Data processing method, electronic equipment and computer readable storage medium

Abstract

The invention discloses a data processing method, electronic equipment and a computer readable storage medium, and relates to the technical field of computers and large models. The method comprises the following steps: acquiring target numerical weather forecast data, target historical power data and target installed capacity data in a target prediction area; a target renewable energy output prediction model is adopted to carry out output prediction on the target numerical weather forecast data, the regional renewable energy output potential is obtained, and the target renewable energy output prediction model is used for carrying out renewable energy output prediction on a designated region according to the given numerical weather forecast data; and generating target renewable energy output based on the regional renewable energy output potential, the target historical power data and the target installed capacity data. The technical problems of high model maintenance cost, inaccurate prediction result and long prediction time existing in renewable energy power generation power prediction in related technologies are solved.

Description

Technical Field

[0001] The present application relates to the fields of computer technology and large model technology, and in particular to a data processing method, an electronic device and a computer-readable storage medium. Background Art

[0002] With the improvement of global environmental awareness and the widespread application of renewable energy, renewable energy generation (such as photovoltaic power generation, wind power generation, hydropower generation, etc.) has become an important part of power supply. Renewable energy power generation power forecasting plays an important role in power dispatching, maintaining power balance of power grid and other scenarios.

[0003] At present, the renewable energy power generation prediction method generally adopts the method of modeling each centralized site separately, or the method of converting benchmark sites based on distributed renewable energy. However, the above methods have the defects of high model maintenance cost, inaccurate prediction results and long prediction time.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present application provide a data processing method, an electronic device, and a computer-readable storage medium to at least solve the technical problems in the related art of renewable energy power generation prediction, such as high model maintenance cost, inaccurate prediction results, and long prediction time.

[0006] According to one aspect of an embodiment of the present application, a data processing method is provided, including: obtaining target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area; using a target renewable energy output prediction model to perform output prediction on the target numerical weather forecast data to obtain regional renewable energy output potential, wherein the target renewable energy output prediction model is used to predict the renewable energy output of a designated area based on given numerical weather forecast data; and generating a target renewable energy output based on the regional renewable energy output potential, the target historical power data, and the target installed capacity data.

[0007] According to another aspect of an embodiment of the present application, a data processing method is also provided, including: obtaining target numerical weather forecast data, target historical power data and target installed capacity data within a target prediction area; using a target photovoltaic output prediction model to perform output prediction on the target numerical weather forecast data to obtain regional photovoltaic output potential, wherein the target photovoltaic output prediction model is used to predict the photovoltaic output of a designated area based on given numerical weather forecast data; and generating a target photovoltaic output based on the regional photovoltaic output potential, the target historical power data and the target installed capacity data.

[0008] According to another aspect of the embodiments of the present application, there is also provided a data processing method, including: obtaining a data processing request through a first application programming interface, where the request data carried in the data processing request includes: target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area; returning a data processing response through a second application programming interface, where the response data carried in the data processing response includes: target renewable energy output, and the target renewable energy output is generated according to the data processing method of any one of the above.

[0009] According to another aspect of the embodiments of the present application, there is also provided a data processing method, including: obtaining a current input data processing dialogue request, where the request data carried in the data processing dialogue request includes: target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area; in response to the data processing dialogue request, returning a data processing dialogue reply, where the information carried in the data processing dialogue reply includes: target renewable energy output, and the target renewable energy output is generated according to the data processing method of any one of the above; displaying the target renewable energy output within a graphical user interface.

[0010] According to another aspect of the embodiments of the present application, there is also provided a data processing method, including: in response to an input instruction acting on an operation interface, displaying target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area on the operation interface; in response to a processing instruction acting on the operation interface, displaying target renewable energy output on the operation interface; where the target renewable energy output is generated according to the data processing method of any one of the above.

[0011] According to another aspect of the embodiments of the present application, there is also provided a model training method, including: obtaining a training data set, where the training data set includes: unmasked sample numerical weather forecast data; training an initial renewable energy output prediction model using the training data set to generate a target renewable energy output prediction model; where the target renewable energy output prediction model is used to predict the output of target numerical weather forecast data within a target prediction area to obtain the regional renewable energy output potential.

[0012] According to another aspect of the embodiments of the present application, there is also provided a data processing system, including: a client for sending target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area; a server connected to the client for using a target renewable energy output prediction model to perform output prediction on the target numerical weather forecast data to obtain the regional renewable energy output potential, and generating target renewable energy output based on the regional renewable energy output potential, target historical power data, and target installed capacity data, wherein the target renewable energy output prediction model is used to predict the renewable energy output of a specified area according to given numerical weather forecast data; the client is further used to output the target renewable energy output.

[0013] According to another aspect of the embodiments of the present application, there is also provided an electronic device, including: a memory storing an executable program; a processor connected to the memory through a bus for running the program, wherein when the program runs, it executes any one of the above data processing methods or model training methods.

[0014] According to another aspect of the embodiments of the present application, there is also provided a computer-readable storage medium, the computer-readable storage medium including a stored executable program, wherein when the executable program runs, it controls the device where the computer-readable storage medium is located to execute any one of the above data processing methods or model training methods.

[0015] According to another aspect of the embodiments of the present application, there is also provided a computer program product, including a computer program, which when executed by a processor, implements any one of the above data processing methods or model training methods.

[0016] In the embodiments of the present application, by obtaining the target numerical weather forecast data, target historical power data, and target installed capacity data within the target prediction area, and then using the target renewable energy output prediction model to perform output prediction on the target numerical weather forecast data to obtain the regional renewable energy output potential, wherein the target renewable energy output prediction model is used to predict the renewable energy output of a specified area according to given numerical weather forecast data, and finally generating the target renewable energy output based on the regional renewable energy output potential, target historical power data, and target installed capacity data, the purpose of efficiently and accurately predicting the renewable energy output within the region is achieved, thereby realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the renewable energy power prediction has a high model maintenance cost, inaccurate prediction results, and long prediction time.

[0017] It is easy to note that the above general description and the following detailed description are only for exemplifying and explaining the present application, and do not constitute a limitation to the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are provided to further understand the present application and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings:

[0019] Figure 1 is a schematic diagram of an application scenario of a data processing method according to an embodiment of the present application;

[0020] Figure 2 is a flowchart of a data processing method according to an embodiment of the present application;

[0021] Figure 3 is a flowchart architecture diagram of two-stage training according to an embodiment of the present application;

[0022] Figure 4 is a schematic diagram of the pressure reconstruction effect according to an embodiment of the present application;

[0023] Figure 5 is a schematic diagram of the comparison of output potential according to an embodiment of the present application;

[0024] Figure 6 is a flowchart of a data processing method according to an embodiment of the present application;

[0025] Figure 7 is a flowchart of a data processing method according to an embodiment of the present application;

[0026] Figure 8 is a flowchart of a data processing method according to an embodiment of the present application;

[0027] Figure 9 is a flowchart of a data processing method according to an embodiment of the present application;

[0028] Figure 10 is a flowchart of a model training method according to an embodiment of the present application;

[0029] Figure 11 is a schematic diagram of the structure of a data processing system according to an embodiment of the present application;

[0030] Figure 12 is a schematic diagram of the structure of a data processing device according to an embodiment of the present application;

[0031] Figure 13 is a schematic diagram of the structure of another data processing device according to an embodiment of the present application;

[0032] Figure 14 is a schematic diagram of the structure of another data processing device according to an embodiment of the present application;

[0033] Figure 15 It is a schematic structural diagram of another data processing device according to an embodiment of the present application;

[0034] Figure 16 It is a schematic structural diagram of another data processing device according to an embodiment of the present application;

[0035] Figure 17 It is a schematic structural diagram of a model training device according to an embodiment of the present application;

[0036] Figure 18 It is a structural block diagram of a computing device according to an embodiment of the present application;

[0037] Figure 19 It is a structural block diagram of an electronic device according to an embodiment of the present application. Detailed implementation manners

[0038] In order to enable those skilled in the art of the present technology to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying 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 of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0039] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances so that the embodiments of the present application described here can be implemented in an order different from those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0040] The technical solution provided by this application is mainly implemented using large model technology. Here, a large model refers to a deep learning model with a large number of model parameters, usually including hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than one quadrillion model parameters. A large model can also be called a Foundation Model. Through pre-training of the large model with a large amount of unlabeled corpus, a pre-trained model with more than one hundred million parameters is produced. This model can adapt to a wide range of downstream tasks and has good generalization ability. For example, large language models (LLMs), multi-modal pre-training models, etc.

[0041] It should be noted that in actual applications, the pre-trained model can be fine-tuned with a small number of samples so that the large model can be applied to different tasks. For example, large models can be widely applied in the fields of natural language processing (NLP), computer vision, speech processing, etc. Specifically, they can be applied to tasks in the field of computer vision such as visual question answering (VQA), image captioning (IC), image generation, etc., and can also be widely applied to tasks in the field of natural language processing such as text-based sentiment classification, text summary generation, machine translation, etc. Therefore, the main application scenarios of large models include but are not limited to digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc. In the embodiments of this application, taking the output prediction of the target renewable energy output prediction model proposed by this application in the renewable energy power generation prediction scenario as an example for explanation.

[0042] First, some nouns or terms that appear in the process of describing the embodiments of this application are applicable to the following explanations:

[0043] Numerical Weather Prediction (NWP): It refers to using mathematical models and computer calculations to predict future weather conditions. This method, based on the current actual situation of the atmosphere, under certain initial values and boundary conditions, solves a series of partial differential equations through large-scale calculations to predict the atmospheric motion state and weather phenomena in a future period of time.

[0044] Masked Auto Encoder (MAE): A pre-training method adopted in deep learning. By randomly occluding or "masking" a part of the information in the input data and then training the model to reconstruct the entire data from the remaining unoccluded information, MAE enables the model to learn the intrinsic structure of the data, which is particularly effective for small-sample learning and feature extraction.

[0045] Vision Transformer (ViT): A relatively powerful image processing model. Instead of using the traditional Convolutional Neural Network (CNN), ViT divides the image into multiple fixed small patches, flattens them, adds positional encoding, and then feeds them into the self-attention Transformer for processing.

[0046] Temporal Convolutional Network (TCN): A neural network structure commonly used for time series prediction. By using one-dimensional convolutional layers and causal convolution to process sequence data, TCN can capture linear and non-linear patterns in time while maintaining flexibility with respect to the length of the input sequence.

[0047] Photovoltaic output: It refers to the electrical output generated by a photovoltaic power generation system, that is, the energy that can be converted into electrical energy by solar panels or photovoltaic power plants under specific time conditions. In the power system, accurately predicting the photovoltaic output is of great significance for power generation scheduling, balancing supply and demand, ensuring the stable operation of the power grid, and optimizing energy management.

[0048] With the improvement of global environmental awareness and the widespread application of renewable energy, the prediction of renewable energy power generation has an important role in scenarios such as power dispatch and maintaining the balance of power consumption in the power grid. Taking photovoltaic power generation as an example, photovoltaic power generation is generally divided into centralized photovoltaic power station generation and distributed photovoltaic power generation. The former is large-scale centralized photovoltaic power generation, and the latter is small-scale but widely distributed photovoltaic power generation, such as rooftop photovoltaic power generation, etc., covering areas from districts and counties to provinces, and the installed capacity is continuously increasing.

