Physical and statistical downscaling for wind power forecasting

The current paper investigates the role and analyses the specific performance of the physical and statistical downscaling techniques for wind power forecasting (WPF). In doing this, a methodology already developed by the authors is compared with a machine learning based methodology. A specific focus on the new challenges and trends in WPF is addressed. Experimental results from a real-world case study permitted to evaluate critically the performance of physical and statistical techniques in performing downscaling of atmospheric model predictions such as the ones provided by the European Centre for Medium-Range Forecasting (ECWMF).