See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/260943725

process simulation of the transport gasifier

Article in Fuel August 2013

CITATIONS READS

0 51

1 author:

Muhammad Tajammal Munir

University of Auckland
34 PUBLICATIONS 110 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Industrial high purity, high recovery distillation : modeling, optimization and control View project

Real-time quality control View project

All content following this page was uploaded by Muhammad Tajammal Munir on 16 October 2017.

The user has requested enhancement of the downloaded file.

 Fuel 115 (2014) 479489

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Process simulation of the transport gasier

Christopher J. Arthur, Muhammad Tajammal Munir, Brent R. Young, Wei Yu
Industrial Information and Control Centre (I2C2), Department of Chemical and Materials Engineering, The University of Auckland, Auckland 1010, New Zealand

h i g h l i g h t s

 A KBR transport gasier has been successfully simulated.

 Three alternate process simulation models were developed.
 Gibbs free energy based model performed the best among the three models.
 Models were validated using real data (i.e. KBR data).

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

Article history: The transport gasier manufactured by Kellogg, Brown and Root (KBR) is reportedly capable of econom-
Received 4 March 2013 ically converting low rank coal (e.g. lignite) to syngas for the production of synthetic chemicals, fuels and
Received in revised form 8 May 2013 energy. However no process simulation of the KBR transport gasier yet exists in the public domain lit-
Accepted 18 July 2013
erature. In this work three alternative process simulation models of the transport gasier were developed
Available online 1 August 2013
using a commercial process simulator combined with Excel/VBA routines. The rst model determined
gasication products on the basis of minimum Gibbs energy. The second model used pseudo-equilibrium
Keywords:
approach and the third model used kinetic expressions. The simulation models were validated with real
Transport gasier
Process simulation
process data. The pseudo-equilibrium model was best able to replicate the data with reasonable process
Gasication assumptions.
TRIG 2013 Elsevier Ltd. All rights reserved.

1. Introduction making the assumption that carbon reacts to completion simplis-

Lignite is a low rank, low value coal that is abundant in New
Zealand [3,32,19]. It is unsuitable for conventional combustion as
it has a low heat of combustion relative to higher ranks of coal.
However Lignite is particularly suited to gasication, with the right
technology, due to its high reactivity. High coal reactivity will al-
low a gasier that can operate at a lower temperature to be used
in a process. The reduced cost of the refractory will increase the
economic case [12].
The transport gasier manufactured by Kellogg, Brown and Root
(KBR) typically operates at temperatures 520 C cooler than the
market leading Shell coal gasication process [12] and is particu-
larly suited to lignite [16,17]. No process simulation for the KBR
transport gasier exists in the public domain literature.
The transport gasier is a recent technology and independent
sources describing the gasication reactor include Higman and
Burgt [12], Rezaiyan and Cheremisinoff [21] and Bell et al. [4]
where complimentary accounts are given. The gasication reactor
is described as having a carbon conversion of approximately 95%
Corresponding author. Tel.: +64 9 373 7599x85027; fax: +64 9 373 7463.
E-mail addresses: tajammal.munir@auckland.ac.nz (M.T. Munir),
auckland.ac.nz (W. Yu).
tic. It is notable that different coal ranks have a large effect on pro-
cess outcome. A more detailed publically available description of
the transport gasier is given by Mann et al. [16]. The review states
that the transport gasier does not expose fresh coal to the oxi-
dant. Re-circulated char is the source of combustion heat not vol-
atiles. This has important implications when dening the
simulation topology. However the review is qualitative and reveals
few quantitative performance attributes.
The transport gasier operates at relatively low temperatures,
making kinetic considerations important. A key publication by
Mann et al. [16] considered the char reaction kinetics in a transport
gasier. It was concluded that the reaction of char with steam was
the dominant carbon reaction and an Eq. (1) was presented that
can determine the conversion rate of carbon in the transport gas-
ier due to steam gasication.
 1:91
dX 2104 CO
1:4  107  e T  P0:63
H2 O  P 0:7
Total  1
ds CO CO2
where X is the carbon fraction, P the pressure (atm), T the temper-
ature (K), s is the residence time (min).
w.yu@ Eq. (1) will allow a model to determine the carbon conversion if
the residence time is known. However the equation was designed

0016-2361/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.fuel.2013.07.068
480 C.J. Arthur et al. / Fuel 115 (2014) 479489
using only three data points making the reliability difcult to
assess.
The KBR transport gasier also operates as a high velocity, cir-
culating, uidised bed reactor. The reactor has no internals or mov-
ing parts and is robust to various coal particle sizes. Solid char
particles circulate through the reactor until converted or ejected
with the ash. The high velocity circulation increases heat and mass
transfer and reduces the operating temperature [12,2,20]. Operat-
ing below the agglomeration temperature of coal ash prevents
refractory damage and the need for complex expansion joints,
thereby reducing the capital cost [8]. A schematic of the transport
gasier can be seen in Fig. 1.
The objective of this study is to create a model using a commer-
cial process simulator that can replicate the data provided by KBR
with reasonable assumptions. A process simulation capable of pre-
dicting mass and energy ows for lignite gasication using the
transport gasier would be a useful tool to evaluate the economic
case for the gasier and to assess sensitivity to process variation.
Previous studies into the gasication of coal gave a critical in-
sight into the capabilities and limitations of process modelling
and the assumptions included in the models. No process models
were found that considered the KBR transport gasier. A selection
of the other previous gasication models are presented below.
Nathen et al. [18] created a model of the Shell gasier using the
process simulator HYSYS. The coal was modelled as a ratio of car-
bon, water and methane and Gibbs energy minimisation formed
the basis of determining the products. Nathen et al. [18] also as-
sumed that the char reacted to completion, that the gasier is adi-
abatic, that equilibrium conditions were achieved quickly and
slagging of the coal ash required no energy. These assumptions will
have to be addressed to differentiate between gasiers. The model
could not account for kinetic limitations. A coal analysis with
greater resolution would be benecial to a gasication model for
chemical production as impurities in the syngas, particularly sul-
phates, affect downstream units [7].
Fig. 1. Transport gasier schematic illustrating the circulating uidised bed.
Gnanapragasam et al. [11] also used Gibbs energy minimisation
but using an ASPEN PLUS simulation platform. Heat loss was eval-
uated in the model and species stable above 873 K were included.
Gnanapragasam et al. [11] also assumed all char was reacted and
that there was no kinetic limitation. In the model nitrates formed
molecular nitrogen whereas ammonia and hydrogen cyanide are
coal gasication products and generally of higher concentrations
[12].
Ardila et al. [1] used a Gibbs energy minimisation simulator
nested within a recycle loop in ASPEN PLUS to simulate a circulat-
ing uidised bed gasier. The simulation demonstrated that using
a yield reactor to convert solid components to specied products
based on an ultimate analysis is a legitimate way of simulating
pyrolysis. The simulation assumed that there was no kinetic limi-
tation to gasication.
Kunze and Spliethoff [15] considered kinetic limitation in an
ASPEN PLUS gasication model by dening gasication as a series
of linearly-independent reactions and specifying a temperature ap-
proach to equilibrium for each reaction. The model was able to clo-
sely replicate empirical results. Departure from equilibrium was
specied to t a known output. The approach temperatures were
not published.
A report by Kramer [14] for the United States Department of En-
ergy published an ASPEN PLUS simulation with specied ap-
proach temperatures. The results showed signicant departures
from thermodynamic equilibrium for the reaction of char with
hydrogen and the water gas shift reaction. The model could simu-
late gasication with a high level of accuracy but the approach
temperatures did not have a theoretical basis.
An introduction to net ow concept and how it can affect the
hydrodynamics in a gasier has also been analysed [33,34]. From
these reported studies, it was concluded that the simulation results
without the net ow concept can deviate signicantly from the
practical results.
This manuscript is organised as follows. After this general intro-
duction, the materials and methods used in this work are explained
and discussed in Section 2. In Section 3 simulation methodology is
explained. In Section 4 results are discussed. Finally in Section 5 re-
sults are summarized, limitations are discussed, and conclusions
are made.
2. Materials and methods
2.1. Coal classication
The transport gasier is suited to young coals particularly lig-
nite. Coal rank with increasing geochemical maturity is shown in
Fig. 2 [5]. Lignite is the lowest rank of coal and the boundary be-
tween lignite and peat is unclear. Lignite resembles its original
plant matter to a greater extent than mature coals. Lignite is por-
ous and contains high fractions of oxygen and therefore functional
groups that affect reactivity. Relative to higher ranks, lignite has a
smaller proportion of aromatic carbon atoms, approximately half
of the total carbon, and a greater hydrogen content, approximately
1:1 (mole:mole) hydrogen to carbon [13,26].
2.2. Coal gasication
Generally, gasication is the conversion of a carbonaceous feed-
stock to a gas with a thermal heating value [12,17]. Gasication
differs from combustion as combustion products have no thermal
value. The gasication process can include: (i) pyrolysis; (ii) partial
oxidation; (iii) char reactions; and (iv) gas phase reactions.
 C.J. Arthur et al. / Fuel 115 (2014) 479489 481

