(3bo) Advanced Dynamic Optimization and Control for Large-Scale Systems | AIChE

(3bo) Advanced Dynamic Optimization and Control for Large-Scale Systems

Authors 

Huang, R. - Presenter, Carnegie Mellon University


Air Separation Unit (ASU) is an energy intensive process. It is economically attractive to apply advanced control techniques to the ASU. Current advances in dynamic optimization algorithms enable us to consider more complicated model and to improve controller’s performance. This work addresses the following problems:  

(1)  Reducing computational delay: Implementing nonlinear model predictive control (NMPC) based on first principle dynamic model exhibit many advantages against traditional linear MPC based on empirical model. However, computational delay associated with solving large-scale optimization problem degrade the performance of the controller. Advanced step NMPC algorithm [1] is utilized to implement NMPC on the ASU problem and the computational delay is reduced by 200 times.

(2)  Offset-free in the presence of plant model mismatch: Offset-free behavior which requires the output of the process to be exactly at the set point is an important requirement in control practice. An NMPC framework based on various state estimators is proposed and implemented on the ASU to achieve offset free even in the presence of plant model mismatch.

(3)  Robust NMPC: It is important to guarantee robust stability of NMPC in the presence of large uncertainties. An multi-scenario formulation is proposed to guarantee the robust stability of NMPC algorithm. It is implemented on the ASU process.   

(4)  Economic-oriented control: With the first-principle dynamic model, it is preferred to directly solved the NMPC problem to maximize the economic profit of the process. An infinite-horizon formulation with guarantee stability is proposed to solve the economic-objective control problem for the ASU.

It is worth pointing out that the proposed methods are not restricted to the ASU problem. On the contrary, they are expected to work for any large-scale processes.

 Reference:

 [1] V. M. Zavala and L.T. Biegler. The advanced step NMC controller, optimality, stability and robustness. Automatica 2009, 45: 86.