[0049] In recent years, the increasing photovoltaic installations have effectively reduced the use of fossil energy while bringing a huge impact on the safe operation of the power grid. Therefore, accurate prediction of photovoltaic power generation is crucial. For centralized photovoltaic power stations, a large number of station predictions are difficult to maintain. For the increasingly widespread distributed photovoltaic power generation, it is even more challenging because it is difficult to collect the actual power output of each household. At the same time, due to the continuous increase in photovoltaic installations, for the increasing number of new power stations and distributed photovoltaics, how to accurately predict in the absence of sufficient historical data is also a problem. On the other hand, from the business side, grid dispatching and power trading often pay more attention to the total photovoltaic power output rather than the prediction of individual power stations.

[0050] Currently, the commonly used algorithms for predicting photovoltaic power generation are statistical analysis methods and machine learning methods. For each centralized power station, a prediction model from the meteorological forecast of the power station to the power generation is established. For distributed photovoltaics, by selecting several benchmark power stations in the region, the predicted power is converted through the installed capacity ratio to obtain the predicted power of distributed photovoltaics.

[0051] The related technologies for predicting photovoltaic power generation based on statistical analysis methods and machine learning methods have the following defects.

[0052] Defect 1: High model maintenance cost. Usually, hundreds of models need to be maintained, and the prediction time is relatively long.

[0053] Defect 2: High data requirements. A good individual power station model requires the power station to have high-quality historical data. It is difficult to train an accurate model for some power stations with poor data quality or less historical data.

[0054] Defect 3: The prediction method for distributed photovoltaics is relatively rough. The meteorology within the region may be significantly different. For example, some areas are cloudy and some areas are sunny. Sometimes, the weather at several locations is difficult to depict the actual situation of the entire region.

[0055] In response to the above defects, no effective solution has been proposed before this application.

[0056] According to the embodiments of this application, a data processing method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0057] Considering that the number of model parameters of large models is huge and the computing resources of mobile terminals are limited, the above method provided by the embodiments of this application can be applied to Figure 1 the application scenarios shown, but not limited to this. InFigure 1 In the application scenario shown, the large model is deployed in server 10. Server 10 can be connected to one or more client devices 20 through a local area network connection, a wide area network connection, an Internet connection, or other types of data networks. Here, client devices 20 can include, but are not limited to: smartphones, tablets, laptops, palmtop computers, personal computers, smart home devices, in-vehicle devices, etc. Client devices 20 can interact with users through a graphical user interface to implement the invocation of the large model, thereby implementing the method provided in the embodiments of the present application.

[0058] In the embodiments of the present application, the system composed of the client device and the server can perform the following steps: The client device performs steps such as obtaining target numerical weather forecast data, target historical power data, and target installed capacity data within the target prediction area, and sending the target numerical weather forecast data, target historical power data, and target installed capacity data to the server. The server performs steps such as using the target renewable energy output prediction model to perform output prediction on the target numerical weather forecast data to obtain the regional renewable energy output potential. Here, the target renewable energy output prediction model is used to perform renewable energy output prediction on a specified area based on the given numerical weather forecast data; based on the regional renewable energy output potential, target historical power data, and target installed capacity data, generate the target renewable energy output, and return the target renewable energy output to the client device. It should be noted that in the case where the operating resources of the client device can meet the deployment and operating conditions of the large model, the embodiments of the present application can be performed in the client device.

[0059] It should be noted that with the rapid development of high-performance computing units, in other application scenarios, the above method provided in the embodiments of the present application can also be applied to a model all-in-one machine. In an alternative embodiment, multiple models are built into the model all-in-one machine. Users can select and adjust one model according to their needs to obtain their own model. Thus, the high-performance computing unit built into the model all-in-one machine can directly call the adjusted model to execute the above method provided in the embodiments of the present application. In another alternative embodiment, a trained model is built into the large model all-in-one machine. Thus, the high-performance computing unit built into the model all-in-one machine can directly call this model to execute the above method provided in the embodiments of the present application.

[0060] Further, when the user needs to train their own model, they can also upload their own dataset through the client. This dataset is sent from the client to the server, enabling the server to adjust the pre-trained model with this dataset to obtain the user's own model, which is then deployed to the production environment. To facilitate the user's model adjustment requirements, the server can provide complete adjustment tools, development frameworks, and processes, supporting multiple adjustment strategies, so that the adjusted model can better adapt to different field applications and achieve high customization.

[0061] Under the above operating environment, the present application provides a data processing method as Figure 2 shown. Figure 2 is a flowchart of a data processing method according to an embodiment of the present application. As Figure 2 shown, the method may include the following steps:

[0062] Step S21, obtain target numerical weather prediction data, target historical power data, and target installed capacity data within the target prediction area;

[0063] Step S22, use the target renewable energy output prediction model to perform output prediction on the target numerical weather prediction data to obtain the regional renewable energy output potential, where the target renewable energy output prediction model is used to predict the renewable energy output of a specified area based on the given numerical weather prediction data;

[0064] Step S23, generate the target renewable energy output based on the regional renewable energy output potential, target historical power data, and target installed capacity data.

[0065] In the embodiment of the present application, the target prediction area can be understood as the specific geographical area where renewable energy output prediction is to be performed. Exemplarily, it can be a certain province, a certain city, or a smaller range, which is not limited here.

[0066] The target numerical weather prediction data, that is, the target NWP data, can be understood as the result obtained by predicting the weather conditions within a certain future time range based on the current atmospheric conditions and historical meteorological data within the target prediction area. Exemplarily, the target numerical weather prediction data may include various meteorological elements, such as temperature, humidity, wind speed, wind direction, cloud cover, solar radiation, etc., which is not limited here.

[0067] The target historical power data can be understood as the power generation data of renewable energy power stations recorded in the target prediction area over a past period of time. Exemplarily, the target historical power data can be recorded as power output per hour, per day, or a longer period, which is not limited here.

[0068] The target installed capacity data can be understood as the total installed capacity of all renewable energy power generation facilities within the target prediction area, that is, the summary of the theoretically maximum power generation capacity.

[0069] Exemplarily, the above-mentioned target numerical weather prediction data, target historical power data, and target installed capacity data can be data after special data processing, such as data after grid processing, which is not limited here.

[0070] In this application, by obtaining three key types of input data within the target prediction area: target numerical weather prediction data, target historical power data, and target installed capacity data, a data basis is provided for subsequent model training and power generation prediction.

[0071] After obtaining the above-mentioned target numerical weather prediction data, target historical power data, and target installed capacity data, a target renewable energy output prediction model is used to predict the output of the target numerical weather prediction data to obtain the regional renewable energy output potential. Among them, the target renewable energy output prediction model is used to predict the renewable energy output of a specified area based on the given numerical weather prediction data. It can be understood that the target renewable energy output prediction model proposed in this application can predict the future power generation of renewable energy power stations in the specified area by analyzing the input NWP data. Exemplarily, taking renewable energy as photovoltaic as an example, the target renewable energy output prediction model proposed in this application can predict the regional photovoltaic output, which is not limited here.

[0072] It should be noted that the target renewable energy output prediction model in this application can uniformly predict the centralized photovoltaic power stations and distributed renewable energies within the specified area, and the predicted power of all power stations and distributed renewable energies within the specified area can be obtained through only one prediction process, with low maintenance cost, strong stability, and high prediction efficiency.

[0073] The target renewable energy output prediction model can be a large model or other deep learning models, which is not limited here.

[0074] The regional renewable energy output potential can be understood as the theoretical value of the maximum power generation that all renewable energy power stations within the target prediction area can generate, that is, assuming that all installed capacities can be fully utilized, without considering the influence of non-weather factors such as equipment failures and maintenance. It can be understood that this regional renewable energy output potential is an estimate of the future power generation capacity, and it still needs to be combined with the actual installed capacity to be converted into an actual power generation prediction.

[0075] In this application, a trained target renewable energy output prediction model is used to predict the output of target numerical weather prediction data, and the potential power generation capacity of the target prediction area under specific meteorological conditions is predicted to obtain the regional renewable energy output potential. Thus, the deep learning model can extract deeper features from NWP data, improving the accuracy of predicting the renewable energy power generation under future weather conditions.

[0076] After obtaining the regional renewable energy output potential, based on the regional renewable energy output potential, target historical power data, and target installed capacity data, the target renewable energy output is generated. Among them, the target renewable energy output can be understood as the specific value finally output by the target renewable energy output prediction model, and the target renewable energy output can be understood as the actual predicted power generation of the target prediction area after considering factors such as current or future weather conditions, installed capacity, and historical power data.

[0077] In this application, the obtained regional renewable energy output potential is used in combination with the target historical power data and the target installed capacity data to conduct actual power generation prediction. Referring to the historical power data can help adjust the prediction results to be closer to the actual power generation behavior. Finally, the target renewable energy output within the target prediction area is generated, that is, the electric energy that all renewable energy power generation facilities in the target prediction area are expected to generate under current or expected weather conditions.

[0078] It can be seen that the target renewable energy output prediction model in this application can output the renewable energy output potential of the entire region at one time, avoiding the need to train models separately for each station, simplifying the prediction process, reducing the complexity and cost of maintaining multiple independent models, and improving the prediction efficiency.

[0079] At the same time, the target renewable energy output prediction model in this application uses deep learning technology and can extract advanced features from numerical weather prediction data, including complex spatio-temporal relationships between meteorological elements, etc., enabling the model to better understand the impact of weather elements on renewable energy output, thus improving the accuracy of predicting renewable energy power generation.

[0080] Furthermore, in this application, the output potential is combined with the installed capacity data and historical power data, enabling the target renewable energy output prediction model to generate more accurate predicted renewable energy output, which helps the dispatching and balancing of the power system and ensures the stable operation of the power grid.

[0081] In addition, the target renewable energy output prediction model of this application not only focuses on individual stations but covers the entire region, including centralized photovoltaic stations and distributed photovoltaics. Thus, it can comprehensively evaluate the renewable energy power generation capacity of a region and has important value for power planning and market transactions.

[0082] The above data processing method provided by the embodiments of the present application can be but is not limited to being applied to application scenarios involving renewable energy output prediction in fields such as e-commerce services, education services, legal services, medical services, conference services, social network services, financial product services, logistics services, and navigation services, and is not limited here.

[0083] By adopting the embodiments of the present application, by obtaining target numerical weather forecast data, target historical power data, and target installed capacity data in a target prediction area, and then using a target renewable energy output prediction model to perform output prediction on the target numerical weather forecast data to obtain the regional renewable energy output potential, where the target renewable energy output prediction model is used to perform renewable energy output prediction on a specified area based on the given numerical weather forecast data, and finally, based on the regional renewable energy output potential, target historical power data, and target installed capacity data, target renewable energy output is generated, thereby achieving the purpose of efficiently and accurately predicting the renewable energy output in the area, thus realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the renewable energy power prediction has a high model maintenance cost, inaccurate prediction results, and long prediction time.

[0084] In an alternative embodiment, in step S21, obtaining the target numerical weather forecast data in the target prediction area includes the following method steps:

[0085] Step S211, obtaining multiple weather elements to be used in the target prediction area;

[0086] Step S212, performing numerical normalization processing on the multiple weather elements to obtain a first processing result;

[0087] Step S213, determining initial numerical weather forecast data based on the first processing result;

[0088] Step S214, performing grid processing on the initial numerical weather forecast data to obtain the target numerical weather forecast data.

[0089] In the embodiments of the present application, when obtaining the target numerical weather forecast data in the target prediction area, multiple weather elements to be used in the target prediction area can be obtained first, such as surface irradiance, diffuse irradiance, temperature, pressure, low-altitude cloud cover, etc., and are not limited here.

[0090] Then, numerical normalization is performed on multiple weather elements to obtain the first processing result. It can be understood that maximum-minimum normalization is performed on each weather element separately, that is, numerical data in different ranges or units is converted to a unified scale range, usually between 0 and 1. Converting the numerical values of various weather elements collected to the same proportional scale helps eliminate the dimension effect and makes different weather elements comparable in model training.

