Evaluation of weather research forecasting-solar radiation schemes for solar power forecasting in equatorial Africa

Among the three primary techniques for solar power forecasting—Machine Learning models, statistical models, and Numerical Weather Prediction (NWP)—NWP techniques, particularly Weather Research Forecasting (WRF), have shown superior performance for hourly to 6-day ahead forecasts. The advancement of WRF has led to the development of WRF-Solar, tailored specifically for solar power forecasting tasks. However, its effectiveness in equatorial Africa hinges on the selection of appropriate radiation schemes, which significantly influence surface solar power forecasts. While previous studies have examined both shortwave and longwave radiation schemes, none have specifically focused on the African continent, particularly in equatorial regions, taking into consideration the different weather. This study aims to fill this gap by evaluating the forecast skill of various shortwave and longwave radiation scheme configurations for solar power forecasting using WRF-Solar across the different weather conditions experienced throughout the year in Nairobi, Kenya. The results obtained showed that a combination of the Rapid Radiative Transfer Model for Global Circulation Models (RRTMG) shortwave and longwave radiation schemes consistently performed better in terms of the Mean Squared Error and Root Mean Squared Error metric while a combination of the Rapid Radiative Transfer Model (RRTM) shortwave and longwave radiation scheme performed optimally in terms of the Mean Absolute Error metric, delivering best performance for two out of four weather seasons. These results emphasize the need for an approach to selecting radiation schemes within the WRF-Solar model, considering the seasonal variability and specific meteorological conditions of the region. • Three key solar forecasting methods are ML, statistical models & NWP, with NWP dominating in medium-term forecasts. • WRF-Solar (WRF tailored for solar forecasting tasks) have shown superior performance for hourly to 6-day ahead forecasts. • This study evaluates WRF-Solar setups for solar power forecasting under varying weather in Nairobi, Kenya. • Results obtained showed RRTMG outperformed in MSE/RMSE across seasons, while RRTM showed better results based on the MAE metric. • Results highlight the importance of selecting WRF-Solar radiation schemes based on seasonal and regional weather patterns.