Vous êtes sur la page 1sur 9

GENETIC OPTIMIZATION OF A PCM ENHANCED STORAGE TANK FOR SOLAR

DOMESTIC HOT WATER SYSTEMS


R. Padovana,* and M. Manzana
a
University of Trieste, v. Valerio 10, 34100, Trieste, Italy
*corresponding author: rpadovan@units.it

ABSTRACT
In this paper Phase Change Materials (PCM)
have been studied in order to provide guidelines
to increase the efficiency of a tank storage
system typically used in Solar Domestic Hot
Water (SDHW) plants. The work stems from
some literature results about the use of PCM in
storage tanks which highlight the scarce impact
of this technology on the overall system.
Some related studies suggest that the
introduction of PCM modules into a storage tank
may be convenient only if the system is
optimized in order to exploit the energy storage
capabilities of the material. In this process
should be taken into account different
parameters such as the complexity of the
system, the PCM characteristics and the
interaction with the climatic conditions.
In this work a PCM improved storage tank,
inserted into a typical Solar Domestic Hot Water
(SDHW) system, has been optimized using
mono and multiobjective genetic algorithms. The
optimization has been carried
with the
modeFRONTIER optimization tool, while the
system plant has been analysed by means of a
modified version of the building energy
simulation code ESP-r.
In parallel with the optimization a sensitivity
analysis has been carried on in order to find out
the relation between the design parameters of
the tank (geometry, temperature of the PCM and
user behaviour) and the performance of the
system.
Thanks to the multiobjective optimization of the
system different solutions have been presented
with different rankings of the optimized
variables.
Keywords: PCM, optimization, solar systems

INTRODUCTION
In the last few decades, the interest aroused by
the Phase Change Materials (PCM) about the
possibility of improving the thermal energy
storage capacity has encompassed several
areas of the building sector.

Inserted in a SDHW system, the PCM has been


often analysed and tested in order to enhance
the solar thermal energy storage: a large
number of numerical implementations and
experimental studies exist in literature.
Mehling et al. [1] implemented a finite difference
scheme and studied the insertion of PCM
modules in a stratified tank in order to enhance
the storage energy density and to compensate
the major heat loss at the upper layers.
Cabeza et al. [2] experimented the same issue
and concluded that the PCM's are a promising
technology since they can provide hot-water for
longer time periods if compared with a traditional
tank . Moreover, the system can be enhanced at
the level to use solar energy supply, without
other heating sources as boilers; finally smaller
PCM tanks could be used to provide the same
hot DHW compared with a traditional larger tank
storage.
In a successive work Ibanez et al. [3] developed
for TRNSYS a new Type consisting in a storage
tank with PCM modules. They tested the new
Type by taking into account the annual solar
fraction achieved by the solar collector, the
quantity of PCM inserted and its phase change
temperature. They declared that to obtain
conclusive results, additional analysis should be
performed, since the results depended also on
factors not considered.
Talmatsky et al. [4] compared PCM and sensible
storage tank performances through MATLAB
simulations. They identified the lack of benefit of
the introduction of PCM modules in the tank due
to two main reasons, the higher mean tank
water temperature with higher heat loss to the
ambient and low solar collector efficiency. A
subsequent work [5] took into account these
results and provided a further statement: the use
of PCM in DHW systems seems not to be
substantially beneficial until a mathematical
optimization at the early stages of the design
process is performed.
Previous works of the same authors [6] focused
on 2 small size tank (73 and 147 l), considering
both traditional and enhanced water tanks with

PCM modules. The results, carried out with


ESP-r, showed that the effectiveness of the
PCM, in therms of saved primary energy, is very
low for the considered configurations. The
same paper underlines that the correct
approach of such a problem should not exclude
an optimization analysis that points out the best
phase change temperature and position of PCM
modules inside the tank.
The purpose of this work is to test different
geometries and characteristics of the tank,
maintaining fixed all the other components of the
SDHW plant, and to find out the best solution(s)
according
to
different
objectives.
The
performance of each plant configuration is
performed by the ESP-r simulation code. The
simulations and the choice of the more
promising configurations are managed by the
modeFRONTIER optimization code.
The main objective of this study is to reduce the
energy taken from the grid to warm the SDHW,
thus to rise up the heat fraction coming out from
the renewable source, by using a prescribed
solar collector. A further target is to figure out if it
is possible to reduce the overall size of the tank.

