Simulation of Flammable and Toxic Gases Released from Condensate Storage Tank in a Gas Plant Based on Elevation Change

Authors

1 Department of Chemical Engineering, Faculty of Engineering, Shahid Bahonar University of Kerman, Kerman, Iran National Iranian Gas Company (NIGC), South Pars Gas Complex (SPGC), Asaluyeh, Iran

2 Department of Chemical Engineering, Faculty of Engineering, Shahid Bahonar University of Kerman, Kerman, Iran

10.22108/gpj.2019.117376.1057

Abstract

The release of hydrocarbon and toxic vapor from atmospheric storage tanks is considered as a hazardous phenomenon in the chemical and process industries. Most often, detailed meteorological data for a particular location are unavailable, making the application of Pasquill’s atmospheric stability table the only choice for running simulations. The simulation of the flammable and toxic vapor release and cloud dispersion in South Pars, Iran in the 4th Gas plant incident in 2012, is run by applying the PHAST 6.53.1s/w. Meteorological data extracted from Pasquill's stability table. A simple methodology is developed for assessing land elevation difference between release source and gas detectors the obtained results are validated against time and location of recorded alarms in the site. The atmospheric parameters applied in this simulation are compared with wind speed and sky cloudiness data received from the Iranian meteorological institute. The Pasquill's parameters applied in this simulation are verified against local meteorological recorded data. The results indicate that if the vapor cloud is close to the earth's surface before any major change in the land elevation, this change can be neglected. Sensitivity analysis reveals that there will be fewer consequences if atmospheric discharge direction has a vertical upward orientation.

Keywords


1. Introduction

Process and chemical facilities like refineries and petrochemical plants are known as high-risk industrial plants (Arunraj, Maiti, 2009). One of the important factors in determining the scale of the potential accident in these facilities is the availability of hazardous materials in the equipment that could be released on an incident (Khan, Abbasi, 1999; Aliso et al., 2014). The storage tanks of crude oil, fuels like gasoline, natural gas condensate and tanks containing toxic substances are the riskiest sources in the process industries. According to (Lees, 1996; Khan, Abbasi, 1999), 3.3% of the reported explosions in chemical industry occur in tank farms.

The main cause of 17.8% of inferno in the chemical plants and related industries are the flammable liquid and gas overflow/release (Norstrom, 1982; Khan, Abbasi, 1999). The gasoline storage tank fire and explosion of Buncefield, England, 2005, fuel storage tank facility fire in Puerto Rico, USA 2009, and oil depot fire in Jaipur, India, 2009, all with severe human, environmental and economic consequences are some of the latest events reported due to hydrocarbon release from storage tanks (Gant, Atkinson, 2011; Joyce et al., 2018). A toxic material release from storage vessels in Bhopal, India in 1984 is still the worst tragedy of chemical plants with the most severe consequences (Khan, Abbasi, 1998; Rad et al., 2017). Despite the advancing knowledge on safety and technology applied in the construction of such facilities, such incidents occur is inevitable, thus the question: Have the consequences of possible major accidents on storage tanks containing toxic/flammable materials, and vapor dispersion been simulated and evaluated correctly to allow appropriate precautions to be considered in the design and construction stages to mitigate probability and consequence of an accident? (Seungkyu et al., 2014)

In this industry as to such incidents, Iran is not an exception: Booali petrochemical plant xylene splitter tower and naphtha storage tank fire in 2016 (Mehr news agency), fire and explosion of propane storage tanks in Khark Petrochemical Company in 2011 (National Iranian Petrochemical company) with a death toll of one person and gas leak incident in the 4th Gas Plant in South Pars, 2012. Because there exist many refineries and processing facilities in the world and the important role of storage tanks therein and the potential of their becoming subject to severe consequences, it is vital to run studies and evaluations in this context, which will, in turn, be contributive in the implementation of safety management systems in reducing the probability of expected consequences.

