Day-ahead solar irradiance prediction method, system, device and storage medium

Abstract

This application discloses a method, system, device, and storage medium for predicting day-ahead solar irradiance. The method comprises the following steps: obtaining historical day-ahead solar irradiance data and determining meteorological data corresponding to the optimal parameterization scheme in the WRF model; determining a weather label for the forecast day based on the historical data; updating parameters of a day-ahead solar irradiance prediction network based on the historical data, meteorological data, and weather label to obtain a trained solar irradiance prediction model; and inputting the historical data into the solar irradiance prediction model to obtain a target day-ahead solar irradiance for the forecast day. This application can be widely applied in the field of artificial intelligence technology.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, system, device and storage medium for predicting day-ahead solar irradiance. Background Art

[0002] The existing WRF model has many configuration options, especially in terms of the radiation transfer model (RTM), microphysics scheme, and convection scheme, which means that users are faced with thousands of possible combinations when configuring the model. This not only increases the difficulty of model optimization, but may also affect the accuracy of the prediction. In addition, most studies focus on GHI predictions on sunny and cloudy days, and there is less research on improving prediction accuracy under complex weather conditions such as rainy and cloudy days. The prediction errors of existing methods under these complex weather conditions are still large, especially in high altitude and complex terrain areas. Therefore, there is currently a lack of adaptive optimization methods for different weather types, resulting in insufficient accuracy and stability of GHI predictions under changeable weather conditions. Therefore, there are still technical problems that need to be solved in related technologies. Summary of the Invention

[0003] The purpose of this application is to solve one of the technical problems existing in the prior art to at least a certain extent.

[0004] To this end, an object of the embodiments of the present application is to provide a method, system, device and storage medium for predicting day-ahead solar irradiance, which can improve the accuracy of GHI prediction.

[0005] In order to achieve the above-mentioned technical objectives, the technical solutions adopted in the embodiments of the present application include: a method for predicting solar irradiance before the day, comprising the following steps: obtaining historical data of solar irradiance before the day and determining the meteorological data corresponding to the optimal parameterization scheme in the WRF model; determining the weather label of the predicted day based on the historical data; updating the parameters of the solar irradiance prediction network before the day based on the historical data, the meteorological data and the weather label to obtain a trained solar irradiance prediction model; inputting the historical data into the solar irradiance prediction model to obtain the target solar irradiance before the day for the predicted day.

[0006] This application can obtain historical data of solar irradiance before the day and determine the meteorological data corresponding to the optimal parameterization scheme in the WRF model; determine the weather label of the forecast day based on the historical data; update the parameters of the solar irradiance prediction network before the day based on the historical data, the meteorological data and the weather label to obtain a trained solar irradiance prediction model; input the historical data into the solar irradiance prediction model to obtain the target solar irradiance before the day for the forecast day. This application can improve the accuracy of the prediction by adding meteorological data.

[0007] In addition, the method for predicting solar irradiance before the day according to the above embodiment of the present invention may also have the following additional technical features:

[0008] Furthermore, in the embodiment of the present application, determining the meteorological data corresponding to the optimal parameterization scheme in the WRF model specifically includes:

[0009] During the combined testing and optimization phase of all parameterization schemes of the WRF model, the parameterization schemes of the surface and planetary boundary layer are systematically tested to obtain several first optimized parameterization schemes;

[0010] Based on fixed surface and planetary boundary layer schemes, testing combinations of fixed radiation transfer models, microphysical models, and convection parameterization schemes, and screening out an optimal parameterization scheme from the plurality of first optimized parameterization schemes; wherein the number of first optimized parameterization schemes is less than the number of all parameterization schemes; and the number of optimal parameterization schemes is less than the number of the first optimized parameterization schemes;

[0011] The WRF model runs the optimal parameterization scheme to obtain corresponding meteorological data.

[0012] Furthermore, in the embodiment of the present application, determining the weather label of the forecast day based on the historical data specifically includes:

[0013] Determining a daily average cloud cover factor based on the historical data;

[0014] Determining a first weather label corresponding to each of the daily average cloud cover coefficients based on a preset neural network model;

[0015] All of the first weather labels are used as all of the weather labels for the predicted day.

[0016] Furthermore, in the embodiment of the present application, determining the daily average cloud cover coefficient based on the historical data specifically includes:

[0017] extracting a first irradiance from the historical data;

[0018] The daily average cloud cover coefficient is determined based on a preset theoretical clear sky irradiance and the first irradiance.

