Application of Nonlinear Model Predictive Control to Optimal Batch Distillation

Abstract

The problem of properly determining and implementing the optimal operational trajectory for batch distillation is discussed. Modeling, control, and optimization aspects are considered. A common modeling approach is shown to lead to significant errors in the solution trajectory. Errors arising from the equilibrium stage assumption show the need for rigorous models with efficient solution methods. Several industrially important control problems are discussed and a control pairing which eliminates some interactions is suggested. Optimization results are shown to be quite sensitive to model parametric (e.g. Murphree efficiency) and product specification changes. Parameter sensitivity studies indicate that a priori parameter estimates must be applicable over the range of operation, or that on-line parameter estimation with reoptimization is required. Nonlinear Model Predictive Control (NMPC) offers an efficient, straightforward method of addressing difficulties in modeling, optimization, and control of batch distillation.