Vous êtes sur la page 1sur 8

Applied Thermal Engineering 73 (2014) 245e252

Contents lists available at ScienceDirect

Applied Thermal Engineering


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

Packed bed thermal energy storage model e Generalized approach


and experimental validation
Florian Opitz*, 1, Peter Trefnger
Offenburg University of Applied Sciences, Badstrae 24, 77652 Offenburg, Germany

h i g h l i g h t s

 A general heterogeneous lumped-element model of heat transfer in packed beds is developed.


 The proposed model expands the application range of existing models.
 Its applicability to different congurations of packed beds is shown.
 The implementation in Modelica favors an easy integration as a sub-model within bigger plant models.

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

Article history: Packed beds serve as thermal energy storages (TES) and heat exchangers (HEX) in different technological
Received 7 April 2014 applications. In this paper, a general heterogeneous model of heat transfer in packed beds is developed. It
Accepted 20 July 2014 is implemented by lumped element formulation in object-oriented modeling language Modelica and is
Available online 29 July 2014
successful validated with data sets taken from two different experiments reported in literature.
The main advantages of the introduced model are the general, theory-based approach and the lumped
Keywords:
element formulation in Modelica. The rst point mentioned above should allow to simulate a packed bed
Thermal energy storage
TES/HEX without the necessity applying measured data for model calibration or to apply specic heat
Packed bed
Object-oriented modeling
transfer correlations with restricted application. The second point establishes the possibility to integrate
Modelica the TES/HEX model within plant models of larger scale without increasing the simulation time
System model drastically.
Lumped element model 2014 Elsevier Ltd. All rights reserved.

1. Introduction integration into the energy system or the industrial plant under
investigation. Sensible heat storage is still dominating in practical
Fixed packed beds are of high importance in technology. There applications of high-temperature thermal energy storage (TES).
are manifold applications of packed beds in reactors, dryers, lters, However, signicant research is also done on packed bed TES
grate furnaces, and many more apparatuses. Packed beds play also a applying phase change material (PCM). Oro  et al. [2] investigated
prominent role as heat exchangers and thermal energy storages. different numerical models of packed bed TES with PCM. They
This paper deals with the modeling of packed beds as thermal compared also several correlations for the convective heat transfer
energy storage, which are of interest as energy storages facilities in between heat transfer uid and PCM capsules. In order to stabilize
concentrating solar power (CSP), compressed air energy storages the outow temperature Zanganeh et al. [3] propose to add a
(CAES), and also in high-temperature industrial processes. A review relative small layer with PCM on the top of a packed bed of rocks.
of thermal energy storages in combination with CSP has been done Design considerations for packed bed TES are derived by means of
by Kuravi et al. [1]. Several authors also worked on the modeling of exergy analysis. Bindra et al. [4] emphasize the effect of axial
packed bed thermal energy storages, whereby modeling shall dispersion and heat losses through the wall on exergy recovery. By
support the design of the thermal energy storage and also its the same authors further [5] the trade-off between thermal exergy
losses and exergy losses through pressure loss is investigated. To
* Corresponding author. Tel.: 49 781 205 302.
improve the exergy recovery the authors propose to add multiple
E-mail address: orian.opitz@hs-offenburg.de (F. Opitz). inlets and outlets which shall be employed according to the actual
1
URL: www.hs-offenburg.de. temperature prole.

http://dx.doi.org/10.1016/j.applthermaleng.2014.07.057
1359-4311/ 2014 Elsevier Ltd. All rights reserved.
246 F. Opitz, P. Trefnger / Applied Thermal Engineering 73 (2014) 245e252

Nomenclature k thermal conductivity, W m1 K1


M measured value, e
m number of observations, e
Dimensionless numbers n number of layers, e
Bi Biot number, e p pressure, Pa
Pe Peclet number, e r radius, m
Pr Prandtl number, e S simulated value, e
Re Reynolds number, e T temperature, K
t time, s
Greek letters U heat transfer coefcient through the wall, W m2 K1
d thickness of wall layer, m V volume, m3
h dynamic viscosity, Pa s w velocity, m s1
l oxygen-fuel ratio, e z axial coordinate of bed, m
r density, kg m3
3 void fraction, e Subscript
0 quantity of the complete packed bed
Latin letters 1 quantity of the core zone
m_ mass ow rate, kg s1 2 quantity of the wall zone
Q_ heat ow rate, W amb quantity of the ambient
A cross-sectional area, m2 bed quantity of the packed bed
cp specic heat capacity, J kg1 K1 f quantity of the uid phase
D diameter of packed bed, m i quantity of i-th layer
d diameter of particles, m in, out quantity at inlet/outlet
fa packing coefcient, - lam laminar
H height of packed bed, m p quantity of the particles
G mass ow rate per unit cross section, kg s1 m2 s quantity of the solid phase
h coefcient of heat transfer, W m2 K1 turb turbulent
h specic enthalpy, J kg1 w quantity of the wall