Sub-
Sub-
Peat
Peat Lignite
Lignite Bituminous
Bituminous Anthracite
Anthracite
bituminous
bituminous

Fig. 2. Coal rank with increasing geochemical maturity.

(i) Pyrolysis: Pyrolysis involves exposing coal to high tempera- Exothermic methanation reactions:
tures under low oxygen conditions. The coal will thermally
Cs 2H2g CH4g  75 kJ=mol 10
decompose and undergo reactions to form a methane rich
volatile phase, liquid tar and a carbon rich solid char. Theo-
retical models that relate actual coal molecules to pyrolysis COg 3H2g CH4g H2 Og  206 kJ=mol 11
products are enormously complex and require knowledge of Exothermic water gas shift reaction:
the specic coal structure [12].
(ii) Partial oxidation: Partial oxidation of coal forms combustion COg H2 Og CO2g H2g  41 kJ=mol 12
products and signicant carbon monoxide. Carbon monox-
Steam reforming:
ide is a desired component of gasication for many applica-
tions, particularly synthetic fuel production and energy. The CH4g H2 Og COg 3H2g 206 kJ=mol 13
oxidation reactions are highly exothermic and a source of
energy for endothermic gasication [12]. The above reactions occur simultaneously. A useful approach to
(iii) Char reaction: In addition to combustion, solid char will calculating the equilibrium conditions when dealing with a simul-
react with steam to form hydrogen and carbon monoxide. taneous series of reactions is to use enthalpies of formation of the
Char will also react with carbon dioxide to form carbon various species. This data can be found in textbooks, for example,
monoxide, and with hydrogen to form methane [12]. by [9,26]. The data is utilised in the property packages used in
(iv) Gas phase reactions: In the gas phase the watergas shift commercial simulators, e.g. ASPEN PLUS, HYSYS, & VMGSim to
(Eq. (2)) and steam reforming (Eq. (3)) reactions occur [12] name a few.
and can be desirable to maximise hydrogen production: Gas phase reactions and combustion can be assumed to achieve
thermodynamic equilibrium under commercial gasication condi-
CO H2 O CO2 H2  41 kJ=mol 2 tions. However, heterogeneous reactions are limited by mass
transfer and the rates of reaction are many orders of magnitude
CH4 H2 O CO 3H2 260 kJ=mol 3 lower than gas phase reactions [12]. The transport gasier operates
at relatively low temperatures which may increase departures
Coal contains organic sulphates and nitrates that under gasica- from thermodynamic equilibrium.
tion conditions form primarily hydrogen sulphide, ammonia, car-
bonyl sulphide and hydrogen cyanide [12]. In the gasication
2.2.2. Gasication kinetics
processes downstream units must reduce minority species to very
Due to the complexity of gasication reactions and the hetero-
low levels, increasing the complexity of the process.
geneous nature of coal, there is no rate equation that denes the
design of a gasier. Gasiers are built on the basis of experience
2.2.1. Gasication thermodynamics
as laboratory scale experiments do not scale successfully to predict
Reviewed gasication models considered pyrolysis and com-
commercial operation. Development of new designs is slow and
bustion/gasication as two distinct sets of reactions [15]. In the
testing has to be on a large scale which is expensive [12].
models pyrolysis was considered to be instant and the separation
It can be seen in an article by Roberts and Harris [22] that, even
of hydrogen from the char total. Higman and Burgt [12] considered
theoretically, incorporation of kinetics into gasication models is
this assumption reasonable at temperatures above 850 C in well
not agreed upon. However, there is consensus that the inherent
mixed gasiers. Subsequent char gasication reactions can there-
reactivity of the coal char plays a signicant role and that char gas-
fore be considered in terms of pure carbon which is a considerable
ication reactions are rate limiting [10,23].
simplication.
Building a kinetic model from reactor engineering principles for
The reactions involving the main species that occur in the gas-
a commercial gasier is impractical as char characteristics, pyroly-
ier post pyrolysis are well established and readily available in lit-
sis, thermodynamic conditions, particle size distributions, particle
erature including Higman and Burgt [12]. The key reactions
interaction, uid mechanics, thermal gradients and inorganic com-
include:
ponents all affect the kinetics and are not knowable with sufcient
Exothermic combustion reactions that are necessary to drive
precision. A kinetically limited simulation can be designed that
pyrolysis and subsequent gasication:
heuristically adjusts a simplied kinetic model to replicate known
Cs 1=2O2 COg  111 kJ=mol 4 data. Eq. (1) proposed by Mann et al. [16] has an empirical basis
using an engineering scale transport gasier and requires only
Cs O2g CO2g  394 kJ=mol 5 temperature, pressure, partial pressures of major species and resi-
dence times.
COg 1=2O2g CO2g  283 kJ=mol 6 Reviewed process models that had the ability to account for ki-
netic limitations presented results that closely resembled experi-
H2s 1=2O2g H2 Og  242 kJ=mol 7 mental data. Kunze and Spliethoff [15] report did not take into
account the specic reactivity of coals but dened gasication as
Endothermic Boudouard reaction:
a series of linearly-independent reactions and estimated an ap-
Cs CO2g 2COg 172 kJ=mol 8 proach temperature or pseudo-equilibrium for each reaction.
Although this method will not produce a model with analytically
Endothermic water gas reaction:
accurate stoichiometry, the overall mass and energy balance
Cs H2 Og COg H2g 131 kJ=mol 9 should be representative. White and Seider [31] demonstrated that
this pseudo-equilibrium method, that is incorporated in
482 C.J. Arthur et al. / Fuel 115 (2014) 479489
commercial process simulators, is a legitimate method of simulat-
ing methane reforming.
A process model of the transport gasier should also take into
account the entry position of the coal and oxygen and the resi-
dence times of the kinetically limiting char in the various sections.
Fig. 3 illustrates that the coal enters the reactor above the region
where the oxygen is present. The coal after pyrolysis is transported
upwards through the riser with the combustion, pyrolysis and gas-
ication products. Principally it is the recycled char that undergoes
combustion and subsequently gasication in the mixing zone. The
residence time of the char in the mixing zone is approximately
45 s, the residence time in the riser is 1 s [16]. It is the recycled
char that dominates gasication and combustion, conserving the
light hydrocarbons that initially evolve from the coal.
2.3. The challenge of coal structure representation
It is difcult to represent the complex and varying structure of
coal in a process simulation. In the reviewed gasication simula-
tions the structure of coal was not modelled despite the effect on
reactivity. Coal was represented from ultimate and proximal anal-
yses [18,11,15]. The computer models were not able to inherently
simulate differences in reactivity arising from structural differ-
ences between coal types. It is therefore important to restrict pre-
dictions from computer models to similar coals to those used to
verify the model.
Fig. 3. Schematic of the combustion zone, mixing zone and riser section of the KBR
transport gasier.
Inorganic material in coal, particularly alkaline elements, can
catalyse gasication reactions [12]. Inorganic content would
greatly increase the complexity of a simulation and was classed
as inert or excluded in the reviewed simulations [18,11,15]. Collot
[7] reviewed the suitability of New Zealand lignites to gasication
and concluded that the high calcium content would improve the
kinetics of char gasication. Clemens et al. [6] demonstrated that
the chars from tested New Zealand lignites were between 1.5
and 1.75 times as reactive as those from equivalent Australian
coals. Therefore, there will be uncertainty in making predictions
for New Zealand lignites on the basis of a simulation veried using
foreign lignites.
2.4. The challenge of coal pyrolysis
The enormous complexity of coal chemistry is manifest in pyro-
lysis. An understanding of the mechanisms involved are critical to
building a process model, as pyrolysis occurs in the initial stages of
gasication and determines the reactants of the key gasication
reactions. Simplications are required to design a practical process
model.
Above temperatures of 320 C, carboncarbon, carbonoxygen
and carbonsulphur bonds in coal can break [12]. Aliphatic bridges
between macrostructures are most prone to breakage. This break-
age leads to the formation of tar and volatiles from the residual
hydrocarbon molecules or metaplasts [30]. Some metaplasts are
unstable radical molecules [27]. Methyl, methoxy and molecular
hydrogen become hydrogen donors [28,29]. At commercial gasi-
cation temperatures a series of decomposition reactions including
hydrogenation and hydrocracking compete with recombination
and condensation reactions resulting in char, volatile gasses such
as carbon monoxide, hydrogen and light hydrocarbons, and tar
[30]. The products of pyrolysis are dependent on the structure of
the coal and the thermodynamic conditions [12]. No single chem-
ical pathway is dominant in the formation of hydrocarbon gasses
from pyrolysis and the mechanism differs between lignite and
bituminous coals [27].
Organic sulphur and nitrogen in the coal structure form car-
bonyl sulphate, hydrogen sulphide, ammonia and hydrogen cya-
nide as metaplasts decompose and become saturated. Nitrogen
distribution depends on whether the nitrogen atom was present
in aromatic rings or amine groups. Prediction of the product there-
fore depends on a knowledge of the coal structure. Sulphur distri-
bution is thermodynamically predictable as it is dependent on the
following equations [12]:
COS H2 H2 S CO 7 kJ=mol 14
COS H2 O H2 S CO2  34 kJ=mol 15
Gasication simulations in literature simplied pyrolysis by
assuming that the process is instantaneous, that char contains only
carbon and that ash is inert. The remaining components were
determined on the basis of the coals proximate and ultimate anal-
ysis and not related to the structure [14,1]. Error due to the simpli-
cation can be corrected heuristically by adjusting the parameters
of subsequent reactions [15,1].
2.5. Validation data
Process information for the transport gasier was available
through published KBR literature as noted below. A summary of
process data can be seen in Table 1.
Ariyapadi et al. [2] detail the ultimate coal analysis for 2 coals
and the corresponding syngas compositions from test runs in an
engineering scale facility. Process variables such as operating pres-
sure, temperatures and oxygen and steam ratios are stated. A US
 C.J. Arthur et al. / Fuel 115 (2014) 479489 483