[0019] Furthermore, in an embodiment of the present application, determining the daily average cloud cover coefficient based on a preset theoretical clear sky irradiance and the first irradiance specifically includes:

[0020] Taking the first irradiance and the theoretical clear sky irradiance as a quotient to obtain a first quotient;

[0021] The first quotient is used as the daily average cloud cover coefficient.

[0022] Furthermore, in an embodiment of the present application, the day-ahead solar irradiance prediction network includes a convolutional neural network, a long short-term memory network, and a Transformer model; the parameters of the day-ahead solar irradiance prediction network are updated based on the historical data, the meteorological data, and the weather label to obtain a trained solar irradiance prediction model, specifically including:

[0023] Inputting the meteorological data into the convolutional neural network to obtain a spatial feature sequence;

[0024] Inputting the spatial feature sequence and the historical data into the long short-term memory network to obtain a first predicted solar irradiance for the predicted day;

[0025] Inputting the historical data, the first predicted solar irradiance, and the weather label into the Transformer model to obtain a second predicted solar irradiance;

[0026] According to the second predicted solar irradiance, the parameters of the day-ahead solar irradiance prediction network are updated for several rounds to obtain a spatial feature sequence.

[0027] Furthermore, in the embodiment of the present application, the method further includes:

[0028] The target day-ahead solar irradiance is compared with a preset day-ahead solar irradiance, and the parameters of the solar irradiance prediction model are adjusted according to the comparison result.

[0029] On the other hand, an embodiment of the present application further provides a day-ahead solar irradiance prediction system, comprising:

[0030] The first processing unit is used to obtain historical data of solar irradiance before the day and determine meteorological data corresponding to the optimal parameterization scheme in the WRF model;

[0031] a second processing unit, configured to determine a weather label for a forecast day based on the historical data;

[0032] a third processing unit, configured to update parameters of a day-ahead solar irradiance prediction network according to the historical data, the meteorological data, and the weather label to obtain a trained solar irradiance prediction model;

[0033] The fourth processing unit is used to input the historical data into the solar irradiance prediction model to obtain the target day-ahead solar irradiance of the prediction day.

[0034] On the other hand, the present application also provides a day-ahead solar irradiance prediction device, comprising:

[0035] at least one processor;

[0036] at least one memory for storing at least one program;

[0037] When the at least one program is executed by the at least one processor, the at least one processor implements a method for predicting solar irradiance before the day as described in any one of the invention contents.

[0038] In addition, the present application also provides a computer-readable storage medium, which stores processor-executable instructions. When the processor executes the processor, the processor-executable instructions are used to execute a day-ahead solar irradiance prediction method as described in any one of the above items.

[0039] The advantages and benefits of this application will be partially given in the following description, and partially become apparent from the following description, or learned through practice of this application:

[0040] This application can obtain historical data of solar irradiance before the day and determine the meteorological data corresponding to the optimal parameterization scheme in the WRF model; determine the weather label of the forecast day based on the historical data; update the parameters of the solar irradiance prediction network before the day based on the historical data, the meteorological data and the weather label to obtain a trained solar irradiance prediction model; input the historical data into the solar irradiance prediction model to obtain the target solar irradiance before the day for the forecast day. This application can improve the accuracy of the prediction by adding meteorological data. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a schematic diagram of the steps of a method for predicting solar irradiance before the day in a specific embodiment of the present invention;

[0042] Figure 2 This is a flow chart of a method for predicting solar irradiance in a specific embodiment of the present invention.

[0043] Figure 3 This is a structural diagram of a day-ahead solar irradiance prediction system in another specific embodiment of the present invention;

[0044] Figure 4 Schematic diagram of the structure of a device for predicting solar irradiance before the day in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following describes the embodiments of the present invention in detail with reference to the accompanying drawings to illustrate the principles and processes of the day-ahead solar irradiance prediction method, system, device and storage medium in the embodiments of the present invention.

[0046] First, the nouns of this application are explained:

[0047] 1. Numerical Weather Prediction (NWP) model: A computer simulation that uses current weather conditions to predict future weather. It is the basis of weather forecasting and uses mathematical models to describe processes in the atmosphere.

[0048] 2.WRF model: Weather Research and Forecasting Model, a widely used regional numerical weather forecast model

[0049] 3. Global Horizontal Irradiance (GHI): The total amount of radiation from the sun received by the ground, including direct radiation and diffuse radiation.

[0050] 4. Parameterization scheme: A simplified method used in numerical models to deal with small-scale processes that cannot be explicitly resolved (such as cloud formation and radiation transfer).

