Modeling for chaotic time series based on linear and nonlinear framework: Application to wind speed forecasting

Abstract

Wind-speed forecasting plays a crucial part in improving the operational efficiency of wind power generation. However, accurate forecasts are difficult owing to the uncertainty of the wind speed. Although numerous investigations of wind-speed forecasting have been performed, many of the previous studies used wind-speed data directly to make forecasts, which were rarely based on the structural characteristics of the data. Therefore, in this study, a hybrid linear-nonlinear modeling method based on the chaos theory was successfully employed to capture the linear and nonlinear factors hidden in chaotic time series. Before the forecast, the noise in the data was removed using a decomposition algorithm. Then, through the phase-space reconstruction, the one-dimensional time series were extended to the multi-dimensional space to determine the utilization form of the data. Finally, Holt's exponential smoothing based on the firefly optimization algorithm and support vector regression were combined to predict the wind speed. The experimental results show that the proposed model is not only better than the comparison models but also has great application potential in the wind power generation system.