Available online at www.sciencedirect.com -

ScienceDirect Ener9y

Procedia

Energy Procedia 4 (2011) 2253-2260 :

www.elsevier.com/locate/procedia

GHGT-10

Modelling of discharge and atmospheric dispersion for carbon dioxide releases including sensitivity analysis for wide range of scenarios

Henk W.M. Witloxa*f, Jan Stenea, Mike Harpera and Sandra Hennie Nilsen

aDNV Software, Palace House, 3 Cathedral Street, London SE1 9DE, UK bStatoil, Research Park, Porsgrunn N0-3908, Norway

Abstract

Projects in carbon capture and storage technologies for energy production involve the transport of vapour, liquid and supercritical CO2 and CO2/hydrocarbon gas mixtures via pipelines and process systems with subsequent injection into wells, e.g. offshore under the seabed. In addition several chemical companies often store and transport large quantities of CO2 and this may also represent a hazard. There is a need to model potential loss of containment scenarios for risk assessment and design purposes for such installations. It is observed that several models used in quantitative risk analyses and hazard assessment studies are not able to take into account modelling of the thermodynamics of CO2 in case of accidental releases from dense or supercritical conditions. Statoil together with DNV therefore initiated a project to further improve the Phast code for modelling of CO2 releases. The work and methodology derived in this project have mainly been developed by Det Norske Veritas (DNV), but with significant co-operation and input by Statoil.

The consequence modelling package Phast examines the progress of a potential incident from the initial release to the far-field dispersion including the modelling of rainout and subsequent vaporisation. The original Phast 6.54 models allow the released chemical to occur only in the vapour and liquid phases. The new Phast 6.6 models were extended to also allow for the occurrence of fluid to solid transition in case of CO2 releases. This applies both for the post-expansion state in the discharge model, as well as for the thermodynamic calculations by the dispersion model. Here it is assumed that no solid deposition occurs on the ground. The current paper documents work regarding modelling by Phast 6.6 of discharge and atmospheric dispersion of carbon dioxide, including a detailed sensitivity analysis for a wide range of scenarios (base cases) including high-pressure cold releases (liquid storage) and high-pressure supercritical releases (vapour storage) from vessels, short pipes or long pipes. The objectives of this work were to examine the effect of input parameters on key output data, to ensure robustness of the models, and to identify further model improvements where deemed to be necessary. © 2011 Published by Elsevier Ltd.

Keywords: C02, consequence modelling, discharge, atmospheric dispersion, thermodynamics

1. Introduction

Carbon capture and storage represents a key technology for reducing CO2 emissions. Statoil has become a world-leader in its development and application. Statoil is currently involved in four large-scale commercial projects involving carbon capture with varying degrees of maturity: the Sleipner area in the North Sea, the Sn0hvit LNG production in Northern Norway, In Salah in Algeria and the carbon dioxide facility at the Mongstad refinery. Projects in carbon capture and storage technologies for energy production involve the transport of vapour, liquid and supercritical CO2 and CO2/hydrocarbon gas mixtures via pipelines and process systems with subsequent injection into oil wells.

Corresponding author. Tel.: +44 20 77166711 ; E-mail address: henk.witlox@dnv.com

ELSEVIER

doi:10.1016/j.egypro.2011.02.114

There is a need to model potential loss of containment scenarios for risk assessment and design purposes. In addition several chemical companies often store and transport large quantities of CO2 and this may also potentially result in a hazard.

The consequence modelling package Phast examines the progress of a potential incident from the initial release to the far-field dispersion including the modelling of rainout and subsequent vaporisation. The original Phast 6.54 discharge and dispersion models allow the released chemical to occur only in the vapour and liquid phases. The new Phast 6.6 models were extended by Witlox et al. [1] to also allow for CO2 releases the occurrence of fluid to solid transition (during expansion to atmopsheric pressure in discharge calculations) and solid to vapour transition (during subsequent dispersion calculations). Furthermore error/warning messages are added to the discharge models in case solid effects are erroneously not accounted for upstream of the orifice. The reader is referred to Witlox et al. [1] for full details of this extension, and therefore only a brief summary of the CO2 modelling is included in this section.

