A Hybrid Wind Power Forecasting Model with XGBoost, Data Preprocessing Considering Different NWPs

In recent years, wind energy has become a competitively priced source of around the world, which created increasing challenges for system operators. Accurate power generation forecasting plays an important role in systems to improve reliable and efficient operation. Therefore, numerous artificial intelligent methods such as machine learning deep have been considered solutions accurate forecasts. addition deterministic forecasting, probabilistic becomes more important, because it indicates level uncertainty. this paper, hybrid model considering different Numerical Weather Prediction (NWP) models XGBoost training is proposed short-term forecasting. The algorithm includes data preprocessing, autoencoder used reduce dimension 20 NWP ensembles. performance method investigated using historical measurements results by Taiwan Central Bureau (CWB); spot speeds from WRFD, RWRF, ensemble WEPS. Based on results, produces better accuracy among other models, reveals importance preprocessing autoencoders use or