A method for predicting total and diffuse radiation on the horizontal surface based on weather types

Abstract

This invention provides a method for predicting total horizontal radiation and diffuse radiation based on weather types. Weather types are classified based on sunshine percentage. For each weather type, the total horizontal radiation calculation model and the direct-diffuse separation model with the smallest prediction error are selected. Meteorological environmental factors with high correlation to weather types are further modified by introducing them into the direct-diffuse separation model, thus improving the accuracy of total horizontal radiation and diffuse radiation prediction. Specifically, because the method of this invention is based on sunshine percentage for weather type classification, it eliminates the need to calculate the clarity index based on total horizontal radiation. Furthermore, by utilizing a cascaded model, the total horizontal radiation predicted by the optimal total horizontal radiation calculation model for the region is used as the input to the next-level optimal direct-diffuse separation model. Therefore, the method of this invention is applicable to areas without total horizontal radiation observation, providing accurate predictions for these areas and thus having wider applicability.

Description

Technical Field

[0001] This invention belongs to the field of photovoltaic power generation technology, specifically relating to a method for predicting total horizontal radiation and diffuse radiation based on weather patterns. Background Technology

[0002] Over the past few decades, with the development of renewable energy, especially the application of solar energy, my country's new energy development has shown strong momentum. my country will strengthen its energy self-sufficiency and continue to expand the utilization of clean energy, with renewable energy power generation becoming the main power source. As of the first half of 2022, China's total installed photovoltaic power generation capacity reached 336.204 million kilowatts, a year-on-year increase of 24.4%. Newly installed photovoltaic power generation capacity reached 30.878 million kilowatts, a year-on-year increase of 137.4%.

[0003] China boasts a vast land area of ​​9.6 million square kilometers, rich in solar energy resources, especially in the western region, which is the richest in solar energy resources in the country. However, my country has relatively few solar radiation observation stations. Compared with more than 2,400 conventional meteorological element observation stations, there are only 99 national-level radiation operation stations. The small number and uneven distribution of these stations result in a serious lack of total horizontal radiation data, making it difficult to conduct a full and effective assessment of solar energy resources.

[0004] Furthermore, scattered radiation data serves as a crucial basis for calculating solar radiation, holding significant importance for solar energy utilization, including calculations of solar radiation on inclined surfaces and predictions of photovoltaic power generation. However, due to limitations in observation equipment and methods, only a few stations can conduct radiation observations, and most stations can only provide data on total horizontal radiation, lacking observation data on direct radiation, scattered radiation, and total inclined surface radiation, thus failing to meet the needs of scientific research and practical applications. Therefore, seeking methods to extrapolate solar scattered radiation from measured observation data is particularly urgent and important.

[0005] Over the past few decades, many researchers have developed a series of direct-scatter separation models to calculate diffuse radiation. These models primarily utilize the correlation between the ratio of diffuse radiation to total radiation on the horizontal surface and conventional meteorological factors (clarity index and sunshine duration) to establish a regression equation; this model is also known as the scattering ratio model. However, the accuracy of these models is limited, and the accuracy varies significantly under different weather types. Summary of the Invention

[0006] This invention addresses the aforementioned problems by providing a method for predicting total horizontal radiation and scattered radiation that can further improve prediction accuracy, while ensuring prediction accuracy even in the absence of accurate historical total horizontal radiation observations. The invention employs the following technical solution:

[0007] This invention provides a method for predicting total and scattered radiation on a horizontal surface based on weather patterns, characterized by the following steps:

[0008] Step S1: Obtain historical solar radiation data, historical astronomical data, and historical meteorological data for the predetermined area, wherein the historical solar radiation data includes at least historical diffuse radiation, and the historical meteorological data includes multiple meteorological environmental factors;

[0009] Step S2: Calculate the sunshine percentage of the region based on the historical astronomical data and the historical meteorological data, and classify the weather of the region into multiple weather types based on the sunshine percentage;

[0010] Step S3: Introduce different meteorological environmental factors into the horizontal total radiation calculation model and perform localized correction. For each weather type, select the corrected horizontal total radiation calculation model with the smallest prediction error as its optimal horizontal total radiation calculation model, and use the optimal horizontal total radiation calculation model to predict the horizontal total radiation of the region.

