Quantitative Risk Assessment of Hydrogen Refueling Station in Cheonan City of South Korea

: The average temperature of the Earth has risen due to the accumulation of greenhouse gases emitted from the usage of fossil fuels. The consequential climate changes have caused various problems, fueling the growing demand for environmentally friendly energy sources that can replace fossil fuels. Batteries and hydrogen have thus been utilized as substitute energy sources for automo-biles to reduce fossil fuel consumption. Consequently, the number of hydrogen refueling stations is increasing due to an increase in the number of hydrogen-powered vehicles. However, several incidents have been reported in the United States of America and Japan where hydrogen refueling stations have been operating for a long time. A risk assessment of hydrogen refueling stations operating in urban areas was performed in this study by calculating the risk effect range using a process hazard analysis tool (PHAST) v8.7 from DNV-GL and a hydrogen risk assessment model (HyRAM) from Sandia National Laboratories (SNL). The societal risk was assessed through a probit model based on the calculation results. The assessment results showed that the risk caused by jet ﬁre and overpressure in an incident is lower than the ‘as low as reasonably practicable’ (ALARP) level.


Introduction
The consumption of fossil fuel has increased drastically since its recognition as one of the main sources of energy for humankind.The emission of greenhouse gases has concomitantly increased, leading to an increase in the average temperature of the Earth.Climate changes due to the rise in the Earth's average temperature have resulted in a variety of issues, thus highlighting the importance of reducing the use of fossil fuels and increasing the utilization of environmentally friendly energy sources.One well-known way of reducing the usage of fossil fuels is by changing the type of fuels used for automobiles.The European Union (EU) announced the decision to limit the operation of automobiles powered by fossil fuels.Batteries and hydrogen are widely used as substitute power sources for automobiles in place of fossil fuels [1].Thus, the number of hydrogen-powered automobiles is constantly increasing.According to statistical data from the Ministry of Land, Infrastructure, and Transport, the number of hydrogen-powered vehicles registered in Korea increased by nearly three-fold, from 11,046 in January 2021 to 32,484 in June 2023 [2].The number of hydrogen refueling stations (HRSs) is increasing due to an increase in the number of hydrogen-powered vehicles.Hydrogen refueling stations are operated mainly in Asian countries such as South Korea, Japan, and China.In European countries, HRSs are operated in Germany, France, the United Kingdom, the Netherlands, and Switzerland.Globally, more than 1070 HRSs have been built as of 2023, of which, 210 HRSs are located in South Korea [3].
Hydrogen refueling stations for HFCVs (Hydrogen Fuel Cell Vehicles) have the advantage of faster charging than electric vehicle charging stations for electric vehicles.Hydrogen vehicles do not generate greenhouse gases in the process of generating power.In addition, hydrogen has the advantage of having a higher calorific value per unit mass than fossil fuels.On the contrary, hydrogen has a very low ignition energy of 0.018 mJ and a wide combustion range of 4-75% in the air.Because hydrogen has a low energy per unit volume, it is stored at high pressure to increase its energy density [4].When hydrogen is leaked from a high-pressure storage facility through a leak hole, a shock wave compresses the air and the temperature of the hydrogen-air mixture rises.Then, self-ignition may occur [5,6].This phenomenon increases the risk associated with storage tanks and tube trailers used at HRSs.
No accidents involving HRSs have occurred in South Korea thus far, but several cases have been reported in the U.S. and Japan where HRSs have been operating for years.In the U.S., 22 incidents occurred from 2004 to 2012, whereas in Japan, 21 incidents occurred from 2005 to 2014 [7].Since there is a risk of high-pressure hydrogen and an accident risk of HRSs, it is necessary to evaluate the risk of high-pressure hydrogen facilities in HRSs.Accordingly, several prior studies have focused on the quantitative risk assessment (QRA) of high-pressure facilities in HRSs [8][9][10].However, most of these studies assessed individual risks of HRSs.This study thus aims to evaluate the risk of an HRS that is currently operational and to lower the societal risk of it.

Materials and Methods
This study assessed the safety of hydrogen refueling stations installed in urban regions of Korea.A consequence analysis (CA) and a QRA were performed based on the information from the HRS analyzed in this study.