Table 1
Transport gasier process data available from Ariyapadi et al. [2].

Input component PRB (sub-bituminous) Freedom (lignite) Syngas component PRB (sub-bituminous) Freedom (lignite)
C (wt.% AR) 51.75 44.21 Temperature 927 C 899 C
O (wt.% AR) 11.52 12.45 Pressure 3600 kPa 3600 kPa
H (wt.% AR) 3.41 2.71 CO (mol.% wb) 39.7 35.6
N (wt.% AR) 0.71 0.68 H2 (mol.% wb) 28.5 25.6
S (wt.% AR) 0.26 0.60 CO2 (mol.% wb) 14.3 17.5
Moisture (wt.% AR) 27.21 29.82 CH4 (mol.% wb) 4.3 6.1
Ash (wt.% AR) 5.13 9.53 NH3 (mol.% wb) 0.4 0.4
HHV (kJ/kg. AR) 2.04E4 1.71E4 H2O (mol.% wb) 12.6 14.4
FC (%) 43.44 29.08 N2 (mol.% wb) 0.09 0.09
O2 ratio (kg/kg MAF coal) 0.75 0.66 H2S (mol.% wb) 0.075 0.2
Steam ratio (kg/kg db coal) 0.3 0.3 HCN (mol.% wb) 0.025 0.027
Moisture (dried AR) 18.89% 18.89% COS (mol.% wb) 0.004 0.01

where AR = As received, PRB = Powder river basin, wb = Wet basis, db = Dry basis, MAF = Moisture and ash free, HHV = Higher heating value, and FC = Fixed carbon.

