A method and system for predicting sunshine duration in the northwest region

Abstract

The application provides a northwest region sunshine duration prediction method and system, and belongs to the technical field of data processing. The method comprises the following steps: obtaining the altitude of a target city and historical data of multiple days to obtain the updated assignment of the relative sunshine duration of each weather type in the first updated seven-point method of the target city and the specific relative sunshine duration of the target city on each day; and obtaining a prediction model of the relative sunshine duration of the target city according to the specific relative sunshine duration of the target city on each day, the actual daily maximum temperature, the actual daily minimum temperature, the actual wind speed, the day serial number, the actual weather type, and the updated assignment of the relative sunshine duration of each weather type. The application aims to solve the problem that the current seven-point method has too strong universality and fails to reflect the unique characteristics of each city and each day, so that the difference between the sunshine duration of the target city predicted by the current seven-point method and the actual sunshine duration is too large.

Description

Technical Field

[0001] This invention belongs to the field of data processing, specifically relating to a method and system for predicting sunshine duration in Northwest China. Background Technology

[0002] In recent years, climate change has attracted widespread attention worldwide, and the resulting water shortage has become increasingly apparent. Developing efficient water-saving irrigation technologies is one of the effective measures to alleviate the water shortage situation in arid northwest regions. Research shows that the accuracy of reference crop evapotranspiration (ETo) prediction is related to the accuracy of crop irrigation forecasts. Currently, there is extensive research on methods for estimating ETO, including mathematical models and empirical formulas. Among them, the PM formula is recommended by the Food and Agriculture Organization of the United Nations as a reliable method for estimating ETO and is recognized worldwide. When using the PM formula for ETO forecasting, future solar radiation values ​​are often required as input. Since it is impossible to accurately measure and obtain future solar radiation values, indirect calculation using available future public weather forecast data is considered. Currently, the main method for obtaining solar radiation based on weather forecast information is by assigning values ​​to relative sunshine duration (n / N) and using sunshine duration as a medium to calculate solar radiation. Currently, crop evapotranspiration is used to determine the effectiveness of crop irrigation at different times and to obtain the optimal irrigation time for crops. Since the daily solar radiation value is the input to the PM formula when calculating daily crop evapotranspiration, this method forecasts the daily solar radiation value for each city in the Northwest region.

[0003] Currently, the relative sunshine duration for each day is obtained based on the weather type predicted by public weather forecasts and the relative sunshine duration assigned to each day type in the current seven-segment method. The actual sunshine duration is then calculated from the relative sunshine duration, and the solar radiation value is calculated using the actual sunshine duration as a medium. However, because the seven-segment method can only assign values ​​to the relative sunshine duration under seven weather types, and because cities in Northwest China differ from Nanjing, the city corresponding to the current seven-segment method, there is a discrepancy between the predicted sunshine duration and the actual sunshine duration for target cities in Northwest China obtained using the current seven-segment method. Therefore, the method for obtaining the predicted sunshine duration for each city in Northwest China needs to be revised. Summary of the Invention

[0004] To address the problem of excessive discrepancies between the predicted and actual sunshine hours of a target city obtained using the current seven-segment method, this invention provides a method for predicting sunshine hours in Northwest China.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Obtain the altitude of the target city, historical forecast data for multiple days, actual relative sunshine hours, the maximum possible sunshine hours for the target city each day, and the altitude of the city corresponding to the current seven-part classification.

[0007] Based on the historical forecast data of the target city over multiple days and the actual relative sunshine hours for each day, the weather types in the first updated seven-class classification of the target city and the assigned values ​​for each weather type are obtained.

[0008] Based on the assigned values ​​of each weather type in the current seven-class classification method and the actual relative sunshine hours and historical forecast data of the target city for each day, normal historical data of the target city are obtained; based on the difference between the altitude of the target city and the altitude of the city corresponding to the current seven-class classification method, the assigned values ​​of each weather type in the first updated seven-class classification method of the target city are adjusted downward multiple times to obtain multiple assigned adjustment values ​​of each weather type in the first updated seven-class classification method of the target city.

[0009] Based on the difference between the actual relative sunshine hours in all normal historical data under the same weather type and the adjustment value of each assigned value in the first update seven-part method of the target city, the updated assignment value of each weather type in the first update seven-part method of the target city is obtained.

[0010] The model is trained by taking the actual relative sunshine hours, historical forecast data, day number, and the updated relative sunshine hours for each weather type in the first update seven-part method as inputs to the target city. The model is then trained to obtain a prediction model for the relative sunshine hours of the target city.

[0011] The forecast data of the target city on the day to be predicted is used as the input to the prediction model of the relative sunshine hours of the target city to obtain the predicted relative sunshine hours of the target city on the day to be predicted; the predicted relative sunshine hours of the target city on the day to be predicted are multiplied by the maximum possible sunshine hours to obtain the predicted actual sunshine hours of the target city on the day to be predicted.

[0012] Furthermore, the specific steps for obtaining the weather types in the first updated seven-part classification of the target city and assigning values ​​to each weather type are as follows:

[0013] According to the target city The forecast weather types for the day are used to obtain the weather type set for the target city; the historical forecast data for the target city for each day before the current number of days includes the forecast weather type, forecast maximum temperature, forecast minimum temperature, and forecast wind speed for each day;

[0014] The weather forecast type for the target city is [number]. The average of the actual relative sunshine hours for all dates under the non-"A to B" weather type is denoted as the first updated seven-part classification of the target city. Assigning values ​​to the relative sunshine duration for weather types other than "A to B";

[0015] The first update of the target city in the seven-part method The specific calculation formula for assigning the relative sunshine duration to each "A to B" weather type is as follows:

[0016]

[0017] In the formula, In the first update of the target city in the seven-part classification, the first... The assignment of relative sunshine hours for weather types other than "A to B" weather. In the first update of the target city in the seven-part classification, the first... The assignment of relative sunshine hours for the first weather type (not "A to B") in the first update seven-classification method of the target city. In the first update of the seven categories of target cities, the first... The assignment of relative sunshine hours for the last weather type (not "A to B") in the first update seven-part classification of the target city. .

[0018] Furthermore, the specific calculation steps for obtaining the normal historical data of the target city are as follows:

[0019] If the target city is in If the absolute value of the difference between the actual relative sunshine duration of a day and the relative sunshine duration assigned to the forecast weather type of that day in the current seven minutes is less than or equal to 0.8, then the target city will be... The forecast weather type and the actual relative sunshine hours for a day are used as two dimensions of normal historical data to obtain a normal historical data.

