The dynamic programming is applied to model predictive control configuration, and a predictive control mathematics model is established for the fuel economy of plug-in hybrid electric vehicles (HEVs) based on spatial-domain. A global optimal method for plug-in HEVs is studied with dynamic programming, and the results indicate that the state of charge (SOC) is reduced slowly from the maximum at a trip’s initiation to the low threshold at the destination. The theoretical SOC slope is proposed to be taken as the reference slope of model predictive control, and the theoretical reference contrail of SOC is modified according to the especial driving modes in future. The results show that the predictive control effect of this method is approached to the calculation result of global optimal and fuel consumption is obviously reduced.