[0091] After that, initial numerical weather forecast data is determined based on the first processing result, that is, the numerical weather forecast data after normalization is obtained. Finally, grid processing is performed on the initial numerical weather forecast data to obtain the target numerical weather forecast data. Among them, grid processing is a preprocessing technology that divides the geographical space into a series of grid cells, and each grid cell contains specific weather data. This can provide information on spatial distribution, enabling the model to more precisely understand and predict the impact of local weather on renewable energy generation.

[0092] Exemplarily, the initial numerical weather forecast data can be divided into multiple grids. For example, the initial numerical weather forecast data is allocated to the corresponding grids according to specific geographical coordinates (such as a longitude-latitude grid of 0.1°×0.1°). The weather parameters within each grid are averaged or aggregated to form a spatially uniform dataset, thereby obtaining the target numerical weather forecast data. In this way, not only can the spatial characteristics of the weather data be retained, but it is also convenient for the deep learning model to extract spatial features, so as to more accurately predict the power generation of different geographical locations.

[0093] Taking photovoltaic power generation as an example, the initial numerical weather forecast data will be processed into grid data in tabular form. The dimension at each moment is C×H×W, where H corresponds to the longitude range in the numerical weather forecast, W corresponds to the latitude range in the numerical weather forecast, and C represents the number of weather elements used. It can be understood that meteorological characteristics closely related to photovoltaic output can be mainly used, including surface irradiance, diffuse irradiance, temperature, pressure, low-cloud amount, etc. Each meteorological element will be separately subjected to maximum-minimum normalization. The grid-processed numerical weather forecast data at each moment will be regarded as a picture, and multiple moments will be regarded as a multi-frame video and input into the model.

[0094] It can be seen that through normalization and grid processing, the consistency and standardization of the input data are ensured, which helps the model learn and predict, and improves the accuracy of prediction. Moreover, grid processing can provide more detailed geographical location information, enabling the model to distinguish weather conditions at different spatial positions, thereby achieving more accurate local power generation prediction. At the same time, the grid and normalized data format simplify the input of the model, improve the learning efficiency of the model, and also facilitate the model to process multi-source heterogeneous data, enhancing the prediction speed and the stability of the model.

[0095] In an optional embodiment, the data processing method further includes the following method steps:

[0096] Step S24, obtaining the initial historical power data and the initial installed capacity data within the target prediction area;

[0097] Step S25, performing grid processing on the initial historical power data to obtain the target historical power data, and performing grid processing on the initial installed capacity data to obtain the target installed capacity data.

[0098] In the embodiment of the present application, when obtaining the target historical power data and the target installed capacity data within the target prediction area, the initial historical power data and the initial installed capacity data within the target prediction area can be obtained first, and then grid processing is performed on the initial historical power data to obtain the target historical power data, and grid processing is performed on the initial installed capacity data to obtain the target installed capacity data.

[0099] It can be understood that the present application can also perform grid processing on the historical power data and the installed capacity data. Thus, the historical power and the installed capacity no longer exist in the form of a single station, but are integrated into grids representing specific geographical locations. Therefore, through grid processing, the data of different stations can be uniformly managed, simplifying the process of data maintenance and update. For newly added renewable energy facilities, only the installed capacity data needs to be updated within the corresponding grid, without the need for large-scale adjustment of the entire prediction system, achieving the technical effect of unified maintenance and update.

[0100] At the same time, the data after grid processing can provide more extensive spatial information, enabling the model to learn the power output rules of spatially similar regions during the training process, which is particularly important for regions lacking historical data or new stations. The model can learn through the data of adjacent regions, improving the small-sample learning ability.

[0101] In addition, the historical power data after grid processing can more accurately reflect the power output characteristics of local areas, which is particularly important for the prediction of distributed renewable energy. By combining the installed capacity data after gridification, the model can more precisely calculate the power generation potential of each grid area, thereby improving the overall prediction accuracy.

[0102] In an alternative embodiment, in step S23, based on the regional renewable energy output potential, the target historical power data, and the target installed capacity data, generating the target renewable energy output includes the following method steps:

[0103] Step S231: Obtain the output potential of multiple grids according to the target historical power data and the target installed capacity data.

[0104] Step S232: Generate the target renewable energy output based on the regional renewable energy output potential and the output potential of multiple grids, where the target renewable energy output includes: centralized renewable energy output and distributed renewable energy output.

[0105] In the embodiment of the present application, when generating the target renewable energy output based on the regional renewable energy output potential, the target historical power data, and the target installed capacity data, the output potential of multiple grids can be obtained first according to the target historical power data and the target installed capacity data. Among them, the output potential of a grid can be understood as the output per unit installed capacity of each grid, that is, the renewable energy power generation that can theoretically be generated in each grid area.

[0106] Then, based on the regional renewable energy output potential and the output potential of multiple grids, the target renewable energy output is generated. Among them, the target renewable energy output includes: centralized renewable energy output and distributed renewable energy output. It can be understood that the regional renewable energy output potential includes the centralized renewable energy output potential and the distributed renewable energy output potential, that is, it can be understood as multiplying the centralized renewable energy output potential and the distributed renewable energy output potential after gridification by the output potential of the corresponding unit grid respectively, so as to obtain the centralized renewable energy output and the distributed renewable energy output, that is, to obtain the target renewable energy output.

[0107] It can be understood that the output potential data of each grid is weighted and summarized, and the weight is determined by the installed capacity of the grid, and finally the renewable energy output prediction of the entire region is obtained. And the renewable energy output prediction of the entire region includes the estimation of centralized and distributed renewable energy output, so that the renewable energy power generation capacity of the entire region can be comprehensively reflected.

[0108] It can be seen that the present application can predict the output of centralized and distributed renewable energy simultaneously within a unified framework, simplifying the complexity of the prediction model. At the same time, there is no need to maintain multiple independent models, reducing the model maintenance cost and operation cost.

[0109] Meanwhile, based on grid processing, the model of the present application can capture the spatial differences and changes in weather conditions more delicately, which is particularly important for distributed renewable energy that is widely distributed and significantly affected by local weather. By combining the installed capacity and historical power data of each grid, the predicted output potential becomes more accurate.

[0110] In an alternative embodiment, in step S231, according to the target historical power data and the target installed capacity data, the output potential of multiple grids is obtained, including the following method steps:

[0111] Step S2311, according to the target historical power data and the target installed capacity data, calculate the sum of historical powers and the sum of installed capacities within the same grid for multiple grids respectively;

[0112] Step S2312, perform numerical normalization processing on the sum of historical powers to obtain a second processing result;

[0113] Step S2313, based on the second processing result and the sum of installed capacities, obtain the output potential of multiple grids.

[0114] In the embodiment of the present application, when obtaining the output potential of multiple grids according to the target historical power data and the target installed capacity data, the sum of historical powers and the sum of installed capacities within the same grid for multiple grids can be calculated respectively according to the target historical power data and the target installed capacity data. Among them, the sum of historical powers can be understood as the total sum of all historical power data points within the same grid after grid processing. The sum of installed capacities can be understood as the total sum of all installed capacity data points within the same grid after grid processing.

[0115] In the present application, by analyzing the target historical power data and the target installed capacity data, all records belonging to the same grid are summed up to obtain the sum of historical powers and the sum of installed capacities of the grid respectively, realizing data aggregation, that is, integrating multiple data points distributed within the grid into a single value for subsequent calculation of output potential.

[0116] Then, perform numerical normalization processing on the sum of historical powers to obtain a second processing result, which can convert the power data into relative values, eliminate the differences in absolute numerical magnitudes, make the historical power data of each grid comparable, and facilitate unified analysis and prediction by the model.

[0117] Finally, based on the second processing result and the installed capacity sum value, the output potential of multiple grids can be obtained. It can be understood that the normalized historical power sum value is combined with the installed capacity sum value to calculate the output potential of each grid. Exemplarily, the normalized historical power sum value can be divided by the installed capacity sum value to obtain the average power generation efficiency per unit installed capacity, and then this efficiency value is applied to the current installed capacity data to estimate the maximum possible power generation of each grid under the current weather conditions, which is not limited herein.

[0118] It can be seen that based on the grid output potential calculation method, the model can learn the correlation of power output between different grids, enhancing the model's ability to process new data. Especially for those grids without sufficient historical data, the model can make predictions based on the performance of adjacent or similar condition grids.

[0119] At the same time, the normalized historical power data can accurately reflect the past power generation efficiency of each grid, unaffected by the absolute value size, which helps the model to more accurately predict the current output potential of the grid.

[0120] In an alternative embodiment, the data processing method further includes the following method steps:

[0121] Step S26: Obtain a first training data set, where the first training data set includes: unmasked sample numerical weather forecast data;

[0122] Step S27: Train the initial renewable energy output prediction model using the first training data set to generate a target renewable energy output prediction model.

[0123] In the embodiment of the present application, a first training data set can be obtained, where the first training data set includes: unmasked sample numerical weather forecast data, that is, the first training data set can be understood as a data set for training the renewable energy output prediction model, containing a large number of multi-frame numerical weather forecasts that have not been randomly masked.

[0124] The unmasked sample numerical weather forecast data can be understood as complete numerical weather forecast data without any random masking process. These data contain information on all predicted weather elements and are the basic inputs for model training.

[0125] After obtaining the first training data set, training the initial renewable energy output prediction model using the first training data set to generate a target renewable energy output prediction model can be understood as training the initial renewable energy output prediction model through the first training data set, optimizing the model parameters, enabling it to more accurately predict the output of renewable energy, and thus obtaining the target renewable energy output prediction model.

[0126] It can be seen that by training on actual historical numerical weather forecast data, the model can learn the complex relationship between weather conditions and renewable energy output, improving the prediction accuracy.

[0127] In an alternative embodiment, the initial renewable energy output prediction model includes: a self-attention encoder and a convolutional network decoder. In step S27, the initial renewable energy output prediction model is trained using the first training dataset to generate the target renewable energy output prediction model, including the following method steps:

[0128] Step S271, using the self-attention encoder to perform feature encoding on the first training dataset to obtain a first encoding result;

[0129] Step S272, using the convolutional network decoder to perform feature decoding on the first encoding result to obtain the centralized predicted output potential and the distributed predicted output potential;

[0130] Step S273, calculating the first target loss based on the centralized predicted output potential, the centralized actual output potential, the distributed predicted output potential, and the distributed actual output potential;

[0131] Step S274, simultaneously updating the parameters of the self-attention encoder and the parameters of the convolutional network decoder according to the first target loss to generate the target renewable energy output prediction model.

[0132] In the embodiment of the present application, the initial renewable energy output prediction model includes: a self-attention encoder (Transformer encoder) and a convolutional network decoder. Among them, the self-attention encoder is used to process sequence data and extract features by calculating the correlation between elements in the sequence (attention mechanism). Specifically in the present application, the self-attention encoder is used to process multi-dimensional spatio-temporal data, such as numerical weather forecast data, which can capture the correlations in time series and space.

[0133] The convolutional network decoder is used to transform the feature representation output by the encoder into a specific prediction result. Specifically in the present application, the convolutional network decoder can process spatio-temporal features and transform them into renewable energy output predictions.

[0134] In the embodiment of the present application, when training the initial renewable energy output prediction model using the first training dataset to generate the target renewable energy output prediction model, the self-attention encoder can be first used to perform feature encoding on the first training dataset to obtain a first encoding result. Among them, the first encoding result can be understood as the intermediate representation obtained by the self-attention encoder after feature extraction from the first training dataset, and these intermediate representations contain the spatio-temporal features extracted from the numerical weather forecast data.