[0020] If the target city is in The forecast weather type for that day does not exist in the current seven-category weather type classification, and the target city is in The forecast weather type for the target city is based on historical data within the past 14 days. The average of the actual relative sunshine hours for all days of the forecast weather type is denoted as the target city's sunshine hours on the 1st day. Standard values ​​for the forecast weather type for the day;

[0021] If the target city is in The actual relative sunshine hours of the day and the target city on the [day] If the absolute value of the difference between the standard values ​​of the forecast weather type for each day is less than or equal to 0.8, then the target city will be on the [number]th [day]. The forecast weather type and the actual relative sunshine duration obtained within a day are recorded as two dimensions within a normal historical data set.

[0022] Furthermore, the specific calculation steps for the multiple assigned adjustment values ​​for each weather type in the first updated seven-part method for the target city are as follows:

[0023] The formula for calculating the step size for each adjustment of the relative sunshine hours of the target city is as follows:

[0024]

[0025] In the formula, This indicates the step size for each adjustment of the relative sunshine hours for the target city. Indicates the altitude of the target city. This indicates the altitude of Nanjing, the city corresponding to the current seven-part classification. Represents the normalization function. Represents the absolute value function;

[0026] The formula for calculating the maximum boundary for adjusting the relative sunshine hours of the target city upwards or downwards is as follows:

[0027]

[0028] In the formula, This indicates the maximum boundary for adjusting the relative sunshine hours of the target city upwards or downwards. Indicates the altitude of the target city. This indicates the altitude of Nanjing, the city corresponding to the current seven-part classification. Represents the normalization function

[0029] like Then The step size is adjusted upwards for the relative sunshine duration assigned to each weather type in the first updated seven-part classification of the target city, and the adjustment range for the relative sunshine duration assigned to each weather type in the first updated seven-part classification of the target city is less than or equal to... ;

[0030] like <0, then The step size is adjusted downwards for the relative sunshine duration of each weather type in the first update seven-part classification of the target city, and the adjustment range for the relative sunshine duration of each weather type is less than or equal to... ;

[0031] like =0, then First, the relative sunshine duration of each weather type in the first updated seven-part classification of the target city is adjusted upwards, and then adjusted downwards, with the adjustment range of the relative sunshine duration of each weather type in the first updated seven-part classification of the target city being less than or equal to 0.1.

[0032] Obtain multiple assigned adjustment values ​​for each relative sunshine duration for each weather type in the first updated seven-part classification of the target city.

[0033] Furthermore, the specific steps for obtaining multiple assigned adjustment values ​​for each weather type in the first updated seven-part classification of the target city are as follows:

[0034] The first update of the target city in the seven-part method The first weather type The specific formula for calculating the probability of updating the relative sunshine hours assignment adjustment value is as follows:

[0035]

[0036] In the formula, In the first update of the seven categories of target cities, the first... The first weather type The value of the relative sunshine duration is the probability of updating the assigned value. This indicates the first update of the seven-category weather forecast for the target city. The number of normal forecast data for each weather type In the first update of the seven categories of target cities, the first... The first weather type The assigned adjustment value for each relative sunshine duration. This indicates the first update of the seven-category weather forecast for the target city. The first weather type The actual relative sunshine hours of a normal historical data point Represents the absolute value function. It is an exponential function with the natural constant as its base;

[0037] The first seven-part classification of the target city Among the possible values ​​for updating the relative sunshine hours of each weather type, the one with the highest probability corresponds to the first update of the target city in the seven-part update method. The assigned adjustment values ​​for all relative sunshine hours for each weather type are denoted as the first updated seven-part classification value for the target city. The weather type is updated and assigned values.

[0038] Furthermore, the specific steps for obtaining the prediction model for the relative sunshine hours of the target city are as follows:

[0039] With the target city The target value is the actual relative sunshine hours within a day, with the target city in The forecast maximum and minimum temperatures, forecast wind speed, day number, and the updated relative sunshine hours under each day's weather type in the first update seven-part classification of the target city are used as input factors. The CNN-LSTM-Attention model is then used to obtain the prediction model for the relative sunshine hours of the target city.

[0040] Furthermore, the specific steps for obtaining the predicted relative sunshine hours of the target city on the day to be predicted are as follows:

[0041] The target city is in The forecast daily maximum temperature, forecast daily minimum temperature, forecast wind speed, day number, and relative sunshine duration for each weather type are updated and assigned as input data to the relative sunshine duration prediction model for the target city. The model's output is then used as the target city's data for the [number of days missing]. The predicted relative sunshine hours for the day.

[0042] Furthermore, the specific steps for obtaining the predicted actual sunshine hours of the target city on the day to be predicted are as follows:

[0043] Record as the target city .

[0044] Furthermore, the specific steps for obtaining the predicted actual sunshine hours of the target city on the day to be predicted are as follows:

[0045] The data acquisition module obtains the altitude of the target city, historical forecast data for multiple days, actual relative sunshine hours, the maximum possible sunshine hours of the target city for each day, and the altitude of the city corresponding to the current seven-part classification.

[0046] The model building module obtains the weather type and the assigned value for each weather type in the first updated seven-part classification of the target city based on the historical forecast data of the target city over multiple days and the actual relative sunshine hours for each day.

[0047] Based on the assigned values ​​of each weather type in the current seven-class classification method and the actual relative sunshine hours and historical forecast data of the target city for each day, normal historical data of the target city are obtained; based on the difference between the altitude of the target city and the altitude of the city corresponding to the current seven-class classification method, the assigned values ​​of each weather type in the first updated seven-class classification method of the target city are adjusted downward multiple times to obtain multiple assigned adjustment values ​​of each weather type in the first updated seven-class classification method of the target city.

[0048] Based on the difference between the actual relative sunshine hours in all normal historical data under the same weather type and the adjustment value of each assigned value in the first update seven-part method of the target city, the updated assignment value of each weather type in the first update seven-part method of the target city is obtained.

[0049] The model is trained by taking the actual relative sunshine hours, historical forecast data, day number, and the updated relative sunshine hours for each weather type in the first update seven-part method as inputs to the target city. The model is then trained to obtain a prediction model for the relative sunshine hours of the target city.