In general, the different modeling approaches can be distin- However, a relative similar arrangement is the packed bed applied
guished between the pseudohomogeneous model and the hetero- in a CSP plant with air as heat transfer uid. Having said this, there
geneous model of packed beds. The pseudohomogeneous model are two main requirements on the model. Firstly, the sub models
treats the packed bed as one continuous phase. Consequently, there for heat transfer, etc. in the model should be general valid to make
is only one temperature for solid and uid phase. Pseudohomoge- the model also applicable with signicant different parameters.
neous models give reasonable results for many industrial applica- Secondly, the model should be applicable with respect to simula-
tions in steady-state operation. Heterogeneous models treat the tion time as part of the system model of a larger system.
solid and the uid phase separately. Thus, a temperature and a
concentration gradient between solid and uid phase can be 2. Modeling of packed beds
dened. In transient operation, as it is the case for thermal energy
storages, a heterogeneous model is required. Most of the presented papers, dealing with the modeling of
Several authors have investigated the application of packed beds sensible TES, use a set of partial differential equations (PDE,
as thermal energy storage for CSP. Zavattoni et al. [6] developed a distributed system) and solve it by application of different nu-
three-dimensional computational uid dynamics model and vali- merical methods. In case of the analysis of more complex systems,
dated it succesfully by means of data of a technical scale reactor. e.g. complete power plants or industrial plants, the use of a lumped
Anderson et al. [7] developed a heterogeneous model which was element model seems to be more suitable, since here ordinary
applied to alumina packed bed with hot air as uid whereby Biot differential equations (ODE) are used. In these bigger models the
numbers Bi have been small. They point out that appropriate uid governing equations of several different components have to be
property models which take temperature dependencies into ac- evaluated, which scales up the computational effort for solving
count and appropriate heat transfer model are required to achieve distributed systems. Thus the evaluation of ODE is advantageous
good agreement with experimental data. Also Modi and Pe rez- over the evaluation of PDE. The object-oriented programming
Segarra [8] proposed a one-dimensional model for a thermocline language Modelica offers a capability to solve a system of different
thermal storage system to be applied within a CSP plant. With lumped models connected to each other [9].
respect to prior modeling efforts they considered e.g. temperature- According to the lumped element approach, the model is
dependent properties of the heat transfer uid and heat losses to divided into n equally spaced layers (Fig. 1). Every layer has the
the ambient. They report a good agreement of simulation results same volume Vi V/n and is further subdivided into different
with experimental data. phases and zones (see sections 2.1 and 2.2).
Several authors state that a key issue is selecting an appropriate
model for the heat transfer between the solid and the uid phase. 2.1. Modeling approaches
The intention of this paper is to develop a model, which is suitable
for system analysis of complete energy systems, namely for steel- The TES under consideration consists of a packed bed of metal
making processes. However, no experimental data are available to ller material; the packing materials of the TES used for validation
validate the thermal energy storage in this particular process. is crushed steatite [10], respectively steel [11]. The heat transfer
F. Opitz, P. Trefnger / Applied Thermal Engineering 73 (2014) 245e252 247

3 Vf/V. It describes the share of uid and solid material in a vol-


ume and takes the value of 3 1 in the case of solely uid phase,
respectively 3 0 if the volume is a solid body. In an innitely
extended, randomly packed bed the void fraction 3 takes a constant
value. At presence of a rigid wall the void fraction changes with the
radial position, the value increases near to the wall and the uid
experiences an increased passability in the wall region [14]. Further
this non-uniformity of the velocity inuences the heat transfer
between the two phases. In distributed models the distribution of
the void fraction is considered by different empirical equations
([15,16,13]). In the present lumped element formulation it can be
accounted for by a simple model proposed by Martin [17]. Herein
the cross-section is subdivided into two zones with different void
fractions 3 1 and 3 2, the mean void fraction of the complete packed
bed is 3 0 (Fig. 2). This distinction results in different pressure drops,
velocities and coefcients of heat transfer in the two zones. The
width of the wall zone is d/2 and thus the radius of the core zone is
r1 D/2d/2. The mean void fraction is given by
 
A1 A1
30 31 1  3 (1)
A A 2

and in the wall zone by

Fig. 1. Division of the packed bed model in n layers.

media in all cases are gases. Therefore both systems can be


described by the theory of uid ow in packed beds. Because of its
great signicance for the understanding of chemical reactors, many
published papers and models over the last decades dealt with the
different aspects of heat transfer, uid ow and mass transfer in
packed beds. Tsotsas [12] divides the available models in lumped
element and distributed system models and the latter in homo-
geneous and heterogeneous. Homogeneous formulation describes
the packed bed as a single phase, whereas in heterogeneous models
the energy equations of the uid and solid phase are considered
separately, interlinked by a heat transfer rate between the two
phases. Hence this approach has the following advantages [13]:

(1) The model is physically more feasible,


(2) Exchange processes between the two phases can be repro-
duced in detail,
(3) Differences in the temperature of the two phases can be
accounted for; in homogeneous models the temperature of
the single phase is regarded as uniform.

