Subscriber access provided by University of Newcastle, Australia
Review
Thermodynamic modeling with equations of state: present challenges with established methods Øivind Wilhelmsen, Ailo Aasen, Geir Skaugen, Peder Aursand, Anders Austegard, Eskil Aursand, Magnus Aa. Gjennestad, Halvor Lund, Gaute Linga, and Morten Hammer Ind. Eng. Chem. Res., Just Accepted Manuscript • DOI: 10.1021/acs.iecr.7b00317 • Publication Date (Web): 10 Mar 2017 Downloaded from http://pubs.acs.org on March 14, 2017
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Industrial & Engineering Chemistry Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Industrial & Engineering Chemistry Research
Thermodynamic modeling with equations of state: present challenges with established methods Øivind Wilhelmsen, Austegard,
‡
∗,†,‡
Ailo Aasen,
Eskil Aursand,
‡
†, ‡
Geir Skaugen,
‡
Magnus Aa. Gjennestad,
Linga,
‡
and Morten Hammer
Peder Aursand,
‡
Halvor Lund,
‡ ‡
Anders Gaute
‡
†Norwegian University of Science and Technology, Department of Energy and Process Engineering, NO-7491 Trondheim, Norway ‡SINTEF Energy Research, NO-7465 Trondheim, Norway E-mail:
[email protected] Abstract
tended corresponding state EoS has a large potential for improvement. The molecular-based
Equations of state (EoS) are essential in the
SAFT family of EoS are preferred when predic-
modeling of a wide range of industrial and nat-
tive ability is important, e.g. for systems with
ural processes. Desired qualities of EoS are ac-
strongly associating uids or polymers where
curacy, consistency, computational speed, ro-
few experimental data are available. We discuss
bustness and predictive ability outside of the
some of their benets and present challenges.
domain where they have been tted.
In this
A discussion is presented on why predictive
work, we review present challenges associated
thermodynamic models for reactive mixtures
with established models, and give suggestions
such as CO2 -NH3 and CO2 -H2 O-H2 S must be
on how to overcome them in the future.
developed in close combination with phase- and
The most accurate EoS available, multipa-
reaction equilibrium theory, regardless of the
rameter EoS, have a second articial Maxwell
choice of EoS. After overcoming present chal-
loop in the two-phase region that gives prob-
lenges, a next-generation thermodynamic mod-
lems in phase-equilibrium calculations and ex-
eling framework holds the potential to improve
clude them from important applications such as
the accuracy and predictive ability in a wide
treatment of interfacial phenomena with mass
range of applications such as process optimiza-
based density functional theory.
tion, computational uid dynamics, treatment
Suggestions
are provided on how this can be improved.
of interfacial phenomena and processes with re-
Cubic EoS are among the most computation-
active mixtures.
ally ecient EoS, but they often lack sucient accuracy. We show that extended correspond-
Introduction
ing state EoS are capable of providing significantly more accurate single-phase predictions than cubic EoS with only a doubling of the
An accurate and robust thermodynamic frame-
computational time.
In comparison, the com-
work is at the base of most higher-level model-
putational time of multiparameter EoS can be
ing and simulation tools. Such a framework is
orders of magnitude larger. For mixtures in the
central for handling problems of practical rele-
two-phase region, however, the accuracy of ex-
vance such as two-phase ow of CO 2 -mixtures
ACS Paragon Plus Environment 1
1
Industrial & Engineering Chemistry Research
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
and design of heat exchangers
24
as well as
Page 2 of 25
will be capable of describing all components or
9
problems of a more academic character, like re-
uids, or be suitable for all applications.
search on the thermal state of the earth's core
process simulations and time-consuming opti-
5
In
or discussions about water cavitation at highly
mization studies, it is advantageous to use com-
negative pressures.
putationally ecient cubic EoS such as Peng
6
The typical framework has
10
11
a collection of several equations of state (EoS),
Robinson
suitable for describing the properties of uids
higher accuracy is needed, corresponding state
or materials of interest, and algorithms to pro-
methods like LeeKesler, extended correspond-
cess the information in the EoS to provide the
ing state methods like SPUNG
appropriate properties for a particular appli-
rameter EoS like GERG-2008 are taken ad-
cation.
vantage of.
For instance, modeling of multiphase
or SoaveRedlichKwong.
1320
12
When
or multipa-
Models with enhanced predic-
heat exchangers or distillation columns requires
tive ability such as Statistical Associating Fluid
phase equilibrium calculations to be performed
Theory (SAFT) are preferred for certain appli-
to obtain the relevant properties.
cations, e.g. for systems with strongly associ-
The accuracy of the thermodynamic model
ating uids or polymers where few experimen-
2123
employed can have a large impact on higher-
tal data are available.
level modeling, and even on the nal conclu-
work is needed to develop EoS for new uids,
sions made in a study.
especially polar and associating components
For instance, the ice-
24
It is clear that more
like structures called hydrates can form and ob-
such as electrolytes.
struct natural gas pipelines in the presence of
we shall discuss some of the fundamental chal-
even trace amounts of water. It is crucial for the
lenges associated with the EoS and the basic
petroleum industry to know the maximum al-
routines found in current state-of-the-art ther-
lowable water content before hydrates can form
modynamic modeling frameworks.
to specify the appropriate dehydration require-
we shall discuss how to overcome these chal-
ments. However, the predicted maximum water
lenges. By overcoming them, a next-generation
content depends strongly on the choice of EoS.
thermodynamic framework holds the potential
7
In this work however,
Moreover,
Hydrate formation is a good example of the
to improve the accuracy and predictive abil-
type of challenges a modern state-of-the-art
ity in a wide range of applications such as
thermodynamic framework should be able to
process optimization, computational uid dy-
handle.
The basis of hydrate equilibrium cal-
namics, treatment of interfacial phenomena and
culations with the van der WaalsPlatteeuw
processes with reactive mixtures such as in dis-
model is an EoS capable of describing non-
tillation columns.
polar, polar and associating components.
8
The
Add-on/utility
EoS should be easy to extend and improve when
Model-indep. routines
new experimental data are available to accommodate for new components and conditions.
Generic EoS interface
Moreover, the EoS should be thermodynam-
Model-specific routines
ically consistent, incorporate known physical constraints, and be able to predict accurately
Temporary storage
outside of the domain where experimental data are available.
Next, phase equilibrium calcu-
Database
lations must be performed. Phase equilibrium calculations can be challenging and often time-
Figure 1:
consuming. In hydrate equilibria, it is possible
structure of a typical thermodynamic frame-
to have simultaneous coexistence of vapor, solid
work
An illustration of the layer-based
and one or more liquid phases. It then becomes challenging to identify the thermodynamically most stable conguration. It is generally accepted that no single EoS
ACS Paragon Plus Environment 2
Page 3 of 25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Industrial & Engineering Chemistry Research
Structure of the thermody-
pressions, but it may also include modeloptimized iterative routines such as den-
namic framework
sity solvers.
Generic EoS interface
From a programming point of view, a modern thermodynamic framework should be modular and easy to maintain and update.
It is
tines.
important to test the accuracy and robustness
then added in.
EoS-
This layer enables the
writing of generic algorithms that can be
Moreover, it should be straight-
ignorant about the underlying model.
forward to add parameters for new EoS or new components.
They simply re-direct a prop-
independent ideal gas contributions are
gramming errors and development of inconsis-
25
wrapper
from the currently chosen model.
Automatic routines
for consistency checks reduce the risk of protent EoS.
are
erty request to the appropriate routine
of EoS to ensure a thermodynamically consistent implementation.
These
routines around the model-specic rou-
Model-independent routines
Post-processing routines to
tines
data are natural parts of a modern thermody-
to do calculations based on principles
namic framework. A natural structure of such a
which are common to all consistent EoS.
framework is illustrated in Fig. 1, which repre-
Examples include iteratively solving for
sents a layer-type structure. Starting from the
bubble/dew points or multi-phase equi-
inner layer, the dierent layers can be described
librium states, or checking for phase sta-
as follows:
bility.
EoS
interface
This layer in-
cludes, but is not limited to:
all available models and components in the framework are stored. This includes
•
model constants, component-specic pa-
Routines for result collection and plotting.
rameters for EoS models and ideal gas
•
correlations, and interaction parameters
Add-on models such as checking for possible
for pairs of components for each model. Since
generic
Add-on and utility routines
This is where the parameters for
Temporary parameter storage
the
rou-
plot phase envelopes and other thermodynamic
Database
utilize
These
formation
of
hydrates
or
other solids.
