Sub-season forecasting system and method for solar irradiance
Solar irradiance forecasting is performed through the fusion of acquisition, preprocessing and convolution with long and short-term memory mapping, which solves the problem of large prediction errors in the prior art, and achieves the prediction effect of high accuracy and small root mean square errors.
Patent Information
- Application Number
- CN202510331605.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-20
- Publication Date
- 2025-08-15
AI Technical Summary
The existing sub-seasonal forecasting system has problems such as large prediction errors and low accuracy in solar irradiance forecasting.
The acquisition device is set up to collect solar irradiance-related data, and the data is pre-processed using the pre-processing device, and the prediction is made by combining the fusion of convolution and long-term memory mapping, and finally visualization is performed through the display module.
The sub-seasonal forecast of solar irradiance with high accuracy and small root mean square error is achieved, and has good generalization effect and parameter optimization capabilities.
Smart Images
Figure CN120492791A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of computer models and systems, and in particular to a solar irradiance sub-seasonal forecasting system and method. Background Art
[0002] Subseasonal forecasts are predictions of weather and climate conditions over a timescale of approximately two weeks to two months. Solar irradiance plays a key role in Earth's energy balance and climate system. Accurate subseasonal forecasts of surface solar irradiance are crucial for renewable energy planning and regional climate research. While there is considerable interest in subseasonal forecasts and some progress has been made, the capabilities of Southeast Asia's dynamical forecast systems for more than two weeks and six weeks remain uncertain, and related forecast accuracy and modeling design issues remain unresolved. Validation of forecasts is necessary to improve these systems.
[0003] Although the European Centre for Medium-Range Weather Forecasts Ensemble System (EC-ENS) model is widely used for sub-seasonal forecasts, this sub-seasonal forecast system has large prediction errors, specifically the root mean square error (RMSE), mean absolute error (MAE) and mean bias error (MBE) of 414.43 W / m2, 380.95 W / m2 and -309.72 W / m2, respectively. There is also the problem of low accuracy in solar irradiance forecasts. Summary of the Invention
[0004] The technical problem to be solved by the present invention is how to overcome the technical defects of the sub-seasonal forecast system in the prior art, such as large prediction error and low solar irradiance forecast accuracy. In order to overcome the above defects of the prior art, the present invention provides a solar irradiance sub-seasonal forecast system and method, which include a solar irradiance sub-seasonal forecast system and a solar irradiance sub-seasonal forecast method.
[0005] The present invention provides a solar irradiance sub-seasonal forecast system, comprising: A collection device for collecting solar irradiance data in the area at a frequency of four times per day; the solar irradiance data is a data vector consisting of solar irradiance, solar zenith angle, azimuth, temperature at 2 meters above ground level, humidity, wind speed, wind direction, and air pressure; a preprocessing device, communicating with the acquisition device, and configured to perform data preprocessing on the solar irradiance related data collected by the acquisition device for 30 days before the current day to obtain a data matrix; The prediction device communicates with the preprocessing device and is configured to use the data matrix to obtain the solar irradiance distribution of the area on the same day and 45 days after the same day by a fusion of convolution and long short-term memory mapping, and visualize the obtained solar irradiance distribution in a statistical chart.
[0006] The solar irradiance sub-seasonal forecast system disclosed in the present invention addresses the technical deficiencies listed above. By providing a collection device to collect solar irradiance-related data in the region, the solar irradiance-related data is a data vector consisting of solar irradiance, solar zenith angle, azimuth, temperature at 2 meters above ground level, humidity, wind speed, wind direction, and air pressure. The data is then pre-processed by a pre-processing device to filter and reduce noise on the acquired data. Finally, a prediction device is provided to obtain the solar irradiance distribution for the region for the current day and 45 days thereafter using a data matrix by combining convolution with long-short-term memory mapping, and the obtained solar irradiance distribution is visualized in a statistical chart. Because the prediction method is a fusion of convolution and long-short-term memory mapping, the system has the technical advantages of high prediction accuracy and ease of parameter optimization, as well as good generalization and a small root mean square error, thereby overcoming the technical deficiencies of existing sub-seasonal forecast systems, such as large prediction errors and low solar irradiance forecast accuracy.
