Content uploaded by Chris Wareing
Author content
All content in this area was uploaded by Chris Wareing on Feb 27, 2018
Content may be subject to copyright.
High pressure CO2CCS pipelines: Comparing
dispersion models with multiple experimental datasets
Christopher J. Wareinga,b,∗, Michael Fairweathera, Samuel A.E.G. Fallec,
Robert M. Woolleya, Abigail M.E. Ward
aSchool of Chemical and Process Engineering, University of Leeds, Leeds LS2 9JT, UK.
bSchool of Physics and Astronomy, University of Leeds, Leeds LS2 9JT, UK.
cSchool of Mathematics, University of Leeds, Leeds LS2 9JT, UK.
Abstract
Carbon capture and storage (CCS) presents the short-term option for signif-
icantly reducing the amount of carbon dioxide (CO2) released into the atmo-
sphere from the combustion of fossil fuels, thereby mitigating the effects of
climate change. Enabling CCS requires the development of capture, storage
and transport methodologies. The safe transport of CO2in CCS scenarios
can be achieved through pipelines or by shipping. Either way, transport and
temporary storage of pressurised liquid CO2will be required and subject to
quantitative risk assessment, which includes the consideration of the low-
risk, low-probability puncture or rupture scenario of such a pipeline, ship or
storage facility. In this work, we combine multiple experimental datasets all
concerned with the atmospheric free release of pure and impure liquid CO2
from CCS-transport-chain-relevant high pressure reservoirs and perform the
first multiple dataset comparison to numerical models for both pure and im-
∗Corresponding author. Tel: +44 113 343 3871. Fax: +44 113 343 5090
Email address: C.J.Wareing@leeds.ac.uk (Christopher J. Wareing)
URL: http://www.maths.leeds.ac.uk/~cjw (Christopher J. Wareing)
Preprint submitted to International Journal of Greenhouse Gas Control August 31, 2016
pure jets in dry ambient air with no water vapour. The results validate the
numerical approach adopted and for the prediction of such releases, high-
light the significance of the mixture fraction at the release point, over the
mixture composition itself. A new method for impure CO2dispersion mod-
elling is introduced and limited preliminary comparisons of impure CO2data
and predictions are performed. No clear difference between pure and impure
releases is found for the cases considered.
Accepted 2016 August 31st. Received in original form 2016 March 30th.
Keywords: CCS, multi-phase flow, experimental measurement,
mathematical modelling, impurities, atmospheric dispersion
1. Introduction
Carbon capture and storage (CCS) refers to a set of technologies designed
to reduce carbon dioxide (CO2) emissions from large industrial point sources
of emission, such as coal-fired power stations, in order to mitigate greenhouse
gas production. The technology involves capturing CO2and then storing it
in a reservoir, instead of allowing its release to the atmosphere, where it con-
tributes to climate change. Once captured, the CO2is generally transported
in a liquified state and stored, typically underground, or used for processes
such as enhanced oil recovery.
The fluid dynamic modelling of liquid CO2poses a unique set of problems
due to its unusual phase transition behaviour and physical properties. Liq-
uid CO2has a density comparable with that of water, but has a viscosity of
magnitude more frequently associated with gases. These properties make the
transport of dense phase CO2an economically viable and attractive propo-
2
sition. However, due to it possessing a relatively high Joule-Thomson ex-
pansion coefficient, calculations and experimental evidence confirm that the
rapid expansion of an accidental release reaches temperatures below 194 K.
Due to this effect, solid CO2forms at temperatures below the triple point
temperature (216.6 K) following a pipeline puncture or rupture, whether di-
rectly from liquid or via a vapour-phase transition. Additionally, CO2sub-
limes at ambient atmospheric conditions, which is a behaviour not seen in
most other solids, and is an additional consideration when modelling flows
such as these. Predicting the correct fluid phase during the discharge pro-
cess in the near-field is of particular importance given the very different
hazard profiles of CO2in the gas and solid states. The safe operation of CO2
pipelines is of paramount importance then, as the inventory associated with
a cross-country pipeline would likely be several thousand tonnes, and CO2
has a toxic effect above 5% concentration and causes hyperventilation above
2% (Connolly and Cusco, 2007; Wilday et al., 2009).
The University of Leeds near-field CO2dispersion mathematical model
(Wareing et al., 2013a), has been developed and validated for free releases
of CO2into air for two data sets; the CO2PipeHaz project (Woolley et al.,
2013a,b) and the National Grid COOLTRANS project (Wareing et al., 2014a).
It has also been validated against small-scale laboratory releases and dry
ice particle behaviour (Wareing et al., 2013b, 2015a), punctures of buried
pipelines (Wareing et al., 2014b) and ruptures of buried pipelines (Wareing
et al., 2015b,c).
In this paper, we perform a comparison to a wider range of experimen-
tal data currently available regarding near-field liquid CO2dispersion. We
3
also introduce an impure equation of state and compare with new impure
near-field CO2dispersion experimental data from multiple sources. In the
next Section we reproduce the relevant background to this area. Following
that in Section 3 we present the details of the experimental data sources.
In Section 4 we briefly present details of our mathematical model and intro-
duce the modifications for impure CO2, including determination of a suitable
equation of state. In Section 5 we describe the numerical methodology used
herein. The comparison between data and predictions is discussed in Section
6. Finally, conclusions are presented in Section 7.
2. Background
In this section, we consider the growing number of recent publications that
have examined the release and dispersion of CO2, revisiting our review from
Wareing et al. (2014a) that summarised the extensive review provided by
Dixon et al. (2012) in the light of new and related additions to the literature.
MMI Engineering presented dispersion simulations (Dixon and Hasson,
2007) employing the ANSYS-CFX computational fluid dynamics (CFD) code.
Solid CO2particles were simulated by a scalar representing the particle con-
centration, avoiding the overhead of full Lagrangian particle tracking. Dixon
et al. (2012) note that this method assumed a constant particle diameter and
temperature at the sublimation temperature of 194.25 K in order to calculate
heat and mass exchange between the particles and the gas phase. Following
this work up, Dixon et al. (2009) used a full Lagrangian particle tracking
method, but still assumed particles to be at the sublimation temperature.
Dixon et al. (2012) noted that since the rate of sublimation increases as par-
4
ticle size decreases, an improved distribution of the source of the CO2gas
resulting from particle sublimation could be obtained by allowing for varying
particle size and for the fact that temperature is expected to fall below the
sublimation temperature in the near-field of a release.
Webber (2011) considered a methodology for extending existing two-
phase homogeneous integral models for flashing jets to the three-phase case
for CO2. Webber noted that as the flow expands from the reservoir conditions
to atmospheric pressure, temperature, density and the jet cross-sectional area
would vary continuously through the triple point, whilst the mass and mo-
mentum would be conserved. This led to the conclusion that there must be a
discontinuity in the enthalpy and CO2condensed phase fraction, in a similar
manner to the energy change associated with passing through a hydraulic
jump. In the development of our composite equation of state for modelling
CO2near-field sonic dispersion (Wareing et al., 2013a), we confirmed this in
a conservative shock capturing CFD code and highlighted the importance of
fully accounting for the solid phase and latent heat of fusion; the near-field
structure of the jet as well as the fraction of solid phase material is different
when this is correctly accounted for.
Witlox et al. (2009, 2011) discussed the application of the software pack-
age PHAST to CO2release and dispersion modelling. Witlox et al. (2009)
described an extension to the model in PHAST (v.6.53.1) to account for the
effects of solid CO2, including the latent heat of fusion. The modifications to
the model consist principally of changing the way in which equilibrium con-
ditions were calculated in the expansion of CO2to atmospheric pressure in
order to ensure that below the triple point, conditions followed the sublima-
5
tion curve in the phase diagram, rather than extrapolating the evaporation
curve (which diverges considerably from reality, hence the limitations of the
Peng and Robinson (1976) and Span and Wagner (1996) equations of state
to above the triple point only). In Witlox et al. (2011), the results of sensi-
tivity tests were reported for both liquid and supercritical CO2releases from
vessels and pipes calculated with the revised PHAST model. The public
release of the CO2PIPETRANS datasets and associated industrial projects,
e.g. (Ahmad et al., 2013), has validated the development of this approach,
which we also adopted in part for our composite equation of state (Wareing
et al., 2013a).
E.ON have published a number of studies (Mazzoldi et al., 2008a,b, 2011;
Hill et al., 2011). Of these, the most relevant to this work are Mazzoldi et al.
