GTRI/DOE Disclaimer Cable Diagnostic Focused Initiative Regional ... CDFI Meeting.pdf · Cable...

1

1CDFI Meeting - Aug. 19-20 San Ramon, CA

Cable Diagnostic Focused InitiativeRegional Meeting

NEETRAC

Hosted byPacific Gas and Electric

San Ramon, CAAugust 19-20, 2009


GTRI/DOE Disclaimer• The information contained herein is to our knowledge accurate and reliable at

the date of publication. • Neither GTRC nor The Georgia Institute of Technology nor NEETRAC will be

responsible for any injury to or death of persons or damage to or destruction of property or for any other loss, damage or injury of any kind whatsoever resulting from the use of the project results and/or data. GTRC, GIT and NEETRAC disclaim any and all warranties both express and implied with respect to analysis or research or results contained in this report.

• It is the user's responsibility to conduct the necessary assessments in order to satisfy themselves as to the suitability of the products or recommendations for the user's particular purpose.

• No statement herein shall be construed as an endorsement of any product or process or provider

• Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Department of Energy

• This material is based upon work supported by the Department of Energy under Award No DE-FC02-04CH1237


PresentersDr. Nigel Hampton is the Program Manager for Reliability work at NEETRAC. He has worked in the Power Cable arena for more than 20 years. Nigel has a PhD in Physics from the University of Bath UK. He is currently the vice-chair of the Insulated Conductor Committee’s subcommittee on diagnostic testing (Subcommittee F).

Dr. Joshua Perkel is a Research Engineer in the Assessment group at NEETRAC. He has worked in the Power Cable arena for more than 5 years. Josh holds a PhD in electrical engineering from the Georgia Institute of Technology.


CDFI ContributorsNEETRAC

Rick Hartlein (PI)Thomas ParkerJoshua Perkel

Jorge AltamiranoTim AndrewsYamille del ValleNigel Hampton (Co-PI)

Georgia Tech - ECEMiroslav BegovicRon HarleyJ.C. HernandezSalman Mohagheghi

IREQJean-Francois Drapeau

2


Day 1

Lunch12:00 – 13:00

SAGE Concept14:30 – 14:45

Diagnostic Testing Technologies (Part I)16:30 – 17:00

Break14:45 – 15:00

Diagnostic Accuracies16:00 – 16:30Case Study: Roswell15:00 – 16:00

Cable System Failure Process14:00 – 14:30CDFI Background/Overview 13:30 – 14:00

NEETRAC Overview13:10 – 13:30Welcome13:00 – 13:10

TopicTime


Day 2

Continental Breakfast07:30 – 08:00

Lunch12:00 – 13:00

Accuracies Really Matter09:30 – 10:00

Selecting a Diagnostic Testing Technology11:25 – 11:45The Things We Know Now That We Did Not Know Before10:15 – 11:20

Review Day 108:00 – 08:15

Summary11:45 – 12:00

Break10:00 – 10:15

Diagnostic Testing Technologies (Part II)08:15 – 09:30

TopicTime


Day 1


Outline• NEETRAC Overview• CDFI Background/Overview• Cable System Failure Process• SAGE Concept • Case Study: Roswell• Diagnostic Accuracies• Diagnostic Testing Technologies• Accuracies Really Matter• The Things We Know Now That We Did Not Know Before• Selecting a Diagnostic Testing Technology• Summary

3


NEETRAC Overview


Background

• Created in 1996 when Georgia Power donated the facilities of its Research Center to Georgia Tech.

• Set up as a self supporting center within the School of Electrical and Computer Engineering of the Georgia Tech.

• NEETRAC is a membership based center, conducting research programs for the Electric Energy Transmission and Distribution Industry.

NEETRAC Overview


NEETRAC Mission & VisionMissionTo provide a venue where NEETRAC Staff, NEETRAC Members and the Georgia Tech Academic community can collaborate to solve problems in the T&D Arena.

VisionWe will build on our expertise to become the leading national Center for collaborative applied and strategic research and development for electric transmission and distribution.

NEETRAC Overview 12CDFI Meeting - Aug. 19-20 San Ramon, CA

Members 2009-20101. 3M2. ABB3. Ameren Services4. American Electric Power5. Baltimore Gas & Electric6. British Columbia Hydro7. Borealis Compounds LLC8. Con Edison9. Cooper Power Systems10. Dominion/Virginia Power11. Dow Chemical Company12. Duke Energy13. Entergy14. Exelon15. First Energy16. Florida Power & Light17. GRESCO Utility Supply

18. Hubbell19. NRECA20. NSTAR21. PacifiCorp22. Prysmian Cables & Systems 23. Public Service Electric & Gas24. S&C Electric Company25. South Carolina Electric & Gas26. Southern California Edison27. Southern Company28. Southern States29. Southwire30. Thomas and Betts/Homac31. TVA32. tyco / Raychem33. Zenergy Power

NEETRAC Overview

4


NEETRAC Membership Growth

20102008200620042002200019981996

40

30

20

10

0

Year

Mem

bers

2009


Members• Utility Members

– Provide > 50% of power sold in the US– Serve over 64,000,000 customers

• Manufacturing Members– Primary suppliers of T&D equipment to electric

utilities in the United States

NEETRAC Overview


Focus Areas Developed

Training/Education

Safety

Power Quality/Grounding

Operation, Installation, Design

System Analysis

Forensics

Condition Assessment

Asset Management

Reliability

System Enhancements

Research

New Product DevelopmentNew Technology/Research

Equipment Spec. & Test Protocol Development

Engineering Analysis & Support

Product Evaluation

Application Research

Hardware/Equipment Testing

FOCUS SEGMENTSPRIMARY FOCUS AREA


Facilities: High Voltage Lab

NEETRAC Overview

5


Facilities: Low Voltage & Mechanical Lab


Ploss

Investment

ΔV

γ r

aγ

Direction of Drop Movement

γ r

aγ

Direction of Drop Movement

NEETRAC Overview


Tensile and Impact Tests on 12-ft Grounding JumpersC-clamp on 2-in pin to Flat-face Clamp on 5/8-in pin, Tension versus Clamp Displacement

-500

0

500

1000

1500

2000

2500

3000

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4Actuator Displacement (inches)

Tens

ion

(lb)

Sample J4 Impact

Sample J4 Tensile

Sample J5, Impact

Sample J5 Tensile

Sample I4, Impact

Sample I5, Impact

Sample I6, Impact


Staff• 25 Research Staff

– Ph.D degees (EE & Physics)– M.S. degrees (EE, IE, & ME)– Bachelors degrees (EE & ME)

• 5 Administrative and IT Support

• 1 Coop Students

NEETRAC Overview

6


This is NEETRAC• 25 Research Staff• 5 Administrative

Support Staff• Academic Faculty• Co-op and Graduate

Students




CDFI Background


• Underground cable system infrastructure is aging (and failing). Much of the system is older than its design life.

• Not enough money / manufacturing capacity to simply replace cable systems because they are old.

• Need diagnostic tools that can help us decide which cables/accessories to replace & which can be left in service.

• Always remember that we are talking about the cable SYSTEM, not just cable.

Cabl

e Fa

ilure

s pe

r Y

ear

20052000199519901985198019751970

1000

800

600

400

200

0

Why do we need diagnostics?

CDFI Background/Overview

7


Composition of US MV systemIn

stal

led

Capa

city

(%

)

UNKNOWNTRXLPEEPRXLPEHMWPEPILC

100

90

80

70

60

50

40

30

20

10

0

25

50

75

CDFI Background/Overview 26CDFI Meeting - Aug. 19-20 San Ramon, CA

Failure Split

Unknown1.1%Terminations

5.6%

Splices37.1% Cable

56.2%



• In the CDFI, NEETRAC worked with 17 utilities, 5 manufacturers and 5 diagnostic providers to achieve the objective of clarifying the concerns and defining the benefits of diagnostic testing.

• Phase 1 has almost exclusively focused on aged medium voltage systems.

• This is the largest coherent study of cable system diagnostics anywhere.

