1

Predictive models for a hydrocracking unit performance

Andreia Filipa Rijo Marianito

Instituto Superior Técnico

October 2015

Abstract: This work was carried out under a traineeship in Plant III in Sines refinery between March

and August 2015 with a view to development a semi-empirical kinetic model that was able to predict, with

reasonable confidence level, the catalytic reactor performance both 1st and 2nd stages of the commercial

unit of Hydrocracking of Sines refinery of Galp Energia. This is a unit with recent operation at the refinery

and high economic incentive, so it is important to understand how the main process variables influence the

different products yield. Ideally, the model should predict the conversion in each reaction stage as a fuction

of the operating conditions (LHSV, temperature, pressure, etc.), as well as of some relevant quality

characteristics of the charge, such as the organic nitrogen concentration, or aromatic content. The model

was validated with experimental data from the commercial unit. It is essential that it also includes a catalytic

deactivation term. It is intended that the model will improve the understanding of the influence of the VGO

physicochemical properties in the unit catalytic performance. Two different models have been developed,

the Model II is more elaborate and realistic, it provides results that describe reasonably the experimental

data and are strongly dependent on the physicochemical charge properties, so the primary purpose of this

work was achieved successfully.

Key-words: Refinery, Hydrocracking, kinetic model, catalytic deactivation, performance, charge

properties

1. Introduction

The hydrocracking unit is composed of

various sections, among which is the reaction

section, which contains two reaction stages in a

high pressure circuit, the HC and HT reaction

sections occur at high temperature and pressure,

always in the presence of hydrogen, at elevated

partial pressure it is necessary to promote the HC

reactions and prevent formation of coke on the

catalyst, while hydrogen is recycled to the loop

reactor to promote a good contact.

The HC unit can operate at temperatures

between 300oC and 440oC and pressures

between 70 and 200 bar, the 1st stage operates

between 400-420oC, while the 2nd stage operates

at temperatures between 370-390oC. Both

reactors are heterogeneous fixed bed, and the

hydrocracking reactions are very exothermic, it is

therefore necessary constant cooling of both

reactors through the quench with hydrogen

streams between each bed. The reactor of first

the stage has six beds, with different catalysts,

while the reactor of the 2nd stage has four beds

with the same catalyst type, only for the catalytic

cracking. The simplified scheme of the

reaction/separation section is in figure 1.

Figure 1 - Simplistic scheme of the reaction/separation HC unit

In this study, it is necessary to focus on

the importance of certain variables that affect

directly or indirectly the process yield. The

properties of the charge to the reactors have a

significant effect in various aspects of the

process, such as the catalyst life, the required

CAT, the hydrogen consumption, yields of

2

product as well as on the properties of these

products.

The system charge is a mixture of several

molecules of different sizes. The charge

composition is dependent on the crude oil used

and the operating conditions of upstream units.

The main charge properties analyzed in the

present study are the nitrogen concentration

and the paraffin/aromatic amount. The

compounds that contain nitrogen are viewed as

a temporary poison for the catalyst, and tend to

exist as basic compounds in the feedstock, which

can be adsorbed on the catalyst acid active sites,

neutralizing them and reducing the cracking

efficacy. On the other hand, polycyclic aromatic

compounds contribute to the formation of coke

on the catalyst surface, they are aromatic

compounds which can be up to several rings, and

tend to be dehydrogenated and make coke.

Therefore, the amount of these compounds in

the charge has a significant effect on the catalyst

activity and fouling rate. Because of the strong

effects of the charge properties in the HC unit

will be taken into account in the development of

the model. It should be noted that other

variables inherent to other procedural areas

influence, equally, the process yield, however,

for the purposes of this study, the emphasis is

placed only on the variables that have a direct

influence in the reaction section.

2. Results

The objective was to develop a robust

model that can predict the performance of both

stages, and to allow the establishment of their

operating conditions, taking into account the

different charge characteristics. The HC unit

receives charges from different sources, it is

necessary to compare the values obtained by

model with the values obtained in the unit in

operation, which constitutes the experimental

data. To do this, we used the data obtained from

analyzes carried out on the charges, VGO, OR and

effluent of both reaction stages.

Among the parameters analyzed to

obtain Model II, the distillation TBP values (°C),

density (kg/m3), and the nitrogen and sulfur

percentage (%) in the sample are used. The same

procedure is make for the Model I with the

addition of a parameter that give the charge

aromaticity account (UOPk).

