Hybrid power systems (HPS) are becoming increasingly important as they leverage complementary features of heterogenous power sources and energy storage to achieve efficient and clean power generation. The ship-board integrated power systems (IPS) and hybrid land vehicles for military applications are representative examples of HPS. The requirements of survivability necessitates real-time failure mode power management and system reconfiguration to deal with unpredictable failures. The nonlinear HPS dynamics, stringent safety constraints and real-time implementation requirements make existing receding horizon control (RHC) computationally prohibitive, given the long prediction horizon and high state/input dimensions. In this paper, we propose a novel approach using reference governor to solve the reconfiguration problem, where the total power demand is governed to enforce constraints. The computational efficiency and load tracking performance of the proposed method as compared to the RHC framework are illustrated using the HPS model as a case study along with the experimental results on a scaled test bed of the HPS.