[0007] In one possible implementation, the pre-processing device is configured to perform the following steps: A1: Clean the solar irradiance data collected by the acquisition device for the 30 days before the current day to remove abnormal values and missing values, thereby obtaining cleaned data; A2: normalizing the cleaned data to obtain standardized data; A3: sorting the standardized data in time to obtain the data matrix; This solution can remove outliers and achieve data standardization, thereby realizing filtering and noise elimination to ensure accurate and reliable prediction results.
[0008] In one possible implementation, the prediction device includes: a depth prediction module, communicating with the preprocessing device, configured to use the data matrix to obtain the solar irradiance distribution of the area on the current day and 45 days thereafter by fusion of convolution and long short-term memory mapping; a display module, communicating with the depth prediction module and configured to visualize the solar irradiance distribution of the area on the current day and 45 days thereafter obtained by the depth prediction module in the form of a statistical chart; This solution uses a deep prediction module to predict solar irradiance distribution and a display module to display the data, ensuring not only the accuracy of the prediction but also the visualization of the results.
[0009] In a possible implementation, the depth prediction module includes a convolutional layer unit, a maximum pooling layer unit, a long short-term memory network layer unit, and a fully connected layer unit sequentially arranged along a running direction; The convolutional layer unit is a network structure formed by connecting several convolutional layers in series, and each convolutional layer is configured to perform the following mapping operations: A=S(C(X)-b), Where, X represents the input of the convolutional layer; C(X)-b represents the intermediate result, which is the difference between the result of the convolution operation on the input X and the bias term b; A represents the output result of the convolutional layer, which is the mapping result of the S-type activation function on the intermediate result; This solution combines convolutional neural networks with long short-term memory networks and outputs results using fully connected layer units, which not only ensures the feasibility and simplicity of model parameter optimization, but also further ensures improved prediction accuracy.
[0010] In a possible implementation, the display module includes: an icon unit, communicating with the fully connected layer unit, and configured to obtain a statistical chart based on the solar irradiance distribution of the region on the current day and 45 days thereafter by using a chart generation algorithm; The large language model unit communicates with the icon unit and is configured to generate explanatory text for the statistical chart by using a large language model algorithm.
[0011] Another technical solution of the present invention is to provide a sub-seasonal forecast method of solar irradiance, which comprises the following steps: S1; retrieve historical data reflecting the relationship between the solar irradiance data of the region 30 days before the current day and the solar irradiance distribution quality inspection data of the current day and the 45 days after the current day from the database; S2: Preprocessing the historical data to obtain preprocessing results, and dividing the preprocessing results into a training set, a validation set, and a test set; S3; optimizing the parameters of the depth prediction module in the prediction device based on the training set, the validation set, and the test set by using a loss function to obtain an optimized prediction device; S4: Collecting the solar irradiance related data of the area through a collection device, and the collection frequency is four times a day; S5: Preprocessing the solar irradiance data obtained by the acquisition device for the 30 days before the current day by a preprocessing device to obtain a data matrix; S6: The optimized prediction device uses the data matrix in a fusion manner of convolution and long short-term memory mapping to obtain the solar irradiance distribution of the area on the day and 45 days after the day, and visualizes the obtained solar irradiance distribution in a statistical chart.
[0012] The method disclosed in the present invention first constructs a training set, a test set, and a validation set, and optimizes model parameters based on a loss function to obtain an optimized prediction device. A collection device is then used to collect solar irradiance-related data for the region. The data is then preprocessed by a preprocessing device to filter and reduce noise on the obtained data. Finally, the prediction device is configured to use the processing results to obtain the solar irradiance distribution for the region for the current day and 45 days thereafter by combining convolution with long-short-term memory mapping, and the obtained solar irradiance distribution is visualized in a statistical chart. Because the prediction method is a fusion of convolution and long-short-term memory mapping, the method has the technical advantages of high prediction accuracy and ease of parameter optimization, as well as good generalization and a small root mean square error, thereby overcoming the technical defects of existing sub-seasonal forecast systems, such as large prediction errors and low solar irradiance forecast accuracy.