•
the
database layer can grow large, it is more
Testing routines such as consistency checks.
practical to work with a more compact
•
structure for the components and the
Framework for tting model parameters to experimental data.
models currently selected, since only a small subset of the database is likely
A thermodynamic framework can be used as
needed for a given application. This layer
part of a process simulation or as part of a
is structured to enable convenient writing
code to perform computational uid dynam-
of high performance property calculation
ics (CFD). When accurate and computationally
routines.
expensive thermodynamic models are needed,
Model-specic routines
e.g. in CFD dealing with compressible ows, These routines im-
thermodynamic calculations can become the
plement the specic thermodynamic mod-
bottleneck of the overall computation.
els.
therefore important that a modern thermody-
Based on the temporary param-
eter storage,
they calculate thermody-
namic framework lends itself easily to paral-
namic properties such as pressure, en-
lelization. Herein, one must consider both
tropy, Gibbs energy, and their derivatives, using a specic model.
It is
1. Enabling simultaneous calls to serial rou-
These routines
tines in the library without causing mem-
are mostly based on explicit analytical ex-
ory collisions between separate instances.
ACS Paragon Plus Environment 3
Industrial & Engineering Chemistry Research
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
2. Enabling routines in the library to utilize
highlight
parallelization internally to solve its prob-
Page 4 of 25
some
important
last
multiparameter EoS must overcome.
lem.
Computational speed/robustness
For the rst point,
hurdles
At present,
a modular structure is
cubic EoS are preferred if high computa-
needed with careful control of any shared mem-
tional eciency is needed. However, they
ory.
This can usually be achieved in a non-
can be inaccurate for predicting certain
intrusive manner, with few implications to the
properties such as the density and the
actual core routines. For the second point, par-
speed of sound in the liquid phase.
allelizing routines requires choosing algorithms
explore the possibility of enhancing the
that scale well in a parallel environment while
accuracy by exploiting the corresponding
minimizing overhead when run on a single pro-
state concept, and also consider how this
cessor. Here, shared memory approaches such
impacts computation time.
as OpenMP oers exibility, low code intrusive-
tions such as steady-state heat exchanger
ness and ease of implementation.
modeling, where the mixture composition is
Models
in
the
constant,
another
We
For applica-
feasible
approach
is to tabulate the values from a high-
thermody-
accuracy EoS in set of grid points (e.g. in
namic framework
the
T P -plane)
and then interpolate.
Predictive capability
Desired qualities of thermodynamic models are
29
For associating uids
accuracy, thermodynamic consistency, compu-
and polymers with few available experi-
tational speed, robustness and predictive abil-
mental data, it becomes necessary to em-
ity outside the region with the data that has
ploy EoS with enhanced predictive abili-
been used for tting the model.
ties such as SAFT and group contribution
The relative
importance of these qualities depend on the
approaches.
application.
dictive thermodynamic models for mix-
Fiscal metering of CO 2 ow has
Moreover, to arrive at pre-
extremely high accuracy demands, while com-
tures such as CO 2 -NH3 and CO2 -H2 O-
putational speed and predictive ability are ir-
H2 S, EoS must be developed in close
relevant. For integrated process optimizations
combination
where the solvent used is also optimized, predic-
equilibrium theory.
tiveness is key since data are scarce for most sol-
must explicitly incorporate the underly-
vents.
ing physical mechanisms such as chem-
26
For large-scale CFD calculations where
with
phase-
and
reaction
A predictive model
the uncertainties in the ow model dominate
ical reactions and association.
those in the thermodynamic model, and where
tively, ignoring association interactions in
ash calculations are called millions of times,
associating mixtures is similar to ignor-
computational speed and robustness take pri-
ing chemical reactions in reacting mix-
ority.
tures.
Since no single thermodynamic model
Qualita-
Although one is often able to
possesses all of these qualities, a exible ther-
t a simple cubic EoS to an associating
modynamic framework caters to these diering
(or reacting) binary system, models ex-
needs by implementing a range of thermody-
plicitly accounting for association (reac-
namic models.