The experimental data used begin to 23

May 2014, after the first planned stop, and cover

the period until 19 May 2015.

2.1. Procedure for sampling

With respect to the effluents samples of

the two reaction stages, there are two sample

collection panels, one for each stage. These are

high pressure sample points, and whenever it is

necessary to his collection, the operator follows

the procedure given by the licensor. Figure 2

illustrates a sampling panel.

The first step to be performed by the

operator is opening the valve V2 in order to

purge the system line with hydrogen for

cleaning, then to opening the valve V1 to the

reactors effluent entry. The sample is cooled in a

heat exchanger with cooling water at 25-30oC.

Ideally, the sample should be cooled to 50oC,

then it goes to a flash balloon, which

depressurizes, and gases are released into the

flare. The liquid sample is collected at the base

for a metallic container and removed for glass

bottle that is sent to the laboratory.

Figure 2 - Scheme for a sampling panel

However, throughout this study, we

detected some consistency problems with the

experimental data after laboratory analysis. The

amounts of TBP distillation, concluded that there

are deficiencies in the sample collection.

Sometimes the cooling water is not used, that is,

3

the valve V3 is not opened, so the sample can’t

has the recommended 50oC, however, the

problem may also be associated with any

deposits that may exist in the heat exchanger

which is in operation since the start of the unit

and which is isolated, that is, until now has never

been cleaned. Another hypothesis is that the

cooling water flow is not sufficient for cooling the

sample flow. Thus, it is concluded that the

sample analysis will tend to overestimate the

heavy ends at the expense of light.

With respect to the charges of both

stages, VGO and OR, are taken at strategic points

so that they can also be monitored. Both are

collected into panels similar to that shown in the

figure 2, but they are obtained in a low-pressure

panels.

2.2. Calculations

2.2.1. Mass balance

To obtain the mass balance, they are seen

as the Input the VGO flow and the make up H2,

and as output the remaining flows referring to

the products.

𝑀𝑎𝑠𝑠 𝐵𝑎𝑙𝑎𝑛𝑐𝑒(%) =𝐼𝑛𝑝𝑢𝑡−𝑂𝑢𝑡𝑝𝑢𝑡

𝐼𝑛𝑝𝑢𝑡× 100 (2.1)

The mass balance is considered an

average in a 24 hour periods. The data

corresponding to days when the mass balance is

out of +/- 5% tolerance were discarded. It

redistributes the mass balance evenly between

the products through the expression 2.1.

𝑀. 𝐵. 𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑖𝑜𝑛 𝑓𝑎𝑐𝑡𝑜𝑟 =100

100−𝑀𝑎𝑠𝑠 𝐵𝑎𝑙𝑎𝑛𝑐𝑒 (2.2)

2.2.2. Yiels

Charge: Knowing the mass flow rates measured

in the unit and assuming a base of 100% for the

VGO, the make up of H2 is calculated by the

following expression.

𝜂𝑚𝑎𝑘𝑒 𝑢𝑝 𝐻2 (%) =

𝑄𝑚𝑎𝑘𝑒 𝑢𝑝 𝐻2

𝑄𝑉𝐺𝑂×100 (2.3)

Products: The yields of the various products are

calculated the same way, with the introduction

of the M.B. correction factor, expression 2.4.

𝜂𝑃𝑟𝑜𝑑𝑢𝑡𝑜𝑖(%) =

𝑄𝑃𝑟𝑜𝑑𝑢𝑡𝑜𝑖×𝑀.𝐵.𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑖𝑜𝑛 𝑓𝑎𝑐𝑡𝑜𝑟

𝑄𝑉𝐺𝑂×100 (2.4)

The mass balance considered all

products, but for the model the lighter products

are excluded, due to the low amount produced

when compared to other products; therefore, it

is necessary to normalize the yield values in

order to discard the lighter products, as the fuel

gas and LPG.

𝜂𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖,𝑛𝑜𝑟𝑚𝑎𝑙𝑖𝑧𝑒𝑑(%) = 𝜂𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖

(%)×100

100−∑ 𝜂𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖,ℎ𝑖𝑔ℎ(%) (2.5)

2.2.3. RCP

RCP is a Recycle Cut Point, it is the cut

point between the heavier product and the

charge of 2nd stage, OR. With the fixed cut point

value the composition of the VGO stream, OR

and effluents are estimated. With the

experimental TBP data and the recovery rates

(data obtained in laboratory), it is to interpolate

the value of the product cut point and is given

the sample composition. The required cut points

are the values shown in Table 1, it is according to

the experimental data.

