ONLINE OPTIMIZATION - A HIERARCHICAL SCHEME FOR DISTILLATION COLUMN CONTROL

This paper presents a hierarchical on-line optimization and nonlinear control of a distillation column using a new semirigorous model which drastically reduces the problem dimensions compared to a rigorous model without significantly compromising with accuracy. Single level control is often found to be inadequate since optimization using highly nonlinear physical models cannot be performed in the very short real times available between controller actions. A two level approach can effectively overcome this difficulty with optimization being carried out using more realistic nonlinear physical models at the supervisory level. The nominal control vector trajectory generated at this level using a nonlinear analytical model predictive control (NAMPC) is passed on to the regulatory level where an LQ type neighbouring optimal control action is generated. Simulation results showed that a two level control can effectively deal with process-model mismatch in the presence of disturbances. Experimentation on pilot scale distillation column vindicated the simulation results and demonstrated the superiority of the hierarchical control scheme.