Evaluating Day-Ahead Solar Radiation Forecasts from ICON, GFS, and MeteoFrance Global NWP Models

Abstract

The accuracy of hourly global horizontal irradiance forecasts provided by the ICON (Icosahedral Nonhydrostatic Weather and Climate Model), GFS (Global Forecast System), and MeteoFrance Global models for Moscow, Russia, over a 36-h forecast horizon, is analyzed. The study investigates the suitability of these freely available global Numerical Weather Prediction (NWP) models for predicting the output of PV power plants, considering the standards of the Russian wholesale electricity and capacity market. The study conducts a comparative assessment of the forecasts from the mentioned models. A notable conclusion drawn from the study is the superior performance of all the NWP models over the naive 36-h persistence forecast, as indicated by the root mean square error, mean absolute error and mean bias error metrics. The overall findings suggest a higher accuracy of hourly global horizontal irradiance forecasts from the ICON model compared to the others under the investigated conditions.