Overview


NEETRAC Members

Non NEETRACMembers Supporters

Dept of Energy

Diagnostic Providers

CDFI


8


Participants

SouthwireSouthern CompanySouthern California EdisonTyco / Raychem Public Service Electric & GasPrysmianOncor (TXU)PEPCOPacific Gas & Electric (added Jan 06)PacifiCorp (added mid 2005)NRECAIMCORPHydro QuebecHV Technologies

CenterPoint Energy

GRESCO

HV Diagnostics

Cablewise / Utilx

Florida Power & Light

Con Edison

HDW Electronics

Georgia Tech

First EnergyExelon (Commonwealth Edison & PECO)

Duke Power CompanyCooper Power Systems

AmerenAmerican Electric Power


CDFI - Primary Activities1) Technology Review2) Analysis of Existing (Historical) Data3) Collection and Analysis of Field (New) Data4) Verification of VLF Test Levels5) Defect Characterization6) Develop Knowledge Based System7) Quantify Economic Benefits8) Reports, Update Meetings and Tech Transfer

Seminars

Analyses are data / results driven



CDFI Activities

CDFI

Analysis Lab Studies

Field Studies Dissemination


CDFI ActivitiesCDFI

Analysis Lab Studies

Field Studies Dissemination

Value / Benefit

Accuracies

Utility Data

IEEE Std Work

VLF Withstand

Tan δ

PD

Georgia Power

Duke

Handbook

Publications

Meetings

Industry

CDFIKnowledge Based Systems


9


CDFI ActivitiesLab

Studies

VLF Withstand Tan δ PD

Test TimeTest VoltageForensics

Time StabilityVoltage Stability

Non-Uniform DegradationNeutral Corrosion

CalibrationPhase Pattern

Feature ExtractionClassification


CDFI ActivitiesField

Studies

Georgia Power XLPE

Jkt & UnJkt21 Conductor Miles

DukeXLPE & Paper

Jkt & UnJkt29 Conductor Miles

Offline PD (0.1Hz)Offline PD (60Hz)

Tan δMonitored Withstand

Offline PD (0.1Hz)Tan δ

Monitored Withstand

Charlotte * 2CincinnatiClemson

Morresville

EvansMaconRoswell



CDFI Activities

Analysis89,000 Conductor Miles

Value / Benefit Accuracies UtilityData IEEE Std Work Knowledge

Based Systems

Economic ModelSAGE

DC WithstandOffline PDOnline PD

Tan δVLF Withstand

400 Omnibus400.2 VLF

SurveyExpert System

Application


CDFI Activities

UtilityData

Con Ed Com Ed PPL Alabama Power Keyspan

DC WithstandOnline PD

VLF Withstand

Offline PD (60Hz)Online PDTan Delta

VLF Withstand

Offline PD (0.1Hz)Tan Delta Online PD Offline PD (0.1Hz)

Tan Delta


10


CDFI Activities

UtilityData

FPL

Offline PD (60Hz)VLF Withstand

PEPCO

Offline PD (60Hz)Offline PD (0.1Hz)

Online PDVLF Withstand

PG&E ONCOR Ameren

Offline PD (60Hz)Online PD

Tan δ

Offline PD (60Hz)Online PD Offline PD (60Hz)


Dataset Sizes

89,000ALLService Performance

Diagnostic

Data Type

-0.3IRC

9,8101.5VLF Withstand

5501.5Tan δ

262-PD Online

4902PD Offline

149-Monitored Withstand

78,105-DC Withstand

Field[Conductor miles]

Laboratory[Conductor miles]Technique



Time [Days]

Log

Cum

ulat

ive

Failu

res

3000200015001000900800

1500

1000

500

100

Time [Days]

Log

Cum

ulat

ive

Failu

res

3000200015001000900800

1500

1000

500

100

Benefits from Diagnostic ProgramsDecreasing failures associated with diagnostics and actions


Program Initiated


At the Start• For many utilities, the usefulness of diagnostic testing was

unclear.

• The focus was on the technique, not the approach.

• The economic benefits were not well defined.

• There was almost no independently collated and analyzed data.

• There were no independent tools for evaluating diagnostic effectiveness.


11


Where we are today (1)1. Diagnostics work – they tell you many useful things, but not

everything.2. Diagnostics do not work in all situations.3. Diagnostics have great difficulty definitively determining the

longevity of individual devices. 4. Utilities HAVE to act on ALL replacement & repair

recommendations to get improved reliability.5. The performance of a diagnostic program depends on

• Where you use the diagnostic• When you use the diagnostic• What diagnostic you use• What you do afterwards


6. Quantitative analysis is complex BUT is needed to clearly see benefits.

7. Diagnostic data require skilled interpretation to establish how to act.

8. No one diagnostic is likely to provide the detailed data required for accurate diagnoses.

9. Large quantities of field data are needed to establish the accuracy/limitations of different diagnostic technologies.

10. Important to have correct expectations – diagnostics are useful but not perfect!


Where we are today (2)


• In the CDFI, NEETRAC worked with 17 utilities, 5 manufacturers and 5 diagnostic providers to achieve the objective of clarifying the concerns and defining the benefits of diagnostic testing.

• We have come a long way wrt the project objective. – Analysis driven by data / results– Developed a good understanding that diagnostic testing can

be useful, but the technologies are not perfect.– Developed ways to define diagnostic technology accuracy and

found ways to handle inaccuracies. – Developed diagnostic technology selection and economic

analysis tools.– Understand that there is yet more to learn.

Overview


QUESTIONS

12




How things fail and what fails have a big impact on the selection of diagnostics

Cable System Failure Process


Failures by Equipment

Unknown - all (%)Terminations - all (%)Splice - all (%)Cable - all (%)

100

80

60

40

20

0

Dis

burs

emen

t of

Fai

lure

s (%

)

Cable System Failure Process 48CDFI Meeting - Aug. 19-20 San Ramon, CA

Failure Rates

100

80

60

40

20

0

Failu

re R

ate

[#/1

00 M

iles/

Yea

r]

Lower Quartile: 1.6Median: 3.5Upper Quartile: 8 Mean: 12Max: 140

Peak at 140


13


Failure Rate Estimates – By Equipment

Unk RateTerm RateSplice RateCable Rate

8.0

7.0

6.0

5.0

4.0

3.0

2.5

2.0

1.5

1.0

0.5

0.0

Failu

re R

ate

- M

onte

Car

lo E

stim

ate

(#/1

00m

iles/

yr)


Major Cable Components

Jacket (Recommended)

Metallic Shield/Neutral

Insulation Shield

Insulation

Conductor

Cable System Failure Processhttp://www.otds.co.uk/cables.php

Conductor Shield

Extruded PILC


1. Cavity at shield(s)2. Cavities due to shrinkage3. Insulation shield defect4. Contaminant (poor adhesion)5. Protrusions at shield(s)6,7 Splinter/Fiber8. Contaminants in insulation or shields

Defect Types in Extruded Cables


Conversion of Water to Electrical Trees

• Acts as a stress enhancement or protrusion (non-conducting)

• Water tree increases local electric field

• Water tree also creates local mechanical stresses

• If electrical and mechanical stresses high enough ⇒electrical tree initiates

• Electrical tree completes the failure path – rapid growthElectrical tree growing

from water treeCable System Failure Process

14


Defect Types in Extruded Cable Accessories


Diagnostics used in Challenging Areas



Summary• Cable system aging is a complex phenomenon.

• Multiple factors cause systems to age.

• Increases in dielectric loss and partial discharge are key phenomenon.

• The aging process is nonlinear.

• Diagnostics must take these factors into consideration.


QUESTIONS

15




SAGE Approach to

Diagnostic Programs


Diagnostic Program Phases - SAGESelectionData compilation and analysis needed to identify circuits that

are at-risk for failure (at-risk population).

ActionDetermine what actions can be taken on circuits based on the

results of diagnostic testing.

GenerationConduct diagnostic testing of the at-risk population.

EvaluationMonitor at-risk population after testing to observe/improve

performance of diagnostic program.SAGE Concept 60

CDFI Meeting - Aug. 19-20 San Ramon, CA

SAGE at WorkFailures [#]

Time

Selection

Action

Generation

Evaluation

Decreasing Failures

Increasing Failures

SAGE Concept

16


Failures [#]

Time

Decreasing Failures

Increasing Failures

Continued Failure Increase

SAGE Concept 62CDFI Meeting - Aug. 19-20 San Ramon, CA

When to deploy diagnostics

Time (Years)

Cabl

e Sy

stem

Per

form

ance

403020100

Operational Stress

Condition AssessmentCommissioning

SAGE Concept


Global ContextComparison with many tests

DatabasesStandards

Context – is important

Local ContextComparisons within one area

DataGeneration from

Diagnostic Measurement

SAGE Concept 64CDFI Meeting - Aug. 19-20 San Ramon, CA

QUESTIONS

17




Case StudyRoswell, GA

November 2008 & January 2009

TDRTan Delta

Monitored WithstandOffline PD


67

Roswell Map

Case Study: Roswell 68CDFI Meeting - Aug. 19-20 San Ramon, CA

SELECTION

Case Study: Roswell

18


Roswell Background Info.• 1980 vintage XLPE feeder cable, 1000 kcmil, 260 mils wall,

jacketed.

• Failures have occurred over the years – no data on source

• Recently experienced very high failure rates of splices on this section: 80 failures / 100 miles / yr.

• Overall there have been 10 -15 failures of these splices in last two years on a variety of GPC feeders.

• Splice replacement may be acceptable if there is a technical basis.


Knowledge Based Selection System

Case Study: Roswell


KBS Demo


Summary for Diagnostic Selection

Replace AccessoriesReplace SegmentReplace Small Portion

TDR

& H

istorical R

ecords ON

LY

PD O

ffline

PD O

nline

Tan Delta

Monitored

Withstand

HV D

C Leakage

VLF 60 Mins

VLF 30 Mins

VLF 15 Mins

DC

Withstand

Diagnostic Technique

ActionScenario

Have a shortlist of three techniques

Case Study: Roswell

19


Economic Details – prior to testing• Complete System Replacement $1,000,000 approx• Complete Splice Replacement $60,000• Test time (determined by switching) 3 - 4 Days• Selection Costs $5,000• Splice Replacement 7 Days• Retest after remediation 1 Day

Monitored Withstand, Offline PD and VLF (30 mins) offer economic benefit over doing nothing.


