Reduced Order Modeling of a Crude Distillation Unit for Online Applications

Abstract

A transport phenomena based process model is a prerequisite for supervisory level online process optimization. For real time applications it is necessary to have an efficient and robust solution procedure for the model equations. A method for the reduction in the number of model equations for a crude petroleum distillation unit is reported here. The predictions from the rigorous, and several reduced order models were compared with the experimental data from an operating refinery. The results of the reduced order model not only compared favorably with those of the rigorous model but the former was also found to be more flexible to tuning with the plant data and showed greater tolerance to noisy measurements as compared to the latter. Use of empirical thermodynamic relations for VLE (vapor-liquid equilibrium) data, and analytical derivatives, choice of independent variables and numbering of stages were found to further reduce computation time significantly.