Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

International Journal of Hydrogen Energy 50 (2024) 1395–1405

Contents lists available at ScienceDirect

International Journal of Hydrogen Energy


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

The effect of natural ventilation through roof vents following hydrogen


leaks in confined spaces
Joonsik Kim a, b, Yangkyun Kim a, *, Byungjik Park a, Unggi Yoon a, Chankyu Kang b, **
a
Hydrogen-infra Research Cluster (HiRC), Korea Institute of Civil Engineering and Building Technology, 64 182 Beon-Gil Mado-Myeon, Hwaseong, 18544, South Korea
b
School of Social Safety System Engineering, Research Center for Safety and Health, Hankyong National University, Anseong, 17579, South Korea

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

Handling Editor: Dr S Nanda Hydrogen has a high diffusion rate and a wide range of combustion, making it essential to take safety precautions
against leakage during storage. We investigate the change in hydrogen concentration in confined spaces due to
Keywords: hydrogen leakage through experimental and numerical methods assuming the worst-case scenario. The vent area
Hydrogen consists of 12 %, 24 %, and 36 % of the floor area and is installed in the ceiling of the concrete enclosure. We
Leakage
confirm that the improvement of ventilation performance is insignificant just by increasing the ventilation area
Natural ventilation
above a certain standard during emergency ventilation. Moreover, the improvement of ventilation performance
Forced ventilation
Performance by forced ventilation is more effective as the natural ventilation area is smaller. Still, the effect decreases as the
natural ventilation area becomes larger. In conclusion, we confirm that ventilation efficiency is determined by
the natural ventilation area rather than forced ventilation.

software, assuming a hydrogen refueling station accident. They


compared the modeled results with experimental data and compiled F–N
1. Introduction curves and safety distances for risk assessments in planning hydrogen
refueling station installations in urban areas. In another study, re­
As the world is moving toward carbon neutrality to reduce green­ searchers combined LOPA and RISKCURVES software to identify the risk
house gas emissions, hydrogen energy is gaining attention as a new eco- that hydrogen refueling stations pose in urban areas and to assess how to
friendly energy source to replace fossil fuels, while its global use is minimize damage in the event of a jet fire or explosion caused by gas
gradually increasing [1–3]. Consequently, hydrogen energy technology leakage [12]. However, for hydrogen-related fires and explosions to
is rapidly developing, and related industries are being revitalized [4]. occur, hydrogen leakage must precede and accumulate an appropriate
However, since hydrogen is highly flammable and poses an explosion concentration [13]. In this process, fundamental safety designs must
hazard, it is necessary to consider the factors that hinder the safe use of prevent hydrogen gas buildup leading to an explosive atmosphere if
hydrogen energy along with guidelines promoting the successful accidents are to be avoided [14].
development of the hydrogen economy [5]. Hydrogen gas has a wide Over the past few years, plentiful research has investigated the
combustion range at concentrations of approximately 4 %–75 % in the hazards posed by hydrogen and the role of ventilation in hydrogen leak
air. Its density at standard state conditions is about 14 times lower than accidents. Since conducting empirical experiments with hydrogen is not
that of air [6], and hydrogen explosions include a high-pressure increase feasible due to safety and cost issues, many studies use computational
rate due to the high laminar flow flame propagation rate [7]. It is also fluid dynamic (CFD) simulations or helium gas [15–17]. Yassine et al.
the lightest substance, with atomic number 1, so it disperses quickly in [18] studied hydrogen behavior concerning the shape and height of
the event of an accidental leak and burns easily in the presence of an explosion vents in case of a leak inside a residential garage through CFD
ignition source [8,9]. Because of these dangerous properties, many simulation. Their results highlighted the importance of vent geometry,
scholars are interested in evaluating the risk of hydrogen fires and ex­ with rectangular and square shapes being more effective than round and
plosions [10]. For example, Park et al. [11] examined the risks of triangular shapes. Further, they reported that the closer the vent is to the
compressed hydrogen release, jet fire, and heat flux using HyRAM

* Corresponding author.
** Corresponding author.
E-mail addresses: kkpp9810@kict.re.kr (J. Kim), yangkyunkim@kict.re.kr (Y. Kim), templer83@kict.re.kr (B. Park), yoonunggi@kict.re.kr (U. Yoon), safetyfirst@
hknu.ac.kr (C. Kang).

https://doi.org/10.1016/j.ijhydene.2023.11.048
Received 23 August 2023; Received in revised form 17 October 2023; Accepted 4 November 2023
Available online 15 November 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 license
(http://creativecommons.org/licenses/by/4.0/).
J. Kim et al. International Journal of Hydrogen Energy 50 (2024) 1395–1405

Nomenclature λ ratio of gas-air mixture burning velocity


Gu sonic flow mass flux of unburned gas-air mixture (kg/m3)
LOPA layer of protection analysis Cd vent flow discharge coefficient
CFD computational fluid dynamic Pmax maximum pressure (bar-g)
SLM standard liter per minutes γb ratio of enclosure pressure before ignition (bar-g)
LFL lower flammable limit P0 enclosure pressure prior to ignition (bar-g)
RANS Reynolds averaged navier stokes Pstat nominal vent deployment or burst pressure(bar-g)
SST shear-stress transport Rj production rate of chemical species j
GCI grid convergence index Sj generation term of chemical species j
p order of convergence Jj mass diffusion velocity term of chemical species j
f i average hydrogen volume fraction Dj,m mass diffusion velocity coefficient
γ grid refinement ratio DT,j thermal diffusion coefficient of chemical species j
Av0 minimum vent area(m2) SCt turbulent Schmidt number
As enclosure internal surface area(m2) Kv venting coefficient
Pred maximum pressure developed in vented enclosure(bar-g) V internal volume of enclosure
Su fundamental burning velocity (m/s) Av vent area (m2)
ρu mass density of unburned gas (kg/m3)

