International Journal of Hydrogen Energy 53 (2024) 200–207

Contents lists available at ScienceDirect

International Journal of Hydrogen Energy

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

CFD model of refuelling through the entire equipment of a hydrogen

refuelling station
H. Ebne-Abbasi *, D. Makarov, V. Molkov
Hydrogen Safety Engineering and Research Centre (HySAFER), Ulster University, Newtownabbey, BT37 0QB, Northern Ireland, UK

A R T I C L E I N F O A B S T R A C T

Handling Editor: D.-V. No. Vo This paper aims at the development and validation of a computational fluid dynamic (CFD) model for simulations
of the refuelling process through the entire equipment of the hydrogen refuelling station (HRS). The absence of
Keywords: such models hinders the design of inherently safer refuelling protocols for an arbitrary combination of HRS
Hydrogen refuelling station (HRS) equipment, hydrogen storage parameters, and environmental conditions. The CFD model is validated against the
Equipment of HRS
complete process of refuelling lasting 195s in Test No.1 performed by the National Renewable Energy Laboratory
CFD model of refuelling
(NREL). The test equipment includes high-pressure tanks of HRS, pressure control valve (PCV), valves, pipes,
Model validation
Fuelling protocols breakaway, hose, and nozzle all the way up to three onboard tanks. The model accurately reproduced hydrogen
Hydrogen safety temperature and pressure through the entire line of HRS equipment. A standout feature of the CFD model,
distinguishing it from simplified models, is the capability to predict temperature non-uniformity in onboard
tanks, a crucial factor with significant safety implications.

1. Introduction fuelling pressure by 125 % of the nominal working pressure (NWP), and
the state of charge (SoC) to 100 %. Exceeding these temperature
The European Union decided to ban the sale of new fossil-fuelled thresholds may result in the degradation of the tank over time, and even
vehicles by 2035, while the UK is aiming to end the sale of new petrol sudden failure of the hydrogen tank, i.e. tank rupture or hydrogen leak
and diesel cars and vans by 2030 [1]. Hydrogen is an emerging energy [14,15]. A heat exchanger (HE) is used to pre-cool hydrogen and keep its
carrier for use in the transportation system which has demonstrated its temperature in the tank below the 85 ◦ C threshold [16,17]. However,
efficacy as a promising substitute for fossil fuels. Hydrogen dispensing is the pre-cooling complicates the fuelling station design, deteriorates its
an important and complex process that should consider the “customer performance, worsens reliability and increases cost [18]. An ideal
acceptable experience” conditions to encourage the adaptation of the fuelling strategy should be able to direct the hydrogen fuelling process in
technology: acceptable refuelling time, a driving range of more than a reasonable duration while keeping the temperature and pressure in­
500 km, technical and economical parity to fossil fuel technologies side the tank within the limits and it should be able to reach an SoC of 90
[2–4]. The US Department of Energy [5] and the European Clean %–100 % in different ambient conditions [19,20].
Hydrogen Partnership [6] propose to target a 3-4-min duration for The SAE J2601 [10] fuelling protocol is limited to light-duty vehicles
refuelling hydrogen-powered cars. To bear high working pressures, (LDV) that can store a maximum of 10 kg of hydrogen inventory. The
onboard tanks use carbon fibre-reinforced polymer (CFRP). These tanks SAE J2601-2 [21] offers only general safety guidelines for heavy-duty
feature a liner that limits permeation to a regulated level [7]. The vehicles (HDV), not protocols. The same is valid for SAE J2601-3 [22]
integrity of onboard storage tanks is endangered during fuelling when aimed to provide recommendations for fuelling hydrogen-powered in­
they are exposed to high temperatures resulting from gas compression, dustrial trucks like tractors, forklifts and pallet jacks, but, unfortunately,
e.g. from 2 MPa to 87.5 MPa, conversion of kinetic energy into internal does not provide a technical specification of fuelling protocols. The
energy [8], and lower thermal conductivity of the materials [9]. The development of a fuelling protocol is a rigorous and intricate process
fuelling protocols mentioned in SAE J2601 [10], SAE J2579 [11], ISO that demands the utilization of modelling tools and experimental vali­
15869 [12] and regulation GTR#13 [13], limit the allowable bulk dation [23].
hydrogen temperature in the tank between − 40 ◦ C and 85 ◦ C, maximum Conducting numerous experiments to develop fuelling protocols for

* Corresponding author.
E-mail address: ebne_abbasi-h@ulster.ac.uk (H. Ebne-Abbasi).

https://doi.org/10.1016/j.ijhydene.2023.12.056
Received 12 May 2023; Received in revised form 4 September 2023; Accepted 7 December 2023
Available online 14 December 2023
0360-3199/© 2023 The Authors. Published by Elsevier Ltd on behalf of Hydrogen Energy Publications LLC. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
H. Ebne-Abbasi et al. International Journal of Hydrogen Energy 53 (2024) 200–207

