You are on page 1of 28

New Temperature-based Models for Predicting Global Solar Radiation

Gasser E. Hassan 1,2,c , M. Elsayed Youssef 1, Zahraa E. Mohamed 3, Mohamed A. Ali 1,3, Ahmed
Hanafy2
1
Computer Based Engineering Applications Dept., Informatics Research Institute, City for Scientific
Research and Technological Applications, New Borg El Arab city, 21934, Alexandria, Egypt
2
Mechanical Engineering Department, College of Engineering and Technology Arab Academy for Science,
Technology and Maritime Transport, Alexandria, Egypt
3
Mathematics Department, Faculty of Science, Zagazig University, Zagazig, Egypt
c
Corresponding Author: Gasser E. Hassan
Email: gasser_hassan@yahoo.com Tel: +2 03 4593415 Fax: +2 03 4593409

Abstract
This study presents new ambient temperature-based models for estimating global solar
radiation as alternatives to the widely-used sunshine based models due to the unavailability of the
sunshine data at all locations around the world. New seventeen temperature-based models are
established, validated and compared with other three models proposed in the literature (Annandale,
Allen and Goodin Models) to estimate monthly average daily global solar radiation on a horizontal
surface. These models are developed using 20-years measured data of global solar radiation for
the case study location (Lat. 30˚ 51 ̀ N and long. 29˚ 34 ̀ E), and then the general formulae of the
new suggested models are examined for ten different locations around Egypt. Moreover, the local
formulae for the models are established and validated for two coastal locations where the general
formulae give inaccurate predictions. The mostly common statistical errors are utilized to evaluate
the performance of these models and identify the most accurate model. The obtained results show
that the local formula for the most accurate new model provides good predictions for global solar
radiation at different locations especially at coastal sites. Moreover, the local and general formula
of the most accurate temperature-based model also perform better than the most accurate two
sunshine-based models from the literature. The quick and accurate estimations of the global solar
radiation using this approach can be employed in the design and evaluation of the performance for
different solar applications.

Keywords: Solar energy; Solar radiation models; Regression analysis; Empirical models;
Statistical errors; Temperature-based models

1. Introduction
Solar energy is the most promising renewable energy source to supply a significant part of
world’s energy demand [1]. An accurate knowledge of solar radiation is considered as the first
step in solar energy availability assessment, also it is the primary input for different solar energy
applications [2,3]. Since the solar radiation measurement are not available due to the high cost as
well as equipment’s calibration and maintenance [4,5], different solar radiation models are
developed to estimate solar radiation. Solar radiation models are proposed to predict solar radiation
1
using different techniques which are based on different types of data such as meteorological data,
geographical data, geostationary satellite images, neural network, time series methods, physically
radiative transfer models, and stochastic weather methods [6–9]. Many studies have been
performed in order to investigate the applicability of different solar models in estimating solar
radiation at several locations. The meteorological data-based models are the most examined and
widely used models around the world [10]. These models are based on empirical correlations
depends on the most common meteorological parameters such as cloud cover, ambient
temperature, relative humidity and sunshine duration which is the most widely-used
meteorological parameter to estimate the solar radiation. The primary sunshine-based model is
proposed by Aͦngström [11] which is modified by Prescott [12] and becomes the most widely used
model for estimating the solar radiation around the world [10,13]. A study of reviewing and
statistical analyzing of different global solar radiation sunshine-based models is performed by
Despotovic et al. [14]. They attempt to evaluate various solar radiation models on global scale that
might be helpful in the selection of the most appropriate and accurate model based on the sunshine
data. Similarly, Al-Mostafa et al. [15] reported the performance of 52 sunshine-based models to
estimate the monthly average global solar radiation on horizontal surfaces in Jouf, Saudi Arabia.
The results show that some models are totally unsuitable for using in this region (Jouf), and others
vary in performance. Mecibah et al. [16] introduces the most accurate model for predicting the
monthly average daily global solar radiation on a horizontal surface for six Algerian cities. The
obtained results confirm the results of the previous studies which demonstrate that the sunshine-
based models generally more accurate than ambient temperature-based models. Also, the best
performances for the six Algerian cities are provided by the cubic and the quadratic regression
models. Robaa [17] modifies Barbaro et al. [18] model to estimate global solar radiation in Egypt,
the predicted values of Robaa model is compared with the predictions of different models. The
results show that Robaa model have the superior performance in estimating global solar radiation
over Egypt. As well, the study of introducing the best model for estimating the monthly average
global solar radiation over six major cities of Iran is carried out by Khorasanizadeh and
Mohammadi [19]. The results illustrate that for all the cities the best model is based on either the
sunshine duration only or a function of sunshine duration, ambient temperature and relative
humidity. Khalil and Shaffe [20] present a comparative study of direct, diffuse and total solar
radiation by using different models on a horizontal and inclined surfaces for Cairo city, Egypt.
They founded that three models from selected models have a good estimation of the total solar
radiation in the selected location. Ajayi et al. [21] developed a new model for estimating daily
global solar radiation over Nigeria showing a good agreement between model predictions and the
measured data. El-Metwally [22] proposed a study of sunshine and global solar radiation

2
estimation at different sites in Egypt, and attempts to present a simple nonlinear model for
evaluating relative sunshine duration and global solar radiation. The results show that the proposed
model gives the best performance compared with experimental measurements for solar radiation.
For the ambient temperature-based models, Hargreaves and Samani [23] proposed a simple
model based only on maximum and minimum temperatures to estimate the solar radiation.
Hargreaves and Samani model is modified by Annandale et al. [24] in order to calculate the effects
of decreased altitude and atmospheric thickness on the global solar radiation by suggesting a
correction factor for Hargreaves and Samani model. Similarly, Allen [25] presents a self-calibrated
model based on Hargreaves and Samani model to calculate the average monthly global solar
radiation. A simple model for estimating daily global solar radiation is proposed by Bristow and
Campbell [26], where the solar radiation is assumed to be an exponential function of temperature
difference. Goodin et al. [27] modifies Bristow and Campbell model by adding the extraterrestrial
solar radiation term in order to behave as scaling factor allowing the temperature difference term
to adapt the greater range of global solar radiation values. El-Metwally also suggests three simple
new models based on minimum and maximum temperature data and cloud cover data for
calculating global solar radiation on a horizontal surfaces in Egypt [28]. Quej et al [29] evaluates
the accuracy and the applicability of thirteen empirical models based on temperature and other
meteorological parameters for estimating global solar radiation in Yucaton Peninsula, Mexico.
Their results show that the new empirical model which is developed as a function of ambient
temperature and other metrological data provides the best estimation for solar radiation at all
stations. Similarly, Mghouchi et al. [30] examined the performance of three solar radiation models
under all sky condition for estimating different components of solar radiation flux on a horizontal
surface in Tetuan city, Morocco. The issue of evaluating the daily global solar radiation spatial
distribution from diurnal temperature ranges over the Tibetan Plateau in China is carried out by
Pan et al. [31], and they concluded that the calibrated Bristow–Camp-bell model performs well for
Tibetan Plateau and can give reasonably accurate global solar radiation estimations. Almorox et
al. [32] calibrates seven existing solar models and develops one new model for predicting global
solar radiation data using temperature data which are measured in seven stations located in Madrid,
Spain. Their results show that the empirical models based on temperature show good predictions
in any location if their coefficients are correctly adjusted. Also, the new suggested model gives the
best estimation for all stations. Besharat et al. [10] reviewed the extensive empirical models
available in the literature and evaluated their accuracy and applicability for estimating the monthly
average daily global solar radiation on a horizontal surface in Yazd city, Iran. Four models namely
El-Metwally [22], Badescu [33], Hargreaves et al. [23], and Chen et al. [34] are selected from
different categories such as sunshine-based models, cloud-based models, temperature-based

3
models and other meteorological-based models, and their performances are compared against the
measured data of Yazd city. The results show that all proposed correlation have good estimations
and El-Metwally [22] sunshine-based model provides the best prediction with a high accuracy.
In general, the sunshine-based models are the most accurate models compared with the
other meteorological parameters-based models [15]. However, sunshine data are not widely
available compared with the ambient temperature data in the standard meteorological stations [35],
as well the ambient temperature is simply measured for the most standard meteorological tasks
[36]. So, the sunshine-based models are difficult to be applied at locations where sunshine data are
not available [22,37]. Therefore, it is feasible to develop new models based on ambient temperature
as alternative models to estimate the solar radiation at different locations around the world.
This study aims to introduce new ambient temperature-based models for estimating global
solar radiation on a horizontal surface with a high accuracy. For this purpose, 20 different ambient
temperature-based models are proposed for estimating monthly average daily global solar radiation
on a horizontal surface; (17 models are introduced in this study as new models while three models
are previously developed in the literature). The performances of the new 17 suggested models are
validated and compared with the performances of the selected three models against the measured
data of global solar radiation at New Borg El-Arab city, Egypt (Lat. 30˚ 51 ̀ N and long. 29˚ 34 ̀
E). Moreover, the generalization capability of the new suggested models is examined for all Egypt
as a case study, and tested over ten Egyptian cities, namely Sidi Barrani, New Borg El-Arab, Siwa,
El-Arich, Nakhel, Hurghada, Cairo, Asyut, Kharga and Aswan. These cities are distributed over
the whole Egypt as shown in Fig. 1, and their geographical locations are donated in Table. 1.
Furthermore, the mostly common statistical errors, such as root mean square error (RMSE), mean
absolute percentage error (MAPE), mean absolute bias error (MABE) and coefficient of
determination (R2), are calculated to evaluate the performance of these models [7,10]. Based on
the validation with the experimental data, the models that have accurate estimations are recognized
and the best model between all the examined models is identified. Moreover, the local formulae
for the models are established and validated for two coastal locations, namely Sidi Barrani and El-
Arich where the general formula usually gives inaccurate predictions. Finally, the performances
of the local and general formula for the most accurate model for the city of Cairo are compared
with those for the most accurate two sunshine-based models from the literature, namely, Robaa
local model and El-Metwally general model.
The main novelty of the current work is the ability to estimate the solar radiation at any
location around the world where there is no equipment for measuring solar radiation. This can be
achieved by using an accurate and reliable ambient temperature-based models since the ambient
temperature is the most available and simply measured or forecasted metrological data. These
4
models can be considered as the backbone of any computer-program for designing different solar
energy systems. Furthermore, another novelty is the adaptation of the empirical coefficients in
order to increase the applicability of the presented models over the whole Egypt. The empirical
coefficients in these models were deduced and summarized, as well the predicted values of the
global solar radiation also are computed using the twenty correlations and compared with
measured values.