ceiling, the more effectively it prevents hydrogen buildup. Lee et al. [19] close to the leakage source for hydrogen removal. In addition, studies on
conducted a CFD simulation of the hydrogen pressure regulation in hydrogen dispersion and removal based on ventilation types were con­
Ulsan City’s hydrogen town in South Korea to investigate small-scale ducted for hydrogen refueling stations [29], nuclear facilities [30], and
leakage scenarios. They suggested an optimal vent location configura­ hydrogen-powered ship fuel cell rooms [31].
tion and compared different ventilation capacities by vent size relative The aforementioned studies confirmed that hydrogen behavior and
to floor area. Additionally, a hydrogen leakage management strategy ventilation performance vary depending on the location and shape of the
was proposed using forced-nitrogen purging. Barley et al. [20] investi­ vent when a hydrogen leak occurs. Additionally, these studies analyzed
gated the relationship between leakage rate, vent design, and hydrogen the ventilation effect in scenarios where natural ventilation is always
concentration through laboratory tests and CFD modeling. Their setup present, with variables comprising only vent shape and location and the
consisted of two vents, a lower and an upper one, while the degree of presence or absence of forced ventilation. However, it is equally
ventilation varied according to the opening of the vents; their experi­ important to consider the possibility of mechanical defects in the safety
mental results were compared to computational data to validate their measures set in advance. From a safety perspective, installing appro­
model. Matsuura [21] also used CFD simulations to study hydrogen priate protective devices and buffers in any system is essential to reduce
dispersion and accumulation in a semi-confined space with natural and the likelihood of an accident or damage caused by one [32]. James
forced ventilation. Various combinations of roof vents and door vents Reason’s Swiss cheese model describes deficiencies in safeguards as the
were considered by changing their size and location, and their effec­ cause of accidents by analogizing them to randomly punctured cheese.
tiveness was compared with that of natural and forced ventilation. Ji In these terms, accidents occur when safeguards deficiencies fall in a
et al. [22] investigated hydrogen leakage and diffusion characteristics straight line. This analogy applies equally well to mitigating damage
by installing a 1.8 m × 2.2 m × 2.9 m room with 0.12 m3 of vents on after an accident has occurred and is compounded to produce results
three sides of the room and ventilation fans on the remaining side. [33,34]. Ultimately, no single safeguard can completely prevent an ac­
Hydrogen concentrations were measured as the vent, ventilation fans, cident, so safeguards should be layered as much as possible to ensure
and supply shut-off valves operated. Moreover, Xie et al. [23] compared safety [35]. Indeed, it is possible scenarios for natural and forced
the effectiveness of different ventilation fan geometries on hydrogen ventilation systems to develop defects and fail to provide adequate
leakage in a garage with vehicles based on CFD simulations. ventilation. Due to the low minimum ignition energy of hydrogen
Studies comparing natural and forced ventilation have also been (mixed with air) (approximately 0.019 mJ), any ventilation method that
actively conducted. Cerchiara et al. [24] evaluated the effect of natural has the potential to release electrical energy, such as forced ventilation
ventilation based on the leakage rate to maintain hydrogen concentra­ with fans, could result in an explosion [19,36].
tion within 2 vol% inside a 25 m3 enclosure, including a fuel cell, and Studies so far have been conducted to create designs in which the
reported an increase in ventilation efficiency with additional forced concentration of hydrogen does not exceed the allowable value in the
ventilation. Prasad [25] compared the natural ventilation and forced event of hydrogen leakage through continuous ventilation. However, as
ventilation effects of compartmentalized spaces during the blowdown described above, consideration should be given to contingency plans to
situation of high-pressure hydrogen tanks, investigating the volume flow safely eliminate leaks in situations where ventilation is restricted (e.g.,
and behavior of released hydrogen. Merilo et al. [26] compared the in the event of ventilation system failure) and concentrations have
natural ventilation and forced ventilation effects in a hydrogen fuel cell already built up. This study aims to experimentally investigate ventila­
vehicle garage concerning hydrogen leakage. They reported that me­ tion effectiveness using natural ventilation by deploying a vent that can
chanical ventilation was efficient in reducing hydrogen concentration in serve as an explosion panel in the event of an emergency; a scenario is
the garage in terms of hydrogen emission rate and ventilation rate. assumed in which a hydrogen leak occurs (due to, for example, irre­
Additionally, Huang et al. [27] compared the ventilation effects of versible damage to a storage container) that results in an accumulating
hydrogen leakage from fuel cell vehicles in an underground parking lot hydrogen concentration. As a result, the ventilation efficiency by size of
using three modes (natural ventilation, forced ventilation after 300 s of the upper vent was compared to derive the appropriate ventilation size
leakage, continuous forced ventilation) and various ventilation layout for accident situations. This provides practical new designs and pa­
configurations. Matsuura et al. [28] investigated the impact of different rameters to prevent and mitigate damage caused by hydrogen leakage
roof ventilation locations and areas on hydrogen accumulation and when designing enclosed spaces where hydrogen can exist. It can also
dispersion in partially open spaces from a natural ventilation perspec­ contribute to the development of a safer hydrogen infrastructure by
tive. They highlighted the effectiveness of selectively opening roof vents proposing new safety measures that are different from previous views.

1396
J. Kim et al. International Journal of Hydrogen Energy 50 (2024) 1395–1405

Fig. 1. (a) Images of the experimental setup, and (b) Schematic of the experimental setup.

