Renewable Energy 33 (2008) 14061412

Technical Note
Evaluation of 12 models to estimate hourly diffuse irradiation on
inclined surfaces
Ali Mohammad Noorian
a
, Isaac Moradi
b,
, Gholam Ali Kamali
a
a
Atmospheric Science and Meteorological Research Center (ASMERC), IRIMO, Tehran, Iran
b
Department of Physical Geography, Faculty of Geography, University of Tehran, Tehran, Iran
Received 26 October 2006; accepted 30 June 2007
Available online 7 September 2007
Abstract
This study evaluates the performance of 12 models to estimate hourly diffuse solar irradiation on inclined surfaces from those
measured on horizontal surfaces. Total solar irradiation incident on a tilted surface consists of three components including: beam, diffuse
and reected from the ground. On a semi-hourly basis, the beam component can be calculated by the ratio of the incidence angle to the
solar zenith angle. The reected component has a small effect on calculations and may be calculated with an isotropic model. In contrast,
models for estimating the diffuse component show major differences, which justify the validation study that this paper discusses. Twelve
models were tested against recorded south- and west-facing slope irradiances at Karaj (35155
0
N; 50156
0
E), Iran. The following models
were included: Badescu [Ba], Tian et al. [Ti], Perez et al. [P9], Reindl et al. [Re], Koronakis [Kr], Perez et al. [P8], Skartveit and Olseth
[SO], Steven and Unsworth [SU], Hay [Ha], Klucher [Kl], Temps and Coulson [TC], and Liu and Jordan [LJ].
The relative root mean square error (RMSE), for the south-facing surface ranges from 10.16% to 54.89% for the SO and TC models,
respectively. For the west-facing surface, RMSE ranges from 30.71% for the P9 model to 63.53% for the TC model. Statistical indices
show that all models produce large errors for the west-facing surface. Statistical indices for the south-facing surface show reasonably
good agreement with measured data.
r 2007 Elsevier Ltd. All rights reserved.
Keywords: Diffuse irradiation; Isotropic models; Anisotropic models; Radiation modeling; Iran
1. Introduction
Solar energy is a sustainable, safe and abundant energy
resource. Estimating solar irradiation incident on inclined
surfaces of various orientations is necessary in order to
calculate the building heat gain from the building envelope
as well as the electric power generated by photovoltaics
(PV) [1], design solar systems and to evaluate their long-
term average performance. Despite the fact that many
meteorological/radiometric stations measure global and
diffuse irradiation received on horizontal surfaces the data
on inclined surfaces are not available and are also
estimated with different models from those measured on
horizontal surfaces [2].
Total radiation incident on a tilted plane consists of
three components: beam radiation, diffuse radiation and
reected radiation from the ground. On an hourly basis the
direct and reected components can be computed with
good accuracy by using simple algorithms but the nature of
the diffuse component is more complicated and the desired
algorithms need assessment and evaluation.
Perez et al. [3,4] developed two new and more accurate
but considerably simpler versions of the original Perez
diffuse irradiance model [5]. The original version has been
used world wide to estimate short time step (hourly or less)
irradiance on tilted surfaces based on global and direct (or
diffuse) irradiance measured on horizontal surfaces. Li and
Lam [6] evaluated the anisotropic models of Klucher [7],
Hay [8] and Perez et al. [3] applied to two-years of
measured data in Hong Kong (19961997). All three
models produced large errors for north-facing surfaces.
Predictions for south-facing surfaces showed reasonably
ARTICLE IN PRESS
www.elsevier.com/locate/renene
0960-1481/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.renene.2007.06.027

Corresponding author. Tel.: +98 21 44580661; fax: +98 21 44580670.

