Constrained multivariable control and real time optimisation of a distillation process

Optimum operation of a process is often not at the ´top of the hill´, but at the intersection of constraints. If the constraints are liable to shift during process operations then a different control strategy may be applicable. This paper describes an application to a real-time simulation of a non-linear model of a BTX (benzene, toluene and xylene) separation process. The application involves the online evaluation of the economic optimum operating point (by a linear programming optimiser) and its achievement through a model based predictive controller which takes into account the constraint conditions of process variables. The LP optimiser in its incremental form is shown to prescribe the best steady state solutions that are consistent with the controller set points. This formulation of the LP optimiser is shown to push the variables towards their constraints and therefore achieving the best economical operating level. The controller is shown to steer to avoid constraint violations at the same time as satisfying overall control objectives.