Predictive control of a direct series resonant converter with active output voltage compensation

Modern high power supplies are based on resonant converters in order to use high frequency reactive elements (for reduced size) without sacrificing converter efficiency. In an effort to achieve compact high power supplies, direct power converter topologies have been considered, since these are mainly characterised by their high power density. A direct resonant converter topology combines matrix converters with conventional resonant converters. This work focused on achieving high quality input/output power and high efficiency. Thus, this thesis presents the research on the control of a direct series resonant converter. Since the resonant converter allows a sinusoidal high frequency output current to be generated, zero current switching (ZCS) was considered to reduce the power losses. Hence, since the converter is switched at every zero crossing of the output current (fixed period), model predictive control was considered. Different predictive approaches for controlling the input and output currents were developed and analysed, however, owing to the characteristics of the topology, these strategies resulted in a suboptimal control. Therefore, in order to improve the input and output qualities (reduce distortion), an output voltage compensation strategy was proposed. This compensation approach is based on adding an H-bridge converter in series between the matrix converter and the resonant tank. This converter improves the voltage applied to the resonant tank, thus, reducing the distortion at the output and, as a consequence, also the distortion at the input. The H-bridge converter utilises only a small capacitor on the dc side in comparison with conventional resonant converters and operates at a low voltage. Simulations were carried out using MATLAB/Simulink and an experimental prototype was built to validate the strategies proposed, achieving a reduction of the input current THD from 4.4% to 2.7%, a reduction of the output current distortion of approximately 40% and an analytically derived efficiency of 89.5%.