Sei sulla pagina 1di 5

Renewable Energy 36 (2011) 1944e1948

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Estimating monthly average daily diffuse solar radiation with multiple predictors: A case study
Huashan Li*, Weibin Ma, Xianlong Wang, Yongwang Lian
Key Laboratory of Renewable Energy and Gas Hydrate, Guangzhou Institute of Energy Conversion, Chinese Academy of Sciences, No. 2 Nengyuan Road, Wushan, Tianhe, Guangzhou 510640, China

a r t i c l e i n f o
Article history: Received 11 June 2010 Accepted 11 January 2011 Available online 5 February 2011 Keywords: Solar radiation Global radiation Diffuse radiation Meteorological parameter Empirical model

a b s t r a c t
Solar radiation measurements are not easily available, especially for the diffuse solar radiation. In this study, two models for estimating the diffuse solar radiation are proposed based on multiple predictors including the clearness index, relative sunshine duration, ambient temperature and relative humidity. One of them aims to increase the estimation accuracy, and the other aims to estimate the diffuse solar radiation direct from other meteorological elements in the absence of the global solar radiation. For a case study, the performance of the proposed models is validated by comparing with eight existing models selected from literature against the measured data at Guangzhou station in China. Through the analysis based on statistical error tests, results show that the two models can estimate the monthly average daily diffuse solar radiation with good accuracy. 2011 Elsevier Ltd. All rights reserved.

1. Introduction Accurate knowledge of the amount of the global solar radiation and its diffuse component is of fundamental importance for studying, planning and designing solar energy systems [1]. However, in most areas of the world, solar radiation measurements are not easily available due to nancial, technical or institutional limitations [2], especially for the diffuse solar radiation [3]. Therefore, it is necessary to develop methods to estimate the diffuse solar radiation from other more readily available meteorological data. Since Liu and Jordans [4] work for estimating the diffuse solar radiation from empirical models, extensive researches have been carried out to elaborate the models [5e17]. For example, Orgill and Hollands [5], Erbs et al. [6], Maxwell [7], Skartveit and Osleth [8], Perez et al. [9] and Boland et al. [10,11], respectively, developed models to estimate the hourly diffuse solar radiation; Iqbal [12], ElSebaii and Trabea [13], Aras et al. [14], Mubiru and Banda [15], Pandey and Katiyar [16] and Jiang [17], respectively, proposed models for calculating the daily diffuse solar radiation. It is noted that, among them, most models relate the diffuse fraction (ratio of diffuse solar radiation to global solar radiation) as a function of the clearness index (ratio of global solar radiation to extraterrestrial radiation), relative sunshine duration or a combination of them

* Corresponding author. Tel.: 86 20 87057792; fax: 86 20 87057791. E-mail address: lihs@ms.giec.ac.cn (H. Li). 0960-1481/$ e see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.renene.2011.01.006

with varying degree order polynomials. Although these models have good estimates, it is worth mentioning that they cannot work with the global solar radiation unavailable, which is a general problem resulted from the sparse distribution of the solar radiation observation. The limitation also restricts the application of the models correlating the diffuse transmittance (ratio of diffuse solar radiation to extraterrestrial radiation) with the clearness index. Besides, although the ambient temperature and relative humidity are the more readily available meteorological elements recorded at many standard meteorological stations [18] and often used to model the global solar radiation [19e22], only a few studies proposed these data as predictors to model the diffuse solar radiation. Reindl et al. [23] used stepwise regression to reduce a set of 28 potential predictors down to four predictors and presented that the clearness index, ambient temperature, relative humidity and solar altitude are the signicant predictors to the diffuse solar radiation. Garrison [24,25] also pointed that atmospheric precipitable water, closely related to the ambient temperature and relative humidity [26], is one of the important factors inuencing the diffuse fraction. In addition, the average values of the best root mean square errors for the models of diffuse fraction-clearness index, diffuse fractionrelative sunshine duration, diffuse transmittance-clearness index and diffuse transmittance-relative sunshine duration type are 0.0233, 0.0347, 0.0420 and 0.0490 MJ/m2, respectively, for estimating the monthly average daily diffuse solar radiation at Ankara, Istanbul and Izmir in Turkey according to [3], while the model correlating the diffuse fraction to the clearness index, ambient