Table 1 - Cut points for the different products

2.2.4. Conversions

The conversion to RCP x(RCP) measuring

the amount of charge that will give product in the

1st stage. The more feedstock is delivered to the

2nd stage, the greater their severity. For this

reason, x(RCP) is used to calculate the conversion

per pass in the 2nd stage. We only wish to

compute the overall conversion for the 1st stage,

i.e., they are considered all the reaction

products, this is considered a RCP of 371oC, and

the RCP established for diesel will therefore

cover all lighter products, excluding the UCO..

The PPC is the percentage of combined

charge, which is converted to products in the 2nd

stage reaction. The PPC is defined by the

expression 2.6.

4

𝑃𝑃𝐶, % =𝑄1×(1−𝑋(𝑅𝐶𝑃)−𝑄𝑈𝐶𝑂)

𝑄2 (2.6)

Q1 and Q2 are the mass flow rates to feed the 1st and

2nd stage, respectively. QUCO the purge mass flow and

x(RCP) is a 1st stage conversion for a RCP to 371oC.

2.2.5. CAT

CAT is the average operating temperature

for each reactor.

2.3. MODEL I

For the first case, some assumptions are

considered in order to describe the studied

system, it is necessary to make some approaches

to thereby simplify the model mathematically.

2.3.1. Assumptions

Tubular and plug flow reactor

Reactors in isothermal operation

Homogeneous reactors

Catalyst with similar characteristics on

all six beds, no difference between

HDM, HDS/HDN and HC – HC-R-01

Apparent kinetic of 1st order – limiting

reagent – VGO

Reaction order 0 relative to H2

Deactivation rate of the catayst is

proportional to the catalytic activity

2.3.2. Kinetic Model

In order to model both reactors, based on

semi-empirical kinetic models considered an

ideal plug-flow tubular reactors in isothermal

operation. This is an approximation to the reality

that produces acceptable results and avoids a

large mathematical complexity which arises from

the fact that the concentration may not be

uniform, for example along the radial direction in

the reactor. So the concentration varies

throughout the reactor, it is necessary to select a

reactor volume element and make the mass

balance this 'slice' of the reactor shown in figure

3.

Figure 3 - Selection of a volume element in order to weigh

the plug flow tubular reactor

𝑄𝐶𝐴|𝑥 − 𝑄𝐶𝐴|𝑥+𝑑𝑥 = 𝑟 × 𝑑𝑉 (2.7)

To be r the reaction rate, given by:

𝑟 = 𝐾 × 𝐶𝐴(1 − 𝑥) × 𝑒−𝛼𝑡. (2.8)

Integrating and replacing the expression 2.7, we

have the expression 2.9, assuming the Arrehnius

law for the pre-exponential factor.

𝒙 = 𝟏 − (𝟏 − 𝒙𝟎) × 𝒆−(𝒌𝟎×𝒆−

𝑬𝒂𝑹𝑻⁄ ×

𝟏

𝑳𝑯𝑺𝑽×𝒆−𝜶𝒕)

(2.9)

Expression 2.9 allows us to obtain the

conversion associated with the model, where k0

is the pre-exponential factor, Ea the activation

energy and LHSV space velocity (time-1). Knowing

that k0 depends of the charge properties and

temperature are taken into account two

possibilities for the value of the pre-exponential

factor, k0.

1st hypothesis: 𝑘0 = 𝑎 (𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡) (2.10)

A first case in which k0 is just a constant

and is not affected by the different charge

properties, is not a realistic hypothesis and that

will not meet the required study.

2nd hypothesis: 𝑘0 =𝑎0×𝑒−𝑎1×𝑈𝑂𝑃𝑘

(1+𝑏×𝑁)𝑛 (2.11)

And a second case where k0 is affected by

the different charge properties such as nitrogen

percentage (%) and its aromaticity. Where a0, a1,

b and n are constant, C the nitrogen percentage

(%)in the charge and UOPk a parameter indicator

of the paraffins/aromatics in the charge.

5

Figure 4 – Calculation scheme. Model I

To better understand the mechanism

developed to obtain the following results, it is

observed in Figure 4, in which each block

represents the steps followed and the respective

input and output parameters.

