This document discusses infrastructure resilience and summarizes Dr. Sarah Dunn's research in this area. Her research aims to improve community resilience to hazards by developing techniques to identify vulnerable infrastructure systems and protect them. She uses network analysis and fragility curves to estimate damage from events like storms to electricity networks. The research forecasts consequences to infrastructure based on hazard intensity, exposure data, and fragility curves derived from historical fault data. This allows identifying areas most likely to be impacted and informing contingency planning.

1. Infrastructure Resilience
Dr Sarah Dunn
Planning for Future Extreme Events
Lecturer in Structural Engineering
“Full” research team: Sean Wilkinson, Russell Adams, Samuel Gonzalez Otalora, Alistair Ford, Hayley Fowler, Nikolas
Kirchner-Bossi, Joana Mendes (MetOffice), Erika Palin (MetOffice)

2. Introduction

3. Introduction
• Sept 2006 – July 2010
MEng Civil and Structural Engineering
• Sept 2010 – Feb 2014
PhD in Civil Engineering
‘An Investigation to Improve Community Resilience using
Network Graph Analysis of Infrastructure Systems’
• March 2014 – March 2015
EPSRC Doctoral Prize Research Fellow
‘Increasing Community Resilience to Climate Change
Impacts through Adaptation of Infrastructure Systems’
• March 2015 – Present
Lecturer Structural Engineering

4. Main Research Area
Aim of Research
To improve the resilience of our communities by developing techniques that can identify
fragile system architectures, recognize vulnerable areas within these systems and
establish methods that can help to protect them from hazard.

5. Fragility of Infrastructure Systems
North American Blackout, 2003
• Affected 50 million people, in 8 US states
• Up to 4 days to restore power
• Economic losses between $7 and $10 billion US Dollars
http://earthobservatory.nasa.gov/images/imagerecords/3000/3719/NE_US_OLS2003227.jpg

6. Fragility of Infrastructure Systems
http://www.huffingtonpost.com/2013/08/14/2003-northeast-blackout_n_3751171.html

7. UK Summer Floods - 2007
This event was caused by a period of extreme rainfall (the wettest since
rainfall records began in 1766).
https://www.metoffice.gov.uk/about-us/who/how/case-studies/summer-2007

8. UK Summer Floods - 2007
The resulting flooding caused direct
damage to over 55,000 home and
businesses.
Transport infrastructure – closure of
roads and railways due to flooding.
Electrical infrastructure – closure of
substations which were affected by floodwater
(including the closure of the Castle Meads
substation which left 42,000 people without
power for up to 24 hours).
Water infrastructure – closure of water
treatment works (including the closure of the
Mythe water treatment works which caused
350,000 people to be without access to mains
water supply for 17 days).
Resulting damage cost the UK economy over £3 billion (~$5.5 billion AUD).
OFWAT (2007). Water and sewerage services during the summer 2007 floods.
Cabinet Office (2008). The Pitt Review: Learning Lessons from the 2007 Floods. Cabinet Office. London.

9. UK Summer Floods - 2007
After this flood event a detailed report was commissioned, the Pitt
Review (2008), which called for ‘a more systematic approach to
building resilience in critical infrastructure’ and highlighted the need
for:
 Improved understanding of the level of
vulnerability to risk to which infrastructure
and hence wider society is exposed;
 More consistent emergency planning for
failures;
 Improved sharing of information at a local
level for emergency response planning.
Cabinet Office (2008). The Pitt Review: Learning Lessons from the 2007 Floods. Cabinet Office. London.

10. What is “resilience”?
‘The ability of a substance or object to spring back into shape’ or ‘the
capacity to recover quickly from difficulties.’
Oxford Dictionary
‘Measures of the persistence of systems and of their abilities to
absorb change and disturbance and still maintain the same
relationships between populations or state variables.’ (Holling 1973)
Ecology
‘The ability of the system to withstand a major disruption within
acceptable degradation parameters and to recover within an
acceptable time and composite costs and risks.’ (Haimes 2009)
Systems and Information Engineering
‘The uncertainty about and severity of consequences of the activity
given the occurrence of any types of events.’ (Aven 2011)
Risk Management

11. What is “resilience”?
The Resistance element is focused on providing
protection. The objective is to prevent damage
or disruption by providing the strength or
protection to resist the hazard or its primary
impact.
The Reliability component is concerned with
ensuring that the infrastructure components are
inherently designed to operate under a range of
conditions and hence mitigate damage or loss
from an event.
The Redundancy element is concerned with the
design and capacity of the network or system.
The availability of backup installations or spare
capacity will enable operations to be switched to
alternative parts of the network in the event of
disruptions to ensure continuity of services.
The Response and Recovery element aims to
enable a fast and effective response to and
recovery from disruptive events. The
effectiveness of this element is determined by the
thoroughness of efforts to plan, prepare and
exercise in advance of events.
Cabinet Office (2011). Keeping the Country Running: Natural Hazards and Infrastructure. Cabinet Office. London.