Scenario Assessment before TestingOffline PD• If 51,000ft is tested• 0.5% fails on test, no customer

interrupted • 1 site / 1,000ft (median)• 40% discharges in cable• Estimate

– 0 fails on test– 51 discharge sites

• 20 cable, • 31 accessories

– 15 splices– <2 failure in 12 months from

test

Monitored Withstand• If 51,000ft is tested• <4% fails on test, no customer

interrupted• 70% of loss tests indicate no

further action• Estimate

– <2 fails on test– 3 assessed for further

consideration by loss – 0.5 failure in 12 months

from test

Case Study: Roswell


ACTION


Initial Corrective Action Options

• Replace splices only – no detailed records assume 12 splices.

• Complete system replacement.

Case Study: Roswell

20


GENERATION


Overhead and Cabinet Terminations

Case Study: Roswell


Tan δ Monitored Withstand


If this had been a Simple Withstand

Length Tested (miles)1086420

18 Segments Tested

No Failures On Test

Case Study: Roswell

21


Monitored Withstand - Stability

Sequence of Lengths Tested (miles)1086420

18 Segments Tested

Case Study: Roswell

Sequence of Lengths Tested (miles)1086420

Pass - Un Stable Loss

Pass - Stable Loss

18 Segments Tested

60 min test

30 min test


Test Results - Local Perspective

Length Along Feeder (ft)

Tip

Up in

Tan

Del

ta {

1.5U

o -

0.5U

o} (

1e-3

) 1000

100

10

1

150001000050000

150001000050000

1000

100

10

1

1 2

3

STABLEUNSTABLE

StabilityMeasurement

7

6

53

2

1

76

53

2

1

76

5

3

21

Panel variable: Phase

Segment ID'sNumbers indicate

Case Study: Roswell


Test Results – Global Perspective

Tip Up 1.5Uo - 0.5 Uo (1e-3)

Tan

Del

ta @

Uo

(1e-

3)

1000100101

100.0

10.0

1.0

0.1

1 150

6

150STABLEUNSTABLE

Stability

ErrorSplice

withstandmonitoredinstability inRange of

Termination Damage


Targeted Offline PD (VLF)

Case Study: Roswell

22


Targeted Offline PD Test – Segment 6

Distance from Cubicle 2 (ft)

Phase

PD

TDR

C - 3

B - 2

A - 1

C - 3

B - 2

A - 1

5000450040003500300025002000150010005000

Positions)(Approx


Phase

PD

TDR

C - 3

B - 2

A - 1

C - 3

B - 2

A - 1

5000450040003500300025002000150010005000

Positions)(Approx

anomalous TDR reflectionsOpen symbols represent the


Phase

PD

TDR

C - 3

B - 2

A - 1

C - 3

B - 2

A - 1

5000450040003500300025002000150010005000

Positions)(Approx

anomalous TDR reflectionsOpen symbols represent the


PD Inception – local perspective

Position from Cubicle 2 (ft)

VLF

Tes

t V

olta

ge (

kV)

232119171513

10

5

048003600240012000

48003600240012000

232119171513

10

5

0

A - 1 B - 2

C - 3

3133

2126

36372126

4088

1681 785

Panel variable: Phase

PD in 1 of 9 splices PD in 1 of 7 splices

PD in 5 of 9 splices

Position of PD (ft)

PD Inception (kV)

Prob

abili

ty o

f Sp

lice

Ince

ptio

n (%

)

201510987

90

80

706050

40

30

20

10

5

3

2

1

Case Study: Roswell


EVALUATION


Evaluation after TestingOffline PD• 15,000ft actually tested• Estimate

– 15 discharge sites • 6 cable, • 9 accessories

– 6 splices– <1 failure in 12 months from

test• Actual

– 7 discharge sites • 0 cable,• 7 accessories

– 25 splices– 0 failure in 7 months since

test

Monitored Withstand• 51,000ft actually tested• Estimate

– 2 fails on test– 3 assessed for further

consideration by loss – 0.5 failure in 12 months

from test

• Actual– 0 fails on test– 6 assessed for further

consideration by stability, tip up & loss

– 1 failure (cable) in 8 months since test

Case Study: Roswell

23


After Testing…• Actions have been performed by GPC.

– Suspect splice investigated, actually broken neutral.– Damaged termination replaced.– Test excavations & Ground Penetrating Radar tests

conducted, concluded that it was not practical to replace splices as planned

• System re-enforcements planned.

• All tested circuits have been left in service and are being monitored by GPC.


QUESTIONS


Break



24


Diagnostic Accuracies


Performance of Diagnostics• Performance evaluation primarily focuses on diagnostic

accuracy.

• Diagnostic accuracies quantify the diagnostic’s ability to correctly assess a circuit’s condition.

• Accuracy must be assessed based on “pilot” type field test programs in which no actions are performed.

• Circuits must be tracked for a sufficient period of time.



Diagnostic Measurements and Failures• Symptoms are difficult to relate to future failures unless they are

in the extremes.

Pro

babi

lity

“Good” “Bad”?


No Failure Failure

Diagnostic Accuracies 96CDFI Meeting - Aug. 19-20 San Ramon, CA

Objective of Diagnostic TestsThe target population contains both “Good” and “Bad” components

– “Good” – Will not fail within diagnostic time horizon– “Bad” – Will fail within diagnostic time horizon

“Bad” Components “Good” ComponentsTarget Population


25


Diagnostic OperationApplying the diagnostic will separate the population into:• No Action Required group• Action Required group

But if the diagnostic is imperfect...

No Action Required Action Required


Complimentary Diagnoses

Online PD Offline PD

VLF TDNo ActionActionNo Test

Category

33.8%

52.1%

14.1%

69.0%

11.3%

19.7%

83.1%

4.2%

12.7%

VLF TD 25% Offline PD 36% Online PD 79%Ratio Action / No Action

3 Service Failures

since testing completedDiagnostic Accuracies


CDFI Accuracies

Action

Agr

eem

ent

ActDont_Act

Agree

Disagree

Agreement btw Online Offline * Online PDAgreement btw Online VLF TD * Online PDAgreement btw Offline VLF TD * Offline PD

Variable

CDFI diagnostic accuracies are based on service performance (failures) not diagnostic agreement.


Perspective• Diagnostics make measurements in the field and find

Anomalies.• Detecting the presence of an Anomaly is, in our view, not

sufficient.• The goal, in our view, is to detect an Anomaly which leads to

reduced reliability (failure in service) or compromised performance (severed neutrals – stray voltage).

In accuracy estimates we have used failures in service and interpreted the diagnostics as “Bad Means Failure.”


26


“Bad Means Failure” Accuracies

1716151413121110987654321

100

80

60

40

20

0

1716151413121110987654321

No Action Accuracy

Dataset

Dia

gnos

tic

Acc

urac

y

Action Accuracy

1716151413121110987654321

100

80

60

40

20

0

1716151413121110987654321

No Action Accuracy

Dataset

Dia

gnos

tic

Acc

urac

y

Action Accuracy


1716151413121110987654321

100

80

60

40

20

0

1716151413121110987654321

No Action Accuracy

Dataset

Dia

gnos

tic

Acc

urac

y

Action Accuracy


Overall AccuracyNo Action AccuracyAction Accuracy

100

80

60

40

20

0

Dia

gnos

tic

Acc

urac

y [%

]


100

80

60

40

20

0

Dia

gnos

tic

Acc

urac

y [%

]

All Accuracies


100

80

60

40

20

0

Dia

gnos

tic

Acc

urac

y [%

]


100

80

60

40

20

0

Dia

gnos

tic

Acc

urac

y [%

]


100

80

60

40

20

0

Dia

gnos

tic

Acc

urac

y [%

]



QUESTIONS



27


Diagnostic Testing Technologies


Introduction• A wide range of diagnostic techniques are commercially

available.

• Tests are performed either offline (circuit de-energized)) or online (energized) and by service providers or utility crews.

• Different voltage sources may be used to perform the same measurement.– DC– 60 Hz. AC– Very Low Frequency (VLF) AC– Damped AC (DAC)



Utility Use of Diagnostics

Diagnostic Testing Technologies 108CDFI Meeting - Aug. 19-20 San Ramon, CA

Diagnostic Survey

• A survey of CDFI participants in 2006 was conducted to determine how diagnostics were employed.

• Survey was updated at the end of 2008.

• Survey results focused CDFI work on technologies currently used in the USA.


