Griffith Research Online

https://research-repository.griffith.edu.au

Heat transfer characteristics for

practical hydrogen pressure vessels
being filled at high pressure
Author
L. Woodfield, Peter, Monde, Masanori, Takano, Toshio

Published
2008

Journal Title
Journal of Thermal Science and Technology

DOI
https://doi.org/10.1299/jtst.3.241

Copyright Statement
Copyright 2008 Japan Society of Mechanical Engineers (JSME). The attached file is reproduced here in
accordance with the copyright policy of the publisher. Please refer to the journal's website for access to
the definitive, published version.

Downloaded from
http://hdl.handle.net/10072/38486
Journal of Thermal Vol. 3, No. 2, 2008

Science and
Technology
Heat Transfer Characteristics for Practical
Hydrogen Pressure Vessels Being Filled at High
Pressure*
Peter L. WOODFIELD **, Masanori MONDE *** and Toshio TAKANO****
** National Institute of Advanced Industrial Science and Technology
744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan
*** Institute of Ocean Energy, Saga University
1 Honjo-machi, Saga, 840-8502, Japan
E-mail:[email protected]
**** JFE Container Company
9-2 Ukisima, Kawasaki-ku, Kawasaki , 210-0862, Japan

Abstract
Experiments have been conducted to measure the rise in temperature of hydrogen
and vessel wall during filling of commercially available, practical tanks to 35 and
70 MPa. Three test vessels with volumes 205, 130 and 39 liters are investigated.
The filling time ranges from 5 to 20 minutes. The heat transfer process is modeled
using a one-dimensional unsteady heat conduction equation for the wall coupled
with a flow and heat balance for the compressed gas. The model requires heat
transfer coefficients between the hydrogen and the wall and the wall and
surrounding air. Values of 500 W/(m2K) during filling, 250 W/(m2K) after filling for
the inside wall and 4.5 W/(m2K) for the outside tank wall are tentatively assumed
based on results from a previous study on a smaller vessel. The measured
temperatures for the hydrogen gas and the wall are in good agreement with the
calculations.

Key words: Hydrogen, Hydrogen Storage, CFRP, Pressure Vessel, Filling Hydrogen

1. Introduction
High pressure gas is presently the most widely used fuel storage mode for hydrogen
vehicles. Typically fiber reinforced plastic (FRP) composite materials are used in order to
reduce the weight of the vessel. While FRP has many desirable properties in relation to
strength and weight, it is necessary to keep the material temperature below 85C to maintain
stringent safety requirements. Since this temperature can be exceeded through the
compression work in the filling process, it is highly desirable to be able to predict the
temperature rise. In the present study, experiments are performed on commercial FRP
hydrogen storage vessels designed for mobile applications. Temperatures and pressures are
measured during charging of three different vessels to 35 MPa or 70 MPa. Temperature
characteristics are also measured for the hydrogen supply vessels.
Previously, the authors developed a thermodynamic model (1) which couples an
energy balance for the gas side with unsteady one-dimensional heat conduction for the wall.
The gas is assumed to be perfectly stirred at all times. The same model is used in the present
study but comparison is made with measurements from actual size vessels. Moreover, the
model is modified to make it applicable to the supply vessels also. Details of the model are
given in the appendix. The main results from the present study, recently presented in a
*Received 18 Sep., 2007 (No. 07-0568) slightly condensed form at a JSME-ASME conference (2), are now expanded and contain
[DOI: 10.1299/jtst.3.241]

241
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

Bank 1
Max: 44 MPa
Max: 1200 L
Thermocouple Flow meter
Thermocouple
Bank 2
Max: 87.5MPa
Max: 660 L
Thermocouple Test vessel
Bank 3
Max: 87.5MPa
Max: 330 L
Thermocouple Controlled environment

Controlled environment

Fig. 1 Pressure vessel filling system (Powertech)

additional discussion concerning the effect of heat transfer in the piping system.

