Abstract
A gust forecasting based on ARX (Autoregressive with Exogenous inputs) model which uses numerical weather prediction and on-site measurement as inputs was proposed and the model parameters were estimated by using non-parametric regression with forgetting factors. The prediction accuracy by using dynamically adaptive model was improved compared to the conventional static MOS (Model Output Statistics) model. It was also shown that the prediction accuracy of maximum gust improves by utilizing the numerical weather prediction with higher horizontal resolution. The predictability of the gust with the maximum wind speed larger than 15m/s was evaluated by using ROC (Receiver Operating Characteristic) curve and AUC (Area Under the Curve) and was improved by the proposed method.
