Hybrid Model Based on Isolated Forest and WT-Catboost for Wind Power Forecasting

A hybrid model based on isolated forests (IF), wavelet transform (WT), and categorical boosting (CatBoost) is proposed for wind energy prediction. In this prediction method, IF is first utilized to catch outliers in the raw wind power time series. Afterward, WT is applied to decompose the wind power sequence processed by outlier detection into several frequencies with more pleasing silhouettes. Finally, CatBoost is employed to learn the nonlinear attributes of these frequencies. The hybrid model is demonstrated on actual wind farm operation data and shows its effectiveness compared with other methods.