A Graphical Performance Tool for Design and Comparative Assessment of Predictive Torque Control Methods in Motor Drive Applications

P. G. Ipoum Ngome,Jin Tao,D. L. Mon-Nzongo,J. Tang,Rodolfo C. C. Flesch,J. Song-Manguelle
DOI: https://doi.org/10.1109/pedg.2019.8807539
2019-01-01
Abstract:In this paper, a graphical performance tool (GPT) is proposed to select, among several existing Predictive Torque Control (PTC) methods, an appropriate approach to drive a motor for given specifications. To achieve a fair trade-off evaluation, switching frequency, current, and torque total demand distortions have been considered depending on PTC parameters. Since, in the conventional PTC method, torque and flux weighting factors (WFs) strongly influence the behavior of the motor. Thus, to avoid an unfair comparison due to a non-optimal tuning, the proposed GPT is used to show each performance criterion when parameters and operating points change. Simulation results of three operating points show that there is only one sharing common zone where current harmonics and torque ripples are near zero. This area has been named the optimal tuning zone for PTC controllers and is used to select their parameter values. After obtaining these desired parameter values, a trade-off performance evaluation is made over a wide operating range to compare different PTC controllers with and without WF that are proposed in the literature.
What problem does this paper attempt to address?