panel

pcm
DHW

tank

source

Figure 1: Analysed system

PROBLEM DEFINITION
Plant model and controls
The system under investigation is reported in
Figure 1 and represents a typical Solar
Domestic Hot Water System.
The key component of the system is the heat
storage system. It has been modelled using a
stratified tank where the temperature varies in
the vertical direction. In the adopted model the
tank can contain PCM modules with a resolution
in both radial and vertical discretization, in the
latter case the discretization reflects the one
used by the tank. The computational model of
the PCM implemented in ESP-r and the
simplification hypothesis are described in [6].
The code simulates the PCM performance and
its interaction with the tank water by using a
finite volume approach. No convection and
subcooling problems are treated, therefore the

adopted model can only be used for those


materials for which these phenomena can be
neglected. In the case of PCM with the addition
of graphite the solid graphite structure increases
the conduction and blocks the convection terms
as reported in [6].
The tank is collocated in an ambient held at a
constant temperature (20C) and is linked to a
flat plate solar collector via 1 inch diameter
pipes. The climatic data used correspond to the
cities of Rome and Trieste and both are
obtained by the IGDG database. Two
commercial flat panels for a total of 4.4 m 2 of
area have been selected; their
technical
parameters are listed in Table 1.
Table 1: Solar collector parameters.
DESCRIPTION
Scoll
0
1
2
m
test

gly
mcoll

VALUE
4.4
0.819
3.125
0.022
0.200
30
180
50
80

The solar collector pump is controlled by the


temperature difference between the collector at
the middle height of the tank water. The control
activates the flow when this value is higher then
5 K and when it falls below 1 K the pump is
turned of. The parameters are common for
SDHW installations, no more advanced logic of
control have been simulated since the present
PCM enhanced tank has been thought as a
simple replacement in standard installations of
the current commercially available tanks.
The storage tank is also connected to a backup
modulating condensing boiler that maintains the
temperature of the tank when the solar collector
is not sufficient. The boiler characteristics are
summarized in Table 2, which lists also the
normalized boiler efficiency performance curve
coefficients computed as:
b =a0 a 1T r a2f L a 3f LT r

(1)

The boiler pump is activated with a flow rate of


4.5E-05 m3/s when the DHW outlet temperature
from the tank falls below 45C and is
deactivated when the temperature rises over
55C.
To maintain the outlet water at a constant value
of 40C an equipercentual diversion control
valve is inserted at the inlet pipe as represented
in Figure 1 at number 4. To represent the DHW
demand the standard esp-r component n 88
water hourly draw profile has been used. The

load may be referred to a family of 4 people


(approximately 240 liters per day) and the hourly
flow distribution is shown in Figure 2.
Table 2: Boiler parameters.
DESCRIPTION
V
H
Tb,max
fb,min
a1
a2
a3
a0

VALUE
3.73E-003
3.50E+007
65
10
0.353
141.45
-2.333
75.076

Tank parameters setting for the optimization


A standard storage tank, which is commercially
available, has been used as a starting point for
the optimization process.
The geometry of the tank and Pcm modules are
changed during the process; Figure 3 reports
the parameters changed during the optimization.
However some geometrical checks have to be
carried on in order to assure the consistency of
the model.

Figure 2: Draw flow rate.


Figure 3: Optimization variables.

TANK OPTIMIZATION
Two optimizations have been performed for the
tank, the former is a one dimensional
optimization and tries to find an optimal
distribution of the PCM modules with a fixed
geometry tank, the latter is a multi-objective
optimization and looks for an optimal
configuration for the whole tank, modifying two
geometric parameters such as its height and
insulation thickness.
Optimization objectives
For both optimizations the primary energy
absorbed by the condensing boiler has been
selected as a minimization objective. The
primary energy can be obtained from the ESP-r
component n 92, condensing boiler with
modulation, which provides as an output the
consumed gas flow rate, therefore the primary
energy can be computed as:
E p =year VH d

(2)

The primary energy is computed simulating the


plant during a whole year.
For the multiobjective optimization only an
additional objective is the minimization of the
gross tank volume, that is the external volume of
the tank considering also the insulation layer, as
presented in Figure 3.