Especially the last point has to be considered because in TES


high temperature differences between the two phases have to be
expected, particularly in terms of transient operation [14]. For the
mentioned reasons - physical feasibility and easy integration within
more complex models - the present TES model is realized as a
lumped element system with heterogeneous phase formulation.

2.2. Accounting for the inuences of void fraction

The main parameter characterizing a packed bed, its thermal


behavior and the resulting uid ow regime is the void fraction Fig. 2. Zonal division of the packed bed model.
248 F. Opitz, P. Trefnger / Applied Thermal Engineering 73 (2014) 245e252

  setup is similar to the experimental conguration of Ha nchen,


2 D=d  7=8
32 1 $1  3 min (2) Brckner & Steinfeld ([10,27]), used for validation in the present
3 D=d  1=2
work. Gnielinski's equation is derived theoretically from Nusselt
with 3 min 0.23. In the present work the void fraction of the core numbers that apply to ow around a single sphere and thus con-
zone 3 1 is chosen in such a way that 3 0 of the model equals the stitutes a more general approach to the problem, which should be
experimentally evaluated mean void fraction. applicable to the data of different experimental congurations.
Both methods give strongly diverging values for the same mass
ow rates [10]. The correlation of Coutier and Farber [24] is:
3. Governing equations
 0:76
700 m_
Assuming conservation of mass, the energy equation of the uid hfs d0:24 (7)
61  3 A
phase at one layer i for both zones 1 (core) and 2 (wall) is
The equation of Gnielinski [26] is averaged over the Nusselt
dTf ;i
3 Vi rf ;i cp;f ;i $ m_ in;i $hin;i m_ out;i $hout;i Q_ fs;i Q_ fw;i : (3) numbers which apply to the laminar and turbulent ow regime
dt around a single sphere. This mean value is corrected by a coefcient
The present energy formulation isn't restricted to a certain ow fa which considers the inuence of the packing:
direction of mass or heat. Following the Modelica sign convention  q
the ow variables m_ and Q_ take positive values if its direction is into Nufs;i fa $ 2 Nu2lam;i Nu2turb;i (8)
a certain model, respectively negative values if outow occurs.
Within one layer perfect mixing is assumed, thus the specic
with
enthalpy at the outlets equals the specic enthalpy inside the layer:
hout,i hf,i. qq
Correspondent the energy equation of the solid material can be Nulam;i 0:644 Rebed;i 3 Prf ;i (9)
formulated as:
and
dTs;i
3  1 Vi rs;i cp;s;i $ Q_ fs;i Q_ sw;i Q_ cond;ax;i Q_ cond;rad;i
dt 0:037 Re0:8
bed;i Prf ;i
Nuturb;i   : (10)
(4) 0:1 2=3
1 2:443 Rebed;i Prf ;i  1
The last two terms on the right hand side account for heat
transfer by thermal conduction of the solid material. Due to the The packing coefcient is
impossibility of solving partial differential equations in Modelica,
fa 1 1:51  3 : (11)
the thermal conduction is approximated by lumped capacitances
(see 3.3). The thermal conductivity of the solid phase of the stag- The Biot number of the investigated packed bed of rocks is
nant packed bed kbed is calculated by a model, proposed by Zehner Bip > 0.1 (experimental data used for validation, given by Hanchen,
and Schlnder [18]. This model is widely accepted in literature Brckner & Steinfeld [10]). Therefore the particle temperature can't
([13,14,16,19]). The two energy Equations (3) and (4) are linked by be regarded to be uniform, the assumptions of the lumped capac-
the heat ow between uid and solid phase Q_ fs;i . The evaluation of itances approach aren't fullled and the heat transfer between the
this term is treated in section 3.1. two phases is reduced. In the equation of Coutier and Farber these
The pressure drop in one layer of the bed can be calculated by effects are already taken into account because it is derived from
the classical Ergun equation [20]: measurements on packed bed of rocks. In contrast, the heat transfer
coefcient calculated by Gnielinski has to be corrected to be able to
1  3 2 hf ;i 1  3 rf ;i 2 use lumped capacitances. Jeffreson [28] suggests the relationship
Dp 150 $wi $Hi 1:75 $wi $Hi (5)
32 d2 33 d
1 1 Bi=5
Allen [21] compares different correlations of pressure drop with : (12)
hfs;p;i hfs;i
experimental data. Though the Ergun equation seems to underes-
timate the pressure drop we rely on it, because the author's sug-
Herein hfs,p,i is the corrected coefcient of heat transfer (for ex-
gested best t method requires extensive experimental data.
amples of applications see Refs. [29,30]). In the present work
Equation (12) is calculated using the Biot number of the packed bed
3.1. Heat transfer between uid and solid phase
Bibed hfs dp/kbed.

In lumped element formulation the heat transfer between uid


3.2. Heat transfer through the wall
and solid phase can be considered by Newton's law of cooling:
  The heat transfer through the surrounding wall can be calcu-
Q_ fs;i hfs;i Ap;i Ts  Tf (6) lated by

The coefcient of heat transfer hfs has been investigated by  


Q_ fw=sw;i Ufw=sw;i Aw;i Tf =s;i  Tw;i : (13)
many researchers over the last decades (eg. Refs. [22e25], an
overview until 1990 is given in Ref. [12]). Depending on the avail-
The overall heat transfer coefcients Ufw,i and Usw,i are calcu-
able experimental data they found different dependencies, usually
lated as sum of the different thermal resistances on the inner side,
correlated by means of a Nusselt number Nu hfs dp/kf. In the
the outer side and within the wall material:
present work a correlation given by Coutier and Farber [24] as well
as one given by Gnielinski [26] is applied for hfs. 1 1 X dw;j 1
The rst correlation of Coutier and Farber was determined using (14)
Ufw=sw;i houtside j
kw;j hfw=sw;i
experimental values measured in a packed bed of rocks. Hence this
F. Opitz, P. Trefnger / Applied Thermal Engineering 73 (2014) 245e252 249

Thus, in the heterogeneous model two separate heat transfer Table 1


coefcients hfw,i and hsw,i for both phases at the inner side of the Parameters of the two test cases.

wall have to be determined. In both cases only the wall zone 2 Parameter Experiment 1 [10] Experiment 2 [11]
participates in the heat transfer between the phase and the wall. Solid material Steatite Steel
Due to the low consideration of the heterogeneous model in liter- Heat transfer uid Air Combustion gas
ature this separate transport parameters aren't accurately known D/m 0.148 0.09
yet [14]. Nevertheless Hein [13] recommends the usage of the d/m 0.02 0.0095
H/m 1.2 0.215
correlations found by Dixon and Cresswell [19]:
tcharge/s 4800 20
30 0.40 0.44
0:06
Nufw;i Pr Re0:75 (15) 31 0.36 0.42
0:52 f ;i bed;i
Gin/kg s1 m2 0.225 240
Tamb/K 298 308
and Tin/K 823 312e357
rs/kg m3 2680 7850
D=2 ks/W m1 K1 2.5 52
Bisw;i 2:12 (16) cp,s/J kg1 K1 1068 500
d n 42 43
According to Ha nchen, Brckner & Steinfeld [10] the coefcient
of heat transfer outside the wall houtside is calculated by a correla-
(M1) The general model described above (section 3) with zonal
tion for heat transfer by free convection at vertical surfaces given by
division, heat conduction and usage of the correlation by
Kast & Klan [31].
Coutier & Farber [24] for heat transfer coefcient hfs.
(M2) Model M1, but application of the correlation by Gnielinski
3.3. Thermal conduction in solid phase
[26] and the introduced Bi-correction (Equation (12)).
(M3) Model M2 without Bi-correction
Tsotsas [12] shows that thermal conduction in solid phase can't
be neglected for low Peclet numbers Pe0 < 100.
The rst two model congurations are validated with data,
Applying too simple models (so called standard model) would nchen, Brckner & Steinfeld [10], the third model
published by Ha
favor misinterpretations in correlating the experimental data and
with data by Srinivasan & Raghunandan [11]. For both cases all
would lead to strongly diverging heat transfer parameters between
boundary conditions are taken from the cited papers of the
different experimental studies. As low Peclet numbers occur in the
research groups (see Table 1). If initial conditions aren't given
experiments under consideration, relevant radial and axial con-
explicitly, they are read out from reported diagrams. Further ad-
duction terms are included in the energy Equation (4). In lumped
justments to the model aren't undertaken. Regarding geometric
element modeling, axial conduction can be considered by
conguration, thermophysical properties and mode of operation,
the two test cases differ strongly. The rst experiment is a packed
A 1  3  
Q_ s;ax;i kbed;i i Ts;i  Ts;i1 (17) bed of rocks, which is charged for 4800 s by air at a constant mass
Hi
ow rate and at a constant inlet temperature. This unit is used as
and radial conduction by pilot-scale experimental set-up in CSP technology. On the other
hand, the second experiment is a packed bed of steel spheres, used
2 p r1 Hi 1  3   in ball bearings. This TES is charged for 20 s by a combustion gas
Q_ s;rad;i kbed;0;i Ts;1  Ts;2 : (18) with time-varying inlet temperature. This unit is used as a compact
D=2
heat exchanger for different applications. The main parameters of
the two experiments are given in Table 1.
3.4. Thermophysical properties of the two materials
The simulation results and experimental measured values are
compared by means of a global mean percentage error' (MPE),
The present model is validated with data from two quite
which gives the average relative error of a test series. For this, the
different experiments. In both cases the thermophysical properties
relative differences between simulation and experiment are sum-
of the solid phase material are assumed to be constant. The used
med up and averaged by the number of observations. The calcu-
typical values are given in Table 1 according to the experiments of
lation of MPE is given in Equation (19).
Hanchen, Brckner & Steinfeld [10] and Srinivasan & Raghunandan
[11].
Due to its high temperature differences in TES applications, air is
100 Xm
Mj  Sj
modeled as an ideal mixture of nitrogen, oxygen, carbon-dioxide MPE $ (19)
m j1 Mj
and argon, dependent on temperature. The specic heat capacity
cp,f and enthalpy hf of the mixture gases are calculated by VDI-
guideline 4670 [32]. The dynamic viscosities hf are calculated by 4.1. Validation of model M1
polynomial equations [33] and thermal conductivities kf by inter-
polation of values given in VDI heat atlas [33]. Accordingly, any The rst model M1 is validated using data published by
other mixture of gases can be described by this approach. Hanchen, Brckner & Steinfeld [10]. There, the correlation of
Coutier & Farber is used for calculation of heat transfer coefcient
4. Simulation results, validation and discussion hfs between the two phases. This correlation was determined at a
packed bed of rocks, which is quite similar to the experimental
As mentioned before the aim of this work is the development of setup of Hanchen, Brckner & Steinfeld. For example, the following
a general model which is applicable to different storage congu- parameters of the two experiments are comparable: Bed height H
rations. In the following, simulation results of three different pa- (1.2 m vs. 0.84 m), particle size d (0.02 m vs. (0.0180.030)m) and
rametrizations are compared to literature data to validate the mass ow rate per unit cross section G (0.225 kg s1 m2 vs. (0.02
general model developed before. The tested congurations are: 4)kg s1 m2).
250 F. Opitz, P. Trefnger / Applied Thermal Engineering 73 (2014) 245e252

Table 2
MPE of model M1.

t/s MPE/%

Core Mix

1200 2.36 4.75


3000 1.81 7.30
4800 1.10 5.20

values lying between the simulated values for the two phases are
reasonable.
As can be seen in the gures, the core temperatures (Fig. 3)
compare better to the measured values than the mixing tempera-
tures (Fig. 4). This nding is supported by Table 2: At any time, the
mean average error MPE of the core temperatures is less than the
MPE of the mixing temperatures. Obviously there is a difference
between the temperatures in the core and the wall zone. In the
present model the effect of bypass ow, observed by Ha nchen,
Brckner & Steinfeld, is taken into account by the zonal model of
Martin. Consequently a correction of the mass ow rate, suggested
by the authors to t the measured temperatures, isn't necessary
here.

4.2. Validation of model M2

Fig. 3. Core uid and solid temperatures of simulation M1 compared to literature data Model M2 is validated with the previous data of Ha nchen,
[10]. Brckner & Steinfeld, too. Compared to M1 the more general cor-
relation for heat transfer of Gnielinski is used instead of the very
specic correlation of Coutier & Farber. Because the Biot number of
Figs. 3 and 4 show the simulation results obtained with model the rock material is Bi p> 0.1, the introduced Bi correction is needed
M1. In Fig. 3 the distribution of uid and solid core temperatures additionally. Fig. 5 shows the simulation result of model M2. Again,
over the bed height for three different charging times are plotted. the core temperatures of both phases are plotted. Due to similar
Additionally Fig. 4 shows the uid and solid temperatures, behavior as in simulation M1, the mixing temperatures aren't
assuming mixing between the two zones. Because the measured shown here again. Analogous to the previous simulation, the MPE
temperatures can't be assigned clearly to one phase, experimental at all times are given in Table 3.

Fig. 5. Core uid and solid temperatures of simulation M2 compared to literature data
Fig. 4. Mixing temperatures of simulation M1 compared to literature data [10]. [10].
F. Opitz, P. Trefnger / Applied Thermal Engineering 73 (2014) 245e252 251

It becomes apparent that the data of Ha nchen, Brckner &


Steinfeld can be reproduced satisfactory by model M2, too.
Compared to model M1 the MPE are in the same range. Here again,
the temperature distribution in the core zone seems to describe the
data better than the distribution of the mixing temperatures.
Nevertheless, the divergence between the two approaches isn't as
obvious as in the rst model any more. At time t 4800 s the
mixing temperature ts the experimental data even better. It's
noticeable that the MPE of the core temperatures increases by time,
while it declines in the rst model.
Nevertheless, the crucial advantage of the second model M2, is
the usage of a general, theory-based correlation for the coefcient
of heat transfer. This approach provides one means to simulate
packed bed thermal storage without the necessity of measured data
or a specic correlation, matching exactly the investigated
conguration.

4.3. Validation of model M3

In the following, a second, quite different experiment of Srini-


vasan & Raghunandan [11] is used for validation of model M3. In
order to check the introduced general model, the coefcient of heat
transfer between the two phases is calculated by the correlation of
Gnielinski again. Due to the usage of steel spheres with high ther-
mal conductivity in the experiment, the Biot number is relative
small and the heat transfer coefcient has not to be corrected.
Fig. 6. Core uid temperatures of simulation M3 compared to literature data [11].
Hence, M3 differs from M2 only in neglecting the Bi correction. The
ndings, gained with the rst two models M1 and M2, support this 5. Conclusion and further work
theory-based approach.
Srinivasan & Raghunandan [11] investigate the dynamic We developed a general model for the description of heat
behavior of a packed bed of steel spheres, which is charged by a hot transfer in packed beds. Due to high temperature differences a
combustion gas. Hence, this experiment differs strongly from the variable uid model, based on VDI guideline 4670, was introduced.
one before regarding the geometric dimensions as well as the For the sake of computational speed and physical feasibility a
thermophysical properties of the material and the selected mode of lumped element model with heterogeneous formulation of the two
operation (see Table 1). The authors aim to derive a new Nu- phases has been chosen. Thus, the model describes the thermal
correlation for this TES conguration. Therefore the temperatures behavior of packed beds by means of ordinary differential
of the uid and solid phase are measured separately.
The heat transfer medium, resulting from the combustion of a
hydrogenenitrogen-mixture with air, can be easily described as a
ideal gas mixture by the introduced uid model (see section 3.4).
Due to missing information from the original work, oxygen-rich
combustion is assumed (l 1.6). The resulting combustion gas
consists of 79.9% nitrogen, 12.4% water vapor, 6.6% oxygen and 1.1%
argon.
Figs. 6 and 7 show simulation results of model M3. Here the
temperature development of the two phases over time for three
positions in the bed is plotted. Additionally in Fig. 6 the experi-
mental inlet temperature is plotted, which serves as a input vari-
able to the simulation. In Table 4 the MPE of the six measurement
series are given.
The different temperatures of the two phases are reproduced
sufciently exact by the heterogeneous model. The MPE values are
even a little lower compared to the previous simulations. Of
particular note are the differences between experiment and
simulation which increase with advancing position z. Therefore,
longer packed beds should be investigated in future work.

Table 3
MPE of model M2.

t/s MPE/%

Core Mix

1200 1.82 3.97


3000 1.89 5.16
4800 3.36 3.07
Fig. 7. Core solid temperatures of simulation M3 compared to literature data [11].
252 F. Opitz, P. Trefnger / Applied Thermal Engineering 73 (2014) 245e252

Table 4 [6] S. Zavattoni, M. Barbato, A. Pedretti, G. Zanganeh, A. Steinfeld, High temper-


MPE of model M3. ature rock-bed TES system suitable for industrial-scale CSP plant e CFD
analysis under charge/discharge cyclic conditions, Energy Procedia 46 (0)
Position MPE/% (2014) 124e133, http://dx.doi.org/10.1016/j.egypro.2014.01.165, 8th Inter-
z/m national Renewable Energy Storage Conference and Exhibition (IRES 2013).
Fluid Solid
[7] R. Anderson, S. Shiri, H. Bindra, J.F. Morris, Experimental results and modeling
0.05 0.08 0.11 of energy storage and recovery in a packed bed of alumina particles, Appl.
0.10 0.71 0.55 Energy 119 (0) (2014) 521e529, http://dx.doi.org/10.1016/
0.15 1.92 1.31 j.apenergy.2014.01.030.
[8] A. Modi, C.D. Perez-Segarra, Thermocline thermal storage systems for
concentrated solar powerplants: one-dimensional numerical model and
equations (ODE). This approach is considered in literature only low, comparative analysis, Sol. Energy 100 (2014) 84e93, http://dx.doi.org/
10.1016/j.solener.2013.11.033.
since most models use distributed systems and partial differential [9] Modelica and the Modelica Association, Modelica Association, 2014 visited on
equations (PDE). The use of ODE instead of PDE scales down the 2014-03-24, www.modelica.org.
[10] M. Ha nchen, S. Brckner, A. Steinfeld, High-temperature thermal storage us-
computational effort, needed to solve the mathematical system.
ing a packed bed of rocks e heat transfer analysis and experimental valida-
The TES model was validated with data from two different ex- tion, Appl. Therm. Eng. 31 (2011) 1798e1806, http://dx.doi.org/10.1016/
periments, which differ strongly regarding geometric congura- j.applthermaleng.2010.10.034.
tion, thermophysical properties and mode of operation. The [11] R. Srinivasan, B. Raghunandan, Experiments on thermal response of low
aspect ratio packed beds at high Reynolds numbers with varying inow
simulation results obtained with a model applying a general, temperature, Exp. Therm. Fluid Sci. 44 (2013) 323e333, http://dx.doi.org/
theory-based correlation for heat transfer coefcient by Gnielinski 10.1016/j.expthermusci.2012.07.004.
[22] meet the experimental data gained with the two different [12] E. Tsotsas, ber die W arme- und Stoffbertragung in durchstro mten Fes-
tbetten, No. 223 in Fortschritts-Berichte VDI, Reihe 3, VDI Verlag, Dsseldorf,
experimental setups. The required parameters (geometry of packed Germany, 1990.
bed, inlet and initial conditions, thermophysical properties of bed [13] S. Hein, Modellierung wandgekhlter katalytischer Festbettreaktoren mit Ein-
material) are summarized in Table 1. In literature many different und Zweiphasenmodellen, No. 593 in Fortschritts-Berichte VDI, Reihe 3, VDI
Verlag, Dsseldorf, Germany, 1999.
correlations for the coefcient of heat transfer between the two
[14] E. Tsotsas, Thermal conductivity of packed beds, in: VDI-GVC (Ed.), VDI Heat
phases are given, which are highly dependent on the experiments, Atlas, Springer, Berlin/Heidelberg, Germany, 2010, pp. 570e580.
the data is taken from. This is rather disadvantageous if a system [15] E. Tsotsas, Heat and mass Transfer in packed beds with uid ow, in: VDI-GVC
should be modeled, for which a tting correlation or own experi- (Ed.), VDI Heat Atlas, Springer, Berlin/Heidelberg, Germany, 2010, pp. 1327e1341.
[16] M. Winterberg, E. Tsotsas, A. Krischke, D. Vortmeyer, A simple and coherent
mental data isn't available. set of coefcients for modelling of heat and mass transport with and without
Further concepts, which contribute to the high generality of the chemical reactions in tubes lled with spheres, Chem. Eng. Sci. 55 (2000)
model, are the zonal approach of Martin [17] and the Bi correction 967e979, http://dx.doi.org/10.1016/S0009-2509(99)00379-6.
[17] H. Martin, Low peclet number particle-to-uid heat and mass transfer in
by Jeffreson [28]. The rst concept accounts for the higher pass- packed beds, Chem. Eng. Sci. 33 (1978) 913e919, http://dx.doi.org/10.1016/
ability of the uid in the wall region due to higher void fraction. The 0009-2509(78)85181-1.
second concept considers the reduced heat transfer due to low [18] P. Zehner, E. Schlnder, W higkeit von Schttungen bei ma
armeleitfa igen
Temperaturen, Chem. Ing. Tech. 14 (1970) 933e941, http://dx.doi.org/
thermal conductivity of the solid ller material. 10.1002/cite.330421408.
The model should be tested with further experimental cong- [19] A. Dixon, D. Cresswell, Theoretical prediction of effective heat transfer pa-
urations expanding the range of packed bed parameters and rameters in packed beds, AIChE J. 25 (1979) 663e676, http://dx.doi.org/