[0013] In one possible implementation, the historical data is the Cartesian product of input data and ideal output data; The input data is the solar irradiance data of the 30 days before the current day, and the collection frequency is four times a day; the solar irradiance data is a data vector consisting of solar irradiance, solar zenith angle, azimuth, temperature at 2 meters above ground level, humidity, wind speed, wind direction, and air pressure; The ideal output data is the actual observation results of the solar irradiance distribution in the area on the day and 45 days after the day; By collecting historical data formed by the above data, we can ensure the accurate correlation between the model and the actual prediction, ensure the rationality of the training, achieve the effect of model parameter optimization, and then ensure that the model after parameter optimization, that is, the optimized prediction device can accurately predict the eye-raising irradiance.
[0014] In a possible implementation, the process of preprocessing the historical data in step S2 includes the following steps: S21: performing data cleaning on the historical data to remove outliers and missing values, thereby obtaining historical cleaned data; S22: normalizing the historical cleaned data to obtain historical standardized data; S23: sorting the historical standardized data by time to obtain the preprocessing result; This solution can remove outliers in historical data to achieve data standardization, thereby realizing filtering and noise elimination, and ensuring the prediction accuracy of model training results.
[0015] In a possible implementation, the normalization process in step S22 is performed by normalizing the historical cleansing data through batch normalization mapping to ensure that the format of the data after normalization is reasonable.
[0016] In a possible implementation, the loss function is a mean square error loss function, which can reduce the mean square error of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 This is a schematic diagram of the structure of a solar irradiance sub-seasonal forecast system disclosed in an embodiment of the present application; Figure 2 Schematic diagram of the depth prediction module structure disclosed in the embodiments of this application; Figure 3 A flow chart of the method disclosed in the embodiments of this application; Figure 4 This is a flowchart of the parameter optimization disclosed in the examples of this application. DETAILED DESCRIPTION
[0018] First, those skilled in the art should understand that these embodiments are merely used to explain the technical principles of the embodiments of the present application and are not intended to limit the scope of protection of the embodiments of the present application. Those skilled in the art may adjust them as needed to suit specific application scenarios.
[0019] In the embodiments of the present application, unless otherwise clearly specified and limited, the communication or communication connection between the first feature and the second feature refers to the transmission of information between the first feature and the second feature. This information transmission can be either unidirectional or bidirectional, and the method of realizing the communication connection can be electrical connection of wires, radio connection, electrical connection of electromagnetic media (such as semiconductors), communication realized by channels, etc.
[0020] In the embodiments of the present application, unless otherwise expressly specified or limited, a first feature being “above,” “below,” “in front of,” or “behind” a second feature may mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Moreover, a first feature being “above,” “above,” or “above” a second feature may mean that the first feature is directly above or diagonally above the second feature, or simply means that the first feature is higher in level than the second feature. A first feature being “below,” “below,” or “below” a second feature may mean that the first feature is directly below or diagonally below the second feature, or simply means that the first feature is lower in level than the second feature. A first feature being “before,” “in front of,” or “in front of” a second feature may mean that the first feature is directly in front of or diagonally in front of the second feature, or simply means that the first feature is prior to the second feature in sequence. A first feature being “behind,” “behind,” or “behind” a second feature may mean that the first feature is directly behind or diagonally behind the second feature, or simply means that the first feature is later in sequence than the second feature.
[0021] The technical solution of the present application is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0022] See also Figures 1 to 4 , the embodiment of the present application discloses a solar irradiance sub-seasonal forecast system, Figure 1 is a schematic diagram of the system structure, as shown in Figure 1 As shown, the system includes a collection device, a preprocessing device and a prediction device. The preprocessing device communicates with the collection device, and the prediction device communicates with the preprocessing device.
