Prediction-based ripple reduction in Direct Torque Control of an induction machine

In this paper, a new prediction technique is presented in order to minimize both torque and flux ripple in direct torque control (DTC) of an induction machine. When the classical DTC scheme is implemented on a discrete-time base using a digital controller, the choice of the appropriate voltage vector is based on calculated values of torque and flux corresponding to the previous sampling moment. This time delay causes a large fraction of the overall ripple when the hysteresis bands are small compared to the maximum torque and flux variations during one sampling period. This paper aims to present a predictive scheme to correct for this time delay. The technique is characterized by its computational simplicity since it only makes use of the calculated effects of voltage vectors during the previous sampling intervals. The scheme does not require additional motor parameters and therefore shows the same robustness towards parameter variations as the conventional DTC scheme. The prediction scheme can be extended to compensate for an additional time delay when the sampling frequency is raised but the overall process time remains unchanged. Simulations and experimental results show the effectiveness of the proposed prediction scheme to reduce both torque and flux ripple.