In this paper, both a baseline galloping piezoelectric energy harvester (GPEH) with a square bluff body and an improved GPEH with an impact bump stop are tested in a wind tunnel in order to determine the system damping, electrical response and limit cycle oscillation (LCO) amplitude. In the baseline GPEH, harvested voltage, LCO amplitude and damping ratio vary with wind velocity and electrical load. They all increase with increasing wind velocity under the same electrical load. Under each wind velocity, the damping ratio increases from the short circuit load, reaches a peak value at the electrical load resulting in a maximum voltage, and reduces the value at the open circuit load. The LCO amplitude shows the opposite trend compared to the damping case. It decreases as the electrical resistance load increases and reaches the minimum value when the damping ratio is highest. A resistance load of 100 kΩ yields a maximum peak power output. The impact stop is introduced to reduce bending stresses and improve the fatigue life of the baseline GPEH. The performance of the improved GPEH depends on the stop design parameters such as gap size, stop location and contact area. Comprehensive tests were conducted to investigate the effect of each parameter on the performance of the improved GPEH and an optimal bump stop configuration was determined. Compared to the expected proportional reduction in both electrical and structural responses, a maximum 70% reduction in LCO amplitude and only a maximum 20% reduction in harvested voltage are achieved in our optimal improved GPEH. The time variable and motion dependent aerodynamic forces acting on the bluff body could contribute to this. In summary, comprehensive experimental evaluations were conducted to characterize the performance of both baseline GPEHs and improved GPEHs. The baseline GPEH service life can be significantly improved by incorporating an impact bump stop. The improved GPEH design provides a practical solution to harvest electricity from wind-induced vibration.