In case of vessel/short-pipe releases calculations in Phast are carried out by the steady-state discharge model DISC or the time-varying model TVDI. These models calculate the initial expansion from the stagnation conditions to the orifice conditions, and the subsequent expansion from the orifice conditions to the atmospheric conditions; see Figure 1. For long pipelines the time-varying models GASPIPE (vapour releases) and PIPEBREAK (liquid releases) are adopted. For time-varying releases (TVDI, GASPIPE, PIPEBREAK), the model TVAV averages mass flow rates, postexpansion velocities and post-expansion solid fractions over a specified duration. The atmospheric dispersion calculations are carried out using the Phast dispersion model UDM based on these averaged discharge data (constant flow rate of finite duration).

The UDM invokes a thermodynamics model for mixing of the released material and the ambient air. This model calculates the phase composition and temperature of the mixture at the cloud centre-line. The default thermodynamic model for a two-phase vapour/liquid release includes modelling of droplet trajectories (starting from the initial droplet size following atmospheric expansion) and droplet rainout. For CO2 the stagnation pressures will be very large and therefore the initial solid particle is expected to be very small (initial fine mist of CO2). Furthermore the atmospheric boiling point is very low (-78.4oC) and therefore the solid particles are expected to evaporate very fast. As a result for the mixing of solid/vapour CO2 with air, the UDM thermodynamics model assumes homogeneous equilibrium without deposition of the solid CO2 onto the substrate. Thus trajectories of solid particles are not modelled.

leak orifice atmosphere

vessel

(stagnation)

CO2 plume

vapour-plume centre-line

SUBSTRATE

Figure 1. Discharge modelling (DISC/TVDI/GASPIPE/PIPEBREAK/ATEX) and dispersion modelling (UDM)

The current paper summarises the results of a detailed sensitivity analysis on Phast 6.6 results for discharge and atmospheric dispersion of carbon dioxide. This includes a wide range of scenarios (base cases) including high-pressure cold releases (liquid storage) and high-pressure supercritical releases (vapour storage) from vessels, short pipes or long pipes. The objectives of this work were to examine the effect of input parameters on key output data, to ensure robustness of the models, and to identify further model improvements where deemed to be necessary.

In this paper first the results of the base cases are presented, with conditions that are relevant for e.g. the Sn0hvit offshore development in the Barents Sea (cold liquid CO2 release) and the In Salah gas field in the Algerian Sahara (hot CO2 vapour release). Figure 2 below shows an illustration of the Sn0hvit installation. Finally the results of input variations for a selected base case are presented. The effect of the input variations on the discharge results (flow rate, post-expansion solid mass fraction) and dispersion results (cloud centre/line height, concentration, temperature, solid mass fraction) is reported and explanations are given to clarify this effect.

2. Base cases

The adopted base cases are related to the Sn0hvit offshore development in the Barents Sea (cold liquid CO2 release)

and the In Salah gas field in the Algerian Sahara (hot CO2 vapour release):

- BC1. (Barents Sea conditions, relevant for Sn0hvit; DISC/TVDI) cold liquid release from leak of tank (storage temperature 10oC, storage pressure 60 barg, inventory 30000 kg, hole size 50 mm)

- BC2. (Barents Sea conditions; DISC/TVDI) cold release from short pipe (length 10 m, diameter 10 cm) attached to above tank

- BC3. (Hot desert conditions, relevant for In Salah; DISC/TVDI) as above, but now hot vapour release (40oC)

- BC4. (Barents Sea conditions; PIPEBREAK) cold release from full-bore rupture in middle of long pipe (10oC, 150barg, 153 km length, 8" diameter, 23kg/s pump flow rate)

- BC5. (Hot desert conditions; GASPIPE): hot vapour release from full-bore rupture in middle of long pipe (75oC, 200barg, 20km length, 8" diameter)

It must be noted that most of the Sn0hvit transport pipeline is below the sea surface, but for the base cases all scenarios

are assumed to be on-shore, above the sea surface.