Department of Energy report [14] estimated oxygen blown cold gas
efciencies in the range of 7283%. Heat loss from the engineering
scale reactor was stated as 10% of the feed energy suggesting the
transport gasier should not be modelled as adiabatic.
3. Simulation methodology
3.1. Process simulation
ASPEN PLUS was used for the process simulator in this study
because Solid Energy Ltd. NZ, project sponsor, use compatible soft-
ware for modelling coal processes. ASPEN PLUS version 2006.5
has the ability to simulate nonconventional solids such as coal
and coal ash. A nonconventional solid is an impure solid which
does not participate in phase and chemical equilibrium calcula-
tions. The solids are dened from empirical data or component
attributes, that in the case of coal is found in the caloric value
and ultimate, proximal and sulphur analyses. Nonconventional sol-
ids must be transformed into conventional components in order to
fully participate in the simulation.
3.1.1. Simulating coal
The composition of the coal was specied in the simulation
according to the coal analysis data published by KBR in the 2008
Twenty-fth Annual International Pittsburgh Coal Conference as
seen in Table 1 [2]. Coal ash was simulated as a nonconventional
component that was initially part of the nonconventional coal
and was inert in the system. Any catalytic effect of the ash may
be taken into account by adjusting the kinetics of the simulation.
It was assumed that all sulphur stated in the ultimate analysis
was organic sulphur and that inorganic sulphur was included in
the ash content.
The property calculators HCCOALGEN and COALIGHT were used
to estimate, thermodynamic and physical properties of the non-
conventional coal in accordance with the Aspen solid modelling
tutorial and the reviewed simulations [14,11,15]. The higher heat
value of the coal was entered into the simulation on a dry basis
as parameter HCOMB. The particle size distribution was estimated
but did not impact the kinetics of the simulations. The porosity,
macrostructure, variance and bound water content of the coal were
not specied in the simulation.
3.1.2. Simulating pyrolysis
Pyrolysis was simulated in two steps using the units displayed
in Fig. 4.
The simulation was required to transform the nonconventional
coal to participating components. This was achieved using a yield
reactor whereby a series of components were specied and the
component attributes of the nonconventional coal were distributed
according to a Microsoft Excel routine embedded in the pro-
gramme. The rst step of pyrolysis involved determining the reac-
tive and inert species via the yield reactor PYRO illustrated in Fig. 4.
Secondly, the Gibbs reactor VOLS, that determines the products on
the basis of minimum Gibbs energy, rearranged the volatile prod-
ucts to form a thermodynamically sound ratio. The simulated spe-
cies produced by pyrolysis were: H2O, ASH, C (Conventional solid
and mixed sub-stream component), H2S, H3N, CH4, O2. Pyrolysis
was assumed to be instantaneous. All light hydrocarbons were rep-
resented by methane. Tars were not included in the simulation. The
inclusion of liquid hydrocarbons would require empirical knowl-
edge of the transient chemical species inside the gasier or a kinetic
model of coal pyrolysis. A Gibbs energy based simulation would not
require detail of interim species other than the ultimate analysis of
the coal. A Kinetic gasication model would require heuristic tun-
ing to compensate for the simplication. In the rst step of pyroly-
sis the proportion of yield reactor products was determined by:
(1) The assumption that all organic nitrogen formed ammonia.
This stems from the chemistry of pyrolysis where initially
HCN is formed as the organic bonds in the coal cleave.
HCN decays to form NHi radicals which saturate to produce
NH3 in the presence of proton donors. This assumption pro-
vides a better result for lower rank coals where nitrogen is
found in amine groups. Higher rank coals that have a greater
proportion of aromatic rings in the structure produce HCN as
the principle nitrogen containing component [12].
(2) All organic sulphur formed H2S. H2S was selected as only 2
elements are involved simplifying the calculations. The dis-
tribution between H2S and HCN can be estimated by a sub-
sequent Gibbs reactor.
(3) The residual hydrogen in the coal formed methane. The tem-
perature, high porosity of lignite and violent internal condi-
tions support the assumption that pyrolysis causes the
cleavage of all volatile CC bonds.
(4) The oxygen within the coal was left in its atomic form as the
distribution between CO, CO2 and water were calculated
subsequently.
(5) It was found that 2% more carbon (mole basis) was required
than was available in the non-xed carbon to stoichiometri-
cally balance the components. The extra carbon was
deducted from the xed carbon proportion of the proximate
analysis. The minimum required carbon to combine with the
oxygen was dened as free carbon and was apportioned to a
MIXED sub stream. The 2% error was tolerated as the KBR
coal analyses were from thermo gravimetric data. Pyrolysis
products of an actual gasier cannot be veried. A theoreti-
cal approach to modelling pyrolysis, that reasonably
matched thermo-gravimetric data, allowed the simulation
484 C.J. Arthur et al. / Fuel 115 (2014) 479489

Fig. 4. Schematic of pyrolysis simulation.

to adjust the re-organising of volatile molecules by specify-

ing inert fractions for components downstream to t empir-
ical data for syngas composition.
(6) The residual carbon became the pure conventional solid
CHAR.
(7) All ash remained as the nonconventional, inert solid ash.
(8) All water contained within the coal became free water.

