Wind Power Forecasting in the Absence of Historical Data

Abstract

Wind power forecasting (WPF) is essential in order to operate power systems with increased wind power penetration in an efficient and secure way. WPF techniques using statistical modeling require the availability of historical data. This paper presents a WPF tool that is self-constructed and self-adaptive. It can, therefore, be implemented from the beginning of a wind farm operation, with very limited or no historical information and it can adapt automatically to any future wind farm enhancements or retrofits. The algorithms converge after a few days of operation, as shown by the application of the method in a real wind farm case-study. The results are compared with a state-of-art wind power prediction model.