Sensitivity of predictive controllers to parameter variation in five-phase induction motor drives

Model predictive control techniques have been recently proposed as a viable control alternative for power converters and electrical drives. The good current tracking, flexible control design or reduced switching losses are some of the benefits that explain the recently increased attention on finite-control-set model predictive control. The performance of the predictive model of the drive, which is the core of the predictive control, highly depends on the parameters of the real system. In this context, most research works assume good agreement between electrical parameters of the predictive model and the real machine, on the basis of nominal values. Nevertheless, this is far from being a real assumption, where non-modeled variables (i.e. the temperature, the magnetic saturation or the deep-bar effect) produce a detuning effect between the real system and its model, which can harm the control performance. The influence of parameter variations on the predictive control has barely been investigated in recent research works, where only conventional three-phase power converter configurations and permanent magnet drives have been taken into account. However, there is a lack of knowledge when different technologies like induction machines or multiphase drives are considered. It is worth highlighting the interest of the industry in induction motors as a mature technology or in multiphase drives as a promising alternative in applications where high overall system reliability and reduction in the total power per phase are required. This paper attempts to fill this gap by examining the impact of parameters mismatch on the finite-control-set predictive control performance of a five-phase induction motor drive, one of the multiphase electromechanical conversion systems with greatest impact in the research community. An exhaustive experimental sensitivity analysis of the close loop system performance based on more than three hundred trials in a test bench is presented. (C) 2017 Elsevier Ltd. All rights reserved.