Apr-2009

Hydrocracker parametric sensitivity study

Product yields and hydrogen consumption were predicted using a discreet lumping model. The model was used to study unit responses to changes in operating conditions, to show that small changes in catalyst loading and recycle ratio can significantly improve yields

Majid Bahmani Tarbiat, Moalem University
Reza Seif Mohaddecy, Sepehr Sadighi and Maryam Mashayekhi, Research Institute of Petroleum Industry

Article Summary

Data were accumulated for an industrial hydrocracker over a period of approximately four years. A discrete lumping model was used to predict the yields of liquefied petroleum gases (LPG), light naphtha, heavy naphtha, kerosene, diesel, residue and the consumed chemical hydrogen. The model’s parameters were obtained by minimising the differences between the model’s predictions and the plant data using non-linear regression algorithms. The validated model was used to study the hydrocracker unit’s responses to changes in operating conditions. Simulations showed that small changes in the catalyst loading and recycle ratio can significantly improve yields.

Hydrocracking is the process of converting vacuum gas oils (VGO) into more valuable lower boiling products, such as naphtha, jet fuel and middle distillates. VGO hydrocracking differs from residue hydroconversion (also called hydrocracking), as the former produces a lot more butane-range products than the latter. In the first process, catalytic reactions are prevalent, while the latter relies primarily on thermal reactions. Hydrocracking is one of the most versatile processing steps currently available to modern refineries.

Hydrocracking has proven to be highly flexible, capable of desulphuris-ing vacuum residua, making lubricating oils, demetalising catalytic cracker feed, cracking gas oil to jet fuel and gasoline, or cracking naphtha to LPG. The yield of products from existing plants can be varied over a fairly broad range in response to changes in demand. Currently, the highest volume product is gasoline. Hydrocracking has the flexibility to meet these changes in product demand. In most of its applications, the relative amount of desired (and undesired) products obtained from its process are crucial to its economic success. A method for predicting these yields has been investigated in this study.

The feedstocks processed in the petroleum industries consist of a large number of components. A typical feed for an industrial hydrocracker unit contains paraffins, iso paraffins, naphthenes and aromatics components. Theses components follow very complicated reaction pathways with carbenium ion intermediates. Modelling such chemical processes is complex due to the extremely large number of reactions and the difficulties in measuring feed and product compositions. The modelling methodologies developed over the years for cracking systems, such as catalytic cracking and hydrocracking, can be classified into two broad categories: lumping models and mechanistic models. In lumped models, the actual reaction network is reduced to a small number of reactions among the lumped species.1-4 The lumps, based on compound types present in feedstock and products such as gas oil, LPG, gasoline and diesel, are often defined by boiling point ranges.5 This approach, also known as discrete lumping, has been utilised in this study.

In hydrocracking, hydrogen is used for the hydrogenation of unsaturated compounds, preventing the formation of undesired products. One of the most important undesired products is coke, which is the result of the oligomerisation reaction of olefinic intermediate compounds formed during hydrocracking reactions. Industrial hydrocracking reactions take place at pressures of around 190 bar, and at such high hydrogen partial pressures coke formation is retarded.

Pure hydrogen streams are not available in refineries, and hydrogen management is subject to ongoing research. Hydrogen plants in refineries use natural gas in a steam reforming process to produce hydrogen. Due to the limited capacity in hydrogen plants, the make-up hydrogen streams in hydrocracking units are also limited. Therefore, a recycle gas stream from the high-pressure separator is used to compensate for the loss of hydrogen in the reaction.

Stangeland developed a kinetic model for predicting hydrocracker yields using correlations based on the boiling point of each of the pseudocomponents that characterise the cut.1 Prior to Stangeland’s work, the methodology for predicting the spectrum of lighter components produced from the hydrocracking of heavier compounds was based on the selection of a small number of products. Various parallel and series reactions were “devised” to produce them. Kinetic parameters were determined to best fit experimental data. The main disadvantage of this approach was that a change in the specification of hydrocracker product, or a number of products, required a redesign of the reactions and a new fitting of the experimental data. The flexibility of this approach made it difficult to use in practical applications.

To overcome the limitations of this early approach, Stangeland developed a kinetic model where the same parameters could be used to describe the yields for similar feeds, even if their boiling ranges were different. Stangeland’s kinetic model depends on only three parameters: one parameter describes the butane yield; the second the type of feed (eg, naphthenic or paraffinic) or the type of catalytic process (selective or random 
cracking); and the third quantifies the cracking reaction rate of each pseudocomponent.

Mohanty, et al, implemented Stangeland’s kinetic model in a computer model for a two-stage commercial-scale VGO hydrocracker.6 For the hydrocracking reactions, these authors assumed pseudo-homogeneous first-order reactions. The parameters of the kinetic model were fitted to literature and plant data.

In this study, data for an industrial hydrocracker unit were accumulated, covering the start-of run (SOR) to the end-of-run (EOR) of the unit. The model’s parameters were optimised using six months worth of data. The validated model was then used to predict the remaining portion of the data. The validated model obtained was subsequently used to carry out a parametric sensitivity analysis. The yields of LPG, naphthas, diesel and kerosene, as well as the chemical hydrogen consumption were studied in terms of the variations in catalyst loadings, and feed and recycle rates.

Model development 
and validation
Since a hydrocarbon cut comprises a number of components (or pseudocomponents), the boiling point range from light gases to the heaviest pseudocomponent of the feed is discretised in N intervals of each 50, as illustrated in Figure 1. Each interval 
is represented by a single pseudocomponent with a single average boiling point.