You are on page 1of 11

Applied Energy 113 (2014) 1338–1348

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Modeling direct steam generation in solar collectors with multiphase


CFD
David H. Lobón a, Emilio Baglietto b, Loreto Valenzuela c,⇑, Eduardo Zarza c
a
Convenio Universidad de Almería-Plataforma Solar de Almería, Crta. Senes, s/n, Tabernas, Almeria E04200, Spain
b
Massachusetts Institute of Technology, Department of Nuclear Science and Engineering, 77 Massachusetts Avenue, Cambridge, MA 02139-4307, USA
c
CIEMAT, Plataforma Solar de Almería, Crta. Senes, s/n, Tabernas, Almeria E04200, Spain

h i g h l i g h t s

 Direct steam generation in parabolic-trough solar collectors.


 CFD package STAR-CCM+ is used for simulating of DSG in solar collectors.
 CFD model is validated with DISS test facility experimental data.

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

Article history: The direct steam generation in parabolic-trough solar collectors, using water as heat-transfer fluid, is an
Received 8 March 2013 attractive option for the economic improvement of parabolic trough technology for solar thermal elec-
Received in revised form 10 August 2013 tricity generation in the multi megawatt range or industrial process heat supply. But the existence of sin-
Accepted 19 August 2013
gle-phase and two-phase flow in the absorber pipes of the solar collectors constitutes a challenge for the
Available online 27 September 2013
development of simulation tools and process control schemes suitable for this type of solar technology.
The computational fluid dynamic package STAR-CCM+ is used to implement an efficient multiphase
Keywords:
model capable of simulating the behavior of direct steam generation in parabolic-trough solar collectors.
Parabolic-trough solar collector
Direct steam generation
This work describes the modeling approach and summarizes the comparison of simulation results with
CFD modeling the measurements taken at a direct steam generation solar test facility located at the Plataforma Solar de
DISS test facility Almería, Spain.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction collectors connected in series, with a total length of 700 m. At


the end of 2012 three new collectors were installed to increase
The feasibility of the direct steam generation (DSG) process in the loop length and thermal power delivered by the solar field. In
parabolic-trough collectors (PTCs) has already been proven in the the old configuration, the one selected in the present work, the first
DISS project [1,2] under real solar condition, at pressures up to two collectors are EuroTrough-type collectors which have a length
100 bar and temperatures up to 400 °C, with more than 10,000 h of 100 m each, while the remaining collectors are modified LS-3-
of operation. In a later project the detailed engineering for a pre- type collectors which have a length of 50 m each, with the excep-
commercial DSG solar thermal power plant was performed [3]. tion of collectors #11 and #12 which are special test collectors
The detailed engineering of the collector field requires accurate with a length of 25 m. All parabolic-trough collectors have a parab-
understanding of the occurring thermo-hydraulic phenomena ola width of 5.76 m and a focal length of 1.71 m. The material of
and of their influence on the stability of the absorber tubes. DSG the former absorber pipes is Ferritic Steel A335 P22.
in parabolic troughs is considered as one of the most feasible op- The collector loop is physically divided into two sections by a
tions for the next generation of solar power plants using this solar water/steam separator. Concerning the DSG process there exist
technology [4–6]. three zones in the collector loop during nominal operation:
The DISS solar field test facility (see Fig. 1), located at the Plata-
forma Solar de Almería (PSA) in Spain, was composed of a single – Preheating section: first collectors (one or two depending
North-South oriented loop composed of 13 parabolic-trough on temperature of feed water) are used to preheat the feed
water up to saturation temperature.
⇑ Corresponding author. Tel./fax: +34 950387934.
– Boiling section: from collector #2 or #3 to collector #10 or
E-mail address: loreto.valenzuela@psa.es (L. Valenzuela).
#11 water is converted into saturated steam.

0306-2619/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.apenergy.2013.08.046
D.H. Lobón et al. / Applied Energy 113 (2014) 1338–1348 1339

Nomenclature

Roman T temperature (°C)


f(u) normalized heat-flux distribution (dimensionless)
~
f external body forces (N) Greek
g gravitational acceleration (m/s2) h angle of incidence (degrees)
hj enthalpy of the species j (J/kg) l liquid viscosity (kg/m s)
hlat enthalpy of vaporization (J/kg) q density (kg/m3)
keff effective conductivity (W/m K) r surface tension (N/m)
m_ ec rate interfacial exchange between the vapor and the li- s transmittance of the absorber glass cover (dimension-
quid (kg/s m2) less)
m_ ew vapor mass generation on the heating surface (kg/s m2) s stress tensor (N/m2)
np Prandtl number exponent seff viscous stress tensor (N/m2)
p pressure (MPa)
qbw surface boiling heat flux (W/m2) Subscripts
t time (s)
in inlet
v
~ velocity vector (m/s) l liquid
Cew rate heat flux fraction used to create vapor bubbles sat saturation
(dimensionless) v vapour
Cl flux concentration ratio (dimensionless)
w wall
Cp specific heat capacity (J/kg K)
Cqw constant dependent on the liquid/surface combination
Acronyms
(dimensionless)
C HTCArea local heat transfer coefficient between the vapor and the CFD Computational Fluid Dynamics
surrounding liquid, multiplied by the interfacial area DISS Direct Solar Steam
DSG Direct Steam Generation
density (W/m2 K)
E energy transfer (J/kg) LHM Local Homogeneous Model
Q heat flux per unit area (W/m2) MIT Massachusetts Institute of Technology
Qloss heat loss per unit area (W/m2) MTD Maximum Temperature Difference
NSE Nuclear Science and Engineering
Pr Prandtl number (dimensionless)
Re Reynolds number (dimensionless) NURBS Non-Uniform Rational B-Spline
Ref mirrors reflectance (dimensionless) PSA Plataforma Solar de Almería
Sct turbulent Schmidt number (dimensionless) PTC Parabolic-Trough solar Collector
SQ Steam Quality
Sh volumetric heat source (W/m3)
Sm mass source (kg)

