Solar Irradiance Prediction Method and Storage Medium

Abstract

The present invention discloses a method for predicting daily radiation exposure and a storage medium, comprising: first, collecting daily historical radiation exposure data and latitude and longitude height data of multiple meteorological observation points in a region, sorting the collected historical radiation exposure data in chronological order, and generating a solar exposure time series for each observation point; then, constructing a spatiotemporal coupling model based on the distribution law of the solar exposure time series and the spatial correlation between the spatial positions of the observation points; thereafter, constructing a training set based on the collected solar exposure time series for each observation point, and training the constructed spatiotemporal coupling model through the training set. Finally, the spatial position parameters and time parameters of the region to be predicted are input into the trained spatiotemporal coupling model to predict the daily radiation exposure for the region to be predicted. This method fully considers the temporal distribution law and spatial correlation of the solar radiation exposure, and can realize the radiation exposure prediction at any spatial position and any time in the region, with high prediction accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing methods specifically suitable for administrative, commercial, financial, management, supervisory or forecasting purposes, and in particular to a daily radiation amount forecasting method and storage medium. Background Art

[0002] Materials like plastics and rubber can age and degrade under intense solar radiation, affecting their performance and service life. This degradation is primarily due to the high energy of ultraviolet light (UV) in sunlight (300–500 kJ / mol), which is absorbed by the dissociation energy of polymer bonds. Therefore, sunlight is sufficient to break down chemical bonds such as Cl-N, C-O, C=C, C-Cl, C-H, and C-N in most polymer materials, disrupting their molecular structure. These broken molecular chains undergo further oxidative degradation under the influence of oxygen and water, leading to the photo-oxidative aging failure process. To guide the environmentally compatible design of products and equipment, numerous environmental data manuals and standards (such as DEF STAN 00-35 and GJB1172) have been published domestically and internationally to understand the distribution and variation of solar radiation extremes across regions.

[0003] The above-mentioned products are derived from statistical analysis of observational data from existing stations. However, my country has a vast territory, and currently only a small number of stations capable of real-time monitoring of solar radiation are available. Therefore, the extreme values of these numerical statistical products are inevitably limited by the number of observation stations and deviate from actual conditions. Furthermore, when mapping the spatial distribution of solar radiation, due to the large distances between stations, large areas between observation points are estimated solely through spatial interpolation, a method that also struggles to reflect the true distribution of solar radiation. Therefore, in order to more accurately map solar radiation extremes and real-time distribution, obtain more precise basic solar radiation data products, meet solar radiation requirements during material testing and evaluation, and effectively support the environmental adaptability design of products and equipment, it is essential to construct an accurate spatiotemporal coupling model to predict solar radiation and its upper and lower limits at any spatial location and time within a region. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the present invention proposes a daily radiation prediction method and storage medium, which can predict the daily radiation at any spatial location and any time in a region.

[0005] A daily radiation prediction method includes the following key aspects:

[0006] Collect daily historical radiation data of multiple observation points in the area, and generate daily radiation time series of each observation point based on the corresponding historical radiation data;

[0007] According to the spatial correlation between the distribution law of the daily radiation time series and the spatial location of the observation points, a spatiotemporal coupling model of daily radiation is constructed.

[0008] The spatiotemporal coupling model is trained based on the time series of daily radiation at each observation point;

[0009] Based on the trained spatiotemporal coupling model, the solar radiation of the area to be predicted is predicted according to the corresponding spatial position parameters and time parameters.

[0010] Furthermore, constructing the space-time coupling model includes:

[0011] The envelope model and relative deviation model of the daily radiation exposure time series are constructed based on the daily temporal variation of the daily radiation exposure data.

[0012] Modeling unknown parameters of the envelope model and the relative deviation model according to the spatial characteristics of the daily radiation, and determining parameter models of the envelope model and the relative deviation model;

[0013] The space-time coupling model is constructed based on the parameter model, the envelope model and the relative dispersion model.

[0014] Furthermore, the envelope model is constructed by using an M-spline function.

