Hierarchical Bayesian Modeling of Solar Irradiance under Extreme Weather in the Tucson Electric Power Region

Accurate forecasting of surface solar irradiance is needed, as it helps in PV power system planning, particularly under extreme weather conditions. Deterministic and persistence-based forecasting methods generally fail under extreme weather conditions. The present study develops a hierarchical Bayesian spatio-temporal model to forecast solar radiation in the Tucson Electric Power (TEP) region, Arizona, United States. Satellite-derived (CERES SYN1deg) and reanalysis (MERRA-2) solar radiation data have been used in the present study to identify variability across the four TEP stations. The hierarchical Bayesian spatio-temporal model outperformed the persistent model. The findings also highlight that, instead of focusing on point forecasts, we should focus on uncertainty-aware forecasts.