[0135] Exemplarily, after the first training dataset is input into the initial renewable energy output prediction model, the first training dataset is sliced into multiple spatio-temporal cubes in the time, H, and W dimensions, and the sliced cubes are directly used as the input of the self-attention encoder, that is, they are input into the self-attention decoder for encoding to obtain the first encoding result. Through the attention mechanism, the self-attention encoder automatically learns and captures the correlations between different time points and different geographical locations in the data, and transforms these complex correlations into a concentrated, multi-dimensional feature representation (i.e., the first encoding result), laying the foundation for subsequent prediction tasks.

[0136] Then, a convolutional network decoder is used to decode the features of the first encoding result to obtain the centralized predicted output potential and the distributed predicted output potential. It can be understood that the convolutional network decoder receives the first encoding result generated by the self-attention encoder, and through a series of decoding and feature recovery operations, finally obtains the centralized predicted output potential and the distributed predicted output potential. Exemplarily, the convolutional network decoder uses a dilated convolutional network to process each cube, thereby outputting the output potential of each grid.

[0137] After that, a first target loss is calculated based on the centralized predicted output potential, the centralized actual output potential, the distributed predicted output potential, and the distributed actual output potential. Among them, the first target loss is used to measure the gap between the output potential predicted by the model and the actual output potential. Exemplarily, the first target loss L fine_tuning can be calculated according to formula (1).

[0138]

[0139] where λ represents a hyperparameter, c represents a centralized station, d represents a distributed station, N represents the number of centralized stations, y represents the true power, represents the predicted power value, represents the centralized actual output potential, represents the centralized predicted output potential, y d represents the distributed actual output potential, represents the distributed predicted output potential.

[0140] Finally, based on the first target loss, the parameters of the self-attention encoder and the parameters of the convolutional network decoder are updated simultaneously to generate the target renewable energy output prediction model. It can be understood that the initial renewable energy output prediction model is fine-tuned based on the first target loss, and the parameters of the initial renewable energy output prediction model are adjusted to optimize the prediction performance of the model, so that the gap between the predicted output potential and the actual output potential is minimized, thereby generating the target renewable energy output prediction model.

[0141] In the embodiment of the present application, during the fine-tuning process, the structure of the self-attention encoder remains unchanged, and the parameters of the pre-trained self-attention encoder are used as its initial values and optimized and trained simultaneously with the parameters of the convolutional network decoder. It can be seen that during the model training process, the encoder parameters in the pre-training stage are used as initialization, which helps the model achieve good performance even on limited training data, that is, the small-sample learning ability, which is particularly important for new stations lacking historical data.

[0142] It can be seen that the parameter update during the training process of the present application ensures the stability of the model and the reliability of the prediction results. Even under complex and changeable weather conditions, the model can maintain a high prediction accuracy. At the same time, the model can handle the prediction of centralized and distributed renewable energy simultaneously, simplifying the prediction process, improving the prediction efficiency, and facilitating later maintenance and upgrade.

[0143] In an alternative embodiment, the data processing method further includes the following method steps:

[0144] Step S28, obtain a second training dataset, where the second training dataset includes: sample numerical weather forecast data randomly masked by a preset ratio;

[0145] Step S29, train the initial numerical weather forecast reconstruction model using the second training dataset to generate a target numerical weather forecast reconstruction model.

[0146] In the embodiment of the present application, a second training dataset can also be obtained, where the second training dataset includes: sample numerical weather forecast data randomly masked by a preset ratio. The second training dataset can be understood as the dataset in the pre-training stage and is used to train the numerical weather forecast reconstruction model. The samples in the second training dataset are usually randomly selected numerical weather forecast data from historical data, that is, numerical weather forecasts randomly masked by a high ratio.

[0147] It can be understood that in deep learning pre-training, random masking is a commonly used method to enhance data, that is, randomly hide some data points to force the model to learn the internal structure and pattern of the data. In the present application, some time series or spatial data points of the numerical weather forecast data in the second training dataset are randomly masked, and the masking ratio is preset, and it is used to train the model on how to predict the missing part according to the context information in the case of missing data.

[0148] After obtaining the second training dataset, the initial numerical weather prediction reconstruction model is trained using the second training dataset to generate a target numerical weather prediction reconstruction model. It can be understood that the initial numerical weather prediction reconstruction model is trained using the second training dataset to optimize the model parameters, thereby generating a target numerical weather prediction reconstruction model. This target numerical weather prediction reconstruction model can not only extract useful information from numerical weather prediction data but also make effective predictions in the case of missing data.

[0149] Exemplarily, during the training process, the model attempts to predict the masked data points based on the unmasked data points, that is, to complete a reconstruction task of numerical weather prediction data. This process is iterated repeatedly, and the model parameters are adjusted according to the accuracy of the model prediction and the value of the loss function after each training until the model can show a high reconstruction accuracy on randomly masked data of various proportions, and finally, the target numerical weather prediction reconstruction model is obtained, which is not limited here.

[0150] It can be seen that by training on randomly masked data, the model can learn the internal structure and rules of numerical weather prediction data, and can make accurate predictions and reconstructions even when facing new and unseen weather data, improving the generalization ability of the model. At the same time, the random masking training in the pre-training stage enhances the model's ability to process incomplete data, enabling the model to make reasonable predictions based on context information even when encountering missing data in practical applications.

[0151] In an optional embodiment, the initial renewable energy output prediction model includes: a self-attention encoder and a self-attention decoder. In step S29, training the initial numerical weather prediction reconstruction model using the second training dataset to generate a target numerical weather prediction reconstruction model includes the following method steps:

[0152] Step S291, performing feature encoding on the second training dataset using the self-attention encoder to obtain a second encoding result;

[0153] Step S292, performing feature decoding on the second encoding result and the masked position embedding using the self-attention decoder to obtain a numerical weather prediction reconstruction value;

[0154] Step S293, calculating a second target loss based on the numerical weather prediction reconstruction value and the numerical weather prediction true value;

[0155] Step S294, updating the parameters of the self-attention encoder and the parameters of the self-attention decoder simultaneously according to the second target loss to generate a target numerical weather prediction reconstruction model.

[0156] In the embodiments of the present application, the initial renewable energy output prediction model includes: a self-attention encoder and a self-attention decoder (Transformer decoder). When training the initial numerical weather prediction reconstruction model with the second training dataset to generate the target numerical weather prediction reconstruction model, the second training dataset can be first feature-encoded by the self-attention encoder to obtain a second encoding result. The second encoding result can be understood as the feature representation obtained after the self-attention encoder encodes the second training dataset, which contains the abstract understanding and potential associations of the masked weather forecast data.

[0157] Exemplarily, after the second training dataset is input into the initial numerical weather prediction reconstruction model, the second training dataset is sliced into multiple spatio-temporal cubes in the time, H, and W dimensions. The sliced cubes are flattened and randomly masked, and the proportion of the masked cubes can be 75% to 95%, which is not limited here. The masked cubes are input into the self-attention decoder for encoding to obtain the second encoding result.

[0158] It can be seen that since the data in the dataset has been randomly masked, this process is actually teaching the encoder how to extract useful spatio-temporal information from the limited unmasked data to prepare for the subsequent reconstruction task.

[0159] Then, the self-attention decoder performs feature decoding on the second encoding result and the masked position embedding to obtain the numerical weather prediction reconstruction value. The masked position embedding can be understood as a special encoding used to mark the positions of the masked data in the Masked AutoEncoder (MAE) architecture. The masked position embedding is used to help the decoder know which positions of the data need to be predicted and the relative positions of these data in the original sequence.

[0160] It can be understood that the output of the self-attention encoder (i.e., the second encoding result) and the masked position embedding are used as inputs and input into the decoder to realize the reconstruction of the numerical weather forecast through the decoder. The self-attention decoder attempts to recover the masked weather forecast data based on the second encoding result and the masked position embedding. The self-attention decoder uses its own attention mechanism to identify and utilize the unmasked data points to predict the masked data points, aiming to learn how to reconstruct the numerical weather forecast in the case of partial information loss, thereby improving the model's processing ability for incomplete data.

[0161] After that, the second target loss is calculated based on the numerical weather prediction reconstruction value and the numerical weather prediction true value. The second target loss can be understood as an index for measuring the error between the predicted value and the true value of the reconstructed numerical weather forecast, and is used to guide the parameter optimization process of the model. Exemplarily, the second target loss Lpre-training It can be calculated according to formula (2).

[0162]

[0163] Among them, Ω represents the set of masked cubic blocks, and x i represents the true value of numerical weather prediction, represents the reconstructed value of numerical weather prediction.

[0164] Finally, based on the second objective loss, the parameters of the self-attention encoder and the parameters of the self-attention decoder are updated simultaneously to generate the target numerical weather prediction reconstruction model. It can be understood that, according to the second objective loss, the parameters of the self-attention encoder and the self-attention decoder are adjusted simultaneously. Exemplarily, the adjustment process can be implemented through an optimization algorithm, which is not limited here. The goal is to enable the model to reconstruct the masked numerical weather prediction data more accurately. After multiple rounds of training and parameter updates, the model is gradually optimized, and the finally generated target numerical weather prediction reconstruction model can effectively recover complete information from limited and partially missing weather data.

[0165] In the embodiment of the present application, during the fine-tuning process, the structure of the self-attention encoder remains unchanged, and the pre-trained self-attention encoder parameters will be used as its initial values and optimized and trained simultaneously with the parameters of the self-attention decoder. It can be seen that during the model training process, the encoder parameters in the pre-training stage are used as initialization, which helps the model to achieve better performance on limited training data, that is, the small-sample learning ability, which is particularly important for new stations lacking historical data.

[0166] It can be seen that in the pre-training stage, by randomly masking numerical weather prediction data, the model has learned how to use the unmasked information to infer the missing information, which is particularly useful for dealing with possible data missing situations in practice. And through the reconstruction task, the self-attention encoder can better learn the spatio-temporal correlation and implicit representation of numerical weather prediction.

[0167] In an optional embodiment, the data processing method further includes the following method steps:

[0168] Summarize the target renewable energy output generated within the target prediction area and the renewable energy outputs generated within multiple areas adjacent to the target prediction area to obtain the total renewable energy output within the preset geographical level range.

[0169] In the embodiment of the present application, after the prediction of the renewable energy output within the target prediction area is completed, the prediction results of the target prediction area and its adjacent multiple areas can be summarized to obtain the total renewable energy output within the preset geographical level range.

[0170] Among them, the preset geographical hierarchical range can be understood as the geographical range for summarizing the output of renewable energy. Exemplarily, it can be the grid coverage area at the municipal level, provincial level or a larger range. The selection of the hierarchical range is based on the predicted business requirements and the needs of grid dispatching, which are not limited here.

[0171] It can be seen that by summarizing the renewable energy output of the target prediction area and adjacent areas, an overview of energy output within a larger geographical range can be provided, thereby realizing an integrated prediction scheme for centralized and distributed photovoltaic power.

[0172] Exemplarily, taking photovoltaic power generation as an example, the solution of this application is mainly divided into three major modules:

[0173] Module 1, data processing module: perform anomaly detection and missing data completion on historical power data, splice historical numerical weather forecasts in the time dimension, and process refined installed capacity data into a grid form.