[0011] Step S4: Introduce the historical weather data into the direct-divergence separation model and perform localization correction on it. For each weather type, select the corrected direct-divergence separation model with the smallest prediction error as its localized direct-divergence separation model.

[0012] Step S5: Perform correlation analysis on the scattering ratio and the historical meteorological data under each weather type, and introduce the relevant meteorological environmental factors into the localized direct-scatter separation model to correct it based on the analysis results, so as to obtain the optimal direct-scatter separation model under each weather type.

[0013] Step S6: Combine the optimal horizontal total radiation calculation model and the optimal direct-scatter separation model into a cascaded model, and use the cascaded model to predict the scattered radiation of the region.

[0014] The method for predicting total horizontal radiation and diffuse radiation based on weather classification provided by this invention also has the following technical feature: In step S2, the formula for calculating the percentage of sunshine is:

[0015]

[0016] In the formula, n represents the number of hours of sunshine, and N represents the number of hours of available sunshine.

[0017] Based on the sunshine percentage, the weather in this region is divided into several weather types, including: when 1 > S p When the value is ≥0.6, the corresponding weather is sunny, sunny turning cloudy, or cloudy turning sunny, which is classified as weather type 1; when 0.6>S pWhen ≥0.1, the corresponding weather is cloudy, overcast turning cloudy, or multi-level turning overcast, classified as weather type 2; when 0.1>S p When the value is ≥0, the corresponding weather is rain, snow, fog, or haze, and it is classified as weather type 3.

[0018] The method for predicting total horizontal radiation and diffuse radiation based on weather classification provided by this invention also has the following technical feature: the formula for calculating the number of hours of sunshine is as follows:

[0019]

[0020] In the formula, δ represents the latitude of the region, and δ represents the declination angle of the region.

[0021] The method for predicting total horizontal radiation and diffuse radiation based on weather types provided by this invention may also have the following technical features, wherein, in step S3, the expression for the optimal total horizontal radiation calculation model for weather type 1 is:

[0022]

[0023] The expression for the optimal horizontal total radiation calculation model for weather type 2 is as follows:

[0024]

[0025] The expression for the optimal horizontal total radiation calculation model for weather type 3 is as follows:

[0026]

[0027] In the formula, H0 is the solar radiation on the horizontal surface outside the atmosphere, also known as astronomical radiation, V is visibility, C is total cloud cover, API is the air pollution index, ΔT is the diurnal temperature range, and R... h denoted as relative humidity, and a0, b0, c0, and d0 are empirical coefficients.

[0028] The method for predicting total horizontal radiation and diffuse radiation based on weather types provided by this invention also has the following technical feature: in step S4, the localized direct and diffuse separation models for weather type 1 and weather type 3 are both modified Jiang models, and their expressions are as follows:

[0029]

[0030] The localized direct dispersion separation model for weather type 2 is the modified El-Sebaii model, and its expression is:

[0031]

[0032] In the formula, H d For scattered radiation, k T The resolution index is represented by a1, b1, c1, d1, e1, f1, and g1, which are empirical coefficients.

[0033] The method for predicting total horizontal radiation and scattered radiation based on weather classification provided by this invention may also have the following technical feature, wherein the formula for calculating the clarity index is:

[0034]

[0035] In the formula, H represents the total radiation of the horizontal plane, which is calculated using the optimal total radiation calculation model for the horizontal plane.

[0036] The method for predicting total and scattered radiation on a horizontal surface based on weather classification provided by this invention may also have the following technical features, wherein, in step S5,

[0037] For weather type 1, based on the corresponding localized direct dispersion separation model, a GPR model is constructed using the GPR algorithm, and the dimensionality of the GPR model is reduced to 3 factors using principal component analysis to establish a PCA-GPR model. Then, the clarity index, sunshine percentage, visibility, relative humidity, and air pollution index of the region are introduced into the PCA-GPR model for training. The trained model is the optimal direct dispersion separation model under weather type 1.

[0038] For weather type 2, based on the corresponding localized direct dispersion separation model, the GPR algorithm is used to construct a GPR model, and the clarity index, sunshine percentage and total cloud cover of the region are introduced into the GPR model for training. The trained model is the optimal direct dispersion separation model under weather type 2.