[0023] In this system, the collection device is used to collect solar irradiance-related data in the area, and the collection frequency is four times a day; the solar irradiance-related data is a data vector composed of solar irradiance, solar zenith angle, azimuth angle, temperature 2 meters above ground level, humidity, wind speed, wind direction, and air pressure. In this embodiment, the collection device is composed of multiple sensors, which are solar irradiance sensors, solar zenith angle sensors, azimuth angle sensors, temperature sensors, humidity sensors, wind speed sensors, wind direction sensors, and air pressure sensors. The number of each sensor can be set according to regional characteristics and distributed in multiple sites.
[0024] In this system, the preprocessing device is configured to perform data preprocessing on the solar irradiance data collected by the collection device for the previous 30 days to obtain a data matrix. In this embodiment, the preprocessing device is configured to perform the following steps: A1: Cleaning the solar irradiance data collected by the collection device for the previous 30 days to remove outliers and missing values, thereby obtaining cleaned data; outliers can be determined based on a normal distribution screening mechanism; A2: Normalizing the cleaned data to obtain standardized data; the normalization method is batch normalization; A3: Time-sorting the standardized data to obtain a data matrix.
[0025] In this system, the prediction device (CNN-LSTM) is set to use the data matrix to obtain the solar irradiance distribution of the area on the same day and 45 days after the same day through the fusion of convolution and long short-term memory mapping, and visualize the obtained solar irradiance distribution in statistical charts. Figure 1 and Figure 2 In this embodiment, the prediction device includes a depth prediction module and a display module. The depth prediction module communicates with the preprocessing device, and the display module communicates with the depth prediction module.
[0026] In the prediction device, the depth prediction module is configured to use a data matrix to obtain the solar irradiance distribution for the current day and the next 45 days in the region by fusing convolution with long-short-term memory mapping. In this embodiment, the depth prediction module includes a convolutional layer unit, a maximum pooling layer unit, a long-short-term memory network layer unit, and a fully connected layer unit arranged in sequence along the direction of operation; the convolutional layer unit is a network structure formed by connecting several convolutional layers in series, and each convolutional layer is configured to perform the following mapping operation:
[0027] A=S(C(X)-b), Where, X represents the input of the convolutional layer; C(X)-b represents the intermediate result, which is the difference between the result of the convolution operation on the input X and the bias term b; A represents the output of the convolutional layer, which is the mapping result of the S-type activation function on the intermediate result.
[0028] Figure 2 This is a schematic diagram of the depth prediction module structure disclosed in this embodiment, which includes an input layer, a convolutional layer, a maximum pooling layer, a long short-term memory network layer unit (LSTM) and a fully connected layer unit, and the fully connected layer unit includes a fully connected layer.
[0029] In the prediction device, the display module is configured to visualize the solar irradiance distribution of the region on the current day and 45 days thereafter, as obtained by the depth prediction module, in the form of statistical charts. In this embodiment, the display module includes an icon unit and a large language model unit. The icon unit communicates with the fully connected layer unit. The icon unit is configured to obtain statistical charts based on the solar irradiance distribution of the region on the current day and 45 days thereafter through a chart generation algorithm. The statistical charts include line charts, box plots, bar charts, and stem-and-leaf plots. The large language model unit communicates with the icon unit. The large language model unit is configured to generate explanatory text for the statistical charts through a large language model algorithm. The method for generating text is the existing technology, which is captured and expanded again.
[0030] See also Figure 3 The following further introduces the forecasting method corresponding to the solar irradiance sub-seasonal forecasting system disclosed in this embodiment. The method includes the following steps: S1; retrieve historical data from the meteorological bureau database reflecting the relationship between the solar irradiance data for the 30 days before the day and the solar irradiance distribution quality inspection data for the day and the 45 days after the day in the region.
[0031] In this embodiment, the historical data referred to in step S1 is the Cartesian product of the input data and the ideal output data. The input data is solar irradiance data for the 30 days preceding the current day, collected four times daily. The solar irradiance data is a data vector consisting of solar irradiance, solar zenith angle, azimuth, temperature at 2 meters above ground level, humidity, wind speed, wind direction, and air pressure. The ideal output data is the actual observations of solar irradiance distribution for the current day and the 45 days following the current day, obtained from the regional meteorological bureau database.
