An Approach To Improving The Physical And Cyber Security ... · An Approach To Improving The...

An Approach To Improving ThePhysical And Cyber Security Of ABulk Power System With FACTS

Mariesa Crow & Bruce McMillinSchool of Materials, Energy & Earth Resources

Department of Computer Science

University of Missouri-Rolla

Stan AtcittyPower Sources Development Department

Sandia National Laboratories

FACTS

FACTSPhysicalAttack

NaturalFaults

Sandia is a multi-program laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United StatesDepartment of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000.

ACKNOWLEDGMENTS

Funded in part by the Energy Storage Systems Program of the U.S. Department Of Energy (DOE/ESS) through Sandia National Laboratories (SNL).

Problem Motivation

Prevent Cascading failures:

2003 BlackoutCauses

Physical & Cyber contingenciesDeliberate disruption

HackersHackersTerrorist ActivityTerrorist Activity

Proposed Solution

Flexible AC Transmission Systems (FACTS)Power Electronic ControllersMeans to modify the power flow through a particular transmission corridorIntegration with energy storage systems

US FACTS Installations

San Diego G&E/STATCOM/100 MVA

Mitsubishi

Eagle Pass (Texas)Back-to-back HVDC

37 MVA/ ABB

CSWS (Texas)STATCOM/ 150

MVA / W-Siemens

Austin EnergySTATCOM/ 100MVA

ABB

AEP/ Unified Power Flow Controller /100 MVA/ EPRI

TVASTATCOM/ 100MVA

EPRI

Northeast Utilities/ STATCOM/ 150 MVA/

Areva (Alstom)

NYPA/ Convertible Static Compensator/

200 MVA

Vermont Electric/ STATCOM/ 130 MVA/ Mitsubishi

Communication and coordination

Scheduling - Distributed Long-Term controlInteraction – Local Dynamic control

Vulnerabilities of the combined physical/ cyber systemRecovery and protection from physical faults and/or cyber attacks and/or human error

