CO2PipeHaz

Dispersion of accidentally released CO2Olav R Hansen J A Melheim O J Taraldset M IchardOlav R. Hansen, J. A. Melheim, O. J. Taraldset, M. IchardGexCon AS

CCS from Cradle to GraveCCS from Cradle to GraveThe technical and safety challenges

Birmingham, UK, 22-23 March 2012g , ,

Outline

• Background• Modelling challenges

FLACS CFD l• FLACS CFD solver• Particle modelling• Wind and dispersion modelling• Wind and dispersion modelling• Further work

Lake Nyos incident Cameroon

• Lake in volcanic crater• Lake in volcanic crater • Supersaturated water• 1986:

• Sudden release of CO2 from deep water• About 1.6 million tons CO2 released from the lake surface

CO fl d i t b ll d kill d b t 1700• CO2 flowed into nearby valleys and killed about 1700 people in a range of 25 kilometers

US statistics

CO d f EOR• CO2 used for EOR• 13 accidents in 22 years in 5600 km of pipelines (1986-2008)• 6 equipment failures 2 operator error 2 corrosion 3 unknown6 equipment failures, 2 operator error, 2 corrosion, 3 unknown • No reported injuries and fatalities

Background

• Hazardous gasg• Pure CO2 is lethal at high concentration. Presence of

impurities such as amines, H2S, CO,…• Pipe-lines will be located in populated areasPipe lines will be located in populated areas

• Unclear regulation • Risk assessment and risk management will be demanded

DNV d d ti 1• DNV recommended practice:1• The risks to people in the vicinity of the pipeline shall be robustly

assessed and effectively managed down to an acceptable level.M j id t h d i k t h ld b f d t• Major accident hazard risk assessment should be performed to provide estimates of the extent and severity…

1 DNV-RP-J202, April 2010. DESIGN AND OPERATION OF CO2 PIPELINES

Modelling challenges

• Supersonic two-phase CO2 jet with bli ti d d iti INERIS FLIE test with sublimation and deposition

• Condensation of humidity in air• Atmospheric boundary layer flow at

S esflashing propane 1.25 kg/s

various stability conditions• Modelling of terrain and vegetation and

their influence on the flow• Thermal stratification caused by cold

and dense cloud

FLACS – GexCon CFD model

• Release & Dispersion (gas, pool-spread & evaporation, flashing) • Explosion (main feature, wide functionality), far field pressure waves

Risk assessments mitigation accident/incident investigation and more• Risk assessments, mitigation, accident/incident investigation and more• Efficient geometry import from dgn-files (and some other formats)• Ventilation (including helideck studies, natural ventilation, HVAC)• About 100 end-users in 20 countries world-wide

GexCon explosion test3 kg methane1.5 bar overpressure

CO2 dispersion demo case

• Constant release of gaseous CO2(100 kg/s at -70°C)(100 kg/s at 70 C)

• Jet release above ground. Horizontal and aligned with the wind direction

• Neutral atmospheric conditions• Neutral atmospheric conditions• Wind: 5 m/s @ 10 m from south• Terrain imported from USGS DEM

and merged with manually added buildings

0.25% isosurface videoGround level concentrations(blue > 0.5%, red > 5%)

CO2 dispersion demo case

• High momentum jet gives good mixing with ambient air

• Cloud follows terrain and it disperses• Cloud follows terrain and it disperses slowly after jet dominated region

• Highly turbulent jet, then decay of turbulence and stratification caused by dense cloud

Two phase jet models

y≥

P0 Ps(T0)P0 Patm

≥>

yi

A1U1αg1

A2U2αg2

Jet expansion -Entrainment

Jet deflection -ImpingementFlashing

1. Homogeneous model• Gas phase is saturatedGas phase is saturated

• Dense gas behaviour of mixture

2. Lagrange particle tracking model2. Lagrange particle tracking model• Real particle diameter distributions

• Model particle motion and psublimation

Lagrange particle tracking

• Track parcels of particles (1 parcel = N particles)p p ( p p )• Drag and gravity forces are modelled• Two-way coupled motion

E ti / bli ti i d ll d• Evaporation/sublimation is modelled• Instantaneous velocity felt by particles must be modelled

(CFD solver provides mean velocity & turbulence parameters)(CFD solver provides mean velocity & turbulence parameters)

From left to right: φ=40μm -φ=80μm - φ=200μm

Particle dispersion model

• Solve Stochastic Differential Equation for instantaneous fluid qvelocity – Langevin Equation (Minier 1995, 2001)- Satisfy turbulence properties (time and length scales, etc)

• Validated for solid particles (Hardalupas et al 1989)• Validated for solid particles (Hardalupas et al., 1989)

Wind and dispersion modelling

• Wind conditions are important in the far-field• Pasquill classes used to characterize

atmospheric boundary layer (ABL) conditions:─ Unstable (A, B and C): Heat from ground ( , ) g

increases turbulence─ Neutral ABL (D): Shear forces dominate─ Stable (E, F, G): Stratified flow, little mixing

• Profiles for velocity, turbulence parameters and temperature are described on inlet boundaries

• Complex terrain and buildings result in more p gcomplex flows; stagnation, wakes, etc.