Some geometrical issues are already taken into


account by the ESP-r model of the storage tank,
as an example the position of the inlets/outlets
are checked against the dimensions of the tank,
so non realistic solutions are avoided. However
the introduction of PCM modules is a daunting
problem since a control on the feasibility of the
position of the modules themselves should be
automated, to allow the optimization process to
variate the input variables. Therefore the PCM
modules are bounded, considering their position
and their dimension, to a physically admissible
configuration. An additional problem is that the
geometric characteristics of the modules are not
independent from each other, as an example a
tall module should not be inserted at a level from
the bottom that lets the top of the module at an
height greater then the tank himself. However
the most difficult geometrical problem is
represented by the number of cylindrical
modules with a specified diameter that can be
contained in the tank or inside the two coils of
the tank, as described in Figure 4. This
geometrical issue has been solved referring to
the packing problem of circles reported in [7].
The dimensions of the tank for the first
optimization, as obtained from a commercial
tank of 300 litres, are: height of 1.54 m, internal
diameter of 0.50 m, 5 cm thick insulation layer

with conductivity 0.04 W/(m K) corresponding to


a mean transmittance U of 0.7 W/(m2 K).
In the second optimization loop also the tank
dimensions are allowed to variate, according to
Table 3 which reports their range of variation.
Table 4 reports the fixed parameters of the tank.

Figure 4: Possible positions of PCM modules.


Table 3: Variables taken for optimization.
DESCRIPTION
ht
sis
nm
hm
Dm
hm,bot
Tpc

1.29
0.01
0
0.05
0.02
0
25

RANGE
1.79
0.075
30
0.3
0.08
1.75
60

Table 4: Optimization cycle PCM tank


parameters.
DESCRIPTION
hin,HX,1
hout,1
nnodes
ts
hin,HX,1
hout,HX,1
hin,HX,2
hout,HX,2
Din,HX,i
Dout,HX,i
Dcoil,HX,i
PHX,i
kHXi
nrad,m
s
l
cs, cl
L
ks,PCM, kl,PCM

VALUE
0.05
ht -0.05
20.00
5.00
1.24
0.88
0.64
0.05
0.027
0.030
0.40
0.05
300
20
880
760
2.100
168.000
2.0

In the optimization cycle the tank inner volume


(insulation layer excluded) has been considered
a variable which changes with the height of the
tank, while maintaining the diameter at a fixed

value of 0.5 m, therefore the inner gross volume


has been considered to variate from 253 to 351
litres. Due to the tank height variation, the DHW
outlet position changes as well since it is located
5 cm beneath the tank top. The coils position is
not affected by the the tank height, since they
are always positioned below the tank minimum
vertical dimension.
All the geometrical feasibility controls, that are
generally managed by the simulation code, are
carried out by the optimization tool and the
solutions which doesn't pass the feasibility test
are rejected.
In all the simulations a fixed value of 20 layers
has been used for tank vertical discretization,
this has been considered a value with sufficient
resolution without increasing excessively the
computational cost [6]. The simulations have
been carried out using 1 minute time step for the
plant and 12 seconds for the tank solution. Due
to the cost of the solutions an explicit approach
has been used for the PCM solution, the time
step is calculated directly by the code taking into
account the stability condition set for the central
cells [6].
Algorithms for optimization
The first optimization adopts a SIMPLEX
algorithm, this is a fast method suitable for
single objective optimization.
The variables in this case are only related with
PCM modules, while the dimensions of the tank
are fixed. The initial population consist in six
designs obtained using the Sobol space filler.
For the multi-object optimization the initial
population have been obtained using a Sobol
space filler with a population of 29 individuals. In
this way the optimization starts with samples
that can fully and uniformly cover the space of
the solutions, in order to allow the optimization
effort to explore all the possibilities.
The target of the optimization cycle is to
minimize the energy consumption and the tank
gross volume. The second parameter is affected
by both the insulation level and the height of the
tank, while the inner volume of the tank changes
with the height of the tank only. For the multiobjective optimization the NSGA-II algorithm has
been selected. This algorithm is very efficient
and allow a continuous variation of input
parameters. The fundamental characteristics of
NSGA-II are the possibility to study multiobjective problems and the robustness in
converging toward the optimal solution. The
optimization is carried for a total of 30
generations.
For the NSGA-II optimizer algorithm a
directional cross-over probability value of 70% a

probability of selection of 5% and a probability


of mutation of 1% have been chosen.

DISCUSSION AND RESULT ANALYSIS


For the single object optimization a true optimal
solution is searched, while for multi-objective
optimization doesn't exist a single optimal value,
but a series of values better than the others.
Single object optimization
Figure 5 reports the history of the optimization
for the Rome climate. The process somewhat
finds an optimal solution, a solution with a lower
primary energy consumed, the solution is
reported in Table 5 for Rome and Trieste.
Although a decrease in primary energy is
obtained, the obtained value (QDHW =1380 kWh)
is very similar to the result obtained without
PCM (QDHW=1400 kWh). The same results have
been obtained for Trieste with an optimized
value of QDHW=2129 kWh against a value
without PCM of QDHW=2164 kWh.

while for others they might be equal; such


solutions form the so called Pareto Frontier.
The results of the optimization are presented in
Figure 6 for Rome and in Figure 7 for Trieste; in
both cases the designs pertaining to the Pareto
frontier are identified in red.