[0032] S2: Preprocess the historical data to obtain preprocessing results, and divide the preprocessing results into training set, validation set and test set.
[0033] In this embodiment, the process of preprocessing historical data in step S2 includes the following steps: S21: performing data cleaning on the historical data to remove outliers and missing values to obtain historical cleaned data; S22: normalizing the historical cleaned data to obtain historical standardized data; the normalization method is to normalize the historical cleaned data through batch normalization mapping; S23: sorting the historical standardized data by time to obtain the preprocessing result.
[0034] S3: Optimize the parameters of the depth prediction module in the prediction device based on the training set, validation set and test set through the loss function to obtain the optimized prediction device. The loss function used in this embodiment is the mean square error loss function to reduce the mean square error of the prediction. The flow chart of the model parameter optimization is as follows: Figure 4 As shown in the table below, the obtained parameters are as follows:
[0035] Parameter name Parameter value Batch size 128 Number of CNN convolutional layers 1 Number of CNN pooling layers 1 Number of LSTM layers 2 LSTM hidden layer nodes 128 Time window length 45 days Learning rate 0.001 Number of training rounds 500 The training method is an existing technology and will not be expanded here.
[0036] S4: Collecting the solar irradiance related data of the area through a collection device, and the collection frequency is four times a day.
[0037] S5: The solar irradiance related data obtained by the acquisition device for the 30 days before the current day are preprocessed by the preprocessing device to obtain a data matrix.
[0038] S6: The optimized prediction device uses the data matrix in a fusion of convolution and long short-term memory mapping to obtain the solar irradiance distribution in the area on the same day and 45 days after the same day, and visualizes the solar irradiance distribution in statistical charts.
[0039] The solar irradiance sub-seasonal forecast system disclosed in this embodiment addresses the technical deficiencies listed above by providing a collection device to collect solar irradiance-related data in the region, where the solar irradiance-related data is a data vector consisting of solar irradiance, solar zenith angle, azimuth, temperature at 2 meters above ground level, humidity, wind speed, wind direction, and air pressure; then, a pre-processing device is provided to pre-process the data to filter and reduce noise, and finally, a prediction device is provided to obtain the solar irradiance distribution of the region for the current day and 45 days thereafter using a data matrix by combining convolution with long-short-term memory mapping, and the obtained solar irradiance distribution is visualized in a statistical chart. Because the prediction method is a combination of convolution and long-short-term memory mapping, it has the technical advantages of high prediction accuracy and ease of parameter optimization, and has good generalization effect and a small root mean square error, thereby overcoming the technical deficiencies of existing sub-seasonal forecast systems, such as large prediction errors and low solar irradiance forecast accuracy.
[0040] In the description of the embodiments of the present application, it should be noted that in the description of the present application, terms such as "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or component must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present application.
[0041] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "in the present embodiment", "specific example", or "some examples" means that the specific features, mechanisms, 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 representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, mechanisms, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.
[0042] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.
Claims
1. A solar irradiance sub-seasonal forecast system, characterized in that: include: A collection device for collecting solar irradiance data in the area at a frequency of four times per day; the solar irradiance data is a data vector consisting of solar irradiance, solar zenith angle, azimuth, temperature at 2 meters above ground level, humidity, wind speed, wind direction, and air pressure; a preprocessing device, communicating with the acquisition device, and configured to perform data preprocessing on the solar irradiance related data collected by the acquisition device for 30 days before the current day to obtain a data matrix; The prediction device communicates with the preprocessing device and is configured to use the data matrix to obtain the solar irradiance distribution of the area on the same day and 45 days after the same day by a fusion of convolution and long short-term memory mapping, and visualize the obtained solar irradiance distribution in a statistical chart.
2. The solar irradiance sub-seasonal forecast system according to claim 1, characterized in that: The pre-processing device is configured to perform the following steps: A1: Clean the solar irradiance data collected by the acquisition device for the 30 days before the current day to remove abnormal values and missing values, thereby obtaining cleaned data; A2: normalizing the cleaned data to obtain standardized data; A3: Time-sorting the normalized data to obtain the data matrix.