(2011) and Hill et al. (2011). These consider atmospheric dispersion from
pipeline and vessel releases. The former paper compared simulations from
the heavy gas model ALOHA to the CFD model Fluidyn-Panache. Only the
gaseous stage of the release was modelled. In the second work (Hill et al.,
2011), the authors presented CFD and PHAST simulations of dense-phase
CO2releases from a 500mm diameter hole in a pipeline, located at an ele-
vation of 5m above level ground. Steady-state flow rates were calculated at
the orifice assuming saturated conditions. CFD simulations were performed
using the ANSYS-CFX code with a Lagrangian particle tracking model for
the solid CO2particles, with three size distributions: 10 to 50 micrometers,
50 to 100 micrometers and 50 to 150 micrometers. Simulations were also
performed without particles. Their results showed that sublimation of the
particles led to a cooling of the CO2plume, affecting dispersion behaviour,
6
although the results were relatively insensitive to particle size. Gas concen-
trations downwind from the release were reportedly somewhat lower using
PHAST (v.6.6) as compared to the CFD results. No comparison to experi-
ment was performed.
Dixon et al. (2012) note that in the Lagrangian model of Hill et al. (2011)
their particle tracks followed closely the plume centreline, rather than being
spread throughout the plume. Dixon et al. (2012) went on: turbulence will
have the effect of bringing particles into contact with parts of the jet at a
higher temperature and lower CO2concentration, thereby tending to increase
the rate of sublimation and increase the radius of the region cooled by the
subliming particles. In their work, Dixon et al. (2012) included turbulent
dispersion effects in the CFX model. Further, they assumed that the solid
particles are much smaller with an initial particle diameter of 5 micrometers.
They made that choice based on an analysis of CO2experiments. In addition,
this particle size distribution is supported by the model recently developed
by Hulsbosch-Dam et al. (2012a,b), which suggested that the particle diame-
ter would be around 5 micrometers for CO2releases at a pressure of 10 MPa,
when the difference between the CO2and ambient temperatures is around
80◦C. They stated that the effect of having smaller particles in their model
was likely to cause more rapid sublimation, which should produce a more
significant reduction in gas temperature in the free jet. Recent examination
of particle size distribution in releases of supercritical CO2from high pres-
sure has shown that even smaller particles immediately post Mach shock are
indeed the case (Liu et al., 2012b), on the order of a few micrometers, which
we confirmed in laboratory releases from the liquid phase (Wareing et al.,
7
2013b).
Dixon et al. (2012) employ a Bernoulli method which they found ”to
provide reasonable predictions of the flow rate for the sub-cooled liquid CO2
releases”. Differences were apparent between the integral model and the
CFD model results. The integral model predicted temperatures that they
noted were too low in the near-field, and which then returned too rapidly to
atmospheric levels (see Dixon et al. (2012) Figure 3.). The CFD model was
noted to be in general better, although in the near-field (<10 m from the
orifice) it was still not clear whether this was the case. Further, the CFD
model appears to under-predict the spreading rate of the jet.
Liu et al. (2014) present simulations of free-jet CO2dispersion from high
pressure pipelines using a non-ideal gas equation of state - specifically the
Peng-Robinson (Peng and Robinson, 1976) equation. They obtained good
agreement compared to the limited data available from the CO2PIPETRANS
datasets (no data is available close to the shock-containing expansion region),
but do not model the solid phase of CO2and hence are limited to predicting
supercritical releases that do not cool below the triple point.
Wareing et al. (2013a) presented a composite equation of state for the
modelling of high pressure liquid CO2releases that accounts for phase changes
and the solid phase and went on to validate against venting releases from the
CO2PipeHaz project (Woolley et al., 2013a,b) and the COOLTRANS re-
search programme (Wareing et al., 2014a). The model demonstrated good
quantitative and qualitative agreement with the experimental data regarding
temperature and concentration in the near- and far-field. More recently, we
have applied the same model to punctures of a buried high pressure dense
8
phase pipeline (Wareing et al., 2015a) and a rupture of a buried 150mm-
diameter pipeline (Wareing et al., 2015b), at a quarter-scale of the full-scale
pipelines intended in the UK White Rose CCS network (Cooper and Barnett,
2014). In both cases, the model shows reasonable agreement with the data,
predicting jet temperatures, structures and behaviour, as well as predicting
particle behaviour.
Woolley et al. (2014a) published a paper linking the elements of the
CO2PipeHaz project together, for the first time numerically modelling a
complete chain rupture and consequent dispersion event in realistic topogra-
phy. This included pipeline decompression linked to near-field sonic shock-
capturing simulation of the flow from the pipe ends through the crater linked
into the far-field, where constant source conditions on a plane above the crater
were taken as input into FLACS and ANSYS-CFX simulations. These were
taken as constant source conditions and the transient nature of the near-field
decompression was not modelled. Gant et al. (2014) had previously con-
sidered the validation of FLACS and ANSYS-CFX in the far-field for this
application, using predictions for the near-field from our composite model
(Wareing et al., 2013a), albeit again modified for input into such commer-
cial software. Wen et al. (2013) presented a number of far-field simulations
of venting and horizontal above ground releases, with successful validation,
also using predictions for the near-field from our composite model (Wareing
et al., 2013a) for input conditions, but with less modification for input into
their software.
In this paper, we return to near-field free dispersion releases above ground
into dry air, and extend previous experimental data comparisons to all the
9
available data in the literature. We also present our extended numerical
model for impure CO2and compare this to impure experimental data, as
well as highlighting similarities and differences compared to pure predictions
and data.
3. Sources of experimental data
3.1. CO2PIPETRANS
Phase 2 of the DNV-GL led CO2PIPETRANS1joint industry project
(JIP) obtained a large amount of data from experiments designed to assist
in the design of CO2pipelines and fill knowledge gaps that were identified
during the execution of CO2PIPETRANS Phase 1, which resulted in the
DNV GL recommended practice document DNV-RP-J2022entitled “Design
and operation of CO2pipelines.”
Relevant to the work herein, CO2PIPETRANS Phase 2 contained dense
phase CO2release modelling validation data from two complimentary pro-
grammes of medium scale CO2release experiments conducted by DNV GL
for BP (data set 1, hereafter DS1) and by DNV GL for Shell (data set 2,
hereafter DS2). Further depressurisation tests on a CO2pipeline (data set
3) and experimental discharge data for large diameter CO2releases (data
set 4) were carried out. Full details can be found in Brown et al. (2014a).
DS1 consisted of tests 1-11 with a repeat of test 8. All were liquid phase
1https://www.dnvgl.com/oilgas/joint-industry-projects/ongoing-
jips/co2pipetrans.html Accessed 2016-Aug-02.
2http://rules.dnvgl.com/docs/pdf/DNV/codes/docs/2010-04/RP-J202.pdf Accessed
2016-Aug-02.
10
releases. Test 1 did not record temperature data in the plume and we there-
fore exclude it. Tests 3 and 6 involved an extension tube and we therefore
exclude them for the reason of not being free releases. Tests 4 and 7 involved
an instrumented target and we therefore exclude them for the reason of not
being free releases. Test 10 was pointed downwards into the ground and we
therefore excluded it for not being a free release. Tests 8, its repeat test 8R
and test 9 all considered releases without buffer pressure and therefore did
not mimic large-scale pipeline, shipping or storage facilities and did not pro-
duce steady-state data. We therefore exclude them. Tests 2, 5 and 11 have
been included here. From DS2, tests 3, 5 and 11 are the only suitable liquid
phase releases that approximate steady state conditions. Again, all are pure
CO2releases and all have been included here. The other tests are transient
releases. Data sets 3 and 4 contain large diameter CO2releases where ac-
curate measurements in the near-field close to the release were not possible.
Hence no data from data sets 3 and 4 have been included in this comparison.
We present the pertinent details of experimental data used in this work in
Table 1. Complete details can be found in the reports accompanying the
CO2PIPETRANS data releases.
3.2. COOLTRANS
National Grid initiated the 3-year TRANSportation of Liquid CO2re-
search programme (COOLTRANS) (Cooper, 2012) at the end of 2010 in or-
der to “address knowledge gaps relating to the safe design and operation of
onshore pipelines for transporting dense-phase CO2from industrial emitters
in the UK to storage sites offshore”. This included developing the capability
for modelling the low-probability, high-impact worst case - an accidental re-
11
lease from a buried pipeline that contains CO2in the dense-phase. Learning
from these studies was subsequently combined with a range of other infor-
mation to develop an appropriate quantified risk assessment (QRA) for a
dense-phase CO2pipeline. The programme included theoretical studies by
University College London (UCL), the University of Leeds and the University
of Warwick, carried out in parallel to provide high-fidelity numerical models
for the pipeline outflow (UCL), near-field dispersion behaviour (University of
Leeds) and far-field dispersion (University of Warwick) behaviour associated
with below-ground CO2pipelines that are ruptured or punctured. Experi-
mental work and studies using currently available practical models for risk
assessment were carried out by DNV GL (Allason et al., 2012). Full details of
the experimental data used herein from this project can be found in Wareing
et al. (2014a). We present the pertinent details of experimental data used in
this work in Table 2.