H. Li et al. / Renewable Energy 36 (2011) 1944e1948

1945

the highest Hd/H value occurs in April at 0.78, and the lowest occurs in November at 0.46. With Kt in most months, as well as the average value, around 0.34, and the average Hd/H about 0.60, it can be deduced that the skies over Guangzhou are partly cloudy during most of the year. 2.2. Models development 2.2.1. Proposed models In this study, to improve the accuracy for estimating the monthly average daily diffuse solar radiation Hd, a model connecting Hd/H with Kt, S/So, Ta and Rh is proposed and dened as follows:

Hd =H a bKt cS=So dTa eRh


Fig. 1. Annual variation of the monthly average daily clearness index Kt, diffuse fraction Hd/H and sunshine duration S, and the monthly average ambient temperature Ta and relative humidity Rh during 1971e2000 in Guangzhou.

(1)

With target to simulate Hd under conditions of H unavailable, we develop another model to calculate the diffuse transmittance Hd/Ho direct from other meteorological elements:

temperature and relative humidity developed by Janjai et al. [26] for Bangkok of Thailand reduced the root mean square error by 8.7% as compared with the diffuse fraction-clearness index model. This implies that to develop a more accurate model for estimating the diffuse solar radiation, these meteorological parameters should be taken into account. Based on multiple predictors, the primary objectives of the present study are to 1) develop a model for estimating the monthly average daily diffuse solar radiation with better accuracy, and 2) propose a model for estimating the monthly average daily diffuse solar radiation direct from other meteorological elements in the absence of the global solar radiation. The multiple predictors include the clearness index, relative sunshine duration, ambient temperature and relative humidity. For a case study, the performance of the proposed models is validated by comparing with eight existing models selected from literature against the measured data at Guangzhou station in China using statistical error tests. 2. Methodology 2.1. Meteorological data and study area The data used in this work are the monthly average daily global solar radiation H, diffuse solar radiation Hd and sunshine duration S, and the monthly average ambient temperature Ta and relative humidity Rh, covering 30 years between 1971 and 2000 at Guangzhou station (latitude 23100 N and longitude 113 200 E). These data are available from the National Meteorological Information Center, China Meteorological Administration. Guangzhou is a representative region (latitude 22 500 e23 560 N and longitude 112 570 e114130 E) characterized by the subtropical monsoon climate in southern China [27], with a hot and wet season from April to September, and a cool and dry season from December to February. Fig. 1 shows that Ta ranges from 13.6  C in January to 28.6  C in July, while the average value is about 22.1  C; Rh is high in all seasons and the average value is approximately 77.5%; and the average S is about 4.4 h with the maximum and minimum values appearing in July and March, respectively. Besides, the diffuse fraction Hd/H is higher than the clearness index Kt throughout all months, caused by the regions relatively high precipitable water content during the wet season, and the aerosol during the dry season [27,28]. The highest Kt value occurs in November at 0.45, and the lowest occurs in March at 0.22, while

Hd =Ho a bS=So clogS=So dTa eRh

(2)

2.2.2. Existing models Similar models incorporating Ta and Rh were proposed by Reindl et al. [23] for ve European and North American locations and Janjai et al. [26] for Bangkok of Thailand dened as follows, respectively:

Hd =H a bKt cTa dRh esina Hd =H a bKt cTa dRh

(3) (4)

In addition, there are many widely used models relating Hd/H and Hd/Ho as a function of Kt, S/So or a combination of them with varying degree order polynomials [3,12e17]. In this paper, we choose the following polynomials for comparative study:

Hd =H a bKt cKt2 dKt3 Hd =H a bS=So cS=So 2 dS=So 3 Hd =H a bKt cKt2 dS=So eS=So 2 Hd =Ho a bKt cKt2 dKt3 Hd =Ho a bS=So cS=So 2 dS=So 3 Hd =Ho a bKt cKt2 dS=So eS=So 2

(5) (6) (7) (8) (9) (10)

In the aforementioned models, Ho is the monthly average daily extraterrestrial radiation, So is the monthly average day length, a is the monthly average solar altitude at solar noon, and a, b, c, d and e are the empirical coefcients. Ho, So and a can be calculated by Dufe and Beckman [29]. 2.3. Performance evaluation The performance of the models is evaluated on the basis of the following statistical error tests: the mean absolute percentage error (MAPE), mean bias error (MBE), root mean square error (RMSE), correlation coefcient (r) and NasheSutcliffe equation (NSE). These indicators can be calculated as follows:

1946

H. Li et al. / Renewable Energy 36 (2011) 1944e1948

Table 1 Regression coefcients and errors of models (1)e(10) at Guangzhou station. No 1 2 3 4 5 6 7 8 9 10 a 0.4461 0.5686 1.4668 1.3686 1.2433 1.4069 0.7463 0.1366 0.1637 0.0493 b 0.4187 0.3724 1.7335 1.6036 3.7455 6.3866 1.2922 0.4425 0.1875 1.4146 c 0.8972 0.2991 0.0121 0.0080 9.3735 18.4206 3.7966 3.8771 1.7268 1.9500 d 0.0049 0.0031 0.3707 0.4912 10.8486 18.3577 0.7285 5.6212 2.2804 0.0306 e 0.3231 0.2035 0.2683 e e e 1.0592 e e 0.1269 MAPE 1.3902 2.1043 2.4054 2.3908 3.5452 4.3621 3.5363 3.5700 3.5571 3.5398 MBE 0.0092 0.0284 0.0112 0.0044 0.0278 0.0455 0.0239 0.0266 0.0244 0.0238 RSME 0.1019 0.1556 0.1724 0.1774 0.2626 0.3283 0.2711 0.2637 0.2729 0.2691 r 0.9952 0.9716 0.9865 0.9864 0.9629 0.9473 0.9637 0.9165 0.9164 0.9178 NSE 0.9950 0.9883 0.9856 0.9847 0.9665 0.9477 0.9643 0.9663 0.9639 0.9649

MAPE MBE

n 1X Hdci Hdmi 100 n i1 Hdmi

(11) (12)

n 1X H Hdmi n i 1 dci

measured values, respectively, and n is the number of observations. For better data modeling, these statistics should be closer to zero, but r and NSE should approach to 1.0 as closely as possible. 3. Results and discussion The measured data at Guangzhou station are used to t the 10 models (Eqs. (1)e(10)). Table 1 shows regression coefcients of the 10 models. Also, the statistical test results of MAPE, MBE, RMSE, r and NSE for these models are shown in Table 1. From Table 1, it can be seen that for all models, MAPE does not go beyond 4.5%, MBE lies between 0.0455 and 0.0112 MJ/m2, RMSE ranges from 0.1019 to 0.3283 MJ/m2, and r and NSE are larger than 0.9, which is good. Except for models 1, 3 and 4, all models have negative values of MBE, i.e., these models generally underestimate Hd. Table 1 also shows that the ambient temperature and relative humidity introduced into the empirical model can generally improve the estimates, which can be seen from the error comparisons between models 1e4 and 5e10. The average values of MAPE, MBE, RMSE, r and NSE for models 1e4 are 2.0727%, 0.0009 MJ/m2, 0.1518 MJ/m2, 0.9849 and 0.9884, respectively, while for models

v u n u1 X RMSE t H Hdmi 2 n i 1 dci Pn i 1 Hdci Hdca ,Hdmi Hdma r v" # " # u u Pn Pn 2 2 t i 1 Hdci Hdca , i 1 Hdmi Hdma Pn H Hdci 2 NSE 1 Pni 1 dmi 2 i 1 Hdmi Hdma

(13)

(14)

(15)

where Hdci and Hdmi are the ith calculated and measured values, respectively, Hdca and Hdma are the average of the calculated and

Fig. 2. Comparison of the estimated value of monthly average daily diffuse solar radiation Hd from models 1e10 with the measured data.