The kinetic parameters (model output

parameters) are optimized using the Excel Solver

tool with the GRC linear non method based on

the minimization of the difference squared

between the conversion given by the model and

experimental conversion, using a least squares

method.

2.3.3. 1st stage results

In the following figures are plotted the

experimental results and the apparent

conversion computed using the model. By

expression 2.9 the apparent conversion was

calculated by the model, values which were

subsequently compared with the apparent

conversions calculated from the experimental

data.

Figure 5 – Graphic comparison between experimental apparent conversion and model conversion. 1st Hypothesis

Figure 6 – Graphic comparison between experimental apparent conversion and model conversion. 2 nd Hypothesis

In the last figures are represented the

experimental data and the final result obtained

by the model, using the two different hypothesis

for the pre-exponential factor described above.

A graphical analysis of the residuals

against the expected values (conversion model)

can also be interesting, as it allows us to observe

if there is any trend or pattern associated with

the errors, or if they are simply randomly

distributed.

Figure 7 - Residue vs expected value. 1st stage

6

From the analysis of the previous figure it

is concluded that the errors associated with the

model are scattered, and there isn’t any

tendency for the residual values. The results

showed the same type behavior for both

hypothesis.

Figure 8 - Comparison of model and experimental apparent conversion apparent conversion. 1st stage

From the analysis of the previous figure, it

follows that once again, both accepted

hypotheses have the same behavior. Table 2

presents the kinetic parameters for the 1st and

2nd hypothesis, for the 1st stage.

Table 2 - Model Kinetic parameters associated with the 1st stage

The value of the cumulative residuals,

corresponding to the 2nd hypothesis, presents a

slight improvement when compared to the 1st

hypothesis, however, it can not be considered a

significant improvement and it does not justify

the use of more parameters, so with a simple

model and with less parameters the same quality

of results can be obtained.

2.3.4. 2nd Stage Results

The conversion is calculated by

expression 2.6, however for the 2nd stage isn’t

subtracted from the product in the charge,

calculate the PPC conversion and the conversion

by the model, not by subtracting the product

"transported" in the charge. The results exhibit

the same behavior observed in 1st stage,

considering the two assumptions the expressions

2.10 and 2.11.

2.4. MODEL II

The model I described above is a simple

model that seeks to describe the conversion at

each stage from an elementary kinetic, when you

consider that the conversion of VGO is the

limiting step. Even introducing some charge

quality parameters, this model fails in general to

describe accurately conversions measures in the

effluent. In addition, there is an evident

dispersion in the experimental points, resulting

from sampling technique the reaction effluent,

making it difficult to develop an accurate model.

The Model II is more elaborate, already

considering a reaction network across product

lumps. Is removed, so the assumption of limiting

step and, more importantly, it is possible to

obtain the overall yield of the various products,

experimental measurement which is more

accurate.

2.4.1. Assumptions

Tubular and plug flow reactor

Reactors in isothermal operation

Homogeneous reactors

Catalyst with similar characteristics on

all six beds, no difference between

HDM, HDS/HDN and HC – HC-R-01

Reaction order 0 relative to H2

Langmuir-Hinshelwood kinetics with

VGO adsorption term

cumulative Deactivation

Law of conservation of mass

(Contribution quench gas (H2)

negligible)

For the possible deployment of the

model, it is necessary to define the cracking

reactions occurring inside the reactor, as is

known to occur numerous reactions it is

impossible to define precisely all of them.

However, we can define a kinetic model in which

the molecules are grouped by lumps. It was,

therefore, a kinetic model involving four lumps:

VGO, diesel, kerosene and naphtha.

7

Figure 9 - Yield vs. Conversion of various products in pilot plant

The reactions considered for the model

development are shown in Figure 10, the choice

of the following reactions is supported by Figure

9.

Figure 10 - Cracking reactions considered to model

development

It was concluded that diesel and kerosene

are primary products, while naphtha is a

secondary product, as the slope at the origin is

practically nil, therefore, initially there will be no

naphtha formation, just don’t admit the reaction

𝑉𝐺𝑂 → 𝑁𝑎𝑝ℎ𝑡𝑎, but only the reactions 4 and 5

of the previous figure, wherein the formation of

primary products stems from naphtha. In figure

11 is shown the calculation scheme followed for

Model II, where they observe the Input and

Output parameters of the model applied to both

stages.