3.3. CO2PipeHaz
The EU FP7-funded CO2PipeHaz (2010-2013) project “addressed the
fundamentally important and urgent issue regarding the accurate predictions
of fluid phase, discharge rate, emergency isolation and subsequent atmo-
spheric dispersion during accidental releases from pressurised CO2pipelines
to be employed as an integral part of a large scale carbon capture and storage
chain”3. More details of the project can be found in Woolley et al. (2014b).
Details of the experimental data used herein from CO2PipeHaz can be found
in Woolley et al. (2012, 2013a) for Tests 6, 7 and 8, and in Woolley et al.
3http://www.co2pipehaz.eu/overview.htm Accessed 2016-Aug-02.
12
(2013b) for Tests 11, 12 and 13, all concerning releases from a liquid phase
pure CO2reservoir. We present the pertinent details of experimental data
used in this work in Table 3, reproduced from Woolley et al. (2012, 2013a,b).
3.4. HSL data set
In Pursell (2012) data are presented from laboratory-scale CO2release
experiments. Measurements were taken of the outflow and near-field disper-
sion behaviour in an expanding CO2jet, from both liquid and gaseous phase
reservoirs. For this work, we consider two tests: HSL Test C and HSL Test
D, both from a pure CO2liquid phase reservoir. Compared to the other data
sets used in this comparison, it is notable that these measurements are taken
on a smaller scale using 2.0 mm and 4.0 mm diameter nozzles. We present
the pertinent details of experimental data used in this work in Table 4.
3.5. CO2QUEST
Following on from CO2PipeHaz, the most recent CO2dispersion data
comes from the EU FP7-funded CO2QUEST project (2013-2016) that in-
volves ”the collaboration of 12 industrial and academic partners in Europe,
China and Canada and focusses on the development of state-of-the-art math-
ematical models along with the use of large scale experiments to identify the
important CO2mixtures that have the most profound impact on the different
parts of the CCS chain”4. A more complete description of the CO2QUEST
project can be found in Brown et al. (2014b). In this work, we take ex-
perimental data from a number of pure and impure releases performed by
4http://www.co2quest.eu/index.php Accessed 2016-Aug-02.
13
INERIS in France (Proust, Hebrard and Jamois, 2016) and DUT in China
(Chen et al., 2016). Specifically, INERIS tests QUEST T12 and QUEST
T14 concern near-field measurements of pure CO2releases. INERIS tests
QUEST T9, T13, T15, T16, T19 and T22-T25 all concern comparable re-
leases of impure CO2and were designed to be directly comparable to the
pure test results, in order to elucidate any differences introduced by CO2
impurities. Both DUT tests concern pure CO2, but are on the largest scale
that we are aware of that measure near-field conditions relevant to this work.
We present the pertinent details of experimental data used in Table 5.
4. Mathematical model and numerical method
The mathematical model and numerical approach is essentially the same
as that adopted and validated in our earlier papers. Complete details of this
model can be found in Wareing et al. (2013a). This model has previously
been validated in the literature in separate works against the COOLTRANS
data (Wareing et al., 2014a), Tests 6, 7 and 8 of the CO2PIPEHAZ data
(Woolley et al., 2012, 2013a) and the remaining CO2PIPEHAZ data (Tests
11, 12 and 13) in a conference paper (Woolley et al., 2013b). In this paper,
the validation is extended from those 7 tests, to three further independent
datasets - CO2PIPETRANS, HSL and CO2QUEST - that contain 17 further
suitable pure CO2tests. Therefore we are able to investigate the validity of
the model and the consistency of the experimental data in a much larger
number of pure CO2tests. Here, as we are also extending this model to
consider impure CO2dispersion, we summarise details of our original com-
posite equation of state (EoS) for pure CO2and present the implementation
14
of an impure CO2EoS in our model. The nature of the new extension is in
the inclusion of complex new equations of state and associated changes to
the homogeneous model, detailed below. The extended number of datasets
introduce 9 new impure CO2tests against which we can validate the new
impure CO2dispersion model.
4.1. Equation of state for pure CO2
Our composite EoS, as described in Wareing et al. (2013a) predicts the
thermophysical properties of the three phases of CO2for the range of tem-
peratures of relevance to CO2dispersion from releases at sonic velocities i.e.
those of interest to the CCS industry. This EoS is designed to be convenient
for computational fluid dynamic applications; the gas phase is determined
from the Peng-Robinson EoS (Peng and Robinson, 1976), and the liquid and
condensed phases from tabulated data generated with the Span and Wagner
EoS (Span and Wagner, 1996). Previously, we modelled the solid phase using
the DIPPR R
Project 801 database 5, academic access to which can be gained
through the Knovel library 6. We now use the J¨ager and Span EoS (J¨ager
and Span, 2012) for solid CO2. Air is modelled by an ideal gas equation of
state with γa= 7/5.
4.2. Impure CO2
Very few forms of EoS are available for the range of pressure and tem-
perature in these near-field simulations. Specifically, we are aware of two
in the literature that have been validated that include the necessary solid
5http://www.aiche.org/dippr/ Accessed 2016-Aug-02.
6http://why.knovel.com Accessed 2016-Aug-02.
15
phase - EOS-CG (Gernert and Span, 2016) embedded in the TREND soft-
ware (Span et al., 2015), and PC-SAFT (see Diamantonis et al., 2013, and
references therein) embedded in the Physical Properties Library (PPL). Oth-
ers are limited to pipeline pressures and temperatures (e.g. Demetraides and
Graham, 2016) and/or do not model the solid phase - e.g. most cubic EoS,
see the comparisons of Diamantonis et al. (2013) and those in the thesis of
Li (Li, 2008) for more details. Here we examine TREND version 2.0 (ob-
tained by private communication, August 2015) and PPL as developed in
CO2QUEST for impure CO2(obtained through the CO2QUEST collabora-
tion, September 2015). For both, we assume a mixture of 96% CO2and 4%
N2, typical of the type of and maximum level of impurity expected in UK
CCS pipelines Cooper and Barnett (2014). It should be noted that this is
not a mature research area: both TREND and PPL are still in development.
Both currently use the J¨ager and Span (2012) EoS in the solid phase and
assume no solubility of impurity in the solid phase.
As neither EoS has been demonstrated to work in the near-field for an im-
pure CO2release, we first looked at simple decompression from high pressure
CO2to low pressure, both with and without a heat source. The inclusion
of a heat source partially mimics the mixing with ambient air that occurs
during the release. This is described by
dU =−P dV −κ(T−Ta),(1)
where U(P, T ) is the internal energy per unit mass, Pis the pressure, V(P, T )
is the specific volume, Tis the temperature, Tathe (constant) ambient tem-
perature and κa constant that determines the amount of heating.
16
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
180 190 200 210 220 230 240
Pressure (MPa)
Solid fraction (kg/kg)
Temperature (K)
Solid frac. - no heating
Solid frac. - with heating
Pressure - with heating
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
180 190 200 210 220 230 240
Pressure (MPa)
Solid fraction (kg/kg)
Temperature (K)
Solid frac. - no heating
Solid frac. - with heating
Pressure - with heating
(a) TREND (b) PPL
Figure 1: Condensed phase CO2and pressure in a decompression computed with TREND
in (a) and computed with PPL in (b). Solid lines: solid fraction during isentropic decom-
pression (κ= 0). Dashed lines: solid fraction during decompression with heating (κ= 1).
Dotted lines: pressure during decompression with heating (κ= 1).
As can be seen comparing TREND in Figure 1(a) to PPL in Figure 1(b),
there is little difference for the isentropic expansion (solid lines). However,
when a heat source is added, TREND produces a smooth evaporation of
solid CO2(dashed line in Figure 1(a)), but with PPL the solid mass fraction
increases rapidly as atmospheric pressure is reached (shown in Figure 1(b)).
Further investigation has revealed this is because there is a discontinuity
at the dew line at low temperatures with PPL, making this version of PPL
unsuitable for the current dispersion simulations. TREND is used for impure
CO2modelling for the rest of this work. As noted, however, these EoS
are undergoing further development with the intention that both should be
applicable to the types of release of interest in CO2pipeline risk assessments,
and for a wide range of impurities.
17
4.3. Homogeneous model
In previous work on dense-phase CO2releases from small nozzles and
punctures (Wareing et al., 2014a,b), particles of solid CO2do not reach equi-
librium with the CO2gas flow in the initial expansion due to the short
distance between release point and Mach shock when compared to particle
thermal and dynamic relaxation times and velocities (Wareing et al., 2013b).
We therefore used a homogeneous relaxation model to take this into account.
Since the EoS with impurities is only available for equilibrium, we use a very
small relaxation time for pure CO2so that it is very close to equilibrium.
