You are on page 1of 5

Solar Energy Vot, 31, No. 3, pp.

313-317, 1983
Prlnled in Great Britain,

0038~)92X/83/0903D-05503.00/0 1983Pergamon Press iAd

STATISTICAL COMPARISON OF MODELS FOR ESTIMATING SOLAR RADIATION ON INCLINED SURFACES

C. C. Y. MA and M. IQBAL Department of MechanicalEngineering,University of British Columbia, Vancouver, Canada


(Received 1 September 1982)

Abstract--The objectives of this study are to compare statistically three models for estimating solar radiation on inclined surfaces and to recommend one that is general and accurate. The isotropic model, and Klucher's and Hay's anisotropic models are chosen for discussion. The hourly and daily formulations of these models are briefly described. Results show that they all produce large errors at steep slopes. The errors are minimum during the summer months. The isotropic model under-estimated consistently throughout the year resulting in the worst performance. Hay's model also under-estimated constantly but by a smaller amount. Klucher's model overestimated in the summer and under-estimated in the winter, but can be considered just as general and accurate as Hay's model for estimating insolation on inclined surfaces.
INTRODUCTION

NATURE OF DATA

There are numerous meteorological stations that provide facilities for measuring insolation arriving on horizontal surfaces. However, measurement of solar radiation incident on inclined surfaces is much less common. Quantitative assessment of radiation incident on a tilted plane is very important to engineers designing solar energy collecting devices, to architects designing buildings, and to agronomists studying insolation on foliage and vegetation on mountain slopes. Over the years many authors have presented models which predict solar radiation incident on inclined surfaces. Some of these models apply to specific cases, some require special measurements and some are limited in their scope. The objectives of this report are to carry out a statistical comparison of three specific models and to recommend one that is general and is the most accurate for estimating solar radiation arriving on an inclined surface. The models chosen for discussion here are for arbitrary sky conditions and are supposed to be applicable anywhere in the world. These methods begin with measured or predicted hourly global and diffuse radiation received on a horizontal surface. These quantities are then transposed onto an inclined plane by mathematical procedure. The transformation of these models for application on a daily time span will be examined. The accuracy of the models are then compared on the basis of statistical error tests and the most accurate model is recommended. It is assumed in this study that the surface is inclined in such a way that it faces the sun, i.e. direct radiation is not striking the back of this plane. Furthermore, total radiation incident on the plane is assumed to have three components: (1) beam radiation, (2) sky-diffuse radiation, and (3) groundreflected radiation. 313

Radiation data recorded at Woodbridge, Ontario fiat. 43.8N) are used in the study. They are minute-by-minute data collected during the year 1977. They provide measured values of the following variables: global, beam and diffuse radiation received on a horizontal surface, and global radiation arriving on a south-facing surface inclined at 30, 60 and 90 to the horizontal. In order to determine the short-term performance of the models and reduce computing time while maintaining a large sample size for statistical error analysis, these data are integrated to obtain hourly values before any testing is carried out. The computation procedure requires ground albedo. It is obtained by taking the ratio of the measurements from two pyranometers mounted back to back, one facing the ground and the other facing the sky. The values of ground albedo used in this report are the ratio of the daily total of the reflected radiation to that of the downward horizontal global radiation.
THE MODELS

Total radiation arriving on an inclined surface has three components: beam radiation, sky-diffuse radiation and ground-reflected radiation. The models discussed here all share the same formulations for the beam and ground-reflected components. It is useful to write down the hourly and daily formulations of these components. The hourly and daily beam radiation received on an inclined surface can be expressed as

IB = (I - Id)(cos O/cosOz)
and
HB = ( H - H d ) R b

(l)

(2)

314

C. C. Y. MA and M. IQBM. face is I,


=

respectively, where Rb is the ratio of the daily radiation incident on an inclined plane to that on a horizontal plane in the absence of the earth's atmosphere. R,, can be expressed as
R 7,277 -

I,[(1
x

cos/3)/2] [1 + F sin3 (/3/2)1

[1 + F cos 2 O sin3 0=]

(8)

where F is the modulating function given by [(27r/360)oJ;] sin 6 sin (4' -/3) + cos 6 cos (4' -/3) sin ~o' [(27r1360)o~] sin 8 sin 4' + cos 8 cos 4' sin ~o~ (3) The beam radiation on a horizontal surface is written as the difference between the global and diffuse values in eqns (1) and (2) because they are more commonly measured using pyranometer. Assuming isotropic reflection, which occurs when the global radiation is composed of primarily diffuse radiation and/or when the ground cover is a perfectly diffuse reflector, and using a common albedo, the hourly and daily ground-reflected radiation can be written as
I , = Io(1 - cos/3)/2
F =

l - (IalI) 2.