Validation cases

• Pure air wind tunnel experiments• Validate solver and turbulence model• Triangular ridge, cube mounted on ground, etc

• Outdoor wind experiments• Validate atmospheric boundary layer model• Askervein Hill, Bolund

Incre

• Outdoor dispersion experiments – passive gas• Validate transport model• Kit Fox, MUST, Prairie Grass, Manhattan

eased co

• Massive dense gas release experiments ─ Validate modelling of the influence of massive

omplexit g

releases on the flow field ─ Burro & Coyote, Maplin Sands, REDIPHEM

ty

• Massive large-scale CO2 release experiments• Most realistic, ultimate validation case• Liaohe Oilfield in 2012 by DUT (WP2.2)

Criteria for a “good” dispersion model

• Model Evaluation Protocol2 (MEP) for LNG is the most comprehensive approach p pp

• The MEP contains a structure for complete model evaluation => all models would need to be validated against key experimental databe validated against key experimental data and is a key part of the model evaluation.

• The parameters compared are:• The parameters compared are:─ maximum concentration along an arc at a

given distance from the sourcel d idth l t i─ cloud width along an arc at a given

distance from the source─ concentrations at specific sensor locations─ temperature at specific sensor locations

• Statistical Performance Measures (SPM) criteria have to be met to pass the MEPp

2http://www.nfpa.org/assets/files/PDF/Research/LNG_database_guide.pdf

Results: Unobstructed testsThe Maplin Sands tests (Shell 1980)

• 34 spills of liquefied gases onto the sea.

• Tests MP 27, 34 and 35 => continuous releases of LNG

• Rates ~20-30 kg/s - Wind from 5.5m/s to 10m/s

• Wind direction along array axis, neutral atmosphere

Results for MP 27: The plume and the maximum averaged concentrations at each row of sensors are shown.

• Very good results are obtained for the 3 tests

– No major improvements needed

Results: Unobstructed tests

The Burro and Coyote tests

• Test performed at the Naval Weapons Center, China Lake, California in the summer of 1980 p p

• 8 + 3 LNG spills onto water, 1.5m below terrain level in a 58m diameter pond.

• The terrain downwind spill pond sloped upwards at about 7 degrees for 80m before leveling outdegrees for 80m, before leveling out.

• Burro tests 3, 7,8, 9 and Coyote tests 3, 5 and 6 are considered. Neutral atmosphere except for Burro 8 (Pasq E)

• Release rates ~ 100 kg/s – Wind 1.8 (B8) to 10.5 m/s (C5)

Coyote 5 simulated plume129 kg/s LNG, 10.5 m/s wind

Results: Unobstructed tests

The Burro test => Initially poor results for Burro 8 (low wind)

• Gas cloud below first two 1m sensor arcs (57m and 140m)

• Good predictions at 400m and 800m arcs

• By simulation actual wind, not average wind, results improved

=> Validation challenging all details from tests are needed=> Validation challenging, all details from tests are needed

Statistical Performance Measures criteriaWind tunnel tests

Field tests

• FLACS generally well within the acceptance criteria. ─ Obstructed Falcon tests challenging, source not well defined?

October 2011: US Department of Transportation accepted FLACS as first CFD tool for LNG studies (NFPA-59A)

CO2 stack releases (not focus in CO2PipeHaz)

CCS facilities need to send CO2 to stack from time to time• Dense gas, will fall to ground during low winds• May often contain toxic components (H2S, amines, …)• Integral models can not predict low winds or geometry / terrain

S f d i b f• «Safe design» can be very unsafe• CFD modeling can investigate challenges and find optimal measures

(change of design, fans, stack extension, heating …)( g g g )

Example 5 m/s wind, 3 m/s wind, 2 m/s wind and 1 m/s wind

Ground level concentrations 2 m/s, 1 m/s wind

Summary

• Dispersion analysis of both accidental and controlled releases will most likely be required in risk assessment y qstudies before a pipeline can transport CO2

• Massive releases of cold CO2 in obstructed areas enforce the need for transient 3D CFD simulationsthe need for transient 3D CFD simulations

• Validation is essential for software predicting dispersion used for risk assessmentsused for risk assessments

• FLACS has shown that it predict dispersion of LNG vapour within the MEP criteria

B tt d l d d f t bl l i d─ Better models are needed for very stable low-wind conditions and obstructed area

• Large-scale validation experiments for massive releases of CO2 are very important

Acknowledgements & Disclaimer

The research leading to the results described in this presentation has received funding from the Europeanpresentation has received funding from the European Union 7th Framework Programme FP7-ENERGY-2009-1 under grant agreement number 241346.

The presentation reflects only the authors’ views and the European Union is not liable for any use that may be made of the information contained thereinof the information contained therein.

Modelling challenges

DNV RP-202: • In many assessments, integral models should provide acceptable

modelling capability, but in areas where the combined effects of topography, buildings, pits, etc. and the heavy gas properties of the released CO2 may have a significant effect on the exposure of people or livestock, more detailed simulations using advanced dispersion tools (e.g. Computational Fluid Dynamics (CFD)) h ld b id dshould be considered.