Applied Energy 133 (2014) 169–182

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Integrating hydroprocessors in refinery hydrogen network optimisation

Blessing Umana a, Abeer Shoaib b, Nan Zhang a,⇑, Robin Smith a
a
Centre for Process Integration, School of Chemical Engineering and Analytical Science, The University of Manchester, PO Box 88, Sackville Street, M60 1QD, UK
b
Department of Chemical and Refining Engineering, Faculty of Petroleum and Mining Engineering, Suez Canal University, Suez, Egypt

h i g h l i g h t s

 Correlations for sulphur in liquid products are modified for hydrotreaters.

 Correlations for light hydrocarbons in hydrotreaters are developed.
 Effects of process and operational variables on H2 production are investigated.
 An NLP algorithm is developed using CONOPT solver, demonstrated with a case study.

a r t i c l e i n f o a b s t r a c t

Article history: Recent developments on sulphur specifications in middle distillate fuels are exerting a higher demand on
Received 11 February 2014 refinery hydrogen systems. Previous methodologies on hydrogen network optimisation have been devel-
Received in revised form 28 June 2014 oped mainly based on the assumption of fixed hydroprocessing performance with constant hydrogen
Accepted 30 June 2014
consumption and light hydrocarbon yields in hydrogen consumer models, in order to reduce the com-
Available online 12 August 2014
plexity of the problem. As a result, interactions among process operating conditions, product quality
and yields, and hydrogen consumption are usually neglected.
Keywords:
The present work is an integrated approach for refinery process and hydrogen network design. Empir-
Process model integration
Hydrogen utilisation
ical correlations for sulphur prediction in liquid products are modified and adopted to predict hydrogen
Hydrogen network consumption in hydrotreaters. The model is validated with different feedstock properties and shows good
Design agreement with experimental data. Light hydrocarbon yields in hydrodesulphurisation reactions are also
Optimisation predicted and integrated in the network model. Modelling and optimisation of the overall network is per-
formed and the effects of process and operational variables on performance indicators and hydrogen pro-
duction requirements are investigated. As a Nonlinear programming model, the overall network model is
optimised with the CONOPT solver in General Algebraic Modelling System (GAMS). As demonstrated in a
case study, by integrating hydrotreating models into multicomponent hydrogen networks, the focus of
refinery hydrogen management can now be shifted from minimising hydrogen consumption to optimis-
ing hydrogen utilisation to improve refining profitability.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction or increase in hydrogen purity for reuse in hydrogen consumers.

The most commonly used processes for hydrogen purification are
A hydrogen system consists of hydrogen production, hydrogen pressure swing absorption, membrane units and cryogenic separa-
consumption, hydrogen purification, hydrogen compression, a fuel tion. Usually, pressure requirements by hydrogen consumers or
system and a distribution network itself. The primary sources of purifiers are satisfied using compressors. Normally, these units
hydrogen in a refinery are catalytic reformers, ethylene plants, are considered highly interactive whole within a distribution
hydrogen plants and hydrogen imports. The most common hydro- framework, otherwise refinery margins are depleted.
gen consumers are hydrotreaters and hydrocrackers. Hydrogen Today’s refineries face an increasing challenge of meeting grow-
containing streams such as hydrotreater off-gases or excess ing demand for cleaner fuels. The need to meet required end prod-
hydrogen streams that are currently being sent to fuel or feed to uct specifications from crude oil has necessitated the increased use
hydrogen plants could be routed to purifiers for hydrogen recovery of hydrogen in hydroprocessing operations, with existing hydrogen
production capacities often being a bottleneck. In recent times,
⇑ Corresponding author. stricter regulations on sulphur specifications and implementation
E-mail address: nan.zhang@manchester.ac.uk (N. Zhang). strategies have been imposed on refiners, hence presenting a major

http://dx.doi.org/10.1016/j.apenergy.2014.06.080
0306-2619/Ó 2014 Elsevier Ltd. All rights reserved.
170 B. Umana et al. / Applied Energy 133 (2014) 169–182

Nomenclature

Ea activation energy (kJ/Kmol) Components

FF feed flow (t/h) C1 methane
LP liquid product (t/h) C2 ethane
MWH2S molecular weight of H2S C3 propane
MWS molecular weight of S C4 butane
PH2 recycle hydrogen partial pressure (bar) C5 pentane
R universal gas constant PC pseudocomponent
Fmix flowrate of mix (t/h)
Fmu flowrate of makeup (t/h) Indices
Fre flowrate of recycle (t/h) i all components
Fri flowrate of reactor inlet (t/h) j hydrogen consumer
Ffeed flowrate of feed (t/h) j1 hydrogen consumer
Ffi flowrate of flash inlet (t/h) k hydrogen producer
Yfs amount of sulphur in the product (ppmw)
Yvap vapour phase purity (mass fraction) Superscripts
Yliq liquid phase purity (mass fraction) L lower bound
Fliq flowrate of liquid (t/h) U upper bound
Fpu flowrate of purge (t/h)
Fpr flowrate of external recycle (t/h)
FSF flowrate of site fuel (t/h)
FH2P flowrate of hydrogen producer (t/h)
FH2C flowrate of hydrogen consumer (t/h)

challenge in deep desulphurisation of petroleum fractions. Refinery networks [3]. Hydrogen pinch analysis gained significance in
processes such as hydrotreating, hydrocracking and catalytic industries and was extended to other areas of research [4].
dewaxing are dependent on the operating severity and availability Some underlying practical constraints exist in the application of
of hydrogen at adequate high pressure and purity. However, these graphical methods to the design of hydrogen distribution net-
processes are often not run under optimised conditions, using works. Hallale and Liu [5] extended Alves [2] Linear Programming
excess or inadequate hydrogen, and allowing hydrogen and associ- (LP) technique to a Mixed Integer Non Linear Programming (MIN-
ated impurities to be sent to purge and used as fuel. Consequently, LP) formulation that takes into account pressure constraints. Liu
effective hydrogen utilisation has become very important to refin- and Zhang [6] developed an automated design superstructure
ery hydrogen management. The present strategy would address approach that demonstrates the choice of purifiers as well as their
two major issues: 1. integrating hydrotreating performance in integration in the hydrogen networks. The objective function for
multicomponent hydrogen network optimisation; 2. evaluating the MINLP problem could be minimum hydrogen utility, operating
the optimum hydrogen requirements for sulphur removal and light costs or the total annualised cost of the network. Ahmad et al. [7]
hydrocarbon production based on quality constraints. Such an extended the MINLP model developed by Liu [8] to multi-period
approach in the design of integrated refinery hydrogen networks hydrogen network designs. Kumar et al. [9] introduced variable
could present opportunities to exploit various additional degrees inlet and outlet pressure configuration to the network model
of freedom in a network for improved hydrogen utilisation. developed by Hallale and Liu [5] to obtain realistic solutions. Liao
et al. [10] proposed a systematic approach for the location of com-
2. Review of previous research pressors and purifiers that accounted for other structural possibil-
ities. Although some of these methods take into consideration
Over the years, the pursuit of hydrogen management programs practical constraints, other constraints such as the concentrations
in the design of refinery hydrogen networks has been studied in of hydrogen sulphide (H2S) in hydrogen streams and their effects
two categories: on processing equipment are neglected. Recently, Zhou et al. [11]
incorporated an H2S removal unit in a hydrogen network to allow
I. Graphical targeting methods reuse of hydrogen rich streams. The targeting and mathematical
II. Mathematical programming methods design approaches have shown the importance of hydrogen saving
in distribution systems.
Towler et al. [1] developed the first systematic approach to ana- A major assumption in the aforementioned approaches is that of
lyse hydrogen distribution systems by proposing a graphical tool to a binary mixture of hydrogen and methane, rather than a multicom-
provide an overview of the costs and values associated with recov- ponent stream consisting of a mixture of impurities such as light
ering hydrogen from hydrogen producing or consuming processes. hydrocarbons, and compounds of sulphur and nitrogen in combina-
Using this concept of cost and value, the driving force for hydrogen tion with hydrogen. Such assumptions could result in loss of accu-
transfer is the difference between the cost of hydrogen available racy and does not account for the influence of impurities on the
and the value added to refinery products. Although, this method hydrogen partial pressure in a system. Singh and Zhang [12] devel-
assesses the economic trade-offs in recovery cost and value added, oped a systematic methodology to incorporate the impact of impu-
it does not provide a systematic approach to the distribution of rities, including hydrogen sulphide (H2S) and methane (CH4,) by
hydrogen in a refinery hydrogen network. Therefore, the concept integrating a high pressure flash model to account for vapour-liquid
of hydrogen recovery was redefined by Alves [2] to target the min- equilibrium characteristics within hydrogen consumers and the
imum hydrogen utility flowrate requirements of hydrogen net- effect on an overall hydrogen network. This method, however,
works using pinch analysis applications of heat exchanger requires a series of iterations between simulation and optimisation
 B. Umana et al. / Applied Energy 133 (2014) 169–182 171

