Remote Sensing of Available Solar Energy Resources in Eastern Asia

KENJI OTANI

Energy Electronics Institute
National Institute of Advanced Industrial Science and Technology (AIST)
1-1-4 Umezono, Tsukuba-shi, Ibaraki 305, JAPAN

ABSTRACT

Solar energy maps that indicate the wide-ranging spatial distribution of solar irradiation are required by the researchers of the solar power systems. However, irradiation measurement networks at ground level are not enough to get reliable information on solar energy distribution especially in developing countries. On the other hands, geostationary satellites have provided the images of clouds over the whole surface of the Earth. The main cause of an irregular change in the irradiation at the ground level is effect of the clouds; therefore the method for estimating the irradiation by using the cloud images may be very useful. The solar irradiation for 14 sites in Japan was estimated by the semi-physical model we have developed. The relative errors in monthly-cumulative global irradiation are approximately within +-10% and the rms (root mean square) errors averaged over 14 stations were 0.0804kWh/m² for hourly estimates, 0.534kWh/m² for daily estimates and 7.14kWh/m² for monthly estimates. An example of solar energy maps for Eastern Asia are presented.

INTRODUCTION

The Knowledge of the available solar irradiation at the Earth's surface is essential to many solar power systems in terms of their design, site selection, and performance efficiency. Solar energy maps that indicate the wide-ranging spatial distribution of the irradiation at the ground are required by the researchers not only of the solar power systems but also of the meteorology and the agronomy. However, irradiation measurement networks at ground level are not enough to get reliable information on the spatial distribution of solar energy resources. On the other hands, geostationary satellites (GMS, GOES, INSAT and METEOSAT) have provided the images of clouds over the whole surface of the Earth. The main cause of an irregular change in the irradiation at the ground level is effect of the clouds; therefore, the methods for estimating the irradiation by using the cloud images may be very useful.

The models for estimating the irradiation from the cloud images have been already proposed by several organizations[1][2][3]. Noia et al. reviewed these models and classified into statistical and physical model[4]. The latter model has the advantage of generality because of the removal of regional differences in model parameters. It may be applied anywhere without knowing actually measured irradiation data on the ground level. This approach seems to be preferable for making the solar energy map.

The authors also have developed the technique for remote sensing of solar irradiation by using cloud images transmitted from a Japanese geostationary satellite, Geostationary Meteorological Satellite (GMS), which covers the spatial ranges from 160W to 80E longitudinally, and from 60N to 60S latitudinally[5][6]. The solar irradiation cumulated hourly, daily and monthly were estimated for 14 sites in Japan by our semi-physical model[7]. Estimated values were calculated with the same regression coefficients for any locations; nevertheless, there was little regional difference of estimate accuracy. The correlation coefficients, averaged over 14 stations, were 0.942 for hourly estimates, 0.959 for daily estimates and 0.985 for monthly estimates; the rms errors were 0.0804kWh/m², 0.534kWh/m² and 7.14kWh/m², respectively.

An example of solar energy maps for Eastern Asia in is presented. These maps may compensate for the lack of solar irradiation measurements, where the solar power systems will develop; such as on the sea and on desert areas.

METHODOLOGY

Estimating hourly irradiation

The GMS images in the visible and in the infrared are available at hourly intervals in the daytime. Hourly irradiation was estimated by using a visible image observed instantaneously by GMS VISSR (Visible Infrared Spin Scan Radiometer), whereas the ground truth data of the hourly irradiation measured at pyranometer stations were obtained by accumulating instant measurements during an hour. Fig. 1. shows the solar irradiation balance in the visible spectral band, and the semi-physical model was induced by this model.

Fig.1: ETL semi-physical model

In the absence of cloud, global irradiation H, received at the ground, is supposed to consist of only direct irradiation that is a residual after extraterrestrial irradiation has been attenuated by atmospheric absorption and scattering. The hourly amount of the direct irradiation transmitted after absorption and scattering under the clear sky is

where, ‚h₀: solar constant (1.367kW/m²), ‚š: solar zenith angle and T: atmospheric transmittance under clear sky. The expression, t=0.5, represents half an hour. This time, we gave the assumption that the atmospheric transmittance T is obtained with following two-parameters empirical relationship:

where, z_noon: solar zenith angle at noon, a₁ and a₂: model parameters. a₂ is the corrective factor of the seasonal variation of the transmittance.