Table 1
Geographical locations for selected cities.
# City Lat. N Long. E
1 Sidi Barrani 31˚ 37 ˋ 25˚ 54 ˋ
New Borg
2 30˚ 55 ˋ 29˚ 41ˋ
El-Arab
3 Siwa 29˚ 12 ˋ 25˚ 29 ˋ
4 El-Arich 31˚ 05 ˋ 33˚ 50 ˋ
5 Nakhel 29˚ 54 ˋ 33˚ 44 ˋ
6 Hurghada 27˚ 09 ˋ 33˚ 43 ˋ
7 Cairo 30˚ 05 ˋ 31˚ 17 ˋ
8 Asyut 27˚ 12 ˋ 31˚ 10 ˋ
9 Kharga 25˚ 27 ˋ 30˚ 32 ˋ
10 Aswan 23˚ 58 ˋ 32˚ 47 ˋ

Fig. 1 Distribution of the selected cities in Egypt [38].


2. Global solar radiation models
The strong relation between the solar radiation and ambient temperature can be explained
by the behavior of the earth surface towards the radiation received from the sun [39]. The solar
energy which reaches the atmosphere in form of short wave electromagnetic radiation is absorbed
by the earth surface causing the earth to warm up. The warm earth surface reemits a part of the
absorbed energy in form of long wave radiation to heat up the surrounding ambient air. The
ambient air is not directly heated by the solar radiation but it is heated due to the contact with the
hot earth's surface. In this operation, there is a phase lag between the solar radiation and
temperature cycles [40], as shown in Fig. 2 [41]. In general, the fluctuation in air temperature can
be affected by radiation balance and air mass advection. The local ambient air temperature and
radiation balance is affected by cloud cover and the nature of surface cover as well as time of day,
day of the year. The systematic variation in incoming solar radiation over the course of a year can
reflects in the annual temperature cycle where a strong relation between solar radiation and
ambient temperature is noticed [39]. Therefore, this work focus on estimating different correlations

5
between solar radiation and ambient temperature which is considered as a significant claim in this
study.
Twenty ambient temperature-based models are analyzed; 17 models are presented as new
models in this study, and the other three models (Goodin et al. model [27], Allen model [25] and
Annandale et al. model [24]) are selected from the literature, as the following:

Model 1: 𝐺⁄ = 𝑎 + 𝑏 𝑇 (Present study)


𝐺0 (1)
Model 2: 𝐺⁄ = 𝑎 + 𝑏 𝑇 + 𝑐 𝑇 2 (Present study)
𝐺0 (2)
Model 3: 𝐺⁄ = 𝑎 + 𝑏 𝑇 + 𝑐 𝑇 2 + 𝑑 𝑇 3 (Present study)
𝐺0 (3)
Model 4: 𝐺 = 𝑎 𝑒𝑥𝑝( 𝑏 𝑇 𝑐 ) (Present study) (4)
Model 5: 𝐺⁄ = 𝑎 ∆𝑇 𝑏 + 𝑐 (Present study)
𝐺0 (5)
Model 6: 𝐺⁄ = 𝑎 𝑇 𝑏 𝐺0 + 𝑐 (Present study)
𝐺0 (6)
Model 7: 𝐺⁄ = 𝑎 𝑇 𝑏 𝐺0 (Present study)
𝐺0 (7)
Model 8: 𝐺⁄ = (𝑎 + 𝑏 ∆𝑇)∆𝑇 𝑐 (Present study)
𝐺0 (8)
Model 9: 𝐺⁄ = (𝑎 + 𝑏 ∆𝑇 + 𝑐 ∆𝑇 2 )∆𝑇 𝑑 (Present study)
𝐺0 (9)
Model 10: 𝐺⁄ = (𝑎 + 𝑏 ∆𝑇 + 𝑐 ∆𝑇 2 )∆𝑇 0.5 + 𝑑 (Present study)
𝐺0 (10)
Model 11: 𝐺⁄ = (𝑎 + 𝑏 ∆𝑇 + 𝑐 ∆𝑇 2 )∆𝑇 𝑑 + 𝑒 (Present study)
𝐺0 (11)
Model 12: 𝐺⁄ = (𝑎 + 𝑏 ∆𝑇 + 𝑐 ∆𝑇 2 + 𝑑 ∆𝑇 3 )∆𝑇 𝑒 + 𝑓 (Present study)
𝐺0 (12)
Model 13: 𝐺⁄ = (𝑎 + 𝑏 𝑇)∆𝑇 𝑐 + 𝑑 (Present study)
𝐺0 (13)
Model 14: 𝐺⁄ = (𝑎 + 𝑏 𝑇)∆𝑇 𝑐 (Present study)
𝐺0 (14)
Model 15: 𝐺⁄ = (𝑎 + 𝑏 𝑇 + 𝑐 𝑇 2 )∆𝑇 𝑑 + 𝑒 (Present study)
𝐺0 (15)
Model 16: 𝐺⁄ = (𝑎 + 𝑏 𝑇 + 𝑐 𝑇 2 )∆𝑇 𝑑 (Present study)
𝐺0 (16)
Model 17: 𝐺⁄ = (𝑎 + 𝑏 𝑇 + 𝑐 𝑇 2 + 𝑑 𝑇 3 )∆𝑇 𝑒 (Present study)
𝐺0 (17)
Model 18: 𝐺⁄ = 𝑎 (1 + 2.7 × 10−5 𝑍) ∆𝑇 0.5 Annandale et al.
𝐺0 (18)
[24]
Model 19: 𝐺⁄ = 𝑎 ∆𝑇 𝑏 Allen [25]
𝐺0 (19)
𝑐
Model 20: 𝐺⁄ = 𝑎 [1 − exp(−𝑏 ∆𝑇 )] Goodin et al. [27] (20)
𝐺0 𝐺0
where 𝑇, 𝑇𝑀𝑎𝑥 , 𝑇𝑀𝑖𝑛 , ∆𝑇, 𝐺0 are the monthly average of daily ambient temperature (˚C),
maximum temperature (˚C), minimum temperature (˚C), temperature difference (˚C) (∆𝑇 =
(𝑇𝑀𝑎𝑥 − 𝑇𝑀𝑖𝑛 )) and extraterrestrial solar radiation on horizontal surface (MJ/m2 day),
respectively. a, b, c, d, e and f are the empirical coefficients.

6
Fig. 2 Solar radiation and air temperature phase lag, [41].

3. Performance evaluation
In this study, the performance of the models is evaluated using the commonly statistical
errors or indicators, namely RMSE, MABE, MAPE and 𝑅 2 [7,42]. The accepted range of the
absolute value of these error is less than 10% [21,43]. The values of the errors are calculated
according to the following equations:
1 1/2
𝑅𝑀𝑆𝐸 = [𝑛 ∑𝑛𝑖=1(𝐺𝑖,𝑐 − 𝐺𝑖,𝑚 )2 ] (21)
1
𝑀𝐴𝐵𝐸 = 𝑛 ∑𝑛𝑖=1|(𝐺𝑖,𝑐 − 𝐺𝑖,𝑚 )|
(22)
1 𝐺𝑖,𝑐 − 𝐺𝑖,𝑚
𝑀𝐴𝑃𝐸 = 𝑛 ∑𝑛𝑖=1 |( ) × 100|
𝐺𝑖,𝑚 (23)
∑𝑛
𝑖=1(𝐺𝑖,𝑚 − 𝐺𝑖,𝑐 )
2
𝑅2 = 1 − 𝑛 ̅̅̅̅̅
∑𝑖=1(𝐺𝑖,𝑚 − 𝐺𝑚 ) 2 (24)
Where 𝐺𝑖,𝑐 is the 𝑖th calculated value and 𝐺𝑖,𝑚 is the 𝑖th measured value. ̅̅̅̅
𝐺𝑚 is the average value
of the measured and the calculated, respectively; 𝑛 is the number of observation.

RMSE values donate information about the short term performance of the model, it is
always positive value. The smaller value means a better performance of model and zero
representing the ideal case. The values of (MBE) give information about the long-term model
performance. The negative value refers to under-estimation while the positive value refers to
overestimation, and the small value is the desired condition. The values of (𝑅 2 ) show information
about the goodness of fit, its values between zero and one (0 ≤ 𝑅 2 ≤ 1), and the largest value is
the desired value.