[0174] Furthermore, during the data processing, numerical weather forecasts, power, and installed capacity will all be processed into grid data such as 0.1°×0.1° and input into the model. First, collect and process grid numerical weather forecasts that can cover the target prediction area. The numerical weather forecasts will be processed from tabular data into grid data. The dimension of each moment is C×H×W, where H and W respectively correspond to the longitude and latitude ranges in the numerical weather forecast, and C represents the number of weather elements used. This application mainly uses meteorological features closely related to photovoltaic output, including surface irradiance, diffuse irradiance, temperature, pressure, low-altitude cloud cover, etc. Each meteorological element will be normalized by maximum-minimum normalization respectively. The grid numerical weather forecast data at each moment will be regarded as a picture, and multiple moments will be regarded as a multi-frame video and input into the model.

[0175] At the same time, historical power and installed capacity data are also gridded. The historical power and installed capacity within the same grid will be added, and the power will be normalized from 0 to 1 to obtain the output per unit installed capacity of each grid, that is, the output potential of the grid.

[0176] Module 2, model training module: The solution of this application is trained in two stages. The first stage is the pre-training stage, which mainly completes the reconstruction of numerical weather forecasts. The second stage is to fine-tune based on the encoder in the first stage, and a decoder based on a convolutional network is used to realize the prediction of regional photovoltaic output.

[0177] Furthermore, Figure 3 is the flow architecture diagram of the two-stage training according to the embodiment of this application, as Figure 3As shown in the figure, model training can adopt a two-stage training method based on the MAE architecture. The first-stage pre-training mainly trains the reconstruction task of numerical weather prediction (raw numerical weather prediction). The input is the numerical weather prediction after multi-frame image cutting and a high proportion of random masking, and the output is the restored complete weather forecast (reconstructed numerical weather prediction).

[0178] The MAE structure adopts an asymmetric encoder and decoder, and its backbone is Transformer. Specifically, the model input is multi-frame numerical weather prediction images (raw numerical weather prediction). The input multi-frame video will be sliced into multiple spatio-temporal cubes in the time, H, and W dimensions. The sliced cubes will be flattened and randomly masked. The proportion of masked cubes is, for example, 75% - 95%, which is not limited here. The masked cubes will be input into the encoder with Vision Transformer (ViT) as the backbone for encoding. The output of the encoder will be used as the input together with the masked position embedding and input into the decoder. Through the decoder, the reconstruction of numerical weather prediction is realized to obtain the reconstructed numerical weather prediction.

[0179] The loss function for the first-stage training is where Ω represents the set of masked cubes, x i and are the true value and the reconstructed value respectively. Through the reconstruction task, the encoder can better learn the spatio-temporal correlation and implicit representation of numerical weather prediction.

[0180] Exemplarily, Figure 4 is a schematic diagram of the pressure reconstruction effect according to an embodiment of the present application, that is, a schematic diagram of the pressure reconstruction effect of the target prediction area. As Figure 4 shown, Figure 4 in figure (a) is the numerical weather prediction, Figure 4 in figure (b) is the schematic diagram of the reconstruction effect diagram.

[0181] The second-stage training is mainly to fine-tune for a specific task dataset. The input in the second-stage fine-tuning stage is the same as that in the first-stage pre-training stage, both are multi-frame numerical weather predictions (raw numerical weather predictions), and the target true value is the grid-connected output potential (regional photovoltaic output potential).

[0182] It can be seen that the two-stage training is different from the one-stage training. In the two-stage training, the encoded cubes are no longer masked, and the cut cubes will be directly used as the input of the encoder. In terms of the model structure, the main difference between the two-stage and one-stage lies in the decoder part. The decoder in the two-stage uses a dilated convolutional network (convolutional network decoder) to process each cube, and the final output is the output potential of each grid (regional PV output potential). In the two-stage fine-tuning task, the structure of the encoder remains unchanged, and the pre-trained encoder parameters will be used as its initial values and optimized simultaneously with the decoder parameters.

[0183] The loss function of the two-stage training is where λ is a hyperparameter, c and d represent centralized and distributed power stations respectively, N represents the number of centralized power stations, and y and represent the true power and predicted power values respectively.

[0184] Module 3: Post-processing module for model output values, i.e., the aggregation and output module of regional output potential: The model output is the grid-based output potential, which needs to be multiplied by the installed capacity corresponding to the grid to obtain the predicted output of each grid. In this application, the processed grid-based centralized PV installed capacity potential and distributed PV installed capacity potential in Module 1 are multiplied by the output of the unit grid obtained in Module 2 to obtain the centralized PV output and distributed PV output respectively. At the same time, this application can also aggregate the grids covered by the target area to obtain the sum of the predicted power of the cities and the sum of the predicted power of the whole province, which is not limited here. Thus, an integrated prediction scheme for centralized PV and distributed PV is realized.

[0185] Exemplarily, Figure 5 is a schematic diagram of output potential comparison according to an embodiment of the present application. As Figure 5 shown, Figure 5 in Figure (a), it is the true grid-based centralized power station power output potential value, Figure 5 and in Figure (b), it is the predicted centralized and distributed grid output potential distribution.

[0186] It can be seen that this application proposes a regional renewable energy power generation prediction algorithm, that is, an end-to-end regional renewable energy power generation prediction method based on MAE pre-training. By pre-training on weather forecasts, the data processing ability is enhanced, and spatio-temporal information is extracted using multi-layer multi-dimensional convolution to achieve unified prediction of all centralized and distributed power stations within a region (such as a city or a province). While ensuring the accuracy, the running time is greatly shortened. That is, this application can achieve high prediction accuracy and efficiency, strong stability in actual implementation scenarios, and can flexibly handle the situation of changes in installed capacity, effectively saving the model maintenance cost and demonstrating excellent prediction performance.

[0187] This application makes full use of the available data in the region, uses a single model to predict both centralized and distributed power stations, has low maintenance costs, high prediction efficiency, strong small-sample learning ability for power stations with less historical data, and the model extracts features from the grid-based numerical weather forecasts of the entire region to obtain small-range predictions for each grid, avoiding the weakness of overgeneralization in traditional methods for distributed power station prediction. At the same time, this application extracts better spatio-temporal representations through pre-training and constructing a deep spatio-temporal convolutional network for multi-frame meteorology, achieving more accurate predictions than traditional methods.

[0188] Taking photovoltaic power generation as an example, this application has established a unified model for predicting centralized and distributed photovoltaic power stations. Through only one prediction process, the power predictions of all power stations and distributed photovoltaics in the region can be obtained, which is easy to promote, easy to maintain, has strong stability, and has small-sample learning ability for newly built power stations.

[0189] At the same time, this application uses a two-stage training method. First, it pre-trains multi-frame NWP images using MAE to obtain better representations of weather data, and then constructs a prediction model from meteorological forecasts to power through a multi-dimensional spatio-temporal convolutional network, with high prediction accuracy.

[0190] It is easy to understand that the beneficial effects of the data processing method provided by this application include the following points.

[0191] Beneficial effect (1), this application can be quickly scaled up and promoted compared with traditional methods. Taking photovoltaic power generation as an example, as the number of solar power stations and households installing rooftop solar energy continues to increase, it becomes impractical to maintain thousands of prediction models, and it is difficult for traditional methods to quickly adapt to newly added grid connections. However, in the method of this application, the algorithm output result is the power generation potential per unit area. As long as the newly installed capacity is obtained, the prediction result can be quickly adapted. Therefore, for newly built centralized photovoltaic power stations, only a small amount of historical data needs to be fine-tuned to obtain relatively accurate prediction results.

[0192] Beneficial effect (2), this application has lower maintenance costs compared with traditional methods. Taking photovoltaic power generation as an example, this application can simultaneously predict centralized and distributed photovoltaics with higher accuracy. Most of the algorithms of traditional methods first model centralized photovoltaic power stations and then convert the results of centralized photovoltaic power stations. The method of this application can simultaneously predict centralized and distributed photovoltaics through a complete set of solutions.

[0193] Beneficial effects (3). Taking photovoltaic power generation as an example, the model of the present application can better model the regional meteorology and photovoltaic power output. Algorithms such as the conversion of centralized photovoltaic power stations or the averaging of regional meteorology in traditional methods cannot well model the details of regional photovoltaic power output. The method of the present application regards the multi-moment NWP meteorology as multi-frame meteorological maps as input, borrows the method of video processing to extract the spatio-temporal correlation between NWPs, obtains a better spatio-temporal representation through the MAE structure, and uses a deep network for spatio-temporal data mining to achieve more accurate prediction.

[0194] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions, and corresponding operation entrances are provided for users to choose to authorize or refuse.

[0195] In addition, it should also be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0196] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc), and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present application.

[0197] According to an embodiment of the present application, there is also provided a Figure 6 data processing method as shown. Figure 6 is a flowchart of a data processing method according to an embodiment of the present application, as shown in Figure 6 The method includes:

[0198] Step S61, obtaining target numerical weather prediction data, target historical power data, and target installed capacity data within a target prediction region;

[0199] Step S62: Use the target photovoltaic output prediction model to perform output prediction on the target numerical weather forecast data to obtain the regional photovoltaic output potential. Herein, the target photovoltaic output prediction model is used to perform photovoltaic output prediction on a specified region based on the given numerical weather forecast data.

[0200] Step S63: Generate the target photovoltaic output based on the regional photovoltaic output potential, the target historical power data, and the target installed capacity data.

[0201] In the embodiments of the present application, the renewable energy may be photovoltaic, that is, the data processing method proposed in the present application can be applied to the scenario of predicting photovoltaic output, which is not limited herein.

[0202] For specific descriptions, reference may be made to the descriptions of the foregoing embodiments, which will not be elaborated herein.

[0203] The above data processing method provided by the embodiments of the present application can be, but is not limited to, applied to application scenarios involving renewable energy output prediction in fields such as e-commerce services, education services, legal services, medical services, conference services, social network services, financial product services, logistics services, and navigation services, which is not limited herein.

[0204] By adopting the embodiments of the present application, by obtaining the target numerical weather forecast data, the target historical power data, and the target installed capacity data in the target prediction region, and then using the target photovoltaic output prediction model to perform output prediction on the target numerical weather forecast data to obtain the regional photovoltaic output potential, wherein the target photovoltaic output prediction model is used to perform photovoltaic output prediction on a specified region based on the given numerical weather forecast data, and finally generating the target photovoltaic output based on the regional photovoltaic output potential, the target historical power data, and the target installed capacity data, the purpose of efficiently and accurately predicting the renewable energy output in the region is achieved, thereby realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the renewable energy power prediction has a high model maintenance cost, inaccurate prediction results, and long prediction time.

[0205] It should be noted that the preferred implementation manners of this embodiment can be referred to the relevant descriptions in the embodiment, which will not be elaborated herein.

[0206] According to the embodiments of the present application, there is also provided a Figure 7 data processing method as Figure 7 shown, which is a flowchart of a data processing method according to the embodiments of the present application. As Figure 7 shown, the method includes:

[0207] Step S71, obtain a data processing request through a first application programming interface. The request data carried in the data processing request includes: target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area.

[0208] Step S72, return a data processing response through a second application programming interface. The response data carried in the data processing response includes: target renewable energy output, and the target renewable energy output is generated according to the data processing method of any one of the above.

[0209] The above first application programming interface and the second application programming interface can be either the same application programming interface or different application programming interfaces. In an optional embodiment, the interface parameters in the above first application programming interface and the second application programming interface may include but are not limited to: interface global identifier, interface signature key, interface timestamp, interface request identifier, system call credential identifier, etc. The above first application programming interface can use GET or POST as the interface request method to obtain the file processing request. The above second application programming interface can use the JSON format to feedback the file processing response.

[0210] In the embodiment of the present application, the first application programming interface (API) can be understood as a specific service interface for receiving data processing requests sent by an external system. In the present application, the first API is specifically used to receive data processing requests from a client or a power dispatching system, that is, renewable energy output prediction requests.

