Load data determination method and device, storage medium and electronic equipment

Abstract

The embodiment of the invention provides a load data determination method and device, a storage medium and electronic equipment, and the method comprises the steps: training a plurality of single prediction models according to first load data, first meteorological data and first time data of a power generation system in a first time period, determining the first prediction precision of each single prediction model after training, wherein each single prediction model is a model for predicting the load of the power generation system; inputting second meteorological data and second time data of the power generation system at the current time into each trained single prediction model to determine second load data, predicted by each trained single prediction model, of the power generation system at the target time, determining a weight corresponding to each single prediction model according to each first prediction precision; and determining target load data of the power generation system at the target time according to the multiple pieces of second load data and the weight corresponding to each single prediction model.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a method and device for determining load data, a storage medium, and an electronic device. Background Art

[0002] At present, the mainstream new energy load forecasting mainly applies: combining historical data analysis and trend analysis, through the collection, cleaning and sorting of historical load data, revealing the seasonal, periodic and random laws of load changes. Statistical analysis methods (such as mean, standard deviation, correlation coefficient, etc.) are used to extract useful information from historical data, and combined with the long-term trend of load changes over time for forecasting.

[0003] The accuracy and completeness of historical data are the basis for the reliability of forecast results. However, due to errors, omissions or outliers in the data collection and recording process, historical data may not be accurate enough, thus affecting the accuracy of forecast results. Trend analysis is mainly used to analyze simple and linear trends, but cannot effectively handle complex nonlinear trends and complex trends caused by the interaction of multiple factors. For new energy load forecasting, due to the influence of multiple factors such as weather, policies, and socio-economic conditions, its change trend is often more complex and changeable.

[0004] Currently, no effective solution has been proposed to the problem of low prediction accuracy of load data predicted by combining historical data analysis and trend analysis in related technologies.

[0005] Therefore, it is necessary to improve the related technology to overcome the above-mentioned defects in the related technology. Summary of the invention

[0006] The embodiments of the present application provide a method and device for determining load data, a storage medium, and an electronic device to at least solve the problem of low prediction accuracy of load data predicted by combining historical data analysis and trend analysis in the related art.

[0007] According to an embodiment of the present application, a method for determining load data is provided, comprising: training multiple single prediction models according to first load data, first meteorological data and first time data of a power generation system in a first time period, and determining a first prediction accuracy of each single prediction model after training, wherein each single prediction model is a model for predicting the load of the power generation system, and the first time period is a time period before a current time; inputting second meteorological data and second time data of the power generation system at the current time into each single prediction model after training to determine second load data of the power generation system at a target time predicted by each single prediction model after training, and determining a weight corresponding to each single prediction model according to each first prediction accuracy, wherein the target time is a time after the current time; determining the target load data of the power generation system at the target time according to multiple second load data and the weight corresponding to each single prediction model.

[0008] In an exemplary embodiment, multiple single prediction models are trained based on first load data, first meteorological data and first time data of a power generation system in a first time period, including: performing data cleaning on the first load data, the first meteorological data and the first time data to process abnormal values ​​and missing values ​​in the first load data, the first meteorological data and the first time data, and obtaining the cleaned first load data, the cleaned first meteorological data and the cleaned first time data; standardizing the cleaned first load data, the cleaned first meteorological data and the cleaned first time data to unify the data formats of the cleaned first load data, the cleaned first meteorological data and the cleaned first time data; and training the multiple single prediction models based on the standardized first load data, the standardized first meteorological data and the standardized first time data.

[0009] In an exemplary embodiment, data cleaning is performed on the first load data, the first meteorological data and the first time data to process abnormal values ​​and missing values ​​in the first load data, the first meteorological data and the first time data, including: identifying missing values ​​in the first load data, the first meteorological data and the first time data through an objective function, and determining the proportion of the missing values ​​in the first load data, the first meteorological data and the first time data; when the proportion is less than a preset proportion, deleting the rows corresponding to the missing values ​​and / or the columns corresponding to the missing values; filling the rows corresponding to the missing values ​​and / or the columns corresponding to the missing values ​​based on a target model; and, identifying abnormal values ​​in the first load data, the first meteorological data and the first time data based on target rules; replacing the abnormal values ​​according to the average value and / or median of neighboring points.

[0010] In an exemplary embodiment, the multiple single prediction models are trained according to the standardized first load data, the standardized first meteorological data and the standardized first time data, and the first prediction accuracy of each single prediction model after training is determined, including: grouping the standardized first load data, the standardized first meteorological data and the standardized first time data to obtain training set data and validation set data; training the multiple single prediction models with the training set data; inputting the meteorological data and time data in the validation set data into each single prediction model after training to obtain prediction results; comparing each load data in the validation set data with each prediction result, and calculating the first prediction accuracy of each single prediction model after training according to the first comparison result.

[0011] In an exemplary embodiment, determining the weight corresponding to each single prediction model according to each first prediction accuracy includes: comparing each first prediction accuracy with an accuracy threshold to determine a second comparison result; obtaining a second prediction accuracy whose first prediction accuracy is greater than or equal to the accuracy threshold from multiple first prediction accuracies according to the second comparison result; and updating the multiple single prediction models according to the second prediction accuracy, wherein the updated multiple single prediction models include: a single prediction model corresponding to each second prediction accuracy.

[0012] In an exemplary embodiment, determining the target load data of the power generation system at the target time according to the plurality of second load data and the weight corresponding to each single prediction model includes: performing weighted averaging on the plurality of second load data according to a first formula to obtain the target load data, wherein the first formula is: Wherein, Y is the target load data, which is i Includes: y1, y2, …, y n ,y i are the plurality of second load data, y1, y2, ..., y n For each second load data, ω i Including: ω1, ω2, …, ω n ,ω i are the weights corresponding to multiple single prediction models, ω1, ω2, ..., ω n are the weights corresponding to each single prediction model.