resulting in increased computational effort, which may limit appli- tions and increasing load on hydrogen distribution systems, it
cability in real systems. Jia and Zhang [13] introduced a more effi- becomes imperative to integrate hydrotreating performance in
cient approach to multicomponent optimisation of refinery the overall optimisation of refinery hydrogen networks. An impor-
hydrogen networks by assuming constant vapour-liquid equilib- tant aspect of this work is the consideration of two key hydrotreat-
rium ratios for slight changes in the flash inlet stream composition. ing process performance indicators, namely sulphur removal and
The Non-Linear Programming (NLP) methodology for multi- light hydrocarbon yield.
component hydrogen network neglects the performance of hyd- Fig. 1 demonstrates schematically the impacts from key process
rotreaters by fixing the hydrogen consumption. Consequently, variable inputs and operating parameters on the outlet conditions
the effect of variations in hydrogen partial pressure, hydrogen– in a typical hydrodesulphurisation (HDS) unit.
oil-ratio, operating conditions and hydrogen consumption on A liquid hydrocarbon feed stream is mixed with a gas rich in
required sulphur specifications, and the overall network perfor- hydrogen, heated and fed to the HDS reactor. The reactor operates
mance is not considered. Moreover, light hydrocarbon productions at a desired hydrogen partial pressure depending on targets for max-
associated with HDS reactions are not integrated in the hydrogen imum sulphur. Hydrogen is consumed in the removal of sulphur to
network model. With the objective being the minimum hydrogen yield light hydrocarbons including methane, ethane, propane,
production flowrate, hydrogen requirements may be misleading butane and pentane, and hydrogen sulphide is produced during
as constraints on maximum allowable sulphur in the product the reaction. The reactor effluent is cooled and routed to a high pres-
stream are ignored. sure flash separator. Part of the gas released from the separator could
This paper aims at providing a framework to assess the interac- be recompressed and purified depending on hydrogen purity and
tions between hydroprocessing reactions and hydrogen distribu- pressure requirements of the reactor. The gas is recycled to the reac-
tion systems by exploiting key degrees of freedom to improve tor inlet and mixed with the make-up stream. The remainder of the
overall network performance. It also demonstrates the impacts of gas stream is purged to prevent build-up of contaminants in the
changes in light hydrocarbon composition in the recycle and purge recycle loop. Normally, the purge stream is either reused as external
streams from hydrogen consuming processes on the hydrogen recycle or sent to the fuel system. The liquid stream containing unre-
requirements. An integrated approach in the design of refinery acted sulphur compounds, dissolved hydrogen, and light hydrocar-
hydrogen networks could maximise the efficient utilisation of bons, is sent to a downstream separator section.
hydrogen and the effective operability of hydroprocessors. In the network, there are three key locations to an HDS process,
namely, reactor inlet, HDS reactor, and flash separator. The interde-
3. Integrated design of refinery process networks pendence of these factors and their intermediate streams in a
hydrogen network are achieved by exploiting the behaviour of
Hydrogen sources such as hydrogen plants, catalytic reformers, the system. Consequently, models of hydrogen consumers that
and off-gas from hydrogen consuming processes affect hydrogen are sufficiently detailed to capture these important operating fea-
consumers by providing hydrogen at different flowrates and puri- tures of the system are required in the early stage of design of a
ties. Changes in composition of liquid hydrocarbon feedstock and refinery hydrogen network. The methodology developed in this
make-up hydrogen could result in changes in the composition of work proposes a generic representation of process models embed-
recycle and purge streams from hydroprocessors. Consequently, ded into a hydrogen network superstructure to yield an integrated
light hydrocarbons and other impurities are formed that could process and network configuration. Three key challenges exist in
affect hydrogen inlet purity, thus affecting process performance. this methodology: development of hydroprocessor models that
In the same manner, changes in operating conditions in the process can predict process performance and hydrogen consumption; inte-
also affect the amount of light hydrocarbons formed and hydrogen gration of process models in the hydrogen consumer model; and
production requirements. With the recent trends in fuel specifica- optimisation of hydrogen networks with process models.

Fig. 1. An extended hydrogen consumer model showing different interactions in the network model.
172 B. Umana et al. / Applied Energy 133 (2014) 169–182

3.1. Development of hydroprocessor models and El Kady et al. [19] respectively. The parameters were obtained
by non-linear least square fitting on a MATLAB platform. The results
Generally, hydroprocessor models are developed to accurately from the model showed a fair fit with experimental data as shown
predict the behaviour of a system from changing input or operating in Figs. 2–4.
variables in a hydrogen consumer. In this work, a process–product The fit obtained may be a result of the model restrictions to
model is modified to represent the characterisation of product highly aromatic feeds that contain large amounts of aromatic sul-
stream composition as a function of liquid feedstock properties, phur. Based on the graphical results from the model in Eq. (1), a
HDS reactivities and operating conditions. Two aspects of hydro- generalised form of the model is produced that is capable of pre-
processing models are studied: namely sulphur removal and light dicting sulphur content in the product to a reasonable accuracy.
hydrocarbon formation. A second parameter b was introduced to the model to reflect the
nature of the feed, structural characteristics of the sulphur
3.1.1. Sulphur removal
Sulphur compounds are one of the most important impurities in
various petroleum fractions that cause deleterious effects on pro-
cesses and the environment. Often sulphur compounds in distillate
and residue fractions could exist as mercaptans, thiols, sulphides,
disulphides, polysulphides, thiophenes and their derivatives such
as benzothiophenes (BTs) and dibenzothiophenes (DBTs). A num-
ber of authors have reported the difficulty in removal of sulphur
from DBTs to mercaptans in descending order and proposed empir-
ical models to predict HDS performance. Shih et al. [14] proposed a
temperature correlation to predict the temperature requirements
for the desulphurisation of distillates to 0.05 wt% sulphur under
typical HDS conditions. Ho [15] developed a property-reactivity
correlation between feedstock quality and HDS reactivity. Chemo-
metric analysis was applied to identify important feedstock prop-
erties that affect HDS reactivity.
Based on the above work, the key variables affecting the compo- Fig. 2. Model fitting with diesel experimental data.
sition of residual sulphur in liquid products are given as follows:
8 9
> Feed properties >
>
> >
>
>
> TS; RS; nitrogen and aromatic content; API >
>
>
> >
>
>
< >
=
Process Operating Conditions
Sproduct ¼
>
> T; P; LHSV; recycle purity; hydrogen  oil ratio >
>
>
> >
>
>
> Process Chemistry >
>
>
> >
>
: ;
HDS reactiv ity; inhibition constants