[0039] For weather type 3, based on the corresponding localized direct dispersion separation model, the GPR algorithm is used to construct a GPR model, and the clarity index and sunshine percentage of the region are introduced into the training of the GPR model. The trained model is the optimal direct dispersion separation model under weather type 3.

[0040] Invention Function and Effect

[0041] The method for predicting total horizontal radiation and diffuse radiation based on weather type according to the present invention classifies weather types based on sunshine percentage. For each weather type, the total horizontal radiation calculation model and the direct-diffuse separation model with the smallest prediction error are selected. Meteorological environmental factors with high correlation to weather type are further modified by introducing them into the direct-diffuse separation model, thus improving the accuracy of total horizontal radiation and diffuse radiation prediction. In particular, since the method of the present invention classifies weather types based on sunshine percentage, it does not require calculating the clarity index based on total horizontal radiation. Furthermore, by using a cascaded model, the total horizontal radiation predicted by the optimal total horizontal radiation calculation model for the region is used as the input to the next-level optimal direct-diffuse separation model. Therefore, the method of the present invention is applicable to areas without total horizontal radiation observation, providing accurate predictions for these areas and thus having greater applicability. Attached Figure Description

[0042] Figure 1 This is a flowchart of the method for predicting total horizontal radiation and scattered radiation based on weather classification in an embodiment of the present invention;

[0043] Figure 2 This is a graph showing the change of the sharpness index with the percentage of sunshine in an embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram comparing the horizontal total radiation prediction of two models in an embodiment of the present invention;

[0045] Figure 4 This is a schematic diagram comparing the scattered radiation prediction of three models in an embodiment of the present invention. Detailed Implementation

[0046] To make the technical means, creative features, objectives and effects of this invention easy to understand, the following describes in detail the method for predicting total horizontal radiation and scattered radiation based on weather classification.

[0047] <Example>

[0048] Figure 1 This is a flowchart of the method for predicting total horizontal radiation and scattered radiation based on weather classification in this embodiment.

[0049] like Figure 1 As shown, the method for predicting total and diffuse radiation on a horizontal surface based on weather patterns includes the following steps:

[0050] Step S1: Obtain historical solar radiation data, historical astronomical data, and historical meteorological data for a certain region;

[0051] Step S2: Calculate the sunshine percentage of the region and classify the weather into multiple weather types based on the sunshine percentage;

[0052] Step S3: Localize the total horizontal radiation calculation model and select the optimal total horizontal radiation calculation model for each weather type.

[0053] Step S4: Localize and modify the direct dispersion separation model, and select the optimal direct dispersion separation model for each weather type.

[0054] Step S5: Introduce highly correlated meteorological factors under each weather type to further modify the direct-divergence separation model and obtain the optimal direct-divergence separation model under each weather type.

[0055] Step S6: Combine the optimal horizontal total radiation calculation model and the optimal direct dispersion separation model into a cascaded model, and use this cascaded model to predict the horizontal scattered radiation of the region.

[0056] The steps described above will be explained in detail below.

[0057] Step S1: Obtain historical solar radiation data, historical astronomical data, and historical meteorological data for the designated area.

[0058] The historical solar radiation data includes at least the historical diffuse radiation and historical direct radiation of the region, and optionally, the historical total horizontal radiation of the region. That is, this method is applicable to regions where historical total horizontal radiation observations are unavailable.

[0059] In this embodiment, taking Beijing as an example, daily-scale solar radiation data (including total horizontal radiation, diffuse radiation, direct radiation, etc.), astronomical data, and meteorological data (including total cloud cover C, visibility V (km), sunshine duration n (h), and maximum temperature T) for Beijing from 2001 to 2012 are obtained. max (°C), lowest temperature T min (°C), daily temperature range ΔT, relative humidity R h (%), Air Pollution Index (API), etc. After obtaining the above data, quality checks and controls are performed, which is a common practice in the field and will not be elaborated upon further.

[0060] The Air Pollution Index (API) is a quantitative scale method for reflecting and evaluating air quality. It simplifies several air pollutants into conceptual index values, mainly including sulfur dioxide, nitrogen oxides, and inhalable suspended particulate matter (PM10). It can comprehensively assess the degree of air pollution and has a significant impact on solar radiation under clear, cloudless conditions.