It can be seen in Fig. 4 that the sub-stream splitter separated the

mixed components (components that participate in phase equilib-
rium) with the conventional and nonconventional solids. This is
because Gibbs reactors are not programmed to handle carbon in
two separate sub-streams if products need to be identied to re-
strict the reactor. The mixed stream or volatile components were
rearranged on the basis of minimum Gibbs energy in the unit VOLS.
A similar approach was taken by Ardila et al. [1]. Ardila et al. [1]
model differs as atomic species were dened in the initial yield
reactor. This simulation aims to reproduce the process of pyrolysis
as closely as feasible; therefore methane rather than atomic hydro-
gen was specied. In pyrolysis, methane is produced initially and
steam reforming subsequently converts some methane to hydro-
gen and carbon monoxide.
3.1.3. Simulating the KBR transport gasier
The transport gasier was simulated by representing different
regions of the reactor with lumped parameter models. Three alter-
native models were designed: (1) Unconstrained Gibbs Energy
Model. (2) Temperature Approach Model. (3) Kinetic Model. Pyro-
lysis and the simulation of the recycle loops where common to all
models but greater resolution was required for the kinetic model
where different sets of reactions were represented by different
units in the model.
The transport gasier uses recycled syngas as a uid to convey
coal into the system [2]. The recycled quantity of syngas was not
specied by KBR. It was assumed that a reasonable assumption
would be 0.23 N m3 per kg coal (From world coal institute: Fuel
conversion facts, September 2012). Fig. 5 shows that the recycled
syngas is not connected to the coal at the inlet as this would re-
quire additional species in the already complex yield reactor PYRO.
The recycled syngas connects with the pyrolysis volatiles before
they are re-organised in the Gibbs reactor VOLS.
All units represent regions within the reactor and are simulated
as isothermal. It is likely that regions of a gasier such as the oxy-
gen inlet will contain steep thermal gradients. KBR literature does
report that the transport gasier has excellent heat transfer prop-
erties due to the high recycle of solid material. However, tempera-
ture readings in the engineering scale test reactor show
temperature differentials up to 93 C [14].
Fig. 5. Annotated screenshot of pyrolysis simulation units.
Net endothermic units were connected to exothermic units
with heat streams. To complete the energy balance, after enough
energy was produced to maintain temperatures and reactions, a
heat stream that represents heat lost to the atmosphere was de-
ned. The reactor was simulated as isobaric at 3600 kPa [2]. No
pump duty was considered for the syngas recycle loop.
The simulation simplied the reactor by not including the lime-
stone system that removes sulphur from the coal [24]. The litera-
ture used to verify the simulation did not include information
regarding the limestone system [2]. COS and H2S molar concentra-
tions in the simulations were, on average, within 65% of the KBR
report.
3.2. Simulation assumptions and topology
3.2.1. Unconstrained Gibbs energy model
The unconstrained Gibbs energy model simulated the reactor
using similar thermodynamic assumptions to Nathen et al. [18]
and Gnanapragasam et al. [11]. The volatile pyrolysis products,
ash, char and recycled material were mixed with the specied oxy-
gen and steam and preceded to the Gibbs reactor GASIFY illus-
trated in Fig. 6. Products were determined on the basis of
minimum Gibbs energy, therefore the rates of reactions were con-
sidered unimportant. The Gibbs energy model can handle compli-
cated and interacting systems of reactions providing
thermodynamic equilibriums are achieved. No iteration or data t-
ting is required to tune the model.
In trial simulations it was noted that the results were implausi-
ble, as all of the coal char was reacted to completion in a single
 C.J. Arthur et al. / Fuel 115 (2014) 479489
Fig. 6. Schematic of process streams to combustion and gasication unit GASIFY in
unconstrained Gibbs energy model.
pass. Therefore, the topology of the simulation had to be adjusted
to accommodate for specied carbon conversion. The model was
altered so that 2.5% of the coal char was removed before entering
the Gibbs reactor.
3.2.2. Temperature approach model
The topology for the temperature approach model was identical
to the unconstrained Gibbs energy model. The difference is in the
specication of the Gibbs reactor GASIFY. The approach is similar
to the simulations published by Kunze and Spliethoff [15] and Kra-
mer [14]. Unlike the unconstrained Gibbs energy model, iteration
is required as the pseudo-equilibrium temperatures have to be
estimated.
In order to build a simulation using approach temperatures, a
series of linearly-independent reactions must be selected that
can dene the system. These equations may not have an empirical
basis. Gasication and combustion involves a near limitless series
of physical stoichiometric options, yet there is no degree of free-
dom in the number of reactions that need to be selected to dene
the simulation. Eight reactions must be specied or the 13 chemi-
cal species in the simulation minus the 5 elements. All input com-
ponents must be specied as reactants and all output components
as products. The only heterogeneous gasication reaction that was
selected was methanation (Eq. (10)). The selected equations did
not imitate the physical process but offered a means to manipulate
the equilibrium conditions to increase or decrease the proportion
of components in the syngas.
The temperature approach model offered the ability to account
for kinetic limitations and also errors in pyrolysis simulation. How-
ever, the model required syngas data to be able to tune the ap-
proach temperatures heuristically. Data for the transport gasier
was not readily available in the public domain and large scale test-
ing is expensive. Only 2 data points were available to verify the
model.
3.2.3. Kinetic model
The complexity of the simulation increased considerably for a
kinetic model. The model aimed to analytically describe what is
happening in the reactor and stoichiometries were specied to rep-
resent the real process.
A stoichiometric reactor can handle multiple specied chemical
reactions and limits the extent of reaction rather than the thermo-
dynamic equilibrium. Extents of reaction are an inappropriate
means to limit the reactions of the transport gasier as they are
not a function of temperature. The available data involved coals
that were tested at different temperatures and a simulation must
be able to adjust the kinetics accordingly. To remedy this problem,
the extent of reaction was calculated from kinetic correlations
485
found in literature and partial pressures and temperatures. Embed-
ded Excel spreadsheets control the stoichiometric reactors: GAS-
IFY, HYDRO and WGS2.
The entry of coal above the mixing and combustion zones, as
described earlier, is critical to the kinetic model. If fresh coal were
combusted, the methane, carbon monoxide and hydrogen that re-
sult from pyrolysis would oxidise preferentially to the solid char.
The transport gasier is designed to conserve valuable hydrogen
and methane and the simulation is designed to represent this.
In the kinetic model, pyrolysis is simulated in a similar fashion
to the previous models. The rst step after pyrolysis is the simula-
tion of the riser section. A schematic of the riser simulation can be
seen in Fig. 7. In the model the recycled syngas, char and volatiles
enter the recycle loop above any simulated combustion or gasica-
tion activity. This is analogous to entering the riser of the gasier
with its short residence time. The rst unit that the material
encounters, after combination with combustion, gasication prod-
ucts and un-reacted material, is the stoichiometric reactor WGS2.
WGS2 models the water gas shift reaction (WGS) (Eq. (9)). In a
gasier the WGS reaction can be assumed to proceed to equilib-
rium at all points as its kinetics are similar to combustion [16].
In the simulation, the units VOLS, GASIFY and WGS2 ensure that
the mixed sub stream is at WGS equilibrium at key nodes. A corre-
lation for the WGS equilibrium constant in terms of partial pres-
sures and temperature was available from Bell et al. [4] and is
shown below. As the WGS is equimolar, the equilibrium constant
is not a function of system pressure.
 
PCO2 PH2 2
ln 1:8907ln T  30:084 lnT 117:982 16
PCO PH2 O
where P is the partial pressure (MPa) and T is temperature (K).
The correlation was claimed to predict the equilibrium constant
to within 0.3% of the calculated values at gasication tempera-
tures. However, the effect of H2S were not accounted for which will
affect the WGS reaction (Smith [25]).
Eq. (16) was required to be converted to an extent of reaction in
order to be useable in a stoichiometric reactor. Furthermore, the
partial pressures required iteration as the node was located within
the innermost of two recycle loops. The complexity increased as
derivation of an extent of reaction from the equilibrium required
polynomial expressions that have more than one solution, and
could not be easily implemented directly. Excel spread sheets were
therefore designed to compare correlated values of the equilibrium
constant (keq) with estimated values and select the extent of reac-
tion (n) when the values converge.
PCO P H
From Eq. (17) a value for PCO2PH O2 or keq could be calculated from
2
system temperature. The extent of reaction is a vector in Excel with
values from 0 to 1 (process knowledge indicates whether the
reaction proceeds forwards or backwards), 1000 columns were
used to minimise discretisation error. The change in partial
Fig. 7. Simulation of riser section of the transport gasier.
486 C.J. Arthur et al. / Fuel 115 (2014) 479489
pressure (x) of a selected component is calculated for each extent
of reaction row. Using the formula:
PCO2 xPH2 x
keq
PCO  xPH2 O  x
The value for keq above was obtained from the partial pressures
in the simulation, which update in real time as the simulation iter-
ates, and compared with the correlation value. Once the values
match, the extent of reaction was exported to the simulator.
The solid components that exit the unit WGS2 were separated
using a sub-stream splitter which simulated the dis-engager and
cyclone of the transport gasier. It was assumed that the ratio of
0.23 N m3 of gas to convey 1 kg of coal is valid within the reactor
as a minimum. From this ratio the minimum quantity of the recy-
cled syngas needed to be determined iteratively by the simulation
and was controlled by an embedded Excel spread sheet. In practice,
not enough steam was available for the gasication reactions if the
recycle rate was set to a minimum. Steam will be in excess in the
gasication zone. The recycle ratio was adjusted to approximately
0.2 to provide sufcient steam.
Prior to the recycled material re-entering the combustion zone,
the ash and limited char was separated using a sub-stream splitter.
All ash was removed per pass. To calculate the separated char com-
ponent, the carbon conversion was translated to a separation ratio
based upon the ow of recycled material using an Excel spread
sheet.
The recycled material owed to the units representing the long
residence time combustion zone and mixing section of the trans-
port gasier. A schematic of the simulation can be seen in Fig. 8.
The reactor COMBUST is a Gibbs reactor as combustion reac-
tions can be assumed to be instantaneous. The reactions were as-
sumed to proceed to completion within the node. Gasication
reactions depend on sub-stoichiometric oxygen for combustion.
To prevent the Gibbs reactor simulating gasication reactions
due to the limited oxygen, which would negate the kinetic aspects
Fig. 8. Schematic of units simulating mixing and combustion zone.
of the model, the output of the reactor was limited to the products:
C, H2O, O2, CO2, CO, H2S, COS, N2, NH3, CHN. Combustion products
and unburned char proceeded to the stoichiometric reactor GAS-
IFY. In GASIFY an Excel spread sheet controlled the water gas shift
17
equilibrium using the method described above. The heterogeneous
char reactions: Steam gasication (Eq. (9)) and the Boudard reac-
tion (Eq. (8)) were simulated in parallel via a further Excel spread
sheet. Using Eq. (1) and the residence time of char in the mixing
zone, the extent of reaction was calculated and exported to the
simulator. The Boudard kinetics were controlled as the Arrhenius
parameters were proposed by Mann et al. [16] and the reaction
rate was assumed to be proportional to the molar carbon dioxide
concentration and rst order. This assumption is supported by Hig-
man and Burgt [12]. Boudard kinetics was controlled using the
equation:
dnC 59;1000:503
A  M CO2 e T 18
ds
where nC is the extent of char consumption (mole fraction), M the
mole fraction, T the temperature (K), s the residence time (min),
and A is the pre-exponential factor
The constant A was unknown, as mole fraction rather than con-
centration is the available measurement. A was determined by t-
ting the curve to a data point. The calculations controlling the unit
GASIFY take the process parameters of the entry stream as an in-
put. The interaction between stoichiometric equations could not
be accounted for in the model.
The stoichiometric reactor HYDRO controlled methanation (Eq.
(10)). An Excel spread sheet controlled the node using the same
method as the spread sheet described previously that determined
the steam gasication reaction. The reaction order of methanation
was assumed to be second order and dependent on the partial
pressure of hydrogen. Eq. (18) determined the reaction extent
and was derived using constants proposed by Mann et al. [16].
dnC 192000:503
A  PH2 2 e T 19
ds
where nC the extent of char consumption (mole fraction), P the par-
tial pressure (atm) T the temperature (K), s the residence time
(min), A is the pre-exponential factor.
The exit stream of the unit COMBUST was the input for the
methanation calculations. Therefore, the methanation reaction in
the simulation was assumed to occur pursuant to other char reac-
tions. This assumption is supported as the kinetics of methanation
are slow relative to steam gasication and the Boudard reaction is
minor [16].
In the kinetic model, the Gibbs reactor VOLS, that determined
volatile pyrolysis products, needed to be adjusted. Methane partic-
ipation was restricted. If pyrolysis products were allowed to
recombine on the basis of Gibbs energy the methane was reduced
to a low level. Consequently more methane needed to be produced
by the methanation reaction (Eq. (10)) consuming char. Recycled
char should not fall below 25% of the char in the reactor [16].
Methane was specied to be 0.35 inert to give a plausible recycle
ratio of 34%. Further process data, precise knowledge of methana-
tion kinetics or validation of pyrolysis would be required to deter-
mine exact departures from equilibrium in VOLS. Restricting
methane participation in VOLS had an analytical basis as methane
results from thermal decomposition and there is some subsequent
steam reforming, Steam reforming is likely to be kinetically re-
stricted as it is usually catalysed at 900 C. Methane restriction
was not necessary in the earlier models as only the kinetic model
required accurate simulation of the internal recycle.
The kinetic model was able to correctly simulate all veriable
process streams as well as the overall material and energy balance.
Treating interacting equations as either sequential or parallel and
 C.J. Arthur et al. / Fuel 115 (2014) 479489 487