7
4. Data collection and location description
The measured data of global solar radiation and temperature for 20 years between 1st July
1983 and 31th Dec 2004 are employed to establish and validate the applicability of models on
estimating the monthly average daily global solar radiation on a horizontal surface. Data obtained
from NASA Surface meteorology and Solar Energy web site [8,44,45]. The monthly average daily
values of global solar radiation and temperature data, as well as extraterrestrial solar radiation
values are calculated using in-house computer program. This program is developed using C#
language [46], the software engineering activities (specification, design, validation, and evolution)
are utilized in the development process. The software engineering is the discipline that concerned
with software development [47,48].
Extraterrestrial radiation, 𝐺0 , is the radiation which originates outside the atmosphere,
and its equation is defined as [49,50]:
24 × 3600 𝐺𝑠𝑐 𝜋𝜔
𝐺𝑜 = 𝑘 [( ) sin(𝐿) sin(𝛿) + cos(𝐿) cos(𝛿) cos(𝜔)] (25)
𝜋 180
where 𝐺𝑠𝑐 presents the solar constant and its value equal 1367 W/m2 [51,52]. k is the eccentricity
correlation factor of the earth’s orbit, 𝜔 is the hour angle at sunset (˚), 𝐿 is the latitude angle (˚),
and δ is the declination angle (˚). k, δ, 𝜔 equations are given as [53]:
360 𝑁
𝑘 = [1 + 0.033 cos ( )] (26)
365
360 (27)
𝛿 = 23.45 𝑠𝑖𝑛 [ (284 + 𝑁)]
365
𝜔 = cos−1 [− tan(𝐿) tan(𝛿)] (28)
where 𝑁 presents day number of the year starting from 1st January.

Egypt’s climate can be classified into three categories. The first category is the coastal
regions either on the Mediterranean or red sea coast such as cities of Sidi Barrani, Borg El-Arab,
El-Arich and Hurghada. The second category is the semi-arid warmer dry climate a broad summer
dry season such as Cairo while the third category is the dry desert climate such as the city of Asyut,
Aswan and Kharga [28]. The selected locations in the study are distributed over the whole Egypt
from the north to south. Generally, Egypt’s climate is characterized by clear sky during the summer
season (from June to August) and partially cloudy skies during the autumn (September to
November) and spring (March to May). On the other side, the winter season for the most of Egypt’s
sites is characterized by cloudy sky especially the northern part, where the amount of the cloud
decreases from north to south. Besides, Egypt is one of countries that located in the most favorable

8
solar belt which enjoys with abundant solar radiation (3,500 – 4,500 hours of sunshine per year,
12–30MJ/m2/day of solar energy magnitude) [22].
For the ambient temperatures, Egypt’s climate is characterized by a cold season from
October to February as well as a hot season from March to September. The northern winds are a
welcome respite from the heat and it keeps the temperatures slightly moderated. Egypt’s weather
also is characterized by especially good wind regimes with excellent locations along the Red Sea
and Mediterranean coasts. Regions with an annual average of 6.0–6.5 m/s have been recognized
along the Mediterranean coast and about 8–10 m/s along the Red Sea coast. Average annual
temperatures also vary from a minimum of 14 ̊ C to a maximum of 37 ̊ C in the coastal locations,
as well as the most humid area is along the Mediterranean coast where the average annual rainfall
is about 200 mm. Rainfall reduces quickly towards the south of Egypt, also in many desert sites it
may rain only once in several years. The temperature values change widely in the inland desert
locations, particularly in summer season (range from 7 ̊ C at night to 43 ̊ C during the day). In
winter season, temperature values in the desert fluctuate less dramatically (range from 0 ̊ C at night
to 18 ̊ C during the day). In general, the average annual temperature increases, moving southward
from the Delta in the north to the south of Egypt, where temperature values are similar to those of
the open deserts to the east and west. As well, through the Delta and the northern Nile Valley,
there are occasional winter cold spells which may be accompanied by light frost and even snow.
There are some sandstorms occurring due to hot spring wind which is known in Egypt as Khamsin
winds and equivalent to the sirocco winds in Europe. These sandstorms may continue for days
during the spring and can lead to temperature rise of 20 ̊ C in 2 hour [38].
5. Results and discussion
The measured data of global solar radiation are divided into two data sets and averaged to
obtain the monthly average daily values. The first data set (1st July 1983 to 31th Dec 2002) is
employed to establish models using regression analysis [10,19,21]. The optimal values of
empirical coefficients corresponding to the actual data at New Borg El-Arab are obtained and
reported in Table. 2. The second data set (1st Jan 2003 to 31th Dec 2004) is employed to evaluate
and validate the developed models using statistical indicators. The predictions of the 20 models
are compared with the measured data of global solar radiation. Table. 2 also summarizes the
statistical errors (RMSE, MAPE, MABE and R2) which are calculated using equations Eqs. (21)-
(24). The acceptable models are identified and the best model is recognized by comparing the
statistical errors associated with all models. The best model is indicated in bold as shown in Table.
2.

9
Table 2
Empirical coefficients and statistical errors for new models (1-17) and selected models (18-20) at Borg El-Arab city.
Model a b c d e f RMSE MِAPE MABE R2 Rank
1 0.34710 0.01218 1.3443 6.2131 1.1326 0.964322 10
2 0.21637 0.02485 -0.00029 1.3302 6.1135 1.1133 0.965069 8
3 0.31469 0.01030 0.00040 -0.00001 1.3233 6.1236 1.1127 0.965426 7
4 1.73780 -3.59174 -0.40530 1.3352 6.1104 1.1140 0.964804 9
5 -5608010.03 -7.40735 0.70119 1.7262 8.4695 1.5393 0.941171 11
6 0.00237 0.38353 0.35394 0.7308 2.8785 0.6159 0.989457 1
7 0.03470 -0.20776 1.7935 9.5227 1.5999 0.936493 12
8 -20868.1684 2255.32473 -3.26103 3.7080 18.0328 2.9800 0.728549 20
9 -0.00086 0.00018 -8.3710E-06 3.86853 2.8636 13.2805 2.3851 0.838105 16
10 -0.76182 0.74262 -0.03377 -9.46160 3.2108 15.7566 2.6756 0.796471 19
11 -0.00068 0.00016 -7.7423E-06 3.85457 -0.21871 2.8665 13.3718 2.3854 0.837780 17
12 -0.64896 -3.03712 0.52194 -0.02090 -0.20542 0.76692 3.1432 15.4269 2.6340 0.804944 18
13 1.34583 0.00656 0.15275 -1.56492 1.1780 5.0689 1.0399 0.972602 2
14 0.10369 0.00255 0.54026 1.1897 5.1479 1.0557 0.972058 3
15 0.12899 -0.00161 8.51339E-05 0.60111 0.00100 1.2094 4.7102 1.0087 0.971124 4
16 0.12952 -0.00162 8.53993E-05 0.60033 1.2095 4.7107 1.0088 0.971121 5
17 0.10501 0.00191 -8.3692E-05 2.60E-06 0.60301 1.2190 4.7305 1.0132 0.970665 6

18 0.18122 1.9150 9.6555 1.6965 0.927599 15


19 0.08362 0.80852 1.8056 7.3727 1.4836 0.935636 13
20 0.66755 4.0437E-08 8.83610 1.8741 9.6451 1.6723 0.930660 14

According to the comparison with the measured data, Models 8, 9, 10, 11 and 12 showed
good RMSE and MABE values in the acceptable range ±10%. On the other hand, their R2 values
are small and the values of MAPE exceed the acceptable range; 18.0328%, 13.2805%, 15.7566%,
13.3718% and 15.4269%, respectively. Also, the performance of Model 9 and 11 are close to each
other, and the worst performance is displayed by Model 8 with the largest MAPE value 18.0328
%, which refer to a bad fitting and also their values are far from the optimum values. Therefore,
these models (Models 8, 9, 10, 11 and 12) are not suitable for estimating monthly average daily
global solar radiation and hence these models can be excluded due to their inaccurate predictions.
On the other side, Model 6 donated the best performance between all other models with
the highest R2 value [7,21,43], followed by Model 13, 14, 15, 16 and 17, successively. Its statistical
errors RMSE, MAPE, MABE, and R2 equal to 0.7308 MJ/m2, 2.8785 %, 0.6159 MJ/m2 and
0.989457, respectively. Also, Models 13, 14, 15, 16 and 17 have a slightly variation in their
performance with an excellent R2 values. Similarly, the performance of Models 1, 2, 3 and 4 are

10
very closer to each other. The statistical results also showed that Allen model (Model 19)
performed the superior performance among the selected models, followed by Goodin et al. model
(Model 20) and Annandale et al. model (Model 18), respectively. Moreover, the other models have
a good agreement between the measured and the predicted solar radiation values where R2 values
are larger than 92% which refer to a well-fitting. Based on the obtained results, all models are
ranked according to their R2 values where the best model has the highest value of R2 [7,21,43], as
introduced in Table. 2. Different predictions of the suggested models are compared with the
measured data as clarified in Fig. 3. The performances of models (1-7) are displayed the subplot
(a) in Fig. 3 and the performances of models (8-12) are illustrated the subplot (b) in the same figure
as well as the performances of models (13-20) are shown in the same figure in the subplot (c). In
addition, the performances of the best new suggested models (Models 6 and 13) and the most
accurate model between the selected models (Allen model) are compared against the measured
data as seen in the subplot (d) of Fig. 3. The results also show that the relations with second and
the third order of temperature, Model 2 and 3 respectively, slightly improve the performance
compared with the linear relation in Model 1, as noticed in Table. 2. Table. 4 summarizes the range
of statistical errors for the new accepted suggested models (Model 1-7, 13-17). It can be noticed
that the best values of statistical errors are presented by Model 6 and the worst errors values are
displayed by Model 7. Although both model 6 and 7 are based on temperature and extraterrestrial
solar radiation, adding the constant term in Model 6 effectively improves the performance
compared with that for Model 7. Moreover, Model 6 gives the best performance among all models.
Figure 4 compares the performance of Model 6 with those of the other models using Taylor
diagram [54]. The main advantages of showing performances of the models in Taylor diagram is
that the models are obviously grouped according to their performances. Fig. 4 shows that Model 6
has the superior performance among other models where it is the nearest model to the reference
point. As well, the representative points of models 1 – 4 and 13 – 17 are very close and overlaped,
which means that their performances are very similar. Similarly, the representative points of the
selected models 18 – 20 (Allen [25], Goodin et al. [27] and Annandale et al. [24] model) are also
close to each other, which refers that these models also have a converged performances. In
opposition to Model 6, the representative points of models 8 – 12 (Rejected models) have the
furthest representative points from reference point, which indicates that these models have the
lower performances compared with those for the other models. As well, their representative points
are near to each other which interoperates that their performances are adjacent.