arbitrary conditions is costly and impractical. So far, no conclusive and whole duration of the fuelling process of 195 s in Test No.1 performed in
transparent fuelling protocol has been established for HDVs through 2021 in NREL by Kuroki et al. [33].
experimental research. The use of thoroughly validated models of
refuelling through the entire equipment of an HRS can greatly expedite 2. Validation experiment: NREL Test No.1
the development of protocols and reduce associated costs by excluding
numerous expensive and hazardous experiments [24]. To achieve this, The experiment used in this study for the CFD model validation was
contemporary models, i.e., CFD models, must be available. CFD is a performed at the Hydrogen Infrastructure Research Facility of NREL by
contemporary research technique that allows one to gain an under­ Kuroki et al. [33]. Fig. 1 presents the piping and instrumentation dia­
standing of underlying physical phenomena. The advantages of CFD gram (PID) of the fuelling line components. The temperature was
models compared to the reduced models include but are not limited to measured in several locations: in each of two HP tanks (TE1) that were
resolution of temperature non-uniformity in onboard storage versus changed once during fuelling, before and after the PCV (TE2 and TE3),
average (bulk) temperature in the reduced integral models, simulation after the HE (TE4), and in the centre of two of the onboard tanks (TE5
of equipment for which heat and mass transfer correlations are not yet and TE6) with an accuracy of ±1.5 K. The pressure was also measured at
available, etc. However, such CFD models are yet to be developed and the exit of the HP tanks (PT1), before valve 4 (PT2) and at one of the
thoroughly validated. One of the challenges for CFD models that is onboard tanks with an accuracy of ±1.0 MPa [33]. The hydrogen flows
addressed in this work is simulation time [25]. There is no intention to from one of two 300-L HP tanks, through almost 62 m of pipe with 24
incorporate CFD models into an automated fuelling system with a short bends (90-degree), PCV, mass flow rate meter (MFM), HE, breakaway,
response time, but the development with the use of validated CFD model hose, nozzle and 5 other valves to the three 36-L onboard storage tanks.
protocols will close this issue. The length-to-diameter (L/D) ratio of the onboard tanks is reported as
So far, CFD models of hydrogen refuelling are focused only on the 3.4 and is unknown for HP tanks. The dimensions, materials and thermal
thermal behaviour of a hydrogen tank system. In many cases, due to the properties for each piping section are available in the appendix section
high computational cost of CFD simulations, reduced models are of the experimental paper [33].
developed and used. The most of reduced models are focused on the Two tests were carried out by the experimentalists [33]. Test. No.1
thermodynamical state of onboard tanks only [18,23,24,26–30]. In was performed without any leak checks. Test No.2 was done along with
2014 Schneider et al. developed the software H2-Fill [31] which cal­ a leak check at the start-up phase and two leak checks during the main
culates pressure and temperature in the onboard tank only while fuelling phase. The simulation of the start-up phase during the Test. No.2
assuming uniform temperature distribution in the tank [32]. In 2015 Air is reported in our previous study [40]. Test No.1 was selected in this
Liquide’s team published model SOFIL [18,25] that assumes homoge­ study to validate the main fuelling process simulations. The initial
neous gas temperature inside the onboard tank and solves mass and temperature, pressure, and average pressure ramp rate (APRR) for Test
energy conservation equations for the piping only from the dispenser to No.1 are shown in Table 1.
the onboard tank. The Engie Lab CRIGEN developed the modelling tool
Hyfill to simulate the fast filling and emptying of hydrogen tanks and
3. CFD model
predict the final hydrogen bulk temperature in a tank [23]. Engie’s
Hyfill is based on the MATLAB/Simulink software and, being a
3.1. Calculation domain and parameters of HRS components
pseudo-1D model, it assumes uniform hydrogen temperature distribu­
tion in onboard tanks [23]. It is limited to modelling the gas dynamics
Fig. 2 presents the calculation domain geometry that exactly follows
from the dispenser to the onboard tanks and 1D heat transfer between
the experimental PID shown in Fig. 1. The only reported parameter for
the gas and the outside [23]. It is worth noting that these models are
the HP tanks is a volume of 300 L each [33]. The HP tanks are assumed
industrial models and are not publicly available. There is only one model
to be cylinders. The HP tank L/D ratio was assumed to be 3 and its total
to the best knowledge of authors that is publicly available and does
wall thickness was estimated as 33 mm combined of linear and CFRP.
include the entire equipment of an HRS: the reduced model H2FillS
The pipe’s dimensions, i.e., length, internal and external diameter, and
developed in 2021 by Kuroki et al. [33]. This reduced model is based on
thermal properties, i.e., density, thermal conductivity, and specific heat,
the Hydrogen Refuelling Station Dynamic Simulation (HRSDS) software
are the ones that were provided by the experimentalists. The total length
of Kyushu University. It can estimate thermodynamic parameters at
of piping is 62 m which includes 24 bends of 90◦ exaclty the same as the
different HRS locations. However, as an integral model, it still has lim­
experiment (see Fig. 2). Static valves are fully open during refuelling.
itations which cannot be overcome without contemporary CFD models.
Due to the inherently complex geometry of PCV, MFM and valves, they
These include the inability to model HRS’s high pressure (HP) tanks
are modelled as pipes of 10 cm lengths with the equivalent internal
thermal behaviour during fuelling, to assess the non-uniformity of
diameters (ID) calculated using the flow coefficient reported in Ref. [33]
hydrogen temperature in onboard storage tanks and difference in tem­
following the procedure described in Ref. [40].
perature in materials of the onboard tank, etc.
Table 2 shows the IDs of the HRS components.
CFD models can surpass these limitations, provide deeper insight
The thermal mass of the valves, PCV and MFM was not reported in
into underlying physical phenomena, and simulate the combined impact
Ref. [33]. Thus, the external diameter of the “numerical valves” is
of diverse equipment of HRS on heat and mass transfer during fuelling to
assumed to be equal to their upstream pipes. It is worth noting that the
develop efficient refuelling protocols, including those without pre-
thermal conductivity of the breakaway and the nozzle were not
cooling. CFD models are usually associated with high 1computational
costs. As a result, all CFD studies of hydrogen fuelling, to mention a few
[8,16–18,32,34–38], were focused on the thermal behaviour of onboard
tanks only and did not account for processes of heat and mass transfer in
the entire refuelling line equipment at HRS [39]. In 2018 Bourgeois et al.
[39] stressed the importance of considering the entire fuelling line when
studying the fuelling procedure. To develop effective and safe refuelling
protocols, it is crucial to include all components of HRS involved in the
process into consideration.
The current study is the follow-up of the author’s previous work in
which a CFD model has been validated for the startup phase of the
fuelling process [40]. It aims to validate the CFD model against the Fig. 1. PID of the NREL experimental facility.