[0051] 5. Radiative transfer model (RTM): A model used to calculate radiative transfer in the atmosphere.

[0052] 6. Microphysics scheme: A parameterized scheme that describes the processes of tiny particles in the atmosphere (such as cloud droplets and raindrops).

[0053] 7. Convection scheme: a parameterized scheme for simulating convection processes in the atmosphere.

[0054] 8. Planetary Boundary Layer (PBL) Scheme: A parameterization scheme that describes the turbulent exchange processes between the Earth's surface and the upper atmosphere, affecting the heat and momentum transport in the lower atmosphere.

[0055] 9. Root Mean Square Error (RMSE): A statistical indicator that measures the deviation between the predicted value and the actual observed value. It is calculated by taking the square root of the average of the squares of the prediction errors.

[0056] 10. Convolutional Neural Network (CNN): A deep learning model that excels at processing data with a grid structure (such as images). In GHI forecasting, CNN is used to extract spatial features, such as cloud distribution and changes, to improve the spatial resolution and accuracy of the forecast.

[0057] 11. Long Short-Term Memory (LSTM): A recurrent neural network (RNN) that can capture long-term dependencies in time series data. In GHI forecasting, LSTM is used to model the temporal dynamics of GHI, improving the accuracy of short- and medium-term forecasts.

[0058] 12. Transformer model: A deep learning model based on the self-attention mechanism that excels at handling long-term dependencies and multivariate input data. In GHI forecasting, the Transformer model is used to capture complex spatiotemporal relationships and improve forecast accuracy.

[0059] 13. Reinforcement Learning (RL): A machine learning method that learns optimal strategies by interacting with the environment.

[0060] 14. Automated Machine Learning (AutoML): A technology that automates the search and optimization of machine learning model hyperparameters and architectures, aiming to reduce the complexity and time cost of manual parameter adjustment.

[0061] The existing WRF model has many configuration options, especially in terms of the radiation transfer model (RTM), microphysics scheme, and convection scheme, which means that users are faced with thousands of possible combinations when configuring the model. This not only increases the difficulty of model optimization, but may also affect the accuracy of the prediction. In addition, most studies focus on GHI predictions on sunny and cloudy days, and there is less research on improving prediction accuracy under complex weather conditions such as rainy and cloudy days. The prediction errors of existing methods under these complex weather conditions are still large, especially in high altitude and complex terrain areas. Therefore, there is currently a lack of adaptive optimization methods for different weather types, resulting in insufficient accuracy and stability of GHI predictions under changeable weather conditions. Therefore, there are still technical problems that need to be solved in related technologies.

[0062] In view of the above-mentioned defects of the prior art, Figure 1 , this application provides a method for predicting solar irradiance before the day. Figure 1 In the method, the method may include the following steps S101-S104.

[0063] S101. Obtain historical data of solar irradiance before the day and determine meteorological data corresponding to the optimal parameterization scheme in the WRF model.

[0064] S102: Determine the weather label for the forecast day based on historical data.

[0065] S103: Update the parameters of the day-ahead solar irradiance prediction network according to historical data, meteorological data, and weather labels to obtain a trained solar irradiance prediction model.

[0066] S104: Input the historical data into a solar irradiance prediction model to obtain the target solar irradiance on the prediction day.

[0067] In some feasible embodiments of the present application, the processor can be connected to the acquisition unit via a wired or wireless connection. After the connection is established, the processor can obtain historical data of the solar irradiance of the day before from the acquisition unit and determine the meteorological data corresponding to the optimal parameterization scheme in the WRF model. Based on the historical data, the processor can determine the weather label of the forecast day. Based on the historical data, the meteorological data and the weather label, the parameters of the solar irradiance prediction network of the day before are updated to obtain a trained solar irradiance prediction model. The historical data is input into the solar irradiance prediction model to obtain the target solar irradiance of the day before for the forecast day.

[0068] It should be noted that the above-mentioned wired connection method may include a connection between a mobile device and a processing module, and may also include a connection between a processing module and a hardware device, as well as a wired connection between other devices currently known or to be developed in the future and the processing module; and the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (Ultra Wide Band) connection, and other wireless connection methods currently known or to be developed in the future.

[0069] This application can obtain historical data of solar irradiance before the day and determine the meteorological data corresponding to the optimal parameterization scheme in the WRF model; determine the weather label of the forecast day based on the historical data; update the parameters of the solar irradiance prediction network before the day based on the historical data, meteorological data and weather labels to obtain a trained solar irradiance prediction model; input the historical data into the solar irradiance prediction model to obtain the target solar irradiance before the day of the forecast day. This application can improve the accuracy of the prediction by adding meteorological data.