10.1002/aic.690250413.
operational parameters. Especially the case of transient operated [20] K.-E. Wirth, Pressure drop in xed beds, in: VDI-GVC (Ed.), VDI Heat Atlas,
long packed beds should be investigated. Springer, Berlin/Heidelberg, Germany, 2010, pp. 1106e1110.
The TES model was implemented by object-oriented modeling [21] K.G. Allen, Performance Characteristics of Packed Bed Thermal Energy Storage
for Solar Thermal Power Plants (M.Sc. dissertation), University of Stellen-
language Modelica. A main focus has been placed on the possibility
bosch, South Africa, 2010.
to integrate the TES model within bigger plant models. This aim is [22] P. Rowe, K. Claxton, Heat and mass transfer from a single sphere to uid
supported by the concept of re-usability in Modelica. For this owing through an array, Trans. Institut. Chem. Eng. 43 (1965) T321eT331.
reason, the introduced model will be used within different plant [23] B. Gillespie, E. Crandall, J. Carberry, Local and average interphase heat transfer
coefcients in a randomly packed bed of spheres, AIChE J. 14 (1968) 483e490,
models like CSP, CAES or other system applications. http://dx.doi.org/10.1002/aic.690140326.
[24] J. Coutier, E. Farber, Two applications of a numerical approach to heat transfer
process within rock beds, Sol. Energy 29 (1982) 451e462, http://dx.doi.org/
Acknowledgement 10.1016/0038-092X(82)90053-6.
[25] C. Wu, G. Hwang, Flow and heat transfer characteristics inside packed and
This research project is funded by the German Federal Ministry uidized beds, J. Heat Transfer 120 (1998) 667e673, http://dx.doi.org/
10.1115/1.2824335.
of Education and Research (BMBF) within the framework concept [26] V. Gnielinski, Fluid-particle heat transfer in ow through packed beds of
IngenieurNachwuchs (fund number 03FH002I2). solids, in: VDI-GVC (Ed.), VDI Heat Atlas, Springer, Berlin/Heidelberg, Ger-
many, 2010, pp. 743e744.
[27] A. Meier, C. Winkler, D. Wuillemin, Experiment for modeling high tempera-
References ture rock bed storage, Sol. Energy Mater. 27 (1991) 255e264, http://
dx.doi.org/10.1016/0165-1633(91)90066-T.
[1] S. Kuravi, J. Trahan, D.Y. Goswami, M.M. Rahman, E.K. Stefanakos, Thermal [28] C. Jeffreson, Prediction of breakthrough curves in packed Beds: I. Applicability
energy storage technologies and systems for concentrating solar power of single parameter models, AIChE J. 18 (1972) 409e416, http://dx.doi.org/
plants, Prog. Energy Combust. Sci. 39 (4) (2013) 285e319, http://dx.doi.org/ 10.1002/aic.690180225.
10.1016/j.pecs.2013.02.001. [29] P. Li, J. van Lew, W. Karaki, C. Chan, J. Stephens, Q. Wang, Generalized charts of
[2] E. Oro , J. Chiu, V. Martin, L.F. Cabeza, Comparative study of different numerical energy storage effectiveness for thermocline heat storage tank design and
models of packed bed thermal energy storage systems, Appl. Therm. Eng. 50 calibration, Sol. Energy 85 (2011) 2130e2143, http://dx.doi.org/10.1016/
(2013) 384e392, http://dx.doi.org/10.1016/j.applthermaleng.2012.07.020. j.solener.2011.05.022.
[3] G. Zanganeh, M. Commerford, A. Haselbacher, A. Pedretti, A. Steinfeld, Stabi- [30] B. Xu, P.-W. Li, C. Chan, Extending the validity of lumped capacitance method
lization of the outow temperature of a packed-bed thermal energy storage for large Biot number in thermal storage application, Sol. Energy 86 (2012)
by combining rocks with phase change materials, Appl. Therm. Eng. 70 (2014) 1709e1724, http://dx.doi.org/10.1016/j.solener.2012.03.016.
316e320, http://dx.doi.org/10.1016/j.applthermaleng.2014.05.020. [31] W. Kast, H. Klan, Heat transfer by free convection: external ows, in: VDI-GVC
[4] H. Bindra, P. Bueno, J.F. Morris, R. Shinnar, Thermal analysis and exergy (Ed.), VDI Heat Atlas, Springer, Berlin/Heidelberg, Germany, 2010, pp. 667e672.
evaluation of packed bed thermal storage systems, Appl. Therm. Eng. 52 [32] VDI-Guideline 4670, Dsseldorf, Germany, Thermodynamic Properties of
(2013) 255e263, http://dx.doi.org/10.1016/j.applthermaleng.2012.12.007. Humid Air and Combustion Gases, 2003.
[5] H. Bindra, P. Bueno, J.F. Morris, Sliding ow method for exergetically efcient [33] M. Kleiber, R. Joh, Properties of pure uid substances e liquids and gases, in:
packed bed thermal storage, Appl. Therm. Eng. 64 (2014) 201e208, http:// VDI-GVC (Ed.), VDI Heat Atlas,, Springer, Berlin/Heidelberg, Germany, 2010,
dx.doi.org/10.1016/j.applthermaleng.2013.12.028. pp. 301e393.

Vous aimerez peut-être aussi