12. Event
Generation /
Trigger
Intensity
Calculation
Exposure
Information
Damage
Estimation
Consequence
Calculation
Consequence Forecast Modelling
Contingency
Actions
Dunn, S., Wilkinson, S. M., Alderson, D., Fowler, H., and Galasso, C., (2018) ‘Fragility curves for assessing the resilience of electricity networks constructed from an
extensive fault database’ Natural Hazards Review. 19(1).

13. • Weather event: wind storm, heatwave, rainfall
• Information: duration, intensity, location
Event
Generation /
Trigger
Consequence Forecast Modelling

14. Event
Generation /
Trigger
Intensity
Calculation
• Calculation of event intensity (wind
speed) at each individual component
location using hi-res weather forecasts
Showing the maximum wind speeds during the Burns Day Storm on the 24th January 1990, shown on a red-green scale,
where red indicates areas of high wind speed and green areas of low wind speed.
Consequence Forecast Modelling

15. Event
Generation /
Trigger
Intensity
Calculation
Exposure
Information
• Data regarding
infrastructure
component type
and location
Consequence Forecast Modelling

16. Event
Generation /
Trigger
Intensity
Calculation
Exposure
Information
Damage
Estimation
Consequence Forecast Modelling

17. Event
Generation /
Trigger
Intensity
Calculation
Exposure
Information
Damage
Estimation
Consequence
Calculation
0
5
10
15
20
25
100 200 300 400 500 600 700 800 900 1000
Liklihood
Estimated Economic Cost (£)
Consequence Forecast Modelling

18. Event
Generation /
Trigger
Intensity
Calculation
Exposure
Information
Damage
Estimation
Consequence
Calculation
Consequence Forecast Modelling
Contingency
Actions

19. Damage Estimation
0
1
2
3
4
5
-4 1 6P(f) Flood Depth (m)
This is achieved through the use of fragility, or vulnerability, curves, which define the
relationship between the magnitude of an event and the probability of failure for
individual components.
0
1
2
3
4
5
-4 1 6
P(f)
Wind speed (m/s)
Fragility curves can be derived using an empirical or analytical methodology and are
essentially the “key” to producing accurate forecasts of damage.

20. Event
Generation /
Trigger
Intensity
Calculation
Exposure
Information
Damage
Estimation
Consequence
Calculation
Consequence Forecast Modelling
Contingency
Actions

21. Application - Electricity Resilience
Wind Storms are the greatest cause of electricity outages and they occur on an almost
annual basis in the UK.

22. Application - Electricity Resilience
“…they could have done more to get
customers reconnected faster and to
keep them better updated on what was
happening…”
“Two electricity distribution companies
were affected more significantly…each
having almost 1100 incidents, affecting a
quarter of a million customers”
“…they had almost 16,000 customers
affected for more than 48 hours and, in
the worst case, some were without
supply for six days. “

23. How do electricity companies currently prepare
for storm related outages?
Weather forecasts provide
information at the scale of
1.5km…and yet we provide
information at the regional
level!
Can we do better?

24. Event – Weather Forecasts

25. Event – Weather Forecasts

26. Exposure Information
Total area covered: 12,456 km2
Total number of consumers: 1,040,174

27. Exposure Information
OverheadLinesUndergroundCables
TowersandPolesSubstations
Number of Substations: 195,000+Length of UGC: ~56,000km
Length of OHL: ~71,000km

28. Fault Information
• We have worked with one DNO in the
UK to develop fragility curves for their
overhead line components
• These were developed using empirical
fault data, held in the NaFIRS database
• This records data regarding the date,
cause of fault, duration, approximate
fault location and number of consumers
affected
• There are around 25,000 faults that are
included in our fragility curves.

29. Damage Estimation – 2 Methods
Method 1 - Umax Method 2 - WSI
We identify “wind storms” as periods of
time where the wind speed exceeded
17m/s.
We then calculate how many faults
occurred in this time period and assume
these faults were caused by the
maximum wind speed recorded in the
storm.
0
2
4
6
8
10
12
0
5
10
15
20
25
30
0 50 100 150 200 250 300
NumberofFaults
WindSpeed(m/s)
Time Series (hours)
Wind Speed
Threshold Value
Wind Storm 1
Wind Storm 2
Number of Faults
More “detailed” methodology which
accounts for the infrastructure density,
spatial location of the highest wind
speeds, duration of wind storm.
Calculated using:
We then plot this WSI value against the
number of faults in the wind storm to
generate the fragility curves.

