An Ultra-Short-Term Wind Power Forecasting Method Based on Data-Physical Hybrid-Driven Model

The ultra-short-term wind power forecasting provides key technical support for unit control and economic dispatch of power systems. To fully exploit the advantages of the data-driven model and the physically driven model in wind power prediction, a data-physical hybrid-driven ultra-short-term wind power forecasting model is proposed. For the time point when wind speed fluctuates violently, the wind-power conversion model is constructed according to the inertial operation characteristics of the unit, which is a physically driven prediction model. Aiming at the smooth power output stage, an ultra-short-term wind power forecasting framework based on a temporal attention network is proposed, which is a data-driven model. Based on the wind speed fluctuation characteristics, a switching mechanism between the physical-driven model and the data-driven model is established. Experimental validation was conducted at a wind farm in Inner Mongolia, China, and the root-mean-square error of the proposed method was less than 13% for prediction in the fourth hour of the future. The results of the study show that the proposed method can effectively improve ultra-short-term wind power prediction accuracy.