Figure 11- Calculation Scheme. Model II

2.4.2. Kinetic Model

�̇�𝑥𝑖|𝑉 − �̇�𝑥𝑖|𝑉+𝑑𝑉 + ∑ 𝑟𝑖𝑑𝑉 = 0 (2.12)

Taking lim 𝑑𝑉 → 0, it has:

−𝑑(�̇�𝑥𝑖) + ∑ 𝑟𝑖𝑑𝑉 = 0 (2.13)

Taking 𝑑�̇� = 0. It has:

�̇�𝑑𝑥𝑖 = ∑ 𝑟𝑖𝑑𝑉 (2.14)

�̇�𝑑𝑥𝑖

𝑑𝑉= ∑ 𝑟𝑖 (2.15)

As 𝑑(𝜌𝑉)̇ = 0 (2.16)

We have 𝑑𝜌�̇� + �̇�𝑑𝜌 = 0 (2.17)

As �̇� e ρ are constants, was replace �̇� by

expression 2.14 for 𝜌�̇� and we have:

𝜌�̇�𝑑𝑥 = ∑ 𝑟 𝑑𝑉 (2.18)

As 𝐿𝐻𝑆𝑉 =�̇�

𝑉𝑐𝑎𝑡 (2.19)

𝑑𝑉 = �̇�𝑑𝜏 (2.20)

Replacing the expression 2.20 in expression 2.18,

we have: 𝜌�̇�𝑑𝑥 = ∑ 𝑟 �̇�𝑑𝜏 (2.21)

So, 𝜌𝑑𝑥 = ∑ 𝑟 𝑑𝜏 (2.22)

Diferencial equation: 𝑑𝑥

𝑑𝜏=

∑ 𝑟𝜌⁄ (2.23)

For the calculation of the reaction rate

Langmuir-Hinshelwood type kinetic is assumed.

𝑟𝑛 =𝑘0×𝑒−

𝐸𝑎𝑅𝑇⁄ ×𝐶𝑖

1+𝐾𝑎𝑑𝑠×𝐶𝑖+𝑘𝑁×𝐶𝑁× ∅𝑛 (2.24)

Where k0 is the pre-exponential factor

and Ea the activation energy, R is the gas

constant, T is the operating temperature (in this

case, is used the CAT considering the isothermal

reactor) Ci reagent composition, CN is the

nitrogen concentration in the charge (VGO)

constant for each day. KADS is a constant

associated to the reagent adsorption on to the

catalyst active sites and KN is a constant

associated with the nitrogen concentration in

the charge. Ø is the activity of the catalyst, is a

cumulative factor that decreases over the unity

of the time working, then Ø is calculated

recursively by the expression 2.25, and τ is the

reactor residence time.

∅𝑛 = ∅𝑛−1−∝× 𝑟𝑛−1 × ∆𝑡 (2.25)

8

ØN is the catalytic activity of the day n (day

study), Øn-1 is the catalytic activity from the

previous day, α is the catalyst deactivation factor,

rn-1 at the previous day's reaction rate, given by

the expression 2.24 and Δt is the interval in days.

Thus, it integrates along the reactor using the

Euler method. The desired result is the effluent

composition at the reactor outlet, to that

integrates to the residence time of each stage,

variable for each day, given by the inverse of

LHSV. It was also studied the integration step,

and it follows that the best was 0.002 hours.

2.4.3. Calculations

After numerical integration by Euler's

method, the compositions of the effluent have

been computed. It was then made a mass

balance to the system of Figure 12, in order to

obtain the mass flow rate of each product outlet,

to thereby calculate the yield by the expression

2.26.

𝑌𝑖𝑒𝑙𝑑 (%) =𝑄𝑝𝑟𝑜𝑑𝑢𝑐𝑡

𝑄𝑉𝐺𝑂× 100 (2.26)

Figure 12 - Scheme system considered for the model. Reaction zone / separation

NAPHTAS Balance

𝑥𝑁,1 × 𝑄1 + 𝑥𝑁,2 × 𝑄2 = 𝑄𝑁 (2.27)

KEROSENE Balance

𝑥𝐾,1 × 𝑄1 + 𝑥𝐾,2 × 𝑄2 = 𝑄𝐾 (2.28)

DIESEL Balance

𝑥𝐷,1 × 𝑄1 + 𝑥𝐷,2 × 𝑄2 − 𝑥𝐷,𝑈𝐶𝑂 × 𝑄𝑈𝐶𝑂 = 𝑄𝐷 (2.29)

UCO Balance

𝑥𝑈𝐶𝑂,1 × 𝑄1 + 𝑥𝑈𝐶𝑂,2 × 𝑄2 − 𝑥𝑈𝐶𝑂,𝑂𝑅 × 𝑄𝑂𝑅 =

𝑥𝑈𝐶𝑂,𝑈𝐶𝑂 × 𝑄𝑈𝐶𝑂 (2.30)

Q1 and Q2 are the mass flow of stream 1 and 2,

respectively, and xi,1 and xi,2 are the compositions of

the different products in respective streams (model

data).