Gant and Atkinson (2011) studied a hydrocarbon storage tank explosion in Buncefield, England. Where, a tank overflow released 180 tons of gasoline and in 23 min, the partial evaporation generated a flammable vapor cloud of about 2 m height covering an area of 200,000 m2 which ultimately led to the explosion. They modeled hydrocarbon vapor cloud dispersion by applying the CFD technique and running sensitivity analysis on the influencing parameters: mesh sizing, turbulence effects, land topology; the cloud spread obstacles and surface roughness. The results indicate that both the fencing and barriers slopes are the most effective factors in vapor distribution, while turbulence and model mesh size had little impact and their simulation was insensitive to the roughness of the surface. Based on the observed vapor cloud expansion at different times obtained from camera images in the area, the considered wind velocity is zero. Due to different factors' simultaneous effects on vapor cloud dispersion, such assumptions may questionable.

Sharma et al. (2013) studied the vapor cloud explosion due to the release of 2116 tons of gasoline from a storage tank for 80 min in Jaipur, India, equivalent to 38 tons of TNT. Dispersion occurred at 1.5 m/s wind speed and a vapor cloud of 2 m height covered an area of 180,000 m2, causing flash fire after the ignition. The overpressure caused by the explosion was simulated in PHAST 6.51 s/w and compared with the affected area and results indicate that the maximum overpressure was 1 bar, which corresponds to the observed damage. The effect of obstacles on vapor cloud dispersion before the explosion is not clear in this study.

Many reliable theoretical models should be introduced to predict the main operating parameters' effect and optimize the key process variables if implementing exponential findings in industrial applications is sought. The models based on the energy and mass balance equations are difficult to solve especially when accompanied by an optimization procedure. The available commercial simulation PHAST s/w based on industrial data could be applied in accurate and reliable simulation, thus, its application in this study.

Next to analytical and numerical models, in this context, CANARY, EFFECTS, PHAST, SAFETI, and ALOHA are developed to evaluate such consequences in industrial facilities. Due to their high simulation speed and no need for expert knowledge, these softwares are widely applied (Arunraj, Maiti, 2009; Mishra, et al., 2013). Provided that the accident scenarios are extremely complex, especially in geometry, and obstacles are present on gas dispersion area or dispersed material is of two phases, the accuracy of results is questionable. Applying the computational fluid dynamics (CFD) technique will be considered as an alternative solution which of course is very time-consuming, needs skilled analysts and in some cases, it is even combined with other simplifying approximations (Pontiggia, et al., 2010; Venetsanos, et al., 2003).

A process hazard analysis software tool PHAST is designed to model the consequences of toxic and flammable gas dispersion, explosion and fire. This software is developed by DNV Company and due to its fairly reliable result yield and high technical support, it is applied by approximately 300 organizations worldwide (Parvini, Kordrostami, 2014). This software contains methods for calculating discharge and dispersion, and toxic or flammable effects (Gant, et al., 2013). PHAST is advantageous because it does not require a large volume of input data and its calculating time is short. Moreover, it provides required data for the risk assessment of equipment and processes (Meysami, et al., 2013).

In this study, flammable and toxic gas dispersions from a condensate storage tank located in South Pars field 4th Gas plant, Iran is assessed and the results are validated with a real release incident. The effects of the influencing parameters on the vapor distribution are discussed by running a sensitivity analysis.

 

 

2. Description of the Process and the Incident

2.1. Stabilization Process Unit

In gas plants, gas condensate is separated from the main gas stream in a slug catcher and sent to the condensate stabilization unit for vapor pressure adjustment. In this unit, light components are separated from the liquid in three stages through pressure reduction and temperature increase. The liquid condensate is stabilized, sweetened (that is, becoming free of hydrogen sulfide) and sent to the condensate storage tanks. If for any reason, the hydrocarbon condensate vapor pressure or the hydrogen sulfide content in the output is more than the acceptable rate, the product is sent to the off-spec condensate storage tank for later processing. In the South Pars 4th Gas plant, this storage tank is of fixed roof type and the design pressure of upper partition is 20 mbarg. If for any reason the vapor pressure in the tank exceeds 7 mbarg, the pressure control valve mounted on the tank roof vents the excess pressure to the atmosphere. If the pressure is still increasing, then the three safety valves installed on the tank open at 15 mbarg to discharge excess gas into the atmosphere.