[0070] Furthermore, in the embodiment of the present application, the step of determining the meteorological data corresponding to the optimal parameterization scheme in the WRF model specifically includes steps S201-S203.

[0071] S201. During the combined testing and optimization phase of all parameterization schemes of the WRF model, the parameterization schemes of the surface and the planetary boundary layer are systematically tested to obtain several first optimized parameterization schemes.

[0072] S202. Based on the fixed surface and planetary boundary layer schemes, test the combination of the fixed radiation transfer model, the microphysical model, and the convection parameterization scheme, and screen out the optimal parameterization scheme from several first optimized parameterization schemes; wherein the number of first optimized parameterization schemes is less than the number of all parameterization schemes; and the number of optimal parameterization schemes is less than the number of first optimized parameterization schemes.

[0073] S203: Run the optimal parameterization scheme to obtain corresponding meteorological data.

[0074] Furthermore, in the embodiment of the present application, the step of determining the weather label of the forecast day based on historical data specifically includes steps S301 to S303.

[0075] S301. Determine the daily average cloud cover coefficient based on historical data.

[0076] S302: Determine a first weather label corresponding to each daily average cloud cover coefficient based on a preset neural network model.

[0077] S303: Use all first weather labels as all weather labels for the forecast day.

[0078] Furthermore, in the embodiment of the present application, the step of determining the daily average cloud cover coefficient based on historical data specifically includes steps S401 and S402.

[0079] S401, extracting the first irradiance in the historical data;

[0080] S402: Determine the daily average cloud cover coefficient based on a preset theoretical clear sky irradiance and a first irradiance.

[0081] Furthermore, in an embodiment of the present application, the step of determining the daily average cloud cover coefficient based on the preset theoretical clear sky irradiance and the first irradiance specifically includes steps S501 and S502.

[0082] S501, taking the first irradiance and the theoretical clear sky irradiance as a quotient to obtain a first quotient;

[0083] S502. Take the first quotient as the daily average cloud cover coefficient.

[0084] Furthermore, in an embodiment of the present application, the day-ahead solar irradiance prediction network includes a convolutional neural network, a long short-term memory network, and a Transformer model; based on historical data, meteorological data, and weather labels, the parameters of the day-ahead solar irradiance prediction network are updated to obtain a trained solar irradiance prediction model. This step specifically includes steps S501 to S504.

[0085] S501, inputting meteorological data into a convolutional neural network to obtain a spatial feature sequence;

[0086] S502, inputting the spatial feature sequence and historical data into a long short-term memory network to obtain the first predicted solar irradiance of the forecast day;

[0087] S503, inputting historical data, the first predicted solar irradiance, and the weather label into the Transformer model to obtain a second predicted solar irradiance;

[0088] S504: Based on the second predicted solar irradiance, update the parameters of the day-ahead solar irradiance prediction network for several rounds to obtain a spatial feature sequence.

[0089] Furthermore, in the embodiment of the present application, the method for predicting solar irradiance before the day also includes step S105.

[0090] S105 : Compare the target day-ahead solar irradiance with the preset day-ahead solar irradiance, and adjust the parameters of the solar irradiance prediction model according to the comparison result.

[0091] The following is combined with Figure 2 The principle of this application is explained.

[0092] Reference Figure 2 , this embodiment proposes a day-ahead solar irradiance (GHI) prediction method based on the adaptive parameter optimization of the numerical model WRF and post-processing of advanced artificial intelligence algorithms. First, high-resolution GHI observation data and numerical weather prediction (NWP) model data are collected and organized. The observation data are derived from the latest meteorological satellites and ground observation stations to ensure that the data have high temporal and spatial resolution (for example, 15-minute intervals and 1 km spatial resolution). The NWP model input uses the latest version of the WRF model (such as WRF 4.5) and is combined with the initial and boundary condition data provided by the European Centre for Medium-Range Weather Forecasts (ECMWF). Data preprocessing includes missing value filling, outlier detection and correction, and data synchronization processing to ensure the integrity and consistency of the input data.