While the composition and amount of

VGO are model inputs, the quantity and OR

composition is determined by the model through

a separation factor in a fractionating column.

However, this separation factor isn’t constant

over the period under study, because it is known

that on the 5 August 2014, due to a reaction

upset the fractionating column, the bottom

plates were damaged and the column lost

separation efficiency of the recycle oil and diesel,

of around 3%. Figure 13 is a graphical

comparison of the experimental composition

and the composition achieved by using the

separation factors mentioned above, which is

visible to the proximity of values.

Figure 13 - Composition of comparison in diesel UCO (experimental) and composition in diesel obtained by model

After optimization of kinetic parameters,

a model sensitivity analysis is done in relation to

the nitrogen values in the charge, that is, the

intention was to confront the data supplied by

the licensor relative to the influence of nitrogen

in the charge, with the values obtained during

the the model application, as shown in the

following figure. It is known from the design

conditions, and what is the expected influence of

the nitrogen concentration in the expected

variation on the CAT value.

9

Figure 14 - Operative curve vs. model curve

It was concluded that in addition the

model developed to capture the variations of

nitrogen in charge, meets with the overall trend

indicated by licensor, which allows us to state

that one of the major objectives is accomplished,

that is, the variation is well described by the

constant inhibition nitrogen adjusted by the

model.

It is now graphically compared the

experimental performance against the yield

calculated by the model data. The following

figures illustrate the yield of each product, such

as diesel, kerosene, naphtha and UCO.

Figure 15 – Experimental yield vs Model yield - DIESEL

Figure 16 - Experimental yield vs Model yield - KEROSENE

Figure 17 - Experimental yield vs Model yield - NAPHTAS

Figure 18 - Experimental yield vs Model yield - UCO

For the above figures, the data points

(days) in which the mass balance didn’t close, or

introduce an error greater than +/- 5%. In the

previous figures, two important periods for the

unit are represented, the first period to 6 August

2014, the incident that occurred in fractionating

column and that meant a loss of yiel of this,

clearly visible for diesel. And a second period to

15 September 2014, when there was a decrease

in 1st stage conversion, increasing by

consequent, the conversion in the 2nd stage, due

to accelerated catalyst deactivation which was

felt in the 1st reaction stage. In the following

tables the kinetic parameters are optimized for

the 1st and 2nd reaction stage.

Table 3 – kinetic parameters for the 1st stage model

Table 4 - kinetic parameters for the 2nd stage model

10

This model is sensitive to nitrogen

variations in the charge (load property

considered for the model development), for the

five reactions considered was obtained values for

the pre-exponential factor and the activation

energy, Ea. The activation energy lies within the

range of expected values, as well as pre-

exponential factors have all the same order of

magnitude. The parameter related to the

catalytic deactivation, α, is much higher in the 1st

stage in Model II, it is considered as a whole, and

as such, the deactivation factor fulfils this entire

period, and it is concluded that the catalytic

deactivation observed in the 1st stage is more

severe than that in the 2nd stage.

It was also considered the 1st stage

effluent compositions (single stage with frequent

samplings) for obtaining the optimum values of

the kinetic parameters. The only effluent for

which analysis is available, is used for

optimization of kinetic parameters, so to "force"

the model to represent what is produced in the

1st stage, and not to be completely randomly.

3. Conclusion

Model I is a relatively simple model and

does not show any significant difference in the

results, when the charges properties change. In

fact the model fits to the experimental data, but

not have in attention the physical and chemical

properties of the charge.

