A Monte Carlo based solar radiation forecastability estimation

Based on the reported literature and commonly used metrics in realm of solar forecasting, a new methodology is developed for estimating metric called forecastability (F). It reveals extent to which radiation time series can be forecasted provides crucial context judging inherent difficulty associated with particular forecast situation. Unlike score given by standard smart persistence model, F bounded between 0% 100% easier interpret, hence making comparisons forecasting studies more consistent. This approach uses Monte Carlo method estimates from error RMSE predictor. measured at six very different locations (with optimized clear sky model) meteorological point view, it shown that varies 25.5% 68.2% exists link errors obtained machine learning prediction methods. The proposed validated 3 parameters may affect estimation (time horizon, temporal granularity, components) 50 relative McClear web service central archive Baseline Surface Radiation Network.