Decentralized Infrastructures

Identify cascading failure scenarios for

test systems

Cascading ScenarioOutage 48-49

45

46

W.Lancst Crooksvl

47

G69

68

G

G

49

66

65

Zanesvll48

50

Philo

MuskngumN

MuskngumS

67

G

Natrium

Kammer

44WMVernon

N.Newark

SpornW

Summerfl

62

64

37

34

36

NwLibrty39

40 41 42

43

S.Kenton

38

S.TiffinWest End Howard

Rockhill

EastLima

Sterling

Portsmth

Bellefnt 74 75SthPoint

CollCrnr23

GTannrsCk

Trenton

24

Portsmth

Hillsbro72

70

71

Sargents73

NPortsmt

WLima35

SpornE

54

51

Cascading ScenarioOutage 48-49

45

46

W.Lancst Crooksvl

47

G69

68

G

G

49

66

65

Zanesvll48

50

Philo

MuskngumN

MuskngumS

67

G

Natrium

Kammer

44WMVernon

N.Newark

SpornW

Summerfl

62

64

37

34

36

NwLibrty39

40 41 42

43

S.Kenton

38

S.TiffinWest End Howard

Rockhill

EastLima

Sterling

Portsmth

Bellefnt 74 75SthPoint

CollCrnr23

GTannrsCk

Trenton

24

Portsmth

Hillsbro72

70

71

Sargents73

NPortsmt

WLima35

SpornE

54

51

Cascading ScenarioOutage 48-49

45

46

W.Lancst Crooksvl

47

G69

68

G

G

49

66

65

Zanesvll48

50

Philo

MuskngumN

MuskngumS

67

G

Natrium

Kammer

44WMVernon

N.Newark

SpornW

Summerfl

62

64

37

34

36

NwLibrty39

40 41 42

43

S.Kenton

38

S.TiffinWest End Howard

Rockhill

EastLima

Sterling

Portsmth

Bellefnt 74 75SthPoint

CollCrnr23

GTannrsCk

Trenton

24

Portsmth

Hillsbro72

70

71

Sargents73

NPortsmt

WLima35

SpornE

54

51

Cascading ScenarioOutage 48-49

45

46

W.Lancst Crooksvl

47

G69

68

G

G

49

66

65

Zanesvll48

50

Philo

MuskngumN

MuskngumS

67

G

Natrium

Kammer

44WMVernon

N.Newark

SpornW

Summerfl

62

64

37

34

36

NwLibrty39

40 41 42

43

S.Kenton

38

S.TiffinWest End Howard

Rockhill

EastLima

Sterling

Portsmth

Bellefnt 74 75SthPoint

CollCrnr23

GTannrsCk

Trenton

24

Portsmth

Hillsbro72

70

71

Sargents73

NPortsmt

WLima35

SpornE

54

51

FACTS Placement and Control

FACTS Control

Distributed Long-Term control algorithms for FACTS settings

Run by each processor in each FACTSAlternatives

MaxMax--flow algorithms flow algorithms Local optimizationsLocal optimizationsAgentAgent--based frameworkbased framework

AssessmentReduction of OverloadsReduction of OverloadsComputabilityComputability

FACTS Placement

PlacementPlace few FACTS in a large network for maximum benefit

Evolutionary Algorithms (EAs) will be used to place FACTS devices in the network

kn

line

line

lineline

yContingencyContingenc S

SPPI ⎟

⎟

⎠

⎞

⎜⎜

⎝

⎛⎟⎟⎠

⎞⎜⎜⎝

⎛= ∑∑

2

maxω

Performance Index Metric

Gradient Descent on PI Metric vs. Maximum Flow

FACTS Interaction Laboratory (FIL)

FIL OverviewConstruct a Laboratory System to Study and Mitigate

Cascading FailuresDeleterious effects of interacting power control devicesCyber Vulnerabilities

Hardware in the Loop (HIL)Real-time Simulation Engine

Simulate Existing Power SystemsSimulate Existing Power SystemsInject Simulated FaultsInject Simulated Faults

Interconnected laboratory-scale UPFC FACTS Device

Measure actual device interactionMeasure actual device interaction

FACTS Interaction LaboratoryArchitecture

FACTS

10 KVA

3332

31 30

35

80

78

74

7966

75

77

7672

8281

8683

8485

156157 161162

vv

167165

15815915544

45160

166

163

5 11

6

8

9

1817

43

7

14

12 13

138139

147

15

19

16

112

114

115

118

119

103

107108

110

102

104

109

142

376463

56153 145151

15213649

4847

146154

150149

143

4243

141140

50

57

230 kV345 kV500 kV

36

69

Simulation Engine

(multiprocessor)

FACTS/ESS

FACTS/ESS

FACTS/ESS

A/D D/A

A/D

D/A

A/D

D/A

Network

HIL Laboratory Interface

Machine 1

D/A output

A/D input

FACTS1

ControllableLoad

Machine 2

Machine 3

Power System Simulation Engine

ControllableLoad

ControllableLoad

FACTS2

FACTS3

D/A output

A/D input

D/A output

A/D input

FACTS – Flexible AC Transmission System Prototype Device

FACTS Interaction Laboratory

HIL Line

UPFC

Simulation

Engine

Cyber Fault Detection

Fault Tolerance

Define correct operation of the power system with FACTS/ESSEmbed as executable constraints into each FACTS/ESS computerFACTS/ESS check each other during operation of distributed control algorithms

Cyber Fault Injection

Attempt to confuse the FACTS embedded computersAttempt to disrupt the communication between FACTS embedded computersConfuse the power system’s operation

Error Coverage of Distributed Executable Correctness Constraints

(Maximum Flow Algorithm)

System Dynamic Control

Power Network Embedded With FACTS Devices

Tie-line flow

CONTROLAREA A CONTROL

AREA B

A decentralized power network embedded with FACTS devices can be viewed as a hybrid dynamical system (Differential-algebraic-discrete-event).

While the FACTS devices offer improved controllability, their actions in a decentralized power network can cause deleterious interactions among them.

Performance of FACTS controllerswith ideal observability

Uncontrollable modes in generator speeds due to device interactions

Project Benchmarks

Construction of HILDemonstration of Cascading FailuresPlacement and ControlHardware/Software ArchitectureCyber Fault DetectionDynamic ControlVisualization

Special Thanks

Imre Gyuk – DOE Energy StorageStan Atcitty – Sandia National LabJohn Boyes – Sandia National Lab