Abstract: In extreme weather conditions,the power of wind power could change drastically within a short-time scale,giving rise to a high-risk ramp event,which seriously threatens the safe and stable operation of the power system. Carrying out the requirement assessment of the ramping reserve will help reduce the adverse effects of wind power output fluctuations and wind power forecast errors on the operation of the power grid. For this reason,this paper proposes a data-driven method for evaluating the requirement for ramping reserve. First,the actual wind power data and the day-ahead forecast data form a multivariate time series,and the accuracy of the forecast results is improved based on the gate recurrent unit（GRU） model.Furthermore, a non-parametric kernel density estimation method is used to estimate the confidence interval of the wind power forecast error,and the wind power forecast interval under a given confidence interval is obtained. Finally,according to the interval prediction results,the ramping event is predicted and the ramping feature is extracted, the ramping reserve requirement evaluation model is established,and the ramping reserve capacity requirement is estimated. An example test was carried out based on the data of a provincial power grid in Northwest China,which has verified the effectiveness of the method in this paper.