rectangular concrete structure with an internal volume of 20.3 m3


Table 1
(Fig. 1). The hydrogen supply system possessed a flow rate range of
Specifications of the experimental setup.
2000–5000 slm (standard liter per minute), a power supply of +15 VDC
Equipment Characteristics to +24 VDC, an operating temperature of 5–50 ◦ C, and a hydrogen
Hydrogen supply system (microbial fuel - Company: MKP, Hwaseong, South concentration sensor (2 V/175 mA); the details are listed in Table 1. A
cell) Korea total of six hydrogen concentration sensors were installed to measure
- Model: TSM-D260 gas concentration at the system’s top, middle, and bottom. The locations
- Flow rate: 2000–5000 slm
- Working pressure: 800 kPa
of these sensors are shown in Table 2.
Hydrogen concentration sensor - Company: SGX Sensortech, Katowice, Three 0.75 m × 1.5 m vents for natural ventilation were installed on
Poland the top of the concrete structure, corresponding to the required mini­
- Model: VQ600 mum vent area (Av0), as shown in Eq. (1)~(2) of NFPA 68 (2013). This
equation is used to calculate explosion protection (deflagration venting),
and this minimum vent area was used as the one vent area in the current
Table 2 experiment. This reason, as mentioned in the introduction, is to consider
Locations of six hydrogen concentration sensors (S1–S6) along three dimensions the vent area that can simultaneously serve as an explosion vent and a
(X, Y, and Z) within a 4.2 × 2.8 × 2.8 m rectangular concrete structure. ventilation vent.
X (m) Y (m) Z (m) [ ( ) ] 1/γb
Pred +1
S1 1.05 1.4 1.95 1− Pmax +1
S2 1.05 1.4 1.1 S u ρu λ
Av0 = As [( )1/γb ] (1)
S3 1.05 1.4 0.25 G u Cd
Pred +1
S4 2.4 1.4 1.95 Pmax +1
− δ
S5 2.4 1.4 1.1
S6 2.4 1.4 0.25
( )γ1
b
PStat +1
P0 +1
− 1
2. Experimental and method section δ= ( )γ1 (2)
b
Pmax +1
− 1
2.1. Experimental setup P0 +1

This experiment aimed to determine the extent to which an internal where Av0 is the vent area calculated in m2, As is the enclosure internal
hydrogen concentration could reach below a reference value in the surface area in m2, Pred is the maximum pressure developed in a vented
event of a hydrogen leak in a confined space that is mitigated via the enclosure during a vented deflagration in barg, Su is the fundamental
deployment of an upper natural vent. The average temperature on the burning velocity of gas-air mixture in m/s, ρu is the mass density of
day of the experiment was 23.5 ◦ C, and the average wind speed was 1.6 unburned gas-air mixture in kg/m3, λ is the ratio of gas-air mixture
m/s. The experimental setup consisted of a 4.2 × 2.8 × 2.8 m burning velocity, Gu is the unburned gas-air mixture sonic flow mass flux

1397
J. Kim et al. International Journal of Hydrogen Energy 50 (2024) 1395–1405

Table 3
Investigated cases in the experiment (Cases E1-E9) and numerical analysis (Cases N1–N4).
Vents in the open position Total natural vent Vent area as a ratio of the Size of leakage hole Leak flow rate Ventilation fan flow rate
area (m2) floor area (%) (inches) (L/min) (m3/min)

Case Center only 1.125 12 1/2 600 –


E1
Case 1/4 200
E2
Case 3/8 400
E3
Case Both sides 2.25 24 1/2 600
E4
Case 1/4 200
E5
Case 3/8 400
E6
Case All 3.375 36 1/2 600
E7
Case 1/4 200
E8
Case 3/8 400
E9
Case Both sides open, and center fan installed 2.25 24 (natural vent) – – 22
N1
Case 44
N2
Case One side was open, a fan was installed on 1.125 12 (natural vent) 22
N3 one vent (center vent closed)
Case 44
N4

in kg/m2-s, Cd is the vent flow discharge coefficient, Pmax is the different parameters. The overall setup for the thirteen cases is sum­
maximum pressure in bar-g, and γb is the ratio of enclosure pressure marized in Table 3.
before ignition in bar-g, P0 is the enclosure pressure prior to ignition in
barg, Pstat is the nominal vent deployment or burst pressure in barg.
The internal space was completely sealed off until the hydrogen 2.2. Conditions for computational analysis
reached the target concentration. The size of each vent comprised 12 %
of the floor area. The experiment was conducted by varying the open 2.2.1. Governing equations
vent area (two and three times the base area, Case E1~E9). Four cases A nonlinear simultaneous differential equation was used to analyze
were selected for the computational analysis. In Cases N1 and N2, both the fluid and energy movements obtained after applying the laws of
vents were opened, a ventilation fan was installed in the center vent, and conservation of mass, momentum, and energy.
the capacity of the ventilation fan was set to different parameters. In As there are variations in velocity and pressure at each point in a
Cases N3 and N4, the center vent was closed, a ventilation fan was general turbulent flow situation, Reynolds Average Navier-Stokes
installed on one vent, and the capacity of the ventilation fan was set to (RANS) equation was utilized to analyze the momentum. In this study,
the k-ω model considering two equations among RANS-based models

Fig. 2. Grid information: (a) overall grid domain, (b) number of volume grid on the vent surface, and (c) size of volume grid on the vent surface.

