Power management in all-electric powertrains has a significant potential to optimally handle the limited energy and power density of electric power sources. Situation-based power management strategies (SB-PMSs), defining optimized solutions related to specific vehicle situations, offer the ability to reduce computational requirements and enhance the solution optimality of simple rule-based algorithms. Moreover, the local optimality of SB-PMSs can be addressed by considering online optimization of the situated solutions for limited horizons. This paper presents a novel PMSs using model predictive control (MPC) to define optimal control strategies based on situated solutions for fuel cell hybrid vehicles. Vehicle states are defined in terms of multiple characteristic variables and power management decisions are optimized offline for each vehicle states. Prediction of vehicle states is conducted using statistical predictive model based on state transitions in a number of driving cycles. Preoptimized solutions related to predicted states are iterated online to achieve better optimality over the look-ahead horizon. Results analysis from online testing revealed the ability of SB-MPC to improve the optimality of situation-based solutions and hence reduce total energy cost in different driving cycles.