[0211] The data processing request can be understood as a request sent by a client or an external system to the renewable energy output prediction system, asking the system to process according to the provided input data and return a prediction result. The request data includes target numerical weather forecast data, target historical power data, and target installed capacity data.

[0212] The second application programming interface can be understood as a service interface for returning a processing result to the requester. In the present application, the second API will return the target renewable energy output prediction generated based on the request data in the first step.

[0213] The data processing response can be understood as a response returned by the system to the requester after processing the data processing request. In this scenario, the response data mainly includes the renewable energy output prediction result of the target area.

[0214] For specific descriptions, reference can be made to the description of the foregoing embodiments, and details are not elaborated here.

[0215] The above data processing method provided by the embodiments of the present application can be, but is not limited to, applied to application scenarios involving renewable energy output prediction in fields such as e-commerce services, education services, legal services, medical services, conference services, social network services, financial product services, logistics services, and navigation services, and is not limited herein.

[0216] By using the embodiments of the present application, a data processing request is obtained through a first application programming interface. Among them, the request data carried in the data processing request includes: target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area. Then, a data processing response is returned through a second application programming interface. Among them, the response data carried in the data processing response includes: target renewable energy output, and the target renewable energy output is generated according to any one of the above data processing methods. Thus, the purpose of efficiently and accurately predicting the renewable energy output within the area is achieved, thereby realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the renewable energy power prediction has a high model maintenance cost, inaccurate prediction results, and long prediction time.

[0217] It should be noted that the preferred implementation manners of this embodiment can be referred to the relevant descriptions in the embodiments, and will not be elaborated herein.

[0218] According to the embodiments of the present application, there is also provided a Figure 8 data processing method as shown in Figure 8 which is a flowchart of a data processing method according to the embodiments of the present application. As shown in Figure 8 the method includes:

[0219] Step S81, obtain the current input data processing dialogue request. Among them, the request data carried in the data processing dialogue request includes: target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area;

[0220] Step S82, in response to the data processing dialogue request, return a data processing dialogue reply. Among them, the information carried in the data processing dialogue reply includes: target renewable energy output, and the target renewable energy output is generated according to any one of the above data processing methods;

[0221] Step S83, display the target renewable energy output in the graphical user interface.

[0222] In the embodiments of the present application, the data processing dialogue request can be understood as a specific request instruction sent by a user or a system to an intelligent device, aiming to obtain or process specific types of data. In the present application, this request instruction is for obtaining the renewable energy output prediction data within a target prediction area.

[0223] A data processing dialogue reply can be understood as a response of an intelligent device to a data processing dialogue request, which contains processed or predicted data results. In this application, the reply will carry the predicted target renewable energy output data.

[0224] A graphical user interface can be understood as a way for users to interact with a computer system, using graphical elements (such as windows, buttons, text boxes, etc.) to display information and receive user commands. In this application, the graphical user interface is used to display the renewable energy output prediction results.

[0225] For specific descriptions, reference can be made to the descriptions of the foregoing embodiments, and details will not be elaborated here.

[0226] The above data processing method provided by the embodiments of this application can be, but is not limited to, applied to application scenarios involving renewable energy output prediction in fields such as e-commerce services, education services, legal services, medical services, conference services, social network services, financial product services, logistics services, and navigation services. There is no limitation here.

[0227] By adopting the embodiments of this application, by obtaining the current input data processing dialogue request, where the request data carried in the data processing dialogue request includes: target numerical weather forecast data, target historical power data, and target installed capacity data within the target prediction area, and then in response to the data processing dialogue request, returning a data processing dialogue reply, where the information carried in the data processing dialogue reply includes: target renewable energy output, and the target renewable energy output is generated according to any one of the above data processing methods, and finally displaying the target renewable energy output in the graphical user interface, thereby achieving the purpose of efficiently and accurately predicting the renewable energy output within the area, thus realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the renewable energy power prediction has a high model maintenance cost, inaccurate prediction results, and long prediction time.

[0228] It should be noted that the preferred implementation manners of this embodiment can be referred to the relevant descriptions in the embodiment, and details will not be repeated here.

[0229] According to the embodiments of this application, there is also provided a data processing method as Figure 9 shown, Figure 9 which is a flowchart of a data processing method according to the embodiments of this application, as Figure 9 shown, and the method includes:

[0230] Step S91, in response to an input instruction acting on the operation interface, display the target numerical weather forecast data, target historical power data, and target installed capacity data within the target prediction area on the operation interface;

[0231] Step S92, in response to a processing instruction on the operation interface, display the target renewable energy output on the operation interface; wherein, the target renewable energy output is generated according to the data processing method of any one of the above.

[0232] In the embodiments of the present application, the operation interface can be understood as the interface for the user to interact with the software system, which can be the interface of a desktop application, a mobile application or a web application, allowing the user to issue instructions and view the system response through clicking, touching or other interaction methods.

[0233] The input instruction can be understood as a command sent by the user to the operation interface to view or operate specific data.

[0234] The processing instruction can be understood as an instruction issued by the user through the operation interface, aiming to start or trigger a specific data processing operation, such as the prediction of renewable energy output.

[0235] For specific descriptions, reference can be made to the descriptions of the foregoing embodiments, and details are not elaborated herein.

[0236] The data processing method provided by the embodiments of the present application can be but is not limited to being applied to application scenarios involving the prediction of renewable energy output in fields such as e-commerce services, education services, legal services, medical services, conference services, social network services, financial product services, logistics services and navigation services, and is not limited herein.

[0237] By adopting the embodiments of the present application, in response to the input instruction on the operation interface, display the target numerical weather forecast data, target historical power data and target installed capacity data in the target prediction area on the operation interface, and then in response to the processing instruction on the operation interface, display the target renewable energy output on the operation interface; wherein, the target renewable energy output is generated according to the data processing method of any one of the above, thereby achieving the purpose of efficiently and accurately predicting the renewable energy output in the region, thus realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems of high model maintenance cost, inaccurate prediction results and long prediction time in the prediction of renewable energy power generation in the related art.

[0238] It should be noted that the preferred implementation manners of this embodiment can be referred to the relevant descriptions in the embodiment, and details are not elaborated herein.

[0239] According to the embodiments of the present application, there is also provided a Figure 10 model training method as shown in Figure 10 is a flowchart of a model training method according to the embodiments of the present application, as shown in Figure 10 shown, the method includes:

[0240] Step S101: Obtain a training data set, where the training data set includes: unmasked sample numerical weather forecast data;

[0241] Step S102: Use the training data set to train an initial renewable energy output prediction model to generate a target renewable energy output prediction model; where the target renewable energy output prediction model is used to perform output prediction on the target numerical weather forecast data in the target prediction area to obtain the regional renewable energy output potential.

[0242] In the embodiments of the present application, the training data set can be understood as a data set used to train the renewable energy output prediction model, which includes unmasked sample numerical weather forecast data.

[0243] In the present application, by using the obtained training data set, the initial renewable energy output prediction model is trained. The model parameters are adjusted so that the model can better predict the output potential from the weather forecast data. Finally, a target renewable energy output prediction model is trained. The target renewable energy output prediction model can perform output prediction on the numerical weather forecast data in the target prediction area to obtain the renewable energy output potential of the area.

[0244] For specific descriptions, reference can be made to the descriptions of the foregoing embodiments, and details are not elaborated herein.

[0245] The above model training method provided by the embodiments of the present application can be but is not limited to being applied to application scenarios involving renewable energy output prediction in fields such as e-commerce services, education services, legal services, medical services, conference services, social network services, financial product services, logistics services, and navigation services. There is no limitation here.

[0246] By adopting the embodiments of the present application, a training data set is obtained, where the training data set includes: unmasked sample numerical weather forecast data, and then the training data set is used to train an initial renewable energy output prediction model to generate a target renewable energy output prediction model; where the target renewable energy output prediction model is used to perform output prediction on the target numerical weather forecast data in the target prediction area to obtain the regional renewable energy output potential. Thus, the purpose of efficiently and accurately predicting the renewable energy output in the region is achieved, thereby realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the renewable energy power prediction has a high model maintenance cost, inaccurate prediction results, and long prediction time.

[0247] It should be noted that the preferred implementation manners of this embodiment can be referred to the relevant descriptions in the embodiment, and details are not elaborated herein.

[0248] According to an embodiment of the present application, a data processing system is further provided. Figure 11 FIG. Figure 11 is a schematic structural diagram of a data processing system according to an embodiment of the present application. As Figure 11 shown, the system includes:

[0249] A client, configured to send target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area;

[0250] A server, connected to the client, configured to perform output prediction on the target numerical weather forecast data by using a target renewable energy output prediction model to obtain regional renewable energy output potential, and generate target renewable energy output based on the regional renewable energy output potential, the target historical power data, and the target installed capacity data, where the target renewable energy output prediction model is used to perform renewable energy output prediction on a specified area according to given numerical weather forecast data;

[0251] The client is further configured to output the target renewable energy output.

[0252] In the embodiment of the present application, the data processing system can be used to execute the data processing method proposed in the foregoing embodiment. For specific descriptions, reference can be made to the descriptions of the foregoing embodiment, and details are not described herein again.

[0253] The above data processing system provided by the embodiment of the present application can be applied to, but is not limited to, application scenarios involving renewable energy output prediction in fields such as e-commerce services, education services, legal services, medical services, conference services, social network services, financial product services, logistics services, and navigation services. There is no limitation here.

[0254] By using the embodiment of the present application, through the data processing system, the purpose of efficiently and accurately predicting the renewable energy output within the region is achieved, thereby realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the renewable energy power prediction has a high model maintenance cost, inaccurate prediction results, and long prediction time.

[0255] It should be noted that the preferred implementation manners of this embodiment can be referred to the relevant descriptions in the embodiment, and details are not described herein again.

[0256] According to an embodiment of the present application, an apparatus embodiment for implementing the above data processing method is further provided. Figure 12 FIG. Figure 12 is a schematic structural diagram of a data processing apparatus according to an embodiment of the present application. As Figure 12 shown, the apparatus includes:

[0257] The first acquisition module 1201 is configured to acquire target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area;

[0258] The first prediction module 1202 is configured to perform output prediction on the target numerical weather forecast data by using a target renewable energy output prediction model to obtain the regional renewable energy output potential, where the target renewable energy output prediction model is used to perform renewable energy output prediction on a specified area according to given numerical weather forecast data;

[0259] The first generation module 1203 is configured to generate target renewable energy output based on the regional renewable energy output potential, target historical power data, and target installed capacity data.

[0260] Optionally, the first acquisition module 1201 is further configured to: acquire multiple weather elements to be used within the target prediction area; perform numerical normalization processing on the multiple weather elements to obtain a first processing result; determine initial numerical weather forecast data based on the first processing result; and perform grid processing on the initial numerical weather forecast data to obtain the target numerical weather forecast data.

[0261] Optionally, the apparatus further includes: a processing module configured to: acquire initial historical power data and initial installed capacity data within the target prediction area; perform grid processing on the initial historical power data to obtain the target historical power data, and perform grid processing on the initial installed capacity data to obtain the target installed capacity data.

[0262] Optionally, the first generation module 1203 is further configured to: acquire the output potential of multiple grids according to the target historical power data and the target installed capacity data; generate target renewable energy output based on the regional renewable energy output potential and the output potential of the multiple grids, where the target renewable energy output includes: centralized renewable energy output and distributed renewable energy output.

