You are on page 1of 4

Solar Energy 81 (2007) 692–695

www.elsevier.com/locate/solener

Short Note

Experimental testing of models for the estimation of hourly


solar radiation on vertical surfaces at Arcavacata di Rende
M. Cucumo, A. De Rosa, V. Ferraro, D. Kaliakatsos *, V. Marinelli
Department of Mechanical Engineering, University of Calabria, 87030 Arcavacata di Rende (CS), Italy

Received 14 June 2005; received in revised form 5 September 2006; accepted 6 September 2006
Available online 16 October 2006

Communicated by: Associate Editor Pierre Ineichen

Abstract

More than 55,000 data of hourly solar radiation on a horizontal surface and on vertical surfaces exposed to the south, west, north and
east, measured at Arcavacata di Rende (CS), were compared with hourly radiation data calculated by various calculation models.
Erbs, Reindl et al. and Skartveit et al. correlations for the split of hourly global radiation in the diffuse and beam components were
used together with the isotropic sky model and three anisotropic sky models.
The agreement between experimental and calculated data is generally good.
Ó 2006 Elsevier Ltd. All rights reserved.

Keywords: Solar radiation; Tilted surfaces; Calculation models validation

1. Introduction et al. (1982), the correlations developed by Reindl et al.


(1990a), denominated, in this paper, in order of increasing
The calculation of hourly solar radiation on tilted and complexity, Reindl1, Reindl2, Reindl3 and the new Skart-
vertical surfaces with various orientations is of great veit et al. (1998) correlation.
importance in many practical applications of solar energy, The most used model for the calculation of incident
such as the calculation of the energy use of buildings, pas- solar radiation on a surface inclined and orientated in
sive heating of buildings, the evaluation of photovoltaic any way is the isotropic model (Cucumo et al., 1994), which
plants and so on. assumes that the sky is an isotropic source of diffuse radi-
Usually hourly global radiation is measured on a hori- ation. In addition to the isotropic model, other models
zontal plane, and radiation on tilted or variably orientated have been developed called anisotropic models (Cucumo
vertical surfaces is calculated by means of composition et al., 1994), in which diffuse radiation is equal to the
models. Global radiation on a horizontal plane is decom- sum of two or three parts: one part called circumsolar, con-
posed, by correlations, into the diffuse and beam compo- sisting of diffuse radiation scattered ahead along the solar
nent and starting from these data the global radiation on rays, concentrated in a cone around the sun and striking
the surfaces involved is reassembled as the sum of beam the tilted surface with an angle of incidence equal to that
and diffuse solar radiation and that reflected from the of the beam radiation; an isotropic part evenly emitted
ground and the surroundings (Cucumo et al., 1994). from the rest of the sky, and a third part called the horizon
Among the best-known split correlations are those of Erbs brightness, more pronounced on clear days, coming from a
thin strip of sky along the horizon.
Among the anisotropic models, in this brief note three
*
Corresponding author. Tel.: +39 0984 494603; fax: +39 0984 494673. models have been considered: the Hay and Davis model
E-mail address: dimitri@unical.it (D. Kaliakatsos). (Hay and Davies, 1980) in which horizon brightness is

0038-092X/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.solener.2006.09.002
M. Cucumo et al. / Solar Energy 81 (2007) 692–695 693

Nomenclature

Dh diffuse solar radiation on a horizontal plane q reflection coefficient of the ground (–)
(W/m2) e mean percentage deviation (%)
Bh beam solar radiation on a horizontal plane RMS root mean square percentage deviation (%)
(W/m2)

not considered and diffuse radiation on a horizontal plane 39°18 0 N, longitude: 16°15 0 E) were compared with the pre-
is the sum of the isotropic and circumsolar component; the diction models.
HDKR model (Reindl et al., 1990b), developed by Reindl, First, the Erbs correlation was used for the division of
extending that of Hay and Davies taking into account hori- the hourly global radiation on a horizontal plane in the dif-
zon brightness on the basis of Klucher’s experimental fuse Dh and beam Bh components and the 0.2 value sug-
observations; and finally, the Perez et al. model (Perez gested by Liu and Jordan (Cucumo et al., 1994) for the
et al., 1986), which is based on a more detailed analysis reflection coefficient q of the ground, the buildings and
of the three components of diffuse radiation and which the surroundings of the testing station was used.
appears to be the most complete since it considers eight Table 1 shows the mean percentage deviations e and the
kinds of sky, ranging from completely clear to completely root mean square percentage deviations RMS between the
covered. hourly global radiation predictions on the variably orien-
tated vertical surfaces obtained with the four models and
2. Experimental testing of models the experimental values.
Examination of the results demonstrates that, for the
The above-mentioned calculation models were com- south direction, the best model is the isotropic, for east the
pared with experimental data taken at the Variable Orien- best model is Hay and Davies, for north the Perez and for
tation Testing Station at the University of Calabria, which west the Hay and Davies. However, notable prediction dif-
is equipped with five Kipp and Zonen pyranometers, one of ferences among the various models are not observed, with
them placed on a horizontal plane and the other four the exception of north direction. Fig. 1a shows the best
placed on the vertical surfaces orientated to the south, east, model for each orientation.
west and north. Since an experimental study carried out by Ineichen
More than 11,000 hourly global radiation data for each et al. in Geneva (Ineichen et al., 1987) pointed out the
orientation taken to Arcavacata di Rende (CS) (latitude: dependence of q on various expositions, with measured

