Direct Model Predictive Control strategy for multi-phase thyristor matrix converters

Abstract

Model Predictive Control (MPC) is a very powerful and increasing popular control method. Due to increasing calculation power of state-of-the-art control hardware platforms and computationally efficient MPC approaches, MPC is already feasible for processes with small time constants as they are common in converter and drive control. In this paper, a 27-to-3-phase thyristor matrix converter (also known as cycloconverter) is considered. Such a converter can be controlled with the standard control angle approach. However the full available system performance cannot be utilized using this standard method, since each of the grid phases is separately controlled. Compared to the standard method, the MPC as a multi-variable control algorithm considers the whole coupled overall system. Due to this, all 273 = 19683 switching states are considered. Besides the proof of the function of a Direct MPC for the 27-to-3-phase thyristor matrix converter, this paper also demonstrates that the commutation frequency can be distinctly reduced by the use of the MPC. Since the switching losses mainly depend on the switching frequency, the converter losses will be significantly reduced by this method.