Nomenclature
as : thermal diffusivity of solid
A : inside surface area of pressure vessel
c : specific heat for solid
Cp : constant pressure specific heat for gas
ha : specific internal enthalpy of the supply
l : total thickness of wall
m : accumulated mass of H2 in vessel
M : mass flow rate into vessel
P : gas pressure
t : Time
Tg : gas temperature
Ts : solid temperature
Tw : wall inside surface temperature
u : specific internal energy

Greek
h : convection heat transfer coefficient from gas to wall
e : convection heat transfer coefficient from wall to outside
: thermal conductivity
: Density

2. Filling System Configuration and Test Vessel Geometry

The experiments were conducted at Powertech in Vancouver, Canada. The configuration
of the hydrogen filling system used in the present study is shown in Fig. 1. The supply
system is divided into three banks having vessels initially at 44 MPa and 87.5 MPa. This
configuration is suitable for charging test vessels to 70 MPa. For 35 MPa runs only bank 1
was required. As shown in Fig. 1, gas temperatures are measured inside representative
vessels in the supply banks and at the inlet to the test tank during filling. Tables 1 and 2 give
the required properties of the material and geometry for the test vessels. Unfortunately, FRP
samples from exactly the same vessels used in the present study were not available to
measure the property data. The data shown in Table 2 is that of similar FRP material. The
listed properties were measured at Saga University using the technique of Monde and

242
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

Manual TC1, 2
valve

TC

TD TA

TE TB

TA, TB, TC, TD: ordinary diameter thermocouples

TE: small diameter thermocouple

Fig. 2 Thermocouple positioning in test vessel

Mitsutake (3). Table 3 gives the geometry of the storage banks.

Figure 2 shows the positioning of the thermocouples in the test vessel. The gas
temperature inside the vessel is measured using four thermocouples and the outside wall
temperature using two thermocouples. Thermocouples TA and TE have different
hot-junction diameters and are located at approximately the same position to check the
effect of the response time of the sensor. For the filling rates investigated in the present
study thermocouples TA and TE gave very similar readings indicating that the ordinary
diameter thermocouples were small enough. Pressures and temperatures of the filling
system, surrounding environment and test vessel are monitored simultaneously during each
test run.

Table 1: Specifications for test vessels

Vessel A B C
Pressure / MPa 35 35 70
Volume / m3 0.205 0.039 0.13
Area / m2 2.33 0.685 1.32
FRP thickness / mm 17 11 43.5
Liner thickness / mm 4.25 3.25 5.25

Table 2: Wall property specifications

Material Conductivity/(Wm-1K-1) Diffusivity/(m2s-1) Density/(kgm-3)
FRP 0.55 0.4510-6 1530
Al alloy 180 74.410-6 2700

Table 3: Geometry of storage banks

Bank 1 Bank 2 Bank 3
Cylinder type D E E
Max. No. of cylinders 8 4 2
Volume for single cylinder / m3 0.15 0.165 0.165
Inside area for single cylinder / m2 1.75 1.98 1.98
FRP thickness / mm 15.5 34 34
Al alloy liner thickness / mm 4.25 4.25 4.25
Heat transfer coefficient / Wm-2K-1 250 250 250
-2 -1
External heat transfer coef. / Wm K 4.5 4.5 4.5

243
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

30
During charging
25 of VESSEL A
1 November

Inlet temperature / (C)

11:36
20

15

10

-5
Bank 1 2 3
-10
0 200 400 600 800
time / (s)

Fig. 3 Measured inlet temperature to test vessel during 5 minute fill

Concerning the Max. No. of cylinders listed in Table 3 it is important to note that not
all cylinders in each bank were used for every test. For any particular run the number of
supply cylinders used can be found by considering the mass supplied to the test vessel and
the corresponding pressure drop in the supply bank.
3. Measurement Results for Test Vessels
3.1 Test vessel inlet temperature
Figure 3 shows measured inlet temperatures to the 205L test vessel VESSEL A during
filling to 35 MPa. Clearly the temperature is far from constant and is influenced strongly by
the finite size of the supply vessels and pressure difference between the supply and test
vessels. Hydrogen is supplied first from Bank 1 at time 170 s, then Bank 2 and finally Bank
3. For the example shown in Fig. 3, initially the inlet temperature rises steeply to a peak of
around 24 C. The temperature rise may be attributed to the negative Joule-Thomson
coefficient for hydrogen as the gas passes through the regulator and partly to compression
work on the gas in the supply line. The temperature change is not instantaneous at the
start of the experiment due to the thermal inertia of the piping system. By about time 200 s
(30 s after filling commenced) the temperature starts to fall and by time t = 360 s, a
minimum of -6 C is reached. Note that this is much lower than the ambient temperature of
about 8 C. The falling temperature after t= 200 s may be attributed almost entirely to the
finite size of the supply bank. The gas inside the supply vessel does work pushing the gas
out and thus the temperature in the vessel falls in a manner approximating adiabatic
expansion. When the cold gas from the supply vessel passes through the regulator,
isenthalpic expansion causes the temperature to rise again but the final temperature reaching
the inlet to the test vessel is still quite low as shown in Fig. 3 for 300 < t < 360 s. The same
pattern is repeated for Banks 2 and 3.

