Post-processing of solar irradiance forecasts from WRF model at Reunion Island

•WRF model is used to forecast hourly global solar radiation at Reunion Island.•Kalman filtering is applied to enhance WRF forecasts.•Hourly post-processing of GHI forecasted with Kalman filtering gave better results than direct WRF output.•rRMSE decreased from 37.2% to 22.24% and rMBE from 6.34% to 0.3%.

An efficient use of solar energy production requires reliable forecast information on surface solar irradiance. This article aims at providing a Model Output Statistics (MOS) method of improving solar irradiance forecasts from Weather Research and Forecasting (WRF) model.The WRF model was used to produce one year of day ahead solar irradiance forecasts covering Reunion Island with an horizontal resolution of 3 km. These forecasts are refined with a Kalman filter using hourly high quality ground measurements. Determination of the relevant data inputs for the Kalman filter method is realized with a bias error analysis. The solar zenith angle and the clear sky index, among others, are used for this analysis.Accuracy of the method is evaluated with a comprehensive testing procedure using different error metrics. The Kalman filtering appears to be a viable method in order to improve the solar irradiance forecasting.