201
H. Ebne-Abbasi et al. International Journal of Hydrogen Energy 53 (2024) 200–207

Table 1 3.2. Governing equations and numerical details

Experimental conditions of Test No.1 [33].
Tinitial Tinitial (HP Pinitial (upstream Pinitial (downstream APRR The governing equations that are used in the CFD model are the
(ambient) tank) of PCV) of PCV) unsteady three-dimensional Favre-averaged mass, momentum and en­
23.0 ◦ C 17.5 ◦ C 88.0 MPa 6 MPa 19.8 ergy conservation equations [41] respectively as follows:
MPa/min
∂ρ ∂ ( )
+ ρ̃uj = 0, (1)
∂t ∂xj
( )
∂ ∂ ( ) ∂p ∂ ∂̃ui ∂̃uj 2 ∂̃uk
(ρ̃
u )+ ρ̃uj ̃ui = − + ( μ + μt ) + − δ + ρgi ,
∂t i ∂xj ∂xi ∂xj ∂xj ∂xi 3 ∂xk ij
(2)
(( ) )
∂ ̃ ∂ ( )
̃ + p) = ∂ μ cp ∂T̃
(ρE) + u (ρE
̃ λ+ t . (3)
∂t ∂xj j ∂xj Prt ∂xj

where xi,xj,xk are the Cartesian coordinates, ui, uj, uk are the velocity
components, t is the time, p is the pressure, ρ is the density, gi is the
gravity acceleration in i-axis direction, μ and μt are the molecular and
Fig. 2. The calculation domain (top view). turbulent dynamic viscosity respectively, δij is the Kronecker symbol, E is
the total energy, T is the temperature, cp is the specific heat at constant
elucidated experimentally and analytically, but were estimated by the pressure, λ is the thermal conductivity, Prt is the turbulent Prandtl
experimentalists based on the temperature rise of hydrogen flowing number. The symbol “overbar” stands for Reynolds averaged parameters
from the breakaway inlet to the nozzle exit [33]. The length and internal and “tilde” for Favre averaged parameters.
diameter of the breakaway, hose and nozzle are the same as those of the Flow turbulence was modelled using the standard k-ε turbulence
experiment, but their external diameter is adjusted and reported by the model [42]:
experimentalists to match the real weight of each component [33]. The ∂(ρk) ∂ ( ) ∂
(( ) )
μ ∂k
specification of the piping section from the manifold to the onboard + ρ̃uj k = μ+ t + Gk + Gb − ρε, (4)
∂t ∂xj ∂xj σk ∂xj
tanks is not reported in Ref. [33] but it is assumed to be the same as in
the H2Fills software demonstration example which is replicating the ∂(ρε) ∂ ( ) ∂
(( )
μt ∂ε
)
ε ε2
NREL refuelling station [33]. ∂t
+
∂xj
ρ̃uj ε =
∂xj
μ+
σ ε ∂xj
+ C1ε (Gk + C3ε Gb ) − C2ε ρ ,
k k
The geometry of the manifold was not reported by the experimen­
(5)
talists. In the simulations, a manifold was designed to split the flow from
the 5.1 mm upstream pipe into three similarly sized 5.1 mm pipes where k is the turbulent kinetic energy, ε is the dissipation rate of tur­
leading to three onboard tanks. The dimensions and the isometric view bulent kinetic energy, μt = ρcμ k2 /ε, GK = μt S2 , Gb = −
of the designed manifold for simulations are given in Fig. 3. In the lack of
gi (μt /ρPrt )(∂p /∂xi ), cm = 0.09, σk = 1.0, σe = 1.3, C1ε = 1.44, C2ε = 1.92,
information for the manifold, its wall thickness and its material are ⃒ ⃒ √̅̅̅̅̅̅̅̅̅̅̅̅
⃒ 2 2 0.5 ⃒
assumed to be the same as its upstream pipes (2.2 mm), and its total C3ε = tanh⃒̃ uy /(̃ uz ) ⃒, S = 2Sij Sij is the mean rate of strain, Sij =
ux + ̃
( )
length is considered to be 100 mm. 1 ∂ui ∂uj
2 ∂xj + ∂xi .
The computational grid is composed of 207,252 hexahedral control
volumes (CV). The minimum orthogonal quality of the mesh is 0.7 with Standard wall functions [42] have been employed for velocity
an average quality of 0.97. The sensitivity analysis for mesh is presented (no-slip condition), temperature, turbulent kinetic energy and dissipa­
in the next section. tion rate on solid surfaces. Buoyancy effects were also considered in the
model by applying the gravitational acceleration normal to the ground
surface. As the ideal gas model cannot accurately predict pressure

Table 2
Calculated equivalent internal diameter of valves, PCV, HE, and MFM using Cv from Ref. [33].
PID component Valve1 Valve2 MFM PCV Valve3 HE Valve4 Valve5

Flow coefficient, Cv [33] 1.3 0.75 1.0 1.0 0.75 1.0 0.75 1.0
Calculated equivalent ID [mm] 6.5 4.99 5.76 5.76 4.99 5.76 4.99 5.76

Fig. 3. Manifold model: plane view (left) and isometric view (right).