[0013] In an exemplary embodiment, after determining the target load data of the power generation system at the target time based on multiple second load data and the weights corresponding to each single prediction model, the method also includes: updating the first meteorological data, the first time data and the first load data, wherein the updated first meteorological data includes: the second meteorological data, the updated first time data includes: the second time data, and the updated first load data includes: the target load data; updating the trained multiple single prediction models according to the updated first meteorological data, the updated first time data and the updated first load data.

[0014] According to another embodiment of the present application, a load data determination device is provided, comprising: a training module, used to train multiple single prediction models according to first load data, first meteorological data and first time data of a power generation system in a first time period, and determine a first prediction accuracy of each single prediction model after training, wherein each single prediction model is a model for predicting the load of the power generation system, and the first time period is a time period before a current time; an input module, used to input second meteorological data and second time data of the power generation system at the current time into each trained single prediction model, so as to determine the second load data of the power generation system at a target time predicted by each trained single prediction model, and determine a weight corresponding to each single prediction model according to each first prediction accuracy, wherein the target time is a time after the current time; a determination module, used to determine the target load data of the power generation system at the target time according to multiple second load data and the weight corresponding to each single prediction model.

[0015] According to another embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when run.

[0016] According to another embodiment of the present application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0017] According to another embodiment of the present application, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0018] Through the embodiment of the present application, multiple single prediction models are trained according to the first load data, first meteorological data and first time data of the power generation system in the first time period before the current time, and the first prediction accuracy of each single prediction model after training is determined, wherein each single prediction model is a model for predicting the load of the power generation system; the second meteorological data and the second time data of the power generation system at the current time are input into each trained single prediction model to determine the second load data of the power generation system at the target time after the current time predicted by each trained single prediction model, and the corresponding weight of each single prediction model is determined according to each first prediction accuracy; the target load data of the power generation system at the target time is determined according to the weight corresponding to the multiple second load data and each single prediction model. That is to say, the embodiment of the present application uses historical data, including the first load data, first meteorological data and first time data of the power generation system in the first time period, to train multiple different single prediction models. Then, at the current time, the second meteorological data and the second time data are input into the multiple trained single prediction models to generate the second load data for the future target time. Through the embodiment of the present application, the problem of low prediction accuracy of load data prediction by combining historical data analysis and trend analysis in the related art can be solved. Furthermore, by integrating the advantages of multiple single forecasting models and dynamically adjusting the forecast based on relevant data, the accuracy and reliability of load forecasting are significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0021] Figure 1It is a hardware structure block diagram of a computer terminal device of a method for determining load data in an embodiment of the present application;

[0022] Figure 2 is a flow chart of a method for determining load data according to an embodiment of the present application;

[0023] Figure 3 is a schematic diagram of a power generation system load prediction system based on a machine learning prediction algorithm according to an optional embodiment of the present application;

[0024] Figure 4 is a flow chart of a method for load prediction of a power generation system based on a machine learning prediction algorithm according to an optional embodiment of the present application;

[0025] Figure 5 is a schematic diagram comparing actual load and predicted load curves of simulation data according to an optional embodiment of the present application;

[0026] Figure 6 is a schematic diagram of the accuracy of load forecasting in the last seven days of simulation data according to an optional embodiment of the present application;

[0027] Figure 7 It is a structural block diagram of a device for determining load data according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0030] The method embodiments provided in the embodiments of the present application can be executed in a computer terminal device or a similar computing device. Taking running on a computer terminal device as an example, Figure 1 1 is a hardware structure block diagram of a computer terminal device of a method for determining load data in an embodiment of the present application. Figure 1 As shown, the computer terminal device may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the above-mentioned computer terminal device may also include a transmission device 106 and an input and output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned computer terminal device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0031] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method for determining load data in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, the above method is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0032] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of a computer terminal device. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0033] In this embodiment, a method for determining load data is provided. Figure 2 is a flow chart of a method for determining load data according to an embodiment of the present application. Figure 2 As shown, the process includes the following steps:

[0034] Step S202, training multiple single prediction models according to first load data, first meteorological data and first time data of the power generation system in a first time period, and determining a first prediction accuracy of each single prediction model after training, wherein each single prediction model is a model for predicting the load of the power generation system, and the first time period is a time period before the current time;

[0035] Step S204, inputting the second meteorological data and the second time data of the power generation system at the current time into each of the trained single prediction models to determine the second load data of the power generation system at the target time predicted by each of the trained single prediction models, and determining the weight corresponding to each of the single prediction models according to each first prediction accuracy, wherein the target time is the time after the current time;

[0036] Step S206, determining the target load data of the power generation system at the target time according to the plurality of second load data and the weight corresponding to each single prediction model.

[0037] Through the above steps, multiple single prediction models are trained according to the first load data, first meteorological data and first time data of the power generation system in the first time period before the current time, and the first prediction accuracy of each single prediction model after training is determined, wherein each single prediction model is a model for predicting the load of the power generation system; the second meteorological data and the second time data of the power generation system at the current time are input into each trained single prediction model to determine the second load data of the power generation system at the target time after the current time predicted by each trained single prediction model, and the corresponding weight of each single prediction model is determined according to each first prediction accuracy; the target load data of the power generation system at the target time is determined according to the weight corresponding to the multiple second load data and each single prediction model. That is to say, the embodiment of the present application uses historical data, including the first load data, first meteorological data and first time data of the power generation system in the first time period, to train multiple different single prediction models. Then, at the current time, the second meteorological data and the second time data are input into the multiple trained single prediction models to generate the second load data for the future target time. Through the embodiment of the present application, the problem of low prediction accuracy of predicting load data by combining historical data analysis and trend analysis in the related art can be solved. Furthermore, by integrating the advantages of multiple single forecasting models and dynamically adjusting the forecast based on relevant data, the accuracy and reliability of load forecasting are significantly improved.