independent was a source of error that was compensated for by
adjusting kinetic constants. Insufcient kinetic data was available
to build a truly analytical model and the parameters of model were
tted to KBR data. Expressions for the minority species were not
explicitly programmed in the kinetic model as the kinetics of sul-
phur and nitrogen distribution in commercial gasiers were not
readily available.
4. Results and discussion
The simulations were run using the coals and corresponding
parameters listed in Table 1. The simulation results were listed
in Table 2. The root of mean square errors (RMSE) between the
KBR data and simulations for the lignite and sub-bituminous coal
from three proposed models were listed at the bottom row in Ta-
ble 2. For the further model evaluation, the material balance was
split between majority species (CO, CO2, H2, and CH4) and sulphur
and nitrogen compounds to allow comparison between models
and with literature. The RMSEs, derived from the absolute error
seen in square brackets in Table 2, were shown separately in Tables
3 and 4 for major and minor species respectively.
It can be seen in Table 2 that the RMSE from the Temperature
Approach model is the lowest, it indicates that the Temperature
Approach model can provide the best predictions among these
three models for both lignite and sub-bituminous coals.
From Table 3, we observed that the kinetic model demonstrates
the lowest major species RMSE (0.21) than the other models in re-
spect to the Freedom Mine Lignite. RMSEs for different feeding
materials: lignite and PRB from the unconstrained Gibbs energy
model are quite different. The same phenomena can be observed
for the kinetic models. These indicated that both models are not ro-
bust for different feed materials: lignite and PRB. The Temperature
Approach is more robust for both feeds since The RMSEs for lignite
and PRB are almost equal. Some absolute errors of the minority
species in Table 1 are not statistically signicant, however these er-
ror would not signicantly affect the performances of model pre-
dictions since they only occupy 1% of total outputs. From RMSEs
of the minority species from the models in Table 3 and the litera-
ture data, we could conclude that the qualities of the model predic-
tions for these minority species are acceptable.
The energy balance will be presented in terms of: (1) The ther-
mal value of the coal (HHV) lost to the atmosphere and (2) The cold
gas efciency (gcold gas). The HHV value of the coal was specied in
Table 1. The models calculated the energy that was required to be
lost from the system as waste heat in order to maintain the com-
ponent units at the specied temperature according to the ASPEN
PLUS property package HCOALGEN. The heat loss was presented
in Table 5 as a percentage of the coal HHV. Cold gas efciency
was calculated as the percentage of the LHV of the feed coal of
the combined LHV of syngas components H2, CO and CH4. Coal
LHV values were estimated using the following correlation:
LHV 9m  91:2H  10:5M  0:34O 20
where H is the hydrogen (%), M the moisture (%), O the oxygen (%);
heating values in (BTU/lb).
Energy balance data was not available for the specic coal tests.
Table 5 shows that cold gas efciencies for all models were within
the range published by the US Department of Energy for the engi-
neering scale reactor. The energy losses were less than the stated
10% [14]. The assumption that the reactor is isothermal would have
affected the heat balance. Without knowledge of the mechanical
design it is not possible to determine the extent of the validity of
this assumption. It can be seen from Table 4 that the minority spe-
cies were not simulated with the accuracy of the major species.
Generally, sulphur compounds proved to be more predictable than
nitrogen compounds. This is due to the equilibrium conditions of
sulphur being dened by Eqs. (2) and (3) whereas nitrogen distri-
bution requires knowledge of the coal structure [12]. Molecular
nitrogen was too high in the predicted syngas compositions as
there was no kinetic restraint of the Gibbs reactor VOLS. The tem-
perature approach model was better able to predict the minority
species distribution as the model has the ability to easily adjust
the approach temperatures of the equations specied in the above
sections. Analytical kinetic expressions for the physical reactions
were not readily available. Therefore, the kinetic model predicted
the distribution of sulphur and nitrogen with only marginally
greater accuracy than the equilibrium model. The models will
therefore be compared on the basis of the major components.
The unconstrained Gibbs energy model was able to approxi-
mately replicate the material balance of the KBR data with an
RMSE error of 1.55. The sub-bituminous coal was simulated with
less than half of the RMSE than the lignite. The components H2,
CO and CO2 were within the range of error published by Ardila
et al. [1]. An advantage of the unconstrained Gibbs energy model
was its simplicity and it involved no tting to data. As the lignite
case was established at a temperature 27 C less than the sub-bitu-
minous case, it is hypothesed that kinetic limitation played a great-
er role during the lignite gasication and Gibbs energy
minimisation was therefore a less suitable simulation approach
for lignite.
The temperature approach model was tuned on the basis of
least residual sum of squares of the two data points. The models
predictions closely resembled KBR process data for the major syn-
gas components. Table 3 shows that the combined RMSE for the
two ranks of coal was 0.54. As the approach temperatures do not
have an analytical basis, and therefore cannot be veried, it is
not possible to conclude that the model makes realistic kinetic
assumptions on the basis of two data points.

