9.6      Enhancing WRF-Solar to provide solar irradiance probabilistic forecasts.

 

Kim, Ju-Hye, Pedro A. Jimenez, National Center for Atmospheric Research, Manajit Sengupta, Jaemo Yang, National Renewable Energy Laboratory, Jimy Dudhia, NCAR, Yu Xie, NREL, and Branko Kosovic, NCAR

 

WRF-Solar was the first numerical weather prediction (NWP) model specifically designed to provide solar irradiance forecast tailored for solar energy applications. Developments focused on improving the cloud-aerosol-radiation physics. New developments are underway to produce probabilistic solar irradiance forecast. Our approach consists of introducing stochastic perturbations of key variables and parameters controlling the surface irradiance. These variables and parameters are being identified using adjoint sensitivity analysis of physics packages responsible for the irradiance variability. We first separate each physics module responsible for cloud-aerosol-radiation processes as a standalone system and then generate tangent linear and adjoint codes using Transformation of Algorithm in Fortran (TAF). Linearity/adjointness tests are made for a correctness verification and sensitive variables are finally being identified by analyzing adjoint sensitivity of irradiance to model variables. Subsequently, the WRF-Solar model will be configured to introduce perturbations on these variables and parameters in order to produce the probabilistic forecasts. This presentation will describe our initial results with the WRF-Solar probabilistic forecasts as well as an overview of these efforts that will convert WRF-Solar into the first NWP model providing probabilistic forecasts specifically tailored for solar energy applications.