[0038] Optionally, the above-mentioned step S202 trains multiple single prediction models according to the first load data, the first meteorological data and the first time data of the power generation system in the first time period, including: performing data cleaning on the first load data, the first meteorological data and the first time data to process the abnormal values ​​and missing values ​​in the first load data, the first meteorological data and the first time data, and obtaining the cleaned first load data, the cleaned first meteorological data and the cleaned first time data; standardizing the cleaned first load data, the cleaned first meteorological data and the cleaned first time data to unify the data format of the cleaned first load data, the cleaned first meteorological data and the cleaned first time data; training the multiple single prediction models according to the standardized first load data, the standardized first meteorological data and the standardized first time data.

[0039] It is understandable that data can be preprocessed, specifically: Data cleaning: Data cleaning aims to remove outliers and missing values ​​from the data set. Outliers may be caused by measurement errors or extreme weather conditions, while missing values ​​may be caused by missing records. Through cleaning, the data set used for training is accurate and the effectiveness of model learning is improved. For example, for outliers, statistical methods such as Z-score or IQR can be used to identify and remove them; for missing values, they can be filled through techniques such as interpolation, forward filling or backward filling.

[0040] Standardization: Data standardization is the process of converting data to the same scale so that data of different dimensions and ranges can be fairly compared in model training. Standardization usually includes data normalization (such as scaling the data to the 0-1 interval) or standard normalization (making the data have zero mean and unit variance). This step is particularly important for machine learning models because it helps the model better learn patterns in the data and avoid training skew caused by differences in feature magnitude.

[0041] Model training: After data cleaning and standardization, multiple single prediction models are trained using the processed first load data (historical load), first meteorological data (historical meteorological conditions), and first time data (historical time information). These models may include linear regression, support vector machines, deep learning models such as BP neural networks, etc. The goal of training is to find the relationship between load and meteorological conditions and time information in the data, so that the model can predict the load output from the input data.

[0042] Through the above process, the quality and consistency of the data are ensured, and the generalization ability and prediction accuracy of the prediction model are improved. In the subsequent real-time prediction, the second meteorological data and the second time data of the current time can be input into the trained model to obtain a more accurate load forecast of the power generation system at the target time, providing decision support for power dispatching and management. This method can effectively cope with the volatility of power load, especially under the influence of weather changes and time periodicity, and improve the operating efficiency and stability of the power system.

[0043] Among them, data cleaning is performed on the first load data, the first meteorological data and the first time data to process abnormal values ​​and missing values ​​in the first load data, the first meteorological data and the first time data, including: identifying missing values ​​in the first load data, the first meteorological data and the first time data through an objective function, and determining the proportion of the missing values ​​in the first load data, the first meteorological data and the first time data; when the proportion is less than a preset proportion, deleting the rows corresponding to the missing values ​​and / or the columns corresponding to the missing values; filling the rows corresponding to the missing values ​​and / or the columns corresponding to the missing values ​​based on a target model; and identifying abnormal values ​​in the first load data, the first meteorological data and the first time data based on target rules; replacing the abnormal values ​​according to the average value and / or median of neighboring points.

[0044] It is understandable that data preprocessing is a crucial step before training a machine learning model for power generation system load forecasting to ensure the quality of the data and the prediction accuracy of the model. To identify and process missing values ​​and outliers in data cleaning, specifically:

[0045] Identify and handle missing values: Objective function Identify missing values ​​and their proportions: Set an objective function that can scan the entire data set, including the first load data (historical load records), the first meteorological data (historical weather conditions, such as temperature, humidity, wind speed, etc.), and the first time data (historical time information, such as date, time, holidays, etc.), and identify missing values. Calculate the proportion of missing values ​​in each data category, that is, the number of missing values ​​divided by the total number of data in that category. This helps to assess the severity of the data integrity problem.

[0046] Delete or fill missing values: Preset ratio: Define a threshold. If the missing value ratio of a feature exceeds the threshold, it may indicate that the feature data record is incomplete. Directly delete the corresponding rows or columns to maintain the integrity of the data set.

[0047] Data filling: For features with a missing value ratio lower than a preset ratio, data filling techniques can be used. The target model here may be a statistical filling method, such as using the mean, median, or predicting missing values ​​by building a small prediction model (such as KNN, decision tree, etc.). Target rule-based filling aims to retain as much valid information as possible while repairing the incomplete parts of the data set.

[0048] Identify and handle outliers: Target rules Identify outliers: Outlier identification is usually based on statistical principles, for example, using Z-score, IQR rule of box plots, or neighborhood-based anomaly detection algorithms. The target rule may refer to determining a reasonable range or threshold, and data points outside this range are considered outliers. When identifying outliers, the distribution characteristics of the data, such as normal distribution, skewed distribution, etc., can be considered to more accurately define the outliers.

[0049] Outlier replacement: After outliers are identified, they are replaced based on the mean and / or median of neighboring points. The selection of neighboring points can be determined by calculating distances, such as using Euclidean distance or Manhattan distance.

[0050] The strategy of replacing outliers is to avoid the influence of extreme values ​​in model training and make the data more consistent with the law of actual load changes, thereby improving the stability and accuracy of the prediction model.

[0051] Through the above data preprocessing steps, the embodiment of the present application can effectively process the missing values ​​and outliers in the data set required for power generation system load forecasting, ensure that the data set used for training is clean and standardized, and provide a solid foundation for subsequent model training and load forecasting. This method not only enhances the generalization ability of the model, but also improves the reliability of the prediction results.

[0052] Among them, the multiple single prediction models are trained according to the first load data after standardization, the first meteorological data after standardization and the first time data after standardization, and the first prediction accuracy of each single prediction model after training is determined, including: grouping the first load data after standardization, the first meteorological data after standardization and the first time data after standardization to obtain training set data and validation set data; training the multiple single prediction models with the training set data; inputting the meteorological data and time data in the validation set data into each single prediction model after training to obtain prediction results; comparing each load data in the validation set data with each prediction result, and calculating the first prediction accuracy of each single prediction model after training according to the first comparison result.

