Effect of lump partitioning on the accuracy of a commercial vacuum gas oil hydrocracking model

Abstract

In an industrial hydrocracking plant, except to the reactor, the separation section can affect the accuracy of the developed mathematical model. In this study, two combined-bed six and seven-lump kinetic models for the hydrocracking of vacuum gas oil in a real refinery are developed and their accuracies for the prediction of product yields are compared. In the six-lump model, the engaged components are selected according to the actual products of the plant, but the seven-lump one is included of lumps which are partitioned based on the specified initial and final boiling points. For both approaches, a complete kinetic network with the catalyst-decay functions is considered. It is found that over a 17-month period of operation from the start of run, the average absolute deviation (AAD%) of the six-lump and the seven-lump models for the yield prediction are 20.51% and 5.21%, respectively for all products. Additionally, the sum of residuals for the seven-lump model has been appreciably decreased in comparison to the six-lump one. Therefore, it is confirmed that partitioning of the products to the lumps with specific boiling points can acceptably decrease the deviation of the prediction. So it is conclusive that including the operation of the separation section in commercial models can effectively increase the accuracy of the model and smooth the prediction error.