Hybrid Models of Distillation Towers for Plant Optimization

Real time optimization of plant operations started in 1980s and expanded rapidly during 1990s and early 2000s.  Such applications have brought significant profit improvements to e.g. ethylene plants or to refineries, as witnessed by the large number of applications.  Benefits from these applications have been made possible by the rigorous models which are able to predict accurately the plant behavior.  At the same time, the need to develop and maintain very accurate rigorous models calls for significant expertise and large effort in implementation and maintenance.

This work explores a possibility to use simpler models for plant optimization.  These models (hybrid models) consist of material and energy balances plus empirical models that relate separation in distillation tows with mass and energy balances for different sections of the tower.  The goal of the research is to develop highly accurate models (within 1% to 2%) of the actual tower behavior, while minimizing the effort required for model building and maintenance.  

We present hybrid models for several different types of distillation towers (other than CDU) encountered in the refinery operations and compare their predictive behavior with rigorous tray to tray models.