Pascal Meyer
In the domain of electronic devices and especially desktop peripherals, there is an industrial trend which consists in removing the cables that pollute our domestic and professional environments. In this sense, wireless communication protocols are already massively widespread while the power supplies still use wires or batteries. To address this problem, alternative solutions must be investigated such as contactless energy transfer (CET). In a broad sense, CET is a process that allows to bring electrical energy from one point to another through a given medium (generally air or vacuum) and at a certain distance. Inductive CET means that the intermediate form of energy is the magnetic induction, generated from primary coils excited by high-frequency alternating currents and collected in secondary coils by induced voltages. Most of existing approaches to design CET systems are applicable to only single applications and do not include an optimization method. For this reason, the present thesis focuses on the modeling, design and optimization of inductive CET systems. Using the coreless transformer as the central part of CET systems, an equivalent electric model is derived from the theory of conventional transformers. The absence of ferrite core gives rise to a specific characteristic, which is to have large leakage inductances compared to the main one. In order to circumvent this issue, using a high frequency together with a resonant circuit allow to enhance the effect of the mutual inductance and to transfer power with an excellent efficiency. Different parts of the coreless transformer are addressed separately. First, an accurate modeling of DC resistances, self and mutual inductances is proposed. Then, the equivalent electric circuit is resolved and the different compensation topologies for the resonant circuit are discussed. Finally, the AC resistance is computed using a 2D finite element modeling that takes into account the skin and proximity effects in the conductors. So as to exploit optimally FEM simulations, a complete output mapping together with a specific interpolation strategy are implemented, giving access to the AC resistance evaluation in a very short time. As a result, all the models are implemented in a way that makes them highly adaptable and low-consuming in term of computing resources. Then a sensitivity analyzis is performed in order to restrict the variation range of different parameters and to provide a general and intuitive understanding of inductive CET. After that, an optimization method using genetic algorithms (GAs) is presented. The main advantage of GAs is that the number of free parameters does not change the complexity of the algorithm. They are very efficient when a lot of free parameters are involved and for optimizations where the computing time is a key factor. As existing GAs failed to converge properly for different tested CET problems, a new one is developed, that allows to optimize two objective functions in the same time. It is thus a multiobjective genetic algorithm (MOGA) and has been successfully applied to the design of different CET systems. Finally, in order to validate the models and optimization methods proposed along the thesis, several prototypes are built, measured and tested. Notably, a CET table that allows to supply simultaneously different peripherals is fabricated. By analyzing in real time the current amplitude in the primary coils, an efficient sensorless detection of the peripherals is implemented. Digital control techniques have enabled the autonomous management of the detection and the local activation of the table. These results contribute to the future development of robust and efficient CET tables.
EPFL2012
