Abstract

In medium-voltage drives, multistep model predictive control (MPC) can lower the total harmonic distortion of the stator currents and thereby reduce losses and improve efficiency. However, from the point of view of implementation, there is still uncertainty as to whether transients have a major adverse effect on achieving this improved steady-state performance. The time-varying nature of machine drives, initialization of the optimization process, and limited computational resources are identified as key factors. This paper analyzes the link between these key factors in detail, thus a suitable reformulation and selective initialization approach is designed to enable the deterministic use of multistep finite-control-set MPC irrespective of the drive system conditions. Guidelines to select the prediction horizon, weighting factor, and minimum switching frequency considering the control platform limitations are presented. The significant impacts of transients on the design and experimental validation, covering several drive conditions, are evaluated in a scaled-down three-level induction machine drive switching at 350 Hz. This paper is accompanied by a video demonstrating the real-time implementation of multistep MPC.