So, it was necessary to develop a more

complex but also more realistic model, which

although essentially become under the same

assumptions, and considering a network reaction

between products lumps. The assumption of

limiting step was removed and it was possible to

obtain the overall yield of the various products,

the products yield is a more correct experimental

measurement. It is an elaborate and realistic

model whose results are to conform to the

experimental data and are strongly dependent

on the physicochemical properties of the charge,

thus fulfilling the fundamental objective of this

work, figure 14 checks that the model captures

the nitrogen charge variations, following the

trend predicted by the licensor. The charge

composition (% diesel and VGO) is also given as

Input to the system, so the model can take into

account that the composition charge and the

nitrogen percentage present in the charge.

References

Site Oficial Galp Energia

http://www.galpenergia.com

Data Book de Segurança, Saúde e Ambiente

2013, Refinaria de Sines, Galp Energia.

Direcção Geral de Energia e Geologia, DGEG;

Ministério do Ambiente, Ordenamento do

Território e Energia; Governo de Portugal.

http://www.dgeg.pt/ Consultado a 14/05/2015

Informação Estatística; Associação Portuguesa

de Empresas Petrolíferas, APETRO.

http://www.apetro.pt/ Consultado a 16/07/2015

Secção Energia, Jornal Expresso Online, noticia

referente a 16 de Janeiro de 2013 às 17h23min.

http://expresso.sapo.pt/economia/economina_

energia/nova-unidade-da-refinaria-de-sines-ja-

produz-gasoleo=f779915

ISOCRACKING, CBI/Lummus Global

http://www.cbi.com/images/uploads/tech_shee

ts/Isocracking-12.pdf

Chevron Lummus Global LLC, “Process Manual,

Chevron Isocracking Unit, Galp Refinery, Sines,

Portugal”, December 2007

Chevron Lummus Global LLC, “Operations

Training Workbook - Chevron ISOCRACKING

Unit”, Galp Refinery; Sines, Portugal; October

2010

www.famousscientists.com.org/rudolf-christian-

karl-diesel/

Consultado a 15/09/2015

www.webserver.dmt.upm.es

Consultado a 15/09/2015

Lemos, Francisco; M. Lopes, José; Ramôa Ribeiro,

Fernando; “Reactores Químicos”, IST PRESS,

Novembro 2002, ISBN: 972-8469-09-8

11

www.earthobservatory.nasa.gov

Consultado a 1/09/2015

www.astm.org

Consultado a 1/09/2015

Nunes dos Santos, A. M., “Reactores Químicos”,

Vol 1, Fundação Calouste Gulbenkian, 1990,

ISBN: 972-31-0519-5

Kayode Coker, A., “Modeling f Chemical Kinetics

and Reactor Design”, Gulf Publishing Company,

Texas, 2001. ISBN: 0-88415-481-5

Cooper, Alfred Ronald, Jeffreys, Godfrey

Vaughan, “Chemical Kinetics and reactor

design”, Oliver and Boyd, 1971, Universidade de

Califórnia. ISBN: 0050021141

Harriott, Peter, “Chemical Reactor Design”, CRC

Press, 2002. ISBN: 0203910230

“Technical Data Book – Petroleum Refining”,

American Petroleum Institute (API), Sixth Edition,

April 1997.

Azita Barkhordari, Shohred Fatemi, Mahdi

Daneshpayeh, Hossain Zamani; “Development of

a discrete kinetic model for modeling of industrial

hydrocracking reaction accompanied with

catalyst deactivation”, School of Chemical

Engineering, Faculty of Engineering, University of

Theran, Iran.

M. Rashidzadeh, A. Ahmad, S. Sadighi; “Studying

of Catalyst Deactivation in a Commercial

Hydrocracking Process (ISOMAX)”, Journal of

Petroleum Science and Technology, Vol.1, No 1,

2011, 46-54.

Couch, Keith A.; Glavin, James P.; Johnson, Aaron

O.; “The Impact of Bitumen-Derived Feeds on the

FCC Unit”, UOP LLC, a Honeywell Company; Des

Plaines, Illinois, USA.

Esteves, Eduardo, “Regressão não-linear

utilizando a ferramenta Solver do Microsoft

Excel”, Instituto Superior de Engenharia,

Universidade do Algarve; 10 de Janeiro de 2011.

https://support.microsoft.com (Solver uses

generalized reduced gradient algorithm)

Mark Harmon, “Step-by-Step Optimization with

Solver Excel, The Excel Statistical Master”, Excel

Master Series, 2011. ISBN: 987-1-937159-11-5.

Costa, Fernando Pestana, “Equações diferenciais

ordinárias”, Março 2001, IST Press. ISBN: 972-

8469-00-4.