Predictive Approach to Increase Efficiency and Reduce Switching Losses on Matrix Converters

The matrix converter stands as an alternative in power conversion. It has no energy storage devices, performing the energy conversion by directly connecting input with output phases through bidirectional switches based on power semiconductors, allowing high-frequency operation. For this reason, it is known as the all-silicon power converter, featuring reduced size and weight. Forced commutations of the high number of semiconductors cause switching losses that reduce the efficiency of the system and imply the use of large heat sinks. This paper presents a novel method to reduce switching losses based on predictive control. The idea is to predict switching losses for every valid switching state of the converter, if applied during the next sampling time, and then, select the optimum state based on an evaluation criterion. The proposed strategy was experimentally tested on an 18-kVA matrix converter driving an 11-kW induction machine, reducing energy losses and increasing efficiency up to 3% compared to the basic strategy. As a consequence, the converter misuses less energy and requires smaller heat sinks.