Beruflich Dokumente
Kultur Dokumente
SOLAR RADIATION
Abstract⎯ The main objective of this study is to develop a linear regression model for estimating radiation for
some selected cities of Jharkhand region. Taking sunshine hours as the prime concern, the three main places
which come under this region, namely—Jamshedpur (longitude 86°11′ E, latitude 22°48′ N), Ranchi (longi-
tude 85°20′ E, latitude 23°21′ N) and Bokaro (longitude 86°09′ E, latitude 23°40′ N) a set of regression con-
stants were obtained in order to develop the linear regression model. The new developed models estimated
the value of regression constant “a” which is ranging between 0.204 to 0.211 and value of regression constant
“b” ranging from 0.489 to 0.514. Results are compared with measured data and some well known models with
the help of statistical test for city Ranchi. Finally, the proposed model was preferred for estimation of solar
radiation in Ranchi, with smallest statistical errors among all models and close agreement with measured
data.
DOI: 10.3103/S0003701X16030099
164
EMPIRICAL MODELS FOR THE ESTIMATION 165
radiation data is still needed for Jharkhand (India) relates monthly average daily global radiation to the
region in the absence of meteorological stations. This average daily sunshine hours, and is given by the fol-
study, therefore, addresses this need. For this we mea- lowing expression:
sured solar energy radiation for the period (2010–
2012) for the cities under consideration. We then sup-
ported our measurements by developing the linear
Hg
H0
=a+b n ,
N ( ) (1)
regression model for the three main cities in this
region. With the help of this developed model, global where H 0 is extraterrestrial radiation on a horizontal
solar radiation has been estimated and compared with surface.
the existing model as well as with the measured data in The daily value H 0 was computed according to the
conjunction with different statistical test to show the following equations:
validation of the proposed model. The error range lies
between the prescribed zones [15]. H 0 = 24 × 3600
π
( )
The main objectives of this paper are: (2)
1. To develop a linear regression model for estimat- × I 0 f cos λ cos δ sin ω s + π ω s sin λ sin δ .
180
ing radiation in some selected cities in Jharkhand.
The eccentricity correction factor (f), solar decli-
2. To estimate the monthly average daily global nation (δ) and the sunrise hour angle (ωs) can be
solar radiation on a horizontal surface at Ranchi using respectively calculated as:
the proposed model, including different empirical
relations.
3. Compare each model with measured data of
(
f = 1 + 0.033 cos 360n ,
365 ) (3)
Ranchi using a statistical test which includes MAPE, ⎡360 ( 284 + n)⎤
MBE, RMSE, NSE and t-stat. δ = 23.45 sin ⎢ ⎥⎦ , (4)
⎣ 365
( )
3. DESCRIPTION OF MODELS FOR
ESTIMATING GLOBAL SOLAR RADIATION + 0.0926h − 0.359 n ,
N
The models use to predict the monthly average
= 0.32 + 0.42 ( n ) .
global solar radiations on a horizontal surface, consid- Hg
(11)
ered for this case study are: H0 N
a. Rietveld model f. Present model
( )
Rietveld [24] examined several published values of Hg
“a” and “b” and noted that “a” is related linearly and = 0.2111 + 0.489 n . (12)
H0 N
“b” hyperbolically to the mean value of n such that
N
this equation is believed to be applicable anywhere in 4. COMPARISON AND VALIDATION OF
the world and yields superior results for cloudy condi- MODELS WITH STATISTICAL ERRORS
tions, for n < 0.4. There are many parameters which deal with the
N assessment and comparison of monthly mean daily
( )
solar radiation estimation models. Here the statistical
Hg
= 0.18 + 0.62 n . (7) parameters like the mean bias error (MBE) and the
H0 N root mean square error (RMSE) helps to calculate the
error or the deviation of the calculated value of the
b. Ogleman model
measured value. Mean percentage error (MPE) and
Ogleman [25] et al. proposed the use of a correla- coefficient of correlation (R) tests the linear relation-
tion which relates the global solar radiation to S in a ship between predicted and measured values. The best
quadratic form as: result is when these statistics are closer to zero, but the
coefficient of correlation, R, should approach to 1 as
( ) ( ).
2
Hg closely as possible for better modeling. To improve the
= 0.195 + 0.675 n − 0.142 n (8)
results and better comparison the Nash–Sutcliffe
H0 N N
equation (NSE) is also selected as an evaluation crite-
c. Akinoglu model rion. A model is more efficient when NSE is closer to 1.