202
H. Ebne-Abbasi et al. International Journal of Hydrogen Energy 53 (2024) 200–207

behaviours at high pressures, the National Institute of Standards and input parameters driving the solution of momentum and energy equa­
Technology (NIST) real gas model was applied. This model uses the NIST tions to avoid computationally expensive resolution of flow through PCV
Thermodynamic and Transport Properties of Refrigerants and Refrig­ and HE. It must be mentioned that the PT2 sensor is located downstream
erant Mixtures Database Version 7.0 (REFPROP v7.0) [43] to evaluate of PCV, after 2.25 m of pipes and HE, in which the experimental length
the transport and thermodynamic properties of fluids. An implicit, of HE is unknown and simulated HE is assumed to be 1 m.
pressure-based solver was used in the simulations, the SIMPLE algorithm In the described CFD model a strategy was implemented to control
was applied for pressure-velocity coupling with the first-order accurate the pressure in the PCV valve via assigning velocity magnitude in the
time stepping. Convective terms were discretised using the first-order PCV control volumes (the momentum conservation equations were not
upwind numerical scheme, and diffusion terms – using the solved in those control volumes). An in-house code was developed to
central-difference scheme. Convergence was controlled in each time step control pressure downstream of PCV (at PT2 location) by changing ve­
using the root mean square (RMS) residual for mass and momentum locity at PCV at each time step (implemented as a user-defined function
conservation equations, which were set to 10− 6. capability of ANSYS Fluent). The procedure is presented in Fig. 5 and
Timestep and mesh sensitivity analysis were done in line with the works as follows:
CFD model evaluation protocol [44]. Timestep equal to TS = 10− 3 s was
applied at the initial stages of the simulation (up to 1 s) and then it was 1. Simulated pressure psim downstream of PCV (at PT2) is acquired from
increased with the simulation advancement to see how it affects the Fluent,
results. The temperature in onboard tank 1 is measured. The tempera­ 2. Pressure psim is compared with the experimental pressure pexp
ture starts to deviate at timestep sizes larger than 0.05 s (Fig. 4-Left). downstream of PCV (PT2),
Fig. 4 (right) shows the sensitivity analysis for the mesh. Three different 3. Dimensionless parameter dpnorm is calculated by Eq. (6), and
meshes were created: one coarse mesh with 104,375 CVs, one with 207, 4. PCV velocity at the previous time step Vi− 1 is multiplied by dpnorm,
252, and a fine one with 540,849 CVs. The temperature in the onboard after which
tank 1 is compared for these cases. The result for coarse mesh deviates 5. The current time PCV velocity value is assigned (Vi-1 = Vi),
from the results of the experiment and the rest of the simulations. For
saving computational resources, the case with 207K CVs is selected for and the whole process is called again at the next time step. The
this study. dimensionless parameter dpnorm is the departure of simulated pressure
The heat conduction through the walls is modelled using the “shell (psim ) from the experimental one (pexp ) introduced as follows:
conduction” capability of ANSYS Fluent. This tool calculates conjugate psim − pexp
heat transfer through the walls in both parallel and normal directions to dpnorm = 1− . (6)
pexp
the walls [45]. The model accounts for the thermal properties, thickness,
and external convective heat transfer. The external convective heat Details of the cooling system (HE) are not available in the experi­
transfer coefficient value of 7 W/m2/K is used in line with the study [9]. mental paper [33]. For the modelling purpose, the HE was represented
All walls are modelled as impermeable and non-slip boundaries. The as a 1.0 m long pipe section with an equivalent internal diameter
initial conditions are those in the experiment (see Table 1): the pressure calculated based on the available HE flow coefficient. A similar to the
is 88 MPa in the HP tanks and down to the PCV, 6 MPa from the PCV and above strategy was implemented here for temperature: temperature in
down to the onboard tanks, initial temperature is 17.5 ◦ C in the HP tank control volumes representing HE was assigned to the temperature
and its walls and 23 ◦ C for the rest of the domain. measured in the experiment. Fig. 6 illustrates how the pressure and
The experimental procedure of switching HP tanks in the simulations temperature control strategies were successful in the reproduction of the
was emulated by stopping the simulation at the specified pressure of experimental data as input for CFD simulations. Fig. 6 (left) compares
63.5 MPa (achieved at time 121.5 s) and re-initialising (“patching”) the the dynamics of the experimental pressure downstream of PCV
temperature and pressure inside the HP tank and the temperature for the (measured in PT2 in PID, Fig. 1) with the simulated pressure at the same
HP tank walls to the initial values. spot location. Fig. 4 (right) compares the experimental temperature after
The simulations were performed using ANSYS Fluent 2020R2 as a the pre-cooler (measurement location TE4 in PID, Fig. 1) with the
CFD engine [45]. The CFD simulations of 195 s of the main fuelling simulated temperature at the same spot. As expected, the simulated
phase take about 2 days on a 32-core AMD Opteron CPU running at 2.3 parameters practically coincide with the experimental ones. It should be
GHz. noted that the pressure regulation technique described in Fig. 4 controls
pressure with an accuracy of ±1 % and any visible deviations are due to
the numerical errors in digitising the experimental data and adding a
3.3. Modelling PCV and HE trendline for the digitized data.

Experimentally measured pressure downstream of PCV (PT2, located

after HE) and temperature immediately after HE (TE4) were simulated
by special means described in this section. They were used as CFD model

Fig. 4. Results for timestep (TS) independency analysis: (Left) pressure in the onboard tank 1 and (Right) temperature in onboard tank 1.

203
H. Ebne-Abbasi et al. International Journal of Hydrogen Energy 53 (2024) 200–207

Fig. 5. Procedure to control pressure downstream of PCV by prescribing flow velocity in “numerical PCV”.

Fig. 6. Experimental and simulated pressure at PT2 (left); experimental and simulated temperature after HE at TE4 (right).

