| Abstract |
There is an ongoing discussion among repetitive process control engineers whether to use multiresonant controllers or iterative learning ones. Power electronics engineers are often in favour of multiresonant controllers whereas motion control designers more often incorporate iterative learning controllers (ILC) into their systems. Both camps of researchers seem to be entrenched and rarely willing to admit that both techniques are capable of introducing exactly the same generating polynomial to the system. Moreover, both techniques in their basic forms suffer from a lack of robustness to a persevering non-zero control error, e.g. due to the physical limitations on the plant side. To render the techniques practical, several tweaks have to be made. Neither of them is also innately immune to the reference and/or disturbance signal frequency fluctuations. Even small frequency variations at the level of 1\% can render the basic ILC impractical in the case of the grid-tied converter discussed in this paper. The paper is supposed to serve as the second chapter of the conversational guide on the similarity and challenges of multiresonant and iterative learning controllers. Alongside sparking the discussion, our main contribution here is a novel conditional learning concept within the ILC, inspired by the variable damping proposed for multiresonant controllers. An improvement gained by augmenting the basic ILC with the conditional learning is illustrated based on numerical simulations. Further possibilities of introducing an adaptation algorithm into the modified controller are suggested. |