Akinoglu and Ecevit [26] suggested a quadratic The errors that have been estimated help to compare
the models, but they do not make the model statisti-
correlation between the ratio of H g H 0 and S to esti- cally significant. The t-statistic allows models to be
mate the values of global solar radiation for 58 loca- compared and at the same time it is carried out to
tions displaced in several countries. This equation, determine statistical significance of the predicted val-
whose coefficients have the same values, respectively, ues by the models.
for all tested locations is
a. The mean bias error
( ) ( )
−2
Hg
= 0.145 + 0.845 n − 0.280 n
n
. (9)
H0 N N MBE = 1
n
∑H i, calc − H i, meas . (13)
1
d. Glover model
This test provides information on long-term per-
Glover and McCulloch [27] attempted to intro- formance. A low MBE value is desired. A negative
duce latitude dependency to one of the Angstrom- value gives the average amount of underestimation in
Prescott coefficients and presented the following the calculated value. So, one drawback of these two
mentioned test is that overestimation of an individual
Hg
H0
= 0.29 cos ϕ + 0.52 n .
N ( ) (10) observation will cancel underestimation in a separate
observation.
Table 1. Regression constant for selected locations The value of RMSE is always positive, representing
Regression constants zero in the ideal case. The normalized root mean
Location square error gives information on the short term per-
a b a+b
formance of the correlations by allowing a term by
Jamshedpur 0.2026 0.514 0.7166 term comparison of the actual deviation between the
Ranchi 0.2111 0.489 0.7001 predicted and measured values. The smaller the value,
Bokaro 0.2039 0.5104 0.7143 the better is the model’s performance.
d. Nash–Sutcliffe equation
b. Mean percentage error n
∑ (H − H i,meas )
2
MPE (%) i,calc
n
⎛ (H i,calc − H i, meas ) ⎞ (14) NSE = 1 − 1
, (16)
∑
n
=1 ⎟ × 100.
⎜
∑ (H − H i,meas )
2
n 1 ⎝ H i,meas ⎠ meas
1
c. Root mean square error
where H meas is the mean measured global radiation. A
12 model is more efficient when NSE is closer to 1.
⎡ n
2⎤
RMSE = ⎢1
⎢⎣n
∑ (H i,calc − H i,meas ) ⎥
⎥⎦
. (15) e. MAPE
1
n
H i,meas − H i.calc
MAPE (%) = 1
n
∑ H i,meas
. (17)
Ratio of global to extraterrestrial
f. Coefficient of correlation
0.60
The coefficient of correlation, r can be used to
solar radiation
( )
Hg MBE shows overestimation and a negative MBE show
= 0.2039 + 0.5104 n . (21) under estimation. In comparison with all the models,
H0 N present model estimates the lowest RMSE having
It is apparent from Eqs. (19–21) that neither “a” 7.82% and the highest one with Gopinathan model
nor “b” vary with latitude or altitude in any systematic (24.9%) followed by globally (20.11%), Rietveld
manner. However, the values of the sum of the regres- (12.5%), Oglemann (9.23%) and Akinoglu (8.47%).
sion constants a + b; which represent the maximum The MBE values obtained from the models are posi-
Table 2. Comparison between measured and estimated monthly average daily global radiation (MJ/(m2-day)) for the city
Ranchi
Rietveld Oglemann Akinoglu Glover Gopinathan Proposed
Month Measured
model model model model model model
Jan 17.079 16.230 16.157 17.204 16.687 14.966 15.630
Feb 19.279 18.427 18.380 19.499 18.917 17.072 17.690
March 21.293 20.623 20.647 21.756 21.198 19.130 20.820
April 24.096 23.258 23.265 24.552 23.798 21.323 22.210
May 23.431 23.001 23.083 24.225 23.688 20.930 21.190
June 16.324 16.973 16.718 18.322 18.945 15.505 16.750
July 14.959 15.683 15.282 17.132 17.952 14.257 14.500
Aug 14.304 15.004 14.610 16.405 17.240 14.074 13.890
Sept 15.408 15.860 15.761 16.946 17.387 14.569 14.900
Oct 16.293 16.272 16.337 17.151 17.128 15.071 15.760
Nov 16.710 16.072 16.061 16.979 16.569 14.721 15.340
Dec 16.211 15.397 15.323 16.324 15.845 14.201 14.680
Measured data
30
Hg MJ/(m2-day)
20
10
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month
Fig. 4. The measured and predicted monthly average daily global solar radiation ( H g ) for the city Ranchi in the generation of
different models.
tive in some cases and negative in others. Values of From the statistical results a new linear model
MBE from all the models except present model indi-
cate an over estimation. Present model has very little
under estimation, i.e. 6.28%. Also, the highest values
Hg
H0 ( )
= 0 . 2111 + 0 . 489 n .