The homogeneous relaxation model is described in Wareing et al. (2013a)
Modifications required for impure CO2are as follows. TREND can be
used to determine the thermophysical properties of impure CO2as functions
of pressure, P, and temperature, T. Since the properties vary rapidly near
the dew line, we use a scaled pressure variable, x, defined by
x= 1 + log e(1−y)+√e2(1−y)−1 y≤1,
x= 1 + log e(1−y)−√e2(1−y)−1 y > 1,
(2)
where
y=P/Pd(T),(3)
and Pd(T) is the pressure at the dew line. This ensures that xvaries rapidly
with Pnear the dew line and linearly elsewhere. Equation (2) can readily
be inverted to give
y= 1 −log [cosh(x−1)] x≤1,
= 1 + log [cosh(x−1)] x > 1.
(4)
18
Given a table in terms of (x, T ), it is simple to generate a table in terms of
(ρ, T ).
A conservative hydrodynamic code works in terms of the total density, ρ,
and total internal energy per unit mass, U, so it is necessary to obtain the
pressure and other quantities given ρand U. Let αbe the mass fraction of
the CO2that is in the condensed phase and βbe the mass fraction of CO2
in a CO2- air mixture. The temperature is found by solving
U=βUCO2(β ρ, T ) + (1 −β)Uair(T),(5)
where UCO2and Uair are the internal energies for CO2and air. The pressure
is then given by
P= (1 −β)RT ρ
[ma(1 −αβρ/ρc)] +PCO2(β ρ, T ),(6)
where ρc(βρ, T ) is the density of condensed phase CO2,mais the molecular
mass of air and α=α(βρ, T ).
This assumes that the CO2- air mixture behaves as if the CO2has density
βρ and temperature T. This is obviously not true at high pressures where the
behaviour of the gaseous phase departs from the ideal gas equation of state.
Fortunately, the mixing between CO2and air occurs at atmospheric pressure
and temperatures significantly below the triple point. In this regime the
gaseous phase does obey the ideal equation of state and one can also neglect
the solubility of nitrogen and oxygen in solid CO2(as TREND does).
19
4.4. Implementation
Both equations of state are implemented via look-up tables in MG, an
adaptive mesh refinement (AMR) RANS hydrodynamic code (Falle, 1991).
The code employs an upwind, conservative shock-capturing scheme and is
able to employ multiple processors through parallelisation with the message
passing interface (MPI) library. Integration in time proceeds according to a
second-order accurate upwind scheme with a Harten Lax van-Leer (van Leer,
1977; Harten et al., 1983) (HLL) Riemann solver to aid the implementation
of complex EoS. The code also uses AMR (Falle, 2005), which reduces the
memory and computation time by an order of magnitude. Further details
can be found in Wareing et al. (2013a) and in the references above.
5. Numerical methodology
In computationally simulating the releases considered below, we employed
the same methodology as Wareing et al. (2013b), solving Favre-averaged,
density-weighted forms of the transport equations for mass, momentum, to-
tal energy (internal energy plus kinetic energy) and scalar transport, closed
using a compressibility-corrected version of the k−turbulence model. We
used a two-dimensional cylindrical polar axisymmetric coordinate system.
Numerical simulations were performed employing the inlet conditions listed
in Table 6 as input conditions in the region defined by r < 0.5D(dimen-
sions are scaled by the vent exit diameter, D) on the z= 0 boundary. The
initial state of the fluid in the domain consists entirely of stationary air at a
pressure and temperature given in Table 6. Conditions in air are calculated
via an ideal gas equation of state with γa= 7/5. The r= 0 axis was treated
20
as symmetric and the other rboundary as free flow, introducing air with
the initial atmospheric condition if an in-flow was detected. This neglects
the effects of a cross-flow in the atmosphere, but this is a reasonable approx-
imation to make over the near-field range, where the momentum from the
release is expected to dominate, and is supported by previous work (Wareing
et al., 2013b; Woolley et al., 2013a). The z= 0 axis was fixed by the input
conditions for r < 0.5Dand as a solid wall outside this region, ignoring any
ability of the release to entrain air from behind the inlet for the purposes of
this work. The other zaxis was free-flow, again only allowing the in-flow of
air with the initial atmospheric condition if in-flow was detected, for exam-
ple as a result of vortices formed before the jet reaches steady state. Given
that vortex structures may be present in the jet as it reaches steady state,
velocities that lead to inflow can occur at the free-flow boundaries. Hence
the boundary conditions are adjusted to ensure that only ambient air can
flow into the domain, with the same properties as the initial condition, and
no CO2.
For the purposes of comparison to experimental data, we show predic-
tions extracted from five numerical simulations in later figures. The inlet
conditions for these simulations are shown in Table 6 and are enforced on
every step at the z= 0 boundary for r < 0.5D. The first three inlet condi-
tions have the same pressure and temperature and only vary liquid fraction
at the nozzle to demonstrate the importance of this key parameter. They are
based on COOLTRANS T7 and we refer the interested reader to Wareing
et al. (2014a) for full details of these inlet conditions. The fourth set of inlet
conditions are matched to QUEST T12 concerning a release of pure CO2in
21
the CO2QUEST project. The fifth set of inlet conditions idealise the fourth
set of conditions for a mixture of 96% CO2and 4% N2, typical of impurity
levels expected in pipeline transport of CO2from power generation (Cooper
and Barnett, 2014; Porter et al., 2015). Specifically, they achieve the same
mass-flow rate and hence are directly comparable to the fourth set of inlet
conditions and also the impure CO2tests from CO2QUEST. Both the fourth
and fifth sets of inlet conditions were obtained from UCL as part of the
CO2QUEST project. We have chosen to idealise the fifth set of inlet con-
ditions in this way in order to explore whether there is a difference between
pure and impure dispersion predictions that otherwise keep the dominating
release factor - mass-flow - the same. There is no such pair of experiments
in the CO2QUEST dataset. Pure Test 12 and impure Tests 14 and 15 are
all similar. In the case of Test 14, the impurity and level of impurity is very
similar to Test 12, but the pressure and temperature are different and hence
mass flow rate is different. In the case of Test 15, the pressure and tem-
perature are similar to the idealised case, but the impurities are different,
even if the total level is very similar. If we had chosen for the fifth set of
inlet conditions to model, e.g. Test 14, the mass-flow rate would have been
different and hence the Mach shock would have been in a different location.
This is a valid thing to do, but outside the scope of this work as we note in
our conclusions.
6. Model comparisons with data
In the following figures we show comparisons of data and dispersion pre-
dictions. The plotted data points indicate experimental measurements of the
22
temperature in the dispersion plume at that location and are the simple aver-
age for that particular sensor during the steady state period. The experimen-
tal data during the steady state period has a variance on each measurement
during the relevant time period of a degree or two. The temperature sensors
used are typically accurate over the observed range to within ±5 K at worst,
hence throughout all the plots shown the experimental data points should
be assumed to have ±5 K error bars, although for clarity these error bars
have not been plotted in the figures. The response time of the sensors is less
than a second in all cases and hence less than the steady state period, which
is at a minimum of 3 seconds, up to tens of seconds for some of the small
diameter, large reservoir releases. Predictions shown are taken from steady
state 2D axisymmetric simulations for the various inlet conditions noted in
the figures, the full details of which were presented in the previous section
and examples of which can be seen in Wareing et al. (2013a,b)
In Figure 2 we compare centreline temperature predictions with varying
liquid fraction at the release point to centreline experimental temperature
data. The first thing to note is that the experimental data is coincident within
the ±5 K error up to 100 D, at temperatures around 190 K. After 100 D, the
different experimental datasets diverge, each warming towards ambient con-
ditions at a different rate. Scaling the centreline distance in nozzle diameters
has created a nozzle-diameter-independent plotting of the experimental data.
The predictions of temperature drop very rapidly through the Mach shock,
which in this figure is at ∼8 D, so is visible by the drop to 166 K pre-shock
and then the rise to 194.25 K post-shock. For the first 100 D the predic-
tions are slightly warm compared to the data. The pure CO2predictions
23
160
170
180
190
200
210
220
230
240
250
260
270
280
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850
Temperature (K)
Distance from nozzle (D)
DS1 Test 2
DS1 Test 5
DS1 Test 11
DS2 Test 3
DS2 Test 5
DS2 Test 11
HSL Test C
HSL Test D
COOLTRANS T7
HAZ T6
HAZ T7
HAZ T8
HAZ T11
HAZ T12
HAZ T13
100% liquid CO2
80% liquid CO2
60% liquid CO2
Figure 2: Experimental data (points) versus predictions (lines) of temperature along the
centreline of an expanding liquid CO2jet release through a nozzle/vent/puncture from/in
a high pressure reservoir. Multiple datasets are identified in the legend and every datapoint
should be assumed to have a ±5 K error, not shown in the figure. Predictions are shown
based on the COOLTRANS T7 inlet conditions for 60%, 80% and 100% liquid fraction at
the nozzle (see Table 6).
24
shown here employ the homogeneous relaxation model in the same way as
used in Wareing et al. (2013b), which allows the gas temperature to vary
compared to the solid temperature beyond the Mach shock. An equilibrium
model would necessarily have a temperature of 194.25 K immediately after
the Mach shock as this is the sublimation temperature at which gas and solid
co-exist at atmospheric pressure. Evaporation of the solid then cools the flow
along the centreline, as noted by us and other authors (see Section 2 for a
discussion of this effect). Within error, the experimental data satisfyingly
falls between these two extremes. The smallest nozzle diameter experiments
(HSL Tests C and D) are the warmest at this point, which is to be expected
as the solid particles are the furthest from equilibrium due to the proximity
of the shock to the release point.
Further than 100 D along the centreline, the 100% liquid fraction predic-
tion clearly remains the coldest of the three predictions into the far-field,
bracketing the coldest extreme of the data at specific centreline distances,
typically CO2PIPETRANS DS2 Test 11. This suggests that DS2 Test 11
was very close to 100% liquid at the release point. At the warm extremes
of the data going into the far-field are HSL Tests C and D. These are cap-
tured by the 60% liquid prediction, although inferred rates in Table 4 are
higher than that. This discrepancy could be related to the true liquid frac-
tion at the nozzle or the fact that these tests are somewhat different to the
other tests shown here, as they are the smallest diameter nozzles and were
also performed in controlled laboratory conditions rather than outdoors. We
have also shown in previous work that the choice of turbulence model in the
numerical prediction affects the plume temperature into the far-field. Here
25
we have used a compressibility corrected k−model described in detail in
Wareing et al. (2013a). So far, we have found this model provides the best fit
to averaged CO2temperature dispersion data. Further work considering the
unsteady jets produced by Reynolds-stress turbulence models will investigate
this further. The 80% liquid prediction falls half-way between the 60% and
100% liquid predictions, suggesting that DS1 Test 11 may have had a liq-
uid fraction at the nozzle around 80%. The variation in experimental data,
though all are releases from high pressure (close to) 100% liquid reservoirs,
can therefore be explained by variation of the liquid fraction at the release
point, as differences in the experimental apparatus range from direct release
from the reservoir, to release along elongated narrow-diameter pipes.
In Figure 3, we show radial comparisons of data and prediction at 80D,
∼100 D, 165 D and 400 D along the centreline of the jet. Whilst we show as
many datasets as possible, here we only show the prediction for 100% liquid
CO2, which as might be expected brackets the lower temperature extreme
of the data. Agreement between the multiple datasets across the wide range
of centreline locations is clear from the figures, as well as a good fit by
the predicted temperature in the dispersion plume. The advantage of this
comparison is that it is now possible to clearly identify experimental data
which is not comparable to the majority. The temperature of 210 K in the
case of COOLTRANS T7 at z=165 D and r=32 D appears to be an outlier
compared to the other measurements, both from that test and others. The
identification of such points should simplify future experimental and model
validation work.
Figure 4 shows a near-field comparison for both pure CO2releases (in
26
170
180
190
200
210
220
230
240
250
260
270
280
290
0 10 20 30 40
Temperature (K)
Radial distance from centreline (D)
(a) Radial comparisons at z=80D
DS1 T1
DS1 T5
DS1 T11
DS2 T3
DS2 T5
DS2 T11
HSL Test C
HSL Test D
HAZ T7
HAZ T8
100% liquid CO2
170
180
190
200
210
220
230
240
250
260
270
280
290
0 10 20 30 40 50
Temperature (K)
Radial distance from centreline (D)
(b) Radial comparisons at z~100D
DS1 T5 (117D)
DS2 T5 (118D)
HSL Test C (106.25D)
HSL Test D (112.5D)
HAZ T6 (111D)
HAZ T13 (100D)
100% liquid CO2
170
180
190
200
210
220
230
240
250
260
270
280
290
0 10 20 30 40 50 60 70
Temperature (K)
Radial distance from centreline (D)
(c) Radial comparisons at z=165D
DS1 T2
DS1 T5
DS1 T11
DS2 T3
DS2 T5
DS2 T11
COOLTRANS T7
HAZ T7
100% liquid CO2
210
220
230
240
250
260
270
280
290
0 20 40 60 80 100 120 140
Temperature (K)
Radial distance from centreline (D)
(d) Radial comparisons at z=400D
DS2 T3
DS2 T5
DS2 T11
HAZ T7
HAZ T11
HAZ T12
100% liquid CO2
Figure 3: Experimental data (points) versus predictions (lines) for radial temperature
distributions at various locations along the centreline of an expanding liquid CO2jet release
through a nozzle/vent/puncture from/in a high pressure reservoir. Multiple datasets are
identified in the legends and every data point should be assumed to have a ±5 K error,
not shown in the figures. Predictions are shown based on the COOLTRANS T7 inlet
conditions for 100% liquid fraction at the nozzle (see Table 6).
27
160
170
180
190
200
210
220
230
240
250
260
270
280
0 5 10 15 20 25 30 35 40
Temperature (K)
Distance from nozzle (D)
DS1 T5
DS2 T5
HSL Test C
HSL Test D
HAZ T8
HAZ T11
HAZ T12
HAZ T13
QUEST T1
QUEST T2
QUEST T6
QUEST T7
QUEST T8
QUEST T12
QUEST T14
DUT Trial 1
DUT Test 1
100% liquid CO2
0 5 10 15 20 25 30 35 40
Distance from nozzle (D)
QUEST T9 - 1.6%CH4
QUEST T13 - 4.4%CH4
QUEST T15 - 4.5%N2
QUEST T16 - 2%CH4,2%N2
QUEST T19 - 4.1%CH4
QUEST T22 - 4.1% CH4
QUEST T23 - 4.5% N2
QUEST T24 - 2.1% N2 1.9% CH4
QUEST T25 - 4.5% N2
Pure CO2 - composite EoS
Impure CO2 - TREND EoS
(a) Pure CO2 (b) Impure CO2
Figure 4: Experimental centreline near-field temperature data (points) for (a) pure and (b)
impure releases of CO2from a high pressure reservoir. Multiple datasets are identified in
the legends and every data point should be assumed to have a ±5K error, not shown in the
figures. Shown also are predictions of the temperature (lines) based on the COOLTRANS
T7 inlet conditions for 100% liquid fraction in (a) and based on the pure and impure CO2
inlet conditions in (b), matching reservoir conditions in QUEST T12 (see Table 6).
28
(a)) and impure CO2releases (in (b)). Near-field data and predictions up
to 40 D from the release point are shown for both cases. In the pure CO2
case, data from all five sources is shown to be remarkably consistent, apart
from the largest field-scale DUT tests, which given the release diameters
of 50 mm and 233 mm are large-flow rate releases at the limit of near-field
measurement capability. Larger scale experiments have been performed, but
with limited near-field measurements using different methods; the numerical
model used has been shown to be able to predict such experiments (Ware-
ing et al., 2014a, 2015a,b). Within the Mach shock structure (z < 8D),
experiment and prediction do not agree well, although the data shows some
indication of upward trending temperature close to the release point. Given
the rapid variation of pressure from tens of atmospheres to fractions of a
percent of atmospheric just before the Mach shock, the rapid expansion and
decrease in density and the large acceleration from typically 100m s−1at
the nozzle to typically 400 m s−1at the shock, it is perhaps not surprising
that thermocouples designed to measure steady-state flows at atmospheric
pressure struggle to capture the extreme gradients in the expanding jet. Be-
yond the Mach shock, temperatures around the sublimation temperature of
194.25 K at atmospheric pressure are again seen. The spread of the data is in
good agreement with this, bar the large-scale DUT Test 1 which has already
been discussed and QUEST T2 which would seem to possess a warming trend
suggestive of low liquid fraction. For future modelling, most of the datasets
would appear to be close to equilibrium, behaving in the way expected from
previous insights - the temperatures post-Mach-shock slowly drop from the
sublimation temperature until all the solid has evaporated and then the dis-
29
persion plume begins to warm. The point at which this occurs is not shown
in this figure and varies with liquid fraction at the nozzle, but can be seen
in Figure 2.
Turning now to releases of impure CO2as shown in Figure 4b, multiple
impure experimental datasets and two predictions are shown for compara-
tive purposes. QUEST Tests 12, 13, 14, 15 and 16 were designed to be di-
rectly comparable for the purposes of differentiating the effects of impurities.
