2015 Volume 36 Issue 4 Pages 31-39
A numerical weather prediction model (NWP) has become a useful tool for a forecast of photovoltaic power generation in an efficient operation of an energy management system (EMS). However, forecast products obtained by the NWP have certainly forecast errors. Before using forecast datasets of solar irradiance (or global horizontal irradiance, GHI), we have to understand error characteristic of GHI forecasts from the NWP.
In this study, we validated GHI forecasts obtained from a local forecast model (LFM) of the Japan Meteorological Agency (JMA) using the surface-observed GHI data at the JMA stations. The LFM is a high-resolution NWP with horizontal grid spacing of 2km. First of all, case studies of GHI forecasts were conducted for different weather conditions. In cases of clear sky conditions and/or spatially large clouds, hourly-forecasted GHI values were significantly close to hourly observed-GHI values. However, significant forecast errors tended to be large in cases of high-level clouds (cirrus) and/or cumulus clouds. Seasonal and regional variations of GHI forecast errors were found from evaluations of forecast errors. Compared to a mesoscale model (MSM, with horizontal grid spacing of 5km) GHI forecasts, regional characteristics of LFM forecast errors were different from MSM forecast results.