3.2 Effect of inlet temperature on gas temperature in test vessel

Figure 4 shows the measured gas temperatures in the test vessel VESSEL A for a five
minute filling time. Experimental data is given by symbols and lines show calculated
results. In particular, the continuous dark line shows the calculated gas temperature using
the present model. Two vertical arrows in Fig.4(b) give the times at which the supply tank
is changed from one bank to another bank. These points correspond to the changes in trend
from a falling temperature to a rising temperature in Fig.3.

244
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

80 80

70 70

60 60

Temperature / (C)
Temperature / (C)
50 50

40 40
Change of banks
30 30

20 TD (meas.) 20 TD (meas.)
TA (meas.) VESSEL A TA (meas.) VESSEL A
TB (meas.) 1 November 11:36 10 TB (meas.) 1 November 11:36
10 TE (meas.)
TE (meas.) Supply temperature Supply temperature
gas (calc.) assumed constant gas (calc.) from TCinlet
0 wall (calc.) 0 wall (calc.)
liner/FRP (calc.) liner/FRP (calc.)
-10 -10
0 250 500 750 0 250 500 750
time / (s) time / (s)
(a) Supply temperature fixed at 12.2C (b) Measured inlet temperature used for supply temperature

Fig. 4 Effect of test vessel inlet temperature on model predictions

In Fig. 4(a) the calculation is based on the assumption of an infinite supply vessel at a
constant temperature of 12.2 C (the initial temperature in the supply vessel for the test).
On the other hand, Fig. 4(b) shows calculations where the inlet temperature was the
measured supply temperature given in Fig. 3. To be consistent, the inlet pressure used for
the calculation shown in Fig. 4(b) was taken to be the pressure of the test vessel rather than
the pressure in the supply bank. This is appropriate because the drop in pressure across the
regulator influences the inlet gas temperature via the Joule-Thomson effect. Clearly the inlet
gas temperature has a significant effect on the model predictions and using the actual
measured inlet temperature as the model input data yielded a better agreement between
calculation and experiment.
There are two reasons why the calculated temperatures shown in Fig. 4(a) are in poorer
agreement with experiment than in Fig. 4(b). Firstly, the actual supply temperature is not a
constant value. This was shown in Fig. 3. However, secondly and more importantly, the
main cause for the higher temperature predictions after about 300 s in Fig. 4(a) is the fact
that the total enthalpy supplied to the vessel in the simulation for Fig. 4(a) is higher than in
the experiment. In other words, the initial supply bank temperature of 12.2C is too high to
be used as an effective average inlet temperature. Reducing the model inlet temperature so
that the average inlet enthalpy over the entire filling time is correct yields predictions for the
gas temperature similar to those shown in Fig. 4(b). This is important to note since using a
constant value may still be useful for predictions where the time history of the inlet
temperature is not accurately known.

3.3 Performance of model for different filling times and different vessels
Figures 5 and 6 show results for different filling times and different vessels. Rather than
using the measured inlet temperature changing with time, an effective constant inlet
temperature was used so that the average inlet enthalpy matched the measured inlet
condition. Figure 5 is for vessels that are filled to 35 MPa and on the whole, the agreement
between model predictions and experiment is quite good.
The general tendency is for the model to over-predict the gas temperature. Note also
that in Figs. 5(b), (d) and Fig. 6 the outside surface temperature of the pressure vessel was
measured at two positions of TC1 and 2 and comparison with model predictions also quite
good in many cases. The poorest agreement between model and experiment for the gas
temperature is in Fig. 6 which shows results for a 130 L vessel charged to 70 MPa. In
contrast to Fig. 5, the gas temperatures in Fig. 6 are around 20 C higher than the measured
temperatures when the vessel was full.

