Abstract: The forecasting of multi\|step ramp is an effective way to ensure security and power quality of power grid with high penetration of wind energy into power grid. A forecasting model of multi\|step wind power ramp based on catastrophe theory is proposed. Firstly, wind power ramp events are regarded as catastrophe phenomena, and the meso\|scale meteorological data is processed by using correlation analysis, principal component analysis and linear weighted accumulation, so as to determine relevant variables of catastrophe ramp models. Thus catastrophe forecasting models for up\|ramps and down\|ramps are built respectively. The multi\|objective optimization problem is solved by increasing the forecasting step and decreasing the forecasting error, by which the parameters are trained and the multi\|step wind power ramp is forecasted. The simulation results show that the proposed method can effectively forecast wind power ramp events, has larger forecasting step and higher accuracy by comparing with that of the statistical method.