Development of a Short-term Solar Irradiance Forecasting using Satellite Image in Combination with Numerical Weather Prediction Model

Abstract

A short-term solar irradiance forecasting method has been developed to provide highly accurate prediction, using satellite image prediction model in combination with numerical weather prediction model. This method was applied and evaluated using the solar irradiance observational site at the CRIEPI Akagi Testing Center (Gunma Prefecture, Japan).

Based on historical forecasting and observation datasets at JMA observation sites in the computational domain, the best blending ratio between the satellite image prediction model and the numerical weather prediction model was selected and used. The developed method has shown high accuracy. Compared to satellite image forecasting, accuracy results averaged between three and six hours of lead-time showed improvements in the RMSE by 44% for the before-sunrise case, 20% for the after-sunrise case, and 8% for the noon to sunset case.