[0015] Furthermore, the envelope model is:

[0016]

[0017] Among them, β0 is the intercept of the envelope signal, t is the time, is the largest integer not exceeding t. K T is the coefficient of the kth basis function of the slowly varying annual mean value of the envelope signal; K D is the coefficient of the kth basis function of the annual amplitude of the envelope signal, K M is the coefficient of the kth basis function of the annual periodic fluctuation term, is the kth basis function in the s-order M-spline family with K basis functions, is an unknown parameter, K T K is the number of basis functions of the spline family in the function that describes the slow change of the mean value of the daily radiation envelope signal. D K is the number of basis functions of the spline family in the annual periodic function of the daily radiation envelope signal. Mis the number of basis functions of the spline family in the vibration amplitude function of the annual periodic signal.

[0018] Furthermore, the relative deviation model is fitted by a Gaussian process.

[0019] Furthermore, fitting the relative deviation model through a Gaussian process includes:

[0020] Analyze the relative deviation sequence between the actual daily radiation time series and the envelope series;

[0021] Calculate the mean of a relative deviation series;

[0022] The generalized extreme value distribution function is used to characterize the mean distribution law of relative deviation series.

[0023] Furthermore, modeling the unknown parameters of the envelope model and the relative deviation model includes:

[0024] The Gaussian process model is used to model the unknown parameters of the envelope model and the relative deviation model. The parameter model is:

[0025]

[0026] Among them, θ i，l is the lth parameter of the exposure time series envelope model at spatial position i, l = 1, ..., L, μ l is the mean vector, σ l Gaussian process random error, Ι is the n×n order unit matrix with 1 on the diagonal and 0 on the other elements, and n is the number of sites;

[0027] K l (X, X) is the kernel function matrix, and its (i1, i2)th element is:

[0028]

[0029] Among them, K l is the parameter that describes the uncertainty of the Gaussian process, γ l is the scale parameter;

[0030] Based on the Gaussian process model, the unknown parameters of the envelope model and relative deviation model at any spatial position are determined as follows:

[0031]

[0032]

[0033] Furthermore, the space-time coupling model is:

[0034]

[0035]

[0036]

[0037] A storage medium stores a computer program, the key point of which is that when the computer program is run, the above-mentioned daily radiation amount prediction method is executed.

[0038] Beneficial Effects: The present invention's daily radiation prediction method and storage medium fully consider the temporal distribution and spatial correlation of daily radiation, enabling high-precision prediction of radiation at any spatial location and time within a region. The predicted daily radiation data can provide additional foundational daily radiation data for data products such as daily radiation extreme values and distribution patterns, compensating, to a certain extent, for the limited precision and accuracy associated with using a single point-based representation of solar radiation distribution patterns. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the specific embodiments of the present invention, the following briefly introduces the drawings required for use in the specific embodiments. In all the drawings, each element or part is not necessarily drawn according to the actual scale.

[0040] Figure 1 A flowchart of a daily radiation prediction method provided by an embodiment of the present invention;

[0041] Figure 2 A flowchart of constructing a spatiotemporal coupling model provided by one embodiment of the present invention;

[0042] Figure 3 A flowchart of constructing a relative deviation model provided by one embodiment of the present invention;

[0043] Figure 4 The figure shows the time series of daily radiation at six observation points in Xinjiang and the upper envelope based on the M-spline function in the embodiment of the present invention;

[0044] Figure 5 The distribution of relative deviations after transformation in six places in Xinjiang and the fitting results using generalized extreme value distribution in the embodiment of the present invention are shown;

[0045] Figure 6 The figure shows the comparison between the fitted value and the actual value of the radiation exposure at the Aksu Observatory from January 1, 2000 to February 28, 2015 in the embodiment of the present invention.

[0046] Figure 7 The diagram shows the distribution cloud map of the radiation exposure prediction envelope in Xinjiang on July 1, 2005, the 90% prediction interval cloud map, and the spatial interpolation cloud map based on actual data in the embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following embodiments of the technical solution of the present invention will be described in detail with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention and are therefore only examples and are not intended to limit the scope of protection of the present invention.

[0048] like Figure 1 As shown, a daily radiation prediction method includes:

[0049] Step 1: Collect daily historical radiation data of multiple observation points, and generate a daily radiation time series for each observation point based on the corresponding historical radiation data;

[0050] Step 2: Based on the spatial correlation between the distribution pattern of the daily radiation time series and the spatial location of the observation points, a spatiotemporal coupling model of daily radiation is constructed;

[0051] Step 3: Train the spatiotemporal coupling model based on the time series of daily radiation at each observation point;

[0052] Step 4: Based on the trained spatiotemporal coupling model, the daily radiation of the area to be predicted is predicted according to the corresponding spatial position parameters and time parameters.