HRS Specifications
The analyzed HRS was selected because it has a high individual risk as it handles a large amount of hydrogen and there is a high population density in the installation area and a high societal risk in the nearby residential areas in the case of hydrogen leakage.The HRS analyzed in this study is in Cheonan City in South Korea.Residential and commercial regions are located 100 m to the west from the HRS.Table 1 presents the specifications for a tube trailer and a storage tank used at the analyzed HRS, while Figure 1 illustrates the process flow diagram (PFD) of the HRS.The analyzed HRS uses a tube trailer pressurized to 350 bar and a storage tank pressurized to 850 bar and refuels hydrogen automobiles at a pressure of 700 bar through a dispenser.Figure 2 shows the location of the HRS and a description of the surrounding region.A wide road is located to the west of the HRS, followed by residential areas.Commercial facilities are located between the two residential areas.A parking facility is located to the north and the east of the HRS, and a park is situated to the south of the HRS.

Accident Scenario
The factors (leak size and leak frequency) required to devise an accident scenario for the HRS were referenced from the "Analyses to Support Development of Risk-Informed Separation Distances for Hydrogen Codes and Standards" of Sandia National Laboratories (SNL) [11].Table 2 presents three types of leak sizes and leakage frequencies proposed in the SNL standard, appropriate for the design pressure of the analyzed HRS.

Weather Conditions
The atmospheric and weather conditions of the worst-case scenario were selected based on the "Technical Guideline for the Selection of Worst and Alternative Accident Scenarios" of the Korea Occupational Safety & Health Agency.Table 3 presents the wind speed, atmospheric temperature, atmospheric stability, and humidity in the worst-case weather conditions.

Hydrogen Risk Analysis
The 'risk effects' from accidents at an HRS include thermal radiation from a jet fire and overpressure from explosions.The HyRAM 5.0 program was used to select the effect range of a jet fire.HyRAM software was developed specifically for the modeling and QRA of hydrogen, methane, and propane systems.Certain studies have been conducted on gas leakage and jet fire modeling in hydrogen storage facilities using HyRAM [13,14].The effect of overpressure was assessed using the process hazard analysis tool (PHAST) v8.7 of DNV-GL.PHAST is capable of modeling the leakage and dispersion of various chemical substances, as well as fire and explosion, and has been used in numerous studies for analyzing the effect of the blast overpressure of chemical substances [15].

Jet Fire
Jet fire can be caused by the combustion of a leaked gas stored under high pressure [16].HyRAM provides five types of notional nozzle models [17][18][19][20][21]. Specifically, this study used the model proposed by Yüceil and Ötügen [17].The conservation of mass for the notional nozzle models is represented by Equation (1).
Here, ρ is the density, ν is the velocity, A is the cross-sectional area, and C D is the discharge coefficient.In the model proposed by Yüceil and Ötügen, momentum and energy are conserved as shown in Equation (2).
A eff can be expressed as Equation (4).
For computing the effective density to select the area in Equation ( 4), the density can be calculated through the conservation of energy using Equation (5) based on the assumption of isentropic expansion.
The 'effect range' for humans exposed to a jet fire was determined based on three levels of thermal radiation caused by the jet fire: 4, 12.5, and 37.5 kW/m 2 .Table 4 presents the effects of each thermal radiation range on humans.An explosion can be triggered by combustion when a leak hole is generated in a hydrogen storage tank.In this study, the risk was assessed based on the assumption that overpressure is caused by an explosion when a flash fire is generated due to ignition delay after a leak hole is created.PHAST v8.7 was used to assess the effect range of an explosion based on the TNO multi-energy (TNO ME) model [24].The TNO ME model is a practical model that adequately represents the mechanics of vapor cloud explosion (VCE) [25].Figure 3 shows the basic form of the TNO ME model.Here, P 0 is the peak overpressure and r 0 is the initial radius.The overpressure and distance can be expressed in terms of dimensionless parameters, as shown in Equations ( 6)- (8).
Here, P s is the maximum blast overpressure, P a is atmospheric pressure, r is the explosion radius, and E is the emission energy of the explosion.The scaled overpressure can be calculated by quantifying E and applying it to the overpressure curvature shown in Figure 4 according to the scaled distance (r').The contour level of the overpressure was set to 1, 3, and 5 psi according to the effects on humans.Table 5 presents the effects of different levels of overpressure on humans.The physical consequences of the continuous leakage of combustible fuels such as hydrogen include gas release, jet fire, flash fire, and explosion [27,28].Figure 5 illustrates the event sequence diagram (ESD) for each physical consequence.The annual frequency of release is calculated for each physical consequence based on the ESD.The probability of failed leakage detection was set to 0.01 based on the probability of failure based on the demand of the system from the RVIM Guideline for Quantitative Risk Assessment, also known as the "Purple Book" [30].The probability of hydrogen ignition according to the release rate provided by the SNL was used as the basic ignition probability, as presented in Table 6.