2.2. Description of the Incident

In this incident, the unstabilized condensate is directed to the stabilization unit. When the tower reboiler is not completely warmed up the condensate stabilization is partial. The operator, knowing the product is off-spec, directs it to the off-spec condensate storage tank, which increases the vapor pressure due to excessive light components in the product and activate the pressure safety valves. Because no toxic or flammable detector is installed on the tank the discharge initiation time is not clear, and until the activation of the first gas detector installed on the adjacent unit, the personnel is not aware of the vapor release. The detectors' activation in other units and lack of investigation in determining the source by fire-fighting teams in these areas, as a misleading factor, led to the late identification of the leakage source, thus more scattered vapor cloud in the area. According to recorded alarms in the central control room, the first flammable gas is detected at 2:22 am and in 32 min, 9 detectors in 4 fire zones become activated. These detectors are of beam gas detector type that detect and report gas concentration by multiplying lower explosive limit (LEL) percentage into the distance between the transmitter and receiver detector. Moreover, 3 toxic gas detectors are activated in two zones that reported hydrogen sulfide concentration in ppm.

3. Methodology

In this study, the dispersion of flammable and toxic hydrogen sulfide gases is simulated by applying PHAST 6.53.1 s/w. The results are compared with the site data. For analysis and simulation of the accident, knowing the composition of released gas, rate, direction and elevation of released source, atmospheric conditions like the wind speed and direction, stability, relative humidity, and temperature, at the time of the occurrence is essential.

3.1. Gas Flow Characteristics

After collecting the required data like the rate of input and output flow, the temperature of the stabilization tower and pressure of different parts of stabilization and storage units from process control system the composition of the released gas and its rate are determined by applying the process simulation HYSYS 7.2 s/w. Based on the obtained compositions, the lower explosive limit is 19220 ppm and the gas flow rate discharged from the tank pressure safety system is 15610 kg/h. Details are as follow: gas release through the pressure control valve and three pressure safety valves are 6554 kg/h and 9056 kg/h, respectively, as recorded in the valve datasheet, therefore, gas release through one pressure safety valve is 3018 kg/h, which corresponds to the valve datasheet.

3.2. Determination of Release Direction

A flame arrester is installed on the outlet of the discharge pipe connected to the pressure safety valve, Fig. (1), which changes the direction of release from being vertical, downwards. The computational fluid dynamics simulation is applied to determine the discharge direction with respect to the flame arrester and the wind direction. The inputs to the CFD model are based on the flame arrester vendor drawing and the data tabulated in Table 1. The results indicate that the direction of the discharge from safety valves is almost downwards.

The direction of gas discharge out of the tailpipe connected to a pressure control valve is horizontal, Fig. (1A).

3.3 Weather Conditions

The values of the Pasquill-Gifford table applied for wind speed and atmospheric stability at the time of the incident are tabulated in Table 2. The relative humidity 41%, the minimum ground temperature is 22.4°C and ambient temperature is 22.8°C, are applied as inputs for simulation. The meteorological data are verified against the corresponding values taken from the nearby station.

 

Table 1. Composition and flow of released gas

The mole percent

Composition

The mole percent

Composition

0.42

Cyclohexane

4.785

Methane

0.06

Toluene

18.86

Ethane

0.8

n-Heptane

31.84

Propane

0.3

n-Octane

9.07

i-Butane

0.08

p-Xylene

15.47

n-Butane

0.08

n-Nonane

4.68

i-Pentane

0.02

n-Decane

4.07

n-Pentane

0.00

C11+

0.165

Mcyclopentan

0.01

p-Xylene

0.11

Benzene

0.49

E-Mercaptan

2.02

n-Hexane

0.10

H2O

1.97

CO2

0.00

Nitrogen

4.6

H2S

15610

Flow (kg/hr)

47.65

Mw (gr/mol)

15

Pressure (mbar)

1.986

Density (kg/m3)

 

 

32

Temperature (˚C)

 

 

Figure 1. Gas discharge pipe connected to pressure relief valve

 

Table 2. Pasquill-Gifford stability conditions

Wind speed at 10 meters height

(m/s)

Nighttime conditions

Anytime Heavy overcast

>4/8

Low cloud

<3/8

Cloudiness

1.5

F

F

D

2.5

E

F

D

3.5

D

D

D

 

 


3.4 Effect of Changes in Altitude

Because at above MSL different sections of a Gas plant is of different heights; at the release point 74 m and detectors location 50 meters the vapor cloud dispersion is studied in the following two cases based on elevation change:

1. Z20: The elevation difference between detectors position is neglected and the ground level at release source is considered as zero in altitude, that is, the release point height is at 20 m from ground level and detectors. Wind speed is 1.5, 2.5 and 3.5 m for stability conditions of D, E and F, respectively, Table 2.

