Tuning of predictive controllers for five-phase drives

Abstract

Model Predictive Control is gaining attention as a versatile tool for multi-phase drives. In the direct digital control configuration the predictive controller can use a finite number of control actions. This allows the use of exhaustive optimization to derive the control action. Using a cost function it is possible to consider several control objectives, such as current, torque or flux tracking. Also, different additional criteria such as switching frequency, harmonic distortion or torque ripple can be considered. Tuning of the controller involves assigning a value to weighting factors used by the cost function. This has been studied in detail for predictive torque controllers. In this paper three design strategies for designing predictive current controllers are compared. The first uses fixed weighting factors, the second eliminates weighting factors by using a ranking based selection and the last considers a local optimization of the weighting factors.