[0093] During the parameterization scheme combination testing and optimization phase, a total of 219 combinations of surface and planetary boundary layer (PBL) parameterization schemes were systematically tested. By fixing the radiative transfer model (RTM), microphysics scheme, and convection scheme, the root mean square error (RMSE) and bias (Bias) of each combination over different test periods (e.g., 15 days) were evaluated. Finally, the surface and PBL scheme combinations that performed best over the majority of test days were selected, such as the NCEP Global Forecast System surface layer scheme, the RUC surface model, and the UW (Park and Bretherton) PBL scheme. Subsequently, based on the fixed surface and PBL schemes, 187 combinations of fixed radiative transfer models, microphysics, and convection parameterization schemes were tested, and 130 combinations that performed successfully in actual operation were selected. The RMSE and bias of these combinations under various weather conditions were compared to identify the best performing parameterization combinations, which have a significant impact on GHI forecast performance.

[0094] Next, based on the observation data and the output of the WRF model (Weather Research and Forecasting Model), advanced artificial intelligence algorithms are used to accurately classify weather types. Specifically, the daily average cloud cover coefficient (Kt ),Right now Among them I observed is the observed or model-predicted irradiance, I clearsky is the theoretical clear sky irradiance. Then, a deep learning model such as convolutional neural network (CNN) is used to calculate K t The forecast days are classified into four categories: sunny, cloudy, overcast and rainy.

[0095] Through training, the AI ​​model can automatically identify and classify complex weather patterns, improving classification accuracy and robustness. Based on the classification results, it dynamically selects the optimal WRF parameterization combination for the corresponding weather type. For example, parameterization A is selected for sunny days, B for cloudy days, C for overcast days, and D for rainy days. This selection process is optimized through a reinforcement learning algorithm, enabling real-time, autonomous parameterization adjustment.

[0096] During the post-processing phase of the AI ​​algorithm, the initial GHI forecast results of the WRF model (Weather Research and Forecasting Model) were further optimized by incorporating the latest AI technologies. First, feature engineering was performed to extract various meteorological variables related to GHI as input features, including temperature, humidity, wind speed, cloud cover, and precipitation. Historical GHI data and real-time observation data were combined to construct a multidimensional feature set. Then, convolutional neural networks (CNN), long short-term memory networks (LSTM), and Transformer models were used to model spatial features, temporal dynamics, and multivariate associations, respectively. This enhanced forecasting capability through multi-level feature extraction and the capture of spatiotemporal relationships. Furthermore, automated machine learning (AutoML) technology was used to automatically search and optimize the AI ​​model's hyperparameters and architecture to ensure optimal performance under different weather conditions. Finally, the trained AI model was applied to real-time forecast data, and the WRF model output was post-processed to generate the final GHI forecast results, significantly reducing RMSE and bias, and improving the accuracy and stability of the forecast.

[0097] During the performance evaluation and feedback optimization phase, forecast performance was evaluated using metrics such as root mean square error (RMSE), mean absolute error (MAE), and bias by comparing the system with high-resolution observational data. The system also compared the forecast performance of adaptive parameterization combinations combined with AI post-processing with a single optimal configuration, traditional hybrid model approaches, and global models such as ECMWF, ensuring that the adaptive approach improves forecast accuracy across different weather types. Based on the evaluation results, the system continuously optimized parameter selection strategies and AI model training using reinforcement learning algorithms and the AutoML framework, forming a closed-loop feedback mechanism to ensure that the system continuously improves forecast performance in practical applications.

[0098] Finally, the entire GHI prediction method is integrated into the energy management and grid dispatching system to support real-time prediction and dynamic adjustment. System integration includes establishing a real-time data interface with meteorological data sources (such as satellites, ground observation stations, ECMWF) to ensure timely acquisition and updating of data; utilizing cloud computing platforms and distributed computing resources to deploy WRF model instances and AI algorithms to support the testing and optimization of large-scale parameterized schemes; and developing a friendly user interface to provide a visual display of real-time GHI prediction results and support user-defined parameter configuration and result analysis. Through the above technical solutions, the present invention can achieve high-precision and stable GHI prediction, meet the high requirements of modern power systems for renewable energy prediction, and has broad application prospects and significant economic benefits.

[0099] Other embodiments of the present invention take Qinghai Province in China as an example to illustrate in detail the specific application process of the technical solution of the present invention.

[0100] First, high-resolution GHI observational data and numerical weather prediction (NWP) model data from 2018 to 2023 were collected and compiled. The observational data were sourced from the latest meteorological satellites and ground-based observation stations, ensuring high temporal and spatial resolution (e.g., 15-minute intervals and 1 km spatial resolution). The NWP model inputs used the latest version of the WRF model (e.g., WRF4.5), combined with initial and boundary condition data provided by the European Centre for Medium-Range Weather Forecasts (ECMWF). Data preprocessing included missing value imputation, outlier detection and correction, and data synchronization to ensure the integrity and consistency of the input data.