ID 987
ID 1192

Figure 6: Scatter Energy/volume for Rome.

ID 750
ID 448
Figure 5: Optimization history for Rome.
Table 5: Optimal solution for fixed size tank.
VARIABLE
nm
hm
Dm
hm,bot
Tpc
QDHW

ROME
23
0.15
0.04
1.37
50
1380

TRIESTE
19
0.20
0.07
1.1458
48
2129

Multi objective optimization


In the multi-objective optimization multiple two
objectives are searched so the process tries to
minimize the overall tank size and the
consumed primary energy defined in Eqn. 2. In
this case not a single solution exists, rather a set
of solutions that dominate the others, that is if a
and b are two designs, a dominates b if at least
in one purpose design a is better then design b,

Figure 7: Scatter energy/volume for Trieste.


In both figures the designs marked in green
refer to the standard configuration tank without
PCM and a volume of 300 litres. The rectangle,
whose top right vertex is indicated by this
design, encloses all the solutions with lower
primary energy consumption and also a reduced
overall volume. In particular, all the solutions
that are located on the rectangle horizontal side
have, for the same overall volume, different
values of primary energy consumption. Similarly,
all the solutions that are located on the rectangle
right vertical side have, for the same energy
consumption, different volumes.
Inspecting Figures 6 and 7, the designs with the
same gross volume of the original tank, ID 987

and ID 750 are highlighted, for Rome and


Trieste respectively. In a similar manner are
highlighted the designs with the same primary
energy absorbed, ID 1192 for Rome and ID 448
for Trieste. The results are reported in Table 6.
Table 6: Optimized designs

Volg
Voln
VolPCM
QDHW
Tpc

ROME
987
1192
470.7 375.5
275.3 261.7
12.3
10.8
1301
1409
45
48

TRIESTE
750
448
464.4 382.0
269.4 259.4
3.3
8.3
2090
2180
45
47

Table 6 reveals that some enhancements can


be obtained starting from the original tank
designs for both climates. The reduction of
primary energy absorbed is rather low with a
value of 7.66% for Rome and 4.13% for Trieste,
confirming the results obtained with the one
objective approach. The effect on the gross
volume appears to be more pronounced with a
reduction of 18.9% for Rome and 17.5% for
Trieste.
However, the volume variation is due both to the
insulation layer thickness and to the tank height.
To understand the correlation between these
two variables and the quantity of the PCM
inserted it is worth considering the scatter
matrices presented in Figure 8-11 for both
climatic conditions.
The value in the coloured boxes represents an
index of the correlation among two variables, a
value equal to one represent a full positive
correlation between two variables, while a value
of zero means that the two values are
uncorrelated, a minus one value means that the
two variables are purely negatively correlated.
The correlation r can be computed for each
couple of variables as [9]:
r=

N xy ( x )( y )

[N x ( x ) ][ N y ( y ) ]
2

(3)

with N the number of couples and x, y the


considered variables.
Figures 8-11 report on the diagonal the
statistical distribution of the variables among the
designs and on the upper half of the matrix the
two dimensional scatter plot of the couple of
variables reported in the top line and in the right
column. In green are also reported the values
pertaining to the Pareto front.
Figures 8 and 10 show that the volume of PCM
has little effect on the primary energy for DHW
with a correlation value of 0.061 and -0.086
respectively, so it appears that the amount of
PCM has little influence on saved energy. This

is also confirmed by the scatter plots reported in


Figures 8 and 10 corresponding to the row
VolPCM and column QDHW, where the designs,
also the ones pertaining to the Pareto front, are
uniformly dispersed. On the contrary the
parameter with greater influence is the insulation
thickness with correlation values of -0.778 and
-0.844 and a scatter plot well correlated for row
sis and column QDHW.
The inspection of Figures 9 and 11 can give an
insight on the relative importance of the
parameters related to the PCM modules inside
the tank. As can be seen, the parameters which
are more correlated to the Primary Energy are
the height of the module, the diameter and,
above all, the position from bottom of the tank,
with values -0.260 and -0.186 respectively: it
means that putting the module in an higher
position gives rise to less energy required. The
phase change temperature appears to have no
impact for both Trieste and Rome on the
primary energy QDHW having correlations of -0.08
for Rome and 0.043 for Trieste.

