Model Predictive Controller Design and Application of Permanent Magnet Machines


Student thesis: Doctoral Thesis

View graph of relations


Related Research Unit(s)


Awarding Institution
Award date20 Jul 2021


Permanent magnet (PM) machine drive systems have been widely applied in the industries because of their high efficiency and compact size. Moreover, the PM machines drive systems with diverse topologies become more and more attractive in recent years, because of the high robustness and efficiency. However, with the diverse topologies, the harmonic currents would increase the total harmonic distortion and cause additional noises. In this thesis, the advanced model predictive controllers in the PM machines with diverse topologies, including the multiphase PM machines and open-winding PM machines are discussed and investigated.

Firstly, the advantages of the model predictive control strategies applied in the PM machines with diverse topologies are proposed. Since there are additional harmonic subspaces in the PM machines with diverse topologies, the tuning process of PI controllers would be difficult and the coupling components between different subspaces would worse the performance of PI controllers. Thus, the model predictive control concept, a multi-input and multi-output control strategy, would be preferred in the application of PM machines with diverse topologies.

Secondly, the predictive model with active compensation strategy is proposed. The inaccurate predictive model and the parameter mismatches would cause large predictive errors and affect the performance of model predictive model. The influence of possible parameter mismatches and the inaccurate predictive model is studied. Then, these predictive errors are compensated by the slide mode observer, the incremental predictive model, and the extended state observer.

Thirdly, the direct modulation pattern control strategy is discussed. The control sets are extended from voltage vectors to the modulation patterns. The semi-controlled modulation patterns are proposed to provide precise torque output with low switching frequency, where the voltage vectors in the multiphase PM machines are reconstructed. Moreover, a two-step optimization process is discussed to reduce the control targets in the cost function.

Finally, another fundamental torque-controlled modulation pattern strategy is proposed to improve the performance of PM machines. The modulation patterns could provide precise fundamental torque output and have different components in the other subspaces. Both the harmonic currents suppression and zero common mode voltage could be realized with this concept.

    Research areas

  • PMSM, Model Predictive Control