The recent rapid development of smart electronic devices, including wearable electronic devices, Internet of Things (IoT) devices, implantable medical devices and remote smart monitors, has raised great research interests on optimising battery power consumption in small-scaled and remotely operational electronic components. For the past ten years, there has been many research works in vibration-based energy harvesting, i.e., converting ambient vibration sources into electrical energy via different transduction methods. In dealing with environmental vibration sources, the most pressing issue for vibration-based energy harvesters (VEHs) from mechanical and structural perspectives is whether the device can efficiently output an optimised power level under realistic (i.e., non-stationary, time-dependent, distributed in frequency spectrum, etc.) excitation sources. Additionally, the performance of VEHs is also restricted in size, complexity of design, and power density issues. Under such design criteria, introducing nonlinearities into VEHs attracted wide attention. This research work aims to investigate the bandwidth performance of vibration-based energy harvesting by subjecting nonlinear phenomena/techniques with internal and externally induced dynamic behaviours into the systems, hence broadening the operational bandwidth and optimising the power level under a wide range of frequencies. The outcomes of this research work yield five peerreviewed journal papers and several international conference papers. The five journal papers are presented in five chapters (Chapter 3 to Chapter 7) as the main contributions of this thesis. Paper 1 presents a magnetic VEH using combined primary and parametric resonances. In order to merge the resonant regions of fundamental primary resonance and parametric resonance as one continuously operational bandwidth, the motion limiter is utilised to induce external hardening effects that aim to eliminate the off-resonance regime between the two resonances. Theoretical investigations, ...