[0101] In the parameterization scheme combination testing and optimization stage, the surface and planetary boundary layer (PBL) parameterization schemes were first systematically tested, with a total of 219 combinations. By fixing the radiation transfer model (RTM), microphysics scheme and convection scheme, the root mean square error (RMSE) and bias (Bias) of each combination during different test periods (such as 15 days) were evaluated. Finally, the surface and planetary boundary layer PBL scheme combination that performed best in most test days was selected, such as the NCEP Global Forecast System surface layer scheme, the RUC surface model and the UW (Park and Bretherton) PBL scheme.

[0102] Subsequently, based on the fixed surface and PBL schemes, 187 combinations of RTM radiation transfer models, microphysics and convection parameterization schemes were tested, and 130 combinations that were successfully operated were screened out. By comparing the RMSE and Bias of these combinations under various weather conditions, the parameterization combination with the best performance was determined, which had a significant impact on the GHI prediction performance.

[0103] Next, based on the observation data (measured under the above model) and the WRF model output, advanced artificial intelligence algorithms are used to accurately classify weather types. Specifically, the daily average cloud cover coefficient (K t ),Right now Among them I observed is the observed or model-predicted irradiance, I clearsky is the theoretical clear-sky irradiance.

[0104] Then, a deep learning model, convolutional neural network (CNN), is used to classify the values ​​and divide the forecast days into four categories: sunny, cloudy, overcast, and rainy.

[0105] Through training, the AI ​​model can automatically identify and classify complex weather patterns, significantly improving classification accuracy and robustness. During the AI ​​algorithm post-processing phase, the latest AI technology is combined to further optimize the WRF model's initial GHI forecast results.

[0106] First, feature engineering is performed to extract various meteorological variables related to GHI as input features, including temperature, humidity, wind speed, cloud cover, precipitation, etc. Historical GHI data and real-time observation data are combined to construct a multi-dimensional feature set. This feature set will serve as input for subsequent AI models to ensure that the models can fully capture the various factors affecting GHI.

[0107] Next, a convolutional neural network (CNN) is used to model spatial features. Through multiple layers of convolution and pooling, CNN extracts local spatial features from high-resolution spatial weather data, such as cloud distribution and terrain influences, thereby capturing the spatial dependencies between geographic locations. These spatial features are processed through multiple layers of convolution kernels to form high-dimensional feature vectors, providing a solid foundation for subsequent time series modeling.

[0108] Subsequently, a long short-term memory (LSTM) network was used to model temporal dynamics. Through its unique gating mechanism, the LSTM effectively captures trends and patterns in GHI over time, particularly excelling in short- and medium-term forecasting. The LSTM receives the spatial feature sequence from the CNN and progressively updates its hidden state, preserving historical information and combining it with current input to generate predictions of future GHI changes.

[0109] Finally, the Transformer model is used to model multivariate correlations. Using its self-attention mechanism, the Transformer can simultaneously process the complex relationships between multiple meteorological variables, identify variables with key impacts on GHI forecasts, and dynamically adjust their weights. Positional encoding ensures the model understands the order and temporal dependencies of the data, while a feedforward neural network further extracts deep-level features. Through its multi-head attention mechanism, the Transformer can integrate information from multiple sources, improving the overall performance of GHI forecasts. Combining the CNN, LSTM, and Transformer models creates a comprehensive feature representation. Through fully connected layers or other fusion mechanisms, the spatial, temporal, and multivariate features extracted from each are integrated to produce the final GHI forecast.

[0110] The GHI output by the LSTM is a short-term forecast based on time series data, focusing on capturing the temporal dependencies between historical GHI and meteorological data. The final GHI forecast is the result of further modeling and optimization of spatial, temporal, and multivariate factors using the LSTM forecast results combined with CNN and Transformer models.

[0111] Furthermore, during the 100 forward propagations of each model, automated machine learning (AutoML) technology was used to automatically search and optimize the hyperparameters and architecture of each model, ensuring optimal performance under different weather conditions. Through this multi-layered, multi-model collaborative approach, the proposed method comprehensively captures the various factors affecting GHI, significantly improving the accuracy and robustness of the forecast and ensuring efficient and accurate GHI forecasts under various weather conditions.

