Predictive Control of Multilevel Inverters for Improving Performance and Efficiency in Energy Distribution and Drive Control
Final Report Abstract
This work shows that with a good problem formulation and by applying several ideas, some of which already known in optimal control theory contexts, one can immensely increase the speed of solving optimization problems for electrical drives. The current distortion optimization problem was chosen to be studied for an induction motor drive. A 5-level inverter was chosen as the power converter due to the fact that distortion optimization gets more complicated and time consuming for multilevel inverters. Based on an extensive literature study, synchronous optimal modulation was chosen as a reference due to its superior performance at low switching frequencies. Then a step-by-step approach was followed to study the current distortion, define the criteria for its optimality, increase the speed of offline optimization and finally apply the ideas from predictive control theory in order to achieve an online optimization. In the first place, the theoretical analysis of current distortion optimization in the αβ coordinate system revealed the relationship between voltage vector sequences of a converter and its current distortion. Based on experimental results, it has been reported that by limiting the switchings to transitions between adjacent voltage vectors, the current distortion decreases. This analysis provides an analytical proof for this statement. In addition, this piece of information made it possible to break down the complicated global current optimization problem to several simpler ones. As a result, a significant improvement was achieved in offline calculation speed. In the next step, the current distortion was formulated as an objective function of a predictive controller. The prediction horizon was set to be equal to 30◦ of the fundamental waveform period. It was known from previous studies that certain voltage vector sequences cause less current distortion. So the objective function optimization became a continuous optimization of switching intervals rather than a successive search for voltage vectors that correspond to the minimum value of the objective function. The continuous optimization problem was finally solved by a modified Gauss-Newton algorithm. Finally, based on measurements at different operating points it was shown that the results of online and offline optimizations are comparable. The results of this work and some other recent studies on fast optimization techniques alongside the fact that calculation power is rapidly increasing imply that many optimal control strategies that were traditionally used for systems with slow dynamics can now, or very soon, be used for electrical drives in real-time. After all, it should be noted that the word optimum must be used with extreme caution since it always implies a set of assumptions that are sometimes neglected. For example, in current distortion optimization, all switching instances are calculated assuming that an ideal converter with ideal switches and without any losses, dead-time effect, capacitor voltage ripple etc. is available. Besides, even harmonics are considered to be absent due to symmetries - which is not the case in practice. The question how the true optimum can be reached is still unanswered, which shows that much work, both practical and theoretical, remains to be done in this area.
Publications
- “An efficient method to calculate optimal pulse patterns for medium voltage converters,” in Industrial Electronics Society, IECON 2014 - 40th Annual Conference of the IEEE, Oct 2014, pp. 1221–1226
R. Fotouhi, L. Leitner, R. Kennel, and H. du Toit Mouton
(See online at https://doi.org/10.1109/IECON.2014.7048658) - “An efficient method to calculate optimal pulse patterns for multi-level converters,” in Power Electronics and ECCE Asia (ICPE-ECCE Asia), 2015 9th International Conference on, June 2015, pp. 533–539
R. Fotouhi, A. Sorokin, R. Kennel, and H. du Toit Mouton
(See online at https://doi.org/10.1109/ICPE.2015.7167836) - “Online optimization of current distortion for mv drives based on stator flux trajectory predictions,” in 2015 IEEE International Symposium on Predictive Control of Electrical Drives and Power Electronics (PRECEDE), Oct 2015, pp. 3–8
R. Fotouhi, A. Sorokin, R. Kennel, and H. d. T. Mouton
(See online at https://doi.org/10.1109/PRECEDE.2015.7395577)
