A new intelligent method based on combination of VMD and ELM for short term wind power forecasting

作者：Abdoos, Ali Akbar^*

来源：Neurocomputing, 2016, 203: 111-120.

DOI：10.1016/j.neucom.2016.03.054

• Extreme Learning Machine (ELM)
• Pattern recognition
• Variational Mode Decomposition (VMD)
• Wind power

Wind power plants are clean energy resources and are increasingly utilizing in power systems. The wind power forecasting is complicated due to the volatile nature of wind speed. Since, accurate wind power forecasting is crucial for wind power plants, a new hybrid pattern recognition method is proposed in this study for short term wind power forecasting. The time series of produced power is estimated through combination of three main steps including: pre-processing, feature selection and regression steps. In the first step, the time series of wind power is decomposed into different modes by using Variational Mode Decomposition (VMD) technique. These modes are then used to construct training patterns and forecasted outputs. In the second step, to eliminate redundant properties, the feature selection method based on Gram-Schmidt Orthogonalization (GSO) is applied on potential candidates. In the last step, Extreme Learning Machines (ELMS) as efficient and fast regression tools are trained using subsets of selected features. Eventually, the power generated by the wind farm is estimated by summing all the predicted modes values. The performance of the proposed wind power forecaster is evaluated using real data collected from two wind farms located in Sotavento Galicia in Spain and Texas in US. The obtained results justify the superiority of the proposed method in accurate forecasting and saving computational time comparing to some previously reported methods.

• 出版日期2016-8-26