An optimal centralized predictive control scheme applied to a Non-Square MIMO benchmark problem

This paper aims at the implementation of advanced control technique such as Generalized Predictive Control (GPC) for non-square Multi-Input-Multi-Output (MIMO) benchmark problem. The process considered is a coupled distillation column (3 input, 2 output) studied by Levein and Morari. The design approach for tuning controller parameters in conventional methods of a non-square system is a big challenge, as it involves computational complexities. In this work, predictive algorithm with constraints on the input and output control signals are developed and compared with the PID controller by retaining its non-minimum phase characteristics. GPC uses Controlled Autoregressive and Integrated Moving-Average (CARIMA) model. Simulation studies demonstrate the superiority of the constrained GPC over GA-PID controller for a non-square multivariable process. The effectiveness of the advance control structure ensures the reduction in computational complexities and fast convergences with better tracking capabilities.