[0050] The acquisition module takes the forecast data of the target city on the day to be predicted as the input to the prediction model of the relative sunshine hours of the target city, and obtains the predicted relative sunshine hours of the target city on the day to be predicted; multiplying the predicted relative sunshine hours of the target city on the day to be predicted by the maximum possible sunshine hours, we obtain the predicted actual sunshine hours of the target city on the day to be predicted.

[0051] The method for predicting sunshine duration in Northwest China provided by this invention has the following beneficial effects:

[0052] Improvement 1: In this application, factors such as past weather types, historical relative sunshine duration, latitude, and topography are comprehensively considered to improve the value assignment of the seven-part method, making it more specific rather than universal, thus improving the accuracy of the assignment. For the blank part of weather type, a solution of taking the average of historical data is proposed. Therefore, the optimized seven-part method for assigning relative sunshine duration can more accurately and comprehensively consider the spatial and topographical differences in relative sunshine duration.

[0053] Improvement 2: Considering that the original seven-part method for establishing the prediction model was based on the quantitative analysis of relative sunshine duration by weather type, the maximum temperature Tmax and the daily temperature range TR were introduced as correction factors. This application proposes to introduce the minimum temperature Tmin, wind speed (u), and day number (J) to improve the accuracy of relative sunshine duration prediction. On the one hand, it considers the close relationship between sunshine duration and wind speed u. On the other hand, the change of sunshine duration is periodic, and the sunshine duration will change with the change of day number.

[0054] Improvement 3: This application combines CNN-LSTM-Attention to construct a deep learning prediction model. The CNN layer can extract spatial features from the input data. In relative sunshine duration prediction, CNN can capture the spatial relationships between input data. Through convolution operations, CNN extracts local features, and through pooling operations, it reduces the feature dimensionality, thus providing a more compact feature representation for subsequent LSTM layers. LSTM excels at handling temporal dependencies in sequential data. In relative sunshine duration prediction, LSTM can capture the changing patterns of relative sunshine duration over time. Through the control mechanisms of input gates, forget gates, and output gates, LSTM can effectively remember or ignore key information in the sequence, thereby avoiding the gradient vanishing problem. The attention mechanism can automatically identify important time steps in the sequence, thereby improving the model's focus on key information. In relative sunshine duration prediction, the attention mechanism helps the model focus on historical time points that have a significant impact on changes in relative sunshine duration, such as consecutive sunny or cloudy days. By calculating the attention weight for each time step, the model can perform a weighted summation of information from the entire sequence, thus more accurately predicting future relative sunshine duration. Attached Figure Description

[0055] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a flowchart of a method for predicting sunshine duration in Northwest China according to Embodiment 1 of the present invention;

[0057] Figure 2 This is a technical roadmap for an improved method of predicting relative sunshine duration using a seven-part method based on deep learning.

[0058] Figure 3 This is a schematic diagram of the CNN-LSTM-Attention model structure. Detailed Implementation

[0059] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0060] Example 1

[0061] This invention provides a method for predicting sunshine duration in Northwest China, specifically as follows: Figure 1 As shown, it includes the following steps:

[0062] Step S001: Obtain the altitude of the target city, historical forecast data for multiple days, actual relative sunshine hours, maximum possible sunshine hours for each day of the target city, and the altitude of the city corresponding to the current seven-part classification.

[0063] It should be noted that research shows a correlation between weather indicators of different cloud cover and weather types (Table 1). In practical applications, the range cannot directly determine the specific value of relative sunshine duration; instead, a specific value needs to be selected. Currently, the empirical methods for assigning relative sunshine duration are mainly the seven-point method and the five-point method. The five-point method defines weather types in weather forecasts as five basic weather types: sunny, sunny with some clouds, cloudy, overcast, and rainy, assigning n / N values ​​of 0.9, 0.7, 0.5, 0.3, and 0.1 respectively. The seven-point method defines them as seven basic weather types: sunny, sunny turning with some clouds, sunny turning cloudy, cloudy, overcast, haze (fog), and rainy, assigning n / N values ​​of 0.920, 0.775, 0.650, 0.525, 0.375, 0.225, and 0.075 respectively (Table 2). Previous studies applied the seven-category method to predict sunshine duration at Nanjing station (31-32°N, plain area), with good prediction results. Compared with the five-category method, the seven-category method provides a more detailed classification of weather conditions, and the relative sunshine duration assignments for each weather type are derived from the average of interval values, making it more applicable. This application unifies weather types including rain and snow into "rain" and proposes an improved method based on the seven-category interval method (Table 1) and the seven-category method (Table 2).

[0064] It should be further noted that the current seven-segment method is based on Nanjing's weather and has strong universality, but it lacks specific characteristics for a particular city. For example, it cannot assign values ​​to weather types unique to Northwest my country, such as floating dust and blowing sand. Table 1 shows the corresponding table for the current seven-segment interval method, and Table 2 shows the corresponding table for the seven-segment method. Therefore, it is necessary to obtain historical information for the target city over multiple days. The current seven-segment method has universality in assigning relative sunshine hours, but it does not consider the spatiotemporal variation of sunshine hours. Spatially, differences in cloud cover between regions, such as altitude, affect cloud cover distribution and thus the magnitude of relative sunshine hours, resulting in inaccurate assignment. Temporally, it does not consider the issue of time sequence; cloud cover also varies between different seasons, meaning that sunshine hours under the same weather type vary in different regions and at different times. Furthermore, Nanjing, the city corresponding to the current seven-segment method, is located on the third step of my country's terrain, with an average altitude of several hundred meters, while Northwest China is located on the second step of my country's terrain, resulting in a significant difference in altitude between cities in Northwest China and Nanjing. Therefore, the altitude of the target city located in the Northwest region is obtained.

[0065] Therefore, this invention aims to address the limitations and low accuracy of the existing single seven-interval method by proposing a new method for predicting sunshine duration in Northwest China. Referring to the weather type and n / NPAR correspondence in Table 1, and considering factors such as topography and latitude of the study area, the changing trends of interval boundary values ​​for different weather types are determined. Combined with the average relative sunshine duration for this basic weather type, an improved relative sunshine duration assignment table is determined. Then, using the highest temperature, lowest temperature, wind speed, and daily ordinal number forecast data available from public weather forecasts, along with the assigned relative sunshine duration, as model inputs, a deep learning model is constructed that integrates Convolutional Neural Networks (CNN), attention mechanisms, and Long Short-Term Memory (LSTM) networks to predict future relative sunshine duration. 80% of the historical data is selected as the training set, and the remaining 20% ​​as the validation set. The MAE (Magnitude of Effect) is used as the evaluation metric to assess model accuracy, ultimately providing a reliable and more robust model for relative sunshine duration prediction.