[0263] Optionally, the first generation module 1203 is further configured to: respectively calculate the sum of historical power and the sum of installed capacity within the same grid for multiple grids according to the target historical power data and the target installed capacity data; perform numerical normalization processing on the sum of historical power to obtain a second processing result; and acquire the output potential of the multiple grids based on the second processing result and the sum of installed capacity.

[0264] Optionally, the apparatus further includes: a first sub-training module configured to: acquire a first training data set, where the first training data set includes: unmasked sample numerical weather forecast data; and train an initial renewable energy output prediction model by using the first training data set to generate a target renewable energy output prediction model.

[0265] Optionally, the initial renewable energy output prediction model includes: a self-attention encoder and a convolutional network decoder. The first sub-training module is further configured to: perform feature encoding on the first training dataset using the self-attention encoder to obtain a first encoding result; perform feature decoding on the first encoding result using the convolutional network decoder to obtain the centralized predicted output potential and the distributed predicted output potential; calculate a first target loss based on the centralized predicted output potential, the centralized actual output potential, the distributed predicted output potential, and the distributed actual output potential; and update the parameters of the self-attention encoder and the parameters of the convolutional network decoder simultaneously according to the first target loss to generate a target renewable energy output prediction model.

[0266] Optionally, the device further includes: a second sub-training module, configured to: obtain a second training dataset, where the second training dataset includes: sample numerical weather forecast data randomly masked by a preset ratio; and train the initial numerical weather forecast reconstruction model using the second training dataset to generate a target numerical weather forecast reconstruction model.

[0267] Optionally, the second sub-training module is further configured to: perform feature encoding on the second training dataset using the self-attention encoder to obtain a second encoding result; perform feature decoding on the second encoding result and the mask position embedding using the self-attention decoder to obtain a numerical weather forecast reconstruction value; calculate a second target loss based on the numerical weather forecast reconstruction value and the numerical weather forecast true value; and update the parameters of the self-attention encoder and the parameters of the self-attention decoder simultaneously according to the second target loss to generate a target numerical weather forecast reconstruction model.

[0268] Optionally, the device further includes: a summarization module, configured to: summarize the target renewable energy output generated within the target prediction area and the renewable energy output generated within multiple areas adjacent to the target prediction area to obtain the total renewable energy output within a preset geographical level range.

[0269] By adopting the embodiment of the present application, by obtaining the target numerical weather forecast data, target historical power data, and target installed capacity data within the target prediction area, and then using the target renewable energy output prediction model to perform output prediction on the target numerical weather forecast data to obtain the regional renewable energy output potential, wherein the target renewable energy output prediction model is used to perform renewable energy output prediction on a specified area based on the given numerical weather forecast data, and finally, based on the regional renewable energy output potential, target historical power data, and target installed capacity data, generate the target renewable energy output, thereby achieving the purpose of efficiently and accurately predicting the renewable energy output within the region, thus realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the renewable energy power prediction has a high model maintenance cost, inaccurate prediction results, and long prediction time.

[0270] It should be noted here that the above first acquisition module 1201, first prediction module 1202, and first generation module 1203 correspond to steps S21 to S23 in the embodiment. The examples and application scenarios implemented by the three modules and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiment. It should be noted that the above modules or units can be hardware components or software components stored in the memory and processed by one or more processors, and the above modules can also run in the server 10 provided in the embodiment.

[0271] According to the embodiment of the present application, another device embodiment for implementing the above data processing method is also provided. Figure 13 It is a schematic structural diagram of another data processing device according to the embodiment of the present application, as Figure 13 shown. The device includes:

[0272] A second acquisition module 1301, configured to acquire target numerical weather forecast data, target historical power data, and target installed capacity data within the target prediction area;

[0273] A second prediction module 1302, configured to use the target photovoltaic output prediction model to perform output prediction on the target numerical weather forecast data to obtain the regional photovoltaic output potential, wherein the target photovoltaic output prediction model is used to perform photovoltaic output prediction on a specified area based on the given numerical weather forecast data;

[0274] A second generation module 1303, configured to generate the target photovoltaic output based on the regional photovoltaic output potential, target historical power data, and target installed capacity data.

[0275] By adopting the embodiment of the present application, by obtaining the target numerical weather forecast data, target historical power data and target installed capacity data within the target prediction area, and then using the target photovoltaic output prediction model to perform output prediction on the target numerical weather forecast data to obtain the regional photovoltaic output potential, wherein the target photovoltaic output prediction model is used to perform photovoltaic output prediction on a specified area according to the given numerical weather forecast data, and finally, based on the regional photovoltaic output potential, target historical power data and target installed capacity data, generate the target photovoltaic output, thereby achieving the purpose of efficiently and accurately predicting the renewable energy output within the area, thus realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the renewable energy power prediction has a high model maintenance cost, inaccurate prediction results and long prediction time.

[0276] It should be noted here that the above-mentioned second acquisition module 1301, second prediction module 1302 and second generation module 1303 correspond to steps S61 to S63 in the embodiment. The examples and application scenarios implemented by the three modules and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should be noted that the above-mentioned modules or units can be hardware components or software components stored in the memory and processed by one or more processors, and the above-mentioned modules can also run in the server 10 provided in the embodiment.

[0277] According to the embodiment of the present application, there is also provided another device embodiment for implementing the above data processing method. Figure 14 It is a schematic structural diagram of another data processing device according to the embodiment of the present application, as Figure 14 shown. The device includes:

[0278] A third acquisition module 1401, configured to obtain a data processing request through a first application programming interface, wherein the request data carried in the data processing request includes: target numerical weather forecast data, target historical power data and target installed capacity data within the target prediction area;

[0279] A first return module 1402, configured to return a data processing response through a second application programming interface, wherein the response data carried in the data processing response includes: target renewable energy output, and the target renewable energy output is generated according to the data processing method of any one of the above.

[0280] By adopting the embodiment of the present application, a data processing request is obtained through a first application programming interface. The request data carried in the data processing request includes: target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area. Then, a data processing response is returned through a second application programming interface. The response data carried in the data processing response includes: target renewable energy output, and the target renewable energy output is generated according to any one of the above data processing methods. Thus, the purpose of efficiently and accurately predicting the renewable energy output within the area is achieved, thereby realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the prediction of renewable energy power generation has a high model maintenance cost, inaccurate prediction results, and long prediction time.

[0281] It should be noted here that the above-mentioned third acquisition module 1401 and the first return module 1402 correspond to steps S71 and S72 in the embodiment. The examples and application scenarios implemented by the two modules and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiment. It should be noted that the above-mentioned module or unit can be a hardware component or a software component stored in the memory and processed by one or more processors, and the above-mentioned module can also run in the server 10 provided in the embodiment.

[0282] According to the embodiment of the present application, there is also provided another device embodiment for implementing the above data processing method. Figure 15 It is a schematic structural diagram of another data processing device according to the embodiment of the present application, as Figure 15 shown. The device includes:

[0283] A fourth acquisition module 1501, configured to acquire a current input data processing dialogue request. The request data carried in the data processing dialogue request includes: target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area;

[0284] A second return module 1502, configured to return a data processing dialogue reply in response to the data processing dialogue request. The information carried in the data processing dialogue reply includes: target renewable energy output, and the target renewable energy output is generated according to any one of the above data processing methods;

[0285] A display module 1503, configured to display the target renewable energy output in a graphical user interface.

[0286] By adopting the embodiment of the present application, a data processing dialogue request is obtained based on the currently input data. The request data carried in the data processing dialogue request includes: target numerical weather forecast data, target historical power data, and target installed capacity data within the target prediction area. Then, in response to the data processing dialogue request, a data processing dialogue reply is returned. The information carried in the data processing dialogue reply includes: target renewable energy output, and the target renewable energy output is generated according to any one of the above data processing methods. Finally, the target renewable energy output is displayed in the graphical user interface, thereby achieving the purpose of efficiently and accurately predicting the renewable energy output within the area, and thus realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the prediction of the renewable energy power generation has a high model maintenance cost, inaccurate prediction results, and long prediction time.

[0287] It should be noted here that the above-mentioned fourth acquisition module 1501, second return module 1502, and display module 1503 correspond to steps S81 to S83 in the embodiment. The examples and application scenarios implemented by the three modules and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiment. It should be noted that the above-mentioned module or unit may be a hardware component or a software component stored in the memory and processed by one or more processors, and the above-mentioned module may also run in the server 10 provided in the embodiment.

[0288] According to the embodiment of the present application, another device embodiment for implementing the above data processing method is also provided. Figure 16 It is a schematic structural diagram of another data processing device according to the embodiment of the present application, as Figure 16 shown. The device includes:

[0289] A first display module 1601, configured to display target numerical weather forecast data, target historical power data, and target installed capacity data within the target prediction area on the operation interface in response to an input instruction acting on the operation interface;

[0290] A second display module 1602, configured to display target renewable energy output on the operation interface in response to a processing instruction acting on the operation interface; wherein, the target renewable energy output is generated according to any one of the above data processing methods.

[0291] By adopting the embodiments of the present application, by responding to the input instruction acting on the operation interface, the target numerical weather forecast data, the target historical power data, and the target installed capacity data in the target prediction area are displayed on the operation interface, and then by responding to the processing instruction acting on the operation interface, the target renewable energy output is displayed on the operation interface; wherein, the target renewable energy output is generated according to the data processing method of any one of the above, thereby achieving the purpose of efficiently and accurately predicting the renewable energy output in the area, thus realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems that the renewable energy power prediction in the related art has a high model maintenance cost, inaccurate prediction results and long prediction time.

[0292] It should be noted here that the above first display module 1601 and second display module 1602 correspond to step S91 and step S92 in the embodiment. The two modules have the same implemented examples and application scenarios as the corresponding steps, but are not limited to the content disclosed in the above embodiments. It should be noted that the above modules or units can be hardware components or software components stored in the memory and processed by one or more processors, and the above modules can also run in the server 10 provided in the embodiment.

[0293] According to the embodiments of the present application, there is also provided another device embodiment for implementing the above model training method. Figure 17 It is a structural schematic diagram of another model training device according to the embodiments of the present application, as Figure 17 shown, the device includes:

[0294] A fifth acquisition module 1701, configured to acquire a training data set, wherein the training data set includes: unmasked sample numerical weather forecast data;

[0295] A training module 1702, configured to train an initial renewable energy output prediction model by using the training data set to generate a target renewable energy output prediction model; wherein, the target renewable energy output prediction model is used to perform output prediction on the target numerical weather forecast data in the target prediction area to obtain the renewable energy output potential in the area.

[0296] By adopting the embodiments of the present application, a training data set is obtained, where the training data set includes: unmasked sample numerical weather forecast data, and then the initial renewable energy output prediction model is trained using the training data set to generate a target renewable energy output prediction model; wherein, the target renewable energy output prediction model is used to perform output prediction on the target numerical weather forecast data in the target prediction area to obtain the regional renewable energy output potential, thereby achieving the purpose of efficiently and accurately predicting the renewable energy output in the region, and thus realizing the technical effects of reducing the model maintenance cost, improving the prediction accuracy and prediction efficiency, and further solving the technical problems in the related art that the renewable energy power prediction has a high model maintenance cost, inaccurate prediction results and long prediction time.

[0297] It should be noted here that the above-mentioned fifth acquisition module 1701 and training module 1702 correspond to steps S101 and S102 in the embodiment. The instances and application scenarios implemented by the two modules and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiment. It should be noted that the above-mentioned module or unit can be a hardware component or a software component stored in the memory and processed by one or more processors, and the above-mentioned module can also run in the server 10 provided in the embodiment.

[0298] It should be noted that the preferred implementation schemes involved in the above embodiments of the present application are the same as the schemes, application scenarios, and implementation processes provided in the embodiments, but are not limited to the schemes provided in the embodiments.

