Empirical mode decomposition and chaos based prediction model for wind speed oscillations

Accurate short-term prediction of wind speed is one of the critical issues faced by wind farm industry so as to plan trading strategies and managing power distribution. In this paper, we demonstrate that empirical mode decomposition (EMD) of the wind speed time series significantly improves prediction accuracy of nonlinear prediction tools. While EMD technique is used to decompose the measured wind speed time series data into its basic components called intrinsic mode functions and residue, nonlinear prediction tool is used to model and forecast each component. Prediction result of each component is summed up to reconstruct the wind speed data into its original form. The Resultant prediction of this hybrid method is compared with the new reference forecast method (NRFM) and local first order method (LFO). The comparison results demonstrate that, prediction accuracy can be remarkably improved by combining EMD and nonlinear model.