Data Enrichment as a Method of Data Preprocessing to Enhance Short-Term Wind Power Forecasting

Wind power forecasting involves data preprocessing and modeling. In pursuit of better performance, most previous studies focused on creating various wind models, but few have been published with an emphasis new types methods. Effective techniques the fusion physical nature called upon as potential future research directions in recent reviews this area. Data enrichment a method has widely applied to problems consumer universe not seen application This study proposes addition existing library methods improve performance. A methodological framework is developed four executable steps: add error features weather prediction sources, at neighboring nodes, time series complementary sources. The proposed takes full advantage multiple commercially available sources continuity wind. It can cooperate any models that inputs. controlled experiments three actual individual farms verified effectiveness method: normalized root mean square (NRMSE) day-ahead forecast XGBoost LSTM 11% 27% lower than without enrichment. future, variations be further explored promising direction enhancing short-term