30. Damage Estimation – 2 Methods
Method 1 - Umax Method 2 - WSI
y = 1E-14x10.051
R² = 0.9889
0
10
20
30
40
50
60
70
0 5 10 15 20 25 30 35 40
AverageNumberofFaultsper1000km
LengthofOHL
Wind Speed (m/s)
All Data
Average Number of Faults

31. Damage Estimation – WSI

32. Damage Estimation – WSI

33. Results to date – Faults (Comparison)

34. Results to date – Faults (WSI)
64%
17%
12%
5.21%
1% 0% 0%
0%
10%
20%
30%
40%
50%
60%
70%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability Number of Faults
Storm 20 area 12, t + 12
39%
19% 19%
13.56%
6%
2% 2%
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 12, t + 24
78%
11%
7%
2.87% 1% 0% 0%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 13, t + 12
78%
11%
7%
2.91% 1% 0% 0%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 13, t + 24
64%
17%
11%
5.23%
2% 0% 0%
0%
10%
20%
30%
40%
50%
60%
70%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 12, t + 0
78%
11%
7%
2.85% 1% 0% 0%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 13, t + 0

35. 78%
10%
7%
3.19% 1% 0% 0%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability Number of Faults
Storm 20 area14, t + 12
78%
10%
7%
3.18% 1% 0% 0%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area14, t + 24
78%
10%
7%
3.27% 1% 0% 0%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area14, t + 0
87%
8%
4% 1.06% 0% 0% 0%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 15, t + 0
79%
10%
6%
2.97% 1% 0% 0%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 15, t + 12
79%
10%
7%
2.91% 1% 0% 0%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 15, t + 24
Results to date – Faults (WSI)

36. 41%
19% 19%
12.83%
6%
2% 2%
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 16, t + 24
48%
20%
17%
9.84%
4%
1% 1%
0%
10%
20%
30%
40%
50%
60%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 16, t + 0
48%
20%
17%
10.12%
4%
1% 1%
0%
10%
20%
30%
40%
50%
60%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability Number of Faults
Storm 20 area 16, t + 12
0%
2%
22%
47.03%
23%
5%
1%
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 17, t + 24
36%
22% 20%
13.28%
5%
2% 2%
0%
5%
10%
15%
20%
25%
30%
35%
40%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 17, t + 12
44%
19% 18%
11.82%
5%
2% 1%
0%
10%
20%
30%
40%
50%
<=5 6-10 11-20 21-40 41-70 71-100 >100
Probability
Number of Faults
Storm 20 area 17, t + 0
Results to date – Faults (WSI)

37. Event
Generation /
Trigger
Intensity
Calculation
Exposure
Information
Damage
Estimation
Consequence
Calculation
Consequence Forecast Modelling
Contingency
Actions

38. Damage Estimation – WSI

39. Results to date – Consumers without Power
Relationship: y = ~300x
Therefore for every fault on the HV network, approx.
300 consumers are without power.

40. Results to date – Consumers without Power
Relationship: y = ~8x
Therefore for every fault on the LV network, approx.
8 consumers are without power.

41. Next Steps…
• The next step in this research is to turn the methodology into “live prediction”
mode and run continuously.
• We are working with the Met Office to turn weather forecasting data, into
infrastructure forecasting models.
• We would like to reduce the scatter in the fragility curves, by considering:
• Material used for OHL (copper or aluminium?)
• Terrain type
• Classification of urban/rural areas
• We would also like to consider other types of hazards (e.g. flooding)

42. Further reading…
Dunn, S., Wilkinson, S. M., Alderson, D., Fowler, H., and Galasso, C., (2018) ‘Fragility curves for
assessing the resilience of electricity networks constructed from an extensive fault database’
Natural Hazards Review. 19(1).

43. Other new papers since my last visit!
Dunn, S., Wilkinson, S. M., (2017) ‘Hazard Tolerance of Spatially Distributed Complex
Networks’ Reliability Engineering and System Safety. 157: 1-12.
Dunn, S., Wilkinson, S. M., (2016) ‘Increasing the resilience of air traffic networks using a
network graph theory approach’ Transportation Research Part E. 90: 39-50.
Dunn, S., Wilkinson, S. M., and Ford, A., (2016) ‘Spatial structure and evolution of
infrastructure networks’ Sustainable Cities and Society. 27: 23-31.

44. Acknowledgements
We would particularly like to thank Western Power Distribution for
access to their fault and asset databases, in order to complete the
analysis for this study. And also the Energy Networks Association for
their insights.
Also to EPSRC and NERC for funding this research.

45. Thank you.