Short-term wind power ramp forecasting with empirical mode decomposition based ensemble learning techniques

Wind is a clean and renewable energy source with huge potential in power generation. However, due to the intermittent nature of the wind, the power generated by wind farms fluctuates and often has large ramps, which are harmful to the power grid. This paper presents algorithms to forecast the ramps in the wind power generation. The challenges of accurate wind power ramp forecasting are addressed. Wind power ramp and power ramp rate are defined. An ensemble method composed of empirical mode decomposition (EMD), kernel ridge regression (KRR) and random vector functional link (RVFL) network is employed to forecast the wind power ramp and the ramp rate. The performance of the proposed method is evaluated by comparing with several benchmark models based on both accuracy and efficiency. Possible future research directions are also identified.