2. Z44: The elevation difference between detectors position in the closest most activated gas detectors is of concern and the plant ground level at detectors position is at zero altitude. The wind speed for change in elevation is corrected based on power equations developed by Hanna et al. (CCPS, 1999), equation (1).

 

(1)

where UZ is the wind speed at any Z height, U10 is the wind speed at 10 meters above the ground and P is a dimensionless coefficient the value of which depends on the stability of the atmosphere and Earth's surface roughness parameters (CCPS, 1999).

Because the accident occurred at 2 A.M., stability parameters D, E and F are applied in the simulation. The surface roughness parameter is considered 50 cm as recommended for mini-refineries (CCPS, 1999).

The wind speed correction is made for the Z44 case in accordance with the height difference of 24 m between the land surface at the dispersions source location and gas detectors position at heights through the following equation. The results are tabulated in Table 3.

 

(2)

Where, U*10 is the wind speed at an altitude of 10 meters height at the in gas detectors location, and U*z is the corrected wind speed at the same height above the ground in dispersion source location Fig. (2).

 

 

 

Figure 2. Plant elevation at different areas and its effect on correcting reference wind velocity

Table 3. Wind speed at 10 meters elevation from local ground level in case Z44 based on the speed of 1.5, 2.5 and 3.5 m/s at a height of 10 meters in the dispersion area, U*10

Wind speed: 3.5 m/s

Wind speed: 2.5 m/s

Wind speed: 1.5 m/s

P

Stability

2.9

2.1

1.25

0.15

D

-

1.6

-

0.35

E

-

1.3

0.76

0.55

F

 

 

 


3.5. Gas Detectors Location

The active detectors in the fire and gas system are identified and their location is marked on the plot plan. Gas concentration is calculated based on the distance between the sender and the receiver. The results are tabulated in Tables 4 and 5. The high and very high concentration alarms correspond to 0.6 and 1.2 LEL.M, respectively. Although the concentrations are not within the explosion range, they could lead to plant emergency shutdown and production halt.

3.6. Simulation

The fixed duration scenario of the software is applied for simulations, where, it is necessary to input the released gas mass and its emission duration. The parameters, like temperature and pressure, discharge height from ground level, discharge angle in relation to the horizontal direction and average time to apply the impact of fluctuations in wind direction on vapor cloud are determined. In this study, the averaging time of 600 s is selected because the activation time of gas detectors is significant (CCPS 1999). Concentrations considered here are selected from Tables 4 and 5. Gas dispersion from the pressure control and pressure safety valves are considered independent and the results are superimposed and validated. The final results are then compared with the results from the activation of flammable gas detectors. All the mentioned steps are expressed in Fig. (3).

4. Results and Discussion

The results of the simulation are tabulated in Tables 6 and 7 for Z20 and Z44 respectively. Acceptable cases are subject to 1.5D and 2.5E for the Z20 and 1.25D, 1.6E and 1.3F for the Z44 atmospheric conditions. These results are in agreement with the actual site data, Figs. (4A to 10A).

 

 

Table 4. Concentrations range identified by the flammable gas detectors

Unit number

Detector Number

Alarm Level

Transmitter-receiver Distance (m)

Expected vapor cloud concentration (ppm)

102

BGD001

High

110

105~210

102

BGD001

Very high

110

More than 210

102

BGD002

High

110

105~210

102

BGD002

Very high

110

More than 210

102

BGD003

not activated

24

Less than 480

101

BGD001

High

84

137~274

101

BGD002

High

90

128~256

101

BGD003

not activated

94

Less than 115

101

BGD004

High

102

123~226

140

BGD001

High

67

172~344

140

BGD002

High

75

154~308

122

BGD001

not activated

46

Less than 250

122

BGD002

High

88

131~262

122

BGD002

Very high

88

More than 262

122

BGD003

not activated

88

less than 131

122

BGD004

High

46

250~500

 

Table 5. Concentration and location of activated toxic gas detectors

Start time

End time

Concentration (ppm)

Detector tag Number

Unit number

2:28

2:29

10~20

TGD001

LER6

2:28

2:30

10~20

TGD003

LER6

2:30

2:34

10~20

TGD008

102

 

