A novel interval forecasting system for uncertainty modeling based on multi-input multi-output theory: A case study on modern wind stations. (January 2021)

• Record Type: Journal Article
• Authors: Liu, Tongxiang; Zhao, Qiujun; Wang, Jianzhou; Gao, Yuyang
• Is Part Of: Renewable energy. Volume 163(2021)
• Page Start: 88
• Page End: 104
• Publication Date: 2021-01
• Subjects: Interval forecasting -- Wind speed -- Data preprocessing -- Feature selection -- Multi-objective optimization algorithm; Renewable energy sources -- Periodicals; Power resources -- Periodicals; Énergies renouvelables -- Périodiques; Ressources énergétiques -- Périodiques; 333.794
• DOI: 10.1016/j.renene.2020.08.139
• Languages: English
• ISSNs: 0960-1481

Abstract

Abstract: With the growing demand for a clean energy source, wind power is drawing increasing attention. However, its intermittence and fluctuation set strict restrictions on its development and applications. Although a vast amount of research has been conducted on this subject, studies have failed to characterize the uncertainties of the growing intervals and have focus only on point prediction. Therefore, this paper proposes an interval prediction system that can effectively avoid the drawbacks of point forecasting. The system is composed of five units: a preprocessing unit, a feature selection unit, an optimization unit, a forecasting unit, and a result evaluation unit. The preprocessing unit, along with the feature selection unit, is applied to obtain the ideal input data. Then, the forecasting unit, whose key parameters are updated by the optimization unit, is used for interval prediction. The experimental results obtained from various evaluation metrics show that the accuracy of the developed system exceeds that of benchmark methods, and also confirm the possibility of applying the proposed method in the effective utilization of wind energy. Highlights: Drawbacks of point forecasting are overcome by interval forecasting. Ideal model input is obtained by phase space reconstruction approach. Pareto-archived evolution strategy is applied to improve the forecasting ability. Multi-input multi-output forecasting method is used for higher efficiency. Various valuation metrics are utilized to verify the effectiveness of the proposed model.