E-mail addresses: [email protected] (A.M. Noorian), isaac_moradi@
yahoo.com (I. Moradi), [email protected] (G.A. Kamali).
good agreement with measured data. Kamali et al. [9]
evaluated eight diffuse models to estimate solar irradiation
on tilted surfaces using daily measured solar irradiation
data in Iran and recommended the Reindl et al. [10] model
for estimating solar irradiation on tilted surfaces. Notton
et al. [11] evaluated the combination of some well-known
models to estimate the hourly global solar irradiation on a
tilted surface from those on a horizontal surface. They
recommend the Klucher [7] model separately or in
combination with other models to estimate diffuse solar
irradiation on surfaces tilted towards the equator.
The objective of this study is to compare the results of 12
widely used models, for estimating total solar radiation on
tilted surfaces, with measurements at Karaj, Iran for west-
and south-facing surfaces inclined at 40

and 45

,
respectively. The models used the same method for
calculating beam and ground reected radiation on the
tilted surface. The only difference was in the treatment of
diffuse radiation.
2. Data
The measuring station was located on the roof-top of the
Faculty of Agriculture, University of Tehran, Karaj
35

55
0
N; 50

56
0
E), Iran. The measurements have been
carried out using two CM5 pyranometers with solarimeter
integrator CC2 manufactured by Kipp and Zonen, Hol-
land. The solar radiation data measured by the authors,
during the period from June to October 2002 were used in
this study. This consisted of semi-hourly global solar
radiation on a horizontal and two tilted surfaces (45

south-facing and 40

west-facing). Also for testing the

diffuse model, semi-hourly global and diffuse radiation on
a horizontal surface was measured by the previously
mentioned radiometers from February to June 2006. This
represents the only measured data available for tilted
surfaces in Iran [12].
3. Theory and models
The hourly total irradiation incident on a tilted surface
G
Th
is composed of direct B
Td
, ground reected R
Th

and sky-diffuse D
Th
components:
G
Th
B
Th
D
Th
R
Th
, (1)
The amount of direct radiation on a surface tilted S
degrees from the horizontal and rotated a
T
degrees from
the northsouth axis can be calculated by multiplying the
direct horizontal irradiation by the ratio of cosy= cosZ,
where y is solar incidence angle on a tilted plane and Z is
solar zenith angle. Also, the measuring station was located
on a roof-top with very low reectance, and the reected
component was very much lower than the direct and the
diffuse components so an isotropic model can be used to
compute the reected component on the tilted surface. So,
Eq. (1) can be written again as follows:
G
Th
B
h
: cosy= cosZ r
d
D
h
G
h
:r:1 cosS=2,
(2)
where B
h
, D
h
and G
h
are hourly direct, diffuse and total
solar radiation on a horizontal surface, either measured
directly or estimated from each other, r
d
is the ratio of the
hourly diffuse irradiation incident on a tilted surface to
that on a horizontal surface, r is ground reectivity, and Z
and y are calculated by well-known formulae:
cosZ sinf sind cosf cosd coso,
cosy sinS sinZ cosa
S
a
T
cosS cosZ, (3)
where f is the latitude of the location, d is the solar
declination, and o is the solar hour angle [13]. Calculation
of the diffuse ratio is more complicated and many
ARTICLE IN PRESS
Nomenclature
B
h
hourly direct solar radiation on a horizontal
surface MJ m
2

B
Th
hourly direct solar radiation on a tilted surface
MJ m
2

D
h
hourly diffuse solar radiation on a horizontal
surface MJ m
2

D
Th
hourly diffuse solar radiation on a tilted surface
MJ m
2

G
h
hourly total solar radiation on a horizontal
surface MJ m
2

G
oh
hourly extraterrestrial solar radiation on a
horizontal surface MJ m
2

G
Th
hourly total solar radiation on a tilted surface
MJ m
2

k
T
atmospheric transmissivity and equals the ratio
of G
h
=G
oh
(dimensionless)
r
b
beam ratio factor, the ratio of cosy= cosZ
(dimensionless)
r
d
ratio of the hourly diffuse solar radiation
incident on a tilted surface to that on a
horizontal surface (dimensionless)
R
Th
hourly reected solar radiation on a tilted
surface MJ m
2

S tilted plane slope angle (deg.)

