A novel combined forecasting system based on advanced optimization algorithm - A study on optimal interval prediction of wind speed. (1st February 2023)

• Record Type: Journal Article
• Authors: Li, Jingrui; Wang, Jiyang; Li, Zhiwu
• Is Part Of: Energy. Volume 264(2023)
• Publication Date: 2023-02-01
• Subjects: Wind speed forecasting -- Data preprocessing technique -- Combined system -- Whale optimization algorithm; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2022.126179
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: Wind speed forecasting is becoming increasingly crucial for an environmentally friendly and sustainable economy because of the renewability and benefits of wind energy. Currently, many scholars have proposed various approaches to forecasting wind speed; however, due to the limitations of point prediction and drawbacks of traditional individual methods, it is challenging to acquire satisfactory results. Compared with the previous articles, this paper combines a data preprocessing strategy by fuzzification, interval estimation, and advanced optimization methods to enhance the accuracy of wind speed forecasting. Moreover, the weight coefficient allocated by a multi-objective algorithm is proved to reach Pareto optimal solution in theory. Based on the comparative experiments and discussion of the performance of the developed combined forecasting system and other control groups, it is revealed that the combined system not only outperforms for predicting wind speed with higher accuracy and stability but also enables a valid assessment of uncertainty than the former studies. It can be convinced that the developed predicting system is an appropriate and efficient tool for further practical applications in energy systems. Highlights: Designing a novel integrated forecasting system to acquire more accurate result. A data decomposition strategy is developed to decrease dimensionality. The interval estimation efficiently solves the cons of point predicting methods. A multi-objective algorithm is theoretically proved to achieve the Pareto optimum.