TS, RS, T, P, and LHSV are total sulphur, refractory sulphur, temper-
ature, pressure and liquid hourly space velocity, respectively. It
could be inferred that process yields and product properties largely
depend on the feed properties, operating conditions and HDS chem-
istry. Such prediction of properties may be inaccurate without the
fundamental chemistry models. Based on the principles of first-
order kinetics, Choudhary et al. [16] derived an empirical equation
to demonstrate that the desulphurization chemistry of heavy oils
was strongly inhibited by three and larger ring aromatic hydrocar-
Fig. 3. Model fitting with vacuum gas oil experimental data.
bon content rather than DBTs in the feed.
  
kj PH2;i;j /
Sprod;j ¼ Sfeed;j  exp ð1Þ
ðKÞ  ð3 þ Rj Þ LHSV j

where Sprod;j = sulphur in the product from hydrotreater j, ppmw;

Sfeed;j = sulphur in the feed to hydrotreater j, ppmw; kj = rate con-
stant of HDS reaction in hydrotreater j, h1; K = 3 + ring aromatic
inhibition constant, 3 þ Rj = 3 + ring core aromatic content to hyd-
rotreater j, ppmw; PH2;i;j = recycle hydrogen partial pressure in hyd-
rotreater j, bar; and a = pressure dependent term. An excellent
correlation was observed between the sulphur conversion (644 K
and 6.9 MPa) for eight different heavy oil feeds and 3 + ring aro-
matic content. The model was tested on diesel, 75% straight run
gas oil (SRGO) and 25% light cycle oil (LCO), vacuum gas oil (VGO)
and kerosene feedstock at reactor operating conditions of 633 K,
60 bar and 1 h1; 653 K, 53 bar and 2 h1; 623 K, 60 bar and 5 h1
respectively. To predict the sulphur content in the product, the
pressure dependence parameter, a was tuned based on the avail-
able experimental data from Knudsen et al. [17], Ancheyta [18] Fig. 4. Model fitting with kerosene experimental data.
 B. Umana et al. / Applied Energy 133 (2014) 169–182 173

Table 1
Parameters obtained from the modified HDS model.

Hydrotreaters Sfeed (ppmw) Sprod (ppmw) 3 + R (ppmw) T (K) P (bar) LHSV (h1) a b
VGO 20,000 8393.58 84,000 653 53 2 907.588 1.738
KHT 2900 208.68 2000 623 60 5 13.329 0.825
DHT 15,200 31.49 32,000 633 60 1 80.735 0.598

Table 2
Comparison of results obtained between the original and modified model.

Sprod (experiment) Sprod (original model) Difference (experiment– Error% Sprod (modified model) Difference (experiment– Error%
ppmw ppmw original) ppmw ppmw modified) ppmw
Diesel
55 100.01 45.01 81.83 53.34 1.66 3.03
45 60.51 15.51 34.48 44.63 0.37 0.83
35 36.62 1.62 4.62 37.34 2.34 6.68
30 22.16 7.84 26.15 31.24 1.24 4.14
28 13.41 14.59 52.12 26.14 1.86 6.64
22 8.11 13.89 63.13 21.87 0.13 0.59
20 4.91 15.09 75.46 18.30 1.70 8.50
Vacuum gas oil
8000 11703.97 3703.97 46.30 8396.69 396.69 4.96
10,000 11831.43 1831.43 18.31 9904.34 95.66 0.96
12,000 11960.28 39.72 0.33 11682.70 317.30 2.64
14,000 12090.53 1909.47 13.64 13780.36 219.64 1.57
16,000 12222.21 3777.79 23.61 16254.67 254.67 1.59
Kerosene
370 687.12 317.12 85.71 387.55 17.55 4.74
310 334.47 24.47 7.89 284.38 25.62 8.26
230 162.81 67.19 29.21 208.68 21.32 9.27
130 79.25 50.75 39.04 153.13 23.13 17.79
100 38.57 61.43 61.43 112.37 12.37 12.37

compounds present in the feed and the product characteristics. original model and the average error is 39% and 5% for the original
With the introduction of b, Sprod is no longer directly proportional and modified model respectively.
to Sfeed. The fit obtained from the modified models are shown in
   Figs. 5–7.
kj PH2;i;j aj
Sproduct;j ¼ Sbfeed;j  exp ð2Þ
ðKÞ  ð3 þ Rj Þ LHSV j 3.1.2. Light hydrocarbon production
Table 1 shows the differences in parameters obtained from the Previous work by Singh and Zhang [12] and Jia and Zhang [13]
three different feedstocks to diesel hydrotreater (DHT), vacuum was based on fixed hydrocarbon production and hydrogen con-
gas oil hydrotreater (VGOHT) and kerosene hydrotreater (KHT). sumption. Light hydrocarbons are produced from hydrotreating
As observed, the b value increases from 0.598 for DHT to 1.738 (example in the desulphurisation of distillates and residues) and
for VGO. Although, it is expected that the b value for KHT is less hydrocracking reactions. Light gas refers to any gaseous or semi-
than that for DHT due to density difference, the amount of sulphur gaseous molecule with a molecular weight that is less than or
in KHT product is high compared to DHT product, thus attributing equal to that of pentane. Hasenberg and Campagnolo [20] imple-
b value to the difficulty in the removal of sulphur in KHT. This mented a light hydrocarbon model to obtain C1–C5 hydrocarbons
trend could be assigned to the decreased residence time and low in an atmospheric residue unit. In this work, an empirical model
reactor temperature in KHT. With reference to other feedstocks, to predict the production of light hydrocarbons during HDS reac-
it is possible to deduce that b increases with the complexity of tions has been developed based on refinery data after several trials
the feed and structural forms of sulphur compound present in and implemented. The model captures key operating conditions in
the feed if operating conditions are maintained. In view of this, a hydrotreater such as sulphur conversion, reaction pressure, LHSV
the introduction of b in the model enhances the accuracy of the and temperature. The light hydrocarbon model in Eqs. (3)–(5) was
empirical model. developed and implemented based on refinery industrial data with
Also, the a value in KHT is small in comparison with the other three consecutive periods having slightly different operating con-
hydrotreaters. An inference could be derived from the resulting ditions. Due to confidentiality issues, it has been difficult to obtain
coefficients which state ‘‘the correlation between the a value and more data points from this refinery or similar industrial data.
polyaromatic content of the feed is directly related.’’ Since the a Experimental data of light hydrocarbon production from hydro-
term is associated with pressure, an extensive study that involves treating process are also quite limited. In future work, the validity
wide range of operating conditions may be conducted to assess of this model would be tried and tested with more industrial data
variations in a value. The modified model was tested on the same as soon as they are available to us.
experimental data, and resulted in a better fit in all feedstock.
   