[0061] Step S2: Calculate the sunshine percentage of the region based on historical astronomical and meteorological data, and classify the weather of the region into multiple weather types based on the sunshine percentage.

[0062] In existing technologies, weather is classified into multiple types based on the clarity index. The clarity index represents the transparency of the atmosphere and is closely related to weather conditions and solar radiation. Its calculation formula is as follows:

[0063]

[0064] In the formula, H represents the total radiation on the horizontal surface, and H0 represents the solar radiation on the horizontal surface outside the atmosphere, also known as astronomical radiation, which is calculated using the following formula:

[0065]

[0066] In the formula, ω s E represents the latitude and sunrise / sunset angle of the region, respectively. SC γ, δ, and δ are the solar constant, the correction value of the solar radiation flux at the upper boundary of the atmosphere caused by the change in the Earth-Sun distance, and the declination angle, respectively. Their calculation formulas are as follows:

[0067] E SC =1367±7W / m 2

[0068]

[0069]

[0070] In the formula, t d It represents the date sequence within a year.

[0071] Existing research classifies weather conditions into three types based on the clarity index:

[0072] When k T When the value is ≥0.5, it is defined as weather type 1, which includes sunny, sunny turning cloudy, and cloudy turning sunny;

[0073] When 0.5 > k T When the value is ≥0.2, it is defined as weather type 2, which includes cloudy, overcast to cloudy, and cloudy to overcast;

[0074] When 0.1 > k T When it is defined as weather type 3, it includes severe weather such as rain, snow, and haze.

[0075] However, due to the lack of historical total horizontal radiation observation data for some areas, the clarity index parameter cannot be obtained. To address this issue, this embodiment uses sunshine percentage to classify weather types.

[0076] The formula for calculating the percentage of sunshine is:

[0077]

[0078] In the formula, n represents the sunshine duration and N represents the available sunshine duration. The calculation formula is as follows:

[0079]

[0080] Figure 2 This is a graph showing the change in clarity index with the percentage of sunshine in this embodiment.

[0081] like Figure 2 As shown, there is a certain correspondence between the sunshine percentage and the clarity index. By establishing a linear equation between the sunshine percentage and the clarity index, the weather type classification method based on the clarity index can be transformed into a weather type classification method based on the sunshine percentage.

[0082] In this embodiment, the sunshine percentage, combined with data from the Beijing area, is used to classify weather types, as shown in Table 1:

[0083] Table 1. Weather Type Classification Based on Sunshine Percentage

[0084]

[0085]

[0086] The segmented intervals in Table 1 are derived from data for the Beijing area. Using the same segmentation method and approach, segmented intervals for other regions can be obtained similarly. When applied to other regions, S p The upper limit remains unchanged at 1, while other thresholds are determined based on S. p and k T The correlation coefficient has changed slightly.

[0087] Step S3: Introduce different meteorological environmental factors into the horizontal total radiation calculation model and perform localization correction. For each weather type, compare the prediction errors of each corrected horizontal total radiation calculation model and select the one with the smallest prediction error as its optimal horizontal total radiation calculation model.

[0088] Among the various models for calculating total horizontal radiation, empirical models are widely used due to their simple structure and high computational accuracy. Most studies on total horizontal radiation are based on models composed of astronomical radiation and sunshine percentage (Angstrom-Prescott model). However, in addition to sunshine percentage, meteorological environmental factors such as air pollutants, visibility, total cloud cover, diurnal temperature range, and relative humidity also significantly influence observed values ​​of total horizontal radiation. Therefore, in this embodiment, based on the Angstrom-Prescott model, meteorological environmental factors highly correlated with various weather types are introduced to construct a simple linear model, which is then used to calculate total horizontal radiation. The localized and corrected calculation model for total horizontal radiation is shown in Table 2 below.

[0089] Table 2 Formulas for Calculating Total Radiation on Localized Horizontal Surface

[0090]

[0091]

[0092] Among them, model GM1 incorporates visibility V and air pollution index API for correction; GM2 incorporates total cloud cover C for correction; and GM3 incorporates diurnal temperature range ΔT and relative humidity R. h The model was modified, and its empirical coefficients a0, b0, c0, and d0 were obtained. A comparative analysis of the prediction errors of the empirical model and the three localized models under various weather types is shown in Table 3 below.