Table 2
Material balance summarised from ASPEN PLUS output. All values are in mole% (dry basis) with absolute error in parentheses.

Compound KBR data Simulation

FL PRB Unconstrained Gibbs energy Temperature approach Kinetic
FL PRB FL PRB FL PRB
H2 30.189 32.834 32.006 [1.817] 33.055 [0.221] 30.766 [0.577] 31.956 [0.878] 29.835 [0.354] 33.106 [0.272]
CO 41.981 45.737 44.114 [2.133] 47.083 [1.346] 42.408 [0.427] 45.668 [0.069] 42.18 [0.199] 43.266 [2.471]
CO2 20.637 16.47 18.446 [2.191] 17.661 [1.191] 20.376 [0.261] 16.88 [0.41] 20.775 [0.138] 18.125 [1.655]
CH4 7.173 4.95 5.414 [1.759] 5.324 [0.374] 6.43 [0.743] 5.486 [0.536] 7.191 [0.018] 5.494 [0.544]
H2S 0.23364 0.08581 0.34229 [0.10865] 0.11569 [0.02988] 0.35723 [0.12359] 0.12586 [0.04005] 0.34732 [0.11368] 0.1295 [0.04369]
COS 0.01168 0.00458 0.02165 [0.00997] 0.00813 [0.00355] 0.0141 [0.00242] 0.00538 [0.0008] 0.02179 [0.01011] 0.00309 [0.00149]
N2 0.10514 0.10297 0.5 [0.39486] 0.44101 [0.33804] 0.29377 [0.18863] 0.28604 [0.18307] 0.50816 [0.40302] 0.43581 [0.33284]
NH3 0.46729 0.45767 0.00853 [0.4588] 0.00756 [0.00024] 0.43713 [0.0302] 0.32037 [0.1373] 0.00816 [0.4591] 0.01622 [0.4415]
HCN 0.03154 0.0286 0.0002 [0.0313] 0.00024 [0.0284] 0.00981 [0.0217] 0.00973 [0.0189] 0.00019 [0.0314] 0.00016 [0.0284]
RMSE 0.997 0.352 0.710
488 C.J. Arthur et al. / Fuel 115 (2014) 479489

Table 3
Major species material balance summary: RMSE between simulations and KBR data.

Model Unconstrained Gibbs energy Temperature approach Kinetic

RMSE 1.55 PRB and lignite (0.64 PRB, 1.34 lignite) 0.54 PRB and lignite (0.55 PRB, 0.53 lignite) 1.08 PRB and lignite (1.52PRB, 0.21 lignite)

Table 4
Minor species material balance summary: RMSE between simulations and KBR data.

Model Unconstrained Gibbs energy Temperature approach Kinetic

N compounds 0.428 PRB and lignite (0.325PRB, 0.35 lignite) 0.122 PRB and lignite (0.133PRB, 0.111 lignite) 0.337 PRB and lignite (0.320PRB, 0.353 lignite)
RMSE
S compounds RMSE 0.057 PRB and lignite (0.021PRB, 0.077 lignite) 0.065 PRB and lignite (0.028PRB, 0.087 lignite) 0.354 PRB and lignite (0.031PRB, 0.081 lignite)

Table 5
Energy balance summary: calculated heat loss to atmosphere and cold gas efciency.

Model Unconstrained Gibbs energy Temperature approach Kinetic

PRB coal Freedom lignite (%) PRB coal (%) Freedom lignite (%) PRB coal (%) Freedom lignite (%)
Heat loss 6.4 6.2 7.2 6.9 7.1 7.7
gCold gas 74.6 76.6 73.8 73.9 75.6 73.1