Ea Pj  e
Table 2 shows the comparison in errors obtained from the two C formed;i;j ¼ ci;j  aconv j  k0;i;j  exp  i;j
RT j LHSV j
models. As seen the maximum percentage error obtained from the
8 i ¼ C1 ! C5 ð3Þ
modified model in all feedstock is 17% compared to 85% for the
174 B. Umana et al. / Applied Energy 133 (2014) 169–182

Fig. 5. Modified model fitting with diesel experimental data.

Fig. 7. Modified model fitting with kerosene experimental data.

fixed sulphur requirements. The unused hydrogen remaining in

the reactor outlet stream is usually recycled with associated light
gases and hydrogen sulphide. From Eq. (3), one of the key con-
straints in the production of light hydrocarbons is the maximum
sulphur specification in the product stream, which sets the amount
of sulphur removed, and subsequently the amount of hydrogen
consumed and light hydrocarbons formed. For hydrogen consum-
ing processes, increasing or decreasing the maximum sulphur in
a hydrotreated product could affect the recycle hydrogen purity
in Eq. (2).

3.2. Integration of reaction correlations in hydrogen network

modelling

An extended hydrogen consumer model is developed from inte-

grating process models for sulphur removal and formation of light
hydrocarbons in the hydrogen consumer. The integration method-
Fig. 6. Modified model fitting with vacuum gas oil experimental data. ology applied to this work opens up opportunities to exploit vary-
ing degrees of freedom that could improve the performance of the
  overall integrated optimisation framework. This work presents the
Sprod;j
aconv j ¼ 1  ð4Þ mathematical formulation and optimisation of integrated multi-
Sfeed;j
component network under variable operating conditions and prod-
uct constraints to satisfy product properties.
MWH2 S  
H 2 Sj ¼  Sfeed;j  FF j  Sprod;j  LP j ð5Þ
MWS 3.2.1. Reactor inlet constraints
where Cformed,i,j = amount of light hydrocarbons i formed in hydrotr- The reactor inlet constraints include the overall mass and mul-
eater j, t/h, ci,j = yield coefficient of light hydrocarbons formed in ticomponent hydrogen balance which ensure that each hydrogen
hydrotreater j, t/h, aconvj = sulphur conversion in hydrotreater j, sink is supplied with the flowrate and purity requirements of the
ko,i,j = temperature dependence parameter, otherwise referred to as hydrogen consumer. The inlet of a reactor is made up of three
the frequency factor for light hydrocarbons i formed in hydrotreater streams, namely liquid feedstock, make-up hydrogen and recycles
j, and e is a pressure dependent term. Eq. (5) describes the formation hydrogen streams. The material balance at the mix point is given
of H2S in hydrotreater j obtained from stoichiometry as in [20]. by:
Fig. 8 shows the amount of light hydrocarbons produced in an
F mix;j ¼ F mu;j þ F re;j 8j ð6Þ
industrial diesel hydrotreater at low operating pressures, approx.
240 Nm3/m3, 592 K, and 2.24 h1.
F mix;j  Y mix;i;j ¼ F mu;j  Y mu;i;j þ F re;j  Y re;i;j 8i 8j ð7Þ
The result from the model showed good agreement with the
refinery data as illustrated in Fig. 9. where F, Y and subscript j represents the flowrate of a stream, pur-
A generalised reduced gradient approach is used to estimate the ity of a stream in mass fraction and hydrotreater respectively. Sub-
parameters as shown in Table 3. script i represents other components in the stream such as H2, H2S,
The yield parameter, ci;j in Table 3 defines the amount of light CH4, C2H6, C3H8, C4H10, C5H12 as well as pseudo-components
hydrocarbons formed. Butane has a high yield compared to other obtained from simulated distillation profile and bulk density of feed
hydrocarbons. This behaviour could be inferred from the tendency streams.
of heavier hydrocarbons to form butanes during cracking, although Similarly, the amount of gas supplied to the reactor inlet must
hydrocracking effect is not considered in the hydrotreating reac- be equal to the flowrate at the mix point and the flowrate of liquid
tions. In the context of this paper, the major function of a hydrotr- feedstock for all hydrotreaters.
eater is to consume hydrogen for desulphurisation reactions at
certain flowrates and purity to produce low sulphur products at F ri;j ¼ F mix;j þ F feed;j 8j ð8Þ
 B. Umana et al. / Applied Energy 133 (2014) 169–182 175

Fig. 8. Light hydrocarbons formed in different periods.

Fig. 9. Model fitting to the refinery data in Fig 8.

Table 3
Parameters obtained from the light hydrocarbon model. shown in Eq. (2). Changing recycle hydrogen purity would also
Light hydrocarbons/model parameters ci,j (t/h) Ko Ea (kJ/Kmol) affect the hydrogen–oil ratio at the reactor inlets as shown in Eq.
(7) and consequently, the light hydrocarbons formed and hydrogen
Methane 4.0066 0.1550 365.3463
Ethane 8.5276 0.0994 452.7053 consumed in Eqs. (10) and (11).
Propane 25.2526 0.1266 407.6503
Butane 28.8097 0.1153 424.7202
F fi;j  Y fi;i;j ¼ F ri;j  Y ri;i;j  RHi;j 8 i ¼ H2 8 j ð10Þ
Pentane 12.4948 0.4589 152.6564
F fi;j  Y fi;i;j ¼ F ri;j  Y ri;i;j þ Z i;j 8 i ¼ C 1 ! C 5 ; H2 S 8 j ð11Þ
The variables RHi;j and Z i;j are defined as consumption of hydrogen
F ri;j  Y ri;i;j ¼ F mix;j  Y mix;i;j þ F feed;j  Y feed;i;j 8i 8j ð9Þ and production of light hydrocarbons including H2 Sj respectively.
The values of RHi;j and Z i;j are obtained from the process models
Since the feed flowrate is constant, and the mass composition of in Eqs. (2)–(5). The amount of H2 Sj formed is calculated in Eq. (5)
hydrogen in the feed is zero, the pure hydrogen flowrate, and the stoichiometric hydrogen consumed from H2 S formation is
F ri;j  Y ri;i;j ; is equivalent to the product of flowrate and purity at derived from the ratio of molecular weights of hydrogen and hydro-
the mix point, 8 i = H2. Initially, the flowrate and purity of the mix- gen sulphide. Similarly, the amount of hydrogen consumed in the
ture are fixed in order to simulate the base case conditions. formation of light hydrocarbons is obtained from stoichiometry.
The total hydrogen consumed is calculated from Eq. (12).
3.2.2. Reaction model RHi;j ¼ RH2 S;i;j þ RC formed;i;j 8 i ¼ H2 8 j ð12Þ
Together with Eqs. (2)–(5), the following constraints define the
overall mass and component balances around the hydrotreater. A where RH2 S;i;j = amount of hydrogen consumed to form H2 Sj and
major feature of this work is the variation in hydrogen consump- RCformed;i;j = amount of hydrogen consumed in the formation of light
tion and light hydrocarbon production as a result of different hydrocarbons in hydrotreater j. From Eq. (2), sulphur requirements
hydroprocessing requirements. These requirements affect the in the product are assigned upper bound and lower bound con-
hydrogen partial pressure (referred to as the product of operating straints in each hydrotreater depending on the scope of optimisa-
pressure and recycle hydrogen purity) of hydrotreating units as tion as in Eq. (13).
176 B. Umana et al. / Applied Energy 133 (2014) 169–182

X X
Y Lfs;j 6 Y fs;j 6 Y Ufs;j 8j ð13Þ ðF H2 P;k;j  Y H2 P;i;k Þ þ ðF H2 C;j1;j  Y H2 C;i;j1 Þ ¼ F mix;j  Y mix;i;j 8j
k j1
As a result, hydrogen–oil ratio is allowed to vary to expand the
ð27Þ
search region of the optimisation, but controlled by maintaining
the recycle flowrate. Also, if there are changes in composition of Eq. (26) shows that the sum of hydrogen production flowrates from
the make-up streams from different hydrogen producing sources, various hydrogen producers, k, externally recycled gas flows from
changes in composition of recycle and purge streams from light other consumers, jl, and internally recycled gas flows within con-
hydrocarbons formed and hydrogen consumed in hydrotreating sumer, j, is equal to the gas flowrate at the reactor inlet mix point
reactions, fixing reactor inlet conditions would not be useful to for consumer, j. Combining with Eq. (27) for multicomponent mass
the hydrogen network optimisation. A key advantage of the present balance, the complete mass balance between hydrogen producers
approach is that hydrogen–oil ratio and hydrogen partial pressure and consumers is obtained.
are allowed to vary, which is beneficial in exploiting optimisation In a hydrogen network, hydrogen consumers require hydrogen
scenarios. at certain flowrates and purities. The purge gas from the high pres-
sure separators of various hydrogen consumers are either reused in
3.2.3. Flash model other consumers or sent to a site fuel system. The relationship
The outlet stream obtained from the reactor is routed to the between hydrogen consumers and site fuel system is represented
high pressure flash separator for the distribution of reactor prod- by the following mass balance equations:
ucts into vapour and liquid phases. The vapour phase could be X
internally recycled in a complete recycle operation or externally F pu;j ¼ F SF;j þ F H2 C;j;j1 j–j1 8 j ð28Þ
j1
recycled in the form of purge to other hydrogen consumers or
sent to the fuel system. The liquid phase are either routed to a X
low pressure separator for further recovery of hydrogen and F pu;j  Y pu;i;j ¼ F SF;j  Y SF;i;j þ ðF H2 C;j;j1  Y H2 C;i;j Þ j–j1 8 j ð29Þ
j1
light gases or prepared for fuel. Constant equilibrium distribu-
tion coefficient (K-value) strategy proposed by Jia and Zhang
Y pu;i;j ¼ Y SF;i;j ¼ Y H2 C;i;j 8i 8j ð30Þ
[13] within narrow composition interval is applied in this work.
Accurate prediction of K-values could provide effective reactor Eq. (28) shows that the purge gas from the flash outlet could be rou-
modelling and improve the accuracy of process simulation and ted to other consumers, as an external recycle or to the fuel gas
performance. Assuming vapour and liquid leaving a flash unit system.
are in equilibrium: Normally, the flowrate from a hydrogen producer are subject to
Y v ap;i;j ¼ Y liq;i;j  K i;j 8i 8j ð14Þ certain maximum or minimum limits as shown in Eq. (31).
X
Mass balance around the flash separator unit is given by the follow- F LH2 P;k 6 F H2 P;k;j 6 F UH2 P;k ð31Þ
j
ing equations:
F fi;j ¼ F re;j þ F liq;j þ F pu;j 8j ð15Þ By combining Eqs. (1)–(31), a process network model is formulated that
comprises non-linear empirical models for prediction of sulphur in
F fi;j  Y fi;i;j ¼ F re;j  Y re;i;j þ F liq;j  Y liq;i;j þ F pu;j  Y pu;i;j 8i 8j ð16Þ product, light hydrocarbon formation and hydrogen consumption,
and the mass balances of hydrogen producers and consumers in a net-
F fi;j ¼ F v ap;j þ F liq;j 8j ð17Þ work. The objective function is to minimise operating cost that accounts
for the hydrogen production cost and fuel gas value as in Eq. (32).
F v ap;j ¼ F re;j þ F pu;j 8j ð18Þ " #
X X
Objectiv e ¼ MIN ðF H2 P;k  U H2 Þ  ðF SF;j  U SF;j Þ ð32Þ
Y v ap;i;j ¼ Y re;i;j ¼ Y pu;i;j 8i 8j ð19Þ k j

X where U H2 and U SF , represent the unit prices of hydrogen and fuel

Y v ap;i;j ¼ 1 8 j ð20Þ gas respectively. Other costs that are integral in refinery process
i
economics have not been considered in this present work. A signif-
X icant aspect of this work is to develop a strategy for hydroprocessor
Y liq;i;j ¼ 1 8 j ð21Þ
integration in multicomponent hydrogen networks and their
i
impacts on hydrogen production requirements and fuel gas loss.
F re;j ¼ F H2 C;j1;j 8 j ¼ j1 ð22Þ The additional process constraints proposed in the formulation of
this methodology are expected to give more realistic solutions.
F pu;j ¼ F pr;j þ F SF;j 8j ð23Þ
4. Integrated optimisation framework for refinery process
F pr;j ¼ F H2 C;j1;j 8 j–j1 ð24Þ networks – A multicomponent strategy

The proposed methodology which incorporates non-linear

3.2.4. Hydrogen network model
empirical hydrotreater models to predict the effect of changing
Normally, the makeup hydrogen to the hydrogen consumers
process variables on hydrotreater performance and its interactions
comes from hydrogen producing sources as in Eq. (25). The rela-
with the multicomponent hydrogen network model is shown in
tionship between hydrogen producers and consumers in a network
Fig. 10.
is represented by the network mass balance:
X The methodology can be summarised into three major steps.
F H2 P;k;j ¼ F mu;j 8j ð25Þ
k 4.1. Integration of process models in a multicomponent framework
X X
F H2 P;k;j þ F H2 C;j1;j ¼ F mix;j 8j ð26Þ The non-linear process models described in Section 3.1 are inte-
k j1 grated in the multicomponent hydrogen framework with flash
 B. Umana et al. / Applied Energy 133 (2014) 169–182 177

integration. Strictly speaking, the ultimate aim of designing a refin- oil ratio and hydrogen partial pressure, more degrees of freedom
ery process network is to produce distillate fuels that meet envi- could be exploited to satisfy the objective function, while obtaining
ronmental, as well as economic objectives. Consequently, optimal process performance.
neglecting the hydroprocessors in the optimisation of hydrogen
networks may result in solutions that are not applicable to real sys-
4.3. Operational optimisation: Integrated multicomponent design of
tems. By considering the integration of process performance in
refinery process networks
hydrogen network optimisation, the interactions between process
requirements and qualitative hydrogen distribution can be esti-
In addition to allowing boundary tolerance in hydrogen oil ratio
mated. In this step, operating conditions and hydrogen–oil ratio
and hydrogen partial pressure, a mix of lower reactor temperature
are fixed for hydrogen consuming processes; the integrated pro-
and purity requirements for the same amount of sulphur in prod-
cess models in the hydrogen consumer define the amount of
uct may be worth exploring for different hydrogen consumers.
hydrogen consumed in the process. Flash calculations using con-
Consequently, an integrated multicomponent hydrogen network
stant K-value strategy for slight compositional changes were inte-
for improved operating conditions could be developed dependent
grated in the network to improve the network behaviour and
on experimental models and limitations. Relaxing operating condi-
reduce complexity. However, changes in feed source, hydrogen
tions in one hydrogen consumer would normally improve the
source could result in large variation in K-values which could affect
interactions between other hydrogen consumers, and subse-
the accuracy of the solutions.
quently benefit hydrogen producer supply for an efficient operat-
ing mix. Therefore, an optimal combination of flowrate, hydrogen
purity and operating conditions is required to meet desired
4.2. Optimisation with varying process operating conditions
specifications.
Normally, operating conditions in refinery process changes are
allowed to vary within specific limits that are suitable for process 5. Case study
improvement. For example, changes in feedstock composition,
hydrogen inlet conditions, and reactor temperature result in vary- 5.1. Base case
ing reaction conversions for different sulphur specifications and
subsequently changes to light hydrocarbon production. The effect The hydrogen network base case is presented here to illustrate
of these slight changes are described in the resulting variation in the applicability of the proposed approach developed for the inte-
hydrogen consumption, flowrate and purity of recycle and purge grated design and optimisation of multicomponent networks. The
streams and overall hydrogen production in the network. By case study describes the simultaneous integration of process mod-
expanding the variable search in the network, such as hydrogen els and multicomponent hydrogen network models in the optimi-
sation of hydrogen network requirements. The objective of this
study is to determine the optimum hydrogen production flowrate
for different study scenarios.
Data The hydrogen network base case consists of two hydrogen pro-
ducers: Hydrogen plant, H2 Plant; catalytic reformer, CCR; and four
hydrotreaters: naphtha hydrotreater, NHT; cracked naphtha hyd-
rotreater, CNHT; diesel hydrotreater, DHT; and vacuum gas oil
hydrocracker, VGOHC. Table 4 shows the detailed feed stream data
Refinery process network for base case.
optimisation with integrated non The components listed in Table 4 are pseudo-components
linear process models for varying derived from the corresponding stream physical properties. Non-
process performance linear process models developed in Section 3.1 are integrated in
the multicomponent hydrogen network under fixed operating con-
ditions as shown in Table 5, fixed hydrogen–oil ratio, and varying
hydrogen consumption for different hydrotreaters based on sul-
phur requirements in the product resulting in an integrated base
case flowsheet in Fig. 11. Hydrogen cost from import is £ 3000/yr
Define range of operating [21].
conditions The hydrogen production flowrate obtained from the integrated
multicomponent process network is 11.771 t/h for fixed inlet
hydrogen conditions. The network diagram in Fig. 11 shows that
required sulphur specifications are not compromised while opti-
mising hydrogen allocation to different hydrotreaters. In the previ-
Optimization with variable ous integrated flash methodology without consideration to process
conditions or environmental demands [13], and therefore fixing hydrogen
consumption, approximately 22% of hydrogen was spent in the fuel
system. This present work demonstrates the utilisation of all
hydrogen produced within the processes; hence no hydrogen
was routed to the fuel system as indicated in Figs. 11–13. However,
hydrogen not required by the process from the integrated con-
Integrated multicomponent
network design under varying sumer models is dissolved in liquid product, which could be sub-
operating conditions ject to further low pressure separation. Integrating the hydrogen
consuming processes in the hydrogen network model opens up
opportunities to exploit various degrees of freedom available to
Fig. 10. Integrated optimisation frameworks for refinery process networks. optimise the process. For example, recycle hydrogen purity and
178 B. Umana et al. / Applied Energy 133 (2014) 169–182

Table 4 reactor inlet hydrogen flowrates have been allowed to vary within
Detailed feed data for base case. bounds, while maintaining the recycle flow as illustrated in Fig. 12.
Hydroprocessors NHT CNHT DHT VGOHC Fig. 12 shows that hydrogen production flow obtained
Feed flowrate (t/h) 175.041 75.886 337.563 260.044 decreases by approximately 2% to 11.547 t/h. As noticed, equilib-
Compositions (mass fraction) rium has been affected, thus resulting in a reduction of dissolved
H2 0.0000 0.0000 0.0000 0.0000 hydrogen in some hydrogen consumers where makeup hydrogen
C1 0.0000 0.0000 0.0000 0.0000 is decreased compared to base case. Note that the amount of light
C2 0.0000 0.0000 0.0000 0.0000
C3 0.0000 0.0000 0.0000 0.0000
hydrocarbons produced is same for both fixed and varying inlet
C4 0.0000 0.0000 0.0000 0.0000 hydrogen flow. With varying inlet configuration for hydrogen con-
C5 0.0000 0.0000 0.0000 0.0000 sumers, makeup hydrogen requirements could be decreased for
H2S 0.0000 0.0000 0.0000 0.0000 some hydrotreaters, resulting in a decrease in hydrogen production
PC1-NHT 0.1627 0.0000 0.0001 0.0000
flowrate. Table 6 describes the comparison between fixed and
PC2-NHT 0.2051 0.0000 0.0002 0.0000
PC3-NHT 0.2207 0.0000 0.0002 0.0000 varying inlet conditions.
PC4-NHT 0.2686 0.0000 0.0003 0.0000
PC5-NHT 0.1429 0.0000 0.0002 0.0000 5.3. Operational optimisation of multicomponent hydrogen network
PC1-CNHT 0.0000 0.0916 0.0000 0.0000
PC2-CNHT 0.0000 0.1675 0.0000 0.0000
Suppose it is required to optimise operating temperature, while
PC3-CNHT 0.0000 0.2422 0.0000 0.0000
PC4-CNHT 0.0000 0.2031 0.0000 0.0000 maintaining the same amount of sulphur in product as described in
PC5-CNHT 0.0000 0.2957 0.0000 0.0000 Fig. 13.
PC1-DHT 0.0000 0.0000 0.0745 0.0000 The resulting multicomponent integrated network shows a
PC2-DHT 0.0000 0.0000 0.1743 0.0000
decrease in H2 import by 9% from the fixed base case. As observed,
PC3-DHT 0.0000 0.0000 0.2630 0.0000
PC4-DHT 0.0000 0.0000 0.3502 0.0000 dissolved hydrogen has been reduced considerably in most hydro-
PC5-DHT 0.0000 0.0000 0.1371 0.0000 gen consumers. Slightly lower hydrogen outlet partial pressures
VGOHC-GA1 0.0000 0.0000 0.0000 0.0062 have been obtained for allowable decrease in reaction severity.
VGOHC-GA2 0.0000 0.0000 0.0000 0.0133 Table 7 describes the changes in the overall network.
VGOHC-GA3 0.0000 0.0000 0.0000 0.0281
Table 7 shows the change in import obtained from varying H2
VGOHC-GA4 0.0000 0.0000 0.0000 0.0192
VGOHC-GA5 0.0000 0.0000 0.0000 0.0140 inlet only compared with sequential temperature optimisation.
VGOHC-NAP1 0.0000 0.0000 0.0000 0.0213 An approximate 7% decrease in H2 import proves the overall ben-
VGOHC-NAP2 0.0000 0.0000 0.0000 0.0358 efit of temperature and purity optimisation.
VGOHC-NAP3 0.0000 0.0000 0.0000 0.0825
VGOHC-NAP4 0.0000 0.0000 0.0000 0.0738
VGOHC-NAP5 0.0000 0.0000 0.0000 0.0394
5.4. Environmental regulatory effects on multicomponent hydrogen
VGOHC-DIE1 0.0000 0.0000 0.0000 0.0756 network
VGOHC-DIE2 0.0000 0.0000 0.0000 0.1242
VGOHC-DIE3 0.0000 0.0000 0.0000 0.1872 Sometimes environmental regulators demand a more stringent
VGOHC-DIE4 0.0000 0.0000 0.0000 0.1831
sulphur specification, especially in diesel hydrotreaters. For exam-
VGOHC-DIE5 0.0000 0.0000 0.0000 0.0964
ple, the case of further sulphur reduction to 6 ppm as in DHT pro-
cess in Table 8. Is it really worthwhile to restrict sulphur limits to
lower levels?
For every reduction in sulphur, there is a simultaneous increase
Table 5
Base case operating conditions of hydroprocessors.
in temperature, as well as slight increases in outlet hydrogen par-
tial pressure. However, the makeup H2 requirement only increases
Hydroprocessors T (K) P (bar) LHSV (h1) insignificantly and thus the H2 import flow. Note that an increase
NHT 623 60 5 in light hydrocarbons formed is likely; hence building up impuri-
CNHT 653 65 1.5 ties in the system as seen in Fig. 14.
DHT 633 60 1
In some cases, the light hydrocarbon slope tends to be steeper
VGOHC 653 52.96 2
depending on the feedstock involved and rate of desulphurisation.
It is usually desirable to operate within an operational safety enve-
lope (OSE) such that a reduction in diesel sulphur specifications to
hydrogen–oil ratio are obtained from the required sulphur conver- a minimum is checked against increase in operating severity, for
sion and process operating conditions, and these set the hydrogen example reaction temperature, build-up of light hydrocarbons in
consumption and makeup hydrogen requirements of the hydro- the recycle loop and the overall effect on hydrogen production
processor, which translates into the hydrogen production flowrate flow. However, note that increase in reaction temperature could
of the network. The amount of total hydrogen consumed, sulphur necessitate a reduction in catalytic activity. If deactivation rate is
consumed and light hydrocarbons formed are mainly based on inversely proportional to the hydrogen partial pressure, then cata-
the HDS and light hydrocarbon process models coupled with stoi- lyst deactivation is expected to follow a downward trend under
chiometric relationships of HDS process for different sulphur dis- increasing hydrogen partial pressure conditions. However, rate of
tribution embedded in the hydrogen consumer. deactivation is still evident due to increasing temperature as in
Fig. 15.
5.2. Optimisation with varying hydrogen inlet conditions The approach adopted in the constrained optimisation is useful
in allowing effective use of hydrogen in hydrogen consuming pro-
For a multicomponent network configuration with fixed hydro- cesses. An optimal integrated multicomponent network have been
gen flowrate at the reactor inlet as in Fig. 11, there are limitations obtained that allows varying hydrogen inlet conditions, hydrogen
to manipulating the degrees of freedom existent in the integrated consumption based on sulphur specification constraints, and a
framework. For example, product requirements could be achieved mix of operating conditions that improves interaction between
at lower make up requirements when there are opportunities to hydrogen use and the overall optimisation of the multicomponent
tune the inlet hydrogen flow for a given property. In this section, hydrogen network.
 B. Umana et al. / Applied Energy 133 (2014) 169–182 179