– Superheating section: is formed by the last two or three flows at the top. Almanza et al. [8] and Eck and Steinmann [9]
collectors of the loop. showed that under these conditions, bending of tubes during direct
steam generation was a realistic possibility. Heat transfer modeling
The solar field can be configured in different operation modes: and temperature fields in the wall of the absorber tubes of para-
bolic-trough collectors have been studied mainly when thermal
– Recirculation mode: boiling and superheating sections are oil is the heat transfer fluid, i.e. there is single-phase flow in the ab-
completely separated and the middle water/steam separa- sorber [10–14]. Concerns temperature fields in the wall of absorber
tor receives two-phase flow coming from the boiling sec- tubes with steam as heat transfer fluid have been investigated by
tion, saturated steam goes to superheating section and Eck and Steinmann, 2005 [9] and Roldan et al. [15]. In these studies
excess liquid water is recirculated to the solar field inlet a commercial CFD code is used to investigate the thermal behavior
by a recirculation pump. of absorber pipes but not the complete thermo-hydraulic behavior
– Once-through mode: the water/steam separator is by- of the DISS test facility. The possible extreme thermal stresses aris-
passed and the end of the boiling section and the beginning ing from stratified flow in the DISS collector when the absorber
of the superheating section are not exactly defined. Feed pipes are heated from one side were only examined analytically
water entering at the inlet of the solar field is directly con- by Eck et al. [7].
verted into superheated steam without any recirculation System transients in the DISS system were investigated initially
through the water/steam separator. by Eck and Steinmann [16], providing details of the techniques
adopted in the Modelica package. A later dynamic modeling ap-
The superheating section is also equipped with water injectors. proach of the DISS facility using Modelica software was given by
In particular, injection of preheated water is done before the last Hirsch et al. [17,18], and more recently by Moya et al. [19] under
collector to control the outlet steam temperature. This injection RELAP environment.
is used both in once-through operation mode and in recirculation Rheinländer and Eck [20] performed a numerical study of the
mode. pressure losses in the DISS system that reports a comparison of
A key area of investigation in parabolic-trough collectors is the experimental pressure drops in the DISS project with results of
thermal stress in the absorber pipes and more important when the empirical pressure drop correlations of Lockhardt–Martinelli–
heat transfer fluid is water [7]. The effect can be compounded if the Thom and Bandel and Friedel. Eck et al. [2] gave a review of the ap-
two-phase flow inside the absorber pipe is stratified, meaning that plied research about the DSG process within the DISS project. With
the liquid water flows at the bottom of the pipe and the steam regard to pressure drops, it was found that use of the Friedel
1340 D.H. Lobón et al. / Applied Energy 113 (2014) 1338–1348

Fig. 1. Schematics of DISS test facility sited at the Plataforma Solar de Almería (Spain).

correlation [21] gives results that over-predict experimental and superheating section). The solar field configured in once-
pressure drops by 20%. It was found that in order to ensure that through mode is used for model validation.
circumferential temperature variations in the absorber tubes are Fig. 2 shows the typical configuration of interconnection be-
kept within design limits, higher water mass flow rates should tween collectors in the experimental DISS test facility. It can be
be used, which was later confirmed by experiment. noted that the interconnections geometrical layout varies between
Given the relevance of the issue, and the limitations shown by collectors; this variation is due to the fact that the DISS plant has a
the modeling approaches adopted in the past, particularly those drop of about 6 m in height across the field. Fig. 2 also shows the
related to pressure loss modeling using empirical or semi-empirical instrumented measurement locations.
correlations, three-dimensional CFD modeling of the DSG process in Fluid temperature measurements were performed with the use
parabolic troughs is considered particularly advantageous and is of thermocouples mounted in thermowells inserted in the piping.
investigated in this work. Accurate CFD modeling could offer clear Thermocouples used are type J class I, which have a nominal accu-
benefits in the detailed design of DSG solar collector fields, allowing racy of ±2 °C, according to the standard IEC 60584.2 (1982). Fur-
identifying critical process conditions that may lead to anticipated thermore some absorber tubes are instrumented by type K
failures as two-phase flow phenomena occurring in the different thermocouples (class 1) each one to measure the wall temperature
hydraulic components forming a DSG solar field. profile at the cross section. Thermocouples in these cross sections
Section 2 details characteristics of the experimental set up. Sec- are distributed as shown in Fig. 3; at the end of each parabolic-
tion 3 is devoted to describe the CFD modeling approach adopted trough collector there is at least one absorber tube instrumented.
to simulate the DISS test facility. Section 4 depicts the boundary The cross-sectional wall temperature profile is closely related to
conditions considered to run simulation tests to reproduce exper- the non-uniform heat flux profile on the absorber tubes. Coplanar
imental data available. And finally Section 5 includes the discus- absolute and differential pressure transmitters were also installed;
sion of results obtained. their accuracy is 0.5% leading to maximum uncertainties of the
pressure measurements of ±0.06 MPa and ±0.004 MPa respectively.
Volumetric flow rates for water and steam were measured with
2. DISS test facility data orifice-plate flow meters from Fisher-Rosemount, which have an
accuracy of 1.6%, resulting in a maximum uncertainty of the
In order to validate the capabilities of the CFD approach, the mass-flow measurements of 0.05 kg/s. Direct normal irradiance
field measurements from the first 10 LS-3 collectors (Fig. 1) are was measured with a pyrheliometer model NIP from Eppley Labo-
used, as this configuration represents the minimum solar field ratory Inc.; this measurement has an maximum uncertainty of
capable of producing the three heating ranges (preheating, boiling ±10 W/m2.