[0066] It should be further noted that, since the purpose of this method is to obtain the predicted actual sunshine hours for each city based on the forecast data, it is necessary to obtain multi-day historical forecast data, the actual relative sunshine hours for each day, and the maximum possible sunshine hours for each day for the target cities located in the Northwest region.

[0067] Specifically, the target city's data up to the current number of days is obtained through the China Meteorological Database and the China Weather Network. Historical forecast data for the day, as well as the actual relative sunshine hours and actual sunshine hours for each day. Obtain the maximum possible sunshine hours for the target city each day. The historical forecast data for the target city up to the current day includes the forecast weather type, forecast high temperature, forecast low temperature, and forecast wind speed for each day. This embodiment sets the data to a preset number of historical days. This example will be used for illustration; other values ​​can be set in other implementations. The table corresponding to the current seven-part division method and the seven-interval division method is obtained as follows:

[0068] Table 1 shows the current seven-interval method:

[0069]

[0070] Table 2 shows the current seven-part classification method:

[0071]

[0072] Furthermore, obtain the altitude of the target city and the altitude of Nanjing, the city corresponding to the current seven-part classification.

[0073] Thus, we have obtained the target city's altitude, historical forecast data, the maximum possible sunshine hours and relative sunshine hours for each day, and the altitude of the city corresponding to the current seven-part classification.

[0074] Step S002: Based on the historical forecast data of the target city over multiple days and the actual relative sunshine hours for each day, obtain the weather type in the first updated seven-part classification of the target city and the assigned value for each weather type.

[0075] It should be noted that due to differences in geographical location and surrounding environment between cities, the weather types and corresponding relative sunshine hours in the current seven-category table may not be entirely applicable to the target city. For example, the northwest region experiences weather types such as dust storms, blowing sand, and sandstorms, which are rarely seen in Nanjing. Therefore, a first updated seven-category method was derived based on multi-day historical forecast data for the target city.

[0076] It should be further explained that weather types are divided into two different types: non-"A to B" weather types, such as sunny, cloudy, overcast, rain or snow, and "A to B" weather types, such as sunny to cloudy or sunny to partly cloudy.

[0077] Furthermore, since the "A to B" weather type consists of two weather types, meaning that these two weather types may alternate within a single day, the relative sunshine duration under the "A to B" weather type in the first updated seven-part classification is assigned based on the relative sunshine duration under the two weather types within the "A to B" weather type. Therefore, based on the average of the forecast relative sunshine duration for all days under each non-"A to B" weather type within the historical forecast data of the target city over multiple days, the relative sunshine duration for non-"A to B" weather types in the target city's first updated seven-part classification is obtained.

[0078] It should be further noted that, as current research demonstrates, the relative sunshine duration under the "A to B" weather type is more significantly influenced by the relative sunshine duration of the preceding weather type than by the relative sunshine duration of the following weather type. Therefore, when assigning the relative sunshine duration value for each "A to B" weather type in the first updated seven-part classification of the target city, a larger influence weight is assigned to the relative sunshine duration of the preceding weather type, and a smaller influence weight is assigned to the relative sunshine duration of the following weather type. This results in the assigned relative sunshine duration value for each "A to B" weather type in the first updated seven-part classification of the target city.

[0079] Specifically, depending on the target city The forecast weather types for the day are used to obtain the weather type set for the target city. The weather type set for the target city does not contain duplicate weather types, and the weather type set for the target city only includes weather types that are not "A to B" and "A to B".

[0080] Furthermore, the weather forecast type for the target city will be... The average of the actual relative sunshine hours for all dates under the non-"A to B" weather type is denoted as the first updated seven-part classification of the target city. The relative sunshine duration is assigned to each weather type that is not "A to B".

[0081] Furthermore, in the first update of the target city's seven-part method, the first... The specific calculation formula for assigning the relative sunshine duration to each "A to B" weather type is as follows:

[0082]

[0083] In the formula, In the first update of the seven categories of target cities, the first... The assignment of relative sunshine hours for weather types other than "A to B" weather. In the first update of the seven categories of target cities, the first... The assignment of relative sunshine hours for the first weather type (not "A to B") in the first update seven-classification method of the target city. In the first update of the seven categories of target cities, the first... The assignment of relative sunshine hours for the last weather type (not "A to B") in the first update seven-part classification of the target city. The preset weights, and This embodiment makes This example is used to illustrate the concept; other values ​​can be set in other implementations.

[0084] Furthermore, based on the weather type set of the target city, the relative sunshine hours for each weather type in the first updated seven-part classification of the target city are assigned, thus obtaining the first updated seven-part classification of the target city.

[0085] Specifically, for the arid Northwest region (taking the Hanghou station in Inner Mongolia as an example), a total of 90 weather types were collected from 2011 to 2023. It was determined whether there were any weather types other than sunny, sunny to partly cloudy, sunny to cloudy, cloudy, overcast, haze (fog), and rain (snow) among all non-"A to B" weather types. It was found that the Hanghou station did not have a "sunny to partly cloudy" weather type, but there were three special weather types: "dust, floating sand, and sandstorm". The multi-year average relative sunshine hours under these three weather types were rounded to two decimal places to 0.55 and used as the value for these three weather types. Therefore, the relative sunshine hours assignment table for the Hanghou station is still the "seven-part method".

[0086] Thus, we obtain the first updated seven-part classification of the target city.

[0087] Step S003: Based on the assigned value of each weather type in the current seven-part classification method and the actual relative sunshine hours and historical forecast data of the target city for each day, obtain the normal historical data of the target city; based on the difference between the altitude of the target city and the altitude of the city corresponding to the current seven-part classification method, and the difference between the actual relative sunshine hours in all normal historical data under the same weather type and the assigned adjustment value in the first updated seven-part classification method of the target city, obtain the updated assigned value of each weather type in the first updated seven-part classification method of the target city.

[0088] It should be noted that because most cities in Northwest China are located on the second step of my country's terrain, while Nanjing is located on the third step, there is a significant difference in altitude between cities in Northwest China and Nanjing. Furthermore, the higher the altitude of a region, the less the cloud cover weakens the intensity of sunlight. Therefore, the relative sunshine duration values ​​for each weather type in the first updated seven-part classification table for target cities in Northwest China were adjusted multiple times to obtain the optimal value for each weather type.

