Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.


Deadlock Avoidance and Resolution for Model Predictive Direct Torque Control


Thomas Burtscher

Master Thesis HS 10 (10037)

For medium voltage ac-drives, model predictive direct torque control (MPDTC) shows excellent performance regarding the switching frequency and the harmonic distortion of the torque and the stator currents. Beside all the advantages of MPDTC compared to classic control schemes (e.g. direct torque control (DTC)), the MPDTC algorithm runs occasionally into infeasible states, also called deadlocks. A particular exit strategy is currently used to resolve these deadlocks, which however causes large peaks in the short-term switching frequency and decreases the overall performance. This master thesis focuses on new methods in order to avoid these deadlocks. The proposed methods are based on terminal constraints or terminal weights, and successfully achieve a strong reduction or even a total avoidance of deadlocks. In the case of a 5-level voltage source inverter topology, there is a significant improvement of the switching frequency and the harmonic distortion of the torque and the stator currents observable, too. Large switching horizons further improve the performance of MPDTC, but they are connected to large computational burdens. To make MPDTC more efficient, the concept of branch and bound is applied, which results in a large reduction of the computational effort. The second part of this master thesis investigates certain heuristics for the branch and bound algorithm. An initial estimate of the optimal switching sequence, based on geometrical considerations, is used to find a tight upper and lower bound for the branch and bound algorithm. 3


Type of Publication:

(12)Diploma/Master Thesis

M. Morari

No Files for download available.
% Autogenerated BibTeX entry
@PhdThesis { Xxx:2011:IFA_4100
Permanent link