International Journal of Greenhouse Gas Control 72 (2018) 26–37

Contents lists available at ScienceDirect

International Journal of Greenhouse Gas Control

journal homepage: www.elsevier.com/locate/ijggc

Dynamic flowsheet simulation for chemical looping combustion of methane T

⁎
Johannes Haus , Ernst-Ulrich Hartge, Stefan Heinrich, Joachim Werther
Hamburg University of Technology, Institute of Solids Process Engineering and Particle Technology, Denickestr. 15, 21073, Hamburg, Germany

A R T I C LE I N FO A B S T R A C T

Keywords: In a Chemical Looping Combustion system, the fuel and air reactors are strongly coupled because of chemical
Flowsheet modeling reactions in both and the circulation of solid oxygen carrier between them. To capture the effects inside the
Process simulation system, a novel dynamic flowsheet simulation environment for solids processes is applied to Chemical Looping
Dynamic simulation Combustion of methane. Flowsheet simulation is a tool for process analysis and optimization covering multiple
Chemical looping combustion
process units and flows in a system.
Gaseous fuel
An experimental 25 kWth pilot plant is operated, and all of its process units are modeled. The modeling
comprises three fluidized bed reactors, two operating in bubbling fluidized bed condition and one as a circu-
lating fluidized bed riser. A cyclone is used for gas-solid separation after the air reactor. The loop seals ensure gas
sealing between the reactors. Fluid mechanics inside the systems are modeled with empirical and semi-empirical
correlations, to enable fast calculations. This approach becomes handy when long-term dynamic effects like
abrasion, start-up, or shut-down procedures as well as load changes are to be modeled.
Chemical reactions for a gaseous fuel and their implications on gas flows were implemented. In addition,
oxidation and reduction of the solid oxygen carrier in the three reactors were part of the simulation. To validate
the simulation results, the pilot plant was operated with methane as fuel. Gas measurements were taken after
both stages of the fuel reactor. Additionally, solid samples were drawn from the hot facility to examine the
oxidation state of the carrier, when fuel is introduced.
A transient simulation of plant operation over a total runtime of 40 min reveals that the solids inventories of
the fluidized bed reactors in the system need only 30 s in the present case to reach a new steady state after a load
change. If the oxidation and reduction reactions of the oxygen carrier are taken into account, however, this
response time extends dramatically to several hundreds of seconds, which can also be seen in the experimental
campaigns. The simulation of such a system behavior requires a powerful simulation tool for flowsheeting,
which has been found here in the dynamic simulation framework.

1. Introduction least two fluidized bed reactors, which are connected by cyclones and
loop seals to fulfill gas-solid separation and prevent gas leakage, re-
The first fluidized bed reactor for Chemical Looping Combustion spectively. A key element for the understanding of the process is
(CLC) operated in 2001 and was run on methane (Lyngfelt et al., 2001). modeling work conducted on the reactors’ fluid mechanics and the
Since then, various process routes have been proposed for Chemical chemical kinetics inside the system.
Looping Conversion processes. Among them, Chemical looping with Process simulation of CLC has been carried out with a variety of
Oxygen Unclouping (CLOU) and internal Gasification Chemical methods and approaches. In literature, no sharp separation of process
Looping Combustion (iG-CLC) are considered to be the most promising simulation tools exists, but one can make some basic classification
technologies. (Adanez et al., 2012). In all these processes, a solid among the time and length scales, which the simulations have taken
oxygen carrier (OC) is used to provide oxygen for the (partial) oxidation into account. One can, for example, differentiate dynamic and steady
of fuel, which can be gaseous, liquid or solid in its state. These processes state simulations. In dynamic or also called transient simulations, the
aim at providing a CO2 rich flue gas, which is suitable for the later operation parameters and, hence, the fluid mechanics in the reactors
compression and storage. Different from the regular air-fired combus- can vary over time, which is closer to reality. Steady state simulations
tion, the OC plays the role of oxidizing agent in fuel oxidation reaction, usually find one “steady state” solution for a set of operation conditions
which lead to a different process layout. To implement the CLC process, and is, therefore, less computationally expensive. Different simulation
most of the operating laboratory and pilot scale plants use a design of at environments and ideas were used to model the whole process of CLC.

⁎
Corresponding author.
E-mail address: [email protected] (J. Haus).

https://doi.org/10.1016/j.ijggc.2018.03.004
Received 17 August 2017; Received in revised form 19 February 2018; Accepted 11 March 2018
Available online 19 March 2018
1750-5836/ © 2018 Elsevier Ltd. All rights reserved.
J. Haus et al. International Journal of Greenhouse Gas Control 72 (2018) 26–37

Nomenclature cv_mf Solids concentration at minimum fluidization velocity [−]

cv_suspension Solids concentration in the dense suspension phase [−]
Abbreviation description a Decay constant [1/m]
Hb Height above distributor, where the dense bottom zone
CLC Chemical looping combustion ends [m]
CLOU Chemical looping with oxygen uncoupling h Height inside reactor [m]
iG-CLC Internal gasification chemical looping combustion Ki,∞ Entrainment constant [kg/m2/s] for each particle size
FR Fuel reactor interval i
AR Air reactor di,p Average diameter of size interval i
OC Oxygen carrier u Superficial gas velocity [m/s]
PFR Plug flow reactor ut,i Terminal velocity of particles in size interval i [m/s]
CSTR Continuously stirred tank reactor ṁOC Mass flow of oxygen carrier [kg/s]
CFD Computational fluid dynamics Xj Solid conversion of component j [−]
CFB Circulating fluidized bed Cd,l Gas concentration of gas component l in dense suspsension
TGA Thermo-gravimetric analysis phase [mol/m3]
AR Archimedes number Cd,l Gas concentration of gas component l in bubble phase
[mol/m3]
Letter/symbol description Cb,l Gas concentration of gas component l in the freeboard
[mol/m3]
Δp Pressure drop [mbar] ud Velocity in dense suspension phase [m/s]
Δmsolids Mass of solids between a certain height interval [kg] rs,j,l Solid reaction rate of solid j with gaseous component l
g Gravitational acceleration [m/s2] [mol/m3]
ρs Density solid [kg/m3] m,j Molar density [mol/m3]
ρf Density fluid [kg/m3] k Kinetic constant for certain temperature and concentra-
Ar Fluidized bed reactor cross section area [m2] tion [1/s]
dv,0 Initial bubble size [m] k0 Pre-exponential factor [1/s]
dv Bubble size [m] EA Activation energy [J/mol]
Vor Volumetric flow rate through orifice [m3/s] J̇Q,l Convective flow of gas component l [mol/m3]
ub Bubble rise velocity [m/s] KQ Convective exchange rate [1/s]
εb Bubble volume fraction [−] rg,l Reaction rate of gaseous compound l [mol/m3]
λ Bubble lifetime [s] kg Gas diffusion resistance [mol/s]
V̇b Visible bubble flow [m/s] D Molar binary diffusion coefficient [m2/s]
θ Scale dependent geometry parameters [−] at Exchange area between bubbles and suspension [1/m]
cv Solids concentration [−]

Table 1
List of process modeling approaches by various research groups, (iG = in-situ Gasification).

Process Fuel Software Fluid mechanics Reaction model Institution Researcher Dynamic

CLC gas IPSEpro 0-D equilibrium reactions Vienna University of Technology Bolhàr-Nordenkampf et al. (2009) no
CLC gas Matlab/Simulink 1-D reaction kinetics Lappeenranta University of Peltola et al. (2013), Peltola et al. (yes)
Technology (2015)
CLC gas AspenPlus 1-D reaction kinetics Heriot-Watt University Edinburgh Porrazzo et al. (2014) no
Syngas- CLC syngas AspenPlus 0-D equilibrium reactions University of Surrey Mukherjee et al. (2015) no
Syngas- CLC syngas AspenPlus 0-D equilibrium reactions Ohio State University Li et al. (2010) no
iG-CLC solid AspenPlus 0-D equilibrium reactions University of Utah Sahir et al. (2014) no
iG-CLC solid AspenPlus 1-D reaction kinetics Technical University of Darmstadt Ohlemüller et al. (2014) no
iG-CLC solid SolidSim 1-D reaction kinetics Hamburg University of Technology Kramp et al., (2012) no

Table 1 gives a short overview of flowsheeting software used for CLC Flowsheeting is a process simulation tool, which takes into account
process simulation by various research groups. the predominant macroscopic effects in a complex process network
Process simulations also vary in the depth of phenomena they take involving multiple process units. For describing these effects, semi-
into account. When looking at smallest structures of fluid mechanics empirical and empirical models are used to characterize the events in
with CFD methods or analyzing detailed chemical kinetics numerically, each and every process unit in order to keep computational time and
then simulations are carried out on micro or even molecular scale. effort within manageable limits.
When some phenomena are averaged or closures are used to describe Two main approaches of solving the model equations in a flowsheet
the micro scale behavior, one can speak of meso-scale simulations. In simulation environment exist. The first is the equation-oriented ap-
macro-scale simulations, only global mechanisms are resolved. This proach, in which all equations of the process units are concluded and
means that for certain unit operations, mass and energy balances are solved simultaneously. The other is the modular-sequential approach,
conducted, whereas in fluidized bed reactors also simplistic models for in which the process units, so called modules, are calculated in series.
describing the solids distribution in the reactors can be used. For the The big advantage of flowsheeting compared to CFD methods is the
different simulation approaches it can be said that the more phenomena short computation time, which makes simulations over a long period of
included, the more computationally expensive the simulation will be. time possible (> 1000s), provided that a dynamic approach is used.
On the other hand, resolving the micro-scale phenomena accurately With this method, effects over a longer time period can be tracked, like
gives obviously more comprehensive results. attrition, start-up and shut-down, and load changes, which cannot be

27
J. Haus et al. International Journal of Greenhouse Gas Control 72 (2018) 26–37

resolved within CFD. approach, which means that all process units are calculated in a row. In
The used approaches deviate considerably in the detail in which CLC, the system usually comprises two interconnected fluidized bed
certain phenomena were resolved. 1-D models consider basic fluid reactors, which are linked via cyclones and loop seals. All of these unit
mechanics of solids and gases inside the fluidized bed reactors. Usually operations were implemented into the calculation system in a way to
one can find a dense zone at the bottom of the reactor, followed by a allow for dynamic changes of the involved solids. This means that the
steady decline of solids concentration with increasing height above the mass holdup inside each unit can change over time, and, depending on
distributor. In two-phase models, the dense zone is further divided into the operation procedure, the solid flux between all process units may
a bubble and a suspension phase. 0-D models do not resolve any of vary as well.
these basic phenomena commonly seen in fluidized bed reactors, but try The main aim of the present work is to show the interdependency of
to mimic the outcome via a series of plug flow reactors (PFR) or con- the fluidized bed reactor system, which have been modeled with the
tinuously stirred tank reactors (CSTR). Recently, in AspenPlus a simple possibility to change their respective mass inventory and a dynamically
fluidized bed reactor module was implemented, but it was concluded adapting mass flow between them. Fluid mechanics were also described
that the model did not describe the hydrodynamics of a circulating for the loop seals, which have the possibility to sustain and release
fluidized bed (CFB) boiler sufficiently (Yörük et al., 2015). excess bed material depending on the overall process situation. The
From the chemical reaction point of view, most researchers from system’s dynamic response to the injection of fuel and the stop of in-
Table 1 used a chemical equilibrium model, which minimizes the Gibbs jection was investigated. The simulated combustion behavior of the
free energy, without taking into account stoichiometry and reaction system was compared to an experiment on a 25 kWth pilot plant. The
rates. Two groups used their own code in Matlab/Simulink and So- pilot plant was operated with methane as fuel and copper oxide on
lidSim, respectively, to implement gasification and char combustion alumina as an OC. Solid samples were taken at certain time periods
kinetics (Peltola et al., 2013; Kramp et al., 2012). Another group de- after a load change to capture the dynamic change of conversion of the
veloped a fluidized bed reactor model and covered these latter reaction OC. The gas concentrations measured in the pilot plant’s off-gas were
schemes by implementing their model in AspenPlus. (Ohlemüller et al., used to fit the kinetic data in the simulations. The conversion of the OC
2014). from the transient simulation was compared to the experimentally
All proposed process models concluded in Table 1 contain no time found ones of the samples.
dependency, so they yield steady state results. One group’s work in-
cluded time dependent variables, but the system itself was used to
model steady state operation (Peltola et al., 2013). Apart from these 2. Methods
simulations covering the whole process, a lot of attention was put to
fuel reactor (FR) modeling, as the FR is crucial for the conversion of the In the present work, the Dyssol software is used to generate a
solid fuel and, hence, the overall efficiency of capturing carbon dioxide. flowsheet of a CLC system. Here, modeling of the hydrodynamics in the
Since the FR is the crucial part of the process, many researchers have system and the conversion of the oxygen carrier by fuel gases is de-
focused solely on the local fluid mechanics and reaction mechanisms in scribed. In the experimental section, a 25 kWth pilot plant is described,
the FR (Abad et al., 2013; Coppola et al., 2015). which is used to gather data on the combustion behavior and compare
In this work, a novel simulation environment “DYnamic Simulation experimental and simulation results. The experimental data is the base
of SOLids processes” (Dyssol) is used to model the entire CLC process for the modeling work with the flowsheeting software.
(Skorych et al., 2017). The basis of this software is a sequential modular The concept of dynamic flowsheet modeling for the solid process of
fluidized bed spray granulation has been described and showed good

Fig. 1. Left: Schematics and right: flowsheet of the pilot plant with 25 kWth rated power as operated at Hamburg University of Technology. (S1/2 = siphon loop seal 1/2; FR = fuel
reactor; SP1,2,3 = standpipe 1,2,3; AR = air reactor).

28
J. Haus et al. International Journal of Greenhouse Gas Control 72 (2018) 26–37

results compared to experimental data (Dosta et al., 2010). Based on the concentration is detected with a paramagnetic cell in the FR exhaust
need of modeling tools for solids processes, the German Research gas. The O2 and CO2 concentration in the AR exhaust gas are measured
Foundation (DFG) started a Priority Program, called SPP 1679, “Dy- by an electro-chemical sensor and a NDIR sensor, respectively.
namic Simulation of Interconnected Solids Processes” in 2013. The core The OC is oxidized in the AR, which is operated as a CFB riser re-
of this project is the development of dynamic process models for a actor at 4–6 m/s superficial velocity. The freshly oxidized OC is sepa-
dynamic flowsheet simulation software called Dyssol. This software was rated from the depleted AR off-gas in the cyclone. Afterwards, it moves
used here to model the process of CLC. Dyssol integrates all parts that into standpipe SP1 and the loop seal siphon S1 towards the upper stage
are developed in the course of the priority program (Skorych et al., of the FR. The two stages of the FR are operated at bubbling conditions
2017). The software uses a sequential-modular approach, meaning that with a gas velocity of uR = 0.2–0.3 m/s. The two-stage system was
all process units are calculated one after the other. demonstrated to be proficient in converting volatiles and gasifying the
For the simulation of the whole process, modules have to be im- fuel char. It is possible to extract sample gas at two different positions of
plemented into the core system. The modules comprise of the necessary the fuel reactor, the first is in the freeboard of the first stage FR1 and the
empirical and semi-empirical correlations, which are used to describe second position is at the outlet pipe of the second stage FR2. Via the two
the motion and conversion of solids and gases. For the CLC process, standpipes SP2 and SP3 the OC is transported to loop seal siphon S2 and
models for the fluidized bed reactors, the loop seals and the cyclone had circulated back to the AR.
to be implemented. The calculation of the differential equations within The pilot plant possesses two solid sampling ports located at the
each unit, as well as the handling of material and energy streams, is upper (S1) and lower loop seal (S2). Through these ports bed material,
done by the software, Dyssol. i.e. the OC, can be extracted during hot operation. The bed material is
sucked via a simplistic vacuum mechanism into a water-cooled con-
2.1. Experimental tainer, where it is cooled down. To prevent further reaction of the bed
material outside of the plant, the container is constantly flushed with
A 25 kWth pilot plant has been operated with an oxygen carrier nitrogen until temperatures below 100 °C are reached. The extracted OC
consisting of CuO as the metallic compound and Al2O3 as the basis. The is then analyzed for its oxygen content by oxidizing a sample mass in a
carrier was produced by impregnating γ-alumina (Puralox NWa-155 by thermogravimetric analyzer.
Sasol) with a solution of copper nitrate. It was aimed for 14 w.-% of
active material, but elementary analysis and TGA data showed that the 2.2. The process flowsheet
active component on the outer surface of the carrier was quickly
abraded, and the active material in the pores of the particle was then A flowsheet of the process was created and included in Fig. 1. In this
about 9 w.-% when operated over a long period in the pilot plant. The flowsheet, all units, i.e. the air reactor (AR), the two FR stages, the loop
concept schematics of the pilot plant, with the two-stage FR system, is seals S1 and S2, and the cyclone are connected via streams. Inside these
shown in Fig. 1. The dimensions of the plant are summarized in Table 2. streams, all information about mass flow, temperature, pressure, as well
The pressure situation is tracked by 32 pressure transducers. Through as the gaseous and solid compounds are shared between the respective
that, the pressure distribution, the fluid mechanic situation and the units. As shown in Fig. 1, the solids from the AR move upwards with the
solid mass distribution can be calculated. The pressure drop Δp between fluidization air (stream 15), and enter the cyclone in stream 1. The
two measurement ports in the FR or in the AR allows an estimation of cyclone separates most of the solids from the gas, and only fine particles
the corresponding solids mass Δmsolids from a force balance: exit with stream 2. The underflow of the cyclone enters the loop seal,
Δm solids⋅g ρs − ρf siphon 1, via stream 3. S1 is fluidized with steam and typically forms
Δp= ⋅ two fluidized chambers. Entering the FR2, stream 4 contains steam
Ar ρs (1)
from fluidization and solids which are carried over. FR 2 is fluidized by
Usually, solid fuels are injected into the fuel reactor in the lower stage the off-gas stream 7 of FR1. Elutriated fines and product gas from the
(FR1), and the gases produced by devolatilization and gasification rise reaction of oxygen carrier and gaseous fuel leave the system via stream
up and react in the upper stage (FR2). Injection from upper stage is also 6. The overflow of FR 2 enters FR 1 via a standpipe, which is expressed
feasible, which provide a longer residence time for char. So far, bitu- by stream 8. FR 1, itself, is fluidized by a mixture of fluidization gas and
minous coal and lignite were used in the plant and experimental data fuel, which in the present work is methane. Coal combustion would be
reported (Haus et al., 2016; Thon et al., 2014). In the present work, the modeled via a separated stream entering the FR stage 1 at a certain
experimental results from a methane CLC experiment are presented and height above the distributor. Parts of the fluidization gas of S2 rise up
used to validate the modeling results. As Fig. 1 shows, methane, to- via the standpipe and contribute to fluidization of FR1. Solids and gases
gether with CO2 and H2O, is injected into FR from the bottom as the from S2 recirculate back to the AR. The whole flowsheet is calculated
fluidized gas. iteratively to solve the whole loop including multiple backflows in the
The volumetric gas concentrations in the FR are measured with non- unit.
dispersive infrared sensors (NDIR) for CH4, CO2 and CO. Hydrogen is For the flowsheet simulation, unit modules, which contain the
measured with a thermal conductivity sensor. The oxygen process models and further correlations for the chemical conversion,

Table 2
Dimensions of the CLC pilot plant as operated.

Dimensions height [m] diameter [m]

air reactor AR 8 0.1

fuel reactor FR 4 0.25
bed height FR 0.6 per stage
cyclone 0.34 0.21

Dimensions width [m] length [m]

siphon S1 0.16 0.13

siphon S2 0.13 0.25

29
J. Haus et al. International Journal of Greenhouse Gas Control 72 (2018) 26–37

had to be developed and introduced into the flowsheet software. The ub = V̇b + 0.71⋅θ⋅ g*dv (4)
used models, fluidized bed reactor, cyclone and loop seal are described
in the next section. With these unit modules inside the process flow- In this equation, V̇b is the visible bubble volumetric flow and θ denotes
sheet, the whole process and the chemical reactions in the reactors can a scale dependent geometry parameter. Within the dense bottom zone,
be calculated. The OC’s reduction with fuel gas CH4 and its decom- solids can only be found in the suspension phase, as the bubbles are
position products CO and H2 takes places in FR 1 and FR 2. Further, in considered to be solid free. The concentration of solids in the suspen-
the AR, the oxidation of OC with O2 from the fluidization air is included sion phase cv,suspension is assumed to be close to the concentration at
in the simulation. minimum fluidization cv,mf. At the border between dense bottom and
dilute upper zone, at height Hb, the concentration cv(Hb) is calculated:

2.3. Fluidized bed reactor module c v (Hb) = (1 − ε b (Hb))⋅c v,suspension forh < Hb (5)
Above the dense phase, a correlation, which assumes an exponential
2.3.1. Fluid mechanics
decay of solids concentration is used (Kunii and Levenspiel, 1991). This
In this work, empirical 1-D models are applied to describe the si-
correlation assumes an exponential decay of the concentration towards
tuation within the units. Oxidation of the OC in the AR is generally fast.
the reactor top. Furthermore, an elutriation effect that differs for each
Complete conversion is usually achieved; therefore, a simple modeling
particle size class i is taken into consideration:
approach is sufficient here.
For all three fluidized bed reactors a single module was developed, c v,i (h) = c v,i, ∞ + (c v,i (Hb) − c v,i, ∞)⋅e−a(h−Hb) forh ≥ Hb (6)
which considers the solid free bubbles and a solid dense suspension
phase (Werther and Hartge, 2004). In the module, the reactor is de- There, cv,i,∞ describes the solids concentration above the transport
scribed by having two zones: a solid rich dense bottom zone and an disengagement height and a is the exponential decay parameter. The
upper region with a lower solids concentration, which is exponentially concentration cv,i,∞ can be calculated from different elutriation rate
shrinking towards the top height of the reactor. Inside the bottom zone, correlations, and a variety of researchers have investigated them in a
the presence of bubbles is assumed for all fluidization velocities. Own broad spectrum of operating regimes of fluidized bed reactors (Chew
research has shown that, also for circulating fluidized bed reactors, the et al., 2015). For fluidization velocities above 3 m/s, i.e. in the AR, a
presence of a dense bottom zone can be detected (Schlichthaerle and correlation by Choi et al. shown in Eq. (7) has been used (Choi et al.,
Werther, 1999). The correlations used for the description of the fluid 2001):
mechanics are shortly summarized below. More details can be found in Ki, ∞⋅di,p 13.1
earlier publications (Werther and Wein, 1994). The initial bubble size = Ar 0.5⋅exp(6.92 − 2.11⋅Fg 0.303 − )
μ Fd 0.902 (7)
in the dense bed is given by (Davidson and Harrison, 1963):
Fg and Fd describe the gravitational and the drag force on the particles,
2 0.2
dv,0
V
= 1.3*⎛⎜ or ⎞⎟ respectively. Elutriation in a fluidized bed at lower velocities at bub-
⎝ g ⎠ (2) bling conditions in the FR are described with the following correlation
(Tasirin and Geldart, 1998):
where Vor denotes the volumetric flow through a single orifice. With
increasing height above the distributor, the bubble diameter dv is de- ut,i
Ki, ∞ = 14.5⋅ρg⋅u2,5⋅exp ⎛−5.4⋅ ⎞
scribed by (Werther and Wein, 1994): ⎝ u ⎠ (8)

d(dv ) 2ε 1/3 dv Here, ut,i denotes the terminal velocity of each respective particle size
= ⎛ b⎞ − class i. With the elutriation rate constant for each particle interval i and
dh ⎝ 9π ⎠ 3λub (3)
the mass fraction it, one can calculate the particle size distribution at
with εb being the bubble volume fraction and λ the average bubble the outlet situated at the top of the reactor. The bed mass inside the
lifetime. The formed bubbles rise with the velocity ub: fluidized bed reactor was implemented in a way to allow for changes

Fig. 2. Left: typical solids distribution inside a fluidized bed reactor. Right: input and output flows inside a fluidized bed reactor, as well as the flows and distributions inside.

30
J. Haus et al. International Journal of Greenhouse Gas Control 72 (2018) 26–37

over time. For each time step, the dense bottom zone and dilute upper on and inside the particles. Further, ρm,j is the molar density of the
dX
zone were calculated as described above to match the total bed mass in solids reactant j and s,j,l is solids conversion rate of reactant j with
dt
the respective fluid bed reactor according to: respect to fuel gas l. The fuel gases included in the simulations are CH4,
CO and H2. The reaction rate for these gases rg,j,l can be determined by
mfluidized bed = Ar⋅ρsolid (∫
0
Hb
∫H
c v dh+
hmax

b )
c v dh
(9)
multiplying the solid conversion rate rs,j,l with the stochiometric factor
stemming from Eqs. (11)–(14).
dX
In this equation, hmax is the total height of the reactor and mfluidizedbed is The solids conversion rate s,j,l is usually determined by experi-
dt
its total inventory. For this calculation, the reactor is discretized into a mentally exposing the oxygen carrier with the active compound j to
defined number of height elements. The fluidized bed reactor module different reacting gases l at various concentrations and temperatures.
allows for two reactor exits: one at the top of the reactor and another at From this data, Arrhenius type reaction rates can be derived. Several
an overflow weir or a standpipe. For each time step, a new bed mass is heterogeneous reaction models were proposed to capture the solids
calculated: conversion, because OCs of different metallic compounds differ in their
structure and, hence, also in their reaction behavior. Abad et al. con-
dmfluidized bed
= ṁ solids,inlet − ṁ elutriation,freeboard − ṁ overflow cluded that a shrinking-core-model, which takes the existence of re-
dt (10)
active grains into consideration, is well suited to describe the perfor-
In a bubbling fluidized bed reactor, solid overflow over the weir is the mance of a CuO/Al2O3 OC (Abad et al., 2010). The general form of
predominant solid discharge and only little elutriation takes place. The reaction kinetics that the Arrhenius models are based on is shown:
number of elutriated particles at the top of the reactor is dependent on
dX s,j,l
the solids concentration there and the gas velocity, whereas the mass = k(Cl n,T)⋅f(X)
dt (16)
flow over the weir depends on the height of the dense bubbling bed
above the weir. When simulating a circulating fluidized bed riser re- There the apparent rate of solid conversion k is a function of the molar
actor, the mass ṁ overflow is set 0 and all the particles exiting the reactor concentration of the reacting gas Cl, the order of reaction n and the
are stemming from elutriation. temperature T. Assuming that the concentration Cl is constant
throughout the course of the reaction in a discretized volume element
2.3.2. Reaction kinetics modeling Ar ∙ dh, the kinetic constant k can be described by an Arrhenius type
In the fluid mechanics section above, it was mentioned that in the equation:
bottom zone, two phases are considered, a solid free bubble phase and a −EA
k= k oe RT (17)
gas-solid suspension phase. Fig. 2 illustrates how this affects the reac-
tions and how they were modeled in general. In the Arrhenius equation, k0 is a pre-exponential factor and Ea is
Heterogeneous gas-solid reactions are assumed to take place pre- the activation energy for the reaction, both are determined experi-
dominantly inside the suspension phase, and only gas phase reactions mentally. The rate of conversion is also a function of the conversion X
can occur inside the bubbles. In the dilute upper region of the reactor, itself. This is often expressed by an algebraic approach. A comprehen-
the solids and gases are assumed to form a dilute suspension. The gases sive summary of the kinetic models and their respective algebraic
are assumed to rise up from the distributor, and back mixing does not equations can be found in (Kruggel-Emden et al., 2011). In the present
occur. Hence, the gas conversion of each species is modeled as in a plug work an algebraic expression for spherical grains and a reaction lim-
flow reactor. The solids are considered to be perfectly mixed inside the itation is used (Adanez et al., 2012):
whole fluidized bed reactor.
f(X) = 1 − (1−X)1/3 (18)
The oxygen carrier used at TUHH is very similar to the one in-
vestigated by the researchers at CSIC Zaragoza using Puralox NWa-155 The molar balances in the discretized volume elements Ar ∙ dh for the
as the base material and impregnation as the production method. gas species are shown in Fig. 3. One can see that in the present case no
Therefore, in this work, at first the reaction kinetics described in lit- homogeneous reaction takes place in the bubble phase. Changes in the
erature, which were determined from TGA experiments are used to molar concentrations inside the bubble phase are solely due to the
describe the gas conversion (Abad et al., 2010; García-Labiano et al.,
2004). In a second simulation, the reaction kinetics were varied in a
way to fit to the measured gas concentrations at the outlet of both FR
stages. The reaction module is designed to allow for different solid re-
action models as well as other reaction rates.
The following reactions were considered for the reduction of the
oxygen carrier in the FR (Abad et al., 2010):

CH4 + CuO − > CO + 2H2 + Cu (11)

CO + CuO − > CO2 + Cu (12)

H2 + CuO − > H2O + Cu (13)

The oxidation in the AR follows:

Cu + ½ O2 − > CuO (14)

The reaction of the solid oxygen carrier j with its respective gas com-
ponent l is described by the reaction rate rs,j,l:
dX s,j,l
rs,j,l = −c v,j⋅ρm,j⋅
dt (15)

In this equation, cv,j denotes the volumetric concentration of solid re-

actant j in the reactor volume element. It is taken by multiplying the Fig. 3. Molar balances solved for each discretized height element dh. The suspension and
the bubble phase are connected via diffusive and convective mass transfer.
total solids concentrations cv and the fraction of active reactant, CuO,

31
J. Haus et al. International Journal of Greenhouse Gas Control 72 (2018) 26–37

diffusive and convective mass transfer. The heterogeneous reactions in

the dense suspension phase will induce a concentration gradient and,
with it, diffusive and convection mass transfer.
The concentration changes of all gas species l at a reactor volume
element Ar ∙ dh are given for the bubble phase b, the suspension phase d
and the freeboard f:
du
dCb,l J̇Q,l − k G·a·(Cb,l − Cd,l) − Cb,l dh
=
dh u−umf (1 − ε b) (19)

dCd,l J̇Q,l + k G·a·(Cb,l − Cd,l) + rg,l

=
dh umf (1 − ε b) (20)
du
dCf,l rg,l − Cf,l dh
=
dh u (21)
In these equations, rg,l denotes the reaction rate of the gas l, induced by
the solid oxygen carrier and based on the reactor volume.
Homogeneous gas–gas reactions could also be included here, but the
chemical reaction scheme described above does not include these re-
actions. Therefore, only heterogeneous gas-solid reactions will have an
effect on the gas concentrations. The chemical reactions inside the re-
actors have an effect on the total molar flow of gases and, with it, the
gas velocity will be changed. Since the velocity inside the suspension
phase is assumed to be constant, produced or consumed gases have to
be accounted for by a convective flow J̇Ql between suspension and
bubble phase. If the gas is consumed in the suspension phase, the flow
J̇Ql is directed from the bubble phase to the suspension phase. If gas is
produced in the suspension phase, a convective flow from the bubbles is
assumed.
K q⋅Cd,l for K q > 0
J̇Q,l = ⎧
⎩ K q⋅Cb,l for K q < 0
⎨ (22)
The convective exchange rate KQ is calculated from all heterogeneous
reactions rates rg,l based on gases in the suspension phase:
R⋅T
KQ =
P
∑ rg,l
(23)
In this equation, R denotes the universal gas constant, T is the tem- Fig. 4. Calculation procedure for the fluidized bed reactor module.
du
perature and p the pressure inside the system. The term Cl dh in the
molar balance equations describes the change of the mass flow of
species l with height due to gas velocity changes. In general, the gas
flow in the fuel reactor is increasing because either a solid fuel is 2.3.3. Calculation procedure
transferred to the gaseous form, or the gaseous fuel CH4 is reacting, In Fig. 4, the steps to calculate the reactor module are shown. At the
creating two H2 molecules and one CO from one molecule of CH4. Both start time, the fixed parameters are read into the program. These
mechanisms will increase the total flow of gases in the FR. In the AR the parameters are not changed during the course of the simulation and,
oxidation of the OC will decrease the molar gas flow. It is assumed that therefore, are only initialized once. These include geometrical values as
the gas passes through the suspension phase close to minimum fluidi- well as the decision which correlations i.e. for the elutriation of parti-
zation velocity. This is the reason why only the gas concentrations cles, are used. Afterwards, the reactor is discretized into a number of
change in the bubble phase and freeboard due to velocity effects. The predefined height elements along the reactor height, which are neces-
chemical reaction in the suspension will lead to differing concentrations sary to provide enough intermediate steps for the calculation of fluid
inside the bubble and suspension phase. Diffusive mass transfer be- mechanics and chemical reactions.
tween the phases is described by the following correlation (Sit and Then, for each time step, the operation conditions, such as fluidi-
Grace, 1981) for the gas diffusion resistance kg: zation velocity and inflows from other process units, are read from the
system. Together with the current state of the reactor itself, i.e. its ac-
umf 4·D·εmf ·ub tual bed mass and its conversion state, the solids distribution in the
kG = +
3 π⋅dv (24) system can be calculated from a fluid mechanic perspective. Knowing
the location of the solids, the chemical reaction scheme can calculate
In this correlation umf describes the minimum fluidization velocity, D is
the conversion of gases for each height element.
the diffusion coefficient, εmf stands for the minimum fluidization voi-
Fluid mechanics and chemical reactions are iterated as long as gas
dage, ub is the bubble rise velocity and dv is defined as the bubble size.
conversion in plug flow and conversion of perfectly mixed solids match
The parameter a describes the ratio of the interfacial area between
each other. Gas conversion is calculated by an explicit Euler algorithm.
bubble and suspension phase to the volume of the reactor element
This means that starting from the bottom of the reactor, the gas con-
under consideration and is calculated with the assumption of spherical
centrations in each discretized height element is calculated from the gas
bubbles from:
concentrations and the derived kinetics in the previous height element.
6ε b The fluid mechanics are calculated with the Sundials DAE solver. When
a=
dv (25) both are iterated and the calculated values converge to a solution

32
J. Haus et al. International Journal of Greenhouse Gas Control 72 (2018) 26–37

within the predefined tolerances, the outlet flow is calculated along methane as fuel. Due to the high concentration of methane in the off-
with a new bed mass. The conversion state and particle size distribution gas, only 15 kWth input of methane was injected to enable stable and
of the solids is taken. The simulation environment Dyssol determines safe operation, even though the plant has a rated power of 25 kWth,
the time step between two timepoints automatically in a way to keep which can be achieved with solid fuels. The methane supply was
the numerical error in the predefined boundaries. switched on and off several times to investigate the systems dynamic
response to this hot start and hot shut-down. Solid samples were
2.4. Loop seal modeling withdrawn during the hot operation at the lower sampling port shown
in Fig. 1. Solid samples were taken after 120, 240, 480 and 720 s after
For the dynamic simulation of the CLC reactor system, the modeling fuel injection was started to get information about the time dependent
of the loop seals is important, as it will act as a hold-up for particulate behavior of the OC mass in the reactor. Additionally, after the plant was
material. The mass, mLS, which is stored in the loop seal, is strongly at steady state operation with fuel, the fuel injection was stopped and
depending on the pressure environment around them and, with it, from solid samples were taken 2, 4 and 8 min after the stop.
the FR and AR. Fig. 5 shows the basic geometry of a loop seal as it is The OC samples were analyzed for their current oxygen content
used in the experimental plant and, therefore, is the geometry that was compared to the maximum amount that the carrier can chemically
modeled into the Dyssol environment. Solids enter the siphon in the bind. A thermogravimetric analyzer (TGA) was used for oxidizing the
supply chamber (SC) pass through the slit, and pile up in the recycle samples.
chamber (RC). In the end, they pass over the weir and return back to the The simulations were carried out to resemble the actual experi-
respective fluidized bed reactor. A loop seal is modeled including the mental performance as close as possible. Kinetic data from TGA ex-
standpipe such that the solids in the standpipe are included in the total periments (Abad et al., 2010) was used to calculate the chemical re-
mass of the SC. The total mass mLS inside the loop seal at any time point actions and conversion in a first simulation run. Afterwards, in another
is the sum of masses in the recycle chamber mRC and the supply simulation run, fitted kinetics were used to match experimental results.
chamber mSC: With the fitted kinetics, the chemical conversion of the OC and the bed
mLS = mSC + mRC (26) mass distribution inside the plant has been investigated. The simulation
began with a fully oxidized OC, what resembles a hot start of a plant,
A loop seal similar to the one operated in Hamburg was investigated by and was conducted over 2400 s. Parameters used during hot operation
Basu and Cheng, which describes the movement of solids through the and simulation are summarized in Table 3.
loop seal (Basu and Cheng, 2000). In their approach, which was in- From the starting point with the bed masses from Table 3, the
tegrated into the present flowsheeting environment, the solid masses system was operated without fuel injection for 100 s. The OC in all units
distribute according to a pressure balance. In contrast to their findings, was started in an entirely oxidized condition. After 100 s the fuel was
it was visible in Hamburg’s cold facility that depending on the pressure ramped from 0 up to 15 kWth power input and the influence on con-
distribution and gas velocities, also a fluidized bed can form in both the version and fluid mechanics was investigated. After 1400 s, the fuel
SC and the RC and not only in the RC. Depending on the gas velocity in supply was stopped.
the SC, the fluid mechanic models for a fluidized bed from the section
above or for a moving bed from Basu and Cheng are applied.
3. Results
2.5. Cyclone modeling
The dynamic flowsheet simulation of all units was conducted for
For gas-solid separation, cyclones are usually used in CLC systems. 2400 s of plant operation. A desktop computer (single core of Intel®
The cyclone in this system is implemented similar to the widely used Core™ i7-4790K) was able to calculate this runtime in less than 1000 s.
Muschelknautz model (Trefz and Muschelknautz, 1993). This model
was validated against data from a variety of hot units and is the stan-
dard method for cyclone design in Germany. The Muschelknautz was
adapted in a way to avoid errors in the population balances for fine
particles (Klett et al., 2005).

2.6. Tracking of secondary attributes of the solid material

For the modeling of CLC two secondary attributes of the particulate

material are taken into account. The first one is its particle size dis-
tribution and the second one is the oxidation state of the OC. Both
values vary over the course of the simulation and have to be tracked
throughout the process flowsheet. To ensure that these attributes are
transported and also manipulated correctly, Dyssol works with trans-
formation matrices. They describe the movement of particles between
so-called attribute intervals for each secondary attribute. Skorych et al.
have described the core system Dyssol and how transformation matrices
transport information between the different process units (Skorych
et al., 2017). To picture the CLC process the oxidation state was de-
scribed in 10 intervals for calculation speed reasons. When the solid
conversion was calculated according to the steps above, solid particles
move from one oxidation class to the next lower one for OC reduction or
the next higher one for OC oxidation.

2.7. Experimental procedure and simulation parameters

Fig. 5. Schematics of the used loop seal model.

The experimental run in the pilot plant was carried out with

33
J. Haus et al. International Journal of Greenhouse Gas Control 72 (2018) 26–37

Table 3 Table 4
Operation parameters in the hot facility and used in the simulation. Velocity data are Measured average volumetric gas concentrations during operation with 15 kWth methane.
given for the locations immediately above the gas distributors, i.e. without consideration Additionally, the inlet concentration is given. Missing percentages are due to leakage
of reaction influences on the gas flows. flows of N2 and O2 from the AR.

parameter unit experimental simulation dry gas concentration unit CH4exp CO2exp H2exp COexp

velocity AR [m/s] 4.3 4.3 inlet FR1 [vol.%] 60 40 0 0

velocity FR [m/s] 0.2 0.2 after lower stage FR1 [vol.%] 32 64 1.1 0.5
velocity S1 [m/s] 0.15 0.15 after upper stage FR2 [vol.%] 4.7 88 0.8 0.4
velocity S2 [m/s] 0.15 0.15
gas AR type air air
gas FR type steam/CO2 steam/CO2
temperature [K] 1073 1073
fuel input [kWth] 15 (methane) 15 (methane)
fuel input [kg/h] 1.1 1.1
mass FR1 [kg] ∼24 24
mass FR2 kg ∼24 24
mass AR [kg] ∼5 5
mass S1 [kg] ∼4 4
mass S2 [kg] ∼8 8

3.1. Gas composition and situation in the FR stages

3.1.1. Experiment
In the experimental campaign with methane, the volumetric gas
concentration after FR stages 1 and 2 were measured. There it was
found that the methane conversion was not complete after the second
stage. This is shown in Fig. 6, where the dry gas concentrations after the
upper stage FR2 and lower stage FR1 are shown. Averaged values for
the concentration measurements are provided in Table 4. The dry inlet
concentrations of carbon dioxide (CO2) and methane (CH4) were ap-
proximately 40 vol.% and 60 vol.% respectively. The CH4 concentra-
tion drops to around 35 vol.% after the lower stage and is further
converted to an outlet concentration of 4.5 vol.% in the second stage.
Carbon monoxide (CO) and hydrogen H2 are detected at the outlet of
the reactor system below 1 vol.%. Their concentrations are slightly
higher after the first stage. The low measured concentrations of H2 and
CO from the CH4 conversion indicate that they are converted fast to-
wards H2O and CO2. This may due to the fact that the CH4 decom-
position can only take place when OC is present and the OC can convert Fig. 7. Simulation of methane conversion inside the FR system with literature kinetics.
the CO and H2 further then due to its vicinity. Nevertheless, the pre- Shown concentrations in the lower dense zone are averaged over the bubble and sus-
pension phase. Comparison to averaged gas concentration measurements at the exit of
sence of H2 and CO show that the methane is not directly reacting to-
each stage.
wards CO2 and H2O, but via CO and H2. This reaction path was pro-
posed by other researchers (Abad et al., 2010) and is confirmed here.
in Fig. 7, the lower stage of the FR converts the methane to around
3.1.2. Simulation 7 vol.%, but does not suffice to convert much of the hydrogen and
Two transient flowsheet simulation runs were carried out for the carbon monoxide from the decomposition. Entering the second stage of
methane conversion. In the first simulation run, literature kinetics ob- the FR, the gases are well mixed into a bed of fresh, i.e. highly oxidized
tained by TGA measurements were used (Abad et al., 2010). As shown OC. In this second stage CO and H2 were converted further to exit levels

Fig. 6. Measurements of the dry gas volumetric concentration inside the CLC system. Methane as fuel was used with an inlet concentration of 60% and carbon dioxide was at 40 vol.% to
enable stable fluidization.

34
J. Haus et al. International Journal of Greenhouse Gas Control 72 (2018) 26–37

of 35 vol.−% CO and 20 vol.−% H2. The region directly above the all bed masses in the system plotted in Fig. 10 arrange themselves in a
dense bottom bed with a relatively high solids concentration leads to a stable operation point within 10–20 s. When fuel injection starts at
jump in overall gas conversion. The sharp change from dense bed to 100 s, bed material moves from both FR stages towards the AR. This is
freeboard will in practice be replaced by a transition zone, i.e. a splash again due to the fact that methane conversion produces a higher gas
zone (Zhang et al., 1995), where the bubbles burst. In the dense bed volume flow, which increases the gas velocity inside both reactor
itself, the methane conversion is comparatively low due to mass stages. Stopping the fuel injection has the directly opposite effect and
transfer limitations by the formation of bubbles. the bed masses arrange themselves basically the same way before fuel
Compared to the simulations with the kinetics by Abad et al., the was inserted.
conversion of the methane is extraordinarily fast at the border between Comparison of experiments and simulations shows that simulations
dense bed and freeboard, because from a calculation perspective, the are able to predict the response of the reactor system with regard to the
gases and solids are perfectly mixed when they enter the freeboard bed mass distribution. In the simulations, the AR increases its bed mass
(Abad et al., 2010). Looking at the gas measurements during the ex- by around 20%, when fuel is added. The FR lose around 5% of their bed
periments, one can see that the methane conversion in the simulation is mass, which is in the range of the experimental results.
overestimated in both stages. For carbon monoxide and hydrogen, it is
the other way around and the conversion found during the experiment 3.3. Conversion of the oxygen carrier
of the two gases is much larger than the simulations show.
The reason for the deviations of the simulation from the experiment The injection of methane converts the OC over a certain period of
are not clear. It could be that the structure of the OC in the present pilot time starting from a state of total oxidation Xs = 0, which is plotted in
scale experiments is different to the one used in the TGA. Another Fig. 11. The solid circulation rate Gs was determined in previous ex-
reason could be the different heat and mass transfer conditions found in periments for the reactor system and was found to be within 15 kg2 and
ms
the TGA and the fluidized bed reactor.
45 kg2 . The results of the chemical conversion of the OC are very sen-
Therefore, the kinetics for the methane conversion by CO and H2 ms
sitive to the solid circulation, because the oxygen carrier leaves the AR
from literature were varied to match the experimental results better.
nearly fully oxidized. The more fully oxidized material enters the FR
This was done by multiplying the kinetic parameter k in Eq. (17) with a
system, the lower is the overall OC conversion in it. The simulation
certain factor for each gas species separately. In Fig. 8 it is shown that
environment predicts a solid circulation of 25 kg2 with the settings from
the exit concentrations can be reproduced accurately, when the kinetic ms
parameter for methane conversion is cut in half and the parameters for Table 3.
the H2 and CO are multiplied by the factor 3 and 5, respectively. It has With the help of the dynamic flowsheet simulation the solid con-
to be stated here that the fitting parameters are the same for both stages version can be tracked and compared to the OC samples taken after
of the FR system. A more detailed investigation of the reaction kinetics 120, 240, 480 and 720 s of fuel injection, which is shown in Fig. 11. Due
of the OC used in the pilot plant is currently under way in the author’s to the big solid holdups inside the system and, further, the circulation of
group and will focus on the phenomena taking place inside a fluidized OC between the reactors, it takes more than 800 s in the simulation to
bed reactor. reach a steady state after injection starts. The final conversion Xs of the
OC inside FR stage 1 is around 0.28. Stopping the fuel supply leads to
3.2. Movement of bed masses inside the system the opposite effect and the conversion Xs of the OC drops again towards
0. Comparing the simulation with the experimental observation, it be-
In a previous work, the dynamic changes of the fluid mechanics and comes clear that the simulation captures the fuel start accurately.
the circulation of bed material during cold operation of the pilot plant
was examined and discussed (Haus et al., 2017). A special focus was put
on the time, which the changes take inside the interconnected system. It
was shown that the fluid mechanics arrange to a new steady state
within 30–40 s, when operation parameters, like the fluidization velo-
city was changed. In this work, the fuel injection and its effect on the
bed masses and fluid mechanics is examined.

3.2.1. Experiment
When fuel is added to the FR system, the volumetric gas flow inside
is increasing, because one methane molecule is reacting to three pro-
duct molecules. This will cause a higher superficial gas velocity in the
reactor and, with it, the bubble volume fraction of the bed will increase
(Werther and Wein, 1994). The height of the bubbling bed is limited by
the standpipe height, so in the end, less solids can be stored in the FR
beds. This effect can be seen in the pressure drops during experiments,
when methane is added, shown in Fig. 9. It can be seen that the excess
material from the FR moves to the AR and increases the solids holdup
there. This effect is enlarged by the oxidation of the OC in the AR,
which will reduce the volumetric gas flow there. When the fuel supply
is stopped, the material moves back from the AR to the FR side. Due to
the strong pressure fluctuations in the plant a precise determination of
the solids holdup is difficult, but one can say that roughly 5–10% of the
bed mass leave the FR system and on the other side the AR bed mass
increases by 10–20%.

3.2.2. Simulation Fig. 8. Simulation of methane conversion inside the FR system with varied kinetic
The transient flowsheet simulation reveals the same behavior of the parameters. Shown concentrations in the lower dense zone are averaged over the bubble
and suspension phase. Comparison to averaged gas concentration measurements.
system that was observed in the experiments. Before the fuel is injected,

35
J. Haus et al. International Journal of Greenhouse Gas Control 72 (2018) 26–37

Fig. 9. Measured pressure drops over the whole reactor height of FR1, FR2 and AR. Methane injection started 13:30 and was stopped 13:55.

Fig. 10. Bed masses inside the system over time with a response to fuel injection at t = 100 s.

accurate prediction of the gas concentration prevailing in the AR over

time. This means that the secondary particle property, the conversion
state Xs, is correctly transferred from the fuel reactors and the AR.

4. Summary and conclusions

Dynamic unit models needed for CLC with methane were developed
and implemented into the novel Dyssol simulation software. The de-
veloped unit models for CLC comprise a CFB riser reactor, a bubbling
bed reactor, a cyclone and the loop seal. The dynamic flowsheet models
were used to simulate the hot operation of an actual CLC plant with
Fig. 11. Conversion of the OC from the lower stage FR1. Simulation of 2400 s of op-
methane used as a fuel over a duration of 40 min. Within this timeframe
eration, fuel injection starts at 100 s and stops at 1400 s. Samples taken during hot op-
the fuel supply was started and stopped again after several minutes.
eration after 120 s, 240 s, 480 s and 720 s after starting fuel injection and 120 s, 240 s,
480 s and 1000 s after fuel was stopped. Simulation with fitted kinetics. Starting the fuel injection inside the experimental plant made the bed
masses rearrange inside the system within around 30 s, which could be
reproduced with the dynamic flowsheet simulation. It took around
Contrarily, in the simulation the reactor system takes considerably
15 min finding a steady state for the solid conversion of the OC after
longer to go back to Xs = 0 after the fuel was stopped. This inaccuracy
fuel injection started.
can be traced back to the assumed mixing behavior of the solids in the
In a first simulation approach, using literature kinetics data for the
lower loop seal. There, a perfect mixing of solids hold-up with the in-
coming solids from the FR is assumed. In the pilot plant, it is probable
that the solids move through like a moving bed without too much
mixing of solids. In a next step, a model will be developed, which de-
scribes the solids movement inside the loop seals and their residence
time more accurately.
The re-oxidization of the depleted OC from the FR utilizes the
oxygen from the air inside the AR. Due to the fast oxidization reaction
of the OC, the amount of oxygen consumed is directly proportional to
the mass flow and conversion state of the OC coming from the FR. In
Fig. 12 the experimentally found oxygen concentrations after starting
the fuel are plotted together with the results from the modeling. It is Fig. 12. Oxygen concentration at the AR outlet. Simulation of the system with fitted
visible that the dynamic tracking of the OC conversion Xs leads to an kinetics compared to gas concentrations found during the start of fuel injection.

36
J. Haus et al. International Journal of Greenhouse Gas Control 72 (2018) 26–37

used OC from TGA measurement, it turned out that the CH4 conversion fuel reactor system for the chemical-looping combustion of lignite and bituminous
was overpredicted and CO and H2 conversion were underpredicted. The coal. Energy Technol. 4 (10), 1263–1273. http://dx.doi.org/10.1002/ente.
201600102.
reasons for these deviations between simulation and measurements Haus, J., Hartge, E.-U., Heinrich, S., Werther, J., 2017. Dynamic flowsheet simulation of
may either be due to differences in the OCs or due to different mass and gas and solids flows in a system of coupled fluidized bed reactors for chemical
heat transfer conditions in TGA and fluidized bed reactor, respectively. looping combustion. Powder Technol. 316, 628–640. http://dx.doi.org/10.1016/j.
powtec.2016.12.022.
A second simulation with adapted kinetics gave a good description of Klett, C., Hartge, E.-U., Werther, J., 2005. Time-dependent behavior of the ash particle
the measurements. It was possible to correctly describe the transient OC size distribution in a circulating fluidized bed system. Proc. Combust. Inst. 30 (2),
conversion inside the interconnected fluidized beds after a start of fuel 2947–2954. http://dx.doi.org/10.1016/j.proci.2004.08.164.
Kramp, M., Thon, A., Hartge, E.-U., Heinrich, S., Werther, J., 2012. Chemical looping
feeding. combustion of solid fuels–modeling and validation. 2nd International Conference on
Computational costs of the used flowsheet simulation are low Chemical Looping 2012.
compared to CFD simulations what makes real-time simulations of in- Kruggel-Emden, H., Stepanek, F., Munjiza, A., 2011. A comparative study of reaction
models applied for chemical looping combustion. Chem. Eng. Res. Des. 89 (12),
dustrial scale combustion systems feasible. These real-time simulations
2714–2727. http://dx.doi.org/10.1016/j.cherd.2011.05.013.
can later be implemented into operation control and predict the be- Kunii, D., Levenspiel, O., 1991. Fluidization Engineering. Wiley, New York.
havior of whole plants at load-changes what still is a major issue during Li, F., Zeng, L., Velazquez-Vargas, L.G., Yoscovits, Z., Fan, L.-S., 2010. Syngas chemical
operation. In a next step, solid fuel conversion will be implemented into looping gasification process: bench-scale studies and reactor simulations. AIChE J. 56
(8), 2186–2199. http://dx.doi.org/10.1002/aic.12093.
the simulation system and compared to existing experimental results Lyngfelt, A., Leckner, B., Mattisson, T., 2001. A fluidized-bed combustion process with
with bituminous coal and lignite. inherent CO2 separation; application of chemical-looping combustion. Chem. Eng.
Sci. 56 (10), 3101–3113. http://dx.doi.org/10.1016/S0009-2509(01)00007-0.
Mukherjee, S., Kumar, P., Yang, A., Fennell, P., 2015. A systematic investigation of the
Acknowledgments performance of copper-, cobalt-, iron-, manganese-and nickel-based oxygen carriers
for chemical looping combustion technology through simulation models. Chem. Eng.
The financial support of DFG (Deutsche Forschungsgemeinschaft) Sci. 130, 79–91. http://dx.doi.org/10.1016/j.ces.2015.03.009.
Ohlemüller, P., Alobaid, F., Ströhle, J., Bernd, E., 2014. Validation of chemical looping
within the priority program SPP 1679: “Dynamic simulation of inter- combustion process models by means of 100 kWth tests. 3rd International Conference
connected solids processes” is gratefully acknowledged; grant number on Chemical Looping 2014.
HA 6935/2-1. This research was supported by a Marie Curie Peltola, P., Ritvanen, J., Tynjälä, T., Pröll, T., Hyppänen, T., 2013. One-dimensional
modelling of chemical looping combustion in dual fluidized bed reactor system. Int.
International Research Staff Exchange Scheme Fellowship within the J. Greenh. Gas Control 16, 72–82. http://dx.doi.org/10.1016/j.ijggc.2013.03.008.
7th European Community Framework Programme. Peltola, P., Ritvanen, J., Tynjälä, T., Hyppänen, T., 2015. Fuel reactor modelling in
chemical looping with oxygen uncoupling process. Fuel 147, 184–194. http://dx.doi.
org/10.1016/j.fuel.2015.01.073.
References
Porrazzo, R., White, G., Ocone, R., 2014. Aspen Plus simulations of fluidised beds for
chemical looping combustion. Fuel 136, 46–56. http://dx.doi.org/10.1016/j.fuel.
Abad, A., Adánez, J., García-Labiano, F., Diego, L.F., de, Gayán, P., 2010. Modeling of the 2014.06.053.
chemical-looping combustion of methane using a Cu-based oxygen-carrier. Combust. Sahir, A.H., Dansie, J.K., Cadore, A.L., Lighty, J.S., 2014. A comparative process study of
Flame 157 (3), 602–615. http://dx.doi.org/10.1016/j.combustflame.2009.10.010. chemical-looping combustion (CLC) and chemical-looping with oxygen uncoupling
Abad, A., Gayán, P., Diego, L.F., de García-Labiano, F., Adánez, J., 2013. Fuel reactor (CLOU) for solid fuels. Int. J. Greenh. Gas Control 22, 237–243. http://dx.doi.org/10.
modelling in chemical-looping combustion of coal: 1. model formulation. Chem. Eng. 1016/j.ijggc.2014.01.008.
Sci. 87, 277–293. http://dx.doi.org/10.1016/j.ces.2012.10.006. Schlichthaerle, P., Werther, J., 1999. Axial pressure profiles and solids concentration
Adanez, J., Abad, A., Garcia-Labiano, F., Gayan, P., de Diego, L.F., 2012. Progress in distributions in the CFB bottom zone. Chem. Eng. Sci. 54 (22), 5485–5493. http://dx.
chemical-looping combustion and reforming technologies. Prog. Energy Combust. doi.org/10.1016/S0009-2509(99)00289-4.
Sci. 38 (2), 215–282. http://dx.doi.org/10.1016/j.pecs.2011.09.001. Sit, S.P., Grace, J., 1981. Effect of bubble interaction on interphase mass transfer ingas
Basu, P., Cheng, L., 2000. An analysis of loop seal operations in a circulating fluidized fluidized beds. Chem. Eng. Sci. 1981 (36), 327–335.
bed. Chem. Eng. Res. Des. 78 (7), 991–998. http://dx.doi.org/10.1205/ Skorych, V., Dosta, M., Hartge, E.-U., Heinrich, S., 2017. Novel system for dynamic
026387600528102. flowsheet simulation of solids processes. Powder Technol. 314, 665–679. http://dx.
Bolhàr-Nordenkampf, J., Pröll, T., Kolbitsch, P., Hofbauer, H., 2009. Comprehensive doi.org/10.1016/j.powtec.2017.01.061.
modeling tool for chemical looping based processes. Chem. Eng. Technol. 32 (3), Tasirin, S.M., Geldart, D., 1998. Entrainment of FCC from fluidized beds—A new corre-
410–417. http://dx.doi.org/10.1002/ceat.200800568. lation for the elutriation rate constants Ki∞*. Powder Technol. 95 (3), 240–247.
Chew, J.W., Cahyadi, A., Hrenya, C.M., Karri, R., Cocco, R.A., 2015. Review of entrain- http://dx.doi.org/10.1016/S0032-5910(97)03343-3.
ment correlations in gas–solid fluidization. Chem. Eng. J. 260, 152–171. http://dx. Thon, A., Kramp, M., Hartge, E.-U., Heinrich, S., Werther, J., 2014. Operational experi-
doi.org/10.1016/j.cej.2014.08.086. ence with a system of coupled fluidized beds for chemical looping combustion of solid
Choi, J.-H., Suh, J.-M., Chang, I.-Y., Shun, D.-W., Yi, C.-K., Son, J.-E., Kim, S.-D., 2001. fuels using ilmenite as oxygen carrier. Appl. Energ. 118, 309–317. http://dx.doi.org/
The effect of fine particles on elutriation of coarse particles in a gas fluidized bed. 10.1016/j.apenergy.2013.11.023.
Powder Technol. 121 (2-3), 190–194. http://dx.doi.org/10.1016/S0032-5910(01) Trefz, M., Muschelknautz, E., 1993. Extended cyclone theory for gas flows with high
00333-3. solids concentrations. Chem. Eng. Technol. 16 (3), 153–160. http://dx.doi.org/10.
Coppola, A., Solimene, R., Bareschino, P., Salatino, P., 2015. Mathematical modeling of a 1002/ceat.270160303.
two-stage fuel reactor for chemical looping combustion with oxygen uncoupling of Werther, J., Hartge, E.-U., 2004. A population balance model of the particle inventory in a
solid fuels. Appl. Energ. 157, 449–461. http://dx.doi.org/10.1016/j.apenergy.2015. fluidized-bed reactor/regenerator system. Powder Technol. 148 (2–3), 113–122.
04.052. http://dx.doi.org/10.1016/j.powtec.2004.09.005.
Davidson, J.F., Harrison, D., 1963. Fluidised Particles. Univ. Press Cambridge. Werther, J., Wein, J., 1994. Expansion behavior of gas fluidized beds in the turbulen-
Dosta, M., Heinrich, S., Werther, J., 2010. Fluidized bed spray granulation: analysis of the tregime. AlChE Symp. Ser. 1994 (90), 34–44.
system behaviour by means of dynamic flowsheet simulation. Powder Technol. 204 Yörük, C.R., Normann, F., Johnsson, F., Trikkel, A., Kuusik, R., 2015. Applicability of the
(1), 71–82. http://dx.doi.org/10.1016/j.powtec.2010.07.018. FB reactor of aspen plus for CFB oxy-fuel combustion of estonian oil shale: gas and
García-Labiano, F., Diego, L.F., de Adánez, J., Abad, A., Gayán, P., 2004. Reduction and solid hydrodynamics. 22nd Fluidized Bed Conversion Conference 2015 in Turku.
oxidation kinetics of a copper-based oxygen carrier prepared by impregnation for Zhang, W., Johnsson, F., Leckner, B., 1995. Fluid-dynamic boundary layers in CFB boilers.
chemical-looping combustion. Ind. Eng. Chem. Res. 43 (26), 8168–8177. http://dx. Chem. Eng. Sci. 50 (2), 201–210. http://dx.doi.org/10.1016/0009-2509(94)
doi.org/10.1021/ie0493311. 00222-D.
Haus, J., Lyu, K., Hartge, E.-U., Heinrich, S., Werther, J., 2016. Analysis of a two-stage

37