Evaluating NOAA’s HRRR v4 and RRFS v1 models as forecasters of day-ahead solar irradiance: Skill, spread, and steps forward

Short-term forecasting—from minutes to hours to days ahead—is essential for electric power grid operations. While Machine Learning techniques are increasingly used, numerical weather prediction (NWP) models still form the backbone of load and renewable generation forecasts used by grid operators. One of the most used NWP models by US grid operators is NOAA’s High Resolution Rapid Refresh (HRRR) model. However, NOAA has been developing a next generation NWP model known as the Rapid Refresh Forecast System (RRFS). Unfortunately, there is limited information available regarding the performance of RRFS for energy applications. Here we evaluate forecasts of global horizontal irradiance (GHI) from NOAA’s HRRRv4 and the experimental RRFSv1 against observations from the SURFRAD network in the Continental US. Our results indicate the HRRR produces more skillful GHI forecasts than the RRFS at most SURFRAD sites, which we trace to a persistent overestimate of GHI in the RRFS. Expanding our analysis to the RRFS’s 5-member perturbed physics ensemble, we make three findings: (1) All RRFS ensemble members are biased towards high GHI forecasts, with two members having significantly larger errors than the deterministic forecast, (2) The RRFS ensemble is under-dispersed, missing both high and low GHI events, and (3) Despite shared biases among RRFS forecast products, the RRFS ensemble-mean produces lower error than the single HRRR deterministic forecast across all sites. We conclude by discussing how the design and tuning of NWP models could evolve to consider emerging applications of weather forecasts in the New Energy Economy.