Energy harvesting devices have been gaining increasing interests, especially in the areas of internet of things (IoTs) and sensor networks. Due to the broadband and random nature of typical vibration energy sources available in the environment, significant amount of research efforts have been put into the bandwidth broadening of the energy harvesters. Utilization of spring nonlinearity has been one of the most studied subject in that regard. In this work, response of an energy harvesting device with curved-beam springs to colored-noise vibration is studied numerically, based on stochastic differential equations. The harvester considered in this study is an electrostatic type with electrets and a balanced comb-drive configuration. The study mainly focuses on the effect of the beam height to the harvested power. The results show that curved-beam springs can increase the harvested electric power by 52% (2.69 mW versus 1.77 mW) in comparison to straight-beam springs of the same dimensions. Buckling-induced rapid snapping of the curved beams is attributed to such a power increase.