Wind power is stochastic and is difficult to reflect its characteristics only by its expectation. Consequently the power grid risk is difficult to analyze. A new model based on the quantile regression analysis theory is proposed， in which a regression function is selected self-adaptively through support vector machine in each wind power fluctuation interval， and the interior point method is adopted to solve this model. The future wind power fluctuation interval analysis is realized. At last， the proposed method is used to analyze the output power of a wind farm in Yantai area power grid. The results show that the proposed method can give a better all-around description of wind power uncertainty， and provides a basis for the risk decision-making in the power system dispatch and control.