[0089] It should be further explained that since the target city's altitude is higher than that of the city corresponding to the current seven-part classification, the cloud cover above the target city reduces the intensity of sunlight less compared to Nanjing. This means a larger upward adjustment is needed to the relative sunshine duration values ​​for each weather type in the first updated seven-part classification for the target city. Therefore, the greater the difference between the target city's altitude and that of Nanjing in the current seven-part classification, the final relative sunshine duration value is obtained by continuously increasing the relative sunshine duration values ​​for each weather type in the first updated seven-part classification table for the target city. Furthermore, the greater the difference between the target city's altitude and that of Nanjing in the current seven-part classification, the larger the adjustment should be for the relative sunshine duration values ​​under each weather type.

[0090] It should be further explained that, since the average relative error index is currently used to reflect the probability that a data point represents all data within a dataset, the average relative error between each data point and the actual relative sunshine duration over multiple days in the target city under a given weather type is calculated to obtain the final relative sunshine duration assignment for each weather type in the first updated seven-part classification of the target city. Therefore, the relative sunshine duration value for each weather type in the first updated seven-part classification of the target city is continuously adjusted to obtain the final relative sunshine duration assignment for each weather type.

[0091] It's important to further clarify that when obtaining forecast data for a target city, the predicted weather type may not reflect the actual weather type for that day. Therefore, relying solely on the predicted weather type and the actual relative sunshine duration for a given day may not accurately reflect the actual relative sunshine duration for that weather type in the target city. For example, if June 19th predicts June 20th to be sunny, but June 20th is not, the forecast data for June 20th might be sunny, but the relative sunshine duration for that weather type would be different from other weather types. In this case, June 20th would not be considered normal historical data. Therefore, normal historical data is obtained by comparing the actual relative sunshine duration for each day in the target city with the predicted weather type from historical forecast data. Furthermore, since the historical data for the target city may contain outliers, if the altitude of the target city is equal to that of Nanjing, the assignments for each weather type in the first updated seven-part classification method for the target city are adjusted multiple times to obtain an optimal assignment for each weather type in the first updated seven-part classification method for the target city.

[0092] It should be further explained that the relative sunshine duration assigned to each weather type in the current seven-category method is highly representative of the relative sunshine duration of other cities under that weather type. In other words, the current seven-category method has strong universality. Therefore, based on the assigned relative sunshine duration for each weather type in the current seven-category method and the actual sunshine duration for each forecast weather type over a multi-day historical period for the target city, outlier historical data is removed. Based on the data after removing outliers, the final relative sunshine duration assigned to each weather type in the first updated seven-category method for the target city is obtained. Since the target city located in the Northwest region may contain weather types not included in the current seven-category method, if a weather type for the target city is not included in the current seven-category method, the average relative sunshine duration over a multi-day historical period for that weather type is used as a standard value to remove outlier data.

[0093] Specifically, if the target city is in the The absolute value of the difference between the actual relative sunshine duration of the day and the relative sunshine duration assigned to the forecast weather type in the current seven minutes is less than or equal to the difference threshold. The target city will be in the 1st The forecast weather type and the actual relative sunshine duration obtained within a day are recorded as two dimensions within a normal historical data set, resulting in a normal historical data set. The preset difference threshold in this embodiment... This example is used to illustrate the concept; other values ​​can be set in other implementations.

[0094] Furthermore, if the target city is in the The forecast weather type for the day does not exist within the current seven-category weather type classification. Therefore, the target city will be... The forecast weather type for the target city is based on historical data within the past 14 days. The average of the actual relative sunshine hours for all days of the forecast weather type is denoted as the target city's sunshine hours on the 1st day. The standard value for the forecast weather type for the day. At this time, if the target city is on the [day number missing]... The actual relative sunshine hours of the day and the target city on the [day] The absolute value of the difference between the standard values ​​of the forecast weather types for each day is less than or equal to the difference threshold. The target city will be in the 1st The forecast weather type and the actual relative sunshine duration obtained within a day are recorded as two dimensions within a normal historical data set, thus obtaining a normal historical data set.

[0095] Furthermore, the relative sunshine duration values ​​for each weather type in the first updated seven-part classification of the target city are adjusted downwards multiple times to obtain the updated sunshine duration for each weather type in the first updated seven-part classification of the target city. In this embodiment, the initial maximum downward adjustment of the relative sunshine duration values ​​is preset to 0.1, with a step size of 0.01; other values ​​can be set in other embodiments.

[0096] The maximum upward or downward adjustment limit for the relative sunshine hours of the target city is:

[0097]

[0098] In the formula, This represents the maximum downward adjustment boundary for assigning relative sunshine hours to the target city. Indicates the altitude of the target city. This indicates the altitude of Nanjing, the city corresponding to the current seven-part classification. This represents the normalization function, which is used for normalization processing in this embodiment.

[0099] The step size for each adjustment of the relative sunshine hours for the target city is as follows:

[0100]

[0101] In the formula, This indicates the step size for each adjustment of the relative sunshine hours for the target city. Indicates the altitude of the target city. This indicates the altitude of Nanjing, the city corresponding to the current seven-part classification. This represents the normalization function, which is used for normalization processing in this embodiment; This represents the absolute value function.

[0102] Furthermore, if Then The step size is adjusted upwards for the relative sunshine duration assigned to each weather type in the first updated seven-part classification of the target city, and the adjustment range for the relative sunshine duration assigned to each weather type in the first updated seven-part classification of the target city is less than or equal to... .

[0103] like <0, then The step size is adjusted downwards for the relative sunshine duration of each weather type in the first update seven-part classification of the target city, and the adjustment range for the relative sunshine duration of each weather type is less than or equal to... .

[0104] like =0, then First, the relative sunshine duration of each weather type in the first updated seven-part classification of the target city is adjusted upwards, and then adjusted downwards. The adjustment range of the relative sunshine duration of each weather type in the first updated seven-part classification of the target city is less than or equal to 0.1.

[0105] Thus, we have obtained multiple assigned adjustment values ​​for each relative sunshine duration for each weather type in the first updated seven-part classification of the target city.