CONCLUSION
Two optimizations have been performed for a
Solar Domestic Hot Water System with a tank
containing PCM modules. The first optimization
takes a commercial tank of 300 litres and tries to
minimize the primary energy required for
providing domestic hot water by the insertion of
PCM modules. Although some improvements
have been obtained the impact is very little with
savings of the order of 2% for Rome and Trieste
climate.
In a second optimization also the geometry of
the tank variates along with the insulation width.
Again
the
results
shown
that
some
improvements of the saved energy can be
obtained, but with scars impact (reduction of the
order 5%). The analysis of the correlation
among variables obtained reducing the tank
gross volume showed that the amount of PCM
inside the tank is of little importance for the
reduction of the energy required, a far higher
impact having different parameters such as the
tank insulation.
From the presented results it appears that the
PCM enhanced tanks dont represent a viable
solution for SDHW systems: this can be due to
the great variation of temperatures inside the
tank, that may destroy the benefits of the latent
heat of PCM with a reduced sensible heat
absorbed.

NOMENCLATURE
a
c
D
E

Boiler coefficients
Specific Heat Capacity [J/(kg K)]
Diameter [m]
Energy

f
h
H
L
k
m
n
P
Q
s
S
T
ts
U
V
Vol

Boiler load factor [%]


Height [m]
Gas heating value at STP [J/m3]
Latent heat [J/kg]
Conductivity [W/(m K)]
Mass [kg]
Number, quantity [-]
Pitch of the coils [m]
Primary Energy [kWh]
Thickness [m]
Surface [m2]
Temperature [C]
Time step
U-value [W/(m2 K)]
Boiler full load gas firing rate [m3/s]
Volume [l]

Greeks

0
1
2

Azimuth []
Elevation from horizontal []
Mass fraction [%]
Efficiency [-]
Constant coeff. of efficiency eq. [-]
Linear coeff. of efficiency eq.[W/(m2 C)]
Quadratic coeff. of efficiency eq.
[W/(m2 C2)]
Density [kg/m3]
Time [s]

Subscripts
b
Boiler
bot
Bottom
coll
Collector
g
Gross
gly
Glycol
HX
Heating coil
in
Inlet
is
Insulation
l
Liquid
L
Load
m
PCM module
min
Minimum
max
Maximum
n
Net
out
Outlet
p
Primary
pc
Phase Change
r
Return lower limit modulating vary
rad
Radial

s
t

Solid
Tank

ACKNOWLEDGMENT
This study was supported by Ministero Sviluppo
Economico with project PIACE (grant no.
00024EE01).

REFERENCES
[1] H. Mehling, L.F. Cabeza, S. Hippeli, S.
Hiebler, PCM-module to improve hot water
heat stores with stratification, Renewable
Energy 28 (2003) 699711
[2] L.F. Cabeza, M. Ibez, C. Sol,
Experimentation with a water tank including a
PCM module, Sol Energy Mater Sol Cells 90
(2006) 12731282
[3] M. Ibez, L.F. Cabeza, C. Sol, J. Roca, M.
Nogus, Modelization of a water tank
including a PCM module, Applied Thermal
Engineering 26 (2006) 13281333
[4] E. Talmatsky, A. Kribus, PCM storage for
solar DHW: An unfullled promise?, Solar
Energy 82 (2008) 861869
[5] T. Kousksou, P. Bruel, G. Cherreau, V.
Leoussoff, T. El Rhaki, PCM storage for
solar DHW: From an unfullled promise to a
real benet, Solar Energy 85 (2011) 2033
2040
[6] R. Padovan, M. Manzan - Development of a
stratified tank storage component for ESP-r
with embedded phase change material
modules Special Issue, Proc IMechE Part
A: J Power and Energy 0(0) (2012), 19
[7] R.L. Graham, B.D. Lubachevsky, K.J.
Nurmela, P.R.J. stergrd, Dense packings
of congruent circles in a circle, Discrete
Mathematics 181 (1998) 139-154
[8] VR. Voller, An overview of numerical
methods for solving phase change problems.
In: WJ Minkowycz and EM Sparrow (eds)
Advances in numerical heat transfer, Taylor
& Francis, Washington DC, 1997.
[9] modeFRONTIER 4 user manual, ESTECO,
Trieste, Italy, 2012, http://www.esteco.com

Figure 8: Matrix representation of he relation among optimization variables for Rome.

Figure 9: Matrix representation of he relation among optimization variables for Rome.

Figure 10: Matrix representation of the relation among variables for Trieste.

Figure 11: Matrix representation of the relation among variables for Trieste.

Vous aimerez peut-être aussi