CI Interdependencies: Real time disaster response capability

1
CI Interdependencies Real time disaster response
capability

• José R. Martí, KD Srivastava, and i2Sim TeamThe
  University of British Columbia

Complex Interdependent Systems Group
1
2
University of British Columbia

• JIIRP project Sponsored by PS (Public Safety
  Canada and NSERC)
• V2010 Olympics Sponsored by DRDC (Defence
  Research and Development Canada)

3
UBCs Multidisciplinary Team

• Electrical and Computer Engineering
• Civil Engineering
• Software Engineering
• Computer Science

• Business
• Geography
• Clinical Psychology
• Graphics and Multi-Media

12 Researchers 12 Graduate Students 2 Post
Doctoral Fellows 2 Research Engineer
4
Our Objective and Motivation
First priority during disaster situations is,
and should be, human survival
4
5
Human Vulnerability
5
6
Panic and Belonging
6
7
San Francisco Bay Earthquake of M6.8or Greater
Due Now!

• A major earthquake on the Hayward Fault, in a
  highly populated section of the San Francisco Bay
  Area, is due.
• The last major earthquake on the Hayward Fault
  was in 1868, 140 years ago
• Research by the U.S. Geological Survey (USGS)
  indicate the past five such earthquakes have been
  140 years apart on average.
• A Hayward Fault EQ will adversely impact up to 5
  Million people
• Damage will likely exceed 1.5 Trillion
• Up to 70 of the loss will be sustained in
  Alameda and Santa Clara Counties - The majority
  of that being in Alameda County

7
8
Vancouvers Juan de Fuca Plate
8
8
Source GSC
9
The Big One (M7-9) due this CenturyMedium ones
(M6-7) Due Now

• Juan de Fucas plate slid into the continental
  coast 400-500 years ago
• It is due time to slide back
• The magnitude is expected to be fairly large
  (VIII to X), THE BIG ONE
• Historically, nine moderate to large earthquakes
  have occurred (Mw 6-7) within 250 km of
  Vancouver in the last 130 years

9
10
Human Needs (Maslow)
10
11
Disaster Mitigation Timeline
1 Preparation
2 Response
3 Recovery
Emergency
Recovery
Months to years
Days to weeks
Hours to days
Days to months
Being
Esteem
Love/Belonging
Safety
Physiological
11
12
Individual Survival Needs Critical Sectors

• SURVIVAL TOKENS
• Water (suitable for drinking)
• Food (adequate for emergency situations)
• Body Shelter (breathable air, clothing,
  temperature, housing)
• Panic Control (hope, political and religious
  leaders, psychologists, media)
• Personal Communication (whereabouts of loved
  ones)
• Individual Preparedness (education)
• Sanitation (waste disposal, washing)
• Medical Care (medicines, physicians, nurses)
• Civil Order (fire fighters, police, army)

• CRITICAL SECTORS (CANADA)
• Energy
• Water
• Food
• Financial
• Communications
• Transport
• Health
• Safety, Order
• Government, Defence
• Manufacturing

13
System of Systems
14
Scope

• Systems Planning
• Time scale of weeks, months
• Statistical models, steady state models,
  long-dynamics models
• Policy planning
• Disaster Response
• Time scale of hours, days
• Urgency of saving human lives
• Infrastructures emergency response plans
• Emergency response management (EOCs)
• Real time models
• First Responders
• Ground zero actions

15
Disaster Response Plans

• During normal times, each infrastructure (power
  grid, telecom system, etc.) knows very well how
  to respond to problems in its own system send
  out repair crews, readjust operation, etc.
• Disaster response plans are normally developed
  assuming the other infrastructures will be
  available
• However, during large-scale disasters, multiple
  infrastructures are damaged simultaneously and
  individual response plans are not sufficient
• Vital survival tokens need to be delivered very
  rapidly to prevent panic

16
Each Infrastructure is Responsiblefor its
Internal Operation

• Each entity, be it a power network or a
  hospital, has its own models and internal modes
  of operation for normal times and for emergency
  times
• Models exist to simulate disaster events, e.g.,
  forest fires, floods, etc.
• We can separate disaster modelling from
  infrastructures operating modes
• i2Sim provides an integration environment to
  optimize the combined actions of the
  interdependent infrastructures
• Solution is very fast for real-time what-if
  scenarios

17
Resources Allocation
Black ? bad decision because hospital cannot
function without water Blue ? good decision to
optimize global objective
18
Fast Survival ResponseTemporary Islands
18
19
Coordination Control (C2)
Choices in redirecting water?
Choices in alternative roads?
19
20
Decision MakingLook-Ahead and Rewind Capability
Real World
A
Alternativeactions
No Action A (I2Sim)
A, B decision pointsDecision A- Take Action
A2 Decision B- Take Action B1
Action A1 (I2Sim)
Action A2 (I2Sim)
B
A
No Action B (I2Sim)
Screens at A- Real World- No Action A (I2Sim)-
Action A1 (I2Sim)- Action A2 (I2Sim)
Action B1 (I2Sim)
Action B2 (I2Sim)
20
21
I2Sim Real Time Platform
22
i2Sim Ontology

