Constrained Long Range Predictive Control of a Heavy Oil Fractionator

This paper is concerned with the evaluation of a constrained long range predictive control algorithm on a simulated multivariable model of a heavy oil fractionator. A multivariable generalized predictive controller (MGPC) is combined with a quadratic programming optimization package to handle the constraints. The results indicate that the constrained MGPC algorithm delivers good servo and regulatory control. The algorithm is also able to maintain this performance in the presence of large model-plant gain mismatch.