Fig. 2. Schematic set up of the three different kinds of the interconnections between solar collectors along of the DISS test facility.
D.H. Lobón et al. / Applied Energy 113 (2014) 1338–1348 1341

Fig. 3. Situation of the thermocouples in the cross-sections along one of the absorber tube of the DISS test facility.

Fig. 4. Cross section elbow, interconnections and solar absorber tube.

3. CFD modeling of the DSG process is inside the logarithmic region of the turbulent boundary layer.
This condition is usually expressed in the form of Y+ requirement
The large size of a DSG solar collector facility represents a par- which should be in the region 30–100, and which was correctly en-
ticular challenge for CFD modeling. The facility is composed by forced for all calculations. The use of incorrect near wall meshes
hundreds of meters of collectors, connected by complex junctions would result in considerable errors in the velocity and pressure
of different diameters. Furthermore both solid and fluid domains drop predictions.
should be represented in the CFD model in order to correctly ana- The modeled solar field was designed to operate at three differ-
lyze temperature and consequently thermal stress distribution on ent conditions, represented by pressure pin = (35, 68, 108) bar, tem-
the components. Constructing an efficient computational mesh is perature Tin = (200, 240, 270) °C which results in a consistent
therefore particularly challenging, and is a good test for modern subcooling of approximately 40 °C; the respective mass flow rates
automatic meshers. It was found that the STAR-CCM+ software range from 0.45 to 0.73 kg/s. In these conditions, the Reynolds
[22] is an optimal approach, in the form of a specialized ‘‘general- number is between 5  104 and 9  104 which indicates that the
ized cylinder mesher’’. The mesher uses a particularly clever tech- flow is always largely turbulent; the interconnection for each col-
nique, based on Non-uniform Rational Basis Spline (NURBS) [23] lector introduces further considerable mixing and turbulence.
surfaces; which uses rational Bézier curves and a non-uniform The very large geometrical domains being modeled put a very
explicitly given knot vector. Therefore, degree, control points, strict requirement on the efficiency of the solution method imple-
weights, and knot vector are needed to specify a NURBS curve mented in CFD. In particular 2-phase flow turbulent simulations
and to generate an optimal flow aligned computational cells. An are necessary which are especially demanding. This work takes
example of the computational mesh is shown in Fig. 4; where it advantage of the specific work performed at the NSE department
is shown that flow aligned prismatic cells are used in the bulk of of MIT on the validation of a fast running 2-phase and boiling flow
the flow domain, while optimal body fitted cells are used in the model [25]. The model details are presented in the following
near wall flow region as well as in the solid region. subsection.
A rigorous grid sensitivity analysis was performed, with models All simulations presented have been run as steady state, where
that ranged from 1.7 to 2.9 million cells per collector. The analysis separate tests demonstrated that no measurable differences are
showed a maximum variation of the computed pressure drop along obtained for transient simulations.
the solar collectors’ loop and temperature of 3% at the end of the
process. The coarser mesh model was then adopted, as a good com- 3.1. Locally homogeneous boiling model
promise between reasonable grid-independence of the solution
and computational cost. For all cases in the sensitivity study the The implemented modeling approach can be described as a Lo-
mesh topology has been produced with high quality treatment of cally Homogeneous two-phase flow Model (LHM), which is based
near wall region, which allows guaranteeing the correct values of on the assumption that all phases have the same velocity and are
the non-dimensional wall distance (Y+) for the near wall computa- in thermodynamic equilibrium at each computational cell. The
tional cells [24]. This requirement is related to the use of wall func- LHM model solves the Reynolds averaged Navier–Stokes equations
tions to model the near wall flow distribution, which entails for continuity, momentum, energy for each phase, and further
ensuring that the location of the first near wall computational node solves for the turbulent kinetic energy k and the turbulent
1342 D.H. Lobón et al. / Applied Energy 113 (2014) 1338–1348

dissipation rate e, which are necessary to describe the turbulent Table 1


viscosity. Therefore only one set of closure equations is solved Boiling parameters adopted in STAR-CCM+.

for the effective mixture fluid, and an additional transport equation Cqw Cew CHTCArea np Sct
is solved for the volume fraction of one of the phases. 0.008 0.1 5  106
1.73 0.9
The government equation for continuity, momentum and en-
ergy can be expressed as follows [26]:

