The Effects of Uncertainty on Quantified Risk Assessment for LNG Facilities

1
The Effects of Uncertainty on Quantified Risk
Assessment for LNG Facilities

• CCPS Meeting
• 24-26 April, 2006

2
Contents

• Introduction
• Worked example
• Sensitivity studies
• Concluding comments

3
Introduction

• Canvey LNG Terminal public inquiry (1982)
• SRD assessment
• British Gas assessment
• Used same approach
• Agreed on list of scenarios
• Estimate frequency and consequences of each
  scenario

4
Introduction

• Conclusions
• Differences of factor of 2 arise through use of
  different consequence models
• Potentially bigger differences in failure
  frequencies
• Large spills dominate risk to society
• Note Much subsequent work has assessed
  consequences not frequencies

5
Introduction

• So what would happen if we repeated the exercise
  now?
• What really matters in deciding individual and
  societal risk
• On-site?
• Off-site?

6
Worked Example
7
Worked Example

• Select scenarios for import terminal
• LNG jetty area
• LNG import pipe work
• LNG storage area
• LNG export pipe work
• Process area
• Natural Gas export area

8
Worked Example

• Add a realistic nearby population
• Loughborough, England

LNG Terminal
9
Worked Example

• Use generic failure frequencies
• Use lengths of large diameter pipe work on site
• Assume size of LNG tanks and transfer lines, size
  of ships
• Implies frequency of ship visits
• Use ACDS/BP model for ship failure frequency
• Use proprietary data to infer jetty release
  frequency

10
Worked Example

• Evaluate Consequences of Release
• Assume wind rose
• Perform calculations typically for 5 wind speeds,
  12 wind directions
• Use different inventory in tanks
• Assume different release directions
• Differentiate between times when ships are
  present and not present
• Differentiate between indoor/outdoor and
  day/night (population distribution)

11
Worked Example

• Combine frequency of event and consequences to
  determine
• Individual Risk
• Societal Risk

12
Worked Example

• Contours of location specific risk (LSR) from all
  scenarios and realizations
• People are present 100 of the time and are
  located outside.

13
Worked Example

• Societal Risks
• F-N curves for each scenario
• Sum over all scenarios

14
Sensitivity Studies

• Examine uncertainty arising from LNG behaviour
• Dense gas behaviour
• Boil-off rate
• Burning rate
• Explosions
• Examine sensitivity to event behaviour
• Size of hole created by initiating event
• Immediate vs. delayed ignition
• Duration of release and mitigation
• Escalation from smaller releases

15
Sensitivity - Dense Gas

• LNG dispersion - dense in near field
• What about further out
• Heat transfer from surface
• Humidity

16
Sensitivity Dense Gas

• Failure of LNG export pipeline
• 5 m/sec and D atmospheric conditions
• Sensitivity to
• Heat transfer modelling
• Humidity

Dispersion
Case
Distance Relative
to Base Case
Base Case
1.00
No heat transfer
1.26
No heat transfer, low
1.38
absolute humidity
17
Sensitivity Boil-off Rate

• Heat transfer from the sea to a spreading LNG
  pool for one size of ship spill
• 50 kW/m2 used for analysis
• 25 and 100 kW/m2 cases examined

18
Sensitivity Boil-off Rate
Figure 6 from paper

• Not too sensitive
• Bigger cloud for shorter time vs. smaller cloud
  for longer
• Middle case is marginally worse

19
Sensitivity Pool Fires

• Mass burning rate for pools
• Back radiation from flame
• Heat from underlying surface
• On land
• Up to 0.14 kg/m2/s has been measured
• On sea
• Heat flux from surface could enhance land value
• Could be approximately double

20
Sensitivity Pool Fires

• Single LNG storage tank release case
• Bigger fire for less time vs. smaller fire for
  longer time
• Low mass burning rate gave slightly higher
  societal risks

LSR Contours
21
Sensitivity - Explosions

• Studies suggest LNG vapour clouds dont detonate
• Produce less risk than equivalent sized LPG cloud
• Less congestion present at LNG import terminal
  compared to an LNG export terminal
• Warning at liquefaction facilities
• Refrigerant Storage
• Large inventories of volatile materials in
  process
• Release in the congested process area can lead to
  potentially significant consequences

22
Sensitivity LNG Behavior

• Conclude
• Effect on risk for a large spill may be
  counter-intuitive
• Cant prejudge outcome by guessing what is worst
  case for risk
• Location specific risk may be very sensitive to
  assumptions
• Societal risk less so

23
Sensitivity Event Behaviour

• Issues
• Size of hole created by initiating event
• Immediate vs. delayed ignition
• Duration of release and mitigation
• Escalation from smaller releases
• These are the parts of the event tree that are
  uncertain

24
Size of Release

• Large ship spill
• Sensitivity of risk to assumed size of release
• One tank 1m2 hole
• One tank 6m2 hole
• All tanks 6m2 hole

25
Size of Release

• Maximum consequence distances change
  significantly
• Reflected in vulnerability of people to event -
  if it happens

26
Size of Release

• . but ACDS suggests 9091 ratio of failure
  sizes
• And catastrophic releases more likely to ignite
  immediately (?)
• So effect on catastrophic risk is less than might
  be thought
• Depends on weighting of N, contribution to PLL
  changes to 49, 36 and 15 respectively, when N2
  is applied

27
Duration of Release

• Spill at Jetty, with 5 sec (small) and 90 sec
  (medium) releases for one loading arm and total
  transfer rate (large) release
• Assume similar 9091 split of smallmediumlarge
• In this case, small releases contribute most to
  expected loss of life

Figure 1 from paper
28
Sensitivity Event Behaviour

• Conclude different ways the event behaves can
  make a big difference
• Need to consider whole of event tree before
  deciding on what is dominant contributor

29
Concluding Comments

• Have we moved forward in 20 years?
• Yes, but
• Important to concentrate on
• Events that cause biggest risk
• Factors that can change this risk
• Not always obvious what impact changes have
  event frequency has linear impact on risk whereas
  modeling parameters may produce less than linear
  changes
• Need to consider causes and effects in holistic
  way
• E.g. Release cause and ignition probability