Mesoscale models have the spinup problem in nowcasting (0 to 2 hour forecast), so extrapolation forecast has become an indispensable technical method for nowcasting. At present, the extrapolation vectors in INCA (Integrated Nowcasting through Comprehensive Analysis system) precipitation extrapolation prediction are deterministic and do not change with forecast time, which results in that the velocity and direction of precipitation movement do not change with the forecast time, which is inconsistent with the reality. It is necessary to consider the variations of extrapolation vectors with time and to make them as the disturbances of extrapolation vectors adding to deterministic extrapolation vectors. Extrapolated vector perturbations are determined by the variations of extrapolated vector with time, which is the deviation of extrapolated vectors at adjacent times. The adjacent 10minute extrapolated vector deviation during July 2017 of the INCA system is statistically calculated to represent the variation of extrapolated vectors with time. Extrapolation vectors are the current and future precipitation moving speed and direction. For easy understanding, if precipitation moves northeast, this paper stipulates that the direction of extrapolation vectors is northeast. The statistical results over the ChangjiangHuaihe Region demonstrate that the deviations are mainly in southwest and northeast. The larger the extrapolation velocities are, the greater the deviations velocities will be. Vast majority of precipitation occurs in front of troughs, so northeast extrapolation vectors account for 73%. 〖JP2〗Based on the statistical samples of extrapolated vector deviation, several extrapolated vector perturbations are generated randomly based on the distribution of extrapolated vector deviation, and several uncertain extrapolated vectors are obtained. The deterministic extrapolated prediction of INCA is changed to 0to2 hour extrapolated ensemble precipitation prediction. The ensemble forecast effects are verified by the Mean Square Error (MSE), TS score, BIAS score and Brier Skill score. The results demonstrate that the results of ensemble forecast are better than that of deterministic forecast with increasing forecast time. Therefore, the accuracy of precipitation extrapolation prediction can be improved by considering the variations of extrapolation vectors with time.〖JP〗