N
of NSE (0.91) as shown in Fig. 9 and lowest values of based on the modified Angstrom model is
t-test results (Table 3) indicate the superiority of the extremely recommended to estimate monthly average
proposed model with respect to others. The daily global solar radiation for Ranchi (Jharkhand)
RMSE(%) value, which is a measure of the accuracy and in elsewhere with similar climate conditions areas
of estimation, has been found to be the lowest for the where radiation data are unavailable. Further, the
present model (7.83%) as shown in Table 3 and plotted in other new proposed models are also being recom-
Fig. 8. The transmissivity of the atmosphere of global mended for estimating the average daily global solar
solar radiation under perfectly clear sky conditions is radiation for Jamshedpur and Bokaro.
given as the sum of the regression coefficients, a + b.
Also, the transmissivity of an overcast atmosphere is
6. CONCLUSIONS
interpreted as the intercept “a.” Hence, the need to
compare present regression relation with others in The objective of this study was to evaluate various
terms of the atmospheric transmissivity values. From models for the estimation of monthly average daily
present regression constants (a = 0.2111 and b = 0.489) global radiation ( H g ) on a horizontal surface from
i.e. 0.7001. The clear-sky transmissivity of most tropical bright sunshine hours for some selected cities of
regions in general seems to lie between 0.68 and 0.75 [18]. Jharkhand and to select the most appropriate model
Table 3. Validation of the models under different statistical test for the city ranchi
Statistical Parameters Present Rietveld Oglemann Akinoglu Glover Gopinathan
Present
R2 Model
Model
Model
Model
Model
MAPE, %
Model
Model
for Jharkhand state. The values of monthly average 3. Okundamiya, M.S., Evaluation of various global solar
global solar radiation are calculated using the models radiation model for Nigeria, Int. J. Green Energy, 2016,
suggested by Rietveld, Ogleman, Akinoglu, Glover, vol. 13, no. 5. . doi 10.1080/15435075.2014.968921
Gopinathan and Present model. The selected models 4. Aras, H., Balli, O., and Hepbasli, A., Global solar radi-
were compared with the present model for estimating ation potential: зart 1: model development, Energy
Sources, 2006, vol. 27. pp. 7–11.
monthly average global solar radiation ( H g ) for Ran-
5. Ahmad, M.J. and Tiwari, G.N., Solar radiation models –
chi, on the basis of statistical error tests such as mean a review, Int. J. Energy Res., 2011, vol. 35, pp. 271–290.
bias error (MBE), the mean percentage error (MPE),
Root mean square error (RMSE), Nash- Sutcliffe 6. Muneer, T., Younes, S., and Munawwar, S., Dis-
courses on solar radiation modeling, Renew. Sust.
equation (NSE), correlation coefficient and the t-test. Energy Rev., 2007, vol. 11, pp. 551–602.
From the statistical results a new linear model 7. Donatelli, M., Bellocchi, G., and Fontana, F., Soft-
ware to estimate daily radiation data from commonly
( )
Hg available meteorological variables, Europ. J. Agronomy,
= 0 . 2111 + 0 . 489 n 2003, vol. 18, pp. 363–367.
H0 N
8. Younes, S. and Muneer, T., Improvements in solar
based on the modified Angstrom model is extremely radiation models based on cloud data, Building Services
recommended to estimate monthly average daily Eng. Res. Technol., 2006, vol. 27, pp. 41–54.
global solar radiation for Ranchi (Jharkhand) and in 9. Angstrom, A., Solar and terrestrial radiation, Q. J. Roy.
elsewhere with similar climate conditions areas where Met. Soc., 1924, vol. 50, p. 121.
radiation data is unavailable. Further, the other new 10. Garg, H.P. and Garg, S.N., Correlation of maonthly
proposed models are also being recommended for average daily global, diffuse and beam radiation with
estimating the average daily global solar radiation for bright sunshine hours, Energy Convers. Manag., 1985,
Jamshedpur, Bokaro. vol. 25, pp. 409–417.