The datasets are remarkably consistent but it should be noted that they are
entirely sourced from INERIS through CO2QUEST. The closest near-field
temperature measurements for small-scale releases have been obtained and
all seem to agree on 194.25 K. The predictions, which model QUEST T12
and its impure idealisation, would indicate the measuring point is very close
to the Mach shock. The pure prediction (solid black line) models the re-
lease conditions from QUEST T12. The impure prediction is idealised to
match the same release temperature and mass-flow as QUEST T12 (for the
reasons set out in Section 5), and so is reasonably comparable to QUEST
T15 with 4.5% N2for the level of impurity and QUEST T16 for the pressure
and temperature and hence mass-flow. Perhaps not surprisingly for matched
mass-flow, the two predictions are very similar - the low-level of impurity
has very little effect and the position of the Mach shock is dominated by the
initial pressure and mass flow rate. The recorded dispersion temperatures
are very similar to that in pure CO2experiments and show no clear difference
between type of impurity and total amount of impurity. Near-field tempera-
ture measurements of the dispersing plume alone clearly cannot differentiate
between pure and impure CO2in this case. This is supported by the close
30
similarity of pure and impure predictions - the experimental accuracy of ±5K
is far greater than the difference between the predictions. This data would
be hard pushed to differentiate between pure and impure releases in the near-
field on that basis. We would like to reassure the reader that the EoS data
used in this work does pick up a difference between pure and impure pre-
dictions - the presence of a two-phase region between the bubble and dew
lines in the tabulated impure EoS data confirms this. Further experiments
with different experimental setups are required to investigate differences be-
tween pure and impure CO2releases, at least for these levels of impurity.
For example, the impure data shown here indicates a difference between the
QUEST datasets T9 to T19 and T22-T25. Beyond the Mach shock, T22-T25
are characteristically warmer than the other tests. These are larger diameter
tests, but in the middle of the range of nozzle sizes shown in previous fig-
ures, so it is not clear why they should be characteristically warmer than the
other QUEST tests. Further work is required here to elucidate the reasons
for these experimental differences, including experiments designed to show
differences between pure and impure that are greater than the experimental
accuracy. Through this work, future modelling efforts should now be able to
refine their experimental validation procedures.
7. Conclusions
In this work we have performed the first multiple-dataset comparison
between experimental data and numerical predictions for the dispersion of
high pressure liquid CO2from CCS transport pipeline scenarios into dry air,
extending previous validation from 7 tests to a total of 24 tests. A simple
31
non-dimensionalisation of experimental data according to nozzle diameter
used here has provided the means to compare these multiple datasets from
different projects and experimental procedures. It has revealed remarkable
consistency between the experimental datasets. Predictions compare well to
the experimental data and highlight the fact that liquid fraction reducing
mass flow rate at the release point is of key importance in modelling CO2
dispersion. It is our hope that the presentation of data from multiple sources
in this extended validation highlights the differences between experimental
tests and will aid researchers looking to validate dispersion models in the
future.
Turning to impure CO2, for the limited range of low-level impurity con-
sidered here, no clear difference to pure CO2releases is discernible in temper-
ature dispersion data. This is reflected by the close agreement between pure
and impure predictions of temperature in the dispersion plume. Differences
between experimental datasets are noted. The publication of this work in
the literature should allow future modelling work to account for these differ-
ences in their validation procedures. Further experimental work is required
to discern any differences between such pure and impure CO2dispersion. It
is worth noting that we chose numerical inlet conditions for the pure and im-
pure cases in order to match mass-flow and explore whether an impurity level
of 4% N2made any difference to the numerical predictions, as this was one
aim of CO2QUEST. Real-life pipelines and transport facilities are likely to
set pressure and temperature specifications. An impurity will therefore alter
the density of the mixture and hence the mass-flow through a given orifice.
Further exploration of such differences between transport-facility conditions
32
is required to quantify the effects of impurities further.
The numerical prediction of impure CO2dispersion requires complex
equations of state. The use of TREND here has shown there is little dif-
ference between pure and impure temperature predictions for low-levels of
impurity (4% N2), but further work is required as the necessary impure equa-
tions of state are still under development and the available experimental data
for validation is limited.
We have not considered the presence of water vapour in the air. The
region in which water vapour will make a difference in the near-field is limited
- the predictions have shown that no air mixes into the centre of the jet
until approximately 40 release diameters downstream from the release point.
Water vapour cannot affect this part of the jet, which includes the near-field
Mach shock. The centreline predictions shown herein will be unchanged.
Where water vapour in the air does become important is on the edges of
the jet. Water droplets will condense once the temperature is below the dew
point - this chiefly defines the visible extent of the jet. Further into the
mixing region, at lower temperatures, water ice will form. Since water has
a latent heat of fusion approximately five times greater than that of CO2, it
will act as an energy sink causing the CO2jet to be less cold (on the order of
a few degrees at most in the edges of the near-field jet, depending on the level
of water vapour in the air). This may change the gradient of the predictions
in our radial temperature predictions (Figure 6), but clearly warming the
prediction in the mixing region close to the edge of these predictions is not
desirable as it will not improve this fit to data. CO2hydrates may also form.
This adds an extra degree of complexity to the equation of state, and EoSs
33
are only now considering this complexity. Here we have presented a first
order modelling solution, which still requires very complex EoS. We leave
the accurate inclusion of water vapour and CO2hydrate formation to future
EoS and future dispersion calculations.
Acknowledgements
The research leading to the results described in this paper has received
funding from the European Union 7th Framework Programme FP7-ENERGY-
2012-1-2STAGE under grant agreement number 309102. The paper reflects
only the authors views and the European Union is not liable for any use that
may be made of the information contained herein. We acknowledge support
from the research groups responsible for TREND and PPL. We thank both
INERIS and DUT for provision of data through CO2QUEST and for per-
mission to include it in this publication. CO2PIPETRANS data sets are
publicly available - see Section 3. COOLTRANS, HSL and CO2PIPEHAZ
datasets are available from the sources noted in Section 3. We thank UCL
(S.Martynov, H.Mahgerefteh) for the provision of inlet conditions. We also
thank the Guest Editors (MF, H. Mahgerefteh, N. Mac Dowell, N. Røkke, R.
Span and S.T. Munkejord) for the invitation to convert our presentation at
the 2nd CCS Forum (2016 Dec. 16-17, Athens, Greece) into this article for
this Special Issue of the International Journal of Greenhouse Gas Control.
The numerical predictions for this paper were performed on the DiRAC Fa-
cility jointly funded by STFC, the Large Facilities Capital Fund of BIS and
the University of Leeds, hosted and enabled through the ARC HPC resources
and support team at the University of Leeds (A. Real, M. Dixon, M. Wallis,
34
M. Callaghan and J. Leng).
References
Ahmad, M., B¨ogemann-van Osch, M., Buit, L., Florisson, O., Hulsbosch-
Dam, C., Spruijt, M., Davolio, F., 2013. Study of the thermohydraulics
of CO2discharge from a high pressure reservoir. Int. J. Greenhouse Gas
Control 19, 63–73.
Allason, D., Armstrong, K., Cleaver, P., Halford, A., Barnett, J., 2012. Ex-
perimental studies of the behaviour of pressurised release of carbon dioxide.
In: IChemE Symposium Series No. 158, IChemE. pp. 142–152.
Brown, J., Holt, H., Helle, K., 2014a. Large Scale CO2Releases for Dispersion
Model and Safety Study Validation. Energy Procedia 63, 2542–2546.
Brown, S., Martynov, S., Mahgerefteh, H., Fairweather, M., Woolley, R.M.,
Wareing, C.J., Falle, S.A.E.G., Rutters, H., Niemi, A., Zhang, Y.C.,
Chen, S., Bensabat, J., Shah, N., Mac Dowell, N., Proust, C., Farret, R.,
Economou, I.G., Tsangaris,D.M., Boulougouris, G.C., Van Wittenberghe,
J., 2014b. CO2QUEST: Techno-economic Assessment of CO2Quality Ef-
fect on Its Storage and Transport. Energy Procedia 63, 2622–2629.
Chen, S., Jianliang, Y., Gui, X., Zhang, Y., Yan, X., Proust, C., Hebrard, J.,
2016. CO2QUEST Deliverable 2.5 Report: ”Experimental studies of the
atmospheric dispersion of CO2at a large scale and near pipeline crack”.
Dalian University of Technology, China; INERIS, France.
35
Connolly, S., Cusco, L., 2007. Hazards from high pressure carbon dioxide
releases during carbon dioxide sequestration processes. In: IChemE Sym-
posium Series No. 153, IChemE. pp. 1–5.
Cooper, R., 2012. National Grid’s COOLTRANS research programme. Jour-
nal of Pipeline Engineering 11, 155–172.
Cooper, R., Barnett, J., 2014. Pipelines for transporting CO2in the UK.
Energy Procedia 63, 2412–2431.
Demetriades, T.A., Graham, R.S., 2016. A new equation of state for CCS
pipeline transport: Calibration of mixing rules for binary mixtures of CO2
with N2, O2and H2. The Journal of Chemical Thermodynamics 93, 294–
304.