28


No TestingTesting - one techniqueTesting - > one technique

27.8%

30.6%

41.7%

Survey of Use of Diagnostics




More than one technique usedNo testingO ne technique used

No TestingOccasional useRegularly usedSome testing

4.0%

96.0%

75.0%

25.0%

No Testing

Testing


Technologies• Simple Dielectric Withstand• Dielectric Loss (Tan δ & Dielectric Spectroscopy)• Time Domain Reflectometry (TDR)• Online Partial Discharge (PD) • Offline Partial Discharge (PD)• Isothermal Relaxation Current (IRC)• Recovery Voltage (RV)• Combined Diagnostics



DatabasesStandards

Context


DataGeneration from



29


Diagnostic Context

OK Not Proven either way

NOT OK

• Extreme conditions are easy to decide what to do about.

• What to do about the ones in the middle?

• How to define the boundaries?


Simple Dielectric Withstand


Simple Dielectric WithstandTest Description• Application of voltage above normal operating voltage for a

prescribed duration.• Attempts to drive weakest location(s) within cable segment to

failure while segment is not in service.

Field Application• Offline test that may use:

– DC– 60 Hz. AC– VLF AC– Damped AC

• Testing may be performed by a service provider or utility crew.

Simple Dielectric Withstand 116CDFI Meeting - Aug. 19-20 San Ramon, CA

Withstand Test Process

HOLDEARLY

Time

Voltage

t = 0 tTest

Voltages and Times for VLF covered in IEEE Std. 400.2

The goal is to have circuit

out of service, test it such that

“imminent”service failures

are made to occur on the

test and not in service

Hold Entry

Ramp Entry


30


VLF Test Voltages

Cable Rating (kV)

Test

Vol

tage

(kV

)

302520151050

60

50

40

30

20

10

0302520151050

Cosine-rectangular Sinusoidal

Peak Voltage (kV) Acceptance

Peak Voltage (kV) InstallationPeak Voltage (kV) MaintenanceRMS Voltage (kV) Acceptance

RMS Voltage (kV) InstallationRMS Voltage (kV) Maintenance

Variable Use


DataGeneration from




Test Sequences

Cumulative Length Tested in One Year (Miles)

Wit

hsta

nd T

est

Out

com

es

140120100806040200

26

20

22

22

26

23

16

26

2730

Time of failure in mins for failures > 15 mins

Simple VLF Withstand to IEEE400.2 Levels




31


Area

Failu

res

on T

est

[% o

f Te

sted

]

4321Overall

35

30

25

20

15

10

Separation with Simple VLF Outcomes

Area 1 is clearly different from the others.


“Early” Phase Matters

Time on Test [Minutes]

Failu

res

on T

est

[% o

f Te

sted

]

10.001.000.100.01

60

50

40

30

20

10

Early Hold

60 % of failures on test occurred during

“Early” phase



Time on Test (Mins)

Failu

re s

On

Test

- F

OT

(% o

f Se

ctio

ns T

este

d)

151050

0.5

0.4

0.3

0.2

0.1

0.0

STA

RT

OF

HO

LD P

HA

SE

151050

STA

RT

OF

HO

LD P

HA

SEDC VLF

1327

VoltageFeeder

0.11%

0.16%

“Early” and “Hold” Phases


Length Adjusted

Difference between VLF and DC is primarily result of “Early” phase


“Early” Phase – Ramp Entry Example

Voltage [U0]

Failu

res

on T

est

[% o

f To

tal T

ests

]

10.01.00.1

80

70

60

50

40

30

20

10

In this case, 60 % of the tests produced a failure before reaching the target test voltage.


32


“Early” and “Hold” Failure Mechanisms (VLF)


Failu

res

on T

est

[% o

f To

tal T

este

d]

100.050.010.05.01.00.50.1

20

15

10

5

1

“Early”Phase

“Hold” Phase


“Early” Phase – Hold Entry


Failu

res

on T

est

[% o

f To

tal T

este

d]

10.05.01.00.50.1

5

1


“Early” phase accounts for 30 % of failures on test.



DatabasesStandards


Withstand Testing Experience


Surv

ivor

s [%

of

Tota

l Len

gths

Test

ed]

706050403020100

100

80

60

40

20

0

IEEE Recommendation

IEEE 400.2 Range

9700 Conductor Miles>2000 Conductor Miles

0.3 Conductor Miles


33


Test Performance for Different Utilities


Failu

res

on T

est

[% o

f 10

00 f

t Se

gmen

ts]

100.0010.001.000.100.01

10

5

3

2

1

0.01

4.4%5.0%

0.5%

5.7%

60

A1A2DI

Utility

1000ft Length Adj.


Service Experience

10000100010010

30

20

10

5

3

2

1

0.1

Time to Failure [Days since test]

Serv

ice

Failu

res

[% o

f To

tal T

este

d]

5%

472

Day

s63

7 Da

ys

10%

1247

Day

s

2247

Day

s

15 Min @ 2.5U030 Min @ 1.8U0

2650 Conductor Miles

224763730 Min @ 1.8 U0

472

Time to Failure5%

[Days]

124715 Min @ 2.5 U0

Time to Failure10%

[Days]Test Conditions



Performance After Test – Pass/No Pass


100010010

20

10

5

3

2

1

0.1

T ime to Failure [Days]

Fa

ilu

res

on

Te

st [

% o

f T

ota

l T

est

ed

]

1

133

215

PassNo Pass - Repaired

Initial Test Result

100010010

20

10

5

3

2

1

0.1

T ime to Failure [Days]

Fa

ilu

res

on

Te

st [

% o

f T

ota

l T

est

ed

]

1

36 1228

PassNo Pass - Repaired

Initial Test Result

15 Minute 30 MinuteNot Length Adjusted


Difference between Passing and not Passing

Time to Failure for 1 % of Tested Segments[Days]

1228363021513315

No Pass - RepairedPass

Test Duration[Min]

Segments that fail on test and subsequently repaired perform better in service.

34


What does this mean for Withstand?• The technique is widely used by utilities

• Tested circuits display improved reliability

• Circuits normally Pass the tests

• Multiple / cascading failures are rare

• IEEE400.2 recommended times (30 mins) and voltages seem to give good service performance


What does this mean for Withstand?• IEEE400.2 recommended times (30 mins) and voltages

seem to give good service performance

• Modifications to IEEE400.2 recommendations need to be considered very carefully

• Voltage & test time cannot be determined independently

• Many test fails occur early in the test, useful information is revealed by tracking of these times / voltages of failure

• More failures on test does not mean fewer service fails


QUESTIONS


Day 2

35




Dielectric Loss (Tan δ)


Dielectric Loss (Tan δ)Test Description• Measures total cable system loss (cable, elbows, splices & terminations).• May be performed at one or more frequencies (dielectric spectroscopy).• May be performed at multiple voltage levels.• Monitoring may be conducted for long durations.


– 60 Hz. AC– VLF AC– Damped AC

• Testing may be performed by a service provider or utility crew.• Step voltage up to pre determined level with post test analysis

Tan δ 140CDFI Meeting - Aug. 19-20 San Ramon, CA

Dielectric Loss (Tan δ)

V

I

RI CI 1tan( ) R

C

IDFI RC

δω

= = =

VRI

ICI

δ

θ

• The cable insulation system is represented by an equivalent circuit.• In its simplest form the equivalent circuit consists of two parameters (IEEE

Std. 400):• Resistor• Capacitor

• When voltage is applied to the cable, the total current is the sum of the capacitor current and resistor current.

Tan δ

36


Cable System Equivalent

T C S C TCable system (cable, splices, and

terminations) is reduced to simple circuit.


DataGeneration from


Tan δ


Tan δ Ramp Test Data

Time [min]

Tan-

delt

a [1

e-3]

543210

100

90

80

70

60

50

40

30

20

10

0.51.01.51.7

[p.u.]Voltage

Time [min]

Tan-

delt

a [1

e-3]

543210

100

90

80

70

60

50

40

30

20

10

0.51.01.51.7

[p.u.]Voltage

Time [min]

Tan-

delt

a [1

e-3]

543210

100

90

80

70

60

50

40

30

20

10

0.51.01.51.7

[p.u.]Voltage

Mean

could be used)Standard Deviation - IQRScatter (represented by

Tip Up

Time [min]

Tan-

delt

a [1

e-3]

543210

100

90

80

70

60

50

40

30

20

10

0.51.01.51.7

[p.u.]Voltage

MeanTip Up



Tan δ

37


Tan δ Data for EPR Cable Systems

Voltage (kV)

Tan

Del

ta (

1e-3

)

1211109876543

25

20

15

10

5 ConcernLowest

ConcernHighest


Segments within a Feeder

Length Along Feeder (ft)

Tip

Up in

Tan

Del

ta {

1.5U

o -

0.5U

o} (

1e-3

)

1600014000120001000080006000400020000

1000

100

10

1

STABLEUNSTABLE

Stable

7

6

5

3

2

1

Phase = 1

Tan δ


Lengths within a Local Region

Length (ft)

Tan

Del

ta (

1e-3

)

50004000300020001500

10

1



DatabasesStandards

Tan δ

38


Testing at Reduced Voltages

Tan-delta @ 2.0 Uo [1E-3]

Tan-

delt

a @

1.5

Uo

[1E-

3]

100.010.01.00.1

100.0

10.0

1.0

0.1

1.2 2.2 4

0.7

1.3

2.3

Regression95% CI95% PI

PI: Prediction IntervalCI: Confidence Interval


Tan δ Interpretation

Tip Up

Tan

Del

ta

-1010-1-

-

150

6

0

No Action

Further Study

Action Required

Based on 258 Conductor Miles

Tan δ


Tan δ Correlation with VLF Withstand

Length (ft)

Tan

D (

1e-3

)

1000100

1000.0

100.0

10.0

1.0

0.11000100

?Filled

Unfilled

Basic Type

Fail Subsequent VLF Withstand Pass Subsequent VLF Withstand


Elasped Time between test and failure in service at May 09 (Month)

Perc

ent

101FOT

40

30

20

10

5

3

2

1

ARFSNA

Action

Tan δ Performance Curves


Perc

ent

101FOT

40

30

20

10

5

3

2

1

ACTION REQUIREDFURTHER STUDYNO ACTION

Action


Perc

ent

101FOT

40

30

20

10

5

3

2

1


Action


Perc

ent

101FOT

40

30

20

10

5

3

2

1


Action


Perc

ent

101FOT

40

30

20

10

5

3

2

1

293236

91011

1.71.92.3

14

4

0.66012 24


Action

Tan δ

39


What does this mean for Tan δ?