245
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

temperature / (C), pressure / (MPa)

temperature / (C), pressure / (MPa)

90 90
VESSEL A Calculation VESSEL A
80 5 min fill 80 gas 10 min fill
pressure Nov. 1
Nov. 1
70 70 wall
liner/FRP
60 60 outside

50 50
40 40
30 30
Experiment
20 Calculation Experiment 20 TD (gas)
gas TD (gas) TA (gas)
10 pressure TA (gas) 10 TB (gas)
wall TB (gas) TE (gas)
0 liner/FRP TE (gas) 0 Tsuface1 (outside)
outside Tsurface2 (outside)
-10 -10
0 250 500 750 1000 0 250 500 750 1000
time / (s) time / (s)
(a) VESSEL A: 5 minute filling time (b) VESSEL A: 10 minute filling time
temperature / (C), pressure / (MPa)

temperature / (C), pressure / (MPa)

90 90
VESSEL A VESSEL B
80 20 min fill 80 5 min fill
Nov. 1 Nov. 2
70 70
60 60
50 50
40 40
30 30
Experiment Calculation
20 20 TD (gas) gas
Calculation Experiment TA (gas)
gas TD (gas) pressure
10 10 TB (gas.) wall
pressure TA (gas) TE (gas) liner/FRP
wall TB (gas)
0 liner/FRP 0 Tsurface1 (outside) outside
TE (gas) Tsurface2 (outside)
outside
-10 -10
0 500 1000 1500 0 250 500 750 1000
time / (s) time / (s)
(c) VESSEL A: 20 minute filling time (d) VESSEL B: 5 minute filling time

Fig. 5 VESSELS A and B with different filling times

3.4 Gas temperature distribution in the vessels

The thermodynamic model in this study assumes the gas is perfectly stirred and thus the
model predictions are best interpreted as the space-averaged gas temperature. Considering
the measured gas temperatures at different positions, it is clear that there is a consistent
trend in the gas temperature distribution within the vessel. In many cases (e.g. see Figs 4
and 5), thermocouple TB, which is furthest from the inlet, shows the highest temperature
reading. Conversely, thermocouple, TD, which is closest to the inlet shows the lowest
temperature. Three factors may contribute to this trend. Firstly, as may be expected from
Fig. 3, since the inlet temperature is quite low, the temperature closer to the inlet may be
influenced by this cooler gas. Secondly, the flow velocity gradually deceases with an
increase in pressure and hardly reaches the furthest area. Thirdly, the gas flow further from
the inlet may be less turbulent resulting in a poorer heat transfer coupling between the wall
and the gas. Thus enhanced heat transfer near the inlet may be a factor contributing to
reduction of the gas temperature for thermocouple TD. The small differences in temperature
readings between thermocouples TA and TE may be attributed to the slightly different
positions and to the difference in size of the hot-junction diameter.

3.5 The influence of the heat transfer coefficient during filling

In all cases shown in Figs. 4 to 6 it was assumed that the heat transfer coefficient was a
constant value of 500 W/(m2K) during filling. After filling the value was assumed to be 250
W/(m2K) and 4.5 W/(m2K) for the outside wall. These values were appropriate averaged
values recommended by Monde et al.(1). In reality, however, one would expect that if filling
is slower convection heat transfer would diminish. Thus for example, the heat transfer
coefficient for Fig. 5(c) should be smaller than that for Fig. 5(a) or 5(b). This may explain

246
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

temperature / (C), pressure / (MPa)

temperature / (C), pressure / (MPa)

100 100
VESSEL C VESSEL C
90 10 min fill 90 20 min fill
80 Nov. 3 80 Nov. 3

70 70
60 60
50 50
40 40
30 30
20 20 Experiment
Experiment
10 TD (gas) 10 TD (gas)
gas TA (gas) gas TA (gas)
0 pressure TB (gas) 0 pressure TB (gas)
wall TE (gas) wall TE (gas)
-10 liner/FRP Tsurface1 (outside) -10 liner/FRP Tsuface1 (outside)
outside Tsurface2 (outside) outside Tsuface2 (outside)
-20 -20
0 250 500 750 1000 0 250 500 750 1000 1250 1500
time / (s) time / (s)