1398
J. Kim et al. International Journal of Hydrogen Energy 50 (2024) 1395–1405

2.2.2. Boundary conditions and grid for numerical analysis


ANSYS FLUENT, a finite volume method-based commercial CFD
software, was employed to conduct a numerical analysis of hydrogen
leaks. Since the equation of chemical species transport is intended to
analyze the leaks, reactions were not considered; however, the con­
vection current and diffusion of chemical species were considered. Since
hydrogen leaks are a time-dependent transient phenomenon, the first-
order implicit was utilized for this purpose, while the coupled tech­
nique was applied to handle the pressure-velocity coupling of the
pressure-based solver. The initial conditions for the analysis were set as
follows: a hydrogen concentration of 0 %, an oxygen volume fraction
(the atmospheric pressure condition) of 21 % and a nitrogen volume
fraction of 79 %, and the size of the time step was set to 0.2 s, the same as
the size of the hydrogen concentration sensor used in the experiment.
An enclosure structure was modeled to analyze the hydrogen leaks
numerically, while a grid was created for the CFD analysis. The structure
was built with the same dimensions as in Fig. 1. Since the structure and
conditions for the analysis were symmetrical, half of the structure was
reflected in the analysis grid. Approximately 440,000 grids were utilized
for the analysis, while tetrahedral lattices were created for utilization.
Here, maximum size of grid in vent is about 9.8e-5 m3, while minimum
Fig. 3. Hydrogen concentration along the centerline of computational domain size of it is about 3.7e-6 m3. Maximum number of volume grid attached
with deferent number of grids. on the surface of vent is about 302, whereas minimum number is about
94. Overall grid information is shown in Fig. 2. As for the numerical
was employed; the Shear-Stress Transport (SST) k-ω model, which has analysis, at the start of the analysis, hydrogen gas flowed in as the ceiling
combined strengths from the k-ω model with an advantage of analyzing vent was blocked entirely for 283 s (when the leak was completed); the
near walls, and the k-ε model with strength in free flow, was employed. inflow of hydrogen gas was completed from 283 s to 900 s; the top vent
As for the hydrogen leaks, chemical species should be additionally was opened to allow ventilation to the exterior; by analyzing this situ­
considered, as well as general flow. The equation of chemical species ation, hydrogen concentrations over time at the exact location as in the
transport can be found in Eqs. (3) and (4). experiment were compared.
∂( ) ( )
ρYj +∇• ρuYj = − ∇ • Jj + Rj + Sj (3) 2.2.3. Grid convergence analysis
∂t
Grid independence analysis was conducted using GCI (Grid
( )
μt ∇T Convergence Index) as has been done by several previous researchers
Jj = − ρDj,m + ∇Yj − DT,j (4) [37–39]. To derive the GCI for the grid ratio of r = 1.5, analysis was
SCt T
performed on 440,000 Medium grids, 293,000 Coarse grids, and 660,
where, Rj, and Sj refer to the production rate and generation term of 000 Fine grids. Analysis was performed under the same conditions for
chemical species j, respectively; Jj is the mass diffusion velocity term of each grid, and the average hydrogen volume fraction data was acquired
chemical species j; Dj,m and DT,j refer to the mass diffusion velocity co­ at 21 points at a height of 2.1 m (Fig. 3), and the order of convergence, p,
efficient and the thermal diffusion coefficient of the chemical species j in was obtained through the average value as Eq. (5).
a mixture, respectively; and Sct represents the turbulent Schmidt
number.

Fig. 4. Hydrogen concentration change over time for a vent area corresponding to 12 % of the floor are: (a) Case E1, (b) Case E2, and (c) Case E3.

1399
J. Kim et al. International Journal of Hydrogen Energy 50 (2024) 1395–1405

Table 4
Time to reach the target concentration by case (s).
12 % 24 % 36 %

Case E1 Case E2 Case E3 Case E4 Case E5 Case E6 Case E7 Case E8 Case E9

4% 108.8 297.6 228.4 43.2 83.8 65.6 47.8 58.8 42


1% 266.6 510.4 434 139.8 156.6 157.6 161.2 134.4 141.2

Fig. 5. Hydrogen concentration change over time for a vent area corresponding to 24 % of the floor size: (a) Case E4, (b) Case E5, and (c) Case E6.
( ) ( )
⃒ ⃒
ln f3 − f2/f − f ln 1.36724 − 1.36676/1.36676 − 1.34769 ⃒1.36724 − 1.36676⃒/
2 1 = ⃒⃒ ⃒ ( −
⃒ 1.5 9.069
)
− 1 ×100% = − 0.036% (6)
p= = = − 9.069 1.36676
ln r ln 1.5
(5) ⃒ ⃒
⃒f2 − f1 ⃒⃒/
GCI1,2 = ⃒⃒ ⃒ (rp − 1) ×100%
Where f 1, f 2, f 3 is average hydrogen volume fraction by location (coarse, f1
medium, fine) respectively. Through the r and p value obtained in
⃒ ⃒
equation (5), the GCI was obtained as follows Eqs. (6) and (7). ⃒1.36676 − 1.34769⃒/
= ⃒⃒ ⃒ ( −
⃒ 1.5 9.069
)
− 1 ×100% = − 1.452% (7)
⃒ ⃒ 1.34769
⃒f3 − f2 ⃒/
GCI2,3 = ⃒⃒ ⃒ p ×100%
f2 ⃒ (r − 1) Grid dependence shown in equation (8) is 0.986, which is approxi­
mately 1. Therefore, it was concluded that results obtained with current
grid structure are within the asymptotic range of convergence.

Fig. 6. Hydrogen concentration change over time for a vent area corresponding to 36 % of the floor size: (a) Case 3, (b) Case 6, and (c) Case 9.

1400
J. Kim et al. International Journal of Hydrogen Energy 50 (2024) 1395–1405

The change in hydrogen concentration for a vent size corresponding


to 12 % of the floor area (the smallest area used in our tests) is shown in
Fig. 4. The times to reach a concentration of LFL were 108.8 s, 297.6 s,
and 228.4 s across three cases, while the times to reach a concentration
of 1 vol% were 266.6 s, 510.4 s, and 434 s (Table 4). This showed a large
variation. The time to mitigate hazardous gas levels in natural ventila­
tion is highly influenced by environmental factors such as external cli­
matic conditions, temperature, humidity, and wind direction. In this
case, the ventilation time cannot be predicted consistently, and the time
to reach LFL and 1 vol% is quite lengthy. This may be because if the area
of the vent is small, it causes ventilation concentrated at a specific point,
and the flow is formed in only one direction, resulting in uneven
ventilation. Thus, we assessed that this vent size would not have
adequate ventilation performance in an emergency.
A plot of the indoor hydrogen concentration changes for a vent size
corresponding to 24 % of the floor area is shown in Fig. 5. In this sce­
nario, the times to reach a concentration of LFL were 43.2, 83.8, and
65.6 s across three cases, and the times to reach a concentration of 1 vol
Fig. 7. Time to target concentration based on the venting coefficient (Kv). % were 139.8 s, 156.6 s, and 157.6 s (Table 4). For Case E4, the venti­
lation time was faster or similar to those of Cases E7, E8, and E9, which
GCI2,3 had larger vent areas. This may be due to the difficulty in controlling the
= 0.986 ≅ 1 (8)
rp × GCI1,2 ambient environment of the test site, which was located outdoors and
exposed to the effect of sudden wind or airflow changes. However,
3. Results and discussion compared to our previous results for the 12 % vent size, the ventilation
time here was significantly reduced with a slight deviation. This is
3.1. Characteristics of natural ventilation because a larger vent area has a more significant effect on constant
fluidity (caused by principles such as the buoyant release of hydrogen
Venting experiments were conducted with different vent areas to and the difference in air pressure between the interior and exterior
measure the time required to reduce explosive hydrogen concentrations. environment), increasing the flow at the roof vent rather than the effect
The vent areas deployed were 1.125 m2, 2.25 m2, and 3.375 m2, which caused by the external environment alone [40].
accounted for 12 %, 24 %, and 36 % of the floor area, respectively. For Fig. 6 displays the results for the largest vent area (36 % of the floor
these experiments, we selected two target hydrogen concentrations to be area). The times to reach a LFL concentration in this scenario were 47.8
reached via ventilation: 4 vol% (LFL) and 1 vol% (safety ventilation s, 58.8 s, and 42 s, whereas it took 161.2 s, 134.4 s, and 141.2 s to reach a
concentration according to NFPA 86, which is 25 % of LFL). The concentration of 1 vol% (Table 4). Compared to the aforementioned two
ventilation time measurement was initiated when the vent was opened, scenarios, the ventilation time in this scenario was the shortest on
and the concentration values indicated by the topmost sensors (S1 and average and the deviation tended to be smaller. However, Case E7 took
S4) were uniform at 15 ± 0.5 %. Taking a conservative stance, from a up to 21.4 s longer to reach 1 vol%, compared with Cases E4, E5, and E6,
safety point of view, ventilation ceased only when the detectable with a 24 % vent area. This was due to the sudden decrease in con­
hydrogen concentration had dropped below the critical threshold centration at S3 and an increase in concentration at S1, which co-
throughout the confined space. In other words, ventilation was deemed occurred at approximately 334 s (Fig. 6a); the hydrogen in the vicinity
adequate when all sensors indicated gas levels below the desired con­ of S3 stayed near S1 and the wall due to an unstable interior flow. These
centration, while the time measurement was terminated when the last results corroborate those of previous studies. It is known that hydrogen
sensor provided this reading. Based on the data from that sensor, a non- inside a naturally ventilated enclosure is subject to large concentration
linear curve fit the LFL. instabilities due to shear layers generated near the vent edges and that