[0053] It is understandable that in the load forecasting of power generation systems, the use of machine learning models for training and verification is a key step to ensure the accuracy and generalization ability of the prediction model. Specifically: Data grouping and preparation, grouping to obtain training sets and validation sets: From the standardized data set, including the first load data (standardized historical load data), the first meteorological data (standardized historical weather data) and the first time data (standardized historical time information), the data is divided into training sets and validation sets by grouping. Typically, most of the data set (such as 70% or 80%) is used to train the model, and the rest (such as 30% or 20%) is used to verify the performance of the model on unseen data.

[0054] Training set is used for model learning: The training set data includes standardized load, weather and time data, which are used to train multiple single prediction models. These models may include linear regression, support vector machine, neural network, etc. Each model learns the relationship between power load and weather conditions and time factors through the training set data.

[0055] Validation set to evaluate model performance: Once the model training is completed, the validation set data is used to evaluate the accuracy and reliability of the model. In this process, the meteorological data and time data in the validation set are input into the trained model to predict the load data, that is, to obtain the prediction results.

[0056] Accuracy calculation and model selection: Comparison of actual and predicted values ​​of the validation set: Compare the actual load data of the validation set with the predicted results of the model. This usually means calculating the difference between the predicted and actual values, such as using mean squared error (MSE), mean absolute error (MAE), or coefficient of determination (R 2 ) and other indicators.

[0057] Calculate the first prediction accuracy: Based on the comparison results, calculate the prediction accuracy of each model. Accuracy can be the inverse of the above error indicators, or a measure of the correlation between the model prediction results and the true results. High accuracy means that the model can accurately predict the load data in the validation set.

[0058] Select the best model: By comparing the first prediction accuracy of different models, the best performing model is selected as the final load forecasting model. The comparison of accuracy helps to identify which model is most effective in handling the load forecasting task, thereby providing the most accurate load forecast for the power system.

[0059] The above embodiments ensure the scientific nature of model training and the reliability of prediction results. Through the performance on the validation set, the prediction ability of the model for unseen data can be evaluated, which is crucial for power dispatching, energy management and demand forecasting. By determining the optimal model, the power system can manage and dispatch resources more effectively to cope with the changing power demand.

[0060] Optionally, the above-mentioned step S204 of determining the weight corresponding to each single prediction model according to each first prediction accuracy includes: comparing each first prediction accuracy with an accuracy threshold to determine a second comparison result; obtaining a second prediction accuracy whose first prediction accuracy is greater than or equal to the accuracy threshold from multiple first prediction accuracies according to the second comparison result; and updating the multiple single prediction models according to the second prediction accuracy, wherein the updated multiple single prediction models include: a single prediction model corresponding to each second prediction accuracy.

[0061] It is understandable that in the machine learning method of power generation system load forecasting, determining the model weight is to fuse the outputs of multiple forecast models to obtain more accurate forecast results. Specifically:

[0062] Accuracy Threshold Setting: Setting the accuracy threshold: The accuracy threshold is a preset performance criterion for screening well-performing models. It is a numerical value that reflects the minimum acceptable accuracy of the predictive model on the validation set. For example, if the accuracy threshold is set to 0.8 (or 80%), only those models that exhibit at least 80% accuracy on the validation set will be further considered.

[0063] Model comparison and screening: Comparison of first prediction accuracy: After model training and preliminary validation, each single prediction model will generate a first prediction accuracy, that is, the prediction accuracy on the validation set. Comparing these first prediction accuracies with the accuracy threshold, we get the second comparison results, which reveal which models' performance exceeds the preset standard.

[0064] Screening the second prediction accuracy: Based on the second comparison results, screen out those models whose first prediction accuracy is greater than or equal to the accuracy threshold from all single prediction models. These screened models will be called models with second prediction accuracy because their prediction performance has reached or exceeded the set accuracy requirement.

[0065] Model update and weight assignment: Update the model set: Update the single prediction model with the second prediction accuracy after screening to a new model set for subsequent prediction fusion. This means removing all models that fail to reach the accuracy threshold and retaining only those models that perform well.

[0066] Determine model weights: Assign a weight to each model based on the second prediction accuracy, i.e. the prediction performance of the selected model. Typically, models with better performance (higher prediction accuracy) will be given greater weights, while models with poorer performance will be given smaller weights. The determination of weights can be based on a direct proportion of prediction accuracy, or it can be optimized through more complex mechanisms such as Bayesian optimization, grid search, and other methods.

[0067] Through the above steps, the embodiment of the present application can effectively screen and optimize the model set, ensuring that only those models that perform well on the validation data will be used for the final load forecast. The allocation of model weights further improves the reliability of the forecast results, because the forecast results are calculated based on the weighted output of the best performing model. This strategy improves the accuracy and stability of the load forecast of the power generation system, which is of great significance for the optimal scheduling of power resources and demand response.

[0068] Optionally, the above step S206 determines the target load data of the power generation system at the target time according to the multiple second load data and the weight corresponding to each single prediction model, including: performing weighted averaging on the multiple second load data according to a first formula to obtain the target load data, wherein the first formula is: Wherein, Y is the target load data, which is i Includes: y1,y2,...,y n ,y i are the plurality of second load data, y1, y2, ..., y n For each second load data, ω i Including: ω1, ω2, ..., ω n ,ω i are the weights corresponding to multiple single prediction models, ω1, ω2, …, ω n are the weights corresponding to each single prediction model.

[0069] It is understandable that in the load forecasting of power generation systems, weighted averaging is a commonly used method to fuse the outputs of multiple forecasting models to obtain more accurate and reliable forecasting results. Specifically: Weighted averaging is a method of averaging in mathematics, in which different values ​​can be assigned different weights to reflect their importance in the final result. In the load forecasting scenario, the prediction results (second load data) of different models may have different prediction accuracy due to the model type, the characteristics of the training data set and other factors. Therefore, through weighted averaging, the outputs of all models can be comprehensively considered, and at the same time, appropriate weights can be assigned to the models according to their prediction accuracy to obtain the final prediction result (target load data).

[0070] Through the embodiments of the present application, the advantages of each model can be fully utilized, and the accuracy and stability of the final prediction results can be ensured through weighted averaging.

