Dp 451 en
Control of ill-conditioned systems
Author: Pich Jiří
The aim of this thesis is to propose modifications of Model Predictive Controller (MPC) that
will improve control of ill-conditioned systems. Based on a background research, an overview
of currently used directionality measure and MPC applications for ill-conditioned systems is
presented in the thesis. The suitable directionality measures are used for system conditioning
analysis. Based on current knowledge and available methods, two extensions of MPC for illconditioned
systems control are proposed in this thesis.
The first of the presented methods extends the MPC cost function by a term penalizing absolute
value of system input. This approach limits the size of the system input absolute value and
prevents its unconstrained rise, which may occur if a system output should track a reference
which requires a step corresponding to the system small singular values.
The second proposed method extends the cost function by a term penalizing the system input
movement in directions corresponding to the small system singular values. This modification
prevents undesirably high system inputs in directions where the input has a small impact on the
reference tracking abilities.
The proposed MPC modifications are applied to a model of a distillation column to show
the reference tracking properties and to check the robust stability of the extended MPC. Robust
stability of the system is checked for different uncertainty representations in order to show
how the additional criteria influence the frequency characteristics for these different uncertainty
representations.
- Jiří Pich,tel: +420 732 238 703, mailto:jiripich@gmail.com
- Prof. Ing. Vladimír Havlena CSc., vedoucí práce, mailto:havlena@fel.cvut.cz
Práce vytvořena ve spolupráci se společností Honeywell.