Table 1
Mean percentage deviations e and root mean square percentage deviations RMS with q = 0.2 and Erbs split correlation
South West North East emed RMSmed
e RMS e RMS e RMS e RMS
Isotropic 0.55 19.2 8.96 26.72 15.58 27.69 12.09 30.49 9.02 26.02
Hay and Davies 1.33 19.28 2.63 22.24 4.35 23.95 3.01 25.85 1.32 22.83
HDKR 2.80 19.66 3.82 24.03 4.13 26.11 9.43 29.14 5.04 24.73
Perez 5.21 19.43 5.98 21.26 4.73 19.45 9.58 26.02 6.37 21.54

Fig. 1. Best models for the four orientations.


694 M. Cucumo et al. / Solar Energy 81 (2007) 692–695

Table 2
Mean percentage deviations e and root mean square percentage deviations RMS for q = 0.15 and Erbs split correlation
South West North East emed RMSmed
e RMS e RMS e RMS e RMS
Isotropic 5.72 18.83 0.22 22.40 4.44 22.03 4.11 25.78 0.76 22.26
Hay and Davies 6.50 20.27 11.37 25.39 15.6 29.20 4.97 26.62 9.61 25.37
HDKR 2.37 19.35 4.92 24.34 7.11 27.64 1.45 27.50 3.24 24.71
Perez 0.04 18.17 2.76 19.35 6.52 18.94 1.61 23.36 1.91 19.96

Table 3
Mean percentage deviations e and root mean square percentage deviations RMS for q = 0.15 and different split correlations
South West North East emed RMSmed
e RMS e RMS e RMS e RMS
Isotropic Erbs 5.72 18.83 0.22 22.40 4.34 22.03 4.11 25.78 0.74 22.26
Reindl1 5.19 18.64 1.81 22.76 6.61 21.86 5.16 26.02 2.10 22.32
Reindl2 4.89 18.70 4.09 24.36 10.50 24.77 7.42 27.46 4.28 23.82
Reindl3 5.18 19.00 4.59 24.75 9.74 26.00 6.21 26.13 3.84 23.97
Skartveit 7.90 18.57 0.21 22.38 5.25 23.29 2.63 26.66 0.05 22.73
Hay and Davies Erbs 6.50 20.27 11.37 25.39 15.60 29.20 4.97 26.62 9.61 25.37
Reindl1 5.96 19.71 10.51 23.65 14.60 26.57 4.48 25.14 8.89 23.77
Reindl2 5.83 19.78 8.88 23.10 11.74 25.59 2.79 24.76 7.31 23.31
Reindl3 6.03 20.21 8.30 23.16 12.19 26.64 3.73 24.80 7.56 23.70
Skartveit 8.08 20.23 10.17 25.39 14.03 28.55 4.61 25.74 9.22 24.98
HDKR Erbs 2.37 19.35 4.92 24.34 7.11 27.64 1.45 27.50 3.24 24.71
Reindl1 1.55 18.89 3.56 22.60 5.46 24.71 2.39 26.12 2.05 23.08
Reindl2 1.38 18.86 1.71 22.51 2.25 24.57 4.27 26.16 0.27 23.03
Reindl3 1.59 19.38 1.10 22.87 2.77 25.84 3.20 25.69 0.57 23.44
Skartveit 1.71 18.93 4.80 24.36 5.59 27.69 1.26 27.11 2.71 24.52
Perez Erbs 0.04 18.17 2.76 19.35 6.52 18.94 1.61 23.36 1.91 19.95
Reindl1 0.54 18.22 1.78 19.36 5.30 18.40 2.19 23.40 1.09 19.84
Reindl2 0.74 18.23 0.46 18.98 3.05 17.61 3.56 23.24 0.20 19.51
Reindl3 0.52 18.32 0.13 18.81 3.59 18.15 2.75 22.59 0.11 19.47
Skartveit 0.03 18.15 2.76 18.85 5.96 18.24 1.53 22.84 1.79 19.52