Fig. 8. Comparison of hydrogen concentrations per monitoring location (a) Left side of enclosure (S1, S2, S3) and (b) Center side of enclosure (S4, S5, S6).

1401
J. Kim et al. International Journal of Hydrogen Energy 50 (2024) 1395–1405

Table 5 3.2. Validation of numerical simulation


Conditions influencing ventilation and ventilation time per case.
Case N1 Case N2 Case N3 Case N4 The numerical analysis and experiment results were compared by
performing the analysis for a total analysis time of 900 s. Fig. 8 indicates
Opened (natural) vent size (%) 24 24 12 12
Fan capacity (m3/min) 22 44 22 44 the hydrogen concentrations over time in the center cross-section of the
Ventilation Time (s) 46.2 44.8 102.6 63.4 enclosure. After 283 s, when the hydrogen leaks stopped, the hydrogen
concentrations were found to continuously decrease after the opening of
the center vent. The same monitoring points as in the experiment were
this possibility increases with the use of larger vents [41]. selected to verify the numerical analysis; hydrogen concentrations over
To compare the ventilation times of all cases, a venting coefficient time are shown in Fig. 8. The changes in the hydrogen concentrations
value was calculated according to the vent size (Fig. 7). The venting showed similar tendencies in the experiment and the numerical analysis.
coefficient (Kv) is one of the design parameters of an explosion panel. It At the calculation time of 283 s, when the leaks were completed,
is a dimensionless value calculated using the outlet area (for pressure or hydrogen concentrations in CFD and the experiment were quantitatively
smoke generated by an explosion or fire) and the internal volume of the compared at each monitoring point; it was found that each hydrogen
enclosed space using the formula Kv = V2/3/Av, where V is the internal concentration showed differences of 1.14 % and 1.52 %, respectively, at
volume of enclosure, and Av is the vent area [42]. the left and center monitoring points at the top, and of 2.25 % and 3.38
Overall, as the vent size increased, the required ventilation time % at the center, indicating that the difference between the numerical
decreased and showed less deviation. Notably, the Kv was 2.21 in the analysis results and the experiment results was relatively significant at
case with the largest vent size, in which a 4 vol% LFL was reached in less the left and center monitoring points at the bottom, rather than those at
than 1 min. Here, even a concentration of 1 vol% was reached relatively the top. There was a deviation in hydrogen concentration of the center
quickly (145.6 s). However, there was no significant difference when bottom between the result of experiment and CFD. This is because all
comparing these results to those obtained at Kv = 3.31. This suggests monitoring points in the CFD simulation are located in the path through
that ventilation performance improves as the vent area increases, but which hydrogen passes. Still, in actual experiments, the sensor’s location
this difference will decrease after a certain point. Thus, we established is not in the path through which hydrogen leaks directly, and the con­
that the size of explosion panels should represent a Kv of 3.31 or less centration of hydrogen is measured through diffusion. However, the
when considering the contribution of natural ventilation in the emer­ monitoring points on the upper parts are only considered in the analysis,
gency use of such a panel. so the hydrogen concentration of the center bottom can be ignored.

Fig. 9. Comparison of hydrogen concentrations per monitoring location for Case N1 and Case N2. (a) Left side of enclosure (S1, S2, S3 of Case N1), (b) Center side of
enclosure (S4, S5, S6 of Case N1), (c) Left side of enclosure (S1, S2, S3 of Case N2), and (d) Center side of enclosure (S4, S5, S6 of Case N2).

1402
J. Kim et al. International Journal of Hydrogen Energy 50 (2024) 1395–1405

Fig. 10. Comparison of hydrogen concentrations per monitoring location for Case N3 and Case N4. (a) Left side of enclosure (S1, S2, S3 of Case N3), (b) Center side
of enclosure (S4, S5, S6 of Case N3), (c) Left side of enclosure (S1, S2, S3 of Case N4), and (d) Center side of enclosure (S4, S5, S6 of Case N4).