Figure 2. Illustration of the CO2 transport pipeline at Sn0hvit

All releases are pointing horizontally in the downwind direction and are at 1m elevation. The dispersion takes place over flat terrain with presumed surface roughness of 0.1m. The surface and ambient air temperature are taken as 10oC (Barents Sea conditions) or 20oC (hot climate conditions). Furthermore neutral stability class D5 is presumed with 70% humidity and ambient pressure of 1 atm. It is indicated above which discharge models are applied for each base case. The results of the time-varying discharge calculations by TVDI, GASPIPE or PIPEBREAK are illustrated by Figure 3. It is seen that the initial release rate is much larger (due to larger release diameter) and the release rate reduces much quicker for the long pipeline releases than for the leak and short pipe releases.

time (s)

Figure 3. Predictions of time-varying release rate for base cases

/12 — 1 - coldleak 2 - spipcold 3 - spiphot — 4 - Ipipcold — 5 - lpiphot

0.36 0.32 0.28

I 0.24

> 0.16

j 0.12

0.08 0.04 0

10 100 Downwind distance (m)

10 100 Downwind distance (m)

10 100 Downwind distance (m)

0.9 0.8

i07 0.6

® 0.5

— 1 - coldleak 2 - spipcold — 3 - spiphot — 4 - lpipcold — 5 - lpiphot

10 100 Downwind distance (m)

Figure 4. UDM dispersion predictions versus downwind distances for base cases

The data input to the dispersion calculations were chosen to be the averaged values over the first 20 seconds for the long pipe base cases, and the initial value for the other base cases.

Figure 4 plots the UDM dispersion predictions versus downwind distance. The CO2 solid fraction is seen to evaporate at a downwind distance which is significantly upwind of the distance (Figure 3b) at which the plume centreline reaches the ground (Figure 3d). This appears to justify the assumption of no solid CO2 deposition on the ground for these base cases. For the cold releases there is a significant amount of initial solid CO2 and the evaporation of the CO2 is seen to lead to significant cooling of the cloud (Figure 3a). The maximum centre-line concentrations for the base cases are seen to increase with the base case flow rates (Figure 3c).

3. Results of sensitivity study

The input data to both the discharge and dispersion models were varied for the above base cases. The effect of the input variations on the discharge results (flow rate, post-expansion solid mass fraction, post-expansion velocity, etc.) and dispersion results (cloud centre/line height, concentration, temperature, solid mass fraction, etc.) was investigated and explanations were given to clarify this effect. Full details of this work are reported by Witlox et al. [2] and only selected results for base case 1 (BC1; cold vessel leak) are included in this section.

Variation of input to discharge models

Figure 5 and Figure 6 include DISC discharge and UDM dispersion predictions with variation of storage pressure. It is seen that with increasing pressure the flow rate and post-expansion solid fraction increase, while a discontinuity appears at the saturated vapour pressure (44 barg) below which the storage phase is vapour and above which it is liquid. The base case value of 60barg is labelled in the legend as '60_BC' in Figure 6. For increasing pressures we have increasing plume solid fractions, increasing plume cooling, and increasing plume concentrations. Plume touchdown is further downwind because of the larger initial momentum.

Gauge pressure (Pa)

Figure 5. BC1 (Variation of storage pressure) - DISC discharge predictions

& 240 £

Downwind distance (m)

— 1barg i

— 2barg

—16barg o

— 32barg 8

-B-60 BC « 0.1

—100barg 1=

—150barg .is

—200barg S

—240barg C0.01

Downwind distance (m)

Downwind distance (m)

— 1barg 2barg 4barg 8barg 16barg 32barg -B-6CLBC

Downwind distance (m)

(b) (d) Figure 6. BC1 (Variation of storage pressure) - UDM dispersion predictions

Figure 7 and Figure 8 consider the case of variation of storage temperature. For increasing storage temperature, the flow rate approximately remains constant for the colder liquid temperatures and decreases for increasing supercritical vapour temperatures. The post-expansion solid fraction decreases. Again there is a discontinuous transition between liquid and vapour at the saturated vapour temperature (296K). The aerosol at colder storage temperature evaporates further downwind but still well before plume touchdown. Higher concentrations occur with the higher flow rates.

■ A Mass release rate (kg/s)

-•-Solid mass fraction of CO2

*......................................................

250 270 290 310 330 350 370 390 410 Temperature (K)

Figure 7. BC1 (V ariation of storage temperature) - discharge predictions

Figure 9 and Figure 10 consider the case of variation of hole size. It is seen that with increasing hole size the flow rate increases, the post-expansion solid fraction remains constant, and the release duration reduces. The TVDI flow rate reduces little with time because of the presumed presence of a pressurizing gas. For increasing hole sizes the increasing flow rates result in a large amount of solid aerosol evaporating less fast and there is more plume cooling. For the large hole sizes the distance at which the aerosol evaporates is seen to be downwind of the point at which the plume centre/line hits the ground. This may mean that there may possibly be solid deposition and therefore our assumption of no deposition may be conservative, i.e. leading to an over-prediction of concentrations. Furthermore with increasing hole sizes (flow rates) the centre-line peak concentration increases.