Probit Model
A probit model is used to calculate the fatality rate of the exposed population with respect to the degree of exposure.The probit function for the acute toxicity of chemicals based on the probit model is widely utilized in the Netherlands' external safety policy [32].In addition, there are probit functions for thermal harm, overpressure harm, and the effects of fragments.The basic form of the probit function is shown in Equation (9).
Here, k 1 and k 2 are constants; V represents the degree and time of exposure.
For the fatality rate of heat flux, the probit function of lethality shown in Equation ( 10) was used in the statistical model of TNO to estimate injury due to heat radiation [24].
The probit function of the lung hemorrhage model due to overpressure shown in Equation ( 11) was used to calculate the fatality rate due to overpressure [33].
Pr Overpressure = 1.47 + 1.37 × ln(P s ) The effect range was calculated by simulating jet fire and explosion overpressure in an accident scenario.The number of fatalities for each accident scenario was calculated by applying the fatality rate derived from the probit function to the floating population of the relevant region.The F-N curve was constructed by applying the accident frequency for each scenario, calculated through ETA, to the estimated number of fatalities for assessing the societal risk.

Results in the Case of a Jet Fire
The effect range of a jet fire was modeled using HyRAM.The heat flux at a height of 1 m and at the center of the jet fire was calculated at twenty different points from 0.01 m to 40 m (Tables 7 and 8).Referring to Table 4, the range of heat flux for 4, 12.5, and 37.5 kW/m 2 , which is the effect on humans, was confirmed.For a tube trailer, it was discovered that the heat flux exceeded 4 kW/m 2 for up to 3 m in the event of a small-sized leak and would thus affect humans.The heat flux exceeded 12 kW/m 2 for up to 10 m and exceeded 4 kW/m 2 for up to 12 m in the event of a medium-sized leak.For a large-sized leak, the heat flux exceeded 37.5 kW/m 2 for up to 27 m and exceeded 4 kW/m 2 at 40 m, thus affecting humans at all twenty points.For a storage tank, it was discovered that when a small-sized leak occurred, the heat flux was 4 kW/m 2 or below at a distance of 3 m, thus having a low impact on humans.In the case of a medium-sized leak, the heat flux exceeded 4 kW/m 2 at a distance of up to 10 m, thus having an effect on humans.For a large-sized leak, the heat flux exceeded 37.5 kW/m 2 at distances up to 22 m and exceeded 36.85 kW/m 2 at 25 m, thus having a fatal impact on humans at a distance of up to 25 m.
The impact range of the heat flux was greater in a tube trailer than in a storage tank with hydrogen under very high pressure.This difference is attributable to the fact that the leak hole size of a tube trailer is set to be larger than that of a storage tank.For both the tube trailer and the storage tank, people within 3 m were exposed to the heat flux caused by a jet fire and were slightly affected when a small-sized leak occurred, but no serious injury or fatal effect was found.When a medium-sized leak occurred, however, people within 10 m of a tube trailer and 8 m of a storage tank could be killed within a few minutes.When a large-sized leak occurred, the heat flux (exceeding 37.5 kW/m 2 ) caused by a jet fire had a fatal impact for up to 27 m for a tube trailer and up to 22 m for a storage tank.

Results in the Case of Overpressure
Table 9 presents the overpressure determined through simulation using PHAST.The simulation data from PHAST did not take the geographical features and firewalls of the HRS into account.No explosion occurred in the event of a small-sized leak, whereas overpressure of 1 psi or above occurred in the case of medium-and large-sized leaks.
For a tube trailer, the range for an overpressure of 3 psi, which can cause serious injury to people, was 8.29 m for a medium-sized leak and 30.49m for a large-sized leak.The range for an overpressure of 5 psi, which can cause fatalities, was 6.01 m for a medium-sized leak and 22.12 m for a large-sized leak.For a storage tank, the range for an overpressure of 3 psi was 6.45 m for a medium-sized leak and 24.73 m for a large-sized leak, while the range for an overpressure of 5 psi was 4.68 for a medium-sized leak and 17.94 m for a large-sized leak.The risk range was greater in a tube trailer where the leak size was set to be larger than that in a storage tank having higher storage pressure, which is similar to the results of the jet fire simulation.