11
30 30
(a) (b)
Monthly average daily global solar radiation (MJ/m^2 day)

Monthly average daily global solar radiation (MJ/m^2 day)


27 27

24
24

21
21
18
18
15
15
12

12
9

9 6

6 3
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month Month
Measured GSRH Model1 Model2
Measured GSRH Model8
Model3 Model4 Model5 Model9 Model10
Model6 Model7 Model11 Model12

30 30
(c) (d) (Best models)
Monthly average daily global solar radiation (MJ/m^2 day)
Monthly average daily global solar radiation (MJ/m^2 day)

27 27

24 24

21 21

18 18

15 15

12 12

9 9

6 6
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month Month
Measured GSRH Model13 Model14
Measured GSRH Model13
Model15 Model16 Model17
Model18 Model19 Model20 Model19 Model6

Fig. 3 Overall performance of the models as well as the best models compared with the measured data.
12
Fig. 4 Taylor diagram for the new presented models 1-17 and selected models 18-20 ( Allen
model, Goodin et al. model and Annandale et al. model) at New Borg El-Arab city.

Based on the comparison between the measured and the predicted values of monthly
average daily global solar radiation, the new suggested models (1-7 and 13-17) give more accurate
estimation than the three existing tested Allen, Goodin et al. and Annandale et al. (Model 18, 19
and 20) as illustrated by statistical performance in Table. 2. The statistical performance for all the
investigated models (17 new suggested models and 3 existing models) is presented in Fig. 5. It is
clearly shown that models (1-7 and 13-17) are more accurate than Allen, Goodin et al. and
Annandale et al. model (18-20) and Model 6 has the best values for statistical indicators among all
models. In addition, it can be seen that five models (8-12) are excluded due to large MAPE values
which exceed the acceptable range ±10%. From the above results, twelve models (1-7 and 13-17)
of new presented models in this study have better performance than the performance of the existing
models in literature Allen [25], Goodin et al. [27] and Annandale et al. [24] model (Model 18- 20).
Moreover, Model 6 proved to be the most suitable model for predicting monthly average daily
global solar radiation on a horizontal surface at New Borg El-Arab city.

13
The obtained results of the new accepted models (1-7 and 13-17) and the three calibrated
models (18 – 20) show that the statistical indicators values are within the acceptable range (±10%)
compared with other previous studies which used the temperature-based models for the estimation
of global solar radiation [32]. In this work, the range of R2 and RMSE values for the new accepted
models are 0.936 – 0.989 and 0.73 - 1.79 (MJ m-2 day-1), respectively, while the range for previous
studies are 0.884 – 0.895 for R2 and 2.69 – 3.367 (MJ m-2 day-1) for RMSE [32]. Moreover,
additional statistical errors such mean percentage error (MPE), mean bias error (MBE) and t-Test
statistic (t) are calculated based on Eqs. 30 – 32 and summarized in Table 3. The obtained results
display that all statistical indicators have good values within the acceptable range. The range of
errors (MBE, MPE and t) of new presented models also are summarized in Table. 4.

Table 3 Table 4
MPE, t-Test and MBE values for new accepted models (1-7 Errors range for new accepted
and 13-17) and selected models (18-20) at Borg El-Arab city. models at New Borg El Arab.
Model t-Test MPE MBE Minimum Maximum
1 0.9730 0.0450 -0.3784 RMSE 0.7308 1.7935
2 0.9294 -0.0842 -0.3589 MAPE 2.8785 9.5227
3 0.9043 -0.0231 -0.3481 MABE 0.6159 1.5999
4 0.9500 -0.1344 -0.3677 R2 0.936493 0.989457
5 1.2174 -2.8479 -0.5948 t-Test 0.7324 2.7115
6 1.4770 -1.0341 -0.2973 MPE -4.1544 0.0449
7 0.7324 -4.1544 -0.3867 MBE -0.7715 -0.2972
13 2.4869 -3.0722 -0.7067
14 2.4018 -2.9045 -0.6978
15 2.6659 -3.0490 -0.7577
16 2.6666 -3.0502 -0.7578
17 2.7115 -3.1294 -0.7715

18 0.1663 2.9566 -0.0959


19 0.7463 0.2190 -0.3964
20 1.1084 -2.5652 -0.5940

14
4.0
3.5
RMSE (MJ/m2 day)

3.0
2.5
2.0
1.5
1.0
0.5

20
18
16
14
MAPE (%)

12
10
8
6
4
2

3.5
3.0
MABE (MJ/m2 day)

2.5
2.0
1.5
1.0
0.5

1.00
0.95
0.90
0.85
R2

0.80
0.75
0.70

Fig. 5 Statistical errors graph for the new models (1-17) and selected models (18-20) at New Borg El-Arab.

15
Furthermore, the measured data at ten Egyptians cities (as seen in Table. 1) are averaged
and employed to examine the generalization capability for the accepted models (1-7 and 13-17) of
new suggested models over the whole Egypt. The general empirical coefficients of these models
are obtained from averaged measured data of ambient temperature and global solar radiation for
the ten cities using regression analysis [10,19,21], as presented in Table. 5. Models predictions are
compared against the measured data for ten cities. Table. 6 shows the statistical indicators for the
estimated values of monthly average daily global solar radiation by different general models at ten
different cities. According to the obtained results, all models show an acceptable performance at
different locations except for Sidi Barrani city, as well as few models at specific locations have
slightly large MAPE values. The low performance is due to the different weather conditions
especially at coastal locations such as Sidi Barrani city [28,43]. This large error is also noticed in
previous studies at the same locations [22]. For each city the best model is identified according to
the statistical errors associated with all models and the models are ranked according to their
performance as listed in Table. 6.
For New Borg El-Arab city; the statistical indicators for all models are in the acceptable
range (±10%) except Model 5 with unacceptable MAPE value of 14.9 %. The performances of
models 1, 2, 3 and 4 are very similar to each other, as well as Model 13 and 14, and also Model 15
and 16, as shown in Table. 6. The lower performance is donated by Model 7. Model 17 displays
the best performance among all models with the highest R2 of 0.98 followed by Model 6 and 14.
For Sidi Barrani city; all models display a low performance where MAPE values exceed
the acceptable range. The bad performance for these models can be explained by different weather
condition at costal locations such as Sidi Barrani [28]. Also, performance of general models are
often lower than the performance of the models that calibrated against the local data [35].
However, the performance of Model 13 may be acceptable where its MAPE value is slightly larger
than the acceptable value with good R2 value.
For Siwa city, all statistical errors are in the acceptable range except the unacceptable
MAPE value of 14.34 % for Model 7. Model 16 displays the best performance among other models
with the highest R2 of 0.97 followed by Model 15, Model 14 then Model 13.
For El-Arich city, six models (Model 1-6) show good predictions with errors values in the
acceptable range, and R2 values larger than 0.91. Model 6 gives the best results among all models
with R2 value of 0.96, followed by Model 5. The other six models (Model7, 13-17) displayed a
bad performance which can be explained by the weather condition at El-Arich city which is
considered as a costal site, as shown in Fig. 1.
16
Table 5
General Empirical coefficients of new models (1-7 and 13-17) for the whole Egypt.
Model a b c d e
1 0.45589 0.00808
2 0.50298 0.00355 0.00010
3 0.52144 0.00084 0.00023 -1.94305E-6
4 34.61726 -4.92844 -0.06804
5 -3.07726 -8.93456 0.64974
6 0.00056 0.65041 0.49158
7 0.05607 -0.35090
13 1.47024 0.00587 0.08675 -1.33823
14 0.24906 0.00384 0.26556
15 0.71199 -0.00886 0.00030 0.20216 -0.45701
16 0.30399 -0.00627 0.00021 0.34735
17 0.12097 0.00203 -0.00008 2.18286E-6 0.58099

