Advances in Solar Power Forecasting

Abstract:

The National Center for Atmospheric Research and partners are building a blended SunCast Solar Power Forecasting system. This system includes several short-range nowcasting models and improves upon longer range numerical weather prediction (NWP) models as part of the “Public-Private-Academic Partnership to Advance Solar Power Forecasting.” The nowcasting models being built include statistical learning models that include cloud regime prediction, multiple sky imager-based advection models, satellite image-based advection models, and rapid update NWP models with cloud assimilation. The team has also integrated new modules into the Weather Research and Forecasting Model (WRF) to better predict clouds, aerosols, and irradiance. The modules include a new shallow convection scheme; upgraded physics parameterizations of clouds; new radiative transfer modules that specify GHI, DNI, and DIF prediction; better satellite assimilation methods; and new aerosol estimation methods. These new physical models are incorporated into WRF-Solar, which is then integrated with publically available NWP models via the Dynamic Integrated Forecast (DICast) system as well as the Nowcast Blender to provide seamless forecasts at partner utility and balancing authority commercial solar farms. The improvements will be described and results to date discussed.