• Provides information on the whole cable system• Most useful features are

– Time Stability– Differential Tan δ (Tip Up)

• Higher loss correlates with increased probability of failure• Comparisons provide very useful information

– Length effects– Adjacent sections / phases

• Existing levels in IEEE Std. 400 are too conservative. Newer (higher) levels to be in IEEE Std. 400.2 revision


Time Domain Reflectometry


Time Domain Reflectometry (TDR)Test Description• Measures changes in the cable impedance as a function of

circuit length by observing the pattern of wave reflections.• Used to identify locations of accessories, faults, etc.

Field Application• Offline test that uses a low voltage, high frequency pulse

generator.• Testing may be performed by a service provider or utility crew.

TDR 156CDFI Meeting - Aug. 19-20 San Ramon, CA

TDR Principles

Near End

TDREquipment

Far End

JointL

Joint

TDR

40


Wet Joint

• Feeder had two splice failures just before the test.

• Water ingress was detected with the TDR.

• Failure on 01/17/2008 at detected water ingress location.

• Water ingress confirmed by tests and repair crew.

Water ingress location as seen by the TDR


TDR Field Measurements

500040003000200010000

A-1

B-2

C-3

Distance from Cubicle 2 [ft]

Phas

e

circles)Anomalous TDR reflections (open

TDR


Lengths Tested

Cable Length - log (ft)

Perc

ent

10.0

7.5

5.0

2.5

0.0100000100001000100

100000100001000100

10.0

7.5

5.0

2.5

0.0

PD Tan D

VLF Withstand

Panel variable: Technique

Median 814 ftMedian 485 ft

Median 3500 ft

Based on diagnostic data supplied to CDFI

Measurements made with TDR


What does this mean for TDR?

• All diagnostics rely on the neutral, TDR helps to establish its condition.

• Length and accessory information are very important in establishing the context of diagnostic findings.

• Unusual TDR traces can diagnose unusual features in their own right.

41


Online Partial Discharge


Online Partial DischargeTest Description• Measurement and interpretation of discharge and signals on

cable segments and/or accessories.• Signals captured over minutes / hours.• Monitoring may be conducted for long durations.

Field Application• Online test that does not require external voltage supply.• Testing typically performed by a service provider.• Different implementations of the overall approach• Assessment criteria are unique to each embodiment of the

technologyOnline PD


DataGeneration from


Online PD 164CDFI Meeting - Aug. 19-20 San Ramon, CA

Discharge Occurrence

No PD PD

Online PD

42




Distribution of PD along Lengths• 5000 ft. portion of sample feeder

• Mixture of different PD levels for different sections and accessories.

Cable Section Accessory

No PDPD

Online PD



DatabasesStandards


Where is PD found?

Accessory54.0%

Cable46.0%

Online PD

43


Variability in PD LocationPe

rcen

tage

of

PD (

%)

AccessoryCable

100

80

60

40

20

0


Diagnostic Results (Overall)

Accessory Cable

52.5%

42.9%

314.8%

266.6%

113.1%

52.3%

41.8%

314.4%

268.0%

113.4%

226 Conductor Miles

Online PD


Level Based Reporting Systems• Level-based (i.e. “1, 2, 3” , “Defer, Repair, Replace”, “Act , Don’t Act”

etc.) reporting systems are increasingly common.

• Level systems, on their own, can have limited meaning for utilities.

• Levels clearly indicate a hierarchy– “5” worse than “4” “Replace” worse than “Defer”

• No sense of the magnitude of the difference– How much worse is “Act” than “Don’t Act” in terms of service

performance?

• Comparisons / interpretation of different level-based reporting systems is difficult.

Need to associate meaning with the levels

Level Based Reporting 172CDFI Meeting - Aug. 19-20 San Ramon, CA

Online PD Performance Curve

Time to Failure (Years)

Perc

ent

20.010.05.01.00.50.1

99

90807060504030

20

10

5

3

2

1

3%

18%

89%

2

345

Level

Level Based Reporting

44


Alternate Interpretation

89518433

< 32<< 31

Alternate Class(based on probability of failure)

Original Level

Class 18 has 6 times poorer endurance than Class 3

Class 89 is 5 times poorer than Class 18

Level Based Reporting 174CDFI Meeting - Aug. 19-20 San Ramon, CA

Probabilistic Approach – Online PD

Days Between Test & Failure

Perc

ent

1000100101

99

9080706050403020

10

532

1

0.01

No PDPD

PD Class

Level Based Reporting


Variability in Diagnostic Results

Perc

ent

Level 5Level 4Level 3Level 2Level 1

90

80

70

60

50

40

30

20

10

0

Level 5Level 4Level 3Level 2Level 1

Accessory Cable


How often is PD found?

PD O

ccur

ence

(#

/100

0 ft

)

All PDAccessory PDCable PD

0.35

0.30

0.25

0.20

0.15

0.10

0.05

0.00

1 PD signal every 4000 ft4.3 signals / 1000 ft

Online PD

45


Estimated Failure Reduction

Time since Start [Days]

Cum

mul

ativ

e Fa

ilure

s [#

]

1000100

100

10

11000100

ACCESSORY CABLE

Levels 4/5 ReplacedAll Segments Left in Service

14 Avoided Fails45 Actions for

23 Avoided Fails52 Actions for


What does this mean for Online PD?

• Highly degraded systems most easily differentiated• Not necessarily easy to deploy – sensor placement and

manhole access can be challenging• Signal analysis is labor intensive• Data for level interpretation is available• Trending is likely to be valuable, incorporating this in a

level-based reporting system can be a challenge• Baseline (when new) studies likely to be valuable• Active failure mechanisms need to involve discharges• Can localize to accessory and cable segments


Offline Partial Discharge


Offline Partial DischargeTest Description• Measurement and interpretation of partial discharge signals

above normal operating voltages.• Signal reflections (combined with TDR information) allows

location to be identified within cable segment.


– 60 Hz. AC service provider – VLF AC utility crew– Damped AC utility crew

• Step voltage up to pre determined level with post test analysisOffline PD

46


DataGeneration from


Offline PD 182CDFI Meeting - Aug. 19-20 San Ramon, CA

PD Pulse

140 mV

180 pC

Offline PD


PD Phase Resolved Pattern

3601800

0

-2.50E-1

-1.25E-1

1.25E-1

-323

-162

162

3232.50E-1

0

Phese [Deg]

Am

plitude [pC]


PD Magnitude

PD Measurement Voltage (Uo)

PD L

evel

(pC

)

2.252.001.751.50

40

35

30

25

20

15

10

5

the fieldMeasurement fromIndividual


PD L

evel

(pC

)

2.252.001.751.50

40

35

30

25

20

15

10

5 Max allowed for current production



PD L

evel

(pC

)

2.252.001.751.50

40

35

30

25

20

15

10

5 Max allowed for current production

Max Limit for 1970's production


Offline PD

47




PD Charge Magnitude Distributions

Apparent Charge Magnitude [pC]

Perc

ent

6005004003002001000

20

15

10

5

0

Apparent Charge Magnitude [pC]

Perc

ent

6005004003002001000

20

15

10

5

0

XLPE

Offline PD


PD Inception Voltage

Apparent Inception Voltage [U0]

Perc

ent

2.42.11.81.51.20.9

18

16

14

12

10

8

6

4

2

0

Apparent Inception Voltage [U0]

Perc

ent

2.42.11.81.51.20.9

18

16

14

12

10

8

6

4

2

0

XLPE



DatabasesStandards

Offline PD

48


189

Location of PD

Termination26.3%

Splice34.3%

Cable39.4%

60.6% of PD sites detected in accessories

222 Conductor Miles


Offline PD Test Sequence• Testing sequence for 16,000 ft.