(9)

In eqn (8), the incidence and zenith angles pertain to a particular hour and do not have a daily equivalent. Therefore, the daily sky-diffuse radiation on an inclined surface is expressed in terms of hourly summations as follows:
day

Hs = ~ Ia[(l + cos/3)f2] [1 + F sin3 (/3/2)1 x [1 + F cos 2 0 sin3 0:]. (10)

(4)

and H,
=

HO(I - cos/3)/2

(5)

respectively. The sky-diffuse component is what distinguishes the three models. Two different concepts are used when treating sky-diffuse radiation on an inclined surface. The concepts are discussed below. (i) Isotropic model In this model the intensity of sky-diffuse radiation is assumed uniform over the skydome, i.e. it is independent of the azimuth and zenith angles. It approximates the completely overcast sky condition. The formulae for the hourly and daily sky-diffuse radiation incident on an inclined plane are
Is = ld(1 + cos/3)/2

When the skies are completely overcast, F = 0, Flucher's model reverts to the isotropic model. (b) Hay's model. This model was developed by Hay[3] using a much different approach. The diffuse radiation on a horizontal surface is assumed to be composed of a circumsolar component coming directly from the direction of the sun and an isotropically distributed diffuse component from the rest of the skydome. These two components are weighted according to the isotropy of the sky. The hourly and daily sky-diffuse radiation on an inclined plane can be written as Is = Ia{[(l - Ia)llo] (cos 0/cos 0:) + [(l + cos/3)/21 [1 - (I -/a)/Io]} and
Hs = Ha{[(H - Ha)/Ho]Rb + [(1 + cos/3)/2]
(11)

x [1 - (H - Hd)/Ho]} (6)

(12)

and
Hs = H.(I + cos/3)I2

respectively. It is evident from the above that when the diffuse radiation approaches its global value Hay's model reduces to the isotropic model. (7)
METHODS OF COMPARISON

respectively. (ii) Anisotropic models This concept approximates partly-cloudy sky conditions. It may vary from clear skies on one extreme to completely cloudy skies on the other. Two of such models are presented below. (a) Klucher's model. This model is based on a study of clear sky conditions by Temps and Coulson[1]. Their model was modified by Klucher[2] who incorporated conditions of cloudy skies. It takes into account the increase in the diffuse radiation in the vicinity of the sun (the circumsolar radiation) and that near the horizon (horizon brightening). Klucher's formulation of the hourly sky-diffuse radiation incident on an inclined sur-

Two statistical tests, root mean square error (RMSE) and mean bias error (MBE), are used to evaluate the accuracy of the models. (i) Root Mean Square Error The root mean square error is defined as
1 l

) I/2

(13) where C~ is the ith calculated value, M~ is the ith measured value, and n is the total number of observations. Generally the lower the RMSE, the more accurate the model is. However, a few large errors in the sum can produce a significant increase in RMSE.

Statistical comparison of models for estimating solar radiation on inclined surfaces (ii) Mean Bias Error The mean bias error is defined as

315

MBE=[~_~(Ci-MI)]/n

(14)

This test provides information with respect to overestimation or under-estimation. A low MBE is desired. A positive value of the MBE gives the average amount of over-estimation in the calculated values and vice versa. One drawback of this test is that over-estimation of individual measurements will cancel under-estimation.
RESULTS AND COMPARISONS

With the integrated hourly radiation data, the RMSEs and MBEs of the models were computed for each month of 1977 at slopes of 30, 60 and 90o. They were calculated as a percentage of the measured mean for the time

period concerned. Preliminary results indicated that errors for the hour after sunrise and the hour before sunset were often large compared to other times of the day. Thus the models were tested again by neglecting these hours. The resulting RMSEs were plotted in Fig. 1 and the MBEs in Fig. 2. From Fig. 1, it is noted that the RMSEs for all three models increase as the slope increases. Regardless of the slope, the models agree quite well with each other during the summer months. They deviate from each other in the winter months when the effects of the difference in the diffuse radiation parameterization is maximum. For some unexplained reasons, all models produced the largest errors in November. The results of this month were discarded. The isotropic model resulted in the largest RMSEs. The highest RMSE of this model is close to 30 per cent. Klucher's model performed very

W O O D B R I DG E. ON TflR ] 0
SLOPE=20 BEG
__,,.-__~__ JSOTROPJC HAY KLUEHER

W~ ILl
U~ I~P,J

t
SLOPE=60

I
BEG

_ VOODBRIDGE.ONTRRIO
__~.._ __~__ ]SOTROPJC HAY KLUCHER

W~
~c~'

?\_

-\ ~ ~
-

/ /

~.x~k..
I I I I
]SO'fR~:'JC
HRY .....

/.
I I I I

/.
t

--~
-

"-

_ ~400DBRIDGE .ONTRR I0
SLOPE---90 D E G
~ ~

KLUCHER

/~X

/
LL1 fit_ ~ /

X_
'

Z---

"~" ~ .

JAN

FEB

MRR

I::IPR

MRY

dUN

JUL

AUG

SEP

OCT

NOV

DEE

Fig. 1. Root mean square errors for global radiation received on inclined surfaces.

