Wind Power Forecasting Focused on Extreme Power System Events

Any small improvement of the wind power forecasting performance can provide additional benefits to the end-users (TSOs, wind farm operators, etc.). Several regimes can be defined based on the different wind power profiles that lead to large forecasting errors and related to specific meteorological events. The regime-switching approach gives the opportunity to predict wind power with a different predictor for each regime, reducing essentially the forecasting error. In this paper, the regime sequence is estimated by a modified ARTMAP and RBFNNs are applied as predictors. A novel adaptive learning method has been developed for the on-line learning of the applied RBFNNs. The proposed model was tested on a real wind farm and was compared with a state-of-art forecasting model.