Orifice diameter (m)

0.35 0.3 ; 0.25 :

0.2 ; 0.15 : 0.1 j 0.05 C 0

(a) TVDI time-varying flow rate

(b) DISC initial flow rate and CO2 solid fraction

Figure 9. BC1 (Variation of hole size) -discharge predictions

Downwind distance (m)

Downwind distance (m)

(b) (d) Figure 10. BC1 (Variation of hole size) -dispersion predictions

Variation of input to dispersion models

The following dispersion input parameters are not input to the discharge models and therefore only have an effect

on the dispersion results:

• Release height. Plume touchdown further downwind with increasing release height. Lower concentrations because of increased jet entrainment for more elevated plume

• Release direction. Plume touchdown further downwind with increased upwards plume release angle. Reduced concentrations and reduced solid CO2 aerosol fractions

• Stability class. In near-field close results because of jet-dominated dispersion. In far-field, for larger wind speed touchdown further downwind and lower concentrations, and for more stable conditions larger concentrations

• Ambient temperature. Quicker evaporation of solid aerosol with increasing temperature, and larger concentrations because of relatively more heavy plume

• Humidity. Faster CO2 evaporation because of water condensation with increasing humidity; limited effect on concentration

• Dispersion on land or water. No effect since CO2 evaporated before plume touchdown

• Substrate temperature. Increasing surface temperature leads to a hotter and less heavy plume, lower concentrations

• Surface roughness. Increasing surface roughness has no effect in near-field, lower concentrations in far-field

The above results are specific to the chosen base case, and other conclusions may apply for other cases. See Witlox et al. [2] for a further detailed analysis and comprehensive results and explanations for the other base cases.

4. Conclusions and recommendations for future work

A detailed sensitivity analysis has been carried out for a wide range of scenarios (base cases) including high-pressure cold releases (liquid storage) and high-pressure supercritical releases (vapour storage) from vessels, short pipes or long pipes. The objectives of this work were to ensure robustness of the models, to examine the effect of input parameters on key output data, and to identify further model improvements where deemed to be necessary.

Overall the models were found to be very robust expect for occasionally limited numerical problems near the critical point (leak BC1), and convergence problems for very large pressure (hot short pipe BC4). The reported dispersion calculations reported in the current paper are all based on a constant flow rate. However for long pipelines time-varying effects of dispersion become more important. A more accurate method of time-dependent dispersion modelling forms part of ongoing work.

It was found from the sensitivity analysis that for elevated horizontal releases solid deposition of the CO2 is very unlikely except for very low elevation heights and/or very large hole sizes or pipe diameters. This justifies the (conservative) assumption that all solid remains in the cloud. The currently adopted method is considered to be conservative related to hazard distances.

It is recommended to await validation against experimental data (to be made available in the public domain), before possibly considering further improved modelling of solids in the discharge and dispersion models. The discharge calculations show the need for experiments close to real scale, since at larger scale CO2 solid deposition and the formation of solid CO2 pools may be more likely. Experimental data should also include varying release directions, including downward directed releases and effect of the jet impinging on obstructions. The currently adopted method may not be conservative in case the cloud resulting from the pressurised CO2 release would be trapped by obstructions. This may lead to a heavy-gas ground-level cloud leading to larger concentrations and consequently increased probability of fatality or injuries.

Acknowledgements

Financial support of the work by DNV Software reported in this paper was provided by Statoil. The contents of this paper including any opinions and/or conclusions expressed, are those of the authors alone and do not necessary reflect the policy of Statoil and/or DNV Software.

References

[1] Witlox, H.W.M., Harper, M., and Oke, A., "Modelling of discharge and dispersion for carbon dioxide releases", J. Loss Prev. Proc. Ind. 2009; 22:795-802

[2] Witlox, H.W.M., Stene, J., and Harper, M., "Modelling by Phast of discharge and atmospheric dispersion of carbon dioxide including detailed sensitivity analysis", Contract 96507000/98509900 by DNV Software for Statoil, December 2008