Fig. 11. Optimised integrated multicomponent hydrogen network under fixed inlet hydrogen conditions.

Fig. 12. Integrated multicomponent network under varying inlet hydrogen conditions.

5.5. NLP hydrogen network optimisation ditions (temperature) for all hydrogen consuming processes at
lower hydrogen purities was explored for the same sulphur speci-
The NLP network model is optimised with the CONOPT solver in fication resulting in an additional benefit of approximately 7%. Fur-
GAMS. With fixed hydrogen–oil ratio and operating conditions, the ther optimisation based on reduction of sulphur target from
hydrogen plant production of the integrated multicomponent 15 ppm to 10 ppm was investigated in DHT process. The result
hydrogen network is 11.77 t/h as shown in Fig. 11. By extending shows that the effect on hydrogen import cost was quite significant
the integrated multicomponent network to capture varying inlet compared to a further decrease to 8 ppm and 6 ppm, which
hydrogen conditions under different sulphur requirements, the incurred zero costs in hydrogen import requirements as presented
hydrogen production flowrate was decreased to 11.55 t/h, result- in Table 8. Hence, there may be no real benefit in exploring lower
ing in a savings of 2%. In another scenario, varying operating con- sulphur requirements.
180 B. Umana et al. / Applied Energy 133 (2014) 169–182

Fig. 13. Integrated multicomponent networks with optimised CNHT.

Table 6
Comparison of fixed and varying H2 inlet conditions.

Cases Fixed and varying inlet H2 for fixed H2 consumption

Hydrogen consumers NHT CNHT DHT VGOHC
Fixed Vary Fixed Vary Fixed Vary Fixed Vary
Fixed reaction conditions 623 623 653 653 633 633 653 653
Fixed sulphur in product (wt%) 0.015 0.015 0.015 0.015 0.0015 0.0015 0.100 0.100
Makeup hydrogen (t/h) 0.556 1.334 0.649 0.827 2.496 2.457 8.404 7.275
External recycle H2 to consumer inlet (t/h) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Chemical hydrogen consumed (t/h) 0.398 0.398 0.426 0.426 1.845 1.845 0.908 0.908
H2 to external recycle/fuel system (t/h) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Dissolved hydrogen (t/h) 0.159 0.937 0.224 0.401 0.651 0.613 7.494 6.368
C1 formed (t/h) 0.021 0.021 0.021 0.021 0.020 0.020 0.011 0.011
C2 formed (t/h) 0.036 0.036 0.038 0.038 0.034 0.034 0.018 0.018
C3 formed (t/h) 0.161 0.161 0.167 0.167 0.153 0.153 0.082 0.082
C4 formed (t/h) 0.152 0.152 0.162 0.162 0.144 0.144 0.078 0.078
C5 formed (t/h) 0.082 0.082 0.085 0.085 0.077 0.077 0.042 0.042
H2 production flowrate – Fixed case (t/h) 11.771
H2 production flowrate – Varying case (t/h) 11.547
Cost of H2 saved from fixed case (£/yr) 5,376,000

Table 7
Comparisons of simultaneous variation of H2 inlet and reaction temperature conditions.