[0106] Furthermore, in the first update of the target city's seven-part classification method, the first... The first weather type The specific formula for calculating the probability of updating the relative sunshine hours assignment adjustment value is as follows:

[0107]

[0108] In the formula, In the first update of the seven categories of target cities, the first... The first weather type The value of the relative sunshine duration is the probability of updating the assigned value. This indicates the first update of the seven-category weather forecast for the target city. The number of normal historical data for each weather type In the first update of the seven categories of target cities, the first... The first weather type The assigned adjustment value for each relative sunshine duration. This indicates the first update of the seven-category weather forecast for the target city. The first weather type The actual relative sunshine hours of a normal historical data point Represents the absolute value function. As an exponential function with the natural constant as its base, this embodiment uses it to represent an inverse proportional relationship.

[0109] It should be noted that, The larger the value, the higher the ranking of the target city in the first update seven-part classification. The first weather type The greater the difference between the assigned adjustment value for relative sunshine hours and the actual relative sunshine hours, the more significant the difference between the first seven-part update of the target city and the second-part update. The first weather type The less accurately the relative sunshine duration is adjusted, the less accurately it reflects the overall situation.

[0110] The table below shows the changes in the average relative error between the assigned and updated values ​​of relative sunshine hours for weather types in the first update seven-part method at Hanghou Station and historical data of the same weather type:

[0111] Table 3 Summary of MAE Changes Before and After Improvement

[0112]

[0113] Furthermore, the first update of the target city in the seven-part classification method... Among the possible values ​​for updating the relative sunshine hours of each weather type, the one with the highest probability corresponds to the first update of the target city in the seven-part update method. The assigned adjustment values ​​for all relative sunshine hours for each weather type are denoted as the first updated seven-part classification value for the target city. The weather type is updated and assigned values.

[0114] The table below shows the first update of the seven-part classification method for the Hangzhou station:

[0115] Table 4. Quantification of Relative Sunshine Hours for the Seven Weather Types After Improvement

[0116]

[0117] Furthermore, this embodiment also provides a method for obtaining the maximum possible sunshine duration and specific relative sunshine duration of a target city each day, the specific steps of which are as follows:

[0118] Specifically, to obtain the first The year's first The formula for calculating the magnetic declination of the sky is as follows:

[0119]

[0120] In the formula, Indicates the first The year's first Magnetic declination of the sky, This represents the sine function in trigonometric functions. For the first The number of days contained in a year. It is 180 degrees. For the first Heaven is in The year's ordinal value.

[0121] Furthermore, obtaining the target city in the first The year's first The specific formula for calculating the sunset angle is as follows:

[0122]

[0123] In the formula, Indicates the target city is in the 1st century. The year's first Sunset angle of the day, Indicates the first The year's first Magnetic declination of the sky, Indicates the latitude of the target city. Represents the tangent function. This represents the inverse cosine function.

[0124] Furthermore, obtaining the target city in the first The year's first The specific formula for calculating the maximum possible sunshine hours per day is as follows:

[0125]

[0126] In the formula, Indicates the target city is in the 1st century. The year's first The maximum possible sunshine hours per day It is 180 degrees. Indicates the target city is in the 1st century. The year's first The angle of sunset.

[0127] Furthermore, obtaining the target city in the first The year's first The formula for calculating the specific relative real number of solar indices per day is as follows:

[0128]

[0129] In the formula, Indicates the target city is in the 1st century. The year's first The specific relative real number of solar radiation for a day, Indicates the target city is in the 1st century. The year's first The maximum possible sunshine hours per day Indicates the target city is in the 1st century. The year's first The actual number of hours of sunshine per day.

[0130] This method also provides a way to calculate wind speed, as follows:

[0131]

[0132] In the formula, It is based on the wind speed value at a height of 2m above the ground. It is based on the wind speed value at a location zm above the ground. The wind speed and altitude were measured at the station.

[0133] Table 5 Wind force level and corresponding wind speed at 10 meters above the ground (m / s)

[0134]

[0135] At this point, the updated values ​​for each weather type in the first seven-part update method for the target city are obtained.

[0136] Step S004: Based on the actual relative sunshine hours, forecast daily maximum temperature, forecast daily minimum temperature, forecast wind speed, day number, forecast weather type, and the updated values ​​of the relative sunshine hours under each weather type in the first update seven-part method for each day, obtain the prediction model of the relative sunshine hours of the target city.

[0137] It should be noted that current forecasts of relative sunshine hours mostly use single regression models. Since the input data do not exhibit a strong correlation, only showing a consistent trend, a single model struggles to capture their underlying features. This invention uses a CNN-LSTM-Attention hybrid model, introducing a CNN convolutional neural network into a single LSTM prediction model to better capture the deep connections between the input data and the forecast target. Simultaneously, an attention mechanism focuses on historical time points that significantly influence changes in relative sunshine hours, such as consecutive sunny or cloudy days. By calculating the attention weight at each time step, the model can perform a weighted summation of information from the entire sequence, thereby more accurately forecasting future relative sunshine hours. Figure 2 This is a technical roadmap for an improved seven-segment method for predicting relative sunshine duration based on deep learning. Among them, Figure 3 This is a schematic diagram of the CNN-LSTM-Attention model structure.

[0138] Specifically, based on the target city The target value is the actual relative sunshine hours within a day, with the target city in Forecast daily high and low temperatures, wind speed, day number, and location within the target city for each day of the day. The forecast weather type for each day of the day is assigned an updated value of the relative sunshine duration under each day's weather type in the first update seven-classification method for the target city as an input factor. A CNN-LSTM-Attention model is then used, selecting historical data. % as the training set, the remaining (1- Using % as the validation set, and NSE, RMSE, R2, and MAE as evaluation indicators to judge the model accuracy, a data expression n / N = f(Tmax, Tmin, J, u, n / NPAR) is established between n / NOBS and five independent variables: n / NPAR, Tmax, Tmin, u, and J. This completes the model building process and yields a prediction model for the relative sunshine hours of the target city. In this embodiment, the preset historical number of days is used as the basis for the prediction. The pre-defined proportion of the training set This will be described using this as an example; other values ​​can be used in other embodiments. The improved technical route of this invention is as follows: Figure 2 As shown, Figure 3 This is the structure of the CNN-LSTM-Attention model.

[0139] Thus, a model for obtaining the predicted relative sunshine hours for the target city on each day is obtained.

