic feature data to form the first input feature data; For the Transformer base model, meteorological sequence data within a first given time window is introduced into the basic feature data to form the second input feature data. For the LSTM base model, in addition to the basic feature data, power generation sequence data and meteorological sequence data within a second given time window are introduced to form the third input feature. The first given time window is greater than twice the size of the second given time window.

5. The hydropower cluster output prediction system based on LightGBM-Transformer-LSTM according to claim 4, characterized in that, The method for determining the time delay feature that has the greatest correlation between power generation data at a given moment and power generation data after a time delay, using historical power generation data, is as follows: The Pearson correlation method is used to calculate the Pearson correlation coefficient between power generation data and power generation data at a given time under different time delays, and the time delay with the largest Pearson correlation coefficient is taken as the time delay feature.

6. The hydropower cluster output prediction system based on LightGBM-Transformer-LSTM according to claim 4, characterized in that, In the power output prediction model, the fused prediction result is specifically as follows: ； in, To integrate the prediction results, Let i be the prediction result of the i-th base model. Let be the weight parameters of the i-th base model. The number of base models; The loss function for the power output prediction model training process is set as follows: ； in, Let n be the loss function and n be the number of samples. This represents the true value of the i-th sample. This represents the prediction value of the j-th base model for the i-th sample; The partial derivative of the calculated loss with respect to the weights of each base model is used to obtain the weight gradient of each base model. The weights of the corresponding base model are then updated based on the weight gradient of each base model.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the hydropower cluster output prediction method based on LightGBM-Transformer-LSTM as described in any one of claims 1 to 3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the hydropower cluster output prediction method based on LightGBM-Transformer-LSTM as described in any one of claims 1 to 3.

Images