Diamantonis, N.I., Boulougouris, G.C., Tsangaris, D.M., El Kadi, M.J.,
Saadawi, H., Nagahban, S., Economou, I.G., 2013. Thermodynamic and
transport property models for carbon capture and sequestration (CCS)
with emphasis on CO2transport. Chemical Engineering Research and
Design 91, 1793–1806.
Dixon, C.M., Gant, S.E., Obiorah, C., Bilio, M., 2012. Validation of disper-
sion models for high pressure carbon dioxide releases, in: IChemE Sympo-
sium Series No. 158, IChemE. pp. 153–163.
Dixon, C.M., Hasson, M., 2007. Calculating the release and dispersion of
gaseous, liquid and supercritical CO2, in: IMechE Seminar on Pressure
Release, Fires and Explosions, London, IMechE.
36
Dixon, C.M., Heynes, O., Hasson, M., 2009. Assessing the hazards associated
with release and dispersion of liquid carbon dioxide on offshore platforms,
in: 8th World Congress of Chemical Engineering, Montreal.
Falle, S.A.E.G., 1991. Self-similar jets. Monthly Notices of the Royal Astro-
nomical Society 250, 581–596.
Falle, S.A.E.G., 2005. AMR applied to non-linear elastodynamics, in: Plewa,
T., Linde, T., Weirs, V.G. (Eds.), Proceedings of the Chicago Workshop
on Adaptive Mesh Refinement Methods, Springer Lecture Notes in Com-
putational Science and Engineering v.41, Springer, New York U.S.A.. pp.
235–253.
Gant, S.E., Narasimhamurthy, V.D., Skjold, T., Jamois, D., Proust, C., 2014.
Evaluation of multi-phase atmospheric dispersion models for application
to Carbon Capture and Storage. Journal of Loss Prevention in the Process
Industries 32, 286–298.
Gernert, J., Span, R., 2016. EOSCG: A Helmholtz energy mixture model for
humid gases and CCS mixtures. The Journal of Chemical Thermodyanmics
93, 274–293.
Harten, A., Lax, P.D., van Leer, B., 1983. On upstream differencing and
Godunov-type schemes for hyperbolic conservation laws. SIAM Review
25, 35–61.
Hill, T.A., Fackrell, J.E., Dubal, M.R., Stiff, S.M., 2011. Understanding the
consequences of CO2leakage downstream of the capture plant. Energy
Procedia 4, 2230–2237.
37
Hulsbosch-Dam, C.E.C., Spruijt, M.P.N., Necci, A., Cozzani, V., 2012a.
As approach to carbon dioxide particle distribution in accidental releases.
Chemical Engineering Transactions 26, 543–548.
Hulsbosch-Dam, C.E.C., Spruijt, M.P.N., Necci, A., Cozzani, V., 2012b.
Assessment of particle size distribution in CO2accidental releases. J. Loss
Prevention in the Process Industries 25, 254–262.
J¨ager, A., Span, R., 2012. Equation of state for solid carbon dioxide based
on the Gibbs free energy. J. Chem. Eng. Data. 57, 590–597.
Li, H., 2008. Thermodynamic Properties of CO2Mixtures and Their Appli-
cations in Advanced Power Cycles with CO2Capture Processes . Royal
Institute of Technology, Stockholm, Sweden.
Liu, Y., Calvert, G., Hare, C., Ghadiri, M., Matsusaka, S., 2012b. Size
measurement of dry ice particles produced from liquid carbon dioxide.
Journal of Aerosol Science 48, 1–9.
Liu, X., Godbole, A., Lu, C., Michal, G., Venton, P., 2014. Source strength
and dispersion of CO2releases from high-pressure pipelines: CFD model
using real gas equation of state. Applied Energy 126, 56–68.
Mazzoldi, A., Hill, T., Colls, J.J., 2008a. CFD and Gaussian atmospheric dis-
persion models: a comparison for leak from carbon dioxide transportation
and storage facilities. Atmos. Environ. 42, 8046–8054.
Mazzoldi, A., Hill, T., Colls, J.J., 2008b. CO2transportation for carbon
capture and storage: sublimation of carbon dioxide from a dry ice bank.
International Journal of Greenhouse Gas Control 2, 210–218.
38
Mazzoldi, A., Hill, T., Colls, J.J., 2011. Assessing the risk for CO2trans-
portation within CCS projects, CFD modelling. International Journal of
Greenhouse Gas Control 5, 816–825.
Peng, D.Y., Robinson, D.B., 1976. A new two-constant equation of state.
Industrial and Engineering Chemistry: Fundamentals 15, 59–64.
Porter, R.T.J., Fairweather, M., Pourkashanian, M., Woolley, R.M., 2015.
The range and level of impurities in CO2streams from different carbon
capture sources. Int. J. Greenhouse Gas Control 36, 161–174.
Proust, C., Hebrard, J., Jamois, D., 2016. CO2QUEST Deliverable 2.4
Report: ”Experimental investigation of the dispersion of impure CO2
in the atmosphere at a small/medium scale”. L’Institut National de
l’Environnement Industriel et des Risques (INERIS), Departmen PHDS,
Parc Technologique ALATA, BP 2, 60550 Verneuil-en-Halatte, France.
Pursell, M., 2012. Experimental investigation of high pressure liquid CO2
release behaviour. In: IChemE Symposium Series No. 158, IChemE. pp.
164-171.
Span, R., Wagner, W., 1996. A new equation of state for carbon dioxide
covering the fluid region from the triple-point temperature to 1100 K at
pressures up to 800 MPa. Journal of Physical and Chemical Reference
Data 25, 1509–1596.
Span, R., Eckermann, T., Herrig, S., Hielscher, S., J¨ager, A., Thol, M.,
2015. TREND. Thermodynamic Reference and Engineering Data 2.0.1..
Lehrstuhl fuer Thermodynamik, Ruhr-Universitaet Bochum.
39
van Leer, B., 1977. Towards the ultimate conservative difference scheme.
IV. A new approach to numerical convection. Journal of Computational
Physics 23, 276–299.
Wareing, C., Woolley, R.M., Fairweather, M., Falle, S.A.E.G., 2013a. A
composite equation of state for the modelling of sonic carbon dioxide jets
in carbon capture and storage scenarios. AIChE Journal 59, 3928–3942.
Wareing, C., Fairweather, M., Peakall, J., Keevil, G., Falle, S.A.E.G., Wool-
ley, R.M., 2013b. Numerical modelling of particle-laden sonic CO2jets
with experimental validation, in: Zeidan, D. (Ed.), AIP Conference Pro-
ceedings of the 11th International Conference of Numerical Analysis and
Applied Mathematics, AIP Publishing. pp. 98–102.
Wareing, C., Fairweather, M., Falle, S.A.E.G., Woolley, R.M., 2014a. Valida-
tion of a model of gas and dense phase CO2jet releases for carbon capture
and storage application. Int. J. Greenhouse Gas Control 20, 254-271.
Wareing, C., Fairweather, M., Falle, S.A.E.G., Woolley, R.M., 2014b. Mod-
elling punctures of buried high-pressure dense phase CO2pipelines in CCS
applications. Int. J. Greenhouse Gas Control 29, 231-247.
Wareing, C., Woolley, R.M., Fairweather, M., Peakall, J., Falle, S.A.E.G.,
2015b. Numerical modelling of turbulent particle-laden sonic CO2jets with
experimental validation. Procedia Engineering 102, 1621-1629.
Wareing, C., Fairweather, M., Falle, S.A.E.G., Woolley, R.M., 2015a. Mod-
elling ruptures of buried high-pressure dense phase CO2pipelines in carbon
40
capture and storage applications - Part I. Validation. Int. J. Greenhouse
Gas Control 42, 701-711.
Wareing, C., Fairweather, M., Falle, S.A.E.G., Woolley, R.M., 2015a. Mod-
elling ruptures of buried high-pressure dense phase CO2pipelines in car-
bon capture and storage applications - Part II. A full-scale rupture. Int.
J. Greenhouse Gas Control 42, 712-728.
Wareing, C., Fairweather, M., Woolley, R.M., Falle, S.A.E.G., 2016a. Mod-
elling particle evolution in turbulent high pressure sonic CO2jets, in: Pro-
ceedings of the 8th International Symposium on Turbulence, Heat and
Mass Transfer, Begell House Inc., New York. pp. in press.
Webber, D.M., 2011. Generalising two-phase homogeneous equilibrium
pipeline and jet models to the case of carbon dioxide. J. Loss Preven-
tion in the Process Industries 24, 356–360.
Wen, J., Heidari, A., Xu, B., Jie, H., 2013. Dispersion of carbon dioxide
from vertical vent and horizontal releases - a numerical study. Proceedings
of the Institution of Mechanical Engineers, Part E: Journal of Process
Mechanical Engineering 227, 125–139.
