Inferential monitoring & optimization of crude units via hybrid models

A new class of models (hybrid models) that can be used to identify the feed distillation curve, predict product properties from tower operating conditions, and predict product properties from feed properties and material and energy balances is put forward. The model predicts target properties within 1% of the rigorous tray to tray model consisting of 10k-20k nonlinear equations. The optimization of tower operating conditions based on the hybrid model and based on the equation-oriented AspenPlus model was presented. The identification of the crude feed distillation curve based on the limited number of tray temperature and the hybrid model variables was put forward, including the prediction of product properties from the available tray temperatures and hybrid model variables. The accuracy of the hybrid model predictions, the almost linear nature of its equations, and its suitability for on-line monitoring were elucidated. This is an abstract of a paper presented at the 2011 AIChE Spring Meeting & 7th Global Congress on Process Safety (Chicago, IL 3/13-17/2011).