Under cloudy sky, two effects of clouds are considered: attenuation of incident irradiation by cloud reflection, and downward reflection referred to as the multiple scattering. The global irradiation is the sum of the direct irradiation and the diffuse irradiation. In our semi-physical model, hourly global irradiation is given by the equation:

where, p_p: GMS-observed albedo, p_s: ground albedo and a₃: model parameter. p_p is given by converting the GMS brightness data with the correlation of the cosine law of illumination, and by averaging over 3 x 3 pixels surrounding their corresponding areas. Since the ground albedo p_s varies according to the location and the season, reference maps of the ground albedo must be produced and be daily updated in order to acquire its exact information. a₃ compensates for the loss of estimated values in terms of sky diffuse irradiation when the solar zenith angle is close to a right angle.

A total of three model parameters, a₁, a₂ and a₃, to be determined by regression, may be independent of the site where the estimation of the irradiation is operated, because of the removal of the regional singularity by considering the ground albedo.

Estimating daily and monthly irradiation

Since GMS provides the visible images ten times per day in summer and eight times per day in winter, more images are usable to estimate daily and monthly irradiation than to estimate the hourly irradiation. However, the images were not available every hour because of the matter of a GMS system, the failure in receiving and so on. In order to fill gaps between the images, an interpolation technique (the trapezoidal integration method) was used for estimating the hourly irradiation. The exact time of sunrise and of sunset for each day were calculated to decide limits of the integration. The monthly cumulative irradiation was calculated as the sum of daily irradiation within a certain month.

Estimating ground albedo

On the clear sky, the dominant element of GMS-observed albedo is ground albedo, which depends on the ground conditions, such as moisture content and vegetation. In order to estimate the irradiation on various ground conditions, the ground albedo must be estimated exactly. If the ground albedo can be detected by GMS images, the semi-physical model may be available for anywhere, with deciding model parameters at a certain site at the least.

Since the ground albedo is much lower than the cloud one (except on a snow-covered surface), and the life-time of cloud is generally shorter than that of snow, it is proposed that the minimum value of the albedo derived at a certain site within a certain period is taken as the index of the ground albedo. Fig. 2. shows, as an example, the plot of daily GMS-observed albedo over Asahikawa, Japan. Solid lines indicate snow depth and estimated ground albedo respectively. The ground albedo was obtained by connecting the minimum value of the GMS-observed albedo for a term of previous 14 days. Agreement with the variation of the ground albedo and of the snow depth indicates the validity of this simple method for estimating the ground albedo. Therefore, estimated ground albedo p_s was assigned the minimum value of p_p observed within a certain period. In terms of the simple technique of remote sensing, not many ground-based measurements must be used. This approach needs only a time series of GMS-observed albedo, and was effective to estimate the ground albedo in spite of its simplicity.

RESULTS

Our semi-physical model was tested at 14 pyranometer stations, which are listed in Table 1. In order to verify that little difference in the value of the model parameters between 14 stations is negligible, each station is assigned common values as the model parameters to calculate the hourly irradiation by equation (2) and (3); which values are determined by regression analysis with the data of all stations. Determined parameters were a₁ = 0.9079, a₂ = -0.2694 and a₃ = 0.0531 in 1992.

Table 1. Geographical coordinates of 14 stations


As shown in Table 2, the resulting correlation coefficient for estimating the hourly irradiation was 0.942, and the rms error was 0.0804kWh/m² (25.1%), on average. Since there was little difference in estimation accuracy between 14 stations, the model parameters may be applicable to anywhere without knowing actually measured irradiation. Accumulation of the hourly irradiation by the trapezoidal integration method for daily estimates increased the correlation coefficient from 0.942 to 0.959, and decreased the rms error from 25.1% to 15.5%, on average. Similarly, accumulation of the daily irradiation for monthly estimates, increased the correlation coefficient from 0.959 to 0.985, and decreased the rms error from 15.5% to 6.84%, on average.

Solar energy distribution in Eastern Asia was calculated by extrapolating the common model parameters mentioned above. Resulting maps have a spatial resolution of approximately 15km and cover the region of 50 longitude and 35 longitude on the center of the map. An example of solar energy maps illustrated in Fig. 4. represents the value of estimated irradiation by eight gray levels.

