Validation of Real-Time Solar Irradiance Simulations Over Kuwait Using WRF-Solar
Abstract
The performance of the WRF-Solar© mesoscale model is evaluated in the desert environment of Kuwait’s Shagaya Park, where various solar technologies are being deployed. WRF-Solar is configured with one large domain (D1) covering the Arabian Peninsula and a nested domain (D2) centered on Kuwait. The daily simulations focus on the day-ahead forecast, and span the whole year of 2017. The initial and boundary conditions are provided by the GFS analysis, and the aerosol information is imported from NASA’s GEOS-5 analysis. WRF-Solar predicts an absence of cloudiness 80% of the time. During such periods, the aerosol optical depth (AOD) predicted by GEOS-5 is found biased compared to local AERONET observations, which negatively impacts the performance of irradiance predictions, particularly for direct irradiance and under high-AOD situations. The impact of the horizontal grid spacing is evaluated. During cloudless conditions, the forecasts obtained for D1 and D2 are found practically identical. During cloudy conditions, however, mixed results are obtained. D2’s results do offer improvement over those from D1 in some cases (depending on type of cloudiness), but not in general. By comparison with high-quality local irradiance observations, it is found that many situations evaluated as cloudless by WRF-Solar are actually cloudy, or vice versa. The performance of the Fast All-sky Radiation model for Solar applications (FARMS), now incorporated into WRF-Solar, is evaluated here in a desert environment for the first time and compared to that of the more established RRTMG model. The two models satisfactorily agree under cloudless conditions, but not always under cloudy conditions. Depending on cloud conditions, one or the other might be closer to observations. In general, neither model is able to forecast the variations of direct irradiance with acceptable accuracy under clouds, which can be a concern for concentrating solar applications.