@q where depending on the sign of the expression (T  Tsat), either con-


v Þ ¼ Sm
þ rðq  ~ ð1Þ
densation or boiling are modeled. The first term, C HTCArea represents
@t
the local heat transfer coefficient between the vapor and the sur-
v
@~ rounding liquid, multiplied by the interfacial area density in order
þ rðq  ~ v Þ ¼ rp þ r  s þ q~g þ ~f
v ~ ð2Þ
@t to represent the cell-wise mass exchange, and T is the temperature
of the mixture. It is clear that in subcooled conditions this conden-
@ 
v  ðqE þ pÞÞ ¼ r  ðkeff rT  Rj hj  J j þ ðs  ~
ðq  EÞ þ r  ð~ v ÞÞ sation term has a significant influence on the local void fraction pre-
@t dictions and was therefore optimized based on available
þ Sh ð3Þ experimental data in the range of operation typical of the DISS facil-
ity. In the boiling sections of the plant, which is the dominant re-
where t represents the time, q is density, ~ v is the velocity vector gion, anyhow this term has much reduced influence.
with respect to 3D coordinate system, Sm is the mass source, p is
 is the stress tensor, ~ Table 1 reports the adopted coefficients which were derived in
the static pressure, s g is the gravitation body
the work of Demarly [25]. In Table 1 Sct represents the turbulent
force, ~f is the external body forces, E is the energy transfer,keff is
Schmidt number (dimensionless) defined as the ratio of viscosity
the effective conductivity, hj is the enthalpy of the species j, ~ J j is
eff is the viscous stress tensor and and mass diffusivity.
the diffusion flux of species j, s
Sh is volumetric heat source.
The k and e equations are not reported here for conciseness 4. Boundary conditions for the CFD modeling of the DSG process
but are fully described in the STAR-CCM + manual [22]. The solu-
tion of this closed set of equations therefore provides first princi- The interconnection tubes in the facility are well insulated and
ple predictions of all flow parameters, including velocities, therefore adiabatic boundary conditions are imposed. For the ab-
temperatures, pressure drop and void fractions. This allows a sorber tubes boundary conditions on the external walls are im-
more general application of the method without the need of posed, which represent the complex heat input from the solar
empirical or semi-empirical models, which is one of the funda- radiation. The heat flux input and effective heat flux input, which
mental advantages of CFD. takes the heat loss in the absorber tubes surface, can be written as:
This approach reduces the computational overhead to a mini-
Q heat flux ¼ C l  DNI  cosðhÞ  Ref  s  f ðuÞ ð7Þ
mum and also extends the robustness of the single-phase CFD
experience. The method is applicable to all flow regimes and in
Q effective heat flux ¼ Q heat flux  Q loss ð8Þ
the extreme of a full resolution of the instantaneous two-phase
flow field could explicitly resolve the slip between phases. where Cl is the ratio between the solar flux intensity achieved after
In order to represent boiling at a wall, again the need for robust- concentration and the incident normal beam insolation for a LS-3
ness and efficiency has driven our selection towards a semi-empir- collector; assuming a non-perfect collector we adopt Cl  45 [28].
ical approach, where rather than mechanistically modeling the DNI (W/m2) represents direct normal irradiance, h angle of inci-
boiling heat partitioning we rely on a traditional correlation as dence, Ref (dimensionless) the mirrors reflectance, and s (dimen-
the base for the estimation of the boiling heat transfer. This model sionless) is the transmittance of the absorber glass cover. Further,
is very general and has been designed to be applicable to a large set f ðuÞ represents the normalized heat-flux distribution at the outer
of conditions. circumference of an absorber tube for parabolic-trough collectors.
The surface boiling heat flux is estimated on the base of the The flux distribution was taken from [14,15] and is displayed in
classic Rohsenow correlation [27]: Fig. 5. The boundary conditions where faithfully represented in
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !3:03 the model as a non-uniform distribution around the outer wall of
gðq1  qv Þ C pl ðT w  T sat Þ the collector pipe. Finally, Q loss represents the thermal losses per
qbw ¼ ll hlat n ð4Þ
r C qw hlat Pr l p unit area in the absorber tubes (W/m2), which were estimated

where ll is the liquid viscosity, C pl is the liquid specific heat capac-


ity, hlat is the latent heat, Prl is the liquid Prandtl number, ql ðv Þ is
the liquid (vapor) density,Tw and Tsat are respectively the wall and
the saturation temperatures, r is the surface tension, g is the gravity
acceleration, np is the Prandtl number exponent and Cqw is a con-
stant dependent on the liquid/surface combination.
The vapor mass generation on the heating surface is given by

C ew qbw
_ ew ¼
m ð5Þ
hlat
where Cew is the heat flux fraction used to create vapor bubbles.
The assembled model must also be extended to include boiling
and condensation in the bulk two-phase flow. This interfacial ex-
change between the vapor and the liquid is simply described as:
C HTCArea ðT  T sat Þ
_ ec ¼
m ð6Þ
hlat Fig. 5. Typical heat flux distribution along the outer surface of an absorber tube.
D.H. Lobón et al. / Applied Energy 113 (2014) 1338–1348 1343

Table 2
Experimental cases data.

Cases Pin (MPa) Tin (°C) Mass flow Direct solar Max heat flux in outer wall Heat loss in Heat loss in Heat loss in
(kg/s) irradiance (W/ influenced by optical preheating boiling collectors superheating
m2) concentration (W/m2) collectors (W/m2)a (W/m2)a collectors (W/m2)a
1 3.42 ± 0.06 205 ± 2 0.47 ± 0.05 822 ± 10 25686 ± 12 1278 ± 15 1828 ± 20 2323 ± 20
2 3.38 ± 0.06 196 ± 2 0.47 ± 0.05 807 ± 10 28184 ± 12 1290 ± 15 1881 ± 20 2356 ± 20
3 6.25 ± 0.06 239 ± 2 0.55 ± 0.05 971 ± 10 30316 ± 12 1974 ± 15 2732 ± 20 3076 ± 20
4 6.23 ± 0.06 235 ± 2 0.50 ± 0.05 850 ± 10 29963 ± 12 1980 ± 15 2765 ± 20 3224 ± 20
5 10.19 ± 0.06 237 ± 2 0.59 ± 0.05 960 ± 10 35219 ± 12 2257 ± 15 3507 ± 20 4606 ± 20
6 10.20 ± 0.06 245 ± 2 0.62 ± 0.05 967 ± 10 34119 ± 12 2123 ± 15 3242 ± 20 4392 ± 20
a
The heat loss is measured in Watt per square meter of absorber pipe surface.