[0112] In addition, refer to Figure 3 ,and Figure 1 Corresponding to the method, an embodiment of the present application further provides a day-ahead solar irradiance prediction system. The system may include a first processing unit 1001, a second processing unit 1002, a third processing unit 1003 and a fourth processing unit 1004. Among them, the first processing unit 1001 can be used to obtain historical data of the day-ahead solar irradiance and determine the meteorological data corresponding to the optimal parameterization scheme in the WRF model. The second processing unit 1002 can be used to determine the weather label of the forecast day based on the historical data. The third processing unit 1003 can be used to update the parameters of the day-ahead solar irradiance prediction network based on the historical data, meteorological data and weather labels to obtain a trained solar irradiance prediction model; the fourth processing unit 1004 can be used to input historical data into the solar irradiance prediction model to obtain the target day-ahead solar irradiance of the forecast day.

[0113] It should be noted that the first processing unit may be any integrated circuit unit or microprocessor unit obtained by integrating a processing chip and its peripheral circuits using existing integration technologies. The first and second processing units may also be any integrated circuit modules or microprocessor modules obtained by integrating a processing chip and its peripheral circuits using existing integration technologies. The first and second processing units may also include one or more memories.

[0114] It should be noted that the contents of the above-mentioned embodiment of the day-ahead solar irradiance prediction method are all applicable to the present day-ahead solar irradiance prediction system embodiment. The functions specifically implemented by the present day-ahead solar irradiance prediction system embodiment are the same as those of the above-mentioned day-ahead solar irradiance prediction method embodiment, and the beneficial effects achieved are also the same as those achieved by the above-mentioned day-ahead solar irradiance prediction method embodiment.

[0115] and Figure 1 Corresponding to the method, the embodiment of the present application also provides a day-ahead solar irradiance prediction device, the specific structure of which can be referred to Figure 4 ,include:

[0116] at least one processor 1011;

[0117] at least one memory 1012, configured to store at least one program;

[0118] When the at least one program is executed by the at least one processor, the at least one processor implements the method for predicting solar irradiance before the day.

[0119] The contents of the above method embodiments are all applicable to the present device embodiments. The functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0120] and Figure 1 Corresponding to the method, an embodiment of the present application further provides a computer-readable storage medium, which stores instructions executable by a processor, and the processor-executable instructions are used to execute the method for predicting solar irradiance before the day when executed by the processor.

[0121] The contents of the above-mentioned embodiment of the day-ahead solar irradiance prediction method are all applicable to the embodiment of this storage medium. The functions specifically implemented by this storage medium embodiment are the same as those of the above-mentioned embodiment of the day-ahead solar irradiance prediction method, and the beneficial effects achieved are also the same as those achieved by the above-mentioned embodiment of the day-ahead solar irradiance prediction method.

[0122] In some optional embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiments presented and described in the flow chart of the present application are provided in an exemplary manner for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operations and logical flows presented herein. Optional embodiments are contemplated in which the order of the various operations is changed and the sub-operations described as a part of a larger operation are performed independently.

[0123] In addition, although the present application is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present application. More specifically, given the properties, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be understood within the routine skills of an engineer. Therefore, a person skilled in the art can implement the present application as set forth in the claims using ordinary techniques without undue experimentation. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present application, which is determined by the full scope of the appended claims and their equivalents.

[0124] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several programs for enabling 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 method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0125] The logic and / or steps represented in a flowchart or otherwise described herein, for example, may be considered as an ordered list of executable programs for implementing the logical functions, and may be embodied in any computer-readable medium for use by, or in conjunction with, a program execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can retrieve and execute a program from a program execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" may be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, a program execution system, apparatus, or device.

[0126] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0127] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable program execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0128] In the above description of this specification, reference to the terms "one embodiment / example," "another embodiment / example," or "certain embodiments / examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples.

[0129] Although the embodiments of the present application have been shown and described, those skilled in the art will appreciate that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and intent of the present application, and that the scope of the present application is defined by the claims and their equivalents.

[0130] The above is a specific description of the preferred implementation of the present application, but the present application is not limited to the embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present application. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present application.

