Direct mean torque control with improved flux control

This paper presents an improved direct torque and flux control algorithm for induction machines based on the Direct Mean Torque Control. The new algorithm combines the high dynamic performance of a direct torque control method with a precise flux control. Summary Servo drive systems for high dynamic applications need a fast torque and flux control. Several methods to control torque and flux of an induction machine are well known: the field oriented control and the some direct torque control algorithms like the Direct Torque Control (DTC), the Direct Self Control (DSR). Here we use the Direct Mean Torque Control (DMTC). DMTC is a model-based, predictive method, which results in a constant switching frequency and is applicable for all kinds of induction machines, especially for servo drives with a very low leakage inductance. Under some particular conditions, the original DMTC algorithm [1], [2] did not provide the maximum dynamic performance. The maximum dynamic performance is necessary for active damping of mechanical oscillations with high resonance frequencies (up to 1kHz). The original algorithm uses one voltage vector and one zero voltage vector for torque and flux control per sampling cycle. At low speed and torque, the duration of active voltage vectors for torque control is too short to avoid a decay of the flux. Then flux supporting voltage vectors have to be chosen which may influence the torque in an adverse way. In the subsequent cycles the adverse effects of the flux supporting voltage vector concerning torque and speed has to be corrected. At the experimental set-up this leads to the excitation of unwanted torque and speed oscillations. But with the improvements presented in this paper this problems can be solved and it is possible to actively damp mechanical oscillations even with high resonance frequency [3]. 1. Direct Mean Torque Control Hysteresis controllers like the DTC can result in extremely short sampling periods if the control of an induction machine with low leakage inductance is realised by a micro processor [4]. This is especially true at low or high speed. At these operating points, the difference between an active voltage vector applied by the inverter and the EMF is large, consequently the rate of change of the stator currents is high.. To obtain a suitable torque ripple, a sampling period of less than 5 μs would be necessary with DTC at our servo machine. This would end in an unacceptable high computational burden for the controller. A significant improvement is achieved by the original DMTC, which is a model-based, predictive method, resulting in sampling periods of e.g. 125μs. DMTC calculates two switching events per sampling period. The switching events are calculated such that the torque time area under and over the torque set point value has the same size in steady state operation [1], [2]. See fig. 1: