Abstract: When there exist disturbances,e.g.,the fractionator feed and components or large load,it is difficult for conventional controller to exhibit better performance.This paper presents a nonlinear predictive control algorithm for distillation component via mechanism model.Aiming at the problem that components cannot be online measured,the extended Kalman filter and accessible temperature are utilized to estimate components,which is further updated by combining with laboratory analysis of the low frequency value.The simulation results show that the proposed algorithm can attain better control performance for model mismatch or feed disturbances.