A Robustification of the Conventional GPC Controller based on the Structured 𝑯∞ Synthesis for a Multivariable Non-Square Process.

Autor: Azzeddine, MENASERIA, Moussa, SEDRAOUI, Mohammed, AIDOUD
Předmět:
Zdroj: Electrotehnica, Electronica, Automatica; Jul-Sep2021, Vol. 69 Issue 3, p100-111, 12p
Abstrakt: This paper deals with the robustification problem based on the higher order Q-Youla parameter, which is used for robustifying an initial polynomial form of the Generalized Predictive Control GPC law. This new proposed robustification is intended for a Multi-Input and Multi-Output MIMO systems whose actual dynamics is modelled using an uncertain non-square model. Accordingly, the challenge is to seek a reduced-order Q-Youla parameter, which is featured by a fixed-order transfer function matrix using the structured 𝐻𝐻∞ synthesis control. As a result, the obtained Q-Youla parameter is then fed into the initial GPC controller to increase its trade-off level appearing between the corresponding Nominal Performance NP margin and corresponding Robust Stability RS margin, i.e., NP-RS. For this purpose, a state-space representation of the desired Q-Youla parameter is firstly preselected by the designer and the finding problem is then expressed as a weighted mixed sensitivity issue in which a large number of frequency-domain requirements are considered. Subsequently, the previous problem is tackled using the MATLAB® function hinfstruct and the resulting Q-Youla parameter is finally employed for building the robustified GPC controller using the Youla parameterization. To validate the proposed robustification strategy, a MIMO ill-conditioned distillation column is controlled in which the resulting performances are compared, in frequency- and time- domains, with those given by the initial GPC controller. The simulation results later show the suitability of the proposed robust GPC controller over the initial GPC one in terms of real-time implementation, NP-RS trade-off in the presence of unstructured multiplicative model uncertainties. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index