 

Figure 3. The schematic flow chart of simulation steps

 

Table 6. Results of the simulation of flammable gas for Z20 case

Total effect PSV & PCV

Min concentration identification at 107 ppm

Wind speed at 10 meters height (m/s)

Not acceptable

acceptable

PSV

PCV

 

2

2

1.5D

 

2

1

1.5F

 

3

3

2.5D

 

2

3

2.5D

 

1

1

2.5F

 

3

3

3.5D

(1)- Gas concentration distribution graphic range is greater than the real range, an unacceptable phenomenon.
(2)- Gas concentration distribution graphic range same as the real range, an acceptable phenomenon.
(3)-  Gas concentration distribution graphic range is smaller than the real range, the detectors non-activation phase.

 

Table 7. Results of the simulation of flammable gas for Z44 case

Total effect PSV & PCV

Min concentration identification at 107 ppm

Wind speed at 10 meters height (m/s)

Not acceptable

Acceptable

PSV

PCV

 

2

3

1.25D

Simulation limit for less than 1 m/s speed

0.76F

 

3

3

2.1D

 

2

3

1.6E

 

2

3

1.3F

 

3

3

2.9D

(1)- Gas concentration distribution graphic range is greater than the real range, an unacceptable phenomenon.
(2)- Gas concentration distribution graphic range same as the real range, an acceptable phenomenon.
(3)-  Gas concentration distribution graphic range is smaller than the real range, the detectors non-activation phase.

 

 

The simulation result of the atmospheric condition in 1.5D, as shown in Fig. (4), completely corresponds with the real conditions. Moreover, the vapor cloud reaches the earth's surface before the reduction of plant-level height. That the mechanism of vapor cloud concentration dilution at the beginning and after discharge from the release source, is mostly due to turbulence caused by the difference between vapor and wind speed is inevitable. This can as well be due to the vapor cloud movement in a direction perpendicular to the earth's surface allowing fresh air entrance. Consequently, disregarding the height reduction effect is acceptable because it occurs after the vapor cloud reaches the earth's surface, although it introduces an error in the simulation. It is observed that conditions for activating a very high alarm level in the above conditions are provided.

Dispersion in the case of 2.5E, which is in accordance with the real situation, is shown in Fig. (2A). Thus, the vapor cloud reaches the ground before the height reduction. However, the vapor cloud reaches the detectors after 196 s, which is faster than that of the 1.5D, 388 s. This is due to greater stability, leading to a lower distribution of gas concentration in perpendicular directions to the wind speed.

The emitted vapor/stream cloud from the pressure control valve rout discharge reaches the ground (in a distance from discharge tank) at a location where the plant level is reduced by 4m, Fig. (2), of course, this 4m in relation to the location of the detectors (20 m below this level) is not significant.

The total effect of gas dispersions in 2.5E, through safety and control valves, is sufficient concerning the alarm sensitivity of the level very high detected by gas detectors. Moreover, the meteorological report indicates a wind speed of 3 m/s, and sky cloudiness of 1 octa at the same time (Iran meteorological organization). This is in accordance with the atmospheric condition in 2.5E.

Gas dispersion subject to atmospheric conditions in 1.25D is shown in Fig. (4A). An elevation difference of 44 m is observed between the release source and the surface where gas detectors are installed like unit No. 102. Vapor cloud reaches the surface where the height is reduced two meters, here vapor cloud moves only 22 m in a direction perpendicular to the ground. Although the volume of 105 ppm of gas concentrations in the subject unit satisfies the high alarm condition for the detectors, the alarm condition of very high in this area and the alarm condition on unit No. 101 are not satisfied. This could be due to vapor cloud dilution as a result of fresh air entering the vapor cloud with a perpendicular orientation of 44 m height.

The explanation on 1.25D is valid for 1.6E and 1.3F cases as well, Figs. (5A and 6A), respectively. The vapor cloud reaches the surface where the reduction in the altitude of the location is only four meters, and, the vapor cloud detection in unit No. 101 occurs in the case of 1.3F, which is the reason for the difference between these cases.

Simulation of toxic gas dispersion is merely run to obtain valid atmospheric conditions of flammable gas dispersion. If the simulation results correspond to the recorded data in the fire and gas system of the refinery, Table 5, for the toxic gas, it would indicate another reason for the verification of the results therein.