(a) 10 minute filling time (b) 20 minute filling time

Fig. 6 VESSEL C with different filling times

temperature / (C), pressure / (MPa)

70 100
VESSEL A VESSEL C
10 min fill 90 10 min fill
60 t = 650 s 2
400 W/m K 80 Nov. 3
t = 800 s Nov. 1
50 t = 450 s 70
Temperature /(C)

60
40
t = 250 s 50
30 40
t = 150 s 30
20
20
t = 100 s Experiment
10 10 TD (gas)
gas TA (gas)
t = 70 s 0 pressure TB (gas)
0 wall TE (gas)
Al alloy -10 liner/FRP Tsurface1 (outside)
Liner FRP outside Tsurface2 (outside)
-10 -20
0 5 10 15 20 0 250 500 750 1000
x /(mm) time / (s)
Fig. 7 Calculated temperature Fig. 8 Model liner thickness
distribution in the wall adjusted to correct liner mass (for
case shown in fig. 6(a))

why during the first 400 seconds in Fig. 5(c), the model predictions are around 5 C lower
than measured gas temperatures. In fact, for example, by reducing the heat transfer
coefficient to 400 W/(m2K) (not shown), much better agreement between the model and
experiment during the first 200 seconds for the case in Fig. 5(b) can be achieved. It is worth
mentioning that recently Woodfield et al.(6) measured heat transfer coefficient during
charging and discharging gas into a vessel and proposed a correlation to predict heat
transfer. When charging hydrogen in the actual vessel at the corresponding velocity, the
correlation gives the heat transfer coefficient to vary from about 500 to 300 W/(m2K) with
decreasing its inlet velocity.
It is worth noting finally that the outside heat transfer coefficient has no effect on the
gas temperature during the first few minutes since time is required for the thermal change to
penetrate to outside wall. This is demonstrated in Fig. 7, which shows the calculated
temperature distribution in the wall of the vessel at various times for the case considered in
Fig. 5(b). In Fig. 7 the outside wall temperature (x = 21.25 mm) does not rise much from the
initial value until time t = 150 s.
Another interesting observation from Fig. 7 is that the temperature in the aluminum
alloy liner is almost uniform, in contrast to the FRP temperature distribution, which is
non-linear. This implies that a simple lumped-capacity model for the FRP wall may be
inappropriate. Note also that after the vessel was full (t = 650 s) the temperature of the liner
starts to fall and the temperature distribution in the FRP starts to even out. This is shown by
the dashed line marked t = 800 s.

247
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

3.6 The influence of thermal capacity of the wall

In general it may be concluded that the gas temperature is strongly influenced by the
capacity of the wall to store and transmit heat. One weakness of the present model is that it
assumes a constant thickness of liner and FRP for the entire pressure vessel. In reality
however, both the liner and the FRP are considerably thicker in the dome regions at either
end of the cylinder. Since the model neglects this extra heat capacity, model predictions
should be conservative (i.e. yield an over-prediction of the maximum gas temperature).

temperature / (C), pressure / (MPa)

temperature / (C), pressure / (MPa)
90 90
VESSEL A VESSEL A
80 5 min fill 80 5 min fill
Nov. 1 Nov. 1
70 70
60 60
50 50
40 40
30 30
20 Calculation Experiment 20 Calculation Experiment
gas TD (gas) gas TD (gas)
10 pressure TA (gas)
10 pressure TA (gas)
wall TB (gas) wall TB (gas)
0 liner/FRP 0 liner/FRP TE (gas)
TE (gas)
outside outside
-10 -10
0 250 500 750 1000 0 250 500 750 1000
time / (s) time / (s)
Fig. 9 Model liner thickness adjusted Fig. 10 Model liner inside surface
to correct liner mass (for case shown area adjusted to correct liner mass
in Fig. 5(a)) (for case shown in Fig. 5(a))
temperature / (C), pressure / (MPa)

90
VESSEL A
80 5 min fill
Nov. 1
70
60
50
40
30
20 Calculation Experiment
gas TD (gas)
10 pressure TA (gas)
wall TB (gas)
0 liner/FRP TE (gas)
outside
-10
0 250 500 750 1000
time / (s)
Fig. 11 Cylindrical coordinates used
for heat conduction in the wall