No PD

PDOffline PD


PD Location

Location [% of Circuit Length]

Perc

ent

9075604530150

18

16

14

12

10

8

6

4

2

0

Terminations

Cable & Splices


PD Sites per Length

PD S

ites

per

100

0 fe

et

5

4

3

2

1

Approx. 1 PD Site/1000 ftMedian = 0.96 PD Sites/1000 ft

Offline PD

49


What does this mean for Offline PD?

• Highly degraded systems most easily differentiated• Signal analysis can be labor intensive• Data for level interpretation could be available• Trending is likely to be very valuable• Incorporating trending in a level-based reporting system

can be a challenge• Baseline (when new) studies likely to be very valuable• Active failure mechanisms need to involve discharges• Can localize to accessory and within short cable length

within a segment


Isothermal Relaxation CurrentRecovery Voltage


Isothermal Relaxation CurrentTest Description• Measures the time constant of trapped charges within the

insulation material as they are discharged.• Discharge current is observed for 15-30 minutes.

Field Application• Offline test that uses DC to charge the cable segment up to

1kV.• Testing is performed by a service provider.

IRC 196CDFI Meeting - Aug. 19-20 San Ramon, CA

Recovery VoltageTest Description• Similar to IRC only voltage is monitored instead of current

Field Application• Offline test that requires initial charging by DC source up to

2kV.• Testing is performed by a service provider.

Recovery Voltage

50


What does this mean for IRC & RV?

• Use limited to evaluation studies in the laboratory• Possibly too sensitive for field use


Combined Diagnostics

Multiple degradation mechanisms mean that two diagnostics are often better than one



No TestingTesting - one techniqueTesting - > one technique

27.8%

30.6%

41.7%


Combined Diagnostics 200CDFI Meeting - Aug. 19-20 San Ramon, CA

Multiple Diagnostics

Tan Delta / PDVLF / Tan Delta

Category

75.0%

25.0%


51


What Diagnostics are Combined

Local

DC Withstand

PD VLF Withstand

DC Leakage

Tan δ

Glo

bal

TDR


Drawbacks of a Single Approach• Each diagnostic looks for symptoms of one failure

mechanism– Voids and water trees cannot generally be detected by

a single technique

• Overlooks short term time evolution of diagnostic measurements

• Technique specific:– Withstand – No idea by how much segment passed– Tan δ – Cannot detect voids or electrical trees– PD – Cannot detect water trees (water filled voids)



BAD

GOOD

Advantage of Multiple Diagnostics

Bad

Diagnostic 1

Goo

d

Good Bad?

?

Diagnostic 2


DataGeneration from



52


Tan δ Ramp

Time [min]

Tan-

delt

a [1

e-3]

54321

1.8

1.7

1.6

1.5

1.4

1.3

1.2

1.1

1.0

1.01.51.7

[p.u.]Voltage



Time [min]

Tan-

delt

a [1

e-3]

1512.5107.552.50

1.8

1.7

1.6

1.5

1.4

1.3

1.2

1.1

1.0

through the 15 minute withstandStability of Tan-delta monitored



Tan δ Ramp & Monitored Withstand

Time [min]

Tan-

delt

a [1

e-3]

76543210

180

160

140

120

100

80

60

40

20

0.51.01.51.7

[p.u.]Voltage

FailureSegment HL_23_22Elbow Failure Hampton Leas


After Repair…

Time [min]

Tan-

delt

a [1

e-3]

76543210

180

160

140

120

100

80

60

40

20

0

0.51.01.51.7

[p.u.]Voltage

Failure

After Failure

Segment HL_23_22Elbow Failure Hampton Leas


53



DatabasesStandards



Cumulative Length Tested in One Year (Miles)

Wit

hsta

nd T

est

Out

com

es

9080706050403020100

U

NST

ABL

E

H

igh

Loss

Hig

h TU

Poor

Sta

bilit

y

27305

Monitored VLF Withstand to IEEE400.2 Levels

Time of failure in mins

IEEE400.2 LevelsSimple VLF Withstand to



QUESTIONS



54


Accuracies Revisited

Why do they matter?


Consequence

Diagnostic Program Costs Cost [$]

Selection

Diagnostic

CorrectiveActions Total Diagnostic

Program Cost

Accuracies Really Matter


Recall the Example...No Action Required Action Required

Avoided Corrective Actions

Avoided service failures

Accuracies Really Matter 216CDFI Meeting - Aug. 19-20 San Ramon, CA

Incorrect Diagnosis

Future service failures Unneeded Corrective Actions



55


Benefit and LossCost [$]

Selection

CorrectiveActions

Diagnostic

ConsequenceAlternate

Program 1

AlternateProgram 2

BENEFIT

LOSS


Considerations

• Diagnostic program economic calculations are based on ability to predict future failures.

• Total diagnostic program cost is more sensitive to certain elements than others.– Failure Rate– Diagnostic Accuracy– Failure Consequence



Uncertainty in Diagnostic Program Costs Cost [$]

Diagnostic

Selection

Consequence

CorrectiveActions

Program Cost

Range


AlternateProgram 1

BENEFIT

LOSS

Program Cost

Range

Cost [$]Uncertainty in Diagnostic Program Costs


56


Diagnostic Accuracy Complications

• Time is a critical factor in the assessment of accuracy.– Failures do not happen immediately after testing.

• Two approaches to computing diagnostic accuracy.– “Bad Means Failure” Approach– “Probabilistic” Approach


Failures Over Time


Year12345



Accuracy Over Time – “Bad Means Failure”

Time[Years]2 4 6 108

Accuracy[%]

100

0

No Action Required Accuracy

Action Required Accuracy?

• System Changes• Additional Aging• Increased Load



Perc

ent

101FOT

40

30

20

10

5

3

2

1

ARFSNA

Action

Probabilistic Approach - Tan δ


Perc

ent

101FOT

40

30

20

10

5

3

2

1


Action


Perc

ent

101FOT

40

30

20

10

5

3

2

1


Action


Perc

ent

101FOT

40

30

20

10

5

3

2

1


Action


Perc

ent

101FOT

40

30

20

10

5

3

2

1

293236

91011

1.71.92.3

14

4

0.66012 24


Action


57


QUESTIONS




The Things We Know Now That We Did Not Know Before


By Diagnostic Technique

VLF DC Tan Delta

PD On PD Off TDR

IRC DAC

No UseOccasionalStandardTesting

Category

CDFI Research

58


CDFI Work in Lab and Field

• Dielectric Withstand– Simple– VLF Laboratory Study

• Dielectric Loss• VLF Tan δ

– Monitored Withstand• Partial Discharge• Offline 60 Hz.

CDFI Research 230CDFI Meeting - Aug. 19-20 San Ramon, CA

CDFIDielectric Withstand

Dielectric Withstand


Dielectric Withstand• Withstand techniques are most widely used diagnostic in

the USA.

• Most utilities use VLF (either sine or cosine-rectangular) in their withstand programs.

• Test duration and voltage are critical to performance on test and in service.

• Explored the concept of “Monitored” Withstand tests.

Dielectric Withstand 232CDFI Meeting - Aug. 19-20 San Ramon, CA

Circuit Length [Conductor ft]

Perc

ent

840007200060000480003600024000120000

30

25

20

15

10

5

0

Median Length = 3500 ft

Length Distribution (Overall)

Wide variability in circuit lengths


59


Length Adjustments• Comparison of withstand failure on test rates must include

length adjustments.

2000 ft.

Choose an appropriate base length


Length Adjustments• Comparison of withstand failure on test rates must include

length adjustments.

2000 ft.

500 ft. 500 ft. 500 ft.500 ft.

Censored

Failure



Length Adjustments

• Base length must be a meaningful length (50 ft is probably not a useful length).

• Two sets of censored segments:– Pass Segments - All segments censored at test

duration– No Pass Segments

• 1 failed segment• remaining segments censored at failure time

• Multiple failure modes must be dealt with appropriately.



Failu

res

on T

est

[%]

100.010.01.00.1

20

10

5

3

2

1

1000 Feet500 FeetNONE

AdjustmentLength

Utility I – Hybrid System


Failu

res

on T

est

[%]

100.010.01.00.1

20

10

5

3

2

1

1000 Feet500 FeetNONE

AdjustmentLength


Failu

res

on T

est

[%]

100.010.01.00.1

20

10

5

3

2

1

30

4.5%

2.4%

17.5%1000 Feet500 FeetNONE

AdjustmentLength

Performance at longer test times can be predicted.

Length Weighted Average FOT

30 Mins 2.7%

60 Mins 5.0%


60


3.53.02.52.01.5

100

80

60

40

20

0

Test Voltage (U0 = Rated Voltage)

Surv

ivor

s [%

of

Test

ed]

IEEE Rec. Level

NEETRAC ExtrudedMixed (PILC and Extruded)ExtrudedPILC

Effect of Test Voltage


VLF Lab Program



239

Overview• Test program combining aging at U0 with multiple applications

of high voltage VLF.

• Uses field aged cable samples - one area within one utility.

• Evaluate the effects of – Voltage and time on the performance on test and – Subsequent reliability during service voltages.