Cases Varying inlet H2 and varying T for same sulphur in product

Vary reaction conditions NHT CNHT DHT VGOHC
Vary H2 inlet Vary T Vary H2 inlet Vary T Vary H2 inlet Vary T Vary H2 inlet Vary T
Reactor operating temperature (K) 623 623 653 652.93 633 633 653 653
Fixed sulphur in product (wt%) 0.015 0.015 0.015 0.015 0.0015 0.0015 0.100 0.100
Makeup hydrogen (t/h) 1.334 1.240 0.827 0.823 2.457 2.457 7.275 6.640
External recycle H2 to consumer inlet (t/h) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Chemical hydrogen consumed (t/h) 0.398 0.398 0.426 0.426 1.845 1.845 0.908 0.908
H2 to external recycle/fuel system (t/h) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Dissolved hydrogen (t/h) 0.937 0.843 0.401 0.397 0.613 0.613 6.368 5.731
Pure recycle H2 at the outlet 6.319 6.319 8.990 8.988 39.679 39.679 10.971 10.971
C1 formed (t/h) 0.021 0.021 0.021 0.021 0.020 0.020 0.011 0.011
C2 formed (t/h) 0.036 0.036 0.038 0.038 0.034 0.034 0.018 0.018
C3 formed (t/h) 0.161 0.161 0.167 0.167 0.153 0.153 0.082 0.082
C4 formed (t/h) 0.152 0.152 0.162 0.162 0.144 0.144 0.078 0.078
C5 formed (t/h) 0.082 0.082 0.085 0.085 0.077 0.077 0.042 0.042
H2 plant flowrate (t/h) – Case I – varying H2 inlet 11.547
H2 plant flowrate (t/h) – Case II – varying H2 inlet, T 10.755
Cost of H2 saved from Case I (£/yr) 19,008,960
 B. Umana et al. / Applied Energy 133 (2014) 169–182 181

Table 8
Effect of sulphur restriction in DHT process on the overall network.

Cases Different sulphur content in product

Hydrogen consumer DHT
Reactor operating temperature (K) 633 639.93 643.48 647.84
Varying sulphur in product (wt%) 0.0015 0.0010 0.0008 0.0006
Makeup hydrogen (t/h) 2.457 2.458 2.458 2.459
External recycle H2 to consumer inlet (t/h) 0.000 0.000 0.000 0.000
Chemical hydrogen consumed (t/h) 1.845 1.845 1.846 1.846
H2 to external recycle/fuel system (t/h) 0.000 0.000 0.000 0.000
Dissolved hydrogen (t/h) 0.613 0.613 0.613 0.613
Pure recycle H2 at the outlet 39.679 39.720 39.743 39.768
H2S formed (t/h) 5.448 5.450 5.451 5.451
H2 import flowrate (t/h) 10.757 10.758 10.758 10.758
Cost of H2 increase (£/yr) 24,000 0 0

6. Conclusions

Until now studies based on refinery hydrogen management

have fixed hydroprocessor performance in the optimisation of a
hydrogen network, thereby neglecting the crucial interactions
between hydroprocessors and multicomponent hydrogen net-
works. The proposed methodology accounts for this deficiency in
the optimisation of integrated multicomponent process networks.
The interactions between varying hydrogen inlet conditions of
hydrogen consuming processes, optimising operating conditions
and constrained process performance indicators have been investi-
gated. The effect on light hydrocarbon generation produced from
HDS processes and catalyst activity have also been analysed. A
Fig. 14. A schematic representation of quality trend in hydrotreaters. potential saving of 2% was realised from varying H2 inlet condi-
tions and a further reduction of 7% was achieved by optimising
temperature. Exploiting such degrees of freedom in the network
opens up opportunities for allocating optimum operating condi-
tions to hydroprocessors, thereby increasing hydrogen utilisation
efficiency and optimising hydrogen production flowrate. By allow-
ing simultaneous consideration of hydroprocessor integration,
multicomponent hydrogen network optimisation, and varying
operating conditions, an actual and effective hydrogen optimisa-
tion methodology has been implemented.

Acknowledgement

The authors would like to acknowledge Petroleum Technology

Development Fund for their financial support granted.

References

Fig. 15. Temperature vs. deactivation rate. [1] Towler GP, Mann R, Serriere AJ, Gabaude CMD. Refinery hydrogen
management: cost analysis of chemically integrated facilities. Ind Eng Chem
Res 1996;35:2378–88.
By integrating hydrotreating models into multicomponent [2] Alves J. Analysis and design of refinery hydrogen distribution systems. Ph.D.
thesis, Department of Process Integration. University of Manchester Institute
hydrogen networks for improved network performance, optimisa-
of Science and Technology, Manchester, UK; 1999.
tion of hydrogen use, and exploitation of operating variables, the [3] Linnhoff B, Mason DR, Wardle I. Understanding heat exchanger networks.
focus of refinery hydrogen management can now be shifted from Comput Chem Eng 1979;1979(3):295.
[4] Zhao Z, Liu G, Feng X. The integration of the hydrogen distribution system with
minimising hydrogen consumption to optimising hydrogen
multiple impurities. Chem Eng Res Des 2007;85:1295–304.
utilisation. [5] Hallale N, Liu F. Refinery hydrogen management for clean fuels production.
A summary of the GAMS calculations for the overall integrated Adv Environ Res 2001;6:81–98.
multicomponent network is as follows: [6] Liu F, Zhang N. Strategy of purifier selection and integration in hydrogen
networks. Chem Eng Res Des 2004;82:1315–30.
[7] Ahmad MI, Zhang N, Megan J. Modelling and optimisation for design of
Computer resource 64-bit operating system, hydrogen networks for multi-period operation. J Clean Prod 2011;19:204–11.
[8] Liu F. Hydrogen integration in oil refineries. PhD. Thesis, Department of
6 GB RAM, 2.4 GHz Intel Core i5-2430M
Process Integration, University of Manchester Institute of Science and
No. of variables 38 Technology, Manchester, UK; 2002.
No. of equations 70 [9] Kumar A, Gautami G, Khanam S. Hydrogen distribution in the refinery using
CPU time 0.68 s (varies for different cases) mathematical modelling. Energy 2010;35:3763–72.
[10] Liao Z, Wang J, Yang Y, Rong G. Integrating purifiers in refinery hydrogen
networks: a retrofit case study. J Clean Prod 2010;18:233–41.
182 B. Umana et al. / Applied Energy 133 (2014) 169–182

[11] Zhou L, Zuwei L, Wang J, Jiang B, Yang Y. Hydrogen sulphide removal process [16] Choudhary TV, Parrot S, Johnson B. Unravelling heavy oil desulphurisation
embedded optimization of hydrogen network. Int J Hydrogen Energy chemistry: targeting Clean fuels. Environ Sci Technol 2008;42:1944–7.
2012;37:18163–74. [17] Knudsen KG, Cooper BH, Topsoe H. Catalyst and process technologies for ultra-
[12] Singh BB, Zhang N. Impact of gas phase impurities on refinery hydrogen low sulphur diesel. Appl Catal A 1999;189:205–15.
network management. 2005 AIChE Spring National Meeting: Conference [18] Ancheyta J. Modelling and simulation of catalytic reactors for petroleum
Proceedings. Atlanta, GA, USA; April 2005. p. 1469–80. refining. John Wiley & Sons 2011.
[13] Jia N, Zhang N. Multi-component optimization for refinery hydrogen networks. [19] El Kady FY, Mohamed LK, Shaban SA. Hydrotreatment of kerosene petroleum
Energy 2011;36(8):4663–70. fraction to improve its properties. Int J Acad Res 2010:2.
[14] Shih SS, Mizrahi S, Green LA, Sarli MS. Deep desulphurisation of distillates. Ind [20] Hasenberg DM, Campagnolo JF. Modelling and simulation of a reaction for
Eng Chem Res 1992;31:1232–5. hydrotreating hydrocarbon oil. Philips Petroleum Company; 1997.
[15] Ho TC. Property-reactivity correlation for HDS of middle distillates. Appl Catal [21] Blenco G. Hydrogen car revolution; November 2009. <www.h2carblog.com>.
A, General 2003;244:115–28.