[0093] Table 3. Prediction Error Table of Localized Horizontal Surface Total Radiation Calculation Model

[0094]

[0095] In the table, the formulas for calculating the Mean Absolute Percentage Error (MAPE), Normalized Root Mean Square Error (NRMSE), and Correlation Coefficient (CORR) are as follows:

[0096]

[0097]

[0098]

[0099] In the formula, N is the number of samples in the set, and P f P is the predicted value. o This is the actual value.

[0100] Since different models perform differently under different weather types, based on the prediction error analysis results, the model with the smallest error under different weather types is selected as the optimal model for calculating total horizontal radiation, and its prediction error is evaluated. The selection and evaluation results are shown in Table 4 below:

[0101] Table 4. Error Analysis of the Optimal Horizontal Total Radiation Calculation Model under Various Weather Types

[0102]

[0103] As shown in Table 4, in weather type 1, which is mainly sunny, the introduction of visibility and air pollution index has a good effect; for cloudy weather (weather type 2), total cloud cover has a good contribution to modeling; for severe weather, the introduction of diurnal temperature range and relative humidity can reduce errors.

[0104] Figure 3 The comparison between the Angstrom-Prescott model, the optimal horizontal total radiation calculation model of this embodiment, and the observed values ​​is shown. It can be seen intuitively that, compared with the Angstrom-Prescott model, the horizontal total radiation calculated by the optimal model of this embodiment is closer to the observed values ​​and has better performance.

[0105] Step S4: Introduce historical meteorological data of the region into the direct-divergence separation model for localization correction. For each weather type, compare the prediction errors of each corrected direct-divergence separation model and select the one with the smallest prediction error as its localized direct-divergence separation model.

[0106] In this embodiment, multiple direct-to-dispersion separation models were locally modified and empirical coefficients a1-g1 were obtained, as shown in Table 5 below:

[0107] Table 5 Formulas for Localized Direct-Dispersion Separation Model

[0108]

[0109] For areas without total horizontal radiation observation, the sharpness index k T The total surface radiation of the water surface can be calculated based on the prediction obtained in step S3.

[0110] The prediction errors of the above models are compared and analyzed, as shown in Table 6 below:

[0111] Table 6 Prediction Error Table of Localized Direct Spray Separation Model

[0112]

[0113]

[0114] Furthermore, based on the prediction error analysis results, the model with the smallest error under different weather types was selected as the optimal localization model, and its prediction error was evaluated. The selection and evaluation results are shown in Table 7 below:

[0115] Table 7 Error Analysis of the Optimal Localized Direct Propagation Separation Model under Various Weather Types

[0116]

[0117] In Tables 5-7, each model refers to the model after localization correction.

[0118] As shown in Table 7, the modified Jiang model and the modified El-Sebaii model performed well overall, that is, the model based on two variables, clarity index and sunshine percentage, performed better than the model based on a single variable.

[0119] Step S5: Perform correlation analysis on scattering ratio and historical meteorological data under each weather type, and based on the analysis results, introduce highly correlated meteorological environmental factors into the localized direct-scatter separation model to correct it, thereby obtaining the optimal direct-scatter separation model under each weather type.

[0120] The scattering ratio, which is historical scattered radiation / historical total horizontal radiation, can be calculated using the above model for areas without total horizontal radiation observation.

[0121] For weather type 1, based on the corresponding localized direct-to-dispersion separation model (the aforementioned localized Jiang model), the GPR algorithm is used to construct the GPR model, and the dimensionality is reduced to 3 factors through principal component analysis, thus establishing the PCA-GPR model. The clarity index, sunshine percentage, visibility, relative humidity, and API of the Beijing area are introduced into the training of the PCA-GPR model, and the trained model is denoted as DM1.

[0122] Principal Component Analysis (PCA) is a mathematical dimensionality reduction method that recombines many correlated and collinear variables into a new set of independent variables. Gaussian Regression (GPR) is a machine learning model that fits a Gaussian distribution to the data, then generalizes this distribution to an infinite-dimensional Gaussian distribution in the continuous domain, resulting in a Gaussian process. Gaussian process regression solves regression problems by fitting a corresponding Gaussian process to finite high-dimensional data. By selecting an appropriate kernel function to determine the covariance matrix, an infinite-dimensional Gaussian distribution in the variable space can be determined based on the existing data, thus obtaining a regression model that can be used for prediction.