The simulation of toxic gas dispersion from the pressure control and pressure safety valves subject to the atmospheric conditions in 2.5E is conducted at the Z20 case. The results of the dispersion from pressure control and pressure safety valves are shown in Figs. (7A and 8A), respectively.

The details of the three detectors activated in the entire area are tabulated in Table 5. Two detectors are activated at the entrance of the ventilation system of local equipment room No.6 (LER 6), though there are three detectors at this point. In unit No. 102, only one of the toxic gas detectors (No. 008) is activated which is de-activated three times in a short period, though there exist 9 detectors in this unit. In this context, it is expected that the concentration of hydrogen sulfide gas is close to the high alarm limit that is at 10 ppm.

The concentrations of 5, 10 and 20 ppm are applied to plot the concentration profile of the toxic vapor cloud.

 

 

 

Figure 4. The gas dispersion subject to atmospheric condition of 1.5D in Z20 from the pressure safety valve

 

The toxic gas dispersion from the pressure safety valve is shown in Fig. (7A). Although the concentrations of 5 ppm are observed in unit No. 101, the required concentration of 10 ppm to activate a high alarm only appears in unit No. 102. The vapor cloud dispersion from the pressure control valve only reaches unit No. 102 at a concentration of 5 ppm, Fig. (8A). As to gas release from the pressure control valve according to countors, there exists no gas cloud with 10 ppm concentration in the site. The simultaneous effect of the released gas from both the sources, the valves, indicates that gas dispersion subject to atmospheric conditions of 2.5E in the case of Z20 is in accordance with alarms recorded in unit No. 102.

As observed in Table 5, the detectors of LER 6 become activated at 2:28 a.m. and dis-activated at 2:30 a.m. Detector 008 in unit No. 102 becomes activated immediately after the detectors become de-activated in LER 6. The records indicate that the wind at 2:28 a.m. is directed towards the LER 6 and then, with a slight fluctuation, diverts towards unit No. 102. This change in wind direction is drawn as a curve.

Toxic gas emission through the safety valve is of only 5 ppm concentration in LER6 area. A concentration of 10ppm does not even get close to this area, Fig. (7A). As to 5ppm, the same holds for the control valve, Fig. (8A). None of these two sources on their own can activate the 001 and 003 detectors in LER6. Their simultaneous effect (superposition) leads to the generation of 10ppm concentration in this area, thus activation of the detectors at these levels.

To make sure that the valid atmospheric conditions at gas release time was 2.5E, the simulation is run at 1.5D as well. According to the toxic gas concentration diagram in Fig. (9A), the obtained results are not valid, because in this case at least 5 detectors would have been activated in unit 102, Fig. (9A). This indicates that gas release occurred at 2.5E atmospheric condition and Z20 case.

4.1. The Sensitivity Analysis of Gas Dispersion

During the simulation of gas dispersion in the previous stages, effective parameters on the results are identified to some extent. Simulation of gas dispersion from both the pressure safety and control valves are of different discharged gas volumes, accordingly, as to the results, the gas discharge angle effect is of high importance.  The effect of changing some parameters, like discharge elevation, atmospheric stability and wind speed on the gas cloud dispersion is assessed during the simulation.

Gas discharge elevation and direction are the only effective controllable parameters on gas dispersion. Based on the main objectives of this study, that is, determining an appropriate solution and possible modification and introducing new design codes for safety systems to mitigate similar incidents, a sensitivity analysis is run the gas discharge angle.

The sensitivity analysis of the gas discharge angle is run on a vertical upward direction for both the sources. The simulation results in Fig 10A indicate that an upward shift in the gas discharge angle from the pressure control valve in the case of 2.5E prevents gas concentration to reach 105 ppm at any point in the plant.

Subject to the same condition, the range of gas concentration discharged upwards from the pressure safety valve is shown in Fig. (11A), and then compared with Fig. (2A), indicating that gas concentration and dispersion and the consequences therein are considerably reduced by changing the discharge angle upwards.