The state-of-the-art modeling
tions) are needed to correctly predict the
approaches within each of the areas are as fol-
behavior of ternary or higher-order mix-
lows:
tures, due to cross eects such as several
Accuracy/consistency thermodynamic
molecules associating (reacting) with the The
models
most for
accurate
same molecule.
predicting
for some
properties of gases and liquids are multiparameter EoS. sistency
27,28
criteria
fullled,
we
We shall see that even mixtures, such as NH 3 -
CO2 , the performance of present EoS is
Although most con-
are
binary
very poor if reactions are neglected.
shall
ACS Paragon Plus Environment 4
Page 5 of 25
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Industrial & Engineering Chemistry Research Cubic EoS Equation (SRK, PR) + α correlations (Classic, Twu, Mathias-Copeman,…) + Mixing rule (Quadratic, Huron-Vidal, Wong-Sandler,…) + Volume shift
In the following, we discuss present research challenges associated with some of the main components framework:
of
a
modern
thermodynamic
thermodynamic models and rou-
tines for calculating phase and reaction equilibria.
Multiparameter EoS (EoS-CG, BWR,…)
In this discussion, we shall deal with
challenges associated with each of the desired
Association contribution
qualities elaborated above.
Corresponding states EoS
An overview of some common thermodynamic models and how they are connected is illus-
PC-SAFT
CPA
trated in Fig. 2. The gure shows that there are
Phase and chemical reaction equilibria
many choices associated with cubic EoS, such as the type of alpha correlations, mixing rules, and incorporation of so-called volume shifts.
Thermodynamic properties
30
Further, the dashed lines elucidate that cubic
Figure 2: Overview of models. Dashed arrows
EoS are used as input in many thermodynamic
represent submodels, and names in parentheses
models such as corresponding state (CSP) mod-
represent choices.
els and the cubic plus association (CPA) EoS. Multiparameter EoS and Perturbed-Chain Statistical Associating Fluid Theory (PC-SAFT),
curacy. Multiparameter EoS have been devised
however, do not rely on cubic EoS. Incorporat-
for single-component uids
ing the association contribution in PC-SAFT
The real uid properties are in multiparameter
and CPA involves choosing the number of asso-
EoS dened in terms of the following reduced
ciation sites, as the gure shows.
Helmholtz energy function:
Finally we mention that an alternative to computing
thermodynamic
EoS, is to use the so-called
properties
γ−φ
αr (δ, τ ) =
from
approach:
Nk δ dk τ tk +
k
X
Here, the vapor phase is modeled with a conventional EoS (e.g.
X
SRK), while an activity
1318
and mixtures.
20
27
X
Nk δ dk τ tk exp −δ lk +
k
Nk δ τ exp −ηk (δ − k )2 − βk (τ − γk )2 dk tk
k (1)
coecient model is used for the liquid phase. These models will not be treated here due to
dk , tk , lk , Nk , k , γk , βk , ηk δ and τ are the
their severe thermodynamic inconsistency, e.g.
where
their inability to predict the existence of a
rameters, and
vapor-liquid critical point.
sity and temperature.
are tted pareduced den-
In addition, Gaussian
terms with prefactors that diverge at the critical point, called non-analytical terms, must
Present
challenges
with
es-
be introduced to reproduce experimental data very close to the critical point.
tablished methods
gions where thermodynamic data are available; however, they have challenges that restrict their
rameter EoS
popularity.
For instance, they give substan-
tially longer computational time than simpler
The desire to represent the available experi-
EoS,
mental data in a compact and precise manner
31
in particular for phase equilibrium cal-
culations, where it is challenging to achieve ro-
has motivated the development of state-of-the-
27
Multiparam-
eter EoS have unparalleled accuracy in the re-
Multiple Maxwell loops in multipa-
art multiparameter EoS.
27
bust and time-ecient calculations.
These equations are
32
The main reason for most of their current
founded on a comprehensive analysis of exper-
drawbacks is the second articial Maxwell loop
imental data and a diligent optimization pro-
in the two-phase region.
cedure, with functional forms optimized for ac-
ACS Paragon Plus Environment 5
27
Figure 3 demon-
Industrial & Engineering Chemistry Research
20 10
;geq ;gs
P [MPa]
;leq
;ls
stable homogeneous phase within the spinodals, ρgs and ρ`s , where subscript s refers to the spinodals and superscripts g and l refer to gas
A Metastable regions ! 1
A
0 -10
@P @;
2
Peq