• Cells (Production Units)
• A hospital cell requires inputs electricity,
  water, doctors, medicines, etc. and produces
  outputs patients healed
• Channels (Transportation Unit)
• The electricity to the hospital is carried by
  wires, the water is carried by pipes, the doctors
  are carried by the transit system
• Tokens (Exchange Unit)
• Quantities that are the inputs and the outputs of
  the cells, e.g., water is a token, a doctor is a
  token, a phone call is a token
• Controls (Distributors, Aggregators)
• Interface the physical layer with the decisions
  making layer, e.g., if electricity supply is
  limited, how much should go to the hospital and
  how much to the water pumping station

23
Basic i2Sim Model
24
Regional Scaling
24
25
Reserve
Cell
Cell
x4(t)
distributor
Hospitalm 70
Power StationOperabilitym 60
x1(t)
x2(t)
x3(t)
x7(t)
x8(t)
x5(t)
x2 (t) a.x1 (t-Tau)
healedpeople
Channelelectrical
aggregator
x8(t)m x7(t)
x6(t)
Channelelectrical
ExternalSource
Only power operator needs details of power station
Cell
Channelwater
Water Stationm 80
ExternalSource
Channelwater
Cell
Residential m 40
Channelelectrical
26
Cell Model
27
Channel Model
28
Internal Details PrivateOnly External Operating
Modes Needed
29
Colour Code by DHS
HRT-090211
29
30
Human Readable Table (HRT)Water Pumping Station
hidden
31
Physical Modes and Resource Modes
Effective Operability
32
Models Granularity

• The HRTs can be built with fine granularity data
  or with very coarse data with no numerical
  problems in the solution
• High granularity data rarely available and not
  really needed for effective emergency response
• Choices by operating models are usually limited
  (e.g., power substation, hospital, etc.)

33
Cells State
Physical Operability (100)
Effective Operability (50) because of lack of
water
Physical Operability (50)
Effective Operability (0) because of lack of
electricity
34
Human Factors

• Can be incorporated the same way as physical
  damage, i.e., as physical operability reduction
• Doctors past their shift time will have slower
  reactions, as a result, the hospital output will
  be reduced
• Human errors can reduce output and also create
  accidents
• Accidents correspond to damage events

35
Events

• An event is an action that changes the
  operability of cells or channels
• Model is independent of what or who produces the
  event
• Damage event degrades operability
• Repair event upgrades operability
• Decisions change resources allocation at output
  distributors

36
Some Math
36
37
Linearized Thévenin Model
37
38
Transportation/Interdependencies Matrix
38
39
UBC CampusI2Sim Interdependencies Matrix
39
40
Sensitivity Analysis

• The well-known Sensitivity Network Approach can
  be directly applied to the interdependencies
  matrix
• Where h is some parameter in T or W

40
41
State Matrix

• System dynamics can be expressed in state-space
  form
• Where state matrix A represents the systems own
  dynamics and matrix B represents the state
  transitions forced by the excitation events
• Matrices A and B can be directly obtained from
  the systems transportation matrix

42
PC-Cluster for Large Systems
42
43
UBC Campus Test Case

• UBCs Vancouver campus is a small municipality
• 2,000 acres
• 50,000 daily transitory occupants
• 10000 full time residents
• own utilities
• Human and Physical layers were classified into
  19 types of cells and 7 types of channels

43
43
44
UBC Buildings Structural Damage
44
44
45
UBC Lifelines
45
45
46
Cells and Channels from Physical Map
46
47
Interdependent Damage Assessment
Overlaid (classical)
Interdependent (new)
47
48
Damage and Casualties
48
49
Multiple Events Simulation Flow

• Events
• Damage by flood
• Change distributor ratio
• Repair asset
• Human error
• Human tiredness
• ...

50
PC-Cluster Simulation
51
Timings

• Closed solution much faster than open iterative
  solutions (e.g., agent-based modelling) by two or
  three orders of magnitude
• As an example, a system of 3,000 cells with 15
  inputs/outputs per cell (45,000 state variables)
  for a 10 hr scenario with delta-t 5 minutes in
  a few seconds of computer time
• Interactive scenario playing is basically
  instantaneous
• Allows for look ahead and rewind for decision
  making in real time

51
52
Summary

• All infrastructures represented
• Models based on operability tables (HRTs)
• HRTs determined by physical damage and resources
  availability
• Decisions determine resources allocation
• Real time environment
• What if capability
• Off-line ? system design
• On-line ? training
• Real-Time ? disaster event management

53

• Thank You!