4. Simulation results and discussion

The experimental data of Test No.1 were digitized from the graphs in
the paper [33] to enable comparison with simulation results. Fig. 7 (left)
shows the dynamics of experimental and simulated pressure in onboard
tanks 1 and 2. The experimental pressure is measured at the tank inlet,
and it is assumed with good confidence that the pressure in the tank is
equal to the pressure in the inlet. In the simulations, the pressure was
measured in the same location. The simulated pressure inside the tank
was indeed the same as the pressure in the inlet. The simulated pressure
reproduces the experimental one with a difference of less than ±1 % Fig. 8. Location of temperature sensor (TE5) in onboard tank 1.
across the whole fuelling duration which is well within the pressure
transducer accuracy of ±1 MPa. The difference in simulated pressure in and introduce some more complexities that could affect the accuracy of
Tank 1 and Tank 2 is less than 1 % and is, most likely, due to the dif­ the methodology, at some component interfaces and near walls, the
ference in pressure losses in manifolds of different geometry. behaviour of flow can significantly differ from bulk conditions, which
The location of thermocouples in the experiment (TE5 and TE6) is contributes to uncertainties in thermal rise measurements, and the
referred to as tanks centre positions [33]. In the simulations, the tem­ experiment might not fully take account for heat losses to the environ­
perature was measured on the tank’s axes and is located in the centre of ment, potentially resulting in an underestimation of the true thermal
each tank as is depicted in Fig. 8. conductivity values. Overall, these results validate the CFD model’s
Fig. 7 (right) compares experimental and simulated temperatures in capability to accurately predict hydrogen temperature in the tank during
Tank 1 and Tank 2. The simulated temperatures are in good agreement the fuelling. As the initial and boundary conditions for both tanks are the
with experimental measurements for both tanks (thermocouples TE5 same, the difference between the measured temperature in the two tanks
and TE6) for the whole duration of the fuelling process with a deviation is assumed to be the difference in pressure losses in manifolds of
of less than 5 ◦ C. The small error in the prediction of temperature in the different geometry, as mentioned in the previous section.
onboard tanks might be due to simplifications in modelling valves, PCV Fig. 9 (left) shows pressure and temperature dynamics in the HP
and MFM, as their thermal masses are not taken into account [33] due to tanks of HRS. The simulation results for HP tank pressure and temper­
lack of available experimental data. So, these components do not absorb ature follow closely the experimental results with a maximum deviation
enough heat during fuelling. Besides, the experimentalists estimated the of less than 1 %. The maximum temperature deviation is right before the
thermal conductivity of materials based on the thermal rise of hydrogen change of tanks and is − 3 ◦ C. The results could have been improved if
while passing through the components. This method may not yield the real geometry of the HP tank, its L/D, and the wall thicknesses were
highly accurate results as there might be some uncertainty during known. Besides, the thermal properties of liner and CFRP for the HP
measurements. The uncertainty includes: the turbulence within the tanks are assumed to be similar to the ones used in the onboard tanks,
hydrogen flow can complicate the heat transfer process to the sensors which in real might be different. Fig. 9 (right) shows experimental and

Fig. 7. Experimental and simulated pressures in onboard tanks (left). Experimental (TE5 and TE6) and simulated temperatures in the onboard tanks (right).

204
H. Ebne-Abbasi et al. International Journal of Hydrogen Energy 53 (2024) 200–207

Fig. 9. Experimental and simulated pressure and temperature in the HP tanks (left). Experimental and simulated temperature upstream (TE2) and downstream (TE3)
of the PCV (right).

simulated hydrogen temperatures before and after the PCV. The tem­ in the fuelling line. Also, the pressure drop in the fuelling line is highest
perature increase after passing the PCV is due to the Joule-Thomson in the middle of the fuelling process (see pressure distribution at t =
effect: at the pressure and temperature of filling, hydrogen has a nega­ 124s) when the mass flow rate in the pipeline is large too.
tive Joule-Thomson coefficient causing the temperature increase (TE2 Fig. 11 compares experimental and simulated mass flow rates across
and TE3) while its pressure decreases in the PCV [33]. Because the exact the PCV. The difference, i.e., oscillations, between measured and
locations of thermocouples TE2 and TE3 in the experiment were not simulated mass flow rates can be attributed to the fact that the experi­
reported, the difference in temperature at various locations from PCV mental mass flow rate was obtained from the measurements of the HP
was estimated from the simulations. It was found that the hydrogen tank mass. In simulations, the mass flow rate is calculated in the MFM
temperature difference at distances from 5 cm to 20 cm from PCV is location, although measuring it in the PCV or other locations showed no
within negligible 0.1◦ . This is in line with the comment of experimen­ difference. The simulated mass flow rate closely follows the experi­
talists that temperature difference along the pipe length for less than 10 mental data trend.
cm is negligible [46]. The CFD model reproduces the experimental As mentioned before, the developed CFD modelling approach is
temperature rise as hydrogen passes the PCV well with a maximum capable of simulating the effect of hydrogen inflow jet on temperature
deviation within ±5 ◦ C. The maximum deviation of the temperature distribution and non-uniformity in the tank. This is especially important
happens after the tank change. The reason for that is probably due to the for hydrogen storage tanks with large L/D ratios [38,47]. In this study,
lower simulated temperature at the HP tank, which cooled down the the temperature non-uniformity was measured. The difference between
fuelling line more (compared to the experiment) and lowered the inlet the maximum spatial temperature and measured temperature in the
temperature which ended in a lower outlet temperature in the PCV. Plus, temperature sensor location in tank 1 is shown in Fig. 12 (left). As is
as mentioned before, the thermal mass of valves, PCV and MFM is not shown in Fig. 12 (left), the temperature non-uniformity in this tank can
taken account, which can adversely affect the results. It is likely that if reach up to 7 ◦ C. The situation will drastically change for larger L/D
the HP tanks were modelled using their actual dimensions and materials, tanks, where non-uniformity can increase to unacceptable from the
the results in the PCV could have been improved further. Yet, the sim­ safety point of view 25–30◦ [47]. This is a serious issue for emerging
ulations results for these parameters are well within the acceptable en­ conformable tanks for onboard storage.
gineering accuracy for such a complicated process of heat and mass
transfer. 5. Conclusions
The CFD model enables us to plot and analyse the pressure distri­
bution along the centreline of the pipes. Fig. 10 depicts the distribution The originality of this work is in the development and validation of
of pressure along the centreline of the pipes starting from the HP tank, the first of its kind CFD model which could simulate the entire HRS
across the system pipes and components, and up to the onboard tank 1 at process with all its components. The model includes HP tanks, all the
different times - at the start of fuelling (t = 0 s), before changing the HP pipes and bends, MFM, PCV, HE, breakaway, hose and nozzle up to three
tank (t = 124 s) and at the end of fuelling (t = 195s). As can be seen, the separate onboard tanks. The simulations were performed using ANSYS
main pressure drop happens in PCV in which the pressure is controlled Fluent as a computational engine and the in-house developed code to
control flow through the PCV and temperature after the HE. The model