[0071] Optionally, after determining the target load data of the power generation system at the target time according to the multiple second load data and the weights corresponding to each single prediction model in the above step S206, the method further includes: updating the first meteorological data, the first time data and the first load data, wherein the updated first meteorological data includes: the second meteorological data, the updated first time data includes: the second time data, and the updated first load data includes: the target load data; updating the trained multiple single prediction models according to the updated first meteorological data, the updated first time data and the updated first load data.

[0072] It is understandable that in the continuous optimization process of power generation system load forecasting, real-time updating of data and retraining of models are important links to maintain forecast accuracy.

[0073] Data update: Update the first meteorological data, the first time data and the first load data: As time goes by, new meteorological data (second meteorological data), time information (second time data) and load data (target load data) are continuously generated. These new data contain the latest status and influencing factors, which are crucial for predicting future loads.

[0074] Specifically, the first meteorological data becomes the second meteorological data after being updated, which means that the latest weather conditions, such as temperature, humidity, wind speed, etc., will be used for subsequent training and prediction of the model. The first time data is updated to the second time data, which includes new timestamps, dates, holiday information, etc. These time information are essential for capturing the periodic and seasonal patterns of the load. The first load data is updated to the prediction result based on weighted average, that is, the target load data. This update reflects the latest results of historical predictions and can be used for retraining and optimization of the model.

[0075] Model update: Update multiple single prediction models after training: After the data is updated, the updated data (updated first meteorological data, first time data and first load data) are used to retrain and optimize the model. The model update process may include the following steps:

[0076] Repartition the data: Repartition the updated data into training and validation sets.

[0077] Retrain models: Each single forecasting model is retrained using the updated training set data to learn the latest load patterns and weather trends.

[0078] Adjust model parameters: Based on the updated validation set data, adjust the model parameters to improve prediction accuracy.

[0079] Evaluate updated models: Recalculate the prediction accuracy of each model to ensure that the model's performance still meets the requirements.

[0080] Through continuous data updates and model retraining, the embodiments of the present application can keep the model up to date and promptly reflect the changing trends of the power system load, especially in the face of special circumstances such as sudden weather changes and holiday load fluctuations, and can provide more accurate prediction results. This dynamic update mechanism is crucial for power dispatching, energy management, and demand response, and can significantly improve the efficiency and stability of power generation system operation, and reduce energy waste and cost increases caused by inaccurate predictions. At the same time, it can also help power companies better plan resources, cope with peak load periods, and ensure the reliability of power supply.

[0081] In order to better understand the process of the above-mentioned load data determination method, the implementation method flow of the above-mentioned load data determination is described below in combination with an optional embodiment, but it is not used to limit the technical solution of the embodiment of the present application.

[0082] The technical problem to be solved by the optional embodiments of the present application is: the optional embodiments of the present application mainly consider various factors such as weather changes and emergencies through machine learning algorithms to establish a more comprehensive and accurate prediction model to solve the current problems of low prediction accuracy, few prediction models, and inability to predict in real time, thereby assisting the power system to more effectively manage and dispatch energy, reduce energy waste, and improve energy utilization.

[0083] Figure 3 is a schematic diagram of a power generation system load prediction system based on a machine learning prediction algorithm according to an optional embodiment of the present application, such as Figure 3 As shown, an optional embodiment of the present application provides a power generation system load forecasting system based on a machine learning prediction algorithm, which mainly includes a data preparation module, a model training module, a model verification and optimization module, and an online operation module, wherein:

[0084] The data preparation module 32 performs data collection, including exporting the historical load data (first load data), weather data (first meteorological data), temperature data, etc. of the past three years from the database, and uploading the real-time data. Then, the collected data is processed to remove missing values ​​and outliers, and the text format data is converted into numerical format. Finally, the processed data is saved in an appropriate format (such as CSVJSON or database) to facilitate subsequent model training and real-time data access.

[0085] That is, collect and obtain historical data of the load in the substation area, historical weather data, historical temperature data, electricity consumption data of major power-consuming enterprises or units, first-time data, etc. First, perform data normalization, data standardization, principal component analysis and other processing on the data, divide the processed data into training group and test group data, and use linear regression, logistic regression, Boost regression, Kalman filtering, BP neural network prediction, and combined prediction models to train these data every day through the big data platform, and use the verification data set to verify the model to obtain the optimal option of the prediction model, obtain load data, weather data, and temperature data online in real time, and combine the prediction model to calculate the predicted value for a certain time in the future. The first meteorological data includes wind speed, wind direction, temperature, air pressure, humidity, etc. The first-time data includes season, date, time period, etc.

[0086] The model training module 34 divides the data processed and saved in the data preparation stage into a training set and a validation set, uses machine learning frameworks such as sk-learn and tensorflow, and uses linear regression, logistic regression, Boost regression, Kalman filtering, BP neural network prediction, and combined prediction models (i.e., multiple single prediction models) for training, and repeatedly verifies and optimizes the model results until the expected prediction accuracy is achieved.

[0087] The model verification and optimization module 36 first uses the forecast model of the day at the beginning of each day, combined with the forecast results of the previous day, real-time load data, and the latest weather and temperature information to make forecasts. During the forecasting process, the forecast results are fine-tuned according to the changes in real-time data, such as considering the immediate impact of sudden weather changes on load.

[0088] The online operation module 38 first deploys the model and deploys the optimal prediction model to the online environment to ensure that the prediction request can be responded to quickly. Then the prediction results are displayed in a visual form to relevant personnel such as power dispatchers. The model operation status is monitored, including indicators such as response time and prediction accuracy, to ensure stable operation of the system.

[0089] It also includes: a power generation system load prediction module 40. The collected historical data is normalized, standardized, and subjected to principal component analysis in order to make the data easier to calculate and eliminate factors that have a small impact on the load (i.e., data dimension reduction). The processed data set is then divided into a training set and a validation set, and interactive training and validation are performed respectively to ensure the accuracy of the prediction model. The prediction accuracy is compared and verified by using different prediction model algorithms. Finally, the optimal result is selected to run in an online environment, and the algorithm is finely adjusted in combination with the actual environment, weather, and important events to obtain the final prediction result.