[0053] Specifically, first, the daily historical data of radiation and longitude and latitude height data of multiple meteorological observation points in a certain area over the years can be collected, and the collected historical data of radiation can be sorted in chronological order to generate a time series of sunshine at each observation point.

[0054] Then, a spatiotemporal coupling model of daily radiation can be constructed based on the spatial correlation between the distribution law of the daily radiation time series and the spatial position of the observation point, and a training set can be constructed based on the collected solar radiation time series of each observation point. The constructed spatiotemporal coupling model is trained with the training set. Finally, the spatial position parameters and time parameters of the area to be predicted are input into the trained spatiotemporal coupling model to obtain the daily radiation of the area to be predicted.

[0055] To validate the prediction method used in this embodiment, the Xinjiang Uyghur Autonomous Region was used as an example. Daily radiation data and latitude, longitude, and altitude information for 11 national meteorological observation stations in the Xinjiang Uyghur Autonomous Region from 2000 to 2015 were obtained from the National Meteorological Information Center. Data from one observation station was selected as test data, and data from the other 10 stations was used as training data to train the spatiotemporal coupling model. After training, the latitude, longitude, and altitude information of the observation stations in the test data were input into the trained spatiotemporal coupling model to obtain predicted daily radiation data for the test stations. The predicted daily radiation data for the test stations were then compared with the actual daily radiation to verify the effectiveness of the model.

[0056] The prediction of solar radiation at each site based on the leave-one-out cross-validation is shown in Table 1 below.

[0057] Table 1 Prediction of solar radiation at each site based on leave-one-out cross-validation

[0058] Serial number Station Name 90% prediction interval coverage Serial number Station Name 90% prediction interval coverage 1 Aksu 0.922 7 Tacheng 0.814 2 Altay 0.847 8 Turpan 0.946 3 Hami 0.894 9 Urumqi 0.833 4 Hotan 0.667 10 Yanqi 0.889 5 Kashgar 0.849 11 Yining 0.849 6 Ruoqiang 0.712

[0059] As can be clearly seen in the table above, the proportion of observed values covered within the 90% prediction interval is close to 90%. This shows that the predicted values are highly consistent with the actual recorded data, not only matching the actual trend of daily radiation exposure, but also with minimal deviation from the actual values. Therefore, the proposed method for predicting the spatiotemporal distribution of daily radiation exposure, based on the daily radiation exposure data input from the observation stations, can predict the daily radiation exposure at any spatial location and time within a region. This can improve the computational sophistication of basic products such as the distribution patterns and extreme values of daily radiation exposure within a region.

[0060] In this embodiment, optionally, constructing the spatiotemporal coupling model includes:

[0061] Step 2-1: Construct an envelope model and a relative deviation model for the daily radiation exposure time series based on the daily temporal variation pattern of the daily radiation exposure data;

[0062] Step 2-2, modeling the unknown parameters of the envelope model and the relative deviation model according to the spatial characteristics of the daily radiation, and determining the parameter models of the envelope model and the relative deviation model;

[0063] Step 2-3: construct the spatiotemporal coupling model based on the parameter model, the envelope model and the relative dispersion model.

[0064] Specifically, first, based on the daily temporal variation of daily radiation exposure data, an envelope model and a relative deviation model can be constructed, including the daily radiation exposure time series. The envelope model can be fitted by an M-spline function, specifically including:

[0065] The daily irradiation at the observation point within a certain period of time is used as input to analyze its changing pattern in time series. In the embodiment of the present invention, taking Xinjiang as an example, Figure 4 As shown in the figure, the upper envelope of daily radiation in Xinjiang gradually increases from the beginning of the year, reaches a peak in the middle of the year, and then gradually decreases, showing an approximately sinusoidal change pattern throughout the year.

[0066] In addition, due to the influence of weather, the actual daily exposure value fluctuates greatly. Therefore, we can first obtain the envelope of the exposure sequence and analyze the envelope to grasp the overall law of exposure changes. Figure 4 It shows that the upper envelope has good regularity, while the lower envelope is not very meaningful. Therefore, in the embodiment of the present invention, the upper envelope is selected for fitting.

