Probabilistic solar irradiance forecasting using numerical weather prediction ensembles over Australia

Uncertainty information accompanying solar irradiance forecasts is increasingly being regarded important to the solar energy industry. We use solar irradiance forecasts from three leading operational numerical weather prediction models (i.e., IFS, GFS and ACCESS) to quantify such information for 12 locations across Australia. Probabilistic forecasts are formed through two ensemble approaches: one is a multiple-model approach which employs deterministic forecasts from the three models and the other is to use the 51 ensemble forecasts from the IFS ensemble model. It was found that the multiple-model approach exhibits an excellent performance, as evaluated by both deterministic and probabilistic metrics, despite it has fewer ensemble members.