values of q close to 0.15, in an environmental context influence of the split correlations is not much important.
apparently similar to that of the Testing Station of the Uni- Fig. 1c summarizes these indications.
versity of Calabria is located, it was considered useful to
repeat the analysis with a value of q = 0.15. Table 2 shows 2.1. Conclusions
the mean percentage deviations and the root mean square
percentage deviations obtained with q = 0.15. For all ori- More than 55,000 hourly radiation data on vertical sur-
entations, the best model appears to be that of Perez. faces orientated to the south, west, north and east were
Since the use of q = 0.15 instead of q = 0.20 produces compared with various isotropic and anisotropic calcula-
a general improvement of predictions, for almost all mod- tion models.
els, the value of q = 0.15 has been chosen and then the At first the Erbs correlation was used for the decompo-
influence on e and RMS of the various split correlations sition of the global radiation on a horizontal plane in the
was explored. The results of the analysis are reported in two diffuse and beam components and the conventional
Table 3 and Fig. 1b shows the best model for each orien- value of the ground reflection coefficient at q = 0.2 was
tation. assumed.
Examining Table 3 it can be noted that, for all direc- Analysis of the results revealed that there are no major
tions, the Perez anisotropic model, independently of the variations among the predictions of the various models.
split correlations, gives the best results, especially for the The difference of the mean deviations being generally con-
RMS. More precisely, for the south direction, the best tained within a 10% range (with exception of the north
results are given by the Perez–Skartveit combination; for direction for which the maximum difference is 20%). For
the west direction by the Perez–Reindl3 combination; for the south direction, the best model is the isotropic, for
the north direction by the Perez–Reindl2 and for the direc- the west that of Hay and Davies, for the north Perez and
tion east by the Perez–Skartveit combination. However, the for the east direction the Hay and Davies model.
M. Cucumo et al. / Solar Energy 81 (2007) 692–695 695

Analysis of the results obtained for q = 0.15 showed References


that, with this value, the Perez model is the best model
for all orientations. Cucumo, M., Marinelli, V., Oliveti, G., 1994. Ingegneria Solare, Principi
The influence on the results of the use of the split corre- ed Applicazioni, Pitagora Editrice, Bologna.
Erbs, D.G., Klein, S.A., Duffie, J.A., 1982. Estimation of the diffuse
lations was studied. For the values of e, maximum differ- radiation fraction for hourly, daily and monthly average global
ences of 6% were observed for the north direction, while radiation. Solar Energy 28, 293–302.
for the other directions they were contained between 1% Hay, J.E., Davies, J.A., 1980. Calculation of Solar Radiation Incident on
and 4%. an Inclined Surface. In: Proceedings of first canadian solar radiation
In particular, for the south direction, the best results data workshop, Ministry of Supply and Service Canada, 59.
Ineichen, P., Perez, R., Seals, R., 1987. The importance of correct albedo
were given by the Perez–Skartveit combination with e ¼ determination for adequately modelling energy received by tilted
0:03% and RMS = 18.15%; for the west direction by surfaces. Solar Energy 39 (No. 4), 301–306.
the Perez–Reindl3 combination with e ¼ 0:13% and Perez, R. et al., 1986. An anisotropic hourly diffuse radiation model for
RMS = 18.81; for the north direction by the Perez–Reindl2 sloping surfaces: description, performance, validation, site dependency
evaluation. Solar Energy 36, 481–575.
e ¼ 3:05% and RMS = 17.61%, for the direction east by the
Reindl, D.T., Beckman, W.A., Duffie, J.A., 1990a. Diffuse fraction
Perez–Skartveit with e ¼ 1:53% and RMS = 22.84%. correlations. Solar Energy 45, 1–7.
In the conclusion, it can be outdrawn that with low pre- Reindl, D.T., Beckman, W.A., Duffie, J.A., 1990b. Evaluation of hourly
cision input data, the importance of the model is not tilted surface radiation models. Solar Energy 45, 9–17.
important. In contrary, to obtain better transposition val- Skartveit, A., Olseth, J.A., Tuft, M.E., 1998. An hourly diffuse fraction
ues, a more accurate model is necessary, and a precise input model with correction for variability and surface albedo. Solar Energy
63, 173–183.
like the albedo is essential.

You might also like