[0140] Step S005: Use the forecast data of the target city on the day to be predicted as the input of the prediction model of the relative sunshine hours of the target city to obtain the predicted relative sunshine hours of the target city on the day to be predicted, and then obtain the predicted actual sunshine hours of the target city on the day to be predicted.

[0141] Specifically, the target city will be in the first The forecast daily high temperature, forecast daily low temperature, forecast wind speed, day number, and the first updated seven-part classification of the target city. The relative sunshine duration of the weather forecast is updated and assigned as the input to the prediction model of the relative sunshine duration of the target city; The output is the target city in the 1st month. The sky .

[0142] Furthermore, the target city will be in the first The sky and The product of the maximum possible sunshine hours for the target city on the 1st day is denoted as the number of days the target city receives. The actual number of hours of sunshine predicted for the day.

[0143] This concludes the embodiment.

[0144] Another embodiment of the present invention provides a sunshine duration forecasting system based on weather forecasting, comprising:

[0145] The data acquisition module obtains the altitude of the target city, historical forecast data for multiple days, actual relative sunshine hours, the maximum possible sunshine hours of the target city for each day, and the altitude of the city corresponding to the current seven-part classification.

[0146] The model building module obtains the weather type and the assigned value for each weather type in the first updated seven-part classification of the target city based on the historical forecast data of the target city over multiple days and the actual relative sunshine hours for each day.

[0147] Based on the assigned values ​​for each weather type in the current seven-class classification and the actual relative sunshine hours and historical forecast data for each day in the target city, normal historical data for the target city are obtained. Based on the difference between the altitude of the target city and the altitude of the city corresponding to the current seven-class classification, the assigned values ​​for each weather type in the first updated seven-class classification of the target city are adjusted downwards multiple times, resulting in multiple adjusted values ​​for each weather type in the first updated seven-class classification of the target city.

[0148] Based on the difference between the actual relative sunshine hours in all normal historical data under the same weather type and the adjustment value of each assigned value in the first update seven-part method for the target city, the updated assignment value of each weather type in the first update seven-part method for the target city is obtained.

[0149] The model is trained using the actual relative sunshine hours, historical forecast data, day number, and updated relative sunshine hours for each weather type in the first seven-part update method for the target city as input. A deep learning model based on CNN-LSTM-Attention is then trained to obtain a prediction model for the relative sunshine hours of the target city.

[0150] The acquisition module takes the forecast data of the target city on the day to be predicted as the input to the prediction model of the relative sunshine hours of the target city, and obtains the predicted relative sunshine hours of the target city on the day to be predicted; multiplying the predicted relative sunshine hours of the target city on the day to be predicted by the maximum possible sunshine hours, we obtain the predicted actual sunshine hours of the target city on the day to be predicted.

[0151] It should be noted that the specific embodiments described above enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. Therefore, although the present invention has been described in detail in this specification and embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention; and all technical solutions and improvements that do not depart from the spirit and scope of the present invention are covered within the protection scope of the patent of the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for predicting sunshine duration in Northwest China, characterized in that, include: Obtain the altitude of the target city, historical forecast data for multiple days, actual relative sunshine hours, the maximum possible sunshine hours for the target city each day, and the altitude of the city corresponding to the current seven-part classification. Based on the historical forecast data of the target city over multiple days and the actual relative sunshine hours for each day, the weather types in the first updated seven-class classification of the target city and the assigned values ​​for each weather type are obtained. Based on the assigned values ​​of each weather type in the current seven-class classification method and the actual relative sunshine hours and historical forecast data of the target city for each day, normal historical data of the target city are obtained; based on the difference between the altitude of the target city and the altitude of the city corresponding to the current seven-class classification method, the assigned values ​​of each weather type in the first updated seven-class classification method of the target city are adjusted downward multiple times to obtain multiple assigned adjustment values ​​of each weather type in the first updated seven-class classification method of the target city. Based on the difference between the actual relative sunshine hours in all normal historical data under the same weather type and the adjustment value of each assigned value in the first update seven-part method of the target city, the updated assignment value of each weather type in the first update seven-part method of the target city is obtained. The model is trained by taking the actual relative sunshine hours, historical forecast data, day number, and the updated relative sunshine hours for each weather type in the first update seven-part method as inputs to the target city. The model is then trained to obtain a prediction model for the relative sunshine hours of the target city. The forecast data of the target city on the day to be predicted is used as the input to the prediction model of the relative sunshine hours of the target city to obtain the predicted relative sunshine hours of the target city on the day to be predicted; the predicted relative sunshine hours of the target city on the day to be predicted are multiplied by the maximum possible sunshine hours to obtain the predicted actual sunshine hours of the target city on the day to be predicted. The specific steps for obtaining the weather types in the first updated seven-part classification of the target city and assigning values ​​to each weather type are as follows: According to the target city The forecast weather types for the day are used to obtain the weather type set for the target city; the historical forecast data for the target city for each day before the current number of days includes the forecast weather type, forecast maximum temperature, forecast minimum temperature, and forecast wind speed for each day; The weather forecast type for the target city is [number]. The average of the actual relative sunshine hours for all dates under the non-"A to B" weather type is denoted as the first updated seven-part classification of the target city. Assigning values ​​to the relative sunshine duration for weather types other than "A to B"; The first update of the target city in the seven-part method The specific calculation formula for assigning the relative sunshine duration to each "A to B" weather type is as follows: ； In the formula, In the first update of the seven categories of target cities, the first... The assignment of relative sunshine hours for weather types other than "A to B" weather. In the first update of the seven categories of target cities, the first... The assignment of relative sunshine hours for the first weather type (excluding "A to B") in the first update seven-classification method of the target city. In the first update of the seven categories of target cities, the first... The assignment of relative sunshine hours for the last weather type (not "A to B") in the first update seven-part classification of the target city. ; The specific calculation steps for the multiple assigned adjustment values ​​for each weather type in the first update seven-point method for the target city are as follows: The formula for calculating the step size for each adjustment of the relative sunshine hours of the target city is as follows: ； In the formula, This indicates the step size for each adjustment of the relative sunshine hours for the target city. Indicates the altitude of the target city. This indicates the altitude of Nanjing, the city corresponding to the current seven-part classification. Represents the normalization function. Represents the absolute value function; The formula for calculating the maximum boundary for adjusting the relative sunshine hours of the target city upwards or downwards is as follows: ； In the formula, This indicates the maximum boundary for adjusting the relative sunshine hours of the target city upwards or downwards. Indicates the altitude of the target city. This indicates the altitude of Nanjing, the city corresponding to the current seven-part classification. Represents the normalization function like Then The step size is adjusted upwards for the relative sunshine duration assigned to each weather type in the first updated seven-part classification of the target city, and the adjustment range for the relative sunshine duration assigned to each weather type in the first updated seven-part classification of the target city is less than or equal to... ; like <0, then The step size is adjusted downwards for the relative sunshine duration of each weather type in the first update seven-part classification of the target city, and the adjustment range for the relative sunshine duration of each weather type is less than or equal to... ; like =0, then First, the relative sunshine duration of each weather type in the first updated seven-part classification of the target city is adjusted upwards, and then adjusted downwards, with the adjustment range of the relative sunshine duration of each weather type in the first updated seven-part classification of the target city being less than or equal to 0.