Z solar zenith angle (deg.)
a
S
solar azimuth angle (deg.)
a
T
tilted plane azimuth angle (deg.)
d solar de clination (deg.)
f latitude of the site and positive for the northern
hemisphere (deg.)
r ground reectivity (dimensionless)
y solar incidence angle on tilted surface (deg.)
o solar hour angle and clockwise from south is
positive (deg.)
A.M. Noorian et al. / Renewable Energy 33 (2008) 14061412 1407
researchers use a simple isotropic model to estimate the
diffuse radiation of tilted surfaces [14]. But we should
consider the anisotropic nature of this component for the
evaluation and sizing of the solar systems. There have been
several algorithms developed that convert the horizontal
diffuse irradiation to the diffuse irradiation on a tilted
surface. We used 12 of these algorithms to compute diffuse
irradiation on tilted surfaces towards the equator (south)
and west.
According to Eq. (2), we need the direct and diffuse
components of global radiation for estimating global solar
radiation on tilted surfaces. For this study, these compo-
nents were not available separately, so we used the Miguel
et al. [15] model to estimate the daily direct and diffuse
components from measured daily global irradiation
(Appendix A, Eqs. (A.6) and (A.7)).
Two statistical tests, root mean square (RMSE) and man
bias error (MBE), were used to compare the models
(Appendix A, Eqs. (A.8)(A.11).
3.1. Diffuse radiation models
The methods used to estimate the ratio of diffuse solar
radiation on a tilted surface to that of a horizontal are
classied as isotropic and anisotropic models. The isotropic
models assume that the intensity of diffuse sky radiation is
uniform over the sky dome. Hence, the diffuse radiation
incident on a tilted surface depends on the fraction of the
sky dome seen by it. The anisotropic models assume the
anisotropy of the diffuse sky radiation in the circumsolar
region (sky near the solar disk) plus and isotropically
distributed diffuse component from the rest of the sky
dome.
The diffuse models chosen for study were the isotropic
models of Badescu (Ba) [16], Tian et al. (Ti) [17],
Koronakis (Kr) [18] and Liu and Jordan (LJ) [19], and
the anisotropic models of Perez et al. (P9) [3], Reindl et al.
(Re) [10], Perez et al. (P8) [4], Skartveit and Olseth (SO)
[20], Steven and Unsworth (SU) [21], Hay (Ha) [8], Klucher
(Kl) [7] and Temps and Coulson (TC) [22]. For more
information about models and mathematical relationships,
the reader is referred to Table 1.
4. Results
The evaluation was carried out on a semi-hourly basis.
The global component on the tilted surfaces was deter-
mined from measured horizontal data using 12 models and
compared with the measured tilted data of the same period.
The ground reectance was determined to be 0.05. The
accuracy of estimating the diffuse component of solar
irradiation from global one [15] was performed using an
independent data set, extending from February to June
2002. The RMSE and MBE were only 10.48% and 2.51%,
respectively, implying overestimation of diffuse radiation
by the model. The RMSE shows that the model can be
used to separate global solar radiation into direct and
diffuse components with good accuracy.
Table 2 shows the statistical results of the models for the
south-facing surface. It can be observed from Table 2 that
the relative root mean square error (RMSE), for the south-
facing surface ranges from 10.16% to 54.89% for SO
model and TC model, respectively. For west-facing
surface (Table 3), RMSE range from 30.71% for P9
model to 63.53% for TC model. The absolute relative mean
bias error (MBE) for south-facing surface ranges from
1.00% (Ba and Ti models) to 31.86% (TC model). For
ARTICLE IN PRESS
Table 1
Details and mathematical relationships of diffuse models selected for current study
Model Year Abb. Type
a
r
d
D
Th
=D
h

Badescu 2002 Ba ISO 3 cos2S=4
Tian et al. 2001 Ti ISO 1 S=180
Perez et al. 1990 P9 ANI 0:51 F
0
1
1 cosS F
0
1
a=b F
0
2
sinS
Reindl et al. 1990 Re ANI B
h
G
oh
r
b
1
B
h
G
oh
_ _
1 cosS=21

