Wind speed forecasting approach based on Singular Spectrum Analysis and Adaptive Neuro Fuzzy Inference System. (October 2018)

• Record Type: Journal Article
• Authors: Moreno, Sinvaldo Rodrigues; dos Santos Coelho, Leandro
• Is Part Of: Renewable energy. Volume 126(2018)
• Page Start: 736
• Page End: 754
• Publication Date: 2018-10
• Subjects: Wind speed -- Hybrid forecasting approach -- Wind farm -- Machine learning -- Singular Spectrum Analysis -- Neuro fuzzy inference system -- Time series forecasting; Renewable energy sources -- Periodicals; Power resources -- Periodicals; Énergies renouvelables -- Périodiques; Ressources énergétiques -- Périodiques; 333.794
• DOI: 10.1016/j.renene.2017.11.089
• Languages: English
• ISSNs: 0960-1481

Abstract

Abstract: As a promising renewable energy source, wind power has environmental benefits, as well as economic and social ones. Due these characteristics, wind farm has grown fast in the last five years, and in some countries, it has already surpassed conventional sources, such as hydro and coal plants. However, owing to the uncertainty of wind speed, it is essential to build an accurate forecasting model for large-scale wind power penetration. This study proposes a hybrid approach that combines the Singular Spectrum Analysis (SSA), which rarely presents application in literature on wind speed forecasting, and a Computing Natural paradigm called Adaptive Neuro Fuzzy Inference System (ANFIS). The SSA decomposes the original wind speed into various components, so these components are pre-processed regarding to the level of original wind series information remained. The main components selected to reconstruct the original series have in their structure the information about trend and harmonic components. The final remaining components grouped are labeled as noise. The ANFIS model uses these two information to construct the model applied to forecasting the next wind speed value. In this way, the hybrid model can learn the trend and the harmonic structure of the wind time series. Experimental results show that prediction errors are significantly reduced using the proposed technique to perform 10min one-step-ahead and k -step-ahead wind speed forecast. Highlights: An hybrid SSA-ANFIS-FCM approach for wind speed forecasting is proposed. The hybridization of SSA and ANFIS-FCM is an original idea on context of Machine Learning applied to wind speed forecast. The model presents excellent results with stable performance.