[0067] The upper envelope signal of the above daily radiation time series can be fitted by the M-spline function, specifically:

[0068]

[0069] Among them, β0 is the intercept of the envelope signal, t is the time, is the largest integer not exceeding t; K T is the coefficient of the kth basis function of the slowly varying annual mean value of the envelope signal; K D is the coefficient of the kth basis function of the annual amplitude of the envelope signal, K M is the coefficient of the kth basis function of the annual periodic fluctuation term, is the kth basis function in the s-order M-spline family with K basis functions, is an unknown parameter, K T K is the number of basis functions of the spline family in the function that describes the slow change of the mean value of the daily radiation envelope signal. D K is the number of basis functions of the spline family in the annual periodic function of the daily radiation envelope signal. M The number of basis functions of the spline family in the amplitude function of the annual periodic signal vibration;

[0070] Used to characterize the slowly changing mean value of the envelope signal; It is used to describe the component of annual periodic changes in the envelope signal; It indicates the slow change of the vibration amplitude of the annual cycle signal;

[0071] Considering the smoothness requirement of the envelope signal and the constraint of model complexity, the third-order M-spline function can be selected to characterize the envelope signal to meet the basic requirements of fitting. In addition, by comparing the fitting effects of different models on existing data, K T =K M =14,K D =8.

[0072] The relative deviation model can be obtained by Gaussian process fitting, including:

[0073] Step 2-1-1: Analyze the relative deviation sequence between the actual daily radiation time series y(t) and the envelope series y~(t) at each observation point. The calculation formula is as follows:

[0074]

[0075] Step 2-1-2: Calculate the mean of the relative deviation sequence. The calculation formula is as follows:

[0076]

[0077] Where m is the amount of actual daily radiation time series data.

[0078] Step 2-1-3: Use the generalized extreme value distribution function to determine the mean distribution law of the relative deviation sequence.

[0079]

[0080] Among them, η is the location parameter, ω is the scale parameter, κ is the shape parameter, and α = {η, ω, κ} is the unknown parameter.

[0081] Then, according to the spatial characteristics of daily radiation, the unknown parameters in the envelope model and relative deviation model are modeled using the Gaussian process model on the two-dimensional coordinate plane of latitude and longitude. The specific steps are as follows:

[0082]

[0083] Among them, θ i，l is the lth parameter of the exposure time series envelope model at spatial position i, l = 0, 1, ..., L, μ l is the mean vector, σ l Gaussian process random error, Ι is the n×n order unit matrix with 1 on the diagonal and 0 on the other elements, and n is the number of sites;

[0084] K l (X, X) is the kernel function matrix, and its (i1, i2)th element is:

[0085]

[0086] Among them, K l is the parameter that describes the uncertainty of the Gaussian process, γ l is a scale parameter used to characterize the effect of the distance between two points in space on the correlation between two test sites.

[0087] Any spatial position The model parameters of the daily radiation time series at can be obtained by interpolation based on the Gaussian process model:

[0088]

[0089]

[0090] Finally, the above parameter model is substituted into the envelope model and the relative deviation model to construct the spatiotemporal coupling model:

[0091]

[0092]

[0093]

[0094] The above-mentioned spatiotemporal coupling model is trained using the collected sunshine time series of each station to determine the model parameters of the spatiotemporal coupling model, and the trained spatiotemporal coupling model is constructed through the spatiotemporal coupling model parameters.

[0095] Specifically, the daily radiation time series obtained from 10 national-level gas phase observation stations in Xinjiang were first substituted into the constructed spatiotemporal coupling model. The time series model parameters for each station were obtained based on the daily radiation time series, as shown in Table 2. Based on the latitude and longitude coordinates of the observation stations, the model parameters of the time series were regressed using a Gaussian process to obtain the spatiotemporal coupling model parameters with spatial correlation. The obtained spatiotemporal coupling model parameters are shown in Table 3.

