Abstract: Ultra-short-term wind power forecasting provides important guidance for unit control and energy economic dispatching. In order to weaken the influence of wind energy volatility on the prediction accuracy of ultra-short-term wind power, a switching output mechanism with wind speed information is proposed to analyze the characteristics of wind speed fluctuation based on the physical model of wind speed and power. For the time point when the fluctuation characteristic exceeds the threshold, according to the inertia operation characteristics of the unit, a wind speed-power conversion model under different wind speed variation scenarios is constructed. For the gentle output stage, considering the inherent limitation of a single model in time series modeling, an ultra-short-term prediction framework under different operating conditions is proposed. The proposed prediction method is applied to a wind farm in Northeast China for example verification. Compared with the prediction results of single model, the proposed method can be utilized to reduce the root mean square error by 1.31%~4.12% on the basis of 13.58%~16.39%. The results show that the proposed method can be utilized to effectively improve the accuracy of ultra-short-term wind power prediction.