Fig. 10. Distribution of pressure across the centreline of the pipes at different
times: (1) At the start of fuelling (t = 0s), (2) Before changing the HP tank, and
(3) At the end of fuelling (t = 195s). Fig. 11. Comparison of experimentally measured and simulated mass flow rate.

205
H. Ebne-Abbasi et al. International Journal of Hydrogen Energy 53 (2024) 200–207

References

[1] UK HM Government. The ten point plan for a green industrial revolution. UK dep
business. Energy Ind Strateg 2020:1–32.
[2] Genovese M, Cigolotti V, Jannelli E, Fragiacomo P. Hydrogen refueling process:
theory, modeling, and in-force applications. Energies 2023;16:2890. https://doi.
org/10.3390/en16062890.
[3] Muratori M, Bush B, Hunter C, Melaina M. Modeling hydrogen refueling
infrastructure to support passenger vehicles. Energies 2018;11:1171. https://doi.
org/10.3390/en11051171.
[4] Kang JE, Brown T, Recker WW, Samuelsen GS. Refueling hydrogen fuel cell
Fig. 12. (Left) The difference between maximum spatial temperature and vehicles with 68 proposed refueling stations in California: measuring deviations
measured temperature in the temperature sensor location in tank1 and (Right) from daily travel patterns. Int J Hydrogen Energy 2014;39:3444–9. https://doi.
temperature distribution in the onboard tank1 at the end of refuelling (t org/10.1016/j.ijhydene.2013.10.167.
[5] Us DOE. Target explanation document: onboard hydrogen storage for light-duty
= 195s).
fuel cell vehicles. US Drive 2017;1–29.
[6] Fuel Cells and hydrogen Joint undertaking (FCH JU). Multi-Annual Work Plan
is highly flexible so that every component of the system can be modified 2014-2020; 2014.
[7] Dadashzadeh M, Kashkarov S, Makarov D, Molkov V. Risk assessment methodology
and/or changed to perform parametric studies to get insights into un­ for onboard hydrogen storage. Int J Hydrogen Energy 2018;43:6462–75. https://
derlying physical phenomena of heat and mass transfer. The benefit of doi.org/10.1016/j.ijhydene.2018.01.195.
the CFD model compared to reduced models is the capability to assess [8] De Miguel N, Acosta B, Baraldi D, Melideo R, Ortiz Cebolla R, Moretto P. The role
of initial tank temperature on refuelling of on-board hydrogen tanks. Int J
temperature non-uniformity inside onboard storage tanks to prevent
Hydrogen Energy 2016;41:8606–15. https://doi.org/10.1016/j.
their failure during refuelling. ijhydene.2016.03.158.
The significance of this study is in the affordable time of simulations [9] Simonovski I, Baraldi D, Melideo D, Acosta-Iborra B. Thermal simulations of a
hydrogen storage tank during fast filling. Int J Hydrogen Energy 2015;40:
that can be used to develop fuelling protocols for arbitrary parameters of
12560–71. https://doi.org/10.1016/j.ijhydene.2015.06.114.
HRS equipment, onboard tanks and atmospheric conditions. This CFD [10] Fuel Cell Standards Committee. J2601_202005, fueling protocols for light duty
model simulates 3 min fuelling process in only 2 days on a 32-cores CPU. gaseous hydrogen surface vehicles. Rev 2020. SAE Int 2020:1–2. https://doi.org/
The model can be used for designing fuelling protocols for the whole 10.4271/J2601_202005.
[11] Fuel Cell Standards Committee. J2579_201806 standard for fuel systems in fuel cell
range of hydrogen-powered vehicles for road, rail, marine and aero­ and other hydrogen vehicles. SAE International Technical Paper; 2018. https://doi.
nautical applications. org/10.4271/J2579_201806.
The rigour of this work is in the validation of the CFD model simu­ [12] ISO/TS 15869 Gaseous hydrogen and hydrogen blends — land vehicle fuel tanks.
https://www.iso.org/standard/52871.html. [Accessed 2 April 2023].
lations against the unique refuelling experiment of NREL (Test No.1) [13] United Nations Economic Commission for Europe. Global registry Geneva
through the entire equipment of HRS. The model can help to improve Switzerland ., 13. A 13: GTRN. Global technical regulation on hydrogen and fuel
the accuracy and efficiency of refuelling simulations, leading to safer cell vehicles. 2013.
[14] Sun K, Li Z. Development of emergency response strategies for typical accidents of
and more cost-effective hydrogen fuelling protocols. hydrogen fuel cell electric vehicles. Int J Hydrogen Energy 2021;46:37679–96.
The model will be validated against new experiments of fuelling as https://doi.org/10.1016/j.ijhydene.2021.02.130.