B
h
=G
h
_
: sin
3
S=2
Perez et al. 1986 P8 ANI 0:51 cosS1 F
0
1
F
0
2
F
0
1
a=c F
0
2
b=d
Koronakis 1986 Kr ISO 1=32 cosS
Skartveit and Olseth 1986 SO ANI B
h
G
oh
r
b
OcosS 1
B
h
G
oh
O
_ _
1 cosS=2 where O max 0; 0:3 2
B
h
G
oh
_ _ _ _ _ _
Stevenand Unsworth 1980 SU ANI
0:51r
b
f1 cosS=2g
1:74
1:26p
sinS S
p
180
_ _
cosS psin
2
S
2
_ _ _ _
Hay 1979 Ha ANI B
h
G
oh
B
h
1
B
h
G
oh
_ _
1 cosS=2
Temps and Coulson 1977 TC ANI
cos
2
S=21 sin
3
S=21 cos
2
ysin
3
Z
Liu and Jordan 1962 LJ ISO 1 cosS=2
Variables are dened in nomenclature section.
a
ISO means isotropic and ANI means anisotropic.
A.M. Noorian et al. / Renewable Energy 33 (2008) 14061412 1408
the west-facing surface the absolute MBE ranges from
2.43% (Kl model) to 38.52% (TC model). For all
the models studied covering south-facing surfaces, except
for SU (47.18%) and TC (54.89%), the RMSE values
do not differ a great deal. The SO, Ha, Re and P9 models
give the most accurate predictions for the south-facing
surface, and P9 model performs better for the west-facing
surface. The MBE shows that all models, with the
exception of SU model, underestimate values for west-
facing surfaces.
Graphical residual analysis of each model (residuals are
estimates of experimental error obtained by subtracting the
observed irradiation from the estimated one) for the south-
facing-surface is reported in Fig. 1 which shows that the
values of residuals for the SO, Ha, Re, P9 and P8 models,
exhibit generally small differences and are very close to
zero. This suggests that these models estimate of irradia-
tion compare well with measured data. Residual analysis
for the west-facing surface (Fig. 2) shows that only the
values of residuals for P9 model show relatively small
values with a distribution near to normal distribution. The
residual analysis for both surfaces conrms the result of the
statistical indices, RMSE and MBE.
For the west-facing surface, although the correlation
coefcients between the measured and estimated irradia-
tion are signicant at 0.05 level (Table 3) the frequency of
the residuals (Fig. 2) shows that the high correlation is due
to the scattering of residuals around zero and high positive
and negative values of the residuals neutralize the effect of
each other. Statistical indices show that all models produce
large errors for the west-facing surface. The statistical and
graphical results for the south-facing surface show reason-
ably good agreement with measured data.
5. Conclusions
An evaluation of the predicted tilted solar irradiance
based on 12 inclined surface models and measured
horizontal solar data in Karaj, Iran has been carried out.
It has been observed that relatively high MBE and RMSE
values are found for the west-facing surfaces, which receive
much less direct radiation than south-facing surfaces; and
the tilted components predicted for south-facing surface
are more accurate. The Skartveit and Olseth [20], Hay [8],
Reindl et al. [10] and Perez et al. [3] models give the most
accurate predictions for the south-facing surface and the
Perez et al. [3] model performs best for the west-facing
surface. The RMSE values for the south-facing surface
range from 10.16% to 54.89% and for west-facing surface
range from 30.71% to 63.53%. In general, the Perez et al.
[3] model shows the best agreement with the measured
tilted data.
Acknowledgments
The authors would like to give their sincere thanks to
Dr. Helfried Scheinger (Vienna) for his helpful advices
and recommendations, Dr. Peter Mayes (NJDEP) for
improving the English text, and anonymous referees for
carefully reading the manuscript and giving a number of
useful suggestions for improvement.
Appendix A
A.1. Computing the semi-hourly total extraterrestrial solar
irradiation, G
oh
For a given time, G
oh
in unit of MJ m
2
is given by [23]:
G
oh

