The dynamic response of fuel cell vehicle is greatly affected by the pressure of reactants. Besides, the pressure difference between anode and cathode will also cause mechanical damage to proton exchange membrane. For maintaining the relative stability of anode pressure, this study proposes a decentralized model predictive controller (DMPC) to control the anodic supply system composed of a feeding and returning ejector assembly. Considering the important influence of load current on the system, the piecewise linearization approach and state space with current-induced disturbance compensation are comparatively analyzed. Then, an innovative switching strategy is proposed to prevent frequent switching of the sub-model-based controllers and to ensure the most appropriate predictive model is applied. Finally, simulation results demonstrate the better stability and robustness of the proposed control schemes compared with the traditional proportion integration differentiation controller under the step load current, variable target and purge disturbance conditions. In particular, in the case of the DC bus load current of a fuel cell hybrid vehicle, the DMPC controller with current-induced disturbance compensation has better stability and target tracking performance with an average error of 0.15 kPa and root mean square error of 1.07 kPa.