Poster Session

Post-optimality analysis of Steady-State Target Calculation of MPC

Abdallah Al-Shammari, PhD student in Chemical Eng. (Process Control)

Supervisor - Dr. J. Fraser Forbes

There are two main approaches to deal with uncertainty in any optimization problem: optimization under uncertainty and post-optimality analysis. In steady-state linear target calculation, some work was done using the first approach while post-optimality analysis is rarely used in model predictive control. In this study, we try to establish a technique using post-optimality analysis to assess and improve economic performance of target calculation (and MPC) and show how sensitivity and stability information can be companied to economically tune the constraints. In other words, the method determines how the constraints can be relaxed to improve the objective function without changing the constraints active set and with the consideration of biases. In addition, it computes variation ranges of basis (and process limits) such that the system is still feasible with the same active constraints (to avoid unnecessarily bouncing of target from vertex to other). At design stage, stability information of economic parameters is used to properly design the system to reduce the possibility of target bouncing and to determine how sensitive the economical model of target calculation to certain changes.