Hybrid system based on a multi-objective optimization and kernel approximation for multi-scale wind speed forecasting

Researching wind speed forecasting is increasingly important given the growing importance of wind energy in power system operation. Precise and reliable wind speed forecasting methods can effectively enhance the operational stability of power grids and improve their economic and environmental benefits. In recent decades, numerous methods have been proposed to reduce the forecasting errors caused by the non-stationary and non-linear properties of wind speed data. However, some traditional forecasting methods are restricted by the limitations of single prediction methods and do not consider the necessity of applying data processing technology and optimization algorithms, which leads to poor prediction accuracy. To overcome the limitations of traditional forecasting methods and provide more precise and stable wind speed prediction results, an advanced hybrid prediction system based on data reconstruction and kernel approximation (random Fourier mapping) is proposed in this paper. The proposed system successfully maximizes the forecasting capabilities of the component methods and effectively improves the wind speed prediction performance. Based on 10-min and 30-min data collected from Shandong province, China, the simulation results indicate that the forecasted values obtained using the proposed strategy are considerably better than those obtained using the compared strategies, which is beneficial for electrical energy conversion and power grid security.