Enhancing WRF-Solar to provide Solar Irradiance Probabilistic Forecasts under All-sky Conditions

Deterministic forecasting of solar irradiance has considerable limitations in cloudy-sky conditions since an under-prediction of clouds leads to significant model biases. It is also difficult to resolve the extensive feedback between processes involving cloud formation and dissipation by solely improving physics schemes when using deterministic methods. An optimized ensemble-based solar forecasting system is developed within a project funded by the US Department of Energy (DOE) to improve the current state-of-the-art WRF-Solar forecasts and provide probabilistic forecasts for grid operations under all-sky conditions. Our approach consists of introducing stochastic perturbations of key variables and parameters controlling the surface irradiance. These have been identified using tangent-linear and adjoint-based sensitivity analysis of five physics packages responsible for the irradiance variability. Subsequently, the WRF-Solar model has been improved to introduce perturbations on these variables and parameters in order to produce probabilistic forecasts. This presentation will describe the WRF-Solar probabilistic forecasting system and compare initial results to a WRF-Solar deterministic forecast. The National Solar Radiation Data Base (NSRDB) will be used for validation of the forecast results. Calibration through these long-term observations from satellites will also provide an improved forecast when used in conjunction with the probabilistic forecasts.