Table 2: Comparison statistics of observed and estimated irradiation for each ground-truth station

Station	Hourly cumulative irradiation					Daily cumulative irradiation					Monthly cumulative irradiation
	Number of cases	correlation coefficient	rms error (kWh/m2)	rms error (%)		Number of cases	correlation coefficient	rms error (kWh/m2)	rms error (%)		Number of cases	correlation coefficient	rms error (kWh/m2)	rms error (%)
Asahikawa	2919	0.936	0.0820	27.04		363	0.963	0.574	17.64		12	0.990	8.53	8.67
Obihiro	2921	0.943	0.0728	24.96		363	0.955	0.512	16.36		12	0.983	5.68	6.01
Morioka	2937	0.915	0.0925	31.68		363	0.942	0.631	20.08		12	0.988	5.24	5.52
Yamagata	2941	0.939	0.0791	26.03		363	0.963	0.491	14.99		12	0.993	4.99	5.03
Fukushima	2942	0.932	0.0848	28.10		363	0.951	0.558	17.11		12	0.991	4.27	4.33
Maebashi	2944	0.954	0.0731	24.28		363	0.960	0.555	17.08		12	0.962	11.02	11.21
Chichijima	2983	0.946	0.0858	22.21		363	0.961	0.512	12.05		12	0.994	4.10	3.19
Matsumoto	2923	0.939	0.0944	26.03		361	0.949	0.701	17.86		12	0.993	12.60	10.67
Kohfu	2945	0.953	0.0740	21.44		363	0.959	0.507	13.54		12	0.988	7.62	6.73
Nagoya	2935	0.957	0.0692	20.58		363	0.970	0.442	12.15		12	0.986	5.68	5.17
Hikone	2922	0.957	0.0705	21.95		362	0.972	0.453	13.03		12	0.987	6.08	5.79
Osaka	2925	0.933	0.0863	27.92		363	0.959	0.559	16.74		12	0.976	10.29	10.18
Nara	2928	0.941	0.0799	25.38		363	0.964	0.494	14.55		12	0.972	8.37	8.15
Saga	2879	0.940	0.0806	24.08		362	0.965	0.488	13.62		12	0.985	5.49	5.08
MAX	2983	0.957	0.0944	31.68		363	0.972	0.701	20.08		12	0.994	12.60	11.21
MIN	2879	0.915	0.0692	20.58		361	0.942	0.442	12.05		12	0.962	4.10	3.19
MEAN	2932	0.942	0.0804	25.12		363	0.959	0.534	15.49		12	0.985	7.14	6.84

Fig.4: Example solar energy map for Eastern Asia in June 1992

CONCLUSION

In this paper, hourly visible images broadcasted by GMS have been used for composing solar energy maps. The method for estimating the irradiation by using GMS images and for composing the solar energy map are described. The methodology, which may have the advantage of generality, was examined for 14 sites in Japan. The resulting correlation coefficients were 0.942 for hourly estimates, 0.959 for daily estimates and 0.985 for monthly estimates, on average; the rms errors were 0.0804kWh/m², 0.534kWh/m² and 7.14kWh/m², respectively. The solar energy maps calculated by above model has high ground resolution (approximately 15 km) and large area of effective mapping coverage in Eastern Asia. For designing solar power systems, these solar energy maps may be useful to where the solar irradiation have not been measured in Eastern Asia.

REFERENCES

[1] D. Cano et al., "A method for the determination of the global solar radiation from meteorological satellite data", Solar Energy, 37, 1986, pp. 31-39.
[2] L. Diabate et al., "Description of an operational tool for determining global solar radiation at ground using geostationary satellite images",Solar Energy, 42, 1989, pp. 201-207.
[3] A. Ben Djemaa et al., "A comparison between one year of daily global irradiation from ground-based measurements versus METEOSAT images from seven locations in Tunisia", Solar Energy, 48, 1992, pp. 325-333.
[4] M. Noia et al., "Solar irradiance estimation from geostationary satellite data", Solar Energy, 51, 1993, pp. 449-465
[5] K. Kurokawa et al., "Statistical formation of solarenergymap by cloud images received from GMS", Pacific Ocean Remote Sensing Conf., PORSEC'94, 1994, pp. 435-442
[6] K. Otani et al., "Estimation of ground albedo by GMS images for solar irradiation monitoring", Solar Energy Materials and Solar Cells, 35, 1994, pp. 395-400
[7] K. Otani et al., "Solar energy mapping by using cloud images received from GMS", First World Conference on Photovoltaic Energy Conversion, WCPEC-1, Hawaii, 1994, to be published