Prediction of Solar Radiation According to Aerosol Optical Depth

The dependencye of aerosol optical depth on wavelength as well as the fit of the humidity, temperature and pressure approximation under atmoshperic condition at Biskra city of Algeria has been investigated. Our work consists of measuring and modeling solar radiation on the horizontal area to create a mathematical model of global solar radiation which depends on the aerosol optical depth data between two wavelengths: 550 and 1250 nm. Simultaneous measurements of global solar radiation were carried out and recorded on the horizontal zone on an urban site (Biskra, Algeria) to characterize the radiative effect of atmospheric aerosols from January to December 2013. In addition, the effect of meteorological parameters such as: humidity, ambient temperature, and time durations were studied. This relationship constitutes an alternative tool to estimate AOD at the routine lighting measurements available at many radiometric stations around the world. Finally, a comparative study was established between the theoretical results and the experimental data which leads at an excellent correlation by a low relative error which is limited by the interval 2 and 15%. doi: 10.5829/ijee.2020.11.04.04


A B S T R A C T
The dependencye of aerosol optical depth on wavelength as well as the fit of the humidity, temperature and pressure approximation under atmoshperic condition at Biskra city of Algeria has been investigated. Our work consists of measuring and modeling solar radiation on the horizontal area to create a mathematical model of global solar radiation which depends on the aerosol optical depth data between two wavelengths: 550 and 1250 nm. Simultaneous measurements of global solar radiation were carried out and recorded on the horizontal zone on an urban site (Biskra, Algeria) to characterize the radiative effect of atmospheric aerosols from January to December 2013. In addition, the effect of meteorological parameters such as: humidity, ambient temperature, and time durations were studied. This relationship constitutes an alternative tool to estimate AOD at the routine lighting measurements available at many radiometric stations around the world. Finally, a comparative study was established between the theoretical results and the experimental data which leads at an excellent correlation by a low relative error which is limited by the interval 2 and 15%. doi: 10 Ghardaïa (Algeria) is a site very recognized for its solar deposit. Guermoui et al. [1] worked on this site and wrote an article on the estimation of solar radiation. They estimated and developed an SVM model to correct the error of solar radiation .
In Finland, Tuononen et al. [2] carried out a forecast study to see the influence of the cloud on the accuracy of solar radiation forecasts. To locate these clouds and see their influence on the accuracy of solar radiation forecasts in Helsinki, they have developed methods and algorithms to evaluate their performance on cloud and radiation forecasts.
There are several prediction models based on the artificial neural network (ANN), support vector regression (SVR), fuzzy logic, and the Gaussian process. Basaran et al. [3] and Ghimire et al. [4] have exploited the learning techniques in predicting the irradiance solar. To estimate the irradiance solar, Basaran et al. [3] have used a set of support vector regression (SVR) and artificial neural networks (ANN). On the other hand, Ghimire et al. [4] used the convolutional neural network and the short-term memory network to determine the global solar radiation (GSR) as a function of time. Beyaztas et al. [5] carried out a study based on the modeling, analysis, and prediction of global solar radiation in Africa and more precisely Burkina Faso. Using the concept of functional data analysis, they found a new mathematical model that predicts daily global solar radiation .
Ghimire et al. [4] used the Queensland site in Australia to predict global solar radiation an artificial neural network (ANN) model. They established models based on the neural network such as the machine learning model, they also used genetics to estimate and calculate solar irradiation. Khatib [7] has deduced a new prediction model based on artificial neural networks. The variables used were humidity, temperature, and energy from the sun. His model gave very satisfactory errors, as an example the mean square error. In Nigeria, more precisely in the city of Makurdi, Kuhe et al. [8] used several forms of artificial neural networks such as the generalized regression model to predict global solar radiation. Wang et al. [9] studied three forms of ANN to predict daily global solar radiation. Wang et al. [9] used three forms of artificial neural networks to predict solar radiation. They used the radial-based neural network, the multilayer perceptron, and the generalized regression neural network. To model daily solar radiation, they applied these three different techniques. In several countries, there is a lack of radiation database. To overcome this problem and predict solar radiation Kumar Yadav and Chandel [10], have used the models found in the literature based on artificial neural networks. To predict solar radiation, Neelamegam and Amirtham [11] used several algorithms. The results obtained by the predictive artificial neural network model used gave very high accuracy .
In Morocco (Marrakech), a new fuzzy model has been established by Iqdour and Zeroual [12]. They predicted daily global solar radiation. The neuro -flou proposed by Parsaei et al. [13], they improved the prediction model by using intelligent algorithms. Voyant et al. [14] employed machine learning methods for the global prediction of solar radiation . Rabehi et al. [15] predicted modeling based on methodologies such as linear regression and the multilayer perceptron. These models have been applied in southern Algeria to predict global solar radiation. The interest in solar energy led us to seek new methods of prediction of solar radiation. Reikard [16] tested some methods like the autoregressive integrated moving average and the networks of artificial neurons (ANN). He found that the ANN model is not optimal, especially in terms of computing time .
In semi-arid regions, Samadianfard et al. [17] studied several models using the adaptive neuro-fuzzy inference, gene expression programming, vector support regression (SVR), and empirical relationships. According to their results, the SVR model is the most efficient. In India, Suthar et al. [18] estimated solar radiation by taking into account atmospheric pollution, they predicted daily global solar radiation independent of location by the exponential quadratic system. They concluded that air pollution is more important than location . Some works were used the different meteorologic parameters such as humidity, temperature, pressure, and angle of sun elevation for determined the global solar radiation [19][20][21][22][23]. The aim of our study was to measure and model solar radiation on the horizontal area to create a mathematical model of global solar radiation which depends on the aerosol optical depth data between two wavelengths: 550 and 1250 nm.