This contributes to explaining the over-prediction of gas temperature for all of the cases
shown in Figs. 5 and 6. In particular, the 70 MPa vessel, VESSEL C has a short aspect ratio
and a very thick wall in the dome section in comparison to the cylindrical section of the
vessel. To see if this extra heat capacity for the liner could explain the large over-prediction
of temperature shown in Fig. 6, the calculation was repeated with the liner thickness
adjusted so that the total mass of the liner was approximately correct. Figure 8 shows the
result of this calculation. Comparing the gas temperature predictions in Figs 6(a) and 8 it is
clear that the extra heat capacity in the dome section of the liner can explain most of the
difference between the model results and experimental data.
For the cases shown in Fig. 5, even without any adjustment for the liner heat
capacity, the agreement between calculation and experiment is quite good. In this case the
cylindrical section of the vessel is large and correcting for the extra heat capacity in the
dome does not have a major effect. Figure 9 shows the results for correcting the liner mass
for the 205 L vessel. The change in gas temperature is not as dramatic as that for Fig. 8, but

248
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

comparing Figs. 9 and 5(a) it is clear that after the vessel is full (t > 450 s), the agreement
between model and experiment is better for Fig. 9. This again confirms the importance of
the total heat capacity of the liner.

3.7 The influence of the inner wall area

Heat transfer from the gas to the wall is also influenced by the inner wall area. The
calculation was repeated with the same wall thickness as given in Table 1 but the wall area
increased so that the heat capacity was close to the actual one including the extra mass in
the dome. In principle this should lead to an underestimate of the gas temperature. Figure
10 shows the result of this calculation. Comparing the gas temperature in Figs. 9 and 10,
one notices that the estimated temperature was improved to be closer to the measured one.
3.8 The influence of the wall curvature
Another point worth noting is that the present model assumes rectangular coordinates
for heat conduction in the wall (see appendix). However, the effect of using cylindrical
coordinates for the wall is not so great for the vessels considered in the present study. Figure
11 shows the same case shown in Fig. 5(a) recalculated with cylindrical coordinates for the
wall. Comparing Figs. 11 and 5(a) it is clear that the effect of the chosen coordinate system
for the wall is small.

4. Prediction of Temperature in Supply Banks

With a small modification as explained in the appendix, the same model used for filling
may be used to predict the temperature during discharge of the supply vessel. The basic
difference is that the enthalpy of the gas leaving the vessel is calculated based on the
instantaneous temperature and pressure in the vessel itself rather than using a constant value

20 20
Bank 1
15 8 cylinders 15
W150H350
Nov. 1
Temperature / (C)
Temperature / (C)

10 10
5 min. test

5 5

0 0

-5 -5
Calculation Experiment Calculation Experiment
-10 gas gas -10 Bank 2 gas gas
wall VESSEL E wall
liner/FRP -15 liner/FRP
-15 adiabatic Nov. 1
outside
outside 5 min. test
expansion
-20 -20
0 500 1000 1500 0 500 1000 1500
time / (s) time / (s)
( ) (a)Bank
B k1 1 (b) Bank 2
20

15
Temperature / (C)

10

-5
Calculation Experiment
-10 gas gas
Bank 3
wall
VESSEL E
-15 liner/FRP
Nov. 1 outside
5 min. test
-20
0 500 1000 1500
time / (s)
(c) Bank 3

Fig. 12 Measurement and prediction of temperatures in supply vessels

249
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

35
theoretical temperature
50 30 for isenthalpic expansion
heat transfer from
0 piping system to
cold gas 25

Temperature / (C)
-50 20

ethalpy / (kJ/kg)
-100 15
temperature at
-150 10 inlet to test
vessel
-200 enthalpy at 5
inlet to test
-250 vessel
0
heat transfer from
-300 warm gas to cold -5
piping system temperature of
-350 -10 supply bank 1
enthalpy of gas
-400 in supply bank 1 -15
-450 -20
100 200 300 400 100 200 300 400
time / (s) time / (s)
(a) Specific enthalpy (b) Temperatures