soon as they are made available and used as a contemporary tool for [15] Lazarenko O, Parkhomenko V-P, Sukach R, Bilonozhko B, Kuskovets A. Design
features and hazards of hydrogen fuel cell cars. Fire Saf 2021;37:52–7. https://doi.
fuelling protocol development. The knowledge acquired during this
org/10.32447/20786662.37.2020.08.
study will be applied to LH2 fuelling/bunkering. [16] Zheng J, Guo J, Yang J, Zhao Y, Zhao L, Pan X, Ma J, Zhang L. Experimental and
numerical study on temperature rise within a 70 MPa type III cylinder during fast
refueling. Int J Hydrogen Energy 2013;38:10956–62. https://doi.org/10.1016/j.
Disclaimer ijhydene.2013.02.053.
[17] Melideo D, Baraldi D, Galassi MC, Ortiz Cebolla R, Acosta Iborra B, Moretto P. CFD
Co-funded by the European Union. Views and opinions expressed are model performance benchmark of fast filling simulations of hydrogen tanks with
pre-cooling. Int J Hydrogen Energy 2014;39:4389–95. https://doi.org/10.1016/j.
however those of the author(s) only and do not necessarily reflect those ijhydene.2013.12.196.
of the European Union or the Clean Hydrogen Partnership. Neither the [18] Bourgeois T, Brachmann T, Barth F, Ammouri F, Baraldi D, Melideo D, Acosta-
European Union nor the Clean Hydrogen Partnership can be held Iborra B, Zaepffel D, Saury D, Lemonnier D. Optimization of hydrogen vehicle
refuelling requirements. Int J Hydrogen Energy 2017;42:13789–809. https://doi.
responsible for them.
org/10.1016/j.ijhydene.2017.01.165.
[19] Reddi K, Elgowainy A, Rustagi N, Gupta E. Impact of hydrogen refueling
configurations and market parameters on the refueling cost of hydrogen. Int J
Declaration of competing interest Hydrogen Energy 2017;42:21855–65. https://doi.org/10.1016/j.
ijhydene.2017.05.122.
The authors declare that they have no known competing financial [20] Reddi K, Elgowainy A, Rustagi N, Gupta E. Impact of hydrogen SAE J2601 fueling
methods on fueling time of light-duty fuel cell electric vehicles. Int J Hydrogen
interests or personal relationships that could have appeared to influence Energy 2017;42:16675–85. https://doi.org/10.1016/j.ijhydene.2017.04.233.
the work reported in this paper. [21] SAE International. SAE J2601-2 - fueling protocol for gaseous hydrogen powered
heavy duty vehicles. SAE Int 2014. https://doi.org/10.4271/J2601/2_201409.
[22] SAE International. SAE J2601-3 - Fueling protocol for gaseous hydrogen powered
Acknowledgements industrial trucks. SAE International Technical Paper; 2022. https://doi.org/
10.4271/J2601/3_201306.
This research has received funding from the Engineering and Phys­ [23] Charolais A, Ammouri F, Vyazmina E, Nouvelot Q, Guewouo T, Greisel M,
Gebhard M, Kuroki T, Mathison S. Protocol for heavy duty hydrogen refueling: a
ical Sciences Research Council (EPSRC) of the UK for funding through modeling benchmark. Int Conf Hydrog Saf 2021:1150–60.
the EPSRC Centre for Doctoral Training in Sustainable Hydrogen [24] Molkov V, Dadashzadeh M, Makarov D. Physical model of onboard hydrogen
“SusHy” (Grant EP/S023909/1), UK National Clean Maritime Research storage tank thermal behaviour during fuelling. Int J Hydrogen Energy 2019;44:
4374–84. https://doi.org/10.1016/j.ijhydene.2018.12.115.
Hub funded by EPSRC grant EP/Y024605/1, and the Fuel Cells and [25] Bourgeois T, Ammouri F, Weber M, Knapik C. Evaluating the temperature inside a
Hydrogen 2 Joint Undertaking (now Clean Hydrogen Partnership) under tank during a filling with highly-pressurized gas. Int J Hydrogen Energy 2015;40:
the European Union’s Horizon 2020 research and innovation pro­ 11748–55. https://doi.org/10.1016/j.ijhydene.2015.01.096.
[26] Hosseini M, Dincer I, Naterer GF, Rosen MA. Thermodynamic analysis of filling
gramme through the SH2APED project under grant agreement compressed gaseous hydrogen storage tanks. Int J Hydrogen Energy 2012;37:
No.101007182. Simulations were performed using the Tier 2 High- 5063–71. https://doi.org/10.1016/j.ijhydene.2011.12.047.
Performance Computing resources provided by the Northern Ireland [27] Bai Y, Zhang C, Duan H, Jiang S, Zhou Z, Grouset D, Zhang M, Ye X. Modeling and
optimal control of fast filling process of hydrogen to fuel cell vehicle. J Energy
High-Performance Computing (NI-HPC) facility, funded by the EPSRC
Storage 2021;35:102306. https://doi.org/10.1016/j.est.2021.102306.
(grant EP/T022175/1, https://www.ni-hpc.ac.uk/Kelvin2/(accessed on
1st Aug 2023)).