Primary MetricSurvival during aging and testing

Secondary Metrics– Before and after each VLF application, PD at U0– Between Phase A & B IRC, PD (AC 2.2U0, DAC), Tan δ


1: No Withstand

2: VLF2.2U015 Min

3: VLF3.6U0

120 Min

4: VLF2.5U060 Min

5: VLF2.2U0

120 Min

6: 60 Hz.3.6U0

0.25 Min

Withstand Testing Periods (variable durations)

Aging Periods

Phase A

End

Failures are the primary metricfor evaluation

61


1: No Withstand

2: VLF2.2U015 Min

3: VLF3.6U0

120 Min

4: VLF2.5U060 Min

5: VLF2.2U0

120 Min

6: 60 Hz.3.6U0

0.25 Min

T1 T2 T3 T4 242CDFI Meeting - Aug. 20-21 San Ramon, CA

1: No Withstand

2: VLF2.2U015 Min

3: VLF3.6U0

120 Min

4: VLF2.5U060 Min

5: VLF2.2U0

120 Min

6: 60 Hz.3.6U0

0.25 Min

T1 T2 T3 T4

No Failures

No Failures

3 VLF FailuresNo Aging Failures


No Failures

2 60 Hz. FailuresNo Aging Failures


1: No Withstand

2: VLF2.2U015 Min

3: VLF3.6U0

120 Min

4: VLF2.5U060 Min

5: VLF2.2U0

120 Min

6: 60 Hz.3.6U0

0.25 Min

T1 T2 T3 T4

No Failures

No Failures



No Failures

2 60 Hz. FailuresNo Aging Failures


Failures on Test

100101

90

50

10

1100101

90

50

10

1

100101

90

50

10

1100101

90

50

10

1

3Uo Rated, Sine

Time of VLF Application (mins)

Prob

abili

ty o

f Fa

ilure

(%

)

15 60

10

71

2.5Uo Rated, Sine15 60

10

134

3Uo Rated, Cosine15 60

10

8

2.5Uo Rated, Cosine15 60

10

63

Uo, Ambient - Phase I & II

2Uo, 45C - Phase III


62


Voltage Effect on Times to Failure

3.02.82.62.42.2

180

160

140

120

100

80

60

40

20

03.02.82.62.42.2

Phase I & II - Uo / RT ageing, Sine

Test Voltage [Uo]

Tim

e to

10%

Fai

lure

[m

ins]

Phase III - 2Uo / 45C ageing, Cosine

Both curves show that higher voltage leads to increased failure rate


Failure Analyses - Trees & Defects in Cables

Distance Along Cable (ft)

Faile

d Sa

mpl

es

2520151050

D

C

B

A

DEFECTLARGE WATER TREEMEDIUM WATER TREESMALL WATER TREE



247

VLF Test Program Summary• Analysis of Phase A is complete.

• Phase B (2U0 aging, 45°C Cosine Rectangular) underway.

• Phases A & B show that no VLF exposed samples have failed under 60 Hz aging @ U0 & 2U0.

• Phase B tests shows two samples without VLF exposure failed during 60 Hz aging @ 2U0.

• VLF failures on test:– Less than 15 mins: 12 % (2 failures)– 15 – 60 mins: 71 % (12 failures)


CDFIDielectric Loss

Tan δ

63


Prevailing View – Tan Delta

Tan δ

Tip Up[2U0 – 1U0]

Importance


CDFI Suggestion – Tan Delta

Tan δTime

Stability

Tip Up[1.5U0 – 0.5U0]

Tan δ[U0]

Importance

Tan δ


Tan δ Time Stability

Tan-delta stdv. @ 0.1 Hz (1.5 Uo) Diagnostic Rank

Brea

kdow

n Pe

rfor

man

ce R

ank

76543210

7

6

5

4

3

2

1

0

0.160.0

[Hz]FrequencyBreakdown


VLF Tan Delta of Cable Systems

Tan Delta at Uo (E-3)

Perc

ent

1000.0100.010.01.00.1

99.9

99

90

8070605040

30

20

10

5

3

2

1

FilledPaperPE

Ins Class

•>650 segments•Mean Length 2000ft•Total length >250 conductor miles

Can segregated based on areas where the curves breakDefine areas that are “normal” and “unusual”

Tan δ

64


Cable System – Global Assessment

Tip Up (1e-3)

Tan

Del

ta (

1e-3

)

-1010-1-

-

150

6

0

Unfilled Polyolefin Insulations

No Action

Further Study

Action Required


Service Performance / Accuracy


Perc

ent

101FOT

40

30

20

10

5

3

2

1

293236

91011

1.71.92.3

14

4

0.66012 24


Action

Tan δ


CDFIPartial Discharge

PD 256CDFI Meeting - Aug. 19-20 San Ramon, CA

CDFI Work

• Analysis of historical PD field test data

• Classification

• Characterization of field samples by PD measurement in laboratory.

• Feature Extraction for Classification

PD

65


PD Charge Magnitude Distributions

540450360270180900-90

40

30

20

10

0

Charge Magnitude [pC]

Perc

ent

Cable_Failed_PCCable_NOFailed_PC

Variable

540450360270180900-90

40

30

20

10

0


Perc

ent


Variable

540450360270180900-90

40

30

20

10

0


Perc

ent


Variable


PD Inception Voltage

2.72.42.11.81.51.20.9

30

25

20

15

10

5

0

Inception Voltage [U0]

Perc

ent

Cable_Failed_IVCable_NOFailed_IV

Variable

2.72.42.11.81.51.20.9

30

25

20

15

10

5

0


Perc

ent


Variable

2.72.42.11.81.51.20.9

30

25

20

15

10

5

0


Perc

ent


Variable

PD


BAD

GOOD

Multi Feature Classification

Bad

Criterion 1

Goo

d

Good Bad?

?

Criterion 2


Classification - PD Magnitude & PDIV

Neighbors Used in Classification [#]

Succ

es R

ate

[% o

f Te

sted

]

3937353331292725232119171513119753

100

90

80

70

60

50

40

30

20

10

0

fail_successnofail_successOverall_success

Variable

Neighbors Used in Classification [#]

Succ

es R

ate

[% o

f Te

sted

]

3937353331292725232119171513119753

100

90

80

70

60

50

40

30

20

10

0

fail_successnofail_successOverall_success

Variable

pC and PDIV are not sufficient to get high classification accuracy

PD

66


Cluster No. Feature Name1 Pos. Phase Range [deg]2 Pos. Mean Phase [deg]

3

Pos. Qmax [pC]Neg. Qmax [pC]Neg. Qmean [pC]Pos. Qmean [pC]Pos. Mean Energy [pC*V]Pos. Max Energy [pC*V]Neg. Max Energy [pC*V]

Neg. Mean Energy [pC*V]

4 Neg. Phase Range [deg]5 Neg. Mean Phase [deg]

6 DMean Energy Ratio

7 Nw [pulses/cycle]

Cluster No. Feature Name1 Pos. Phase Range [deg]2 Pos. Mean Phase [deg]

3

Pos. Qmax [pC]Neg. Qmax [pC]Neg. Qmean [pC]Pos. Qmean [pC]Pos. Mean Energy [pC*V]Pos. Max Energy [pC*V]Neg. Max Energy [pC*V]

Neg. Mean Energy [pC*V]

4 Neg. Phase Range [deg]5 Neg. Mean Phase [deg]

6 DMean Energy Ratio

7 Nw [pulses/cycle]

PD Lab Data - Cluster Variable Analysis

261PD 262CDFI Meeting - Aug. 19-20 San Ramon, CA

Partial Discharge Diagnostic Features

Sim

ilari

ty L

evel

[%

]

Nw [p

ulses

/cycl

e]

Mean En

ergy

Rati

oD

Neg.

Mean Ph

ase [d

eg]

Neg.

Phas

e Ran

ge [d

eg]

Neg. M

ean E

nergy [

pC*k

V]

Neg.

Max En

ergy

[pC*

kV]

Pos.

Max En

ergy

[pC*

kV]

Pos.

Mean E

nergy [

pC*k

V]

Pos.

Qmea

n [pC

]

Neg. Q

mea

n [pC

]

Neg.

Qmax [p

C]

Pos.

Qmax

[pC]

Pos.

Mean Ph

ase [d

eg]

Pos.

Phas

e Ran

ge [d

eg]

15.18

43.45

71.73

100.00


Sim

ilari

ty L

evel

[%

]

Nw [p

ulses

/cycl

e]

Mean En

ergy

Rati

oD

Neg.

Mean Ph

ase [d

eg]

Neg.

Phas

e Ran

ge [d

eg]

Neg. M

ean E

nergy [

pC*k

V]

Neg.

Max En

ergy

[pC*

kV]

Pos.

Max En

ergy

[pC*

kV]

Pos.

Mean E

nergy [

pC*k

V]

Pos.

Qmea

n [pC

]

Neg. Q

mea

n [pC

]

Neg.

Qmax [p

C]

Pos.

Qmax

[pC]

Pos.

Mean Ph

ase [d

eg]

Pos.