For Nakhel city; the best performance is presented by Model 14 with R2 value of 0.986.
Model 13 and 6 illustrate the second and the third-best performance and the statistical performance
of Model 13 is very close to the performance of the best model (Model 14). The remaining models
introduce a good performance expect for Model 5, where the unacceptable MAPE value of 11.12
% is presented. The eleven models have good R2 values which are higher than 0.90.
For Hurghada city; Model 7 demonstrates the worst performance among the models, where
the MAPE value of 11.01 % exceeds the acceptable value. Model 6 presents the best performance
with R2 value of 0.94, followed by Model 5. The remaining models also show good performances
with acceptable values for the statistical errors, as appeared in Table. 6.
For Cairo city; the most accurate prediction is obtained by Model 2 with R2 value of 0.985,
followed by Model 3 and then Model 1. The worst estimations are predicted by Model 7 with
unacceptable MAPE value of 13.31 %. The remaining models give a good estimations for global
solar radiation.
Unlike the cities located at coastal regions, for Asyut city; the values of all statistical
indicators for all models are in the acceptable ranges. Model 1, 2, 3 and 4 have very slightly
differences in their performance, where R2 values larger than 0.94. Based on the preference of the
statistical indicators, Model 13 provided the best performance among all models with R2 value of
0.964 followed by Model 14.
For Kharga and Aswan cities; similar to Asyut city, all models give good estimations and
all statistical indicators are in the acceptable range except Model 17 for Aswan city which gives
MAPE value of 11.44 %. The most accurate performance is presented by Model 6 with the highest

17
R2 values of 0.97 and 0.96. For both the two cities the worst performance is displayed by Model
17 with R2 value of 0.82 for 0.41 for Kharga city and Aswan city respectively.
Table 6
Statistical performances for generalized models (1-7 and 13-17) at ten cities in Egypt.
Site Model RMSE MAPE MABE R2 Rank Site Model RMSE MAPE MABE R2 Rank
1 2.8472 21.1383 2.4665 0.839713 6 1 1.4462 5.8121 1.1542 0.929291 5
2 2.8400 21.1379 2.4626 0.840528 5 2 1.4114 5.6249 1.1246 0.932648 3
3 2.8393 21.1262 2.4608 0.840609 4 3 1.4115 5.6226 1.1252 0.932639 4
4 2.8676 21.2103 2.4817 0.837411 7 4 1.5002 6.0766 1.1939 0.923906 7
5 3.7972 28.2490 3.3077 0.714904 11 5 1.3629 4.4420 1.0385 0.937200 2
Sidi Barrani

Hurghada
6 2.6847 19.7144 2.4184 0.857494 3 6 1.3441 5.7669 1.1091 0.938919 1
7 3.7080 15.0526 2.7345 0.728145 10 7 2.2684 11.0066 2.0241 0.826023 11
13 2.5134 11.1720 2.0194 0.875098 2 13 1.4778 6.1681 1.2723 0.926168 6
14 2.4477 11.8518 2.0092 0.881543 1 14 1.5393 6.4215 1.3330 0.919887 8
15 3.3166 11.8123 2.4727 0.782509 9 15 1.7113 6.3125 1.4373 0.900987 9
16 3.2001 12.2119 2.4385 0.797527 8 16 1.7856 6.5462 1.4972 0.892198 10
17 5.2406 21.1580 4.2667 0.456977 12 17 2.8336 10.0914 2.3438 0.728531 12

1 1.4994 8.3714 1.3084 0.955612 8 1 0.7211 3.0900 0.6015 0.984206 3


2 1.5220 8.5739 1.3410 0.954264 10 2 0.7035 3.0933 0.6007 0.984967 1
3 1.5211 8.5796 1.3412 0.954318 9 3 0.7047 3.1021 0.6027 0.984917 2
4 1.4781 8.0902 1.2639 0.956865 7 4 0.7629 3.3251 0.6352 0.982322 4
Borg El-Arab

5 2.5039 14.9130 2.2323 0.876223 12 5 1.5330 8.5818 1.3594 0.928620 11


Cairo

6 1.1715 6.8671 1.0689 0.972905 2 6 0.7876 2.8387 0.5839 0.981156 5


7 2.1563 9.3174 1.7551 0.908203 11 7 2.7181 13.3072 2.4297 0.775581 12
13 1.2282 7.0308 1.0668 0.970216 4 13 1.0480 3.8478 0.7938 0.966641 7
14 1.2277 7.0576 1.0684 0.970243 3 14 1.0462 3.8039 0.7869 0.966756 6
15 1.2435 7.6096 1.1398 0.969470 6 15 1.1144 4.3030 0.8558 0.962279 8
16 1.2433 7.6292 1.1406 0.969481 5 16 1.1161 4.3095 0.8578 0.962165 9
17 0.9861 5.0632 0.8292 0.980804 1 17 1.5099 5.2288 1.1232 0.930748 10

1 1.4513 6.3830 1.3262 0.944465 8 1 1.4867 5.8997 1.2437 0.944819 7


2 1.4302 6.2972 1.3133 0.946061 6 2 1.4871 5.9118 1.2344 0.944789 8
3 1.4306 6.2953 1.3133 0.946034 7 3 1.4891 5.9260 1.2370 0.944645 9
4 1.4878 6.5006 1.3446 0.941635 9 4 1.4904 5.8632 1.2540 0.944548 10
5 1.9786 5.9167 1.4606 0.896774 11 5 2.0692 9.6121 1.8433 0.893111 12
Asyut

6 1.5770 7.3197 1.4590 0.934422 10 6 1.4057 5.5400 1.2061 0.950666 6


Siwa

7 3.0415 14.3436 2.6895 0.756070 12 7 2.0266 7.8820 1.6107 0.897468 11


13 1.1533 5.4497 1.0956 0.964926 4 13 1.2007 5.3419 1.0537 0.964007 1
14 1.1529 5.4843 1.0994 0.964949 3 14 1.2142 5.3278 1.0535 0.963196 2
15 0.9832 4.6193 0.9244 0.974509 2 15 1.2637 5.2379 0.9551 0.960133 3
16 0.9824 4.6244 0.9226 0.974552 1 16 1.2773 5.2425 0.9581 0.959272 4
17 1.2578 5.8075 1.0727 0.958282 5 17 1.3890 4.7801 0.9452 0.951837 5

18
Table 6 (Continued)
Site Model RMSE MAPE MABE R2 Rank Site Model RMSE MAPE MABE R2 Rank
1 1.8867 6.2885 1.5165 0.921269 4 1 0.9903 3.4833 0.7180 0.964467 4
2 1.9432 6.4855 1.5645 0.916483 5 2 0.9747 3.2981 0.6939 0.965580 3
3 1.9435 6.4880 1.5650 0.916461 6 3 0.9745 3.2864 0.6921 0.965590 2
4 1.8047 6.2061 1.4733 0.927968 3 4 1.0217 3.7549 0.7518 0.962175 5
5 1.7956 8.1542 1.5596 0.928692 2 5 1.2544 5.1427 1.0200 0.942991 8
El-Arich

Kharga
6 1.4060 5.3763 1.1976 0.956276 1 6 0.8673 3.2301 0.6647 0.972747 1
7 2.9443 14.4831 2.6407 0.808267 7 7 1.6614 7.4742 1.4403 0.899993 11
13 4.5067 19.5439 4.1554 0.550790 9 13 1.2129 4.5161 1.0010 0.946701 6
14 4.3267 18.3267 3.9459 0.585955 8 14 1.2532 4.6082 1.0305 0.943099 7
15 5.3901 22.8905 4.9149 0.357428 11 15 1.5368 5.8055 1.2932 0.914430 9
16 5.2334 21.9138 4.7406 0.394233 10 16 1.5822 5.9227 1.3266 0.909294 10
17 7.2967 32.9122 6.8482 -0.177570 12 17 2.2068 7.3836 1.7375 0.823554 12
1 1.4592 5.7558 1.2224 0.952085 8 1 1.1236 3.9560 0.8652 0.945199 3
2 1.5206 6.0025 1.2615 0.947966 9 2 1.1959 4.2239 0.9405 0.937919 5
3 1.5208 6.0159 1.2607 0.947954 10 3 1.1890 4.1951 0.9316 0.938634 4
4 1.3950 5.3486 1.1649 0.956205 7 4 1.0386 3.6114 0.7677 0.953175 2
5 2.0814 11.1166 1.8644 0.902508 11 5 1.2883 5.0189 1.1386 0.927955 6
Nakhel

Aswan
6 1.0983 4.2179 0.8510 0.972852 3 6 0.9653 3.7074 0.7642 0.959552 1
7 2.4064 9.4801 1.8995 0.869688 12 7 2.0931 8.9371 1.7674 0.809832 9
13 0.8015 3.6075 0.6806 0.985543 2 13 1.6099 5.6114 1.3466 0.887493 7
14 0.7963 3.6138 0.6791 0.985729 1 14 1.7021 5.9031 1.4215 0.874245 8
15 1.1447 5.9476 0.9914 0.970511 5 15 2.5492 7.9923 1.9896 0.717915 10
16 1.1394 5.8934 0.9791 0.970785 4 16 2.6672 8.3048 2.0737 0.691200 11
17 1.2816 4.5583 0.9272 0.963040 6 17 3.6789 11.4449 2.8720 0.412528 12

Based on the previous results, the general formulas of new suggested models (Model 1-7
and 13-17) displayed a good performance at different locations in Egypt, except for Sidi Barrani
city. This can be explained by the weather condition at costal sites such as Sidi Barrani city
[22,28,43]. However, some models at Sidi Barrani provided acceptable errors with slightly
increase in MAPE values, such as Model 13 and 14. In addition, it can be seen that Model 6 gives
the best estimations for four cities namely El-Arich, Hurghada, Kharga and Aswan, where its R2
values are higher than 0.93 at all cities except Sidi Barrani city. The predicted results for Model 6
in its generalized form for the ten sites are consistent with the previous results of the local Model
6 at New Borg El-Arab city which demonstrates that Model 6 has the best accuracy among other
models, as indicated in Table. 2.