H. Li et al. / Renewable Energy 36 (2011) 1944e1948

1947

5e10 are 3.6851%, 0.0287 MJ/m2, 0.2780 MJ/m2, 0.9374 and 0.9623, respectively. Based on these statistical indicators, it is concluded that the accuracy of the models for estimating Hd can be improved by incorporating Ta and Rh. Besides, the error comparisons between models 3 and 4 show that the accuracy of the model is not improved by the solar altitude introduced. More importantly, it is found that model 1 shows the best results among all models. This is due to model 1 has the lowest value of MAPE and RMSE and the highest values of r and NSE compared with the other models. The values of MAPE, RMSE, r and NSE for model 1 are 1.3902%, 0.1019 MJ/m2, 0.9952 and 0.9950, respectively. These are considered excellent indicators in that model 1 can estimate Hd with better accuracy in comparison with other models. In addition, with the values of MAPE and RMSE being the second lowest, respectively, at 2.1043% and 0.1556 MJ/m2, and NSE being the second highest at 0.9883, model 2 shows second best performance. This means that model 2 can be used to estimate Hd in the absence of H with good accuracy, which is better than that of model 9, the only existing model selected from literature able to estimate Hd without the global solar radiation. For clarity, the values of Hd obtained from models 1e10, are compared with the measured data in Fig. 2. As can be seen from the gure, the agreement between the values obtained from model 1 and the measured values is better than the eight models from literature, as is the agreement for model 2. These agreements can be considered as indications of the applicability of models 1 and 2, respectively. It is worthwhile to mention that the performance of models 1 and 2 is better than that of models 3 and 4 owing to the fact that the incorporation of S/So, the most inuential parameter next to Kt [30], improves the estimates. Models 5e10 cannot accurately account for Hd throughout the year except for November, and the maximum errors occur in February overestimating Hd approximately by 7%. Hence, the addition of Ta and Rh to exert a signicant inuence on Hd modeling is reasonable. 4. Conclusions Meteorological observations such as the ambient temperature and relative humidity are commonly measured at many stations around the world. However, models for estimating the diffuse solar radiation using these elements are rare. The present study develops two models with multiple predictors including the clearness index, relative sunshine duration, ambient temperature and relative humidity aiming to increase the estimation accuracy and estimate the diffuse solar radiation direct from other meteorological elements in the absence of the global solar radiation, respectively. For a case study, the measured meteorological data at Guangzhou station in China are used to validate the performance of the two models. Among our main results, we nd that the monthly average temperature and relative humidity introduced into the empirical model can generally improve the estimates. Further, based on the statistical error tests such as MAPE, MBE, RMSE, r and NSE, it is concluded that the proposed models 1 and 2 can be used to estimate the monthly average daily diffuse solar radiation with good accuracy. Therefore, the two models can be condently used to estimate the monthly average daily diffuse solar radiation in regions like Guangzhou characterized by the subtropical monsoon climate, where the data are missing or not available. Admittedly, a limitation of this study is that the validation of the proposed models is constrained at one station as a case study, indicating that our results should be used cautiously in regions subject to other climates. Future efforts should explore the performance of the two models in other climate regions, and then develop models for general application.

Acknowledgments The research is supported by the Key Laboratory of Renewable Energy and Gas Hydrate, Chinese Academy of Sciences (No. y107jc), and the National Natural Science Foundation of China (No. 50506025). We sincerely thank Dr. Liang Zhao at the State Key Laboratory of Multiphase Flow in Power Engineering, Xian Jiaotong University for his helpful suggestions on this paper. We would also like to thank the National Meteorological Information Center, China Meteorological Administration for its data support. Nomenclature