206
H. Ebne-Abbasi et al. International Journal of Hydrogen Energy 53 (2024) 200–207

[28] Kuroki T, Sakoda N, Shinzato K, Monde M, Takata Y. Dynamic simulation for [38] Ramasamy V, Richardson ES. Thermal response of high-aspect-ratio hydrogen
optimal hydrogen refueling method to Fuel Cell Vehicle tanks. Int J Hydrogen cylinders undergoing fast-filling. Int J Heat Mass Tran 2020;160:120179. https://
Energy 2018;43:5714–21. https://doi.org/10.1016/j.ijhydene.2018.01.111. doi.org/10.1016/j.ijheatmasstransfer.2020.120179.
[29] Johnson T, Bozinoski R, Ye J, Sartor G, Zheng J, Yang J. Thermal model [39] Bourgeois T, Ammouri F, Baraldi D, Moretto P. The temperature evolution in
development and validation for rapid filling of high pressure hydrogen tanks. Int J compressed gas filling processes: a review. Int J Hydrogen Energy 2018;43:
Hydrogen Energy 2015;40:9803–14. https://doi.org/10.1016/j. 2268–92. https://doi.org/10.1016/j.ijhydene.2017.11.068.
ijhydene.2015.05.157. [40] Molkov V, Ebne-Abbasi H, Makarov D. CFD model of refuelling through the entire
[30] Deng S, Li F, Luo H, Yang T, Ye F, Chahine R, Xiao J. Lumped parameter modeling HRS equipment: the start-up phase simulations. Hydrogen 2023;4:585–98. https://
of SAE J2601 hydrogen fueling tests. Sustainability 2023;15:1448. https://doi.org/ doi.org/10.3390/hydrogen4030038.
10.3390/su15021448. [41] Ansys Inc. Ansys Fluent theory guide. 2020.
[31] Schneider J, Meadows G, Mathison SR, Veenstra MJ, Shim J, Immel R, Wistoft- [42] Launder BE, Spalding DB. The numerical computation of turbulent flows. Comput
Ibsen M, Quong S, Greisel M, McGuire T, et al. Validation and sensitivity studies for Methods Appl Mech Eng 1974;3:269–89. https://doi.org/10.1016/0045-7825(74)
SAE J2601. the Light Duty Vehicle Hydrogen Fueling Standard 2014;3. https://doi. 90029-2.
org/10.4271/2014-01-1990. [43] Lemmon EW, Bell IH, Huber ML, McLinden MO. NIST standard reference Database
[32] Carrere P, Lodiar G, Vyazmina E, Ammouri F, Charolais A. CFD simulations of the 23: reference fluid thermodynamic and transport properties-REFPROP, version
refueling of long horizontal H2 tanks. Int Conf Hydrog Saf 2021:1175–84. 10.0. Gaithersburg: National Institute of Standards and Technology, Standard
Edinburgh. Reference Data Program; 2018. https://doi.org/10.18434/T4/1502528.
[33] Kuroki T, Nagasawa K, Peters M, Leighton D, Kurtz J, Sakoda N, Monde M, [44] Tolias IC, Giannissi SG, Venetsanos AG, Keenan J, Shentsov V, Makarov D,
Takata Y. Thermodynamic modeling of hydrogen fueling process from high- Coldrick S, Kotchourko A, Ren K, Jedicke O, et al. Best practice guidelines in
pressure storage tank to vehicle tank. Int J Hydrogen Energy 2021;46:22004. numerical simulations and CFD benchmarking for hydrogen safety applications. Int
https://doi.org/10.1016/j.ijhydene.2021.04.037. –17. J Hydrogen Energy 2019;44:9050–62. https://doi.org/10.1016/j.
[34] Melideo D, Baraldi D. CFD analysis of fast filling strategies for hydrogen tanks and ijhydene.2018.06.005.
their effects on key-parameters. Int J Hydrogen Energy 2015;40:735–45. https:// [45] Ansys Inc. Ansys® fluent 2020 R2 user manual. 2020. https://ansyshelp.ansys.
doi.org/10.1016/j.ijhydene.2014.10.138. com/account/secured?returnurl=/Views/Secured/corp/v202/en/. [Accessed 4
[35] Melideo D, Baraldi D, Acosta-Iborra B, Cebolla RO, Moretto P. Cfd investigation of February 2023].
filling and emptying of hydrogen tanks. Proc. Int. Conf. Hydrog. Safety, Yokohama, [46] Kuroki T, Sakoda N, Shinzato K, Monde M, Takata Y. Prediction of transient
Japan. 19–21 Oct. 2015. n.d. temperature of hydrogen flowing from pre-cooler of refueling station to inlet of
[36] Melideo D, Baraldi D, Galassi MC, Ortiz Cebolla R, Acosta Iborra B, Moretto P. vehicle tank. Int J Hydrogen Energy 2018;43:1846–54. https://doi.org/10.1016/j.
Assessment of a CFD model for the simulation of fast filling of hydrogen tanks with ijhydene.2017.11.033.
pre-cooling. Int Conf Hydrog Saf 2021:1–11. [47] Gebhart TMJ, Spiller M, Çelik H, Dahlmann R, Hopmann C. Experimental analyses
[37] Heitsch M, Baraldi D, Moretto P. Numerical investigations on the fast filling of of the spatial varying temperature development of type-IV hydrogen pressure
hydrogen tanks. Int J Hydrogen Energy 2011;36:2606–12. https://doi.org/ vessels in cyclic tests considering different length to diameter ratios. Int J
10.1016/j.ijhydene.2010.04.134. Hydrogen Energy 2023;48:27304–18. https://doi.org/10.1016/j.
ijhydene.2023.03.353.

207