Evaluation of Fatality through Probit Model
The percentage of fatalities was calculated according to distance in each scenario using the probit function (Figures 6-9).In accidents involving a tube trailer, the maximum distance at which fatalities may have occurred was 36 m for a jet fire and 33 m for overpressure.The maximum distance at which fatalities may have occurred in accidents involving a storage tank was 29 m for a jet fire and 27 m for overpressure.

Societal Risk
Table 10 presents the calculated frequency of jet fires and flash fires with respect to the leak size in a tube trailer and a storage tank.The floating population near the HRS evaluated herein was analyzed using the commercial area analysis system provided by the Ministry of SMEs and Startups [34].The floating population was estimated by this particular system in increments of 50 m 2 based on the call volumes of mobile phones.Table 11 shows the floating population within 300 m of the analyzed HRS by time period as of April 2023.The societal risk assessment was performed with the assumption that the risk is highest between 17:00 and 21:00 when the floating population is the highest; Figure 10 shows the resulting F-N curve.The maximum number of fatalities caused by leakage and fire at the HRS was estimated to be 7.95, and the expected accident frequency was 4.60 × 10 −10 /year.The F-N curve was positioned lower than the 'as low as reasonably practicable' (ALARP) level, thus demonstrating a low level of risk.This curve did not take into account the wall protection of the HRS.Considering the existence of the wall, the default value of 0.001 was applied as the defense coefficient of the blast-wall [35].The storage tank had a wall in 75% of directions, and the tube trailer had a wall in 25% of directions.Figure 11 shows the results of the F-N curve considering the burst-wall.

Conclusions
This study assessed the risk range for accidents occurring at the HRS in operation and performed a safety evaluation of the HRS by assessing the societal risk; the following conclusions can be drawn.

1.
Jet fire was modeled using HyRAM.The results showed that the range of fatal impact is at a maximum when a large-sized leak occurs, specifically at a distance of 27 m for a tube trailer and 22 m for a storage tank.When a medium-sized leak occurs, the distance of fatal impact is 8 m for a tube trailer and 5 m for a storage tank.When a small-sized leak occurs, there is no fatal impact.2.
The analysis performed using PHAST demonstrates a fatal impact for an overpressure radius of up to 11.08 m.The fatal impact of overpressure is not wider than that of HRS facilities.Therefore, the effect of overpressure is expected to be low in fatal effects, except for the workers of the HRS.

3.
The area of the analyzed HRS is 1997 m 2 .There is no fatal impact outside the HRS in the cases of small-and medium-sized leaks.Heat flux due to jet fire can have an effect even outside the HRS in the case of a large-sized leak, but the actual risk range may be smaller because the geographical features and firewalls of the HRS were not taken into consideration in modeling the effects of the jet fire and overpressure.
Future studies will focus on all the independent protection layers (IPLs) such as firewalls and water curtains installed at the HRS which would significantly lower the societal risk.

Figure 2 .
Figure 2. Location and regional status of the HRS.

Figure 3 .
Figure 3. Basic form of vapor cloud expansion as basis of multi-energy model.

Figure 4 .
Figure 4. Peak dynamic overpressure for combustion energy scaled distance.

Figure 6 .
Figure 6.Fatality due to thermal radiation from jet fire at tube trailer.

Figure 7 .
Figure 7. Fatality due to thermal radiation from overpressure at tube trailer.

Figure 8 .
Figure 8. Fatality due to thermal radiation from jet fire at storage tank.

Figure 9 .
Figure 9. Fatality due to thermal radiation from overpressure at storage tank.

Table 2 .
Accident scenario data for HRS.

Table 7 .
Heat flux of jet fire for tube trailer.

Table 8 .
Heat flux of jet fire in storage tank.

Table 10 .
Frequency of fire in each scenario.

Table 11 .
Floating population around the HRS.