[0090] Figure 4is a flow chart of a method for load forecasting of a power generation system based on a machine learning prediction algorithm according to an optional embodiment of the present application, such as Figure 4 As shown:

[0091] Step S401, collecting temperature historical data, load historical data, and weather historical data.

[0092] Historical data: Export the historical load data, weather data (such as wind speed, sunshine duration, precipitation, etc.), temperature data (daily maximum temperature, daily minimum temperature, average temperature, etc.) and time history data of the power substation area in the past three years from the database or data warehouse. Real-time data: Ensure that the necessary sensors are installed in the online environment to report load, weather (such as through the weather station interface), and temperature data in real time.

[0093] Step S402, normalize and standardize the data.

[0094] Remove missing values ​​and outliers (such as extreme load values ​​and unreasonable temperature values).

[0095] Data standardization or normalization is performed to ensure that data of different dimensions can be treated fairly in the same model.

[0096] Convert text-based data into numeric format, such as converting date and time into days or specific time period codes.

[0097] Step S403: principal component factor analysis.

[0098] Save the cleaned data in an appropriate format (such as CSV, JSON or database) to facilitate subsequent model training and real-time data access.

[0099] Step S404: split the data into a training data set and a validation data set.

[0100] Step S405, performing linear regression training, Boost regression training, Kalman filter training, BP neural network training and combined prediction model training according to the data.

[0101] (1) Select and configure the environment.

[0102] Install necessary Python libraries (such as numpy, pandas, scikit-learn, tensorflow, etc.).

[0103] Configure a GPU environment (if available) to accelerate the training of deep learning models.

[0104] (2) Data division.

[0105] Divide historical data into training set, validation set and test set.

[0106] (3) Model training.

[0107] Use linear regression, logistic regression (although not commonly used for continuous value prediction, it can be used for classification problems), gradient boosting regression (such as XGBoost), Kalman filtering (for BP neural networks (for smooth prediction of time series), complex nonlinear relationship modeling), and combined prediction models for training.

[0108] Adjust hyperparameters and optimize model performance through methods such as cross-validation.

[0109] A weighted average combination model is used to add and combine the prediction results of different single prediction models according to certain weights. By assigning larger weights to models with higher prediction accuracy, the advantages of multiple models are combined through weighted averaging to improve prediction accuracy.

[0110] Assume that there are n single prediction models, and their prediction results are y1, y2, …, y n , the corresponding weights are ω1, ω2, …, ω n (the sum of the weights is 1), then the prediction result of the weighted average combination model can be expressed as:

[0111] Step S406, using the validation data set and prediction results to determine the weight of each single prediction model.

[0112] Use real-time data as a new validation set every day to evaluate the prediction accuracy of the current model.

[0113] Adjust model parameters or structure based on validation results. Try to integrate prediction results from different models and improve overall prediction performance through weighted averaging, stacking, and other methods.

[0114] Step S407, perform load forecasting and determine the final forecasting result.

[0115] (1) Model deployment.

[0116] Deploy the optimal prediction model to the online environment to ensure rapid response to prediction requests.

[0117] (2) Real-time prediction and adjustment.

[0118] At the beginning of each day, a forecast is made using the forecast model for that day combined with the forecast results from the previous day, real-time load data, and the latest weather and temperature information.

[0119] During the forecasting process, the forecast results are fine-tuned according to changes in real-time data, such as considering the immediate impact of sudden weather changes on load.

[0120] (3) Result output and monitoring.

[0121] The prediction results are displayed in a visual form to relevant personnel such as power dispatchers.

[0122] Monitor the operation status of the model, including indicators such as response time and prediction accuracy, to ensure stable operation of the system.

[0123] Figure 5 is a schematic diagram comparing actual load and predicted load curves of simulation data according to an optional embodiment of the present application, Figure 6 is a schematic diagram of the accuracy of the load forecast in the last seven days of a simulation data according to an optional embodiment of the present application, such as Figure 5 , Figure 6 It can be seen that (1) the forecast of renewable energy load involves a large amount of historical data and multiple influencing factors (such as meteorological conditions, socio-economic conditions, etc.). Machine learning algorithms can automatically learn and adapt to complex patterns in the data, accurately capture the laws of load changes through feature extraction and modeling, and thus improve the accuracy of forecasts. (2) Compared with traditional statistical methods, machine learning algorithms (such as neural networks, support vector machines, etc.) are better at dealing with nonlinear and time-dependent problems, and can more accurately establish relationship models between load and multiple influencing factors. (3) Machine learning forecasting algorithms can realize real-time data processing and forecasting, enabling energy suppliers to understand the supply of renewable energy in a timely manner and make corresponding scheduling and management decisions to improve the flexibility of the system. (4) Based on accurate load forecasting results, energy suppliers can accurately dispatch renewable energy equipment to ensure that energy utilization is maximized and energy waste is reduced while meeting demand. (5) Load forecasting technology also helps to formulate demand-based dynamic pricing plans. By predicting electricity demand in different time periods in the future, competitive and fair electricity prices can be formulated to optimize resource allocation and promote the healthy development of the energy market.

[0124] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0125] In this embodiment, a load data determination device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0126] Figure 7 is a structural block diagram of a device for determining load data according to an embodiment of the present application, such as Figure 5 As shown, the device comprises:

[0127] A training module 72 is used to train multiple single prediction models according to first load data, first meteorological data and first time data of the power generation system in a first time period, and determine a first prediction accuracy of each single prediction model after training, wherein each single prediction model is a model for predicting the load of the power generation system, and the first time period is a time period before the current time;

[0128] An input module 74 is used to input the second meteorological data and the second time data of the power generation system at the current time into each of the trained single prediction models to determine the second load data of the power generation system at the target time predicted by each of the trained single prediction models, and determine the weight corresponding to each of the single prediction models according to each first prediction accuracy, wherein the target time is the time after the current time;

[0129] The determination module 76 is used to determine the target load data of the power generation system at the target time according to the plurality of second load data and the weight corresponding to each single prediction model.

