Adaptive local learning techniques for multiple-step-ahead wind speed forecasting

A massive deployment of wind energy in power systems is expected in the near future. However, a still open issue is how to integrate wind generators into existing electrical grids by limiting their side effects on network operations and control. In order to attain this objective, accurate short and medium-term wind speed forecasting is required.This paper discusses and compares a physical (white-box) model (namely a limited-area non hydrostatic model developed by the European consortium for small-scale modeling) with a family of local learning techniques (black-box) for short and medium term forecasting. Also, an original model integrating machine learning techniques with physical knowledge modeling (grey-box) is proposed.A set of experiments on real data collected from a set of meteorological sensors located in the south of Italy supports the methodological analysis and assesses the potential of the different forecasting approaches.

► We discuss and compare a physical model with a family of local learning techniques for wind power forecasting. ► We propose an original model integrating machine learning techniques with physical knowledge modeling (grey-box). ► A set of experiments on real data collected from a set of meteorological sensors supports the methodological analysis. ► Future work will extend this analysis by considering multiple and spatial time series.