3.3. Characteristics of forced ventilation concentration, like the method used in the previous experiment, the
time when all monitoring points were below the target concentration
Based on the simulation that confirmed validity, the results of the was measured, as shown in Table 5.
simultaneous application of natural and forced ventilation were shown In the case in which a fan with a capacity of 22 m3/min (Case N1)
below. The final target concentration was set as 4 vol% which is the LFL was installed in the center vent within a natural ventilation area cor­
of hydrogen, in the computational analysis. To measure the target responding to 24 % of the floor area, it took 46.2 s to reach the target

Fig. 11. Overall comparison. (a) Concentration changes of the sensor [C-top] that fell below 4 % at the latest per case, and (b) Comparison of ventilation time below
4 % per case.

1403
J. Kim et al. International Journal of Hydrogen Energy 50 (2024) 1395–1405

concentration (LFL) (Table 5, Fig. 9). This finding is close (only 3 s 4. Conclusions
difference) to the fastest ventilation time (43.2 s) in cases E4, E5, and E6,
in which only natural ventilation (24 % of the floor area) was applied in This study experimentally investigated the efficiency of natural
the previous experiment. Considering that the simulation was analyzed ventilation in mitigating the effects of a hydrogen concentration buildup
conservatively, it is judged that the internal ventilation pattern (intake in a confined space with limited ventilation by deploying a vent that
and exhaust) formed under the influence of forced ventilation increased serves as an explosion panel. We reached the following conclusions.
ventilation efficiency. The formation of such a ventilation pattern can be
expected to maintain a constant ventilation efficiency, unlike natural i) As for experimental ventilation time, it was difficult to predict the
ventilation, which varies depending on the surrounding environment general ventilation time required to reach safe conditions in the
[43,44]. However, there was no considerable decrease in ventilation case of the smallest vent area (Kv = 6.62) because it was highly
time due to the addition of forced ventilation. affected by the external environment.
Under Case N2, similar to Case N1, in which a fan with a capacity of ii) The required ventilation time tended to decrease as the vent area
44 m3/min was installed in the center vent within a natural ventilation increased. But the ventilation time (to LFL) for the medium and
area corresponding to 24 % of the floor area, it required 44.8 s to reach large vents was similar. These results revealed that selecting an
the LFL (Table 5, Fig. 9). Compared to this result with Case N1 with the area below Kv = 3.31 is reasonable when designing a roof vent.
capacity of 22 m3/min, a slight difference of 1.4s was found. This may be iii) As a result of the numeric analysis, the time required for venti­
due to the saturation of ventilation efficiency. In the previous experi­ lation decreased as fan capacity increased. We found that this
mental part, the ventilation efficiency of natural vents, which have an effect was larger when the natural ventilation area was smaller.
area equivalent to 24 % and 36 % of the floor area, was similar. This This implies that the ventilation performance depended on forced
indicates that when an appropriate natural ventilation area is secured in ventilation factors like the capacity of fans as the natural venti­
forced ventilation, the natural ventilation area is dominant in deter­ lation area decreased.
mining ventilation efficiency rather than forced ventilation factors such iv) Higher ventilation efficiency can be expected with forced venti­
as fan capacity. Therefore, if natural ventilation area had already pro­ lation under the condition that the vent area is larger than Kv =
vided optimal ventilation efficiency, further increases in forced venti­ 3.31; however, in the case where it is smaller than Kv = 3.31, the
lation capacity may not have a real effect. ventilation time will be dominated by natural ventilation area,
Under Case N3, in which there was a fan with a capacity of 22 m3/ which would make it difficult to expect a higher ventilation ef­
min within a natural ventilation area corresponding to 12 % of the floor ficiency after adding forced ventilation.
area, it took 102.6 s to reach the LFL (Table 5, Fig. 10). Compared with
Case N1, with the fan with the same capacity, and the difference in the This study proposed a practical and novel method to design safety
natural ventilation area by 12 %, two cases showed a difference of 56.4 measures for areas where hydrogen is stored. This method can be
s; this finding is close (6.2 s difference) to the fastest ventilation time applied as an emergency measure when designing a ventilation system
(108.8 s) in Cases E1, E2, and E3, in which only natural ventilation (12 for a confined space. However, it may be challenging to maximize
% of the floor area) was applied in the previous experiment. Compared ventilation efficiency by using only the method proposed in this study as
to the results of 297.6 s and 228.4 s, it was 2.9 times smaller. It was also an emergency ventilation system. Designs such as the addition of forced
confirmed that the ventilation time, compared to Case N1, was more ventilation with nitrogen proposed in Ref. [19] can help design safer
influenced by the natural ventilation area than the fan capacity. hydrogen facilities. Therefore, further research is needed to increase
Case N4, similar to Case N3, in which there was a fan with a capacity ventilation efficiency by applying additional methods.
of 44 m3/min within the natural ventilation area corresponding to the
floor area, required 63.4 s to reach the LFL (Table 5, Fig. 10). Compared
to the difference between Cases N1 and N2 with 12 % additional natural Declaration of competing interest
ventilation areas, the difference between Cases N3 and N4 significantly
increased by approximately 37.8 s. This finding confirms that depen­ The authors declare that they have no known competing financial
dence on forced ventilation via fan increased in a smaller natural interests or personal relationships that could have appeared to influence
ventilation area. If the area of the natural ventilation where air is being the work reported in this paper.
drawn in is small, hydrogen tends to move downward due to the pro­
motion of global circulation within the confined space, making it diffi­ Acknowledgements
cult for the hydrogen to be discharged through the roof vent [41]. This
directly affects the ventilation rate, and the more stringent these re­ This work is supported by the Ministry of the Interior and Safety
strictions become, the more efficient it is to apply forced ventilation. (20018517). The funding agency had no role in study design; in the
Therefore, if the vent size was 12 % of the floor area, forced ventilation collection, analysis and interpretation of data; in the writing of the
(fan capacity) greatly affected the ventilation efficiency. report; and in the decision to submit the article for publication.
Fig. 11 indicates the comparison of entire cases. We found that the
last point falling below 4 % among six monitoring points in all cases was
the center-top point (C-top), which is revealed in Fig. 11 (a), while the References
total ventilation time can be found in Fig. 11 (b). As for the overall
[1] Derwent RG. Global warming potential (GWP) for hydrogen: sensitivities,
comparison, there was a tendency in which ventilation efficiency uncertainties and meta-analysis. Int J Hydrogen Energy 2023;48(22):8328–41.
increased as fan capacity increased; however, the impact of the fan https://doi.org/10.1016/j.ijhydene.2022.11.219.
[2] Momirlan M, Veziroglu TN. Current status of hydrogen energy. Renew Sustain
considerably decreased as the natural ventilation area increased.
Energy Rev 2002;6(1–2):141–79. https://doi.org/10.1016/S1364-0321(02)00004-
Furthermore, applying forced ventilation can reduce the overall venti­ 7.
lation time compared to the application of natural ventilation only. [3] Rosen MA, Koohi-Fayegh S. The prospects for hydrogen as an energy carrier: an
However, as mentioned above, ventilation performance tends to vary overview of hydrogen energy and hydrogen energy systems. Energ Ecol Environ
2006;1:10–29.
depending on the natural ventilation area. Therefore, it is ideal to secure [4] International Energy Agency. The future of hydrogen. Seizing today’s
the area of the natural vent when applying the emergency vent in case of opportunities. 2019. https://iea.blob.core.windows.net/assets/9e3a3493-b9a6-
hydrogen leakage. Still, it is reasonable to use additional forced vent if 4b7d-b499-7ca48e357561/The_Future_of_Hydrogen.pdf. [Accessed 18 June 2021].
[5] Scovell MD. Explaining hydrogen energy technology acceptance: a critical review.
only a small area can be secured. Int J Hydrogen Energy 2022;47(19):10441–59. https://doi.org/10.1016/j.
ijhydene.2022.01.099.

