Wind speed forecasting based on variational mode decomposition and improved echo state network. (February 2021)

• Record Type: Journal Article
• Authors: Hu, Huanling; Wang, Lin; Tao, Rui
• Is Part Of: Renewable energy. Volume 164(2021)
• Page Start: 729
• Page End: 751
• Publication Date: 2021-02
• Subjects: Artificial intelligence -- Wind speed forecasting -- Echo state network -- Variational mode decomposition -- Differential evolution; Renewable energy sources -- Periodicals; Power resources -- Periodicals; Énergies renouvelables -- Périodiques; Ressources énergétiques -- Périodiques; 333.794
• DOI: 10.1016/j.renene.2020.09.109
• Languages: English
• ISSNs: 0960-1481

Abstract

Abstract: Accurate wind speed forecasting is conducive to power system operation, peak regulation, security analysis, and energy trading. This study proposes a hybrid model named VMD-DE-ESN incorporating variational mode decomposition (VMD) and differential evolution (DE) and echo state network (ESN) for wind speed forecasting. In the proposed model, VMD is applied to decompose the wind speed series to eliminate noise and mine the main features of original series, DE is utilized to optimize three important parameters of echo state network, and the improved ESN is used to forecast each decomposed subseries. The final forecasting results are obtained by summarizing the forecasting results of all subseries. To validate the accuracy and stability of the proposed model, four wind speed datasets collected from Sotavento wind farm in Galicia, northwestern Spain are used for forecasting. Mean absolute percentage errors of the proposed model in four datasets are 2.0161%, 3.4153%, 2.1544%, and 2.8478% respectively, which are much lower than those of nine comparative models. Therefore, the proposed model has satisfactory performance and is suitable for wind speed forecasting. Highlights: Propose a hybrid model (VMD-DE-ESN) based on VMD and improved ESN. VMD-DE-ESN is firstly applied to wind speed forecasting. Use mean ratio of residual to original series to determine the optimal mode number. VMD-DE-ESN outperforms nine comparative models in four wind speed datasets.