[0123] For weather type 2, based on the corresponding localized direct dispersion separation model (the aforementioned localized El-Sebaii model), the GPR algorithm is used to construct a GPR model, and the clarity index, sunshine percentage and total cloud cover of Beijing area are introduced into the training of the GPR model. The trained model is denoted as DM2.

[0124] For weather type 3, based on the corresponding localized direct-to-dispersion model (the aforementioned localized Jiang model), the GPR algorithm is used to construct a GPR model, and the clarity index and sunshine percentage of the Beijing area are introduced into the training of the GPR model. The trained model is denoted as DM3.

[0125] The DM1 to DM3 models are the optimal direct-divergence separation models for weather types 1 to 3, respectively.

[0126] Error analysis was performed on models DM1 to DM3, and the results are shown in Table 8 below:

[0127] Table 8. Error Analysis of the Optimal Direct Separation Model for Each Weather Type

[0128]

[0129] Figure 4 The comparison between the original Jiang model, the original El-Sebaii model, the optimal localized direct-scatter separation model of this embodiment, and the observed values ​​is shown. It can be seen intuitively that, compared with the two existing models, the scattered radiation calculated by the optimal localized model of this embodiment is closer to the observed values ​​and has better performance.

[0130] Step S6: Combine the optimal localized total horizontal radiation calculation model (obtained in step S3) and the optimal direct-scatter separation model (obtained in step S5) under each weather type into a cascaded model, and use the cascaded model to predict the total horizontal radiation and scattered radiation of the region.

[0131] In this embodiment, the total horizontal radiation and scattered radiation of the Beijing area are calculated using a cascaded model, and error analysis is performed on them. The results are shown in Table 9 below:

[0132] Table 9 Error Analysis of Cascaded Models for Each Weather Type

[0133]

[0134] In this embodiment, the parts not described in detail are well-known technologies in the art.

[0135] Functions and effects of the embodiments

[0136] The method for predicting total horizontal radiation and diffuse radiation based on weather types provided in this embodiment classifies weather types based on sunshine percentage. For each weather type, the total horizontal radiation calculation model and the direct-diffuse separation model with the smallest prediction error are selected. Meteorological environmental factors with high correlation to weather types are introduced into the direct-diffuse separation model for further correction, thus improving the accuracy of total horizontal radiation and diffuse radiation prediction. In particular, since the method of this invention classifies weather types based on sunshine percentage, it does not require calculating the clarity index based on total horizontal radiation. Furthermore, by using a cascaded model, the total horizontal radiation predicted by the optimal total horizontal radiation calculation model for the region is used as the input to the next-level optimal direct-diffuse separation model. Therefore, the method of this invention is applicable to areas without total horizontal radiation observation, providing accurate predictions for these areas and thus having greater applicability.

[0137] In this embodiment, the weather is divided into three categories based on the percentage of sunshine. Based on the characteristics of each weather type, different meteorological environmental factors are introduced into the model for targeted correction. Therefore, better prediction accuracy can be achieved under various weather types. Compared with several existing horizontal total radiation calculation models and direct radiation separation models, the model in this embodiment further improves the prediction accuracy.

[0138] The above embodiments are only used to illustrate specific implementations of the present invention, and the present invention is not limited to the scope of the description of the above embodiments.

Claims

1. A method for predicting total and scattered radiation on a horizontal surface based on weather patterns, characterized in that, Includes the following steps: Step S1: Obtain historical solar radiation data, historical astronomical data, and historical meteorological data for the predetermined area, wherein the historical solar radiation data includes at least historical diffuse radiation, and the historical meteorological data includes multiple meteorological environmental factors; Step S2: Calculate the sunshine percentage of the region based on the historical astronomical data and the historical meteorological data, and classify the weather of the region into multiple weather types based on the sunshine percentage; Step S3: Introduce different meteorological environmental factors into the horizontal total radiation calculation model and perform localized correction. For each weather type, select the corrected horizontal total radiation calculation model with the smallest prediction error as its optimal horizontal total radiation calculation model, and use the optimal horizontal total radiation calculation model to predict the horizontal total radiation of the region. Step S4: The historical meteorological data is introduced into the direct-divergence separation model for localization correction. For each weather type, the corrected direct-divergence separation model with the smallest prediction error is selected as its localized direct-divergence separation model. Step S5: Perform correlation analysis on the scattering ratio and the historical meteorological data under each weather type, and introduce the relevant meteorological environmental factors into the localized direct-scatter separation model to correct it based on the analysis results, so as to obtain the optimal direct-scatter separation model under each weather type. Step S6: Combine the optimal horizontal total radiation calculation model and the optimal direct-scatter separation model into a cascaded model, and use the cascaded model to predict the scattered radiation of the region.