1404
J. Kim et al. International Journal of Hydrogen Energy 50 (2024) 1395–1405

[6] Barbir F. Safety issues of hydrogen in vehicles. Energy Partners; 1999. p. 1–5. [26] Merilo EG, Groethe MA, Colton JD, Chiba S. Experimental study of hydrogen
Available from: https://courses.engr.illinois.edu/npre470/sp2019/web/readings/ release accidents in a vehicle garage. Int J Hydrogen Energy 2011;36:2436–44.
Hydrogen%20safety%20issues.pdf. https://doi.org/10.1016/j.ijhydene.2010.04.056.
[7] Molnarne M, Schroeder V. Hazardous properties of hydrogen and hydrogen [27] Huang T, Zhao M, Ba Q, Christopher DM, Li X. Modeling of hydrogen dispersion
containing fuel gases. Process Saf Environ Protect 2019;130:1–5. https://doi.org/ from hydrogen fuel cell vehicles in an underground parking garage. Int J Hydrogen
10.1016/j.psep.2019.07.012. Energy 2022;47:686–96. https://doi.org/10.1016/j.ijhydene.2021.08.196.
[8] Swain MR, Filoso P, Grilliot ES, Swain MN. Hydrogen leakage into simple [28] Matsuura K, Nakano M, Ishimoto J. The sensing-based adaptive risk mitigation of
geometric enclosures. Int J Hydrogen Energy 2003;28(2):229–48. https://doi.org/ leaking hydrogen in a partially open space. Int J Hydrogen Energy 2009;34:
10.1016/S0360-3199(02)00048-4. 8770–82. https://doi.org/10.1016/j.ijhydene.2009.08.013.
[9] Denkevits A. Explosibility of hydrogen–graphite dust hybrid mixtures. J Loss Prev [29] Agranat V, Cheng Z, Tchouvelev A. CFD modeling of hydrogen releases and
Process Ind 2007;20(4–6):698–707. https://doi.org/10.1016/j.jlp.2007.04.033. dispersion in hydrogen energy station. In: Proceedings of the 15th world hydrogen
[10] Wei R, Lan J, Lian L, Huang S, Zhao C, Dong Z, Weng J. A bibliometric study on energy conference: citeseer; 2004. Available at: https://citeseerx.ist.psu.edu/doc
research trends in hydrogen safety. Process Saf Environ Protect 2022;159:1064–81. ument?repid=rep1&type=pdf&doi=b656610f4eeda9027e58ea43fe7855a73f5b
https://doi.org/10.1016/j.psep.2022.01.078. 484f.
[11] Park B, Kim Y, Hwang I. An experimental study on the explosion hazards in the fuel [30] Taveau J. Explosion hazards related to hydrogen releases in nuclear facilities.
cell room of residential house. J Korean Surg Soc 2021;36(4):71–9. https://doi. J Loss Prev Process Ind 2011;24:8–18. https://doi.org/10.1016/j.jlp.2010.08.002.
org/10.14346/JKOSOS.2021.36.4.71. [31] Gitushi K, Blaylock M, Klebanoff L. Hydrogen gas dispersion studies for hydrogen
[12] Park B, Kim Y, Paik S, Kang C. Numerical and experimental analysis of jet release fuel cell vessels II: fuel cell room releases and the influence of ventilation. Int J
and jet flame length for qualitative risk analysis at hydrogen refueling station. Hydrogen Energy 2022;47:21492–505. https://doi.org/10.1016/j.
Process Saf Environ Protect 2021;155:145–54. https://doi.org/10.1016/j. ijhydene.2022.04.263.
psep.2021.09.016. [32] Duijm NJ, Markert F. Safety-barrier diagrams as a tool for modelling safety of
[13] Baraldi D, Papanikolaou E, Heitsch M, Moretto P, Cant RS, Roekaerts D, hydrogen applications. Int J Hydrogen Energy 2009;34(14):5862–8. https://doi.
Dorofeev S, Koutchourko A, Middha P, Tchouvelev AV, Ledin S, Wen J, org/10.1016/j.ijhydene.2009.02.002.
Venetsanos A, Molkov V. Gap analysis of CFD modelling of accidental hydrogen [33] Hosseinian SS, Torghabeh ZJ. Major theories of construction accident causation
release and combustion. 2010. https://publications.jrc.ec.europa.eu/reposit models: a literature review. Int J Adv Eng Technol 2012;4(2):53. Available at: http
ory/bitstream/JRC58011/jrc58011_new_jrc58011.pdf. s://www.researchgate.net/profile/Seyyed-Shahab-Hosseinian/publication/2684
[14] Park B, Kim Y, Lee K, Paik S, Kang C. Risk assessment method combining 39084_Major_theories_of_construction_accident_causation_models_A_literature_re
independent protection layers (IPL) of layer of protection analysis (LOPA) and view/links/59ea640ba6fdccef8b08d4b2/Major-theories-of-construction-accident
RISKCURVES software: case study of hydrogen refueling stations in urban areas. -causation-models-A-literature-review.pdf.
Energies 2021;14(13):4043. https://doi.org/10.3390/en14134043. [34] Reason J, Hollnagel E, Paries J. Revisiting the Swiss cheese model of accidents.
[15] Swain MR, Grilliot ES, Swain MN. The application of a hydrogen risk assessment J Clin Eng 2006;27(4):110–5. https://www.eurocontrol.int/sites/default/files/li
method to vented spaces. Adv Hydrogen Energy 2002:163–73. https://doi.org/ brary/017_Swiss_Cheese_Model.pdf.