11. Raja, I.A. and Twidell, J.W., Distribution of global
insolation over Pakistan, Solar Energy, 1990, vol. 44,
REFERENCES pp. 63–71.
12. Gopinathan, K.K. and Soler, A., A sunshine dependent
1. Massaquoi, J.G.M., Global solar radiation in Sierra global insolation model for latitudes between 60N and
Leone (West Africa), Solar Wind Technol., 1998, vol. 5, 70N, Renew. Energy, 1992, vol. 2, pp. 401–404.
pp. 281–283.
13. Katiyar, A.K. and Chanchal Kumar Pandey, Simple
2. Ibrahim, S.M.A., Predicted and measured global solar correlation for estimating the global solar radiation on
radiation in Egypt, Solar Energy, 1985, vol. 35, horizontal surfaces in India, Energy, 1994, vol. 35,
pp. 185–188. pp. 5043–5048.
14. Namrata, K., Sharma, S.P., and Seksena, S.B.L., 22. Shukla, K.N., Rangnekar, S., and Sudhakar, K., Com-
Comparision of different models for estimation of dif- parative study of isotropic and anisotropic sky models
fuse solar radiation in Jharkhand (India) region, Appl. to estimate solar radiation incident on tilted surface: a
Solar Energy, 2015, vol. 51, no. 3, pp. 65–70. case study for Bhopal, India, Energy Rep., 2015, vol. 1,
15. Sukhatme, S.P., Solar Energy Principles of Thermal Col- pp. 96–103.
lection and Storage, Tata McGraw-Hill, 1997. 23. Namrata, K., Sharma, S.P., and Seksena, S.B.L.,
16. Solar Radiation Handbook, Solar Energy Centre, Comparison of different models for estimation of global
MNRE. solar radiation in Jharkhand (India) Region, Smart
17. Quansah, E., Amekudzi, L.K., Preko, K., et al., Grid Renew. Energy, 2013, vol. 4, pp. 348–352.
Empirical models for estimating global solar radiation 24. Rietveld, M.R., A new method for estimating the
over the Ashanti Region of Ghana, J. Solar Energy, regression coefficients in the formula relating solar
2014. doi 10.1155/2014/897970 radiation to sunshine, Agricult. Meteorol., 1978, vol. 19,
18. Das, A., Park Jin-Ki and Park Jong-Hwa, Estimation pp. 243–252.
of available global solar radiation using sunshine dura-
tion over south Korea, J. Atmosph. Solar-Terrestrial 25. Ogelman, H., Ecevit, A., and Tasdemiroglu, E., A new
Phys., 2015, vol. 134, pp. 22–29. doi method for estimating solar radiation from bright sun-
10.1016/j.jastp.2015.09.001 shine data, Solar Energy, 1984, vol. 33, pp. 619–625.
19. Doost, A.K. and Akhlaghi, M., Estimation and com- 26. Akinoglu, B.G. and Ecevit, A., A further comparison
parison of solar radiation intensity by some models in a and discussion of sunshine based models to estimate
region of Iran, J. Power Energy Eng., 2014, vol. 2, global solar radiation, Solar Energy, 1990, vol. 15,
pp. 345–351. pp. 865–872.
20. Manzano, A., Martın, M.L., Valero, F., and Armenta, C., 27. Glover, J. and McGulloch, J.D.G., The empirical rela-
A single method to estimate the daily global solar radi- tion between solar radiation and hours of sunshine, Q.
ation from monthly data, Atmosph. Res., 2015, vol. 166, J. Roy. Meteorol. Soc., 1958, vol. 84, pp. 172–175.
pp. 170–182. doi 10.1016/j.atmosres.2015.06.017
28. Gopinathan, K.K. and Soler, A., A sunshine dependent
21. Mousavi, S.M., Mostafavi, E.S., Jaafari, A., et al., global insolation model for latitudes between 60N and
Using measured daily meteorological parameters to 70N, Renew. Energy, 1992, vol. 2, pp. 401–404.
predict daily solar radiation, Measurement, 2015,
vol. 76, pp. 148–155.