[0096] Table 2 Time series model parameters

[0097]

[0098] Table 3 Parameters of the spatiotemporal coupling model

[0099]

[0100] Substituting the daily radiation data collected from various national gas phase observation stations into the spatiotemporal coupling model of the component, the envelope of the daily radiation series of six places in Xinjiang based on the M-spline function is obtained. The specific situation is as follows Figure 4 As shown in Figure 2, the envelope function based on M-spline can well describe the changing trend of the solar radiation envelope. Figure 5 , it can be seen that the generalized extreme value distribution fits the distribution of relative deviation after transformation very well. The constructed spatiotemporal coupling model is used to predict the daily radiation at Aksu Observatory from January 1, 2000 to February 28, 2015. The comparison between the predicted value and the actual value is shown in Figure 6 ,It can be seen that the model constructed by the present invention well describes the temporal variation of daily radiation, and the 90% prediction interval covers the actual exposure level well.

[0101] Based on the above time series results and the constructed spatiotemporal coupling model, the envelope value, 90% prediction interval and spatial interpolation results of the daily radiation in Xinjiang on July 1, 2005 were obtained as follows: Figure 7 As shown, it can be seen that the constructed model better describes the temporal and spatial distribution characteristics of solar radiation distribution.

[0102] A storage medium stores a computer program, which executes the above-mentioned daily radiation amount prediction method when running.

[0103] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and description of the present invention.

Claims

1. A method for predicting daily radiation, characterized in that: include: Collect daily historical radiation data of multiple observation points in the area, and generate daily radiation time series of each observation point based on the corresponding historical radiation data; According to the spatial correlation between the distribution law of the daily radiation time series and the spatial location of the observation points, a spatiotemporal coupling model of daily radiation is constructed. The spatiotemporal coupling model is trained based on the time series of daily radiation at each observation point; Based on the trained spatiotemporal coupling model, the solar radiation of the area to be predicted is predicted according to the corresponding spatial location parameters and time parameters; Constructing the space-time coupling model includes: The envelope model and relative deviation model of the daily radiation exposure time series are constructed based on the daily temporal variation of the daily radiation exposure data. Modeling unknown parameters of the envelope model and the relative deviation model according to the spatial characteristics of the daily radiation, and determining parameter models of the envelope model and the relative deviation model; Constructing the spatiotemporal coupling model according to the parameter model, the envelope model and the relative dispersion model; The envelope model is: Among them, β0 is the intercept of the envelope signal, t is the time, is the largest integer not exceeding t; is the coefficient of the kth basis function of the slowly varying annual mean value of the envelope signal; is the coefficient of the kth basis function of the annual amplitude of the envelope signal, is the coefficient of the kth basis function of the annual periodic fluctuation term, is the kth basis function in the s-order M-spline family with K basis functions, is an unknown parameter, K T K is the number of basis functions of the spline family in the function that describes the slow change of the mean value of the daily radiation envelope signal. D K is the number of basis functions of the spline family in the annual periodic function of the daily radiation envelope signal. M The number of basis functions of the spline family in the amplitude function of the annual periodic signal vibration; The space-time coupling model is:

2. The method for predicting daily radiation exposure according to claim 1, wherein: The relative dispersion model is fitted by a Gaussian process.

3. The method for predicting daily radiation exposure according to claim 2, wherein: The step of fitting the relative deviation model by a Gaussian process comprises: Analyze the relative deviation sequence between the actual daily radiation time series and the envelope series at each observation point; Calculate the mean of a relative deviation series; The generalized extreme value distribution function is used to characterize the mean distribution law of relative deviation series.

4. The method for predicting daily radiation exposure according to claim 2, wherein: Modeling unknown parameters of the envelope model and the relative deviation model includes: The Gaussian process model is used to model the unknown parameters of the envelope model and the relative deviation model. The parameter model is: Among them, θ i，l is the lth parameter of the exposure time series envelope model at spatial position i, l = 1, ..., L, μ l is the mean vector, σ l Gaussian process random error, Ι is the n×n order unit matrix with 1 on the diagonal and 0 on the other elements, and n is the number of sites; K l (X, X) is the kernel function matrix, and its (i1, i2)th element is: Among them, K l is the parameter that describes the uncertainty of the Gaussian process, γ l is the scale parameter; Based on the Gaussian process model, the unknown parameters of the envelope model and relative deviation model at any spatial position are determined as follows:

5. A storage medium storing a computer program, characterized in that: When the computer program is running, the daily radiation exposure prediction method according to claim 1 is executed.

Images