3. The solar irradiance sub-seasonal forecast system according to claim 2, characterized in that: The prediction device comprises: a depth prediction module, communicating with the preprocessing device, configured to use the data matrix to obtain the solar irradiance distribution of the area on the current day and 45 days thereafter by fusion of convolution and long short-term memory mapping; The display module communicates with the depth prediction module and is configured to visualize the solar irradiance distribution of the area obtained by the depth prediction module on the current day and 45 days after the current day in a statistical chart.
4. The solar irradiance sub-seasonal forecast system according to claim 3, characterized in that: The depth prediction module includes a convolutional layer unit, a maximum pooling layer unit, a long short-term memory network layer unit and a fully connected layer unit arranged in sequence along the running direction; The convolutional layer unit is a network structure formed by connecting several convolutional layers in series, and each convolutional layer is configured to perform the following mapping operations: A=S(C(X)-b), Where, X represents the input of the convolutional layer; C(X)-b represents the intermediate result, which is the difference between the result of the convolution operation on the input X and the bias term b; A represents the output result of the convolutional layer, which is the mapping result of the S-type activation function on the intermediate result.
5. The solar irradiance sub-seasonal forecast system according to claim 4, characterized in that: The display module includes: an icon unit, communicating with the fully connected layer unit, and configured to obtain a statistical chart based on the solar irradiance distribution of the region on the current day and 45 days thereafter by using a chart generation algorithm; The large language model unit communicates with the icon unit and is configured to generate explanatory text for the statistical chart by using a large language model algorithm.
6. A method for sub-seasonal solar irradiance forecasting, characterized in that: The solar irradiance sub-seasonal forecast system according to any one of claims 1 to 5 comprises the following steps: S1; retrieve historical data reflecting the relationship between the solar irradiance data of the region 30 days before the current day and the solar irradiance distribution quality inspection data of the current day and the 45 days after the current day from the database; S2: Preprocessing the historical data to obtain preprocessing results, and dividing the preprocessing results into a training set, a validation set, and a test set; S3; optimizing the parameters of the depth prediction module in the prediction device based on the training set, the validation set, and the test set by using a loss function to obtain an optimized prediction device; S4: Collecting the solar irradiance related data of the area through a collection device, and the collection frequency is four times a day; S5: Preprocessing the solar irradiance data obtained by the acquisition device for the 30 days before the current day by a preprocessing device to obtain a data matrix; S6: The optimized prediction device uses the data matrix in a fusion manner of convolution and long short-term memory mapping to obtain the solar irradiance distribution of the area on the day and 45 days after the day, and visualizes the obtained solar irradiance distribution in a statistical chart.
7. The method for sub-seasonal solar irradiance forecasting according to claim 6, characterized in that: The historical data is the Cartesian product of the input data and the ideal output data; The input data is the solar irradiance data of the 30 days before the current day, and the collection frequency is four times a day; the solar irradiance data is a data vector consisting of solar irradiance, solar zenith angle, azimuth, temperature at 2 meters above ground level, humidity, wind speed, wind direction, and air pressure; The ideal output data is the actual observation results of the solar irradiance distribution in the area on the day and 45 days after the day.
8. The solar irradiance sub-seasonal forecast method according to claim 6 or 7, wherein: The process of pre-processing the historical data in step S2 includes the following steps: S21: performing data cleaning on the historical data to remove outliers and missing values, thereby obtaining historical cleaned data; S22: normalizing the historical cleaned data to obtain historical standardized data; S23: Time-sorting the historical standardized data to obtain the preprocessing result.
9. The method for sub-seasonal solar irradiance forecasting according to claim 8, characterized in that: The normalization process in step S22 is performed by normalizing the historical cleansing data through batch normalization mapping.
10. The method for sub-seasonal solar irradiance forecasting according to claim 9, characterized in that: The loss function is a mean square error loss function.