Analysis and optimisation of energy-harvesting micro-generator systems
This thesis investigates electro-mechanical generator systems which harvest energy from their environment. Such systems are needed to create maintenance-free sensor nodes for use in autonomous wireless sensor networks which have applications in health monitoring. Inertial microgenerators are investigated in detail. Inertial micro-generators produce electrical energy when subjected to acceleration. Three architectures of inertial micro-generator were identi ed as suitable for implementation using MEMS technology. Two of these architectures, both resonant in nature, have been reported in the existing literature. The third, a non-resonant type, is new. The architectures have been analysed and compared within a common framework, based on sinusoidal driving signals and a common set of normalisation factors. A simple procedure for the design process of micro-generators was established. Within the analytical framework, the non-resonant generator achieved the highest power density of the three architectures when powered from large amplitude motion, making it the most suitable for powering implanted medical devices. Comparing the performance of the three architectures on measured acceleration data from human subjects showed that this result is more widely applicable than for simple sinusoidal driving motions. Bio-compatibility of microgenerator systems has not been addressed in this work. The non-resonant architecture was investigated in more detail. To maximise nal energy yield taking into account interactions between various generator sub-systems, a uni ed simulation of a power supply system built around the non-resonant generator was developed and includes detailed models of the required semiconductor devices. A prototype generator was used to verify the behaviour of the model. The concept of system effectiveness was introduced which accounts for both the ef ciency of the energy conversion stages and the success in coupling energy from what is assumed to be a large and free original source.