1. Obtain multiple assigned adjustment values ​​for each relative sunshine duration for each weather type in the first updated seven-part classification of the target city.

2. The method for predicting sunshine duration in Northwest China according to claim 1, characterized in that, The specific calculation steps for obtaining the normal historical data of the target city are as follows: If the target city is in If the absolute value of the difference between the actual relative sunshine duration of a day and the relative sunshine duration assigned to the forecast weather type of that day in the current seven minutes is less than or equal to 0.8, then the target city will be... The forecast weather type and the actual relative sunshine hours for a day are used as two dimensions of normal historical data to obtain a normal historical data. If the target city is in The forecast weather type for that day does not exist in the current seven-category weather type classification, and the target city is in The forecast weather type for the target city is based on historical data within the past 14 days. The average of the actual relative sunshine hours for all days of the forecast weather type is denoted as the target city's sunshine hours on the 1st day. Standard values ​​for the forecast weather type for the day; If the target city is in The actual relative sunshine hours of the day and the target city on the [day] If the absolute value of the difference between the standard values ​​of the forecast weather type for each day is less than or equal to 0.8, then the target city will be on the [number]th [day]. The forecast weather type and the actual relative sunshine duration obtained within a day are recorded as two dimensions within a normal historical data set.

3. The method for predicting sunshine duration in Northwest China according to claim 1, characterized in that, The specific steps for obtaining multiple assigned adjustment values ​​for each weather type in the first updated seven-part method for the target city are as follows: The first update of the target city in the seven-part method The first weather type The specific formula for calculating the probability of updating the relative sunshine hours assignment adjustment value is as follows: ； In the formula, In the first update of the seven categories of target cities, the first... The first weather type The value of the relative sunshine duration is the probability of updating the assigned value. This indicates the first update of the seven-category weather forecast for the target city. The number of normal forecast data for each weather type In the first update of the seven categories of target cities, the first... The first weather type The assigned adjustment value for each relative sunshine duration. This indicates the first update of the seven-category weather forecast for the target city. The first weather type The actual relative sunshine hours of a normal historical data point Represents the absolute value function. It is an exponential function with the natural constant as its base; The first seven-part classification of the target city Among the possible values ​​for updating the relative sunshine hours of each weather type, the one with the highest probability corresponds to the first update of the target city in the seven-part update method. The assigned adjustment values ​​for all relative sunshine hours for each weather type are denoted as the first updated seven-part classification value for the target city. The weather type is updated and assigned values.

4. The method for predicting sunshine duration in Northwest China according to claim 1, characterized in that, The specific steps for obtaining the prediction model for the relative sunshine hours of the target city are as follows: With the target city The target value is the actual relative sunshine hours within a day, with the target city in The forecast maximum and minimum temperatures, forecast wind speed, day number, and the updated relative sunshine hours under each day's weather type in the first update seven-part classification of the target city are used as input factors. The CNN-LSTM-Attention model is then used to obtain the prediction model for the relative sunshine hours of the target city.

5. The method for predicting sunshine duration in Northwest China according to claim 1, characterized in that, The specific steps for obtaining the predicted relative sunshine hours of the target city on the day to be predicted are as follows: The target city is in The forecast daily maximum temperature, forecast daily minimum temperature, forecast wind speed, day number, and relative sunshine duration for each weather type are updated and assigned as input data to the relative sunshine duration prediction model for the target city. The model's output is then used as the target city's data for the [number of days missing]. The predicted relative sunshine hours for the day.

6. The method for predicting sunshine duration in Northwest China according to claim 1, characterized in that, The specific steps for obtaining the predicted actual sunshine hours of the target city on the day to be predicted are as follows: The product of the predicted relative sunshine hours and the maximum possible sunshine hours for the target city on the day to be predicted is denoted as the predicted actual sunshine hours for the target city on the day to be predicted.

7. The system for predicting sunshine duration in Northwest China as described in any one of claims 1-6, characterized in that, include: The data acquisition module obtains the altitude of the target city, historical forecast data for multiple days, actual relative sunshine hours, the maximum possible sunshine hours of the target city for each day, and the altitude of the city corresponding to the current seven-part classification. The model building module obtains the weather type and the assigned value for each weather type in the first updated seven-part classification of the target city based on the historical forecast data of the target city over multiple days and the actual relative sunshine hours for each day. Based on the assigned values ​​of each weather type in the current seven-class classification method and the actual relative sunshine hours and historical forecast data of the target city for each day, normal historical data of the target city are obtained; based on the difference between the altitude of the target city and the altitude of the city corresponding to the current seven-class classification method, the assigned values ​​of each weather type in the first updated seven-class classification method of the target city are adjusted downward multiple times to obtain multiple assigned adjustment values ​​of each weather type in the first updated seven-class classification method of the target city. Based on the difference between the actual relative sunshine hours in all normal historical data under the same weather type and the adjustment value of each assigned value in the first update seven-part method of the target city, the updated assignment value of each weather type in the first update seven-part method of the target city is obtained. The model is trained by taking the actual relative sunshine hours, historical forecast data, day number, and the updated relative sunshine hours for each weather type in the first update seven-part method as inputs to the target city. The model is then trained to obtain a prediction model for the relative sunshine hours of the target city. The prediction value acquisition module takes the forecast data of the target city on the day to be predicted as the input of the prediction model of the relative sunshine hours of the target city, and obtains the predicted relative sunshine hours of the target city on the day to be predicted; multiplying the predicted relative sunshine hours of the target city on the day to be predicted by the maximum possible sunshine hours, we obtain the predicted actual sunshine hours of the target city on the day to be predicted.

Images