Table 3
Experimental data and simulation results for case #1 and case #2 (p = 3 MPa).

Fluid temperature Fluid temperature STAR- Temperature relative Pressure Pressure STAR-CCM+ Pressure relative Length
experimental (°C) CCM+ (°C) error (%) experimental (MPa) (MPa) error (%) (m)
Case 1
205 ± 2 205.0 0.00 3.42 ± 0.06 3.420 0.00 0
241 ± 2 240.4 0.25 3.41 ± 0.06 3.409 0.03 51
240 ± 2 240.3 0.13 3.40 ± 0.07 3.400 0.00 102
240 ± 2 240.0 0.00 3.39 ± 0.07 3.384 0.18 153
240 ± 2 240.0 0.00 3.35 ± 0.08 3.361 0.33 204
240 ± 2 239.8 0.08 3.33 ± 0.08 3.332 0.06 255
238 ± 2 239.7 0.71 3.30 ± 0.08 3.297 0.09 306
238 ± 2 239.6 0.67 3.26 ± 0.09 3.255 0.15 357
238 ± 2 239.5 0.63 3.22 ± 0.08 3.209 0.34 408
237 ± 2 239.5 1.05 3.17 ± 0.10 3.183 0.41 459
247 ± 2 248.2 0.49 3.14 ± 0.10 3.162 0.70 510
Pressure drop 0.28 ± 0.16 0.258 7.86
Case 2
196 ± 2 196.0 0.00 3.37 ± 0.06 3.370 0.00 0
240 ± 2 238.2 0.75 3.36 ± 0.06 3.363 0.09 51
240 ± 2 238.0 0.83 3.35 ± 0.07 3.355 0.15 102
240 ± 2 238.0 0.83 3.34 ± 0.07 3.341 0.03 153
239 ± 2 237.8 0.50 3.31 ± 0.08 3.321 0.33 204
240 ± 2 237.7 0.96 3.29 ± 0.08 3.295 0.15 255
238 ± 2 237.6 0.17 3.26 ± 0.08 3.263 0.09 306
237 ± 2 237.4 0.17 3.22 ± 0.09 3.225 0.16 357
237 ± 2 237.5 0.21 3.18 ± 0.09 3.183 0.09 408
236 ± 2 237.5 0.64 3.13 ± 0.10 3.160 0.96 459
262 ± 2 260.3 0.65 3.12 ± 0.10 3.139 0.61 510
Pressure drop 0.25 ± 0.16 0.231 7.60

according to the data available for the heat-loss functions DISS-LS-3 of the modeled solar collector loop. Input data and boundary con-
collectors [29] for the three different sections of preheating, boiling ditions considered for the simulation are those detailed in Table 2.
and superheating (see Table 2). Due to the different cooling behavior that the diverse fluid flow
Table 2 reports all the necessary inputs for the operating and patterns existing inside the absorber tubes present and to the non-
boundary conditions used in this work. The data were obtained uniform heat flux profile in their outer surface, the wall absorber
at the DISS facility, and were recorded within the period from temperature distribution will be influenced by these phenomena.
March 2000 until August 2003. Five days have been selected, which Since there is no direct insight into the absorber tube one has to
represent different and wide operational conditions for working distinguish indirectly that the best agreement with boiling model
pressures of 3 MPa, 6 MPa and 10 MPa. is observed when the maximum temperature gradient measured
in the cross section of any absorber pipe and calculated are consis-
tent. Figs. 12–17 show the calculated and measured maximum
5. Results temperature gradients computed and measured along the solar
collector loop for the cases considered in Table 2. These figures also
As previously mentioned, experimental data for three different depict the steam quality profile calculated. No sensors are installed
nominal pressures are available, and are reported in Tables 3–5. in the experimental test facility to measure the steam quality, so
The tables include both experimental measurements and CFD re- this output cannot be validated directly. CFD tool can even get
sults for various locations along the collector loop. Looking at the steam quality, velocity, pressure profiles at any cross section of ab-
temperature column in all these tables of results, it is evident that sorber tube or interconnection piping along the collector loop.
two-phase flow conditions are present in most of the collector Fig. 18 shows the flow pattern at 100 m from the collector loop in-
loop, as indicated by the measured (and modeled) saturated let for case #1, i.e. collector loop’s inlet pressure of 3.42 MPa and
temperature. inlet temperature of 205 °C, and Fig. 19 shows the flow pattern
Figs. 6–11 depict the computer and experimental pressure and at 200 m from the inlet for case #3, i.e. inlet pressure of
temperature values along the collectors from the inlet to the outlet 6.25 MPa and inlet temperature of 239 °C.
1344 D.H. Lobón et al. / Applied Energy 113 (2014) 1338–1348

Table 4
Experimental data and simulation results for case #3 and case #4 (p = 6 MPa).

