Seasonal and Regional Variations of the Range of Forecast Errors of Global Irradiance by the Japanese Operational Physical Model

• Published in: Energy procedia Vol. 57; pp. 1247 - 1256
• Main Authors: Ohtake, Hideaki, Shimose, Ken-ichi, Foncseca, Joao Gari da Silva, Takashima, Takumi, Oozeki, Takashi, Yamada, Yoshinori
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 2014
• Subjects: annual changes; forecast errors; GHI; physical model; PV forecast; regional changes; seasonal changes; PV forecast; annual changes; GHI; regional changes; forecast errors; seasonal changes; physical model
• ISSN: 1876-6102; 1876-6102
• DOI: 10.1016/j.egypro.2014.10.114

For an energy management using large amount of photovoltaic (PV) systems installed in Japan islands, the forecast of a global horizontal irradiance (GHI) based on a numerical weather prediction model (NWP) will be necessary. The Japan meteorological Agency (JMA) has been developed the operational NWP with the horizontal grid spacing of 5km. In this study, the range of forecast errors of GHI values obtained from the NWP are investigated. Seasonal changes of forecast errors and the dependency on the weather conditions using surface-measured GHI data are researched during the period from 2008 to 2011. Results of the validation of hourly GHI forecasts for a station showed that overestimations in relatively higher clearness index (CI) (slightly cloudy weather) were found while underestimations in relatively lower CI (cloudy weather) were found. On the other hands, ranges of forecast errors in the clear sky or the deep cloudy conditions tend to be relatively small. From seasonal changes of the range of forecast errors, it was found that underestimations of GHI in summer tended to be significant. Annual changes of the range of forecast errors were not large during the period. In order to estimate the effect of decreasing forecast errors by the spatial-averaging method, we also analyzed forecast errors of GHI values for the relatively large area in the Kanto region (near Tokyo), located in central Japan. Compared the ranges of forecast errors by the point analysis for Tsukuba station with those for the relatively large area, the ranges of forecast errors by the spatial-averaging method were decreased up to about 70% compared with the range of a point analysis for Tsukuba station.