Claims

1. A method for predicting solar irradiance before the day, characterized in that: The following steps are involved: Obtain historical data of solar irradiance before the day and determine the meteorological data corresponding to the optimal parameterization scheme in the WRF model; Determining the meteorological data corresponding to the optimal parameterization scheme in the WRF model specifically includes: in a stage of combined testing and optimization of all parameterization schemes of the WRF model, systematically testing the parameterization schemes of the surface and the planetary boundary layer to obtain a plurality of first optimized parameterization schemes; based on fixed surface and planetary boundary layer schemes, testing a combination of a fixed radiation transfer model, a microphysical model, and a convection parameterization scheme to screen out an optimal parameterization scheme from the plurality of first optimized parameterization schemes; wherein the number of first optimized parameterization schemes is less than the number of all parameterization schemes; and the number of optimal parameterization schemes is less than the number of the first optimized parameterization schemes; and running the optimal parameterization scheme to obtain corresponding meteorological data; Determining a weather label for the forecast day based on the historical data; determining the weather label for the forecast day based on the historical data specifically includes: determining a daily average cloud cover coefficient based on the historical data; determining a first weather label corresponding to each daily average cloud cover coefficient based on a preset neural network model; and using all of the first weather labels as all of the weather labels for the forecast day; updating the parameters of the day-ahead solar irradiance prediction network according to the historical data, the meteorological data, and the weather label to obtain a trained solar irradiance prediction model; The historical data is input into the solar irradiance prediction model to obtain the target day-ahead solar irradiance of the prediction day.

2. The method for predicting solar irradiance before the day according to claim 1, wherein: Determining the daily average cloud cover coefficient based on the historical data specifically includes: extracting a first irradiance from the historical data; Based on the preset theoretical clear sky irradiance and the first irradiance, the daily average cloud cover coefficient is determined.

3. The method for predicting solar irradiance before the day according to claim 2, wherein: The determining of the daily average cloud cover coefficient based on the preset theoretical clear sky irradiance and the first irradiance specifically includes: Taking the first irradiance and the theoretical clear sky irradiance as a quotient to obtain a first quotient; The first quotient is used as the daily average cloud cover coefficient.

4. The method for predicting solar irradiance before the day according to claim 1, wherein: The day-ahead solar irradiance prediction network includes a convolutional neural network, a long short-term memory network, and a Transformer model. The method of updating the parameters of the day-ahead solar irradiance prediction network based on the historical data, the meteorological data, and the weather label to obtain a trained solar irradiance prediction model specifically includes: Inputting the meteorological data into the convolutional neural network to obtain a spatial feature sequence; Inputting the spatial feature sequence and the historical data into the long short-term memory network to obtain a first predicted solar irradiance for the predicted day; Inputting the historical data, the first predicted solar irradiance, and the weather label into the Transformer model to obtain a second predicted solar irradiance; According to the second predicted solar irradiance, the parameters of the day-ahead solar irradiance prediction network are updated for several rounds to obtain a spatial feature sequence.

5. The method for predicting solar irradiance before the day according to claim 1, wherein: The method further comprises: The target day-ahead solar irradiance is compared with a preset day-ahead solar irradiance, and the parameters of the solar irradiance prediction model are adjusted according to the comparison result.

6. A day-ahead solar irradiance prediction system, characterized in that: include: The first processing unit is used to obtain historical data of solar irradiance before the day and determine meteorological data corresponding to the optimal parameterization scheme in the WRF model; Determining the meteorological data corresponding to the optimal parameterization scheme in the WRF model specifically includes: in a stage of combined testing and optimization of all parameterization schemes of the WRF model, systematically testing the parameterization schemes of the surface and the planetary boundary layer to obtain a plurality of first optimized parameterization schemes; based on fixed surface and planetary boundary layer schemes, testing a combination of a fixed radiation transfer model, a microphysical model, and a convection parameterization scheme to screen out an optimal parameterization scheme from the plurality of first optimized parameterization schemes; wherein the number of first optimized parameterization schemes is less than the number of all parameterization schemes; and the number of optimal parameterization schemes is less than the number of the first optimized parameterization schemes; and running the optimal parameterization scheme to obtain corresponding meteorological data; a second processing unit configured to determine a weather label for a forecast day based on the historical data; wherein determining the weather label for the forecast day based on the historical data specifically includes: determining a daily average cloud cover coefficient based on the historical data; determining a first weather label corresponding to each daily average cloud cover coefficient based on a preset neural network model; and using all of the first weather labels as all of the weather labels for the forecast day; a third processing unit, configured to update parameters of a day-ahead solar irradiance prediction network according to the historical data, the meteorological data, and the weather label to obtain a trained solar irradiance prediction model; The fourth processing unit is used to input the historical data into the solar irradiance prediction model to obtain the target day-ahead solar irradiance of the prediction day.

7. A device for predicting solar irradiance before the day, characterized in that include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the day-ahead solar irradiance prediction method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing instructions executable by a processor, characterized in that: The processor-executable instructions are used to execute a method for predicting solar irradiance before the day as described in any one of claims 1 to 5 when executed by the processor.

Images