A regional pretraining-classification-selection forecasting system for wind power point forecasting and interval forecasting

Wind power forecasting is extremely crucial for power market transactions and power system operation. Although a lot of researches have concentrated on wind power forecasting, the forecasting performances were confined and the forecasting models were only suitable for providing good performance at a few sites of investigation. To bridge these gaps, a novel regional pretraining-classification-selection wind power forecasting system is proposed in this paper based on four modules-pretraining module, classification module, point forecasting module, and interval forecasting module, which effectively improves forecasting performance and extends the applicability to different data characteristics. 10-min wind power data obtained from 20 datasets are used to verify the forecasting ability of the proposed forecasting system. The experimental analyses and discussions reveal that the proposed forecasting system is accurate and reliable for achieving high-quality wind power point and interval forecasting results. Thus, it could provide useful references for wind producers and managers in power system dispatch and operation. (C) 2021 Elsevier B.V. All rights reserved.

Automated summary

Wind power forecasting is crucial for power market transactions and system operation. The proposed system in the paper has shown accurate and reliable results in experiments, providing useful references for wind producers and managers.