[0299] The embodiments of the present application can provide a computing device. Figure 18 It is a structural block diagram of a computing device according to an embodiment of the present application. As Figure 18 shown, the computing device A may include: one or more ( Figure 18 only one is shown in the figure) processors 1802, a memory 1804, a storage controller, and a peripheral interface. The peripheral interface can be connected to a radio frequency module, an audio module, a display screen, etc., which are not limited here.

[0300] The above-mentioned computing device A can be understood as an integrated intelligent terminal, including but not limited to a server, a desktop computer, a PC (Personal Computer), a model all-in-one machine, etc., and the model described in the above embodiments of the present application can be preset in the computing device.

[0301] Specifically, the computing device A can pre-set various types of models, including but not limited to models in the fields of natural language processing, visual processing, speech processing, code processing, multi-modal task processing, etc., so as to provide diverse model selections. In different product forms, the computing device A can support one or more model usage methods, including but not limited to model training, model invocation, model fine-tuning, model deployment, model inference and application, etc. In some product forms, the computing device A also supports model management, including but not limited to multi-type model management (supporting the management of various types of models such as discriminative and generative models), model version control (supporting the control of different model versions), model evaluation (evaluating the performance and effect of the model based on model evaluation tools), etc. In other product forms, the computing device A can also create applications based on the model, provide API invocation capabilities, and can call the model into the created application through the API interface, while providing application management tools to achieve the management and monitoring of the application.

[0302] Furthermore, the computing device A can also include data management (supporting the creation and management of model tuning data sets), a training center (providing rich training resources to help users learn and master AI technologies), and basic control capabilities (providing enterprise-level basic control capabilities to ensure the security and efficient operation of the system). Through the above functions, a comprehensive and integrated AI development, training, deployment, and application device is provided.

[0303] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and devices in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implements the methods in the above embodiments. The memory can include high-speed random access memory, and can also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory can further include a memory remotely set relative to the processor, and these remote memories can be connected to the terminal through a network. Examples of the above network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and their combinations.

[0304] The processor can call the executable program stored in the memory through the transmission device to execute the method described in any one of the above embodiments.

[0305] Those of ordinary skill in the art can understand that the structure shown Figure 18 is only schematic, and the computing device A can also be a terminal device such as a smart phone, a tablet computer, a handheld computer, and a Mobile Internet Device (MID), a PAD, etc. The Figure 18It does not limit the structure of the above computing device. For example, computing device A may further include more or fewer components (such as network interfaces, display devices, etc.) than those shown in the Figure 18 , or have a different configuration from that shown in the Figure 18 .

[0306] Embodiments of the present application can provide an electronic device. Figure 19 is a structural block diagram of an electronic device according to an embodiment of the present application. As Figure 19 shown, the electronic device may include: an input / output device 192; a memory 194, and a processor 196. Among them, the processor 196 is connected to the input / output device 192 and the memory 194 through a bus 198.

[0307] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implements the methods in the above embodiments. The memory may include a high-speed random access memory, and may further include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory may further include a memory remotely provided relative to the processor, and these remote memories can be connected to the terminal through a network. Examples of the above network include but are not limited to the Internet, enterprise intranets, local area networks, mobile communication networks, and their combinations.

[0308] The processor can call the executable program stored in the memory through a transmission device to execute the method described in any one of the above embodiments.

[0309] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware of the terminal device through a program. The program can be stored in a computer-readable storage medium. The storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0310] Embodiments of the present application also provide a computer-readable storage medium. Optionally, in this embodiment, the above computer-readable storage medium can be used to save the program code executed by the data processing method or model training method provided in the above embodiments.

[0311] Optionally, in this embodiment, the above computer-readable storage medium can be located in any one of the computer terminals in a computer terminal group in a computer network, or in any one of the mobile terminals in a mobile terminal group.

[0312] Embodiments of the present application also provide a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements any of the above data processing methods or model training methods.

[0313] In the above embodiments of the present application, the descriptions of the respective embodiments have their own focuses. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0314] In the several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of units or modules can be in an electrical or other form.

[0315] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0316] In addition, the functional units in the various embodiments of the present application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0317] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable 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 various embodiments of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), mobile hard disks, magnetic disks, or optical discs.

[0318] The above are only the preferred embodiments of this application. It should be noted that for those of ordinary skill in the art, without departing from the principle of this application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of this application.

Claims

1. A data processing method, characterized in that, Including: Obtaining target numerical weather forecast data, target historical power data, and target installed capacity data within a target prediction area; Using a target renewable energy output prediction model to perform output prediction on the target numerical weather forecast data to obtain regional renewable energy output potential, where the target renewable energy output prediction model is used to perform renewable energy output prediction on a specified area based on given numerical weather forecast data; Generating target renewable energy output based on the regional renewable energy output potential, the target historical power data, and the target installed capacity data.

2. The data processing method according to claim 1, wherein Obtaining the target numerical weather forecast data within the target prediction area includes: Obtaining multiple weather elements to be used within the target prediction area; Performing numerical normalization processing on the multiple weather elements to obtain a first processing result; Determining initial numerical weather forecast data based on the first processing result; Performing grid processing on the initial numerical weather forecast data to obtain the target numerical weather forecast data.

3. The data processing method according to claim 1, characterized in that The data processing method further includes: Obtaining initial historical power data and initial installed capacity data within the target prediction area; Performing grid processing on the initial historical power data to obtain the target historical power data, and performing grid processing on the initial installed capacity data to obtain the target installed capacity data.

4. The data processing method according to claim 1, characterized in that Generating the target renewable energy output based on the regional renewable energy output potential, the target historical power data, and the target installed capacity data includes: Obtaining the output potential of multiple grids according to the target historical power data and the target installed capacity data; Generating the target renewable energy output based on the regional renewable energy output potential and the output potential of the multiple grids, where the target renewable energy output includes: centralized renewable energy output and distributed renewable energy output.

5. The data processing method according to claim 4, wherein Obtaining the output potential of the multiple grids according to the target historical power data and the target installed capacity data includes: Respectively calculating the sum of historical power and the sum of installed capacity within the same grid among the multiple grids according to the target historical power data and the target installed capacity data; Performing numerical normalization processing on the sum of historical power to obtain a second processing result; Obtaining the output potential of the multiple grids based on the second processing result and the sum of installed capacity.

6. The data processing method according to claim 1, characterized in that, The data processing method further includes: Obtaining a first training data set, where the first training data set includes: unmasked sample numerical weather forecast data; Training an initial renewable energy output prediction model using the first training data set to generate the target renewable energy output prediction model.

7. The data processing method according to claim 6, wherein The initial renewable energy output prediction model includes: a self-attention encoder and a convolutional network decoder. Training the initial renewable energy output prediction model using the first training data set to generate the target renewable energy output prediction model includes: Performing feature encoding on the first training data set using the self-attention encoder to obtain a first encoding result; Feature decode the first encoding result using the convolutional network decoder to obtain the centralized predicted output potential and the distributed predicted output potential; Calculate a first target loss based on the centralized predicted output potential, the centralized actual output potential, the distributed predicted output potential, and the distributed actual output potential; Update the parameters of the self-attention encoder and the parameters of the convolutional network decoder simultaneously according to the first target loss to generate the target renewable energy output prediction model.

8. The data processing method according to claim 7, wherein The data processing method further includes: Obtain a second training data set, where the second training data set includes: sample numerical weather forecast data randomly masked by a preset ratio; Train an initial numerical weather forecast reconstruction model using the second training data set to generate the target numerical weather forecast reconstruction model.

9. The data processing method according to claim 8, wherein The initial renewable energy output prediction model includes: the self-attention encoder and the self-attention decoder. Training the initial numerical weather forecast reconstruction model using the second training data set to generate the target numerical weather forecast reconstruction model includes: Feature encode the second training data set using the self-attention encoder to obtain a second encoding result; Feature decode the second encoding result and the mask position embedding using the self-attention decoder to obtain a numerical weather forecast reconstruction value; Calculate a second target loss based on the numerical weather forecast reconstruction value and the numerical weather forecast true value; Update the parameters of the self-attention encoder and the parameters of the self-attention decoder simultaneously according to the second target loss to generate the target numerical weather forecast reconstruction model.

10. The data processing method according to claim 1, wherein The data processing method further includes: Summarize the target renewable energy output generated in the target prediction area and the renewable energy output generated in multiple areas adjacent to the target prediction area to obtain the total renewable energy output within a preset geographical level range.

11. A data processing method, characterized in that, including: Obtain target numerical weather forecast data, target historical power data, and target installed capacity data in the target prediction area; Use a target photovoltaic output prediction model to perform output prediction on the target numerical weather forecast data to obtain the regional photovoltaic output potential, where the target photovoltaic output prediction model is used to perform photovoltaic output prediction on a specified area according to given numerical weather forecast data; Generate target photovoltaic output based on the regional photovoltaic output potential, the target historical power data, and the target installed capacity data.

12. A data processing method, characterized in that, including: Obtain a data processing request through a first application programming interface, where the request data carried in the data processing request includes: target numerical weather forecast data, target historical power data, and target installed capacity data in the target prediction area; Return a data processing response through a second application programming interface, where the response data carried in the data processing response includes: target renewable energy output, and the target renewable energy output is generated according to the data processing method described in any one of claims 1 to 10.

13. A data processing method, characterized in that, including: Obtain the current input data processing dialogue request, wherein the request data carried in the data processing dialogue request includes: target numerical weather forecast data, target historical power data, and target installed capacity data within the target prediction area; In response to the data processing dialogue request, return a data processing dialogue reply, wherein the information carried in the data processing dialogue reply includes: target renewable energy output, and the target renewable energy output is generated according to the data processing method described in any one of claims 1 to 10; Display the target renewable energy output within the graphical user interface.

14. A data processing method, characterized in that, Comprising: In response to an input instruction acting on the operation interface, display the target numerical weather forecast data, target historical power data, and target installed capacity data within the target prediction area on the operation interface; In response to a processing instruction acting on the operation interface, display the target renewable energy output on the operation interface; Wherein, the target renewable energy output is generated according to the data processing method described in any one of claims 1 to 10.

15. A model training method, characterized in that, Comprising: Obtain a training data set, wherein the training data set includes: unmasked sample numerical weather forecast data; Use the training data set to train an initial renewable energy output prediction model to generate a target renewable energy output prediction model; Wherein, the target renewable energy output prediction model is used to perform output prediction on the target numerical weather forecast data within the target prediction area to obtain the regional renewable energy output potential.

16. A data processing system, characterized in that, Comprising: A client for sending the target numerical weather forecast data, target historical power data, and target installed capacity data within the target prediction area; A server connected to the client, for using the target renewable energy output prediction model to perform output prediction on the target numerical weather forecast data to obtain the regional renewable energy output potential, and generating the target renewable energy output based on the regional renewable energy output potential, the target historical power data, and the target installed capacity data, wherein the target renewable energy output prediction model is used to perform renewable energy output prediction on a specified area according to the given numerical weather forecast data; The client is further used to output the target renewable energy output.

17. An electronic device, characterized in that, Comprising: A memory storing an executable program; A processor for running the program, wherein when the program runs, it executes the data processing method described in any one of claims 1 to 14 or the model training method described in claim 15.

18. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein when the executable program runs, it controls the device where the computer-readable storage medium is located to execute the data processing method described in any one of claims 1 to 14 or the model training method described in claim 15.

19. A computer program product, characterized in that, Comprising a computer program, which when executed by a processor implements the data processing method described in any one of claims 1 to 14 or the model training method described in claim 15.

Images