The optimised approach temperatures tted to both data points
were:
C O2 CO2 T  C
CO 1=2O2 CO T  C
CO S COS T 110  C
H3 N CO HCN H2 O T 0  C
C 2H2 CH4 T  6 C
CO H2 O CO2 H2 T  40  C
H2 S H2 S T 90  C
0:5N2 1:5H2 NH3 T  510  C
where T is the system temperature (C).
The approach temperatures evident above indicate that the set
of equations cannot be taken at equilibrium to accurately simulate
gasication in the transport gasier. The model can be adjusted to
replicate gasication of a particular coal very closely by adjusting
an approach temperature to increase or decrease the equilibrium
constant. The corresponding energy balance is calculated on the
basis of the difference in Gibbs energy between products and reac-
tants at the specied reference conditions.
As the available kinetic data was based upon lignite, the kinetic
model was tuned to the lignite data point. Table 3 shows that the
RMSE in the prediction was 1.08 with respect to both data points
but 0.21 with respect to lignite alone. As the temperature approach
model required different approach temperatures to precisely
match the different coal ranks, the kinetic characteristics of the dif-
ferent coal ranks were different. This is in agreement with litera-
ture [16]. A kinetic model may have the ability to simulate a
narrow range of coals with known kinetics. To conrm this, a sta-
tistically signicant number of data points of similar coals would
be required. The cost of large scale gasication trials is limiting.
It would be preferable to have based the kinetic model on an en-
tirely analytical basis but the necessary process detail is not in
the public domain.
The kinetic model calculated that the proportion of char that
was steam gasied per pass was 25.3% in the lignite case and
24.6% in the sub-bituminous case. These values are close to the
upper gure of 20% proposed by Mann et al. [16]. Only the kinetic
model is capable of correctly simulating the circulating uidised
region of the gasier from a mechanical point of view, as without
explicit kinetic restriction on the char, the char reacts to comple-
tion in a single pass. Therefore, a kinetic model would be useful
to analyse aspects of the process other than the overall mass and
energy balance. A kinetic model is veriable, as unlike approach
temperatures, the kinetic constants should have an empirical basis
if the process parameters are known with sufcient resolution.
The kinetically limited models could simulate the syngas com-
position of individual coal samples with RMSEs less than the equi-
librium model. Therefore, the equilibrium conditions within the
Transport Reactor are kinetically limited. As char reactions were
explicitly restricted in the kinetic model, it is probable that the het-
erogeneous char reactions are rate limiting.
21 Although the kinetically limited models could very precisely rep-
licate the KBR data for a single test sample, it is not possible to con-
clude that the assumptions necessary to replicate the data are
plausible as discrepancies could be manifest in the kinetic limitations.
In the case of the temperature approach model the pseudo-equilib-
rium cannot be veried analytically and needs to be tested using a sta-
tistically signicant number of data points. The kinetic model has
reaction assumptions and kinetic constants that were also tted to
limited data and was not solely analytical. It is the unconstrained
Gibbs energy model that forms the best analytical basis to conclude
the KBR data can be simulated with reasonable process assumptions.
It is believed that the pseudo-equilibrium model is undoubtedly
the most precise when it comes to replicating the empirical data.
The downside is that it must be tuned to empirical data. The results
used a set of approach temperatures that are common to both coal
types. It is easy to tune the model to get an exact match to a single
coal type. However, predictions can only be made with caution un-
less a statistically signicant number of data points are available to
set the parameters.
5. Conclusions
All three simulations replicated the cold gas efciency of the
syngas within the range published by the US Department of En-
 C.J. Arthur et al. / Fuel 115 (2014) 479489
ergy, providing between 6.2% and 7.7% of the higher caloric value
of the coal is lost to the atmosphere. The rst model replicated the
syngas molar composition of the components H2, CO, CO2 and CH4,
for lignite and a sub-bituminous coal, with a combined RMS error
of 1.55. The second and third models replicated the data on the
same basis with errors of 0.54 and 1.08 respectively.
On the basis of minimum Gibbs energy predicting the syngas
composition with an RMSE of 1.55, it was concluded that the
KBR data can be replicated with reasonable process assumptions
including: The reactor is not adiabatic; the process can be repli-
cated using a lumped parameter, isothermal model and carbon
conversion is 97.5%.
A lumped parameter equilibrium reactor model matched the re-
ported process data with an RMS error of 1.55. Lumped parameter
kinetically limited reactor models provided improvements for
matching the reported process data with RMSEs between 0.54
and 1.08. Therefore, the kinetics of gasication in the transport
gasier limits thermodynamic equilibrium.
Kinetically limited models could not precisely match the results
of the lignite and sub-bituminous coal cases without adjusting the
kinetic parameters. Different coal ranks will require different ki-
netic representations. Due to the high alkaline metal content and
unique kinetics of New Zealand lignite any predictive model
should be specically validated.
Acknowledgement
Support of Dr. Stefan Harms from Solid Energy Ltd. New Zea-
land for discussion and advice.
References
[1] Ardila YC, Figueroa JEJ, Lunelli BH, Filho RM, Wolf Maciel MR. Syngas
production from sugar cane bagasse in a circulating uidized bed gasier
using Aspen Plus: modelling and Simulation. In: Ian David Lockhart B,
Michael F, editors. Computer aided chemical engineering, vol.
30. Berlin: Elsevier; 2012. p. 10937.
[2] Ariyapadi S, Shires P, Bhargava M, Ebbern D. KBRS Transport Gasier (TRIG)
an advanced gasication technology for sng production from low-rank coals.
In: 25th Annual international Pittsburg coal conference, Pittsburg, PA; 2008.
[3] Barry JM, Duff SW, Macfarlane DAB. Coal resources of New Zealand. Resource
information report 16. Energy and resources division. New Zealand, Ministry of
Commerce; 1994.
[4] Bell DA, Towler BF, Fan M. Gasiers. Coal gasication and its
applications. Boston: William Andrew Publishing; 2011. p. 73100 [chapter 4].
[5] Bielowicz B. A new technological classication of low-rank coal on the basis of
Polish deposits. Fuel 2012;96:497510.
[6] Clemens T, Gong D, Pearce S. Study on the suitability of New Zealand coals for
hydrogen production. Int J Coal Geol 2006;65(34):23542.
[7] Collot A-G. Matching gasication technologies to coal properties. Int J Coal
Geol 2006;65(34):191212.
View publication stats
489
[8] Dahlin RS, Dorminey JR, Peng W, Leonard RF, Vimalchand P. Preventing ash
agglomeration during gasication of high-sodium lignite. Energy Fuels
2008;23(2):78593.
[9] Elliott MA. Chemistry of coal utilization. Second supplementary volume; 1981.
[10] Everson RC, Neomagus HWJP, Kasaini H, Njapha D. Reaction kinetics of
pulverized coal-chars derived from inertinite-rich coal discards: gasication
with carbon dioxide and steam. Fuel 2006;85(78):107682.
[11] Gnanapragasam NV, Reddy BV, Rosen MA. Hydrogen production from coal
gasication for effective downstream CO2 capture. Int J Hydrogen Energy
2010;35(10):493343.
[12] Higman C, Burgt Mvd. Gasication. USA: Gulf Professional Publishing, Elsevier;
2003, ISBN 978-0-7506-7707-3.
[13] Kabe T, Ishihara A, Qian EW, Putu I. Methods of classication and
characterization of coal. Studies in surface science and catalysis, vol.
150. Amsterdam: Elsevier; 2004. p. 179.
[14] Kramer S. Gasication plant cost and performance optimization. Washington,
D.C., U.S. Department of Energy, Ofce of Scientic and Technology
Information; 2004 [DE-AC26-99FT40342].
[15] Kunze C, Spliethoff H. Modelling of an IGCC plant with carbon capture for
2020. Fuel Process Technol 2010;91(8):93441.
[16] Mann MD, Knutson RZ, Erjavec J, Jacobsen JP. Modeling reaction kinetics of
steam gasication for a transport gasier. Fuel 2004;83(1112):164350.
[17] Nagvecker M. Optimising hydrogen and chemicals production from the KBR
transport gasier. In: Technology conference 2009, Dubai; 2009.
[18] Nathen SV, Kirkpatrick RD, Young BR. Gasication of New Zealand Coals: a
comparative simulation study. Energy Fuels 2008;22(4):268792.
[19] Natush D. Relevance of the liquid fuels trust board work to present-day energy
issues: realising New Zealands mineral potential. In: NZ minerals conference,
Auckland; 2005.
[20] Park C-L, Kim B-G. The optimization of low-rank coal grinding for transport
coal gasication by robust design. Fuel 2012;95:2826.
[21] Rezaiyan J, Cheremisinoff NP. Gasication technologies: a primer for engineers
and scientists. Boca Raton: Taylor & Francis; 2005.
[22] Roberts DG, Harris DJ. A kinetic analysis of coal char gasication reactions at
high pressures. Energy Fuels 2006;20(6):231420.
[23] Roberts DG, Ilyushechkin AY, Harris DJ. Linking laboratory data with pilot scale
entrained ow coal gasication performance. Part 1: laboratory
characterisation. Fuel Process Technol 2012;94(1):8693.
[24] Smith PV. Initial operation of the PSDF transport gasier. Fuel Energy Abstr
2002;43(4):249.
[25] Smith RJ, Loganathan M, Shantha MS. A review of the water gas shift reaction
kinetics. Int J Chem React Eng 2010;8.
[26] Speight JG. Handbook of coal analysis. Hoboken, NJ, USA: John Wiley & Sons,
Inc.; 2005.
[27] Stock LM. Coal pyrolysis. Acc Chem Res 1989;22(12):42733.
[28] Suuberg EM, Peters WA, Howard JB. Product composition and kinetics of
lignite pyrolysis. Ind Eng Chem Process Des Dev 1978;17(1):3746.
[29] Suuberg EM, Peters WA, Howard JB. Product compositions in rapid
hydropyrolysis of coal. Fuel 1980;59(6):40512.
[30] Wanzl W. Chemical reactions in thermal decomposition of coal. Fuel Process
Technol 1988;20:31736.
[31] White CW, Seider WD. Computation of phase and chemical equilibrium, part
IV: approach to chemical equilibrium. AIChE J 1981;27(3):46671.
[32] Wilkins RWT, George SC. Coal as a source rock for oil: a review. Int J Coal Geol
2002;50(14):31761.
[33] Yan H-m, Heidenreich C, Zhang D-k. Mathematical modelling of a bubbling
uidised-bed coal gasier and the signicance of net ow. Fuel 1998;77(9
10):106779.
[34] Yan HM, Heidenreich C, Zhang DK. Modelling of bubbling uidised bed coal
gasiers. Fuel 1999;78(9):102747.