10.1007/0-306-46922-7_13. [35] Hollnagel E. Is safety a subject for science? Saf Sci 2014;67:21–4. https://doi.org/
[16] Pitts WM, Yang JC, Fernandez MG. Helium dispersion following release in a 1/4- 10.1016/j.ssci.2013.07.025.
scale two-car residential garage. Int J Hydrogen Energy 2012;37(6):5286–98. [36] Ono R, Nifuku M, Fujiwara S, Horiguchi S, Oda T. Minimum ignition energy of
https://doi.org/10.1016/j.ijhydene.2011.12.008. hydrogen–air mixture: effects of humidity and spark duration. J Electrost 2007;65
[17] Gupta S, Brinster J, Studer E, Tkatschenko I. Hydrogen related risks within a (2):87–93. https://doi.org/10.1016/j.elstat.2006.07.004.
private garage: concentration measurements in a realistic full scale experimental [37] Roache PJ. Quantification of uncertainty in computational fluid dynamics. Annu
facility. Int J Hydrogen Energy 2009;34(14):5902–11. https://doi.org/10.1016/j. Rev Fluid Mech 1997;29:123–6. https://doi.org/10.1146/annurev.fluid.29.1.123.
ijhydene.2009.03.026. [38] Zhong H-Y, Sun J, Lin C, Wu S-H, Shang J, Kikumoto H, et al. CFD simulation of
[18] Hajji Y, Bouteraa M, ElCafsi A, Belghith A, Bournot P, Kallel F. Natural ventilation pumping ventilation in a three-story isolated building with internal partitioning:
of hydrogen during a leak in a residential garage. Renew Sustain Energy Rev 2015; effects of partition widths, heights and locations. Building Simulation. Springer;
50:810–8. https://doi.org/10.1016/j.rser.2015.05.060. 2023. p. 1–18. https://doi.org/10.1007/s12273-023-1068-3.
[19] Lee J, Cho S, Cho H, Cho S, Lee I, Moon I, Kim J. CFD modeling on natural and [39] Zhang X, Weerasuriya AU, Wang J, Li CY, Chen Z, Tse KT, Hang J. Cross-ventilation
forced ventilation during hydrogen leaks in a pressure regulator process of a of a generic building with various configurations of external and internal openings.
residential area. Process Saf Environ Protect 2022;161:436–46. https://doi.org/ Build Environ 2022;207:108447. https://doi.org/10.1016/j.
10.1016/j.psep.2022.03.065. buildenv.2021.108447.
[20] Barley CD, Gawlik K. Buoyancy-driven ventilation of hydrogen from buildings: [40] Swain M, Swain M. Passive ventilation systems for the safe use of hydrogen. Int J
laboratory test and model validation. Int J Hydrogen Energy 2009;34(13): Hydrogen Energy 1996;21:823–35. https://doi.org/10.1016/0360-3199(96)
5592–603. https://doi.org/10.1016/j.ijhydene.2009.04.078. 00002-X.
[21] Matsuura K. Effects of the geometrical configuration of a ventilation system on [41] Matsuura K. Effects of the geometrical configuration of a ventilation system on
leaking hydrogen dispersion and accumulation. Int J Hydrogen Energy 2009;34 leaking hydrogen dispersion and accumulation. Int J Hydrogen Energy 2009;34
(24):9869–78. https://doi.org/10.1016/j.ijhydene.2009.09.044. (24):9869–78. https://doi.org/10.1016/j.ijhydene.2009.09.044.
[22] Ji HW, Lee H, Hwang I, Jang H. Safe ventilation methods against leaks in hydrogen [42] Wang Q, Luo X, Wang C, Liu Y, Zhou P, Li B. Experimental study on external
fuel cell rooms in homes. Energies 2022;15(15):5434. https://doi.org/10.3390/ explosion for vented hydrogen deflagration in a rectangular tube with different
en15155434. vent coefficients. Process Saf Environ Protect 2022;158:331–9. https://doi.org/
[23] Xie H, Li X, Christopher DM. Emergency blower ventilation to disperse hydrogen 10.1016/j.psep.2021.12.002.
leaking from a hydrogen-fueled vehicle. Int J Hydrogen Energy 2015;40(25): [43] Chu C-R, Su Z-Y. Natural ventilation design for underground parking garages. Build
8230–8. https://doi.org/10.1016/j.ijhydene.2015.03.14. Environ 2023;227:109784. https://doi.org/10.1016/j.buildenv.2022.109784.
[24] Cerchiara GM, Mattei N, Schiavetti M, Carcassi MN. Natural and forced ventilation [44] Flores-Velazquez J, Montero JI, Baeza EJ, Lopez JC. Mechanical and natural
study in an enclosure hosting a fuel cell. Int J Hydrogen Energy 2011;36:2478–88. ventilation systems in a greenhouse designed using computational fluid dynamics.
https://doi.org/10.1016/j.ijhydene.2010.03.124. Int J Agric Biol Eng 2014;7:1–16. https://doi.org/10.3965/j.ijabe.20140701.001.
[25] Prasad K. High-pressure release and dispersion of hydrogen in a partially enclosed
compartment: effect of natural and forced ventilation. Int J Hydrogen Energy 2014;
39:6518–32. https://doi.org/10.1016/j.ijhydene.2014.01.189.

1405

You might also like