Fluid temperature Fluid temperature STAR- Temperature relative Pressure Pressure STAR-CCM+ Pressure relative Length
experimental (°C) CCM+ (°C) error (%) experimental (MPa) (MPa) error (%) (m)
Case 3
239 ± 2 239.0 0.00 6.25 ± 0.06 6.250 0.00 0
278 ± 2 277.4 0.22 6.23 ± 0.06 6.231 0.02 51
278 ± 2 277.4 0.22 6.22 ± 0.07 6.227 0.11 102
277 ± 2 277.3 0.11 6.21 ± 0.07 6.221 0.18 153
277 ± 2 277.2 0.07 6.20 ± 0.08 6.212 0.19 204
278 ± 2 277.1 0.32 6.19 ± 0.08 6.199 0.15 255
276 ± 2 277.0 0.36 6.18 ± 0.08 6.179 0.02 306
276 ± 2 277.0 0.36 6.16 ± 0.09 6.156 0.06 357
277 ± 2 276.9 0.04 6.13 ± 0.09 6.133 0.05 408
277 ± 2 276.8 0.07 6.11 ± 0.10 6.121 0.18 459
287 ± 2 285.9 0.38 6.10 ± 0.10 6.111 0.18 510
Pressure drop 0.15 ± 0.16 0.139 7.33
Case 4
236 ± 2 236.0 0.00 6.23 ± 0.06 6.230 0.00 0
277 ± 2 277.2 0.07 6.21 ± 0.06 6.215 0.08 51
277 ± 2 277.1 0.04 6.20 ± 0.07 6.212 0.19 102
276 ± 2 277.1 0.40 6.20 ± 0.07 6.205 0.08 153
277 ± 2 277.0 0.00 6.19 ± 0.08 6.195 0.08 204
277 ± 2 276.9 0.04 6.18 ± 0.08 6.184 0.06 255
276 ± 2 276.8 0.29 6.17 ± 0.08 6.171 0.02 306
276 ± 2 276.7 0.25 6.16 ± 0.09 6.156 0.06 357
277 ± 2 276.5 0.18 6.14 ± 0.09 6.138 0.03 408
291 ± 2 292.2 0.41 6.12 ± 0.10 6.127 0.11 459
316 ± 2 314.4 0.51 6.11 ± 0.10 6.118 0.13 510
Pressure drop 0.12 ± 0.16 0.112 6.67

Table 5
Experimental data and simulation results for case #5 and case #6 (p = 10 MPa).

Fluid temperature Fluid temperature STAR- Temperature relative Pressure Pressure STAR-CCM+ Pressure relative Length
experimental (°C) CCM+ (°C) error (%) experimental (MPa) (MPa) error (%) (m)
Case 5
237 ± 2 237.0 0.00 10.19 ± 0.06 10.190 0.00 0
287 ± 2 288.0 0.35 10.18 ± 0.06 10.182 0.02 51
310 ± 2 311.3 0.42 10.17 ± 0.07 10.171 0.01 102
311 ± 2 311.2 0.06 10.17 ± 0.07 10.169 0.01 153
311 ± 2 311.1 0.03 10.16 ± 0.08 10.165 0.05 204
311 ± 2 311.1 0.03 10.16 ± 0.08 10.161 0.01 255
311 ± 2 311.0 0.00 10.16 ± 0.08 10.155 0.05 306
311 ± 2 310.7 0.10 10.15 ± 0.09 10.148 0.02 357
334 ± 2 330.1 1.17 10.14 ± 0.09 10.141 0.01 408
349 ± 2 344.4 1.32 10.14 ± 0.10 10.138 0.02 459
369 ± 2 365.2 1.03 10.13 ± 0.10 10.135 0.05 510
Pressure drop 0.06 ± 0.16 0.055 8.33
Case 6
246 ± 2 246.0 0.00 10.20 ± 0.06 10.200 0.00 0
292 ± 2 290.0 0.68 10.20 ± 0.06 10.198 0.02 51
310 ± 2 310.2 0.06 10.19 ± 0.07 10.188 0.02 102
312 ± 2 312.1 0.03 10.18 ± 0.07 10.184 0.04 153
311 ± 2 311.4 0.13 10.18 ± 0.08 10.180 0.00 204
311 ± 2 311.1 0.03 10.18 ± 0.08 10.176 0.04 255
311 ± 2 311.0 0.00 10.17 ± 0.08 10.171 0.01 306
311 ± 2 311.6 0.19 10.17 ± 0.09 10.164 0.06 357
325 ± 2 323.2 0.55 10.16 ± 0.09 10.157 0.03 408
338 ± 2 335.8 0.65 10.15 ± 0.10 10.151 0.01 459
357 ± 2 355.3 0.48 10.14 ± 0.10 10.147 0.07 510
Pressure drop 0.06 ± 0.16 0.053 11.67

Looking at the results shown in Fig. 6, which correspond to tests Figs. 12 and 13 display the typical maximum wall temperature
performed at a working pressure of 3 MPa (case #1 and #2), in gen- distribution along the DISS solar field which is influenced by the
eral the agreement between CFD results and measurements are flow pattern profiles inside the absorber tube as mentioned previ-
consistent due to the relative error between both pressure losses ously. Computed maximum temperature gradients and measure-
from collector loop inlet to outlet are around 7.86% for case #1 ments have a mean relative error of around 11%.
and 7.60% for case #2 (see Table 2). Fig. 7 shows that good agree- Fig. 8 shows pressure profile along the collector loop com-
ment is also achieved with regard to fluid temperature along the puted and measured for cases #3 and #4, which correspond to
DISS solar field for these two cases, i.e. CFD computed and mea- tests performed at 6 MPa. For this working pressure, the agree-
sured outlet steam temperature are very consistent as relative ment between CFD and measurements is slightly lower than
error is 0.4% for case #1 and 0.8% for case #2 (Table 2). the corresponding to working pressure of 30 bar, although we
D.H. Lobón et al. / Applied Energy 113 (2014) 1338–1348 1345

Fig. 6. Computed pressure with STAR-CCM+ vs. experimental data (p = 3 MPa).


Fig. 9. Computed fluid temperature with STAR-CCM+ vs. experimental data
(p = 3 MPa).