Phas

e Ran

ge [d

eg]

15.18

43.45

71.73

100.00

PD Lab Data - Cluster Variable Analysis

262Partial Discharge Diagnostic Features

Sim

ilari

ty L

evel

[%

]

Nw [p

ulses

/cycl

e]

Mean En

ergy

Rati

oD

Neg.

Mean Ph

ase [d

eg]

Neg.

Phas

e Ran

ge [d

eg]

Neg. M

ean E

nergy [

pC*k

V]

Neg.

Max En

ergy

[pC*

kV]

Pos.

Max En

ergy

[pC*

kV]

Pos.

Mean E

nergy [

pC*k

V]

Pos.

Qmea

n [pC

]

Neg. Q

mea

n [pC

]

Neg.

Qmax [p

C]

Pos.

Qmax

[pC]

Pos.

Mean Ph

ase [d

eg]

Pos.

Phas

e Ran

ge [d

eg]

15.18

43.45

71.73

100.00


Sim

ilari

ty L

evel

[%

]

Nw [p

ulses

/cycl

e]

Mean En

ergy

Rati

oD

Neg.

Mean Ph

ase [d

eg]

Neg.

Phas

e Ran

ge [d

eg]

Neg. M

ean E

nergy [

pC*k

V]

Neg.

Max En

ergy

[pC*

kV]

Pos.

Max En

ergy

[pC*

kV]

Pos.

Mean E

nergy [

pC*k

V]

Pos.

Qmea

n [pC

]

Neg. Q

mea

n [pC

]

Neg.

Qmax [p

C]

Pos.

Qmax

[pC]

Pos.

Mean Ph

ase [d

eg]

Pos.

Phas

e Ran

ge [d

eg]

15.18

43.45

71.73

100.00


Sim

ilari

ty L

evel

[%

]

Nw [p

ulses

/cycl

e]

Mean En

ergy

Rati

oD

Neg.

Mean Ph

ase [d

eg]

Neg.

Phas

e Ran

ge [d

eg]

Neg. M

ean E

nergy [

pC*k

V]

Neg.

Max En

ergy

[pC*

kV]

Pos.

Max En

ergy

[pC*

kV]

Pos.

Mean E

nergy [

pC*k

V]

Pos.

Qmea

n [pC

]

Neg. Q

mea

n [pC

]

Neg.

Qmax [p

C]

Pos.

Qmax

[pC]

Pos.

Mean Ph

ase [d

eg]

Pos.

Phas

e Ran

ge [d

eg]

15.18

43.45

71.73

100.00

50 % Similarity Level

1 2 3 4 5 6 7

3a 3b

PD


QUESTIONS



67


Selecting a Diagnostic TechnologyKnowledge-Based System

Selecting a Diagnostic Technology 266CDFI Meeting - Aug. 19-20 San Ramon, CA

KBS• Selecting the right diagnostic is not easy.

• No one diagnostic covers everything.

• How you measure is influenced by what you do with the results.

• The KBS captures the experience and knowledge of people who have been operating in the field

Selecting a Diagnostic Technology


Knowledge Based Systems• Knowledge-Based Systems are computer systems that are programmed to imitate human problem-solving.

• Uses a combination of artificial intelligence and reference to a database of knowledge on a particular subject.

• KBS are generally classified into:– Expert Systems– Case Based Reasoning– Fuzzy Logic Based Systems– Neural Networks


Extruded Cable Diagnostics


68


KBS Example


Short Listing of Diagnostic Approaches

Exp

erts

Rec

omm

endi

ng a

Dia

gnos

tic

Tech

niqu

e (%

) 100

80

60

40

20

0

Simple Withstand Monitored WithstandDischarge

Simple Withstand Historical DataDischargeDielectric

Simple WithstandMonitored Withstand

EXPERTSBY FEWESTRECOMMENDED

BY MOST EXPERTSRECOMMENDED



Impact of Remedial Action

• Hybrid Cable System• Most service failures occur in Accessories• Usual remediation is by replacement of cable sections

High

Low

Medium

Service Failure Rate

40 - 5025Paper

0 - 1042EPR

20 - 3033PE

Age[yrs]

Portion [%]

System Component


Expe

rt R

ecom

men

dati

on (

%)

Replace Small Section

Replace Segment

Replace Accessories

Simple Withstand

Monitored Withstand

History & TDRDischarge

Dielectric

Simple Withstand

Monitored W ithstand


Dielectric

Simple Withstand



Dielectric

100

80

60

40

20

0

Hybrid Cable System

Expe

rt R

ecom

men

dati

on (

%)


Replace Segment

Replace Accessories

Simple Withstand

Monitored Withstand


Dielectric

Simple Withstand



Dielectric

Simple Withstand



Dielectric

100

80

60

40

20

0

Expe

rt R

ecom

men

dati

on (

%)


Replace Segment

Replace Accessories

Simple Withstand

Monitored Withstand


Dielectric

Simple Withstand



Dielectric

Simple Withstand



Dielectric

100

80

60

40

20

0


Expe

rt R

ecom

men

dati

on (

%)


Replace Segment

Replace Accessories

Simple Withstand

Monitored Withstand


Dielectric

Simple Withstand



Dielectric

Simple Withstand



Dielectric

100

80

60

40

20

0

Most Recommended

69


QUESTIONS




Summary


What we have learned about diagnostics (1)1. A developing database of field failure diagnostic data shows

that different diagnostic techniques can provide someindication about cable system condition.

2. Even if the diagnostics themselves are imprecise, diagnostic programs can be beneficial.

3. Benefits can be quantified, however this is not simple and requires effort.

4. Many different data analysis techniques, including some non conventional approaches, are needed to assess diagnostic effectiveness.

5. Utilities HAVE to act on ALL replacement/repair recommendations to get improved reliability.

Summary

70


What we have learned about diagnostics (2)6. PD, VLF, DC and Tan δ & VLF withstand tests detect problems

in the field and can be used to improve system reliability.

7. It is difficult to predict whether or not the problems/defects detected by PD or Tan δ will lead to failure.

8. PD assessments are good at establishing groups of cable system segments that are not likely to fail.

9. Tan δ measurements provide a number of interesting features for assessing the condition of cable systems.

10.Tan δ & PD measurements require interpretation to establish how to act.

Summary 278CDFI Meeting - Aug. 19-20 San Ramon, CA

11. Interpretation of PD measurements is more complex than interpretation of Tan δ measurements.

12. IRC & RV are particularly difficult to deploy in the field.

What we have learned about diagnostics (3)

Summary


DC

Voltage

Time

Diagnostic Information

Info

rmat

ion

Con

tent

High

Low

Ease of Utility ImplementationSimple Some Skill Difficult

VLFDAC


Voltage

Time

Info

rmat

ion

Con

tent

High

Low


VLFDAC

TDRamp

PDRamp



DC

71


DC

Voltage

Time

Info

rmat

ion

Con

tent

High

Low


VLFDAC

TDRamp

PDRamp



Preset Test ProtocolsAnalysis of data performed after

tests are completed.

Adaptive Test ProtocolsTests adjusted in real time according to

analysis/decisions made during each test.


Voltage

Time

Info

rmat

ion

Con

tent

High

Low

Simple Some Skill Difficult

VLF

DC

DAC

TDRamp

PDRamp

TD/PD Monitored Withstand

Ease of Utility Implementation



?


Voltage

Time

Info

rmat

ion

Con

tent

High

Low


VLFDAC

TDRamp

PDRamp

TD Ramp w/TD Monitored Withstand

PD Ramp w/PD Monitored Withstand





?

DC


Info

rmat

ion

Con

tent

High

Low


VLFDAC

TDRamp

PDRamp

TD & PD Ramp TD & PD Monitored Withstand


Voltage

Time



TD Ramp w/TD Monitored Withstand

PD Ramp w/PD Monitored Withstand


?

DC

72


Reflections• Approach to data analysis established in CDFI

• Many questions answered, there still remain gaps in our understanding of:– Benefits– Distinguishing anomalies from weaknesses

• Answers will come with continued analysis of field test data (diagnostic tests followed by circuit performance monitoring) as well as controlled laboratory tests.

• The potential value of continued analysis is high.

Summary 286CDFI Meeting - Aug. 19-20 San Ramon, CA

CDFI Phase 1 Extension

Schedule: October 1, 2009 - September 30, 2010

Tasks• VLF Withstand• Defects• Field Surveys• Regional Meetings

Summary


CDFI Phase IISchedule: January 2010? (3 Year Duration)

Tasks• High Voltage Testing • Commissioning Tests• Field Demonstrator• Field Testing

– Revisits/Trending– Challenging Utility Regions

• Diagnostic Reference Handbook• Knowledge-Based System

Summary

Phase II ParticipantsCurrent CDFI

EPRIDOE

New entrants


QUESTIONS

GTRI/DOE Disclaimer Cable Diagnostic Focused Initiative Regional ... CDFI Meeting.pdf · Cable...

Documents

Transcript of GTRI/DOE Disclaimer Cable Diagnostic Focused Initiative Regional ... CDFI Meeting.pdf · Cable...