316
I I I

C.C.Y. MA and M. IQBAL


I I I I I I I

WOODBRIDGE.ONTRRIO
SLOPE=30 DEGREES

_._
~

JSOTR~]C
H,qY KLUCHER

_ _ ~.__

P~

~t.n cc ! w

./

"/

I
SLOPE=riO

I
_~__

VOODBRIDGE.
DEGREES

ONTFIR I 0

]SOTRDP]C HflY KLUCHER

~t

.1

I
SLOPE--90

I
/ z

I
x

I
_._
~
__.__

I
HRY KLUCHER

VOODBRIDGE
D~GREES

.ONTARI~
~ . . . . ~q'x

]SOTRO~JE

CE (Z3 zu ~Z

I RPR

I MRY

l JUN

I JUL

I RUG

I 5EP

I DCT

I N~V DEC

b~N

FEB

MRR

Fig. 2. Mean bias errors for global radiation received on inclined surfaces. much like Hay's model especially at gentle slopes and the largest RMSE is about 20 per cent. With maximum RMSEs of less than 15 per cent for slopes of up to 60, Klucher's and Hay's models are both very good estimators of insolation on inclined planes. The MBE plots in Fig. 2 again show the direct relationship between error magnitude and surface slope. The isotropic model under-estimated almost consistently. Its performance was excellent during the summer months but is relatively poor for the rest of the year. Hay's model under-estimated consistently but only by no more than 7 per cent of the measured mean. The difference in its seasonal performance is very little. Klucher's model over-estimated in the summer and under-estimated in the winter, both by less than 5 per cent for slopes of up to 60. The magnitude of deviation increases as slope increases. At the 90 slope, the behaviour of these models became erratic and the plots are not as smooth as for gentler slopes. Klucher's and Hay's models are definitely better than the isotropic model for slopes of up to 60. Even though the plots indicate that there are differences between Klucher's and Hay's models, the small magnitude of error suggests that one is as good as the other for estimating insolation on inclined surfaces.
CONCLUSIONS

The RMSEs and MBEs of the models were plotted and compared for slopes of 30, 60 and 90. Generally the magnitude of both errors increases with the slope of the surface. The models produced very similar RMSEs during the summer months but Klucher's and Hay's models resulted in much lower RMSEs for the rest of the year. Excluding the month of November, the maximum RMSE

Statistical comparison of models for estimating solar radiation on inclined surfaces for Klucher's and Hay's models is less than 15 per cent for slopes of up to 60 . The M B E plots indicate that the isotropic model under-estimated almost consistently and by a larger amount than Hay's model. Hay's model also has a smaller difference in seasonal performance. Klucher's, model over-estimated in the summer and under-estimated in the winter, both by less than 5 per cent for slopes of up to 60 . Klucher's and Hay's models are definitely more superior than the isotropic model. For these two models, the small magnitude of the errors suggests that one works just as well as the other. They are both recommended for predicting solar radiation on inclined surfaces.

317

Acknowledgements--Financial support of the Natural Sciences and Engineering Research Council of Canada is gratefully acknowledged. Thanks are due to Dr. D. Wardle for supplying the Woodbridge data.
H Hd Ha HR Hs H0 NOMENCLATURE daily global radiation incident on a horizontal surface, MJm -2 day i daily diffuse radiation incident on a horizontal surface, MJm -2 day i daily beam radiation incident on an inclined surface, MJm 2 day -I daily ground-reflected radiation incident on an inclined surface, MJm-" day -t daily sky-diffuse radiation incident on an inclined surface, MJm -~ day -~ extraterrestrial daily radiation incident on a horizontal surface, MJm : day

I hourly global radiation incident on a horizontal surface, kJm-2 hr -I lb hourly beam radiation incident on a horizontal surface, kJm -2 hr-I Id hourly diffuse radiation incident on a horizontal surface, kJm-'- hr-i IB hourly beam radiation incident on an inclined surface, kJm 2hr-I IR hourly ground-reflected radiation incident on an inclined surface, kJm 2 hr-~ Is hourly sky-diffuse radiation incident on an inclined surface, kJm -2 hr t Io extraterrestrial hourl~ radiation incident on a horizontal surface, kJm -2 hrRb ratio of extraterrestrial daily radiation incident on an inclined surface to that on a horizontal surface /3 surface slope from the horizontal, degrees declination, degrees to~ sunrise hour angle, degrees co~ sunrise hour angle for a tilted surface, degrees tb latitude, degrees p ground albedo 0 incidence angle, the angle between beam radiation and the surface normal, degrees 0= zenith angle, the angle between beam radiation and the vertical, degrees REFERENCES 1. R. C. Temps and K. L. Coulson, Solar radiation incident upon slopes of different orientations. Solar Energy 19(2), 179-184 (1977). 2. T. M. Klucher, Evaluation of models to predict insolation on tilted surfaces. Solar Energy 23(2), 111-114 (1979). 3. J. E. Hay, Study of shortwave radiation on non-horizontal surfaces. Rep. 79-12, Atmospheric Environment Service, Downsview, Ontario (1979).

You might also like