According to the obtained results of the generalized models, the empirical coefficients of
the new proposed models can be adjusted against the local data for the cities where the models
show poor performance, namely Sidi Barrani and El-Arich cities. The empirical coefficients and

19
statistical errors for Sidi Barrani city and El-Arich city are calculated and summarized in Table. 7
and Table. 8, respectively. All models are ranked according to their performance as shown in the
two tables.
The obtained results for Sidi Barrani city and El-Arich city show that Models 8-12 are also
excluded from estimating solar radiation due to the unacceptable MAPE values. Similarly, the old
models Allen [25], Goodin et al. [27] and Annandale et al. [24] model (Model 18- 20) as well as
the new suggested models Models 1-5 at Sidi Barrani city and Model 7 at El-Arich city are also
excluded for the same reason (MAPE > ±10%). On the other hand, the best performance is
presented by Model 6 with an excellent R2 values of 0.98 and 0.99 for Sidi Barrani and El-Arich
cities respectively, also the others models donated a good estimations. In general, the results show
that the adjusted models against local data for each site effectively increase the accuracy of the
models, as seen in statistical errors for local and general models. Moreover, Model 6 shows the
best estimations when its empirical coefficient adjusted against local data for a certain site.
Table 7
Empirical coefficients and statistical errors for new models (1 -17) and selected models (18-20) at Sidi Barrani city.
Model a b c d e f RMSE MِAPE MABE R2 Rank
1 0.31132 0.01238 1.8936 13.2142 1.7029 0.929106 11
2 0.23429 0.01990 -0.00018 1.8804 13.0007 1.6735 0.930087 10
3 0.05370 0.04662 -0.00146 0.00002 1.8748 12.9283 1.6638 0.930503 9
4 2.44664 -3.73243 -0.31142 1.8737 12.8889 1.6587 0.930582 8
5 -240132381.72 -12.91027 0.62096 2.7872 18.4997 2.2674 0.846403 14
6 0.00114 0.56911 0.35253 1.0876 7.7297 0.9785 0.976613 1
7 0.03178 -0.20776 1.6696 8.8634 1.4181 0.944881 7
8 -14614.04617 3041.99304 -3.89350 5.9146 38.0684 4.8110 0.308321 20
9 -0.07285 0.02873 -0.00249 3.25390 4.7779 23.8854 3.9086 0.548641 16
10 0.30690 3.71298 -0.34317 -21.61652 4.8931 28.7575 4.1461 0.526602 19
11 0.00058 0.00027 -4.4E-05 4.88746 -0.47338 4.7871 23.2705 3.8810 0.546903 17
12 -0.01194 -0.00135 0.00242 -0.00026 3.24392 0.05289 4.7895 23.1116 3.8727 0.546446 18
13 1.14858 0.00898 0.15477 -1.20373 1.4912 8.1704 1.2322 0.956031 2
14 0.13151 0.00507 0.47765 1.5175 8.4426 1.2581 0.954467 3
15 0.28678 -0.01492 0.00043 0.65157 0.00101 1.5655 8.0860 1.2698 0.951542 5
16 0.28768 -0.01495 0.00044 0.65039 1.5654 8.0849 1.2696 0.951551 4
17 0.13166 0.00206 -7.9E-05 3.94E-06 0.62262 1.5727 7.1894 1.2315 0.951096 6
18 0.23497 2.4987 15.7529 2.2152 0.876548 13
19 0.23321 0.50487 2.4969 15.7108 2.2130 0.876731 12
20 0.58900 0.05000 8.83610 2.9218 21.0101 2.6232 0.831203 15

20
Table 8
Empirical coefficients and statistical errors for new models (1 -17) and selected models (18-20) at El-Arich city.
Model a b c d e f RMSE MِAPE MABE R2 Rank
1 0.45402 0.01005 0.8775 3.8060 0.7490 0.982970 5
2 0.08733 0.04766 -0.00093 0.8124 3.6484 0.6907 0.985401 3
3 0.57328 -0.02831 0.00294 -0.00006 0.7928 3.4416 0.6538 0.986099 2
4 19.58070 -4.45101 -0.09063 0.8441 3.5881 0.7117 0.984242 4
5 -1140557.58 -21.56848 0.66697 1.6533 8.2806 1.4337 0.939547 10
6 0.00067 0.65219 0.49351 0.5905 2.5425 0.5175 0.992289 1
7 0.03994 -0.24792 2.8101 14.4504 2.4803 0.825342 15
8 -150.57956 34.55465 -1.71377 5.8927 25.0685 4.9635 0.232013 18
9 0.00084 0.00008 -2E-05 4.63233 6.1247 25.8377 5.1305 0.170346 20
10 0.06999 1.48641 -0.11849 -8.98573 5.9732 25.2029 4.9089 0.210887 19
11 -0.05361 0.01301 -0.00081 4.34709 13.70424 5.7826 25.0057 5.1225 0.260445 17
12 0.58884 -0.30813 0.04992 -0.00259 2.98242 7.25642 5.7731 24.9958 5.1228 0.262869 16
13 1.42941 0.00834 0.10627 -1.30663 1.0649 4.0547 0.8889 0.974917 8
14 0.24247 0.00557 0.31713 1.0692 4.0503 0.8908 0.974718 9
15 0.16620 0.01501 -0.00022 0.28943 0.00119 1.0127 3.8626 0.8475 0.977317 6
16 0.16706 0.01502 -0.00022 0.28894 1.0127 3.8626 0.8475 0.977316 7
17 0.15093 0.00216 -7.6E-05 3.02E-06 0.63677 1.7805 7.0489 1.4738 0.929881 13
18 0.25215 1.9932 8.0879 1.7031 0.912127 14
19 0.37789 0.29297 1.7327 8.0010 1.5612 0.933595 12
20 0.66697 9.195943 2.28550 1.6533 8.2806 1.4337 0.939546 11

In order to investigate the accuracy of Model 6, the performance of local and general
formula of Models 6 is compared with predictions of some existing sunshine-based models in
literature for Cairo city (Egypt’s capital). The measured data for Cairo city between 1st Jan 2010
and 31th Dec 2010 is used for validation. Measured data of global solar radiation and sunshine
hours are retrieved from World Radiation Data Center website and other data are obtained from
Reliable Prognosis website [14,22,55–58]. Based on previous studies, sunshine-based models are
the most used solar radiation models compared with the models which depend on other
meteorological parameters due to their relatively high accuracy [10,15]. Additional statistical
errors are calculated, such as relative percentage error (𝑒) for each month, mean percentage error
(𝑀𝑃𝐸), mean bias error (𝑀𝐵𝐸) and t-Test statistic (𝑡) are defined as [10]:
𝐺𝑖,𝑐 − 𝐺𝑖,𝑚
𝑒=( ) × 100 (29)
𝐺𝑖,𝑚
1 𝐺𝑖,𝑐 − 𝐺𝑖,𝑚
𝑀𝑃𝐸 = 𝑛 ∑𝑛𝑖=1( ) × 100 (30)
𝐺𝑖,𝑚
1
𝑀𝐵𝐸 = 𝑛 ∑𝑛𝑖=1(𝐺𝑖,𝑐 − 𝐺𝑖,𝑚 ) (31)
1/2
(𝑛−1)(𝑀𝐵𝐸)2 (32)
𝑡 = [(𝑅𝑀𝑆𝐸)2 −(𝑀𝐵𝐸)2 ]

21
As discussed before, Robaa [17] examines the performance of all models available for
estimating the monthly average daily global solar radiation on a horizontal surface, 𝐺, in Egypt.
The results showed that the local formula of Robaa model provides the most accurate predictions
of G for all the nine stations, and hence it is the recommended model for calculating 𝐺 at any site
around Egypt. As well, Besharat et al. [10] evaluates the applicability of global solar radiation
models for calculating the monthly average global solar radiation on a horizontal surface in Iran.
They conclude that El-Metwally [22] sunshine-based model has the best estimations with high
accuracy. Based on these results, the predictions of the new general formula of Model 6 is
compared with prediction of general formula of El-Metwally model [22] at Cairo city. As well,
the local formula of new suggested models for Cairo city are established and the best model is
recognized and its predictions are compared with those for local Robaa model [17] at Cairo city.
The values of the empirical coefficients of Robaa model, Eq.33, and El-Metwally model, Eq.34,
are obtained by their own authors as follow, [17] and [22],
𝐺 = 13.9 (𝑠)1.24 (ℎ)−0.19 + 10550(sin(ℎ))2.1 + 300(sin(ℎ))3 (33)

𝐺/𝐺0 = 0.713(1/𝑆) (34)

where ℎ is the solar elevation at solar noon on the 15th day of the month, 𝑠 is the monthly
average daily bright sunshine hour (hour) and 𝑆 is the monthly average daily relative sunshine.
The empirical coefficients and statistical errors for Cairo city are computed and reported
in Table. 9. As shown in the table, the obtained results are clearly consistent with the previous
results for New Borg El-Arab, El-Arich and Sidi Barrani which illustrate that the new proposed
Model 6 provide the best estimations and Models 8-12 are rejected. As well, the new presented
models perform better than the three selected old models (Allen, Goodin et al. and Annandale et
al.) except Model 17 and 7. Table 10 denotes the comparison between the results of the local and
general formula of Model 6 and the results of the local Robaa model and the general El-Metwally
Model. The obtained results show that the new suggested Model 6 (local model) provides better
performance than the performance of Robaa model, with excellent values of statistical indicators
where R2 value is higher than 0.99 and very close to the unity (desired value). Also, the other
errors values are very close to zero (desired value). Similarly, the general formula of new suggested
Model 6 introduced the better performance than El-Metwally model with R2 value of
approximately 0.96.