Fig. 7. Computed fluid temperature with STAR-CCM+ vs. experimental data


(p = 3 MPa). Fig. 10. Computed pressure with STAR-CCM+ vs. experimental data (p = 10 MPa).

Fig. 8. Computed pressure with STAR-CCM+ vs. experimental data (p = 6 MPa).


Fig. 11. Computed fluid temperature with STAR-CCM+ vs. experimental data
(p = 10 MPa).

consider is still quite good as the relative error between CFD pres-
sure loss results and measurements is around 10.67% for case
#3 and 6.67% for case #4 (see Table 4). Fig. 9 displays the fluid along the collector loop length. Mean relative error is in the range
temperature along the collector loop. Agreement between CFD of 12% in these cases.
and measurements continues to be very good as the relative error Finally Figs. 10 and 11 show test results at a working pressure of
between both computed and measured outlet temperature values 10 MPa. With regard to pressure loss, the agreement between CFD
is 0.3% in case #3 and 0.7% in case #4. Figs. 14 and 15 show the and experimental data continues to be quite good as the relative
steam quality calculated and the maximum temperature differ- error is around 8.33 in case #5 and 11.67 in case #6. Same con-
ence both measured and computed in the wall absorber tube clusion is stated for the outlet steam temperature (see Fig. 11) as
1346 D.H. Lobón et al. / Applied Energy 113 (2014) 1338–1348

Fig. 12. Calculated steam quality and calculated maximum wall temperature Fig. 15. Calculated steam quality and calculated maximum wall temperature
difference vs. measurements in case #1 (p = 3 MPa). difference vs. measurements in case #4 (p = 6 MPa).

Fig. 13. Calculated steam quality and computed maximum wall temperature Fig. 16. Calculated steam quality and calculated maximum wall temperature
difference vs. measurements in case #2 (p = 3 MPa). difference vs. measurements in case #5 (p = 10 MPa).

Fig. 17. Calculated steam quality and calculated maximum wall temperature
Fig. 14. Calculated steam quality and calculated maximum wall temperature difference vs. measurements in case #6 (p = 10 MPa).
difference vs. measurements in case #3 (p = 6 MPa).

decreases along the preheating zone, increases during evaporation


the relative error between CFD output and experimental data is 1% and smoothly increases during overheating.
in case #5 and 0.6 in case #6. Figs. 16 and 17 depict the steam qual- The presented comparison evidences that overall, a mean dis-
ity and maximum wall temperature gradient for both cases. crepancy lower than 12% between pressure loss measurements
Looking at the computational results in detail, it seems reasonable and CFD is obtained. Furthermore it should be noticed that several
to conclude that for all experimental cases considered the agreement factors could be limiting the agreement between CFD and mea-
between CFD outputs and measurements are quite satisfactory. With surements; for instance, a measurement error of 0.05 kg/s in the
regard to maximum wall temperature gradients both simulation and mass-flow rate measurements in comparison with the mass-flow
experiments agree that there are three slightly different behavior rate used for validation (0.51–0.73 kg/s), would result in uncertain-
along the solar collector loop, i.e. wall temperature gradient ties between 9.8% and 6.8%. Moreover, during the DISS testing
D.H. Lobón et al. / Applied Energy 113 (2014) 1338–1348 1347

Fig. 18. Cross sectional view of the computed steam quality at 100 m from the start of the first collector (case #1).

Fig. 19. Cross sectional view of the computed steam quality at 200 m from the start of the first collector (case #3).

other factors could have been present; including a non-perfect of a larger number of CPUs and with reasonable computational
sun-tracking of the collector, the cleanliness of the mirrors or the turnaround times.
glass cover of these particular collectors could not be homogenous
during these tests.
The calculations have been performed on a modern Linux Clus- 6. Conclusions
ter, utilizing 16 Intel Xeon E5620, 2.4 GHz CPUs. Fully converged
solutions are achieved in roughly 4 h per collector. The computa- In the present study, experiments and CFD simulations were
tions show an excellent parallel performance indicating that they carried out in order to evaluate several different operating condi-
could efficiently be extended to larger domains taking advantage tions and the simulation predictive accuracy for a solar test facility
1348 D.H. Lobón et al. / Applied Energy 113 (2014) 1338–1348

response. In particular, the thermal hydraulic behavior of absorber [6] European Solar Thermal Electricity Association (ESTELA). Strategic Research
Agenda 2020–2025, December 2012. [chapter 6] <http://www.estelasolar.eu>.
tubes using liquid water and steam as heat transfer fluid has been
[7] Eck M, Steinmann WD, Rheinländer J. Maximum temperature difference in
investigated. horizontal and tilted absorber pipes with direct steam generation. Energy
A fast running, locally homogeneous, boiling model has been 2004;29(5):665–76.
implemented in the commercially available STAR-CCM+ code and [8] Almanza R, Jiménez G, Lentz A, Valdés A, Soria A. DSG under two-phase and
stratified flow in a steel receiver of a parabolic trough collector. J Sol Energy
robustly adopted for the DISS facility simulations without any Eng 2002;124(2):140–4.
parameter tuning. [9] Eck M, Steinmann WD. Modelling and design of direct solar steam generating
According to pressure loss and maximum wall temperature gra- collector fields. J Sol Energy Eng 2005;127(3):371–9.
[10] Forristall R. Heat transfer analysis and modeling of a parabolic trough solar
dients computed, which agrees quite well with measurements, it receiver implemented in engineering equation solver. Technical report NREL/
can be concluded that the adopted boiling model shows excellent TP-550-34169; 2003.
performance. Most importantly its prediction accuracy could be [11] García-Valladares O, Velázquez N. Numerical simulation of parabolic trough
solar collector: improvement using counter flow concentric circular heat
further improved if more experimental data are available to tune exchangers. Int J Heat Mass Transf 2009;52:597–609.
some of the empirical model parameters. [12] Cheng ZD, He YL, Xiao J, Tao YB, Xu RJ. Three-dimensional numerical study of
Results show a consistently good agreement between the CFD heat transfer characteristics in the receiver tube of parabolic trough solar
collector. Int Commun Heat Mass Transf 2010;37(7):782–7.
and the measured temperatures for all the tested input power/inlet [13] Tao YB, He YL. Numerical study on coupled fluid flow and heat transfer process
velocity combinations, with an average error of approximately 3 °C in parabolic trough solar collector tube. Sol Energy 2010;84:1863–72.
(2%). Further, the pressure loss predictions show a maximum er- [14] Muñoz J, Abánades A. Analysis of internal helically finned tubes for parabolic
trough design by CFD tools. Appl Energy 2011;88:1139–49.
ror around 0.02 MPa (10%).
[15] Roldan MI, Valenzuela L, Zarza E. Thermal analysis of solar receiver pipes with
The preliminary results presented in this work for the direct superheated steam. Appl Energy 2013;103:73–84.
steam generation in parabolic-trough solar collectors show that [16] Eck M, Steinmann WD. Dynamic behaviour of the direct solar steam
CFD can very robustly reproduce the plant behavior over a wide generation in parabolic trough collectors: a simulation study. In: Proc. of the
10th SolarPACES int. symp. on solar thermal concentrating technologies,
regime of operational conditions. Further development of the pre- Sydney, Australia; 2000. p. 101–6.
sented approach is recognized to be valuable as it will represent a [17] Hirsch T, Steinmann W, Eck M. Simulation of transient two-phase flow in
fundamental tool for design and optimization of concentrating parabolic trough collectors using Modelica. In: Proceedings of the 4th
international modelica conference, Hamburg, Germany; 2005. p. 403–12.
solar systems using liquid water and steam as heat transfer fluid. [18] Hirsch T, Eck M. Simulation of the start-up procedure of a parabolic trough
collector field with direct solar steam generation. In: Proceedings of the 5th
Acknowledgments international modelica conference, Vienna, Austria; 2006. p. 135–43.
[19] Moya SL, Valenzuela L, Zarza E. Numerical study of the thermal-hydraulic
behavior of water–steam flow in the absorber tube of the DISS system using
This work is supported by the Spanish Ministry of Economy and RELAP. In: Concentrating solar power and chemical energy systems
Competitiveness in the framework of GEDIVA Project (No. (SolarPACES), Granada, Spain; 2011.
[20] Rheinlander J, Eck M. Numerical modelling of pressure losses. research project
ENE2011-24777). The first author wishes to thank the Plataforma on direct solar steam generation (DSG) in parabolic trough collectors. Task
Solar de Almería and the University of Almería for the financial 400-DSG Applied Research. Technical report DISS-SC-MI-03. Plataforma Solar
assistance/scholarship awarded for his PhD research period, during de Almería, Tabernas, Spain; 2002.
[21] Friedel L. Modellgesetz fur den Reibungsdruckverlust in der
which this work was conducted and also thanks Massachusetts
Zweiphasenstromung. VDI-Forschungsheft 1975; 572.
Institute of Technology, Department of Nuclear Science and Engi- [22] STAR-CCM+. <http://www.cd-adapco.com>; 2013.
neering for providing access to the computational facilities used [23] Lai YL. Tool-path generation of planar NURBS curves. Robot Comput Integ
for the present study. Manuf 2010;26(5):471–82.
[24] Salim SM, Cheah SC. Wall y+ strategy for dealing with wall-bounded turbulent
flows. In: Proceedings of the international multiconference of engineers and
References computer scientists (IMECS), Hong Kong; 2009.
[25] Demarly E, Baglietto E. Development of a new approach for full scale Steam
[1] Zarza E, Valenzuela L, León J, Hennecke K, Eck M, Weyers HD, et al. Direct Generator simulations with CFD. Internship final report, INSTN, Saclay, France;
steam generation in parabolic troughs: final results and conclusions of the DISS 2013.
project. Energy 2012;29:635–44. [26] Batchelor GK. An introduction to fluid dynamics. Cambridge University Press;
[2] Eck M, Zarza E, Eickhoff M, Rheinländer J, Valenzuela L. Applied research 1967.
concerning the direct steam generation in parabolic troughs. Sol Energy [27] Rohsenow WM. A method of correlation heat transfer data for surface boiling
2003;74:341–51. of liquid. Trans ASME 1952;74:969–76.
[3] Zarza E, Rojas E, González L, Caballero J, Rueda F. INDITEP: the first pre- [28] Kalogirou SA, Lloyd S, Ward J, Eleftheriou P. Design and performance
commercial DSG solar power plant. Sol Energy 2006;80:1270–6. characteristics of a parabolic-trough solar-collector system. Appl Energy
[4] Price H, Lüpfert E, Kearney D, Zarza E, Cohen G, Gee R, et al. Advanced in 1994;47:341–54.
parabolic trough solar power technology. J Sol Energy Eng [29] Valenzuela L, Zarza E, Berenguel M, Camacho EF. Control scheme for direct
2002;124(2):109–25. steam generation in parabolic troughs. Sol Energy 2005;78(1):301–11.
[5] Montes MJ, Rovira A, Muñoz M, Martínez-Val JM. Performance analysis of an
integrated solar combined cycle using direct steam generation in parabolic
trough collectors. Appl Energy 2011;88:3228–38.

You might also like