Fig. 13 Effect of heat transfer in the piping system between supply and test vessel

from the supply temperature and pressure. Figure 12 shows predictions and experimental
data for the 5 minute filling case discussed previously in connection with Figs. 3 and 4.
Note that experimental data was only collected for the gas temperature at a single point in
one vessel from each bank. The dot-dash line in Fig. 12(a) is the estimated gas
temperature change based on adiabatic expansion, namely no heat loss. The heat transfer
coefficient during discharging was assumed to be a constant value of 250 W/(m2K) for all of
the supply banks. Using this value, comparison between the measured gas temperature
(symbols) and calculated gas temperature (dark line) is excellent. It is worth mentioning
that in a short time after discharging, the measured temperature agrees with the dot-dash
line so that it behaves as adiabatic expansion, and after that the temperature is gradually
shifted upward by the influence of the heat transfer from the wall.
It is also interesting to compare Fig. 12 with Fig. 3, which shows the gas temperature at
the inlet to the test vessel for the same run. Note that the temperature in Bank 1 (Fig. 12(a))
becomes much colder (-12 C) than the minimum temperature at the inlet to the test vessel
(-6 C). The temperature rise between the supply vessel and the inlet to the test vessel is due
to the negative Joule-Thomson coefficient for hydrogen as it passes through the regulator.

5. Enthalpy Difference between the Supply Bank and Test Vessel

The enthalpies at the exit of the supply tank and inlet of the test vessel can be calculated
from the pressure and either the estimated or measured temperature. Figure 13(a) shows
the enthalpy change with time during which hydrogen gas is supplied from the bank 1.
Enthalpies are calculated relative to the reference point 25 C, 0.1013 MPa where the
enthalpy is taken to be zero. Initially and even after filling starts, the specific enthalpy of the
gas in the supply bank is higher than the specific enthalpy at the inlet to the vessel. The
reason for this is that heat transfer occurs between the warm gas and the cold pipeline
before the hydrogen enters the vessel. When the gas passes through the regulator the
enthalpy should be almost constant resulting in a temperature rise due to the negative
Joule-Thomson coefficient. Heat is then transferred from the warmer gas to the colder
pipeline so that the enthalpy of the gas at the inlet to the vessel is lower. However as filling
progresses, the temperature of the gas in the supply bank decreases due to the work done
pushing the gas out of the vessel. The temperature of the supply bank is shown by the black
dash-dot line in Fig. 13(b). At some point in time (t 220 s) the temperature in the supply
bank is sufficiently low so that even with the increase in temperature due to the negative
Joule-Thomson coefficient, the gas temperature downstream of the regulator becomes lower
than the temperature of the piping system. After this, heat is transferred from the piping

250
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

system to the colder gas. This explains the observation in Fig. 13(a) that after about t = 220
s, the specific enthalpy of the hydrogen at the inlet to the test vessel is higher than the
specific enthalpy of the hydrogen in the supply bank.
Figure 13(b) compares the measured temperature at the inlet, the measured temperature
in the supply bank and the theoretical temperature for isenthalpic expansion. The
temperature calculated from isenthalpic expansion is always greater than the temperature in
the supply bank because of the negative Joule-Thomson coefficient. The difference between
the blue dashed line and the red solid line in Fig. 13(b) shows the importance of heat
transfer between the gas and the supply piping system.

6. Conclusions
The main conclusions from this fast-fill study conducted at Powertech in Canada are as
follows.
1. The temperature of the hydrogen during charging of three different vessels was
measured and compared well with predictions from the analytical model.
2. The model also successfully predicted the temperatures in the supply vessels during
discharge.
3. Differences between predictions and experiment can be explained by the fact that the
model neglects the extra thermal capacity of the liner and FRP in the dome sections of
the cylinders.
4. Heat transfer in the pipeline has an important influence on the temperature of the gas
entering the test vessel.

Acknowledgement
The authors would like to express appreciation to New Energy and Industrial
Technology Development Organization (NEDO) for financial support to this research.