2. The method for predicting total and scattered radiation on a horizontal surface based on weather patterns according to claim 1, characterized in that: in, In step S2, the formula for calculating the percentage of sunshine is: In the formula, n represents the number of hours of sunshine, and N represents the number of hours of available sunshine. Based on the sunshine percentage, the weather in this region is divided into several weather types, including: When 1 > S p When the value is ≥0.6, the corresponding weather is sunny, sunny turning cloudy, or cloudy turning sunny, which is classified as weather type 1; When 0.6 > S p When the value is ≥0.1, the corresponding weather is cloudy, overcast turning cloudy, or multi-layered turning overcast, which is classified as weather type 2; When 0.1 > S p When the value is ≥0, the corresponding weather is rain, snow, fog, or haze, and it is classified as weather type 3.

3. The method for predicting total horizontal radiation and diffuse radiation based on weather patterns according to claim 2, characterized in that: in, The formula for calculating the available lighting hours is as follows: In the formula, δ represents the latitude of the region, and δ represents the declination angle of the region.

4. The method for predicting total and scattered radiation on a horizontal surface based on weather patterns according to claim 2, characterized in that: in, In step S3, the expression for the optimal horizontal total radiation calculation model for weather type 1 is: The expression for the optimal horizontal total radiation calculation model for weather type 2 is as follows: The expression for the optimal horizontal total radiation calculation model for weather type 3 is as follows: In the formula, H0 is the solar radiation on the horizontal surface outside the atmosphere, also known as astronomical radiation, V is visibility, C is total cloud cover, API is the air pollution index, ΔT is the diurnal temperature range, and R... h denoted as relative humidity, and a0, b0, c0, and d0 are empirical coefficients.

5. The method for predicting total and scattered radiation on a horizontal surface based on weather patterns according to claim 4, characterized in that: in, In step S4, the localized direct-to-indirect separation model for both weather type 1 and weather type 3 is the modified Jiang model, and its expression is: The localized direct dispersion separation model for weather type 2 is the modified El-Sebaii model, and its expression is: In the formula, H d For scattered radiation, k T The resolution index is represented by a1, b1, c1, d1, e1, f1, and g1, which are empirical coefficients.

6. The method for predicting total and scattered radiation on a horizontal surface based on weather patterns according to claim 5, characterized in that: in, The formula for calculating the sharpness index is as follows: In the formula, H represents the total radiation of the horizontal plane, which is calculated using the optimal total radiation calculation model for the horizontal plane.

7. The method for predicting total and scattered radiation on a horizontal surface based on weather patterns according to claim 5, characterized in that: in, In step S5, For weather type 1, based on the corresponding localized direct dispersion separation model, a GPR model is constructed using the GPR algorithm, and the dimensionality of the GPR model is reduced to 3 factors using principal component analysis to establish a PCA-GPR model. Then, the clarity index, sunshine percentage, visibility, relative humidity, and air pollution index of the region are introduced into the PCA-GPR model for training. The trained model is the optimal direct dispersion separation model under weather type 1. For weather type 2, based on the corresponding localized direct dispersion separation model, the GPR algorithm is used to construct a GPR model, and the clarity index, sunshine percentage and total cloud cover of the region are introduced into the GPR model for training. The trained model is the optimal direct dispersion separation model under weather type 2. For weather type 3, based on the corresponding localized direct dispersion separation model, the GPR algorithm is used to construct a GPR model, and the clarity index and sunshine percentage of the region are introduced into the training of the GPR model. The trained model is the optimal direct dispersion separation model under weather type 3.

Images