37:595
d
2
cos fcos dsin o
2
sin o
1

p
180
o
2
o
1
sin fsin d
_ _
,
(A.1)
where d is the earthsun distance in astronomical units, f is
latitude, d is solar declination, o
1
is the hour angle for the
beginning time, o
2
is the hour angle for the ending time
and all angles are given in degrees. The solar constant is
taken as 1367 Wm
2
[24,25].
ARTICLE IN PRESS
Table 2
Comparison of regression (model a b measured) and correlation
(R) coefcients, root mean square (RMSE) and mean bias errors (MBE) of
different hourly diffuse models for south-facing surface
Model a b R RMSE MBE %RMSE %MBE
LJ 0.07 0.93 0.99 0.07 0.03 13.4 6.16
Kr 0.08 0.95 0.99 0.08 0.05 14.89 9.54
Ba 0.06 0.88 0.99 0.08 0.01 13.95 1.00
Ti 0.06 0.88 0.99 0.08 0.01 13.95 1.00
Ha 0.06 0.94 1.00 0.06 0.02 10.37 4.47
SO 0.06 0.94 1.00 0.06 0.02 10.16 4.27
Re 0.06 0.95 1.00 0.06 0.03 10.87 6.07
SU 0.09 1.23 0.99 0.26 0.22 47.18 40.62
P8 0.02 0.94 0.99 0.07 0.01 12.44 2.07
P9 0.04 0.94 0.99 0.06 0.01 11.17 1.44
Kl 0.08 0.96 0.99 0.08 0.06 15.43 10.40
TC 0.1 0.5 0.94 0.3 0.18 54.89 31.86
RMSE and MBE are in units of MJ m
2
.
Table 3
Same as Table 2 but for west-facing surface
Model a b R RMSE MBE %RMSE %MBE
LJ 0.12 0.72 0.89 0.24 0.04 42.49 7.70
Kr 0.12 0.74 0.89 0.24 0.03 42.10 5.20
Ba 0.1 0.69 0.89 0.25 0.08 44.17 13.45
Ti 0.1 0.68 0.89 0.26 0.08 44.57 14.46
Ha 0.08 0.76 0.92 0.21 0.06 37.18 10.04
SO 0.08 0.76 0.92 0.21 0.06 37.25 10.23
Re 0.09 0.76 0.92 0.21 0.05 36.85 8.95
SU 0.13 1.03 0.95 0.24 0.15 40.90 26.26
P8 0.02 0.76 0.95 0.24 0.16 41.80 28.23
P9 0.05 0.81 0.95 0.18 0.07 30.71 11.50
Kl 0.11 0.78 0.91 0.22 0.01 37.97 2.43
TC 0.07 0.5 0.89 0.37 0.22 63.53 38.52
A.M. Noorian et al. / Renewable Energy 33 (2008) 14061412 1409
d
180
p
0:006918 0:399912 cos G 0:070257 sin G
0:006758 cos 2G 0:000907 sin 2G
0:002697 cos 3G 0:00148 sin 3G, A:2
1
d
2
1:00011 0:034221 cos G 0:00128 sin G
0:000719 cos G 0:00077 sin 2G, A:3
G
360
365
d
n
1, (A.4)
o 18012 TST=12, (A.5)
where d
n
is day of the year, e.g. 1 Jan 1, 20 Feb 51, etc.
and TST is true solar time.
A.2. Computing hourly diffuse and direct components from
hourly global irradiance
A simple, physically based method proposed by Miguel
et al. [15] was used for estimating hourly diffuse and direct
components from hourly global irradiance. For three different
ranges of atmospheric transmissivity k
T
G
h
=G
oh
, the
resulting correlations are given by the following expression:
D
h
G
h

0:995 0:081k
T
if k
T
o0:21;
0:724 2:738k
T
8:32k
2
T
4:967k
3
T
if 0:21pk
T
p0:76;
0:180 if k
T
40:76:
_

_
(A.6)
Then B
h
can be calculated as follows:
B
h
G
h
D
h
. (A.7)
ARTICLE IN PRESS
Fig. 1. Percent frequency of the deviations of the daily global radiation on south-facing tilted surface, calculated with different models to the experimental
data.
A.M. Noorian et al. / Renewable Energy 33 (2008) 14061412 1410
A.3. Statistical methods of comparison
The statistical error tests are as follows [26]:
RMSE