Appendix: Thermodynamic Model for Charging Hydrogen into Vessel and

Discharging from Vessel
Figure A1 gives an overview of the present model. The supply enthalpy, ha is calculated
either using a fixed supply temperature and pressure (infinite source size) or as a function of
time using the measured inlet temperature and pressure.
Assuming the gas temperature is uniform throughout the vessel, conservation of energy
for the gas side is given by Eq. (A1).

h A(Tw Tg ) + Mha = (mu (P,Tg ))

d
(A1)
dt
h is the convection heat transfer coefficient to the wall, A is the inside surface area of
the vessel, Tw is the wall inside surface temperature and Tg is the gas temperature. The
density required to calculate the mass, m in the vessel is given by the real-gas equation of
state specified in Eq. (A2). The compressibility, Z, is calculated using a polynomial fit to
hydrogen gas data generated the Lee-Kesler method (4). This data agreed well with tabulated
compressibility data (5) for hydrogen in the range of temperatures considered in this study.

(P, Tg ) =
P (A2)
Z (P,Tg )RTg
It is worth noting that for discharge gas, the enthalpy for the discharged gas in Eq.(A1)

251
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

is identical to that in the vessel.

The wall is assumed to behave as a one-dimensional solid, thus conservation of energy
for the wall may be described by unsteady heat conduction, Eq. (A3).

large source
zero contact
Mha resistance
Tw Tn

P,Tg h(Tg-Tw)
Te e(Tn-Te)
H2 gas
perfectly liner FRP
stirred heat loss to and
through wall

convection
heat loss
Fig. A1 Schematic diagram of thermodynamic model

Ts 2T
= as 2s (A3)
t x
The boundary conditions for Eq. (A3) are given by Eqs. (A4) and (A5)

Ts
s |x = 0 = h (Tg Ts |x = 0 ) (A4)
x
T
s s |x = l = e (Ts |x = l Te ) (A5)
x
Here Tw Ts|x=0 and l is the total thickness of the wall. The initial condition is taken to be a
uniform temperature. It should be mentioned that the temperature and heat flux at the
interface between liner and FRP have the identical values for them and the thermal
properties are also used for the liner and FRP, respectively.
The system of equations is closed by specifying the measured pressure in the vessel as a
function of time until the vessel is full. After filling, the mass in the vessel is assumed
constant and the pressure is calculated as the vessel cools. The heat transfer coefficients, h
and e for inside and outside surfaces of the vessel are assumed to have values of h=500
W/(m2K) during filling, h=250 W/(m2K) after full and e=4.5 W/(m2K) for all conditions.
During discharging h is assumed to be 250 W/(m2K), the sign of mass in Eq. (A1) becomes
negative and the enthalpy leaving the vessel, ha is calculated based on the instantaneous
temperature and pressure in the vessel.
The above equations are solved simultaneously using the finite-volume method to
discretize Eq. (A3) and a bisection search algorithm is used to couple this result with energy
conservation for the gas side, Eq. (A1).

References
(1) M. Monde, Y. Mitsutake, P. L. Woodfield, S. Maruyama, Characteristics of Heat Transfer and
Temperature Rise of Hydrogen during Rapid Hydrogen Filling at High Pressure, Heat
Transfer Asian Research, Vol. 36 (2007) pp. 13-27.
(2) P. L. Woodfield, T. Takano, M. Monde, Characteristics of Heat Transfer for Hydrogen and
Wall During Filling Hydrogen into Actual Tank at High Pressure, HT2007-32550,
Proceedings of ASME-JSME Thermal Engineering Summer Heat Transfer Conference,
Vancouver, July 2007.

252
Journal of Thermal Vol. 3, No. 2, 2008
Science and Technology

(3) M. Monde, Y. Mitsutake, A new estimation method of thermal diffusivity using analytical
inverse solution for one-dimensional heat conduction, International Journal of Heat and
Mass Transfer, Vol. 44 (2001) pp. 3169-3177.
(4) B. I. Lee, M. B. Kesler, A generalized thermodynamic correlation based on three-parameter
corresponding states, AIChE Journal, Vol. 21 (1975) pp. 510-527.
(5) R. H. Perry, D. W. Green, J. O. Maloney, Perrys Chemical Engineers Handbook, 7th
Edition, McGraw-Hill, USA, pp. 2-224 2-245, 1997.
(6) P. L. Woodfield, M. Monde, Y. Mitsutake, Measurement of average heat transfer coefficients
in high-pressure vessel during charging with hydrogen, nitrogen or argon gas, J. of
Thermal Science and Engineering, Vol.2, No.2, (2007), pp.180-190.

253