Table 11 shows the values of relative percentage error for all month which are calculated
for the local and general models. The results show that the maximum value for relative error of

22
local Model 6 does not exceed 5.7% while the maximum value for Robaa local model exceeds the
acceptable value (±10%) for November, December and January with values of -17.8%, -19.4%
and -19.6%, respectively . Also, the relative errors for general formula of Model 6 (All Egypt)
have acceptable values except for June and July with slightly high values of 10.5% and 10.8%
respectively. On the other hand, relative error for El-Metwally general model exceeds the
acceptable value for December and February with values of 11.4% and 11.2% respectively. Fig. 6
shows the overall performance for local and general formula of the new suggested Model 6
compared with the Robaa local model and El-Metwally general model against the measured data
at Cairo city.

Table 9
Empirical coefficients and statistical errors for new models (1 -17) and selected models (18-20) at Cairo city.
Model a b c d e f MRSE MِAPE MABE R2 Rank
1 0.44961 0.00731 0.7330 2.7698 0.5367 0.983681 4
2 0.41739 0.01050 -0.00007 0.7297 2.9175 0.5524 0.983827 3
3 0.63366 -0.02218 0.00151 -0.00002 0.7060 2.7451 0.5308 0.984860 2
4 18.42449 -4.26725 -0.07341 0.7345 3.1029 0.5751 0.983614 6
5 -1349665.08 -7.19093 0.64337 1.1417 5.0404 0.9785 0.960409 10
6 0.00030 0.79028 0.49159 0.5813 2.4064 0.4600 0.989737 1
7 0.05319 -0.35540 2.2555 11.9449 1.9220 0.845471 15
8 -1152.38593 139.07555 -2.30873 3.6378 17.5160 3.1840 0.598020 18
9 -0.00086 0.00018 -8.37104E-06 3.86853 4.2767 20.2393 3.5373 0.444443 19
10 -0.71800 0.34952 -0.01494 -2.89789 3.5808 17.0831 3.0928 0.610533 17
11 -0.00068 0.00016 -7.74239E-06 3.85457 -0.21871 4.2802 20.2907 3.5326 0.443519 20
12 0.52334 -0.12089 0.00983 -0.00027 1.97087 -1.2953 3.4986 15.2548 2.9142 0.628198 16
13 -0.00208 0.00037 1.17055 0.49292 0.7349 3.0450 0.5955 0.983597 7
14 0.26447 0.00394 0.21938 0.7336 3.1701 0.5971 0.983651 5
15 0.28544 0.00092 6.72512E-05 0.23148 0.00121 0.7443 3.1374 0.5988 0.983172 9
16 0.28652 0.00092 6.73956E-05 0.23095 0.7443 3.1367 0.5987 0.983175 8
17 0.10501 0.00191 -8.36928E-05 2.159E-06 0.60301 1.4340 7.4477 1.2998 0.937538 14

18 0.17469 1.2639 5.2178 1.0853 0.951475 12


19 0.28693 0.30445 1.1438 4.9004 0.9822 0.960258 11
20 0.61836 8.43735 1.73960 1.2670 6.1632 1.0738 0.951240 13

Table 10
Statistical errors for new proposed Model 6 compared with Robaa model and El-Metwally model at Cairo.
Model Type Model t-test MPE MBE MRSE MِAPE MABE R2 Rank
Model 6 0.9313 0.4277 0.1302 0.4816 1.9832 0.3803 0.992956 1
Local Model
Robaa Model 4.0207 -6.9889 -1.0478 1.3583 7.3055 1.1248 0.943963 2

Model 6 (All Egy) 4.4668 4.4666 0.9557 1.1903 4.5302 0.9634 0.956961 1
General Model
El-Metwally (All Egy) 3.8198 6.5676 0.9947 1.3173 7.1703 1.1601 0.947287 2

23
30 30

Monthly average daily global solar radiation (MJ/m^2 day)


Monthly average daily global solar radiation (MJ/m^2 day)

Cairo local Model 6 Cairo general Model 6


27 27

24 24

21 21

18 18

15 15

12 12

9 9

6 6
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month Month
Measured GSRH Model 6 (Local) S.M. Robaa Model Measured GSRH Model 6 (All Egy) El-Metwally (All Egy)

Fig. 6 Overall performance for local and general formula of new suggested Model 6 comparing with Robaa
model [17] and El-Metwally model [22] against the measured data at Cairo city, Egypt.
Table 11
Relative error (𝑒) for new proposed Model 6 compared with Robaa model and
El-Metwally model at Cairo city.
Local Model General Model
Month
Model 6 (Local) S.M. Robaa Model Model 6 (All Egy) El-Metwally (All Egy)
Jan -3.8 -19.6 -0.9 8.5
Feb -0.5 -8.8 3.1 11.2
Mar -5.1 -7.9 -1.2 3.0
Apr -0.5 -0.9 4.1 2.0
May 1.4 -1.6 6.4 1.9
Jun 5.4 -0.8 10.5 3.2
Jul 5.7 0.7 10.8 3.5
Aug 4.9 1.2 9.6 5.3
Sep 2.4 -0.9 6.7 8.1
Oct -0.7 -8.4 2.9 9.0
Nov -3.8 -17.8 -0.8 11.4
Dec -1.0 -19.4 1.8 9.3

From the above discussion, it can be concluded that the new suggested models in this study
(Model 1-7 and 13-17) can be employed for estimating global solar radiation with a higher
accuracy. Consequently, the new suggested models are adequate for predicting monthly average
daily global solar radiation on a horizontal surface at any site in Egypt. On the other hand, these
models are site-dependent, therefore it is better to adjust their empirical coefficients against the
local data when they are utilized in other locations.
24
Moreover, using new suggested models, estimating solar radiation can easily be determined
with sufficient reliability at locations where there is no equipment for measuring solar radiation,
particularly in developing countries. Where ambient temperatures data are already measured very
easily and continuously for other goals, this considers the main advantage of the proposed
technique in the present work. In addition, the high applicability of the temperature-based solar
radiation models can be achieved by coupling these models with different short/long term weather
forecast techniques which are mainly used to accurately forecast the weather temperature. Using
these accurate future estimations for temperature as inputs to the temperature-based models in
order to predict the accurate future estimations for solar radiation can be considered as a valuable
tool for future design and performance analysis for different solar energy systems.

6. Conclusion
This study aims to present new solar radiation models based on ambient temperature as alternatives
to widely-used sunshine-based models since the ambient temperature is the most available and
simply measured metrological data at any location around the world. To achieve this purpose,
twenty different ambient temperature-based models are proposed for estimating monthly average
daily global solar radiation on a horizontal surface; 17 new models and three selected models from
the literature. The values of monthly average daily extraterrestrial solar radiation and solar
elevation are calculated using in-house computer program that is developed using C# language.
The models are established and validated using 20-years measured data of global solar radiation
for ten different studied locations. The generalization capability are examined and the local
formulae for the models are validated for two coastal locations where the general formulae give
inaccurate predictions. Moreover, the results of the most accurate new model are compared with
those for the most accurate two sunshine-based models from the literature. The study illustrates
that twelve new models perform better than the three selected models from the literature and the
most accurate results are predicted by the new model (Model 6). In addition, the general formula
of the most accurate new model (Model 6) provides the best estimations for different four cities
namely El-Arich, Hurghada, Kharga and Aswan with R2 values higher than 0.93 for all cities
except for Sidi Barrani city. For coastal sites such as Sidi Barrani city and El-Arich cities, where
the general formulas commonly give inaccurate predictions, the local formula of Model 6 gives
the best estimations with excellent R2 value of 0.98. In general, Model 6 shows the best
performance for either the general formula with acceptable R2 values or the local formula when
the empirical coefficients are adjusted against the local data for a certain location especially for
coastal sites. Moreover, the local and general formulas of Model 6 give better performances with
higher R2 values than the most accurate two sunshine-based models (Robaa local model and El-
Metwally general model) according to the measured data of the city of Cairo. Therefore, the new
temperature-based models, especially Model 6, can be considered as alternatives to the sunshine-
based models for predicting global solar radiation on a horizontal surface at different locations.
According to the results, the suggested new temperature-based models are significant and
applicable for quick and accurate estimation of monthly average daily global solar radiation on a
25
horizontal surface. The models and the computer code can be considered as the backbone of any
computer-program for designing different solar energy systems. Moreover, the high applicability
of the temperature-based solar radiation models can be achieved by coupling these models with
different weather temperature forecast techniques in order to perform future design and
performance analysis for different solar energy systems. Furthermore, the presented models should
be further examined and applied at different locations around the world with different climate
conditions to extend the global applicability of these new models.