SC
i
M
i

2
n

, (A.8)
MBE

C
i
M
i

n
, (A.9)
where C
i
and M
i
are the ith calculated and measured
values on a tilted surface, respectively, and n is the total
number of observations for a specic period of time.
Relative RMSE (%RMSE) and MBE (%MBE), which
are a dimensionless measure of RMSE and MBE, can be
dened as follows.
%RMSE
RMSE
M
100, (A.10)
%MBE
MBE
M
100, (A.11)
where M is the mean of measured values on the tilted
surface.
References
[1] Cheng CL, Chan CY, Chen CL. An empirical approach to estimating
monthly radiation on south-facing tilted planes for building applica-
tion. Energy 2006;31:294057.
[2] Burlon S, Bivona S, Leone C. Instantaneous hourly and daily
radiation on tilted surfaces. Sol Energy 1991;47:839.
[3] Perez R, Ineichen P, Seals R. Modelling daylight availability and
irradiance components from direct and global irradiance. Sol Energy
1990;44:27189.
[4] Perez R, Seals R, Ineichen P, Stewart P, Menicucci D. A new
simplied version of the Perez diffuse irradiance model for tilted
surfaces. Sol Energy 1987;39:2213.
ARTICLE IN PRESS
Fig. 2. Same as Fig. 1 but for west-facing surface.
A.M. Noorian et al. / Renewable Energy 33 (2008) 14061412 1411
[5] Perez R, Stewart R, Arbogast C, Seals R, Scott J. An anisotropic
hourly diffuse radiation model for sloping surfaces, description,
performance validation, site dependency evaluation. Sol Energy 1986;
36:48197.
[6] Li DHW, Lam JC. Evaluation of slope irradiance and illuminance models
against measured Hong Kong data. Build Environ 2000;35:5019.
[7] Klucher TM. Evaluation of models to predict insolation on tilted
surfaces. Sol Energy 1979;23:1114.
[8] Hay JE. Calculation of monthly mean solar radiation for horizontal
and inclined surfaces. Sol Energy 1979;23:30130.
[9] Kamali GA, Moradi I, Khalili A. Estimating solar radiation on tilted
surfaces with various orientations: a study case in Karaj (Iran). Theor
Appl Climatol 2006;84:23541.
[10] Reindl DT, Beckman WA, Dufe JA. Evaluation of hourly tilted
surface radiation models. Sol Energy 1990;45:917.
[11] Notton G, Poggi P, Cristofari C. Predicting hourly solar irradiations
on inclined surfaces based on the horizontal measurements:
performances of the association of well-known mathematical models.
Energy Convers Manage 2006;47:181629.
[12] Moradi I. Estimating solar radiation on tilted surfaces of various
orientations (Theoretical survey and experimental results in the
climatic conditions of Karaj). M.Sc. thesis, Faculty of Water and Soil
Engineering, University of Tehran, Karaj, Iran; 2003.
[13] Iqbal M. An introduction to solar radiation. New York: Academic
Press; 1983.
[14] Jimenz JI, Castro Y. Solar radiation on sloping surfaces with
different orientations in Granda, Spain. Sol Energy 1982;28:25762.
[15] Miguel A, Bilbao J, Aguiar R, Kambezidis H, Negro E. Diffuse solar
radiation model evaluation in the north Mediterranean belt area. Sol
Energy 2001;70:14353.
[16] Badescu V. 3D isotropic approximation for solar diffuse irradiance
on tilted surfaces. Renew Energy 2002;26:2213.
[17] Tian YQ, Davies-Colley RJ, Gong P, Thorrold BW. Estimating solar
radiation on slopes of arbitrary aspect. Agric For Meteorol 2001;
109:6774.
[18] Koronakis PS. On the choice of the angle of tilt for south facing
solar collectors in the Athens basin area. Sol Energy 1986;36:
21725.
[19] Liu BYH, Jordan RC. Daily insolation on surfaces tilted towards the
equator. Trans ASHRAE 1962;67:52641.
[20] Skartveit A, Olseth JA. Modelling slope irradiance at high latitudes.
Sol Energy 1986;36:33344.
[21] Steven MD, Unsworth MH. The angular distribution and intercep-
tion of diffuse solar radiation below overcast skies. Q J R Meteorol
Soc 1980;106:5761.
[22] Temps RC, Coulson KL. Solar radiation incident upon slopes of
different orientations. Sol Energy 1977;19:17984.
[23] Bird RE, Riordan CJ. Simple solar spectral model for direct and
diffuse irradiance on horizontal and tilted planes at the Earths
surface for cloudless atmospheres. J Climate Appl Meteorol 1986;25:
8797.
[24] Roderick ML. Estimating the diffuse component from daily and
monthly measurements of global radiation. Agric For Meteorol 1999;
95:16985.
[25] Roderick ML. Methods for calculating solar position and day length
including computer programs and subroutines. Division of Resource
Management, Technical Report #137, Western Australian Depart-
ment of Agriculture, Perth; 1993.
[26] Nijmeh S, Mamlook R. Testing of two models for computing global
solar radiation on tilted surfaces. Renew Energy 2000;20:7581.
ARTICLE IN PRESS
A.M. Noorian et al. / Renewable Energy 33 (2008) 14061412 1412