monthly average daily diffuse solar radiation on horizontal surface (MJ/m2) H monthly average daily global solar radiation on horizontal surface (MJ/m2) monthly average daily extraterrestrial radiation (MJ/m2) Ho monthly average daily clearness index Kt S monthly average daily sunshine duration (h) monthly average day length (h) So monthly average ambient temperature ( C) Ta monthly average relative humidity (%) Rh a monthly average solar altitude at solar noon ( ) a, b, c, d, e empirical coefcients MAPE mean absolute percentage error (%) MBE mean bias error (MJ/m2) RMSE root mean square error (MJ/m2) r correlation coefcient NSE NasheSutcliffe equation Hd References
[1] El-Sebaii AA, Al-Hazmi FS, Al-Ghamdi AA, Yaghmour SJ. Global, direct and diffuse solar radiation on horizontal and tilted surfaces in Jeddah, Saudi Arabia. Applied Energy 2010;87(2):568e76. [2] Wu GF, Liu YL, Wang TJ. Methods and strategy for modeling daily global solar radiation with measured meteorological data e A case study in Nanchang station, China. Energy Conversion and Management 2007;48(9):2447e52. [3] Ulgen K, Hepbasli A. Diffuse solar radiation estimation models for Turkeys big cities. Energy Conversion and Management 2009;50(1):149e56. [4] Liu BYH, Jordan RC. The inter-relationship and characteristic distribution of direct, diffuse and total solar radiation. Solar Energy 1960;4(3):1e19. [5] Orgill JF, Hollands KGT. Correlation equation for hourly diffuse radiation on a horizontal surface. Solar Energy 1977;19(4):357e9. [6] Erbs DG, Klein SA, Dufe JA. Estimation of the diffuse radiation fraction for hourly, daily and monthly average global radiation. Solar Energy 1982;28 (4):293e302. [7] Maxwell EL. A quasi-physical model for converting hourly global horizontal to direct normal insolation. Solar Energy Research Institute, Golden, CO; 1987. Report SERI/TR-215-3087. [8] Skartveit A, Olseth JA. A model for the diffuse fraction of hourly global radiation. Solar Energy 1987;38(4):271e4. [9] Perez R, Ineichen P, Maxwell E, Seals R, Zelenka A. Dynamic global-to-direct irradiance conversion models. ASHRAE Transactions, Research Series 1992;98:354e69. [10] Boland J, Ridley B, Brown B. Models of diffuse solar radiation. Renewable Energy 2008;33(4):575e84. [11] Ridley B, Boland J, Lauret P. Modelling of diffuse solar fraction with multiple predictors. Renewable Energy 2010;35(2):478e83. [12] Iqbal M. A study of Canadian diffuse and total solar radiation datadI Monthly average daily horizontal radiation. Solar Energy 1979;22(1):81e6. [13] El-Sebaii AA, Trabea AA. Estimation of horizontal diffuse solar radiation in Egypt. Energy Conversion and Management 2003;44(15):2471e82. [14] Aras H, Balli O, Hepbasli A. Estimating the horizontal diffuse solar radiation over the Central Anatolia Region of Turkey. Energy Conversion and Management 2006;47(15e16):2240e9. [15] Mubiru J, Banda EJKB. Performance of empirical correlations for predicting monthly mean daily diffuse solar radiation values at Kampala, Uganda. Theoretical and Applied Climatology 2007;88(1e2):127e31. [16] Pandey CK, Katiyar AK. A comparative study to estimate daily diffuse solar radiation over India. Energy 2009;34(11):1792e6.

1948

H. Li et al. / Renewable Energy 36 (2011) 1944e1948 [24] Garrison JD. Estimation of atmospheric precipitable water over Australia for application to the division of solar radiation into its direct and diffuse components. Solar Energy 1992;48(2):89e92. [25] Garrison JD, Adler GP. Estimation of precipitable water over the United States for application to the division of solar radiation into its direct and diffuse components. Solar Energy 1990;44(4):225e41. [26] Janjai S, Praditwong P, Moonin C. A new model for computing monthly average daily diffuse radiation for Bangkok. Renewable Energy 1996;9 (1e4):1283e6. [27] Zhou K, Ye YH, Liu Q, Liu AJ, Peng SL. Evaluation of ambient air quality in Guangzhou, China. Journal of Environmental Sciences 2007;19(4):432e7. [28] Liu AJ, Du YD, Wang HY. Climatic characteristics of haze in Guangzhou. Meteorological Monthly 2007;30(12):68e71 [in Chinese]. [29] Dufe JA, Beckman WA. Solar engineering of thermal processes. 3rd ed. New Jersey: John Wiley & Sons, Inc; 2006. [30] Munawwar S, Muneer T. Statistical approach to the proposition and validation of daily diffuse irradiation models. Applied Energy 2007;84 (4):455e75.

[17] Jiang YN. Estimation of monthly mean daily diffuse radiation in China. Applied Energy 2009;86(9):1458e64. [18] Skeiker K. Correlation of global solar radiation with common geographical and meteorological parameters for Damascus province, Syria. Energy Conversion and Management 2006;47(4):331e45. [19] Maghrabi AH. Parameterization of a simple model to estimate monthly global solar radiation based on meteorological variables, and evaluation of existing solar radiation models for Tabouk, Saudi Arabia. Energy Conversion and Management 2009;50(11):2754e60. [20] Ododo JC, Usman A. Correlation of total solar radiation with common meteorological parameters for Yola and Calabar, Nigeria. Energy Conversion and Management 1996;37(5):521e30. [21] Ertekin C, Yaldiz O. Estimation of monthly average daily global radiation on horizontal surface for Antalya, Turkey. Renewable Energy 1999;17(1):95e102. [22] Trabea AA, Shaltout MAM. Correlation of global solar radiation with meteorological parameters over Egypt. Renewable Energy 2000;21(2):297e308. [23] Reindl DT, Beckman DT, Dufe JA. Diffuse fraction correlations. Solar Energy 1990;45(1):1e7.

Potrebbero piacerti anche