5. Conclusions and Recommendation

In the absence of in-situ meteorological atmospheric parameters applying their equivalents from Pasquill-Gifford table yields valid results in simulation. Gas dispersion intensity in the area is subject to ground level change, neglecting this elevation change between the release area and the location of gas detectors has no significant error on the simulation results if the vapor cloud reaches the ground surface before the occurrence of any considerable ground elevation changes. Atmospheric stability is a more sensitive parameter in vapor cloud dispersion in comparison with discharged gas elevation. The results indicate that at the time of the accident, the atmospheric conditions were at 2.5E. An upward change in vapor discharge angle leads to a significant reduction in vapor cloud dispersion.

Acknowledgement

Appreciations are extended to South Pars Gas Complex Co. for financial support and allowing having access to required data.

Alison M.G., Ju L.S., D.L., M.W., M.B., 2014. A risk assessment methodology for high pressure CO2 pipelines using integral consequence modeling. Process Saf. Environ. Prot. 92, 17-26.

Arunraj, N.S., Maiti, J., 2009. A methodology for overall consequence modeling in chemical industry. J Hazard Mater. 169, 556-574.

CCPS (Center for Chemical Process Safety), 1999. Guidelines for Consequence Analysis of Chemical Releases, A.I.O.C. Engineers,

Gant, S.E., Atkinson, G.T., 2011. Dispersion of the vapour cloud in the Buncefield Incident. Process Safety and Environmental Protection. 89(6), 391-403.

Gant, S.E., Kelsey A., McNally K., Witlox H. W. M., Bilio M., 2013. Methodology for global sensitivity analysis of consequence models. J Loss Prev Process Ind. 26, 792-802.

I.R. of Iran meteorological organization (http://irimo.ir/eng/wd/720-Products-Services.html)

Joyce T., Artur. Z., FrancesconiSávio.S.V., Viann.V., 2018. Modelling of source term from accidental release of pressurised CO2. Process Saf. Environ. Prot. 113, 88-96.

Khan, F.I., Abbasi, S.A., 1998. Techniques and methodologies for risk analysis in chemical process industries. J Loss Prev Process Ind. 11, 261-277.

Khan, F.I., Abbasi, S.A., 1999. Major accidents in process industries and an analysis of causes and consequences. J Loss Prev Process Ind. 12, 361-378.

Lees, F. P., 1996. Loss prevention in chemical process industries.London: Butterworth.

Meysami, H., Ebadi, T., Zohdirad, H., Minepur, M., Worst-Case identification of gas dispersion for gas detector mapping using dispersion modeling. J Loss Prev Process Ind. 26, 1407-1414.

Mishra, K.B., Wehrstedt, K.D., Krebs, H., 2013. Lessons learned from recent fuel storage fires. Fuel Process. Technol. 107, 166-172.

Mehr news agency (www.mehrnews.com/ news/3708114).

National Iranian petrochemical company (www.aparat.com/v/kAvGb/)

Norstrom, G.P., 1982. Fire/explosion losses in the CPI. Chem. Eng.Prog. 8, 78-80.

Parvini, M., Kordrostami, A., 2014. Consequence modeling of explosion at Azad-Shahr CNG refueling station. J Loss Prev Process Ind. 30, 47-54.

Pontiggia, M., Derudi, M., Alba, M., Scaioni, M., Rota, R., 2010. Hazardous gas releases in urban areas: Assessment of consequences through CFD modelling. J. Hazard. Mater. 176, 589-596.

Rad. A., Rashtchian. D., Badri. N., 2017. A risk-based methodology for optimum placement of flammable gas detectors within open process plants. Process Saf. Environ. Prot. 105, 175-183.

Seungkyu D., Chang J., Jeongpil P., Dongil S., En S, 2014. Quantitative risk analysis of fire and explosion on the top-side LNG-liquefaction process of LNG-FPSO. Process Saf. Environ. Prot. 92, 430-441.

Sharma, R.K., Gurjar, B.R., Wate, S.R., Ghuge, S.P. Agrawal, R., 2013. Assessment of an accidental vapour cloud explosion: Lessons from the Indian Oil Corporation Ltd. accident at Jaipur, India. J Loss Prev Process Ind. 26, 82-90.

Venetsanos, A.G., Bartzis, J.G., Würtz, J., Papailiou, D.D., 2003. DISPLAY-2: a two-dimensional shallow layer model for dense gas dispersion including complex features. J. Hazard. Mater. 99, 111-144.