An online short-term wind power prediction considering wind speed correction and error interval evaluation

In this paper, a rolling ultra-short term wind power prediction (WPP) based on an online sequential extreme learning machine (OS-ELM) algorithm is presented. Two issues are specially considered to improve the wind power forecasting effect: wind speed sequence correction and error interval evaluation. Three independent OS-ELM based neural networks are accordingly designed: the single wind turbine generator model, the wind speed correction model and the error interval evaluation model, in which the OS-ELM´s fast learning speed characteristics are well utilized. Case studies on a real off-shore wind farm in China with two years´ history data prepared are performed to verify the effectiveness of the proposed method.