Wilday, A.J., McGillivray, A., Harper, P., Wardman, M., 2009. A compar-
ison of hazard and risks for carbon dioxide and natural gas pipelines. In:
IChemE Symposium Series No. 155, IChemE. pp. 392–398.
Witlox, H.W.M., Harper, M., Oke, A., 2009. Modelling of discharge and
atmospheric dispersion for carbon dioxide release. J. Loss Prevention in
the Process Industries 22, 795–802.
41
Witlox, H.W.M., Stene, J., Harper, M., Nilsen, S.H., 2011. Modelling of
discharge and atmospheric dispersion for carbon dioxide releases including
sensitivity analysis for wide range scenarios. Energy Procedia 4, 2253–
2260.
Woolley, R.M., Proust, C., Fairweather, M., Falle, S.A.E.G., Hebrard, J.,
Jamois, D., Wareing, C.J., 2012. Experimental Measurement and RANS
Modelling of Multiphase CO2 Jet Releases, in: Hanjalic, K., Nagano, Y.,
Borello, D. and Jakirlic, S. (Eds.), Proceedings of the 7th International
Symposium on Turbulence, Heat and Mass Transfer, Begell House Inc.,
New York. pp. 661–664.
Woolley, R.M., Fairweather, M., Wareing, C.J., Falle, S.A.E.G., Proust, C.,
Hebrard, J., Jamois, D., 2013a. Experimental measurement and Reynolds-
averaged Navier-Stokes modelling of the near-field structure of multi-phase
CO2jet releases. Int. J. Greenhouse Gas Control 18, 139–149.
Woolley, R.M., Proust, C., Fairweather, M., Falle, S.A.E.G., Wareing, C.J.,
Hebrard, J., Jamois, D., 2013b. Measurement and RANS Modelling
of Large-Scale Under-Expanded CO2 Releases for CCS Applications, in:
Simos, T.E., Psihoyios, G., and Tsitouras, C. (Eds.), AIP Conference Pro-
ceedings 1558: Proceedings of the 11th International Conference of Numer-
ical Analysis and Applied Mathematics 2013 (ICNAAM 2013), American
Institute of Physics. pp. 107–111.
Woolley, R.M., Fairweather, M., Wareing, C.J., Proust, C., Hebrard, J.,
Jamois, D., Narasimhamurthy, V.D., Storvik, I.E., Skjold, T., Falle,
42
S.A.E.G., Brown, S., Mahgerefteh, H., Martynov, S., Gant, S., Tsan-
garis,D.M., Economou, I.G., Boulougouris, G.C., Diamantonis, N.I..,
2014a. An integrated, multi-scale modelling approach for the simulation
of multiphase dispersion from accidental CO2pipeline releases in realistic
terrain. Int. J. Greenhouse Gas Control 27, 221–238.
Woolley, R.M., Fairweather, M., Wareing, C.J., Falle, S.A.E.G.,
Mahgerefteh, H., Martynov, S., Brown, S., Narasimhamurthy, V.D.,
Storvik, I.E., Saelen, L., Skjold, T., Economou, I.G., Tsangaris,D.M.,
Boulougouris, G.C., Diamantonis, N.I.., Cusco, L, Wardman, M., Gant,
S., Wilday, J., Zhang, Y.C., Chen, S., Proust, C., Hebrard, J., Jamois, D.,
2014b. CO2PipeHaz: Quantitative Hazard Assessment for Next Genera-
tion CO2Pipelines. Energy Procedia 63, 2510–2529.
43
Table 1: Initial conditions for the CO2PIPETRANS tests included here, reproduced from
CO2PIPETRANS test descriptions.
Description DS1 T2 DS1 T5 DS1 T11 DS2 T3 DS2 T5 DS2 T11
Pressure (barg) 155.5 157.68 82.03 147.3 148.8 80.3
Temperature (degC) 7.84 9.12 17.44 9.8 17.8 -0.2
CO2fraction 1.00 1.00 1.00 1.00 1.00 1.00
Condensed phase fraction 1.00 1.00 1.00 1.00 1.00 1.00
Orifice diameter (mm) 11.94 25.62 11.94 12.7 25.4 12.7
Atmospheric conditions
Pressure (mbara) 958.2 985.4 960.2 1017 905 995
Temperature (degC) 7.5 5.8 11.6 11.2 9.0 3.6
Relative humidity (%) 96 97 94 66 91 78
44
Table 2: Initial conditions for the COOLTRANS test included here, reproduced from
(Wareing et al., 2014a).
Description COOLTRANS T7
Pressure (MPa) 15.00
Temperature (degC) 7.45
CO2fraction 1.00
Condensed phase fraction 1.00
Orifice diameter (mm) 25.4
Atmospheric conditions
Pressure (MPa) 0.1
Temperature (degC) 7.45
45
Table 3: Initial conditions for the CO2PipeHaz tests included here, reproduced from
(Woolley et al., 2012, 2013a,b).
Description HAZ T6 HAZ T7 HAZ T8 HAZ T11 HAZ T12 HAZ T13
Reservoir Pressure (bar) 95.0 85.0 77.0 83.0 77.0 69.0
Reservoir Temperature (degC) 3.0 4.0 6.0 3.0 3.0 3.5
CO2fraction 1.00 1.00 1.00 1.00 1.00 1.00
Condensed phase fraction 1.00 1.00 1.00 1.00 1.00 1.00
Orifice diameter (mm) 9.0 12.0 25.0 12.0 25.0 50.0
Observed flow rate (kgs−1) 7.7 24.0 40.0
Atmospheric conditions
Pressure (MPa) 0.1 0.1 0.1 0.1 0.1 0.1
Temperature (degC) 3.0 6.0 4.0 3.0 3.0 3.5
Relative humidity (%) >95.0>95.0>95.0>95.0>95.0>95.0
46
Table 4: Initial conditions for the HSL tests included here, reproduced from (Pursell,
2012).
Description HSL Test C HSL Test D
Average feed pressure (bar) 54.5 3.55
Nozzle pressure (bar) 46.9 36.7
Nozzle temperature (degC) 11.6 2.2
CO2fraction 1.00 1.00
Condensed phase fraction 0.86 0.84
Orifice diameter (mm) 2.0 4.0
Laboratory conditions
Assumed pressure (MPa) 0.1 0.1
Assumed temperature (degC) 20 20
47
Table 5: Initial conditions for the CO2QUEST tests included here.
Description Pres Tres Mixture Dnoz Patmos. Tatmos.
(bar) (degC) (mm) (MPa) (degC)
Pure CO2tests
QUEST T12 56 16-18 100% liquid CO26.0 0.1 18
QUEST T14 37 -5 - -2 100% liquid CO26.0 0.1 10
DUT Trial 1 53 20 100% CO2, two-phase 233.0 0.1 36
DUT Test 1 42-54 10-20 100% gas CO250.0 0.1 1.5
Impure CO2tests
QUEST T9 73 23-30 98.4% CO2, 1.6% CH46.0 0.1 26
QUEST T13 63 16 95.6% CO2, 4.4% CH46.0 0.1 18
QUEST T15 65 13 95.5% CO2, 4.5% N26.0 0.1 13
QUEST T16 57 8 95.6% CO2, 2.2% N2, 2.2% CH46.0 0.1 11
QUEST T19 51 6 95.9% CO2, 4.1% CH46.0 0.1 9
QUEST T22 56 10 95.9% CO2, 4.1% CH412.0 0.1 14
QUEST T23 63 11 95.5% CO2, 4.5% N212.0 0.1 14
QUEST T24 57 10 96.0% CO2, 2.1% N2, 1.9% CH412.0 0.1 10
QUEST T25 63 11 95.5% CO2, 4.5% N212.0 0.1 12
48
Table 6: Inlet conditions considered here.
Description: Liquid CO2tests Pure CO2Impure CO2
Based on: COOLTRANS T7 QUEST T12 adapted QUEST T12
Reference: Wareing et al. (2013b)
Reservoir conditions
Pressure (MPa) 15.00 5.70 5.70
Temperature (degC) 7.45 15.2 3.0
Mixture 100% liquid CO2100% liquid CO296% CO2, 4% N2
Atmospheric conditions
Pressure (MPa) 0.1 0.1 0.1
Temperature (degC) 7.45 18.0 18.0
Inlet conditions
Diameter (mm) 24.3 6.0 6.0
Pressure (MPa) 4.14 1.469 1.498
Temperature (degC) 6.85 -29.15 -33.15
Mean velocity (m s−1) 105.60 136.5 135.7
Liquid fraction (kg/kg) 1.00 / 0.80 / 0.60 0.684 0.693
Density (kg m−3) 883.5 / 392.1 / 252.0 111.86 112.68
Mass-flow (kg s−1) 43.3 / 19.2 / 12.3 0.432 0.432
49