Acknowledgements
The authors would like to thank the editor, assistant editors and the reviewers for
their valuable inputs, comments and suggestions, which improve the quality of the paper
in a very positive way to.
References
[1] Şenkal O, Kuleli T. Estimation of solar radiation over Turkey using artificial neural network and
satellite data. Appl Energy 2009;86:1222–8.
[2] Janjai S, Pankaew P, Laksanaboonsong J. A model for calculating hourly global solar radiation
from satellite data in the tropics. Appl Energy 2009;86:1450–7.
[3] Wong LT, Chow WK. Solar radiation model. Appl Energy 2001;69:191–224.
[4] El-Sebaii a. a., Al-Hazmi FS, Al-Ghamdi a. a., Yaghmour SJ. Global, direct and diffuse solar
radiation on horizontal and tilted surfaces in Jeddah, Saudi Arabia. Appl Energy 2010;87:568–76.
[5] Dorvlo ASS, Jervase JA, Al-Lawati A. Solar radiation estimation using artificial neural networks.
Appl Energy 2002;71:307–19.
[6] Amrouche B, Le Pivert X. Artificial neural network based daily local forecasting for global solar
radiation. Appl Energy 2014;130:333–41.
[7] Li H, Ma W, Lian Y, Wang X. Estimating daily global solar radiation by day of year in China.
Appl Energy 2010;87:3011–7.
[8] Fadare DA. Modelling of solar energy potential in Nigeria using an artificial neural network
model. Appl Energy 2009;86:1410–22.
[9] Voyant C, Darras C, Muselli M, Paoli C, Nivet ML, Poggi P. Bayesian rules and stochastic models
for high accuracy prediction of solar radiation. Appl Energy 2014;114:218–26.
[10] Besharat F, Dehghan A a., Faghih AR. Empirical models for estimating global solar radiation: A
review and case study. Renew Sustain Energy Rev 2013;21:798–821.
[11] Aͦngström A. Solar and terrestrial radiation. Q J R Meteorol Soc 1924;50:121–5.
[12] Prescott JA. Evaporation from water surface in relation to solar radiation. Trans R Soc Aust
1940;46:114–8.
[13] Almorox J, Benito M, Hontoria C. Estimation of monthly Angstrom-Prescott equation coefficients
from measured daily data in Toledo, Spain. Renew Energy 2005;30:931–6.
[14] Despotovic M, Nedic V, Despotovic D, Cvetanovic S. Review and statistical analysis of different
global solar radiation sunshine models. Renew Sustain Energy Rev 2015;52:1869–80.
[15] Al-Mostafa ZA, Maghrabi AH, Al-Shehri SM. Sunshine-based global radiation models: A review
and case study. Energy Convers Manag 2014;84:209–16.
[16] Mecibah MS, Boukelia TE, Tahtah R, Gairaa K. Introducing the best model for estimation the
monthly mean daily global solar radiation on a horizontal surface (Case study: Algeria). Renew
Sustain Energy Rev 2014;36:194–202.
[17] Robaa SM. Validation of the existing models for estimating global solar radiation over Egypt.
Energy Convers Manag 2009;50:184–93.
[18] Barbaro S, Coppolino S, Leone C SE. Global solar radiation in Italy. Sol Energy 1978;20:431–5.

26
[19] Khorasanizadeh H, Mohammadi K. Introducing the best model for predicting the monthly mean
global solar radiation over six major cities of Iran. Energy 2013;51:257–66.
[20] Khalil S a., Shaffie a. M. A comparative study of total, direct and diffuse solar irradiance by using
different models on horizontal and inclined surfaces for Cairo, Egypt. Renew Sustain Energy Rev
2013;27:853–63.
[21] Ajayi OO, Ohijeagbon OD, Nwadialo CE, Olasope O. New model to estimate daily global solar
radiation over Nigeria. Sustain Energy Technol Assessments 2014;5:28–36.
[22] El-Metwally M. Sunshine and global solar radiation estimation at different sites in Egypt. J Atmos
Solar-Terrestrial Phys 2005;67:1331–42.
[23] Hargreaves GH, Samani ZA. Estimating Potential Evapotranspiration. J Irrig Drain Div
1982;108:225–30.
[24] Annandale J., Jovanovic N., Benadé N, Allen R. Software for missing data error analysis of
Penman-Monteith reference evapotranspiration. Irrig Sci 2002;21:57–67.
[25] Allen RG. Self-Calibrating Method for Estimating Solar Radiation from Air Temperature. J
Hydrol Eng 1997;2:56–67.
[26] Bristow KL, Campbell GS. On the relationship between incoming solar radiation and daily
maximum and minimum temperature. Agric For Meteorol 1984;31:159–66.
[27] Goodin D, Hutchinson J. Estimating solar irradiance for crop modeling using daily air temperature
data. AGROCLIMATOLOGY 1999;91:845–51.
[28] El-Metwally M. Simple new methods to estimate global solar radiation based on meteorological
data in Egypt. Atmos Res 2004;69:217–39.
[29] Quej VH, Almorox J, Ibrakhimov M, Saito L. Empirical models for estimating daily global solar
radiation in Yucatán Peninsula, Mexico. Energy Convers Manag 2016;110:448–56.
[30] El Mghouchi Y, El Bouardi A, Sadouk A, Fellak I, Ajzoul T. Comparison of three solar radiation
models and their validation under all sky conditions – case study: Tetuan city in northern of
Morocco. Renew Sustain Energy Rev 2016;58:1432–44.
[31] Pan T, Wu S, Dai E, Liu Y. Estimating the daily global solar radiation spatial distribution from
diurnal temperature ranges over the tibetan plateau in China. Appl Energy 2013;107:384–93.
[32] Almorox J, Hontoria C, Benito M. Models for obtaining daily global solar radiation with measured
air temperature data in Madrid (Spain). Appl Energy 2011;88:1703–9.
[33] Badescu V. Correlations to estimate monthly mean daily solar global irradiation: Application to
Romania. Energy 1999;24:883–93.
[34] Chen R, Ersi K, Yang J, Lu S, Zhao W. Validation of five global radiation models with measured
daily data in China. Energy Convers Manag 2004;45:1759–69.
[35] Li H, Cao F, Wang X, Ma W. A Temperature-Based Model for Estimating Monthly Average
Daily Global Solar Radiation in China. Sci World J 2014.
[36] Yacef R, Mellit A, Belaid S, Şen Z. New combined models for estimating daily global solar
radiation from measured air temperature in semi-arid climates: Application in Ghardaïa, Algeria.
Energy Convers Manag 2014;79:606–15.
[37] Kumar R, Aggarwal RK, Sharma JD. Comparison of regression and artificial neural network
models for estimation of global solar radiations. Renew Sustain Energy Rev 2015;52:1294–9.
[38] Robaa S. Evaluation of sunshine duration from cloud data in Egypt. Energy 2008;33:785–95.
[39] I B R A H I M DINCER, SUKRAN DILMAC IET and ME. A simple technique for estimating
solar radiation parameters and its application for Gebze. Energy Convers Manag 1996;37:183–98.
[40] Seco J, Calvo A, Egido M, Egido A. SOLAR RADIATION AND AIR TEMPERATURE : A
STUDY OF SANTIAGO DE CHILE AND SALAMANCA. Atmos Environ 1993;27A:457–62.
[41] Atkins N. Daily Temperature Variations - Diurnal Variation of Solar Radiation. Yndon State
Collage Surv Meterology n.d.
http://apollo.lsc.vsc.edu/classes/met130/notes/chapter3/daily_trend3.html (accessed May 25,
2016).

27
[42] Sözen A, Arcaklioǧlu E. Solar potential in Turkey. Appl Energy 2005;80:35–45.
[43] Khorasanizadeh H, Mohammadi K. Prediction of daily global solar radiation by day of the year in
four cities located in the sunny regions of Iran. Energy Convers Manag 2013;76:385–92.
[44] NASA. NASA Surface meteorology and Solar Energy n.d. https://eosweb.larc.nasa.gov/cgi-
bin/sse/daily.cgi (accessed April 10, 2015).
[45] Tadros MTY, Mustafa MAM. Estimation of the Global Horizontal Solar Radiation in Iraq. Int J
Emerg Technol Adv Eng 2014;4:587–605.
[46] Microsoft. Microsoft C# Language n.d. https://msdn.microsoft.com/en-
us/library/aa289180(v=vs.71).aspx (accessed January 1, 2015).
[47] Vahid Garousi, Ahmet Coşkunçay, Aysu Betin-Can OD. A survey of software engineering
practices in Turkey. J Syst Softw 2015;108:148–77.
[48] Sommerville I. Software Engineering 8. 8th ed. Boston, MA, USA: Addison-Wesley Longman
Publishing Co; 2007.
[49] Ayodele TR, Ogunjuyigbe ASO. Prediction of monthly average global solar radiation based on
statistical distribution of clearness index. Energy 2015;90:1733–42.
[50] Jiang Y. Estimation of monthly mean daily diffuse radiation in China. Appl Energy
2009;86:1458–64.
[51] Khalil S a., Shaffie AM. A comparative study of total, direct and diffuse solar irradiance by using
different models on horizontal and inclined surfaces for Cairo, Egypt. Renew Sustain Energy Rev
2013;27:853–63.
[52] Camacho E, Berenguel M, Rubio F, Martínez D. Chapter 1-Solar Energy Fundamentals. Control
Sol. energy Syst., London: Springer London; 2012, p. 1–23.
[53] Karakoti I, Das PK, Singh SK. Predicting monthly mean daily diffuse radiation for India. Appl
Energy 2012;91:412–25.
[54] Despotovic M, Nedic V, Despotovic D, Cvetanovic S. Evaluation of empirical models for
predicting monthly mean horizontal diffuse solar radiation. Renew Sustain Energy Rev
2016;56:246–60.
[55] WRDC. World Radiation Data Center n.d. http://wrdc.mgo.rssi.ru/ (accessed April 10, 2015).
[56] Winslow JC, Hunt ER, Piper SC. A globally applicable model of daily solar irradiance estimated
from air temperature and precipitation data. Ecol Modell 2001;143:227–43.
[57] Tadros MTY. Uses of sunshine duration to estimate the global solar radiation over eight
meteorological stations in Egypt. Renew Energy 2000;21:231–46.
[58] rp5. Reliable Prognosis n.d. http://rp5.ru/Weather_in_the_world (accessed April 10, 2015).

28

You might also like