Solar irradiation
We chose the total solar irradiance equation in the form of Siberian and variable according to the angle of solar elevation accompanied by a constant.
We noticed that each constant of the chosen equation is variable by month change, and for this, we tried to connect it with the optical spray according to the wavelengths of 550 and 1250 microns taking into account the climate values, for example, moisture and temperature .
About the impact of the aerosol optical depth onto solar radiation; we trying to find the relationship between forecast the variation of global solar radiation on a horizontal area as a function of the Meteorology parameters based on Equation  At this point, we tried to create a new approach corresponding to weather parameters and aerosols optical depth between two wavelengths 550 nm and 1240 nm, and another parameter as a function to many months; For determined Y0 easily taken out of Tables 1 and 2 help us to give all constants which relate by each month of the year. Table 2 summarized the average aerosol optical depth corresponding to two wavelengths.   Table 3 reported the constants corresponding to Y0 according to the relationship between the metrology values.
The constants corresponding to A and ε according to the relationship between the number of months are given in Table 4.
The constants corresponding to aerosols optical depth of BC, DU, OM, SS, and SU with µ =550 nm are reported in Table 5.      months. According to experimental data we established desired model. Through the two curves, we notice a convergence and convergence between the two curves, which makes the chosen model ideal. We can see that June and July estimated the symmetrical axes of the global solar radiation which partitions the values of the number of the months in the year. The global solar radiation started with minimum variation in January by 600 W/m² and continues to increase to June and July with 1100 W/m² and decrease to respectively to December with 600 W/m². Figure 4 represents the relative errors of global solar radiation as a function to numbers of the day. Remark that the maximum relative error has been selected in the number of the day 33, 125, and 300 which estimated the values of 11%, 13%, and 15%, respectively. The minimum relative error has been selected in the number of the day 75, and 200 which estimated the values of 5% and 2%. That means the prediction model gives a perfect curve of the global solar radiation.

CONCLUSION
The researchers today need to predict some parameters before they do the experimental study, such as the simulation or numerical modeling of some physic phenomena. So at these points trying to create a good approximate of global solar radiation with a perfect relationship between many parameters such as weather climate and some components of pollution in the atmosphere like the aerosols optical depth.
In conclusion that has a good result between measured data and predicted model; which resulted in a perfect approximation with a great convergent.
The result conducted that the global solar radiation has a relationship with the aerosol optical depth in both wavelengths 550 nm and 1240 nm. When the global solar radiation on the horizontal area takes a maximum value is parallel with a maximum aerosol optical depth in both waves the opposite is true.
Finally, we made the global solar radiation varying between many parameters; such as temperature, relative humidity, wind speed, air pressure, and rainfall, and aerosols optical depth. AODs increases with decreasing wavelength for all meteorological conditions.