[0130] Through the above device, multiple single prediction models are trained according to the first load data, first meteorological data and first time data of the power generation system in the first time period before the current time, and the first prediction accuracy of each single prediction model after training is determined, wherein each single prediction model is a model for predicting the load of the power generation system; the second meteorological data and the second time data of the power generation system at the current time are input into each trained single prediction model to determine the second load data of the power generation system at the target time after the current time predicted by each trained single prediction model, and the corresponding weight of each single prediction model is determined according to each first prediction accuracy; the target load data of the power generation system at the target time is determined according to the weight corresponding to the multiple second load data and each single prediction model. That is to say, the embodiment of the present application uses historical data, including the first load data, first meteorological data and first time data of the power generation system in the first time period, to train multiple different single prediction models. Then, at the current time, the second meteorological data and the second time data are input into the multiple trained single prediction models to generate the second load data for the future target time. Through the embodiment of the present application, the problem of low prediction accuracy of load data prediction by combining historical data analysis and trend analysis in the related art can be solved. Furthermore, by integrating the advantages of multiple single forecasting models and dynamically adjusting the forecast based on relevant data, the accuracy and reliability of load forecasting are significantly improved.

[0131] In an exemplary embodiment, the training module 72 is also used to perform data cleaning on the first load data, the first meteorological data and the first time data to process abnormal values ​​and missing values ​​in the first load data, the first meteorological data and the first time data, and obtain the cleaned first load data, the cleaned first meteorological data and the cleaned first time data; standardize the cleaned first load data, the cleaned first meteorological data and the cleaned first time data to unify the data format of the cleaned first load data, the cleaned first meteorological data and the cleaned first time data; train the multiple single prediction models according to the standardized first load data, the standardized first meteorological data and the standardized first time data.

[0132] In an exemplary embodiment, the training module 72 is also used to identify missing values ​​in the first load data, the first meteorological data and the first time data through an objective function, and determine the proportion of the missing values ​​in the first load data, the first meteorological data and the first time data; when the proportion is less than a preset proportion, delete the rows corresponding to the missing values ​​and / or the columns corresponding to the missing values; fill the rows corresponding to the missing values ​​and / or the columns corresponding to the missing values ​​with data based on the target model; and, based on target rules, identify outliers in the first load data, the first meteorological data and the first time data; and replace the outliers according to the average value and / or median of the neighboring points.

[0133] In an exemplary embodiment, the training module 72 is also used to group the first load data after standardization, the first meteorological data after standardization and the first time data after standardization to obtain training set data and validation set data; train the multiple single prediction models with the training set data; input the meteorological data and time data in the validation set data into each single prediction model after training to obtain prediction results; compare each load data in the validation set data with each prediction result, and calculate the first prediction accuracy of each single prediction model after training based on the first comparison result.

[0134] In an exemplary embodiment, the input module 74 is also used to compare each of the first prediction accuracy with an accuracy threshold to determine a second comparison result; obtain a second prediction accuracy whose first prediction accuracy is greater than or equal to the accuracy threshold from multiple first prediction accuracies based on the second comparison result; and update the multiple single prediction models based on the second prediction accuracy, wherein the updated multiple single prediction models include: a single prediction model corresponding to each second prediction accuracy.

[0135] In an exemplary embodiment, the determination module 76 is further configured to perform weighted averaging on the plurality of second load data according to a first formula to obtain the target load data, wherein the first formula is: Wherein, Y is the target load data, which is i Includes: y1, y2, …, y n ,y i are the plurality of second load data, y1, y2, ..., y n For each second load data, ω i Including: ω1, ω2, …, ω n ,ω i are the weights corresponding to multiple single prediction models, ω1, ω2, …, ω nare the weights corresponding to each single prediction model.

[0136] In an exemplary embodiment, the determination module 76 is also used to update the first meteorological data, the first time data and the first load data, wherein the updated first meteorological data includes: the second meteorological data, the updated first time data includes: the second time data, and the updated first load data includes: the target load data; and the trained multiple single prediction models are updated according to the updated first meteorological data, the updated first time data and the updated first load data.

[0137] It should be noted that the above modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or the above modules are located in different processors in any combination.

[0138] An embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0139] Optionally, in this embodiment, the storage medium may be configured to store program codes for executing the following steps:

[0140] S1, training multiple single prediction models according to first load data, first meteorological data and first time data of the power generation system in a first time period, and determining a first prediction accuracy of each single prediction model after training, wherein each single prediction model is a model for predicting the load of the power generation system, and the first time period is a time period before the current time;

[0141] S2, inputting the second meteorological data and the second time data of the power generation system at the current time into each of the trained single prediction models to determine the second load data of the power generation system at the target time predicted by each of the trained single prediction models, and determining the weight corresponding to each of the single prediction models according to each first prediction accuracy, wherein the target time is the time after the current time;

[0142] S3, determining the target load data of the power generation system at the target time according to the plurality of second load data and the weight corresponding to each single prediction model.

[0143] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0144] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0145] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0146] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0147] S1, training multiple single prediction models according to first load data, first meteorological data and first time data of the power generation system in a first time period, and determining a first prediction accuracy of each single prediction model after training, wherein each single prediction model is a model for predicting the load of the power generation system, and the first time period is a time period before the current time;

[0148] S2, inputting the second meteorological data and the second time data of the power generation system at the current time into each of the trained single prediction models to determine the second load data of the power generation system at the target time predicted by each of the trained single prediction models, and determining the weight corresponding to each of the single prediction models according to each first prediction accuracy, wherein the target time is the time after the current time;

[0149] S3, determining the target load data of the power generation system at the target time according to the plurality of second load data and the weight corresponding to each single prediction model.

[0150] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any one of the above method embodiments are implemented.

[0151] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0152] An embodiment of the present application also provides a computer program, which includes computer instructions, which are stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in any one of the above method embodiments.

[0153] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0154] S1, training multiple single prediction models according to first load data, first meteorological data and first time data of the power generation system in a first time period, and determining a first prediction accuracy of each single prediction model after training, wherein each single prediction model is a model for predicting the load of the power generation system, and the first time period is a time period before the current time;

[0155] S2, inputting the second meteorological data and the second time data of the power generation system at the current time into each of the trained single prediction models to determine the second load data of the power generation system at the target time predicted by each of the trained single prediction models, and determining the weight corresponding to each of the single prediction models according to each first prediction accuracy, wherein the target time is the time after the current time;

[0156] S3, determining the target load data of the power generation system at the target time according to the plurality of second load data and the weight corresponding to each single prediction model.

[0157] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0158] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0159] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for determining load data, characterized in that: include: Training a plurality of single prediction models according to first load data, first meteorological data and first time data of the power generation system in a first time period, and determining a first prediction accuracy of each single prediction model after training, wherein each single prediction model is a model for predicting the load of the power generation system, and the first time period is a time period before the current time; Input the second meteorological data and the second time data of the power generation system at the current time into each of the trained single prediction models to determine the second load data of the power generation system at the target time predicted by each of the trained single prediction models, and determine the weight corresponding to each of the single prediction models according to each first prediction accuracy, wherein the target time is the time after the current time; The target load data of the power generation system at the target time is determined according to the plurality of second load data and the weight corresponding to each single prediction model.

2. The method according to claim 1, characterized in that Training a plurality of single prediction models according to first load data, first meteorological data and first time data of the power generation system in a first time period includes: Performing data cleaning on the first load data, the first meteorological data, and the first time data to process abnormal values ​​and missing values ​​in the first load data, the first meteorological data, and the first time data, and obtaining cleaned first load data, cleaned first meteorological data, and cleaned first time data; Standardizing the cleaned first load data, the cleaned first meteorological data, and the cleaned first time data to unify the data formats of the cleaned first load data, the cleaned first meteorological data, and the cleaned first time data; The multiple single prediction models are trained according to the standardized first load data, the standardized first meteorological data and the standardized first time data.

3. The method according to claim 2, characterized in that Performing data cleaning on the first load data, the first meteorological data, and the first time data to process abnormal values ​​and missing values ​​in the first load data, the first meteorological data, and the first time data includes: identifying missing values ​​in the first load data, the first meteorological data, and the first time data through an objective function, and determining a proportion of the missing values ​​in the first load data, the first meteorological data, and the first time data; When the ratio is less than a preset ratio, deleting the row corresponding to the missing value and / or the column corresponding to the missing value; Filling the rows corresponding to the missing values ​​and / or the columns corresponding to the missing values ​​with data based on the target model; and identifying abnormal values ​​in the first load data, the first meteorological data, and the first time data based on a target rule; The outliers are replaced according to the mean and / or median of the neighboring points.

4. The method according to claim 2, characterized in that: The plurality of single prediction models are trained according to the standardized first load data, the standardized first meteorological data, and the standardized first time data, and a first prediction accuracy of each trained single prediction model is determined, including: Grouping the first load data after the standardization processing, the first meteorological data after the standardization processing, and the first time data after the standardization processing to obtain training set data and verification set data; Using the training set data to train the multiple single prediction models; Inputting the meteorological data and time data in the validation set data into each single prediction model after training to obtain a prediction result; Each load data in the validation set data is compared with each prediction result, and a first prediction accuracy of each single prediction model after training is calculated according to the first comparison result.

5. The method according to claim 1, characterized in that Determining a weight corresponding to each single prediction model according to each first prediction accuracy includes: Comparing each of the first prediction accuracies with an accuracy threshold to determine a second comparison result; According to the second comparison result, obtaining a second prediction accuracy from a plurality of first prediction accuracies, the second prediction accuracy of which is greater than or equal to the accuracy threshold; The multiple single prediction models are updated according to the second prediction accuracy, wherein the updated multiple single prediction models include: a single prediction model corresponding to each second prediction accuracy.

6. The method according to claim 1, characterized in that Determining the target load data of the power generation system at the target time according to the plurality of second load data and the weight corresponding to each single prediction model includes: The target load data is obtained by weighted averaging the plurality of second load data according to a first formula, wherein the first formula is: Wherein, Y is the target load data, which is i Includes: y1,y2,...,y n ,y i are the plurality of second load data, y1, y2, ..., y n For each second load data, ω i Including: ω1, ω2, ..., ω n ,ω i are the weights corresponding to multiple single prediction models, ω1, ω2, …, ω n are the weights corresponding to each single prediction model.

7. The method according to claim 1, characterized in that After determining the target load data of the power generation system at the target time according to the plurality of second load data and the weight corresponding to each single prediction model, the method further includes: updating the first meteorological data, the first time data and the first load data, wherein the updated first meteorological data includes: the second meteorological data, the updated first time data includes: the second time data, and the updated first load data includes: the target load data; The trained multiple single prediction models are updated according to the updated first meteorological data, the updated first time data and the updated first load data.

8. A device for determining load data, characterized in that: include: a training module, configured to train a plurality of single prediction models according to first load data, first meteorological data and first time data of a power generation system in a first time period, and determine a first prediction accuracy of each single prediction model after training, wherein each single prediction model is a model for predicting the load of the power generation system, and the first time period is a time period before a current time; An input module, used for inputting the second meteorological data and the second time data of the power generation system at the current time into each of the trained single prediction models, so as to determine the second load data of the power generation system at the target time predicted by each of the trained single prediction models, and determining the weight corresponding to each of the single prediction models according to each first prediction accuracy, wherein the target time is the time after the current time; A determination module is used to determine the target load data of the power generation system at the target time according to the plurality of second load data and the weight corresponding to each single prediction model.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method of any one of claims 1 to 7 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 7 through the computer program.

Images