cdfi icc education spring 2009 minutes - used slides & … · 2009. 7. 16. · 6 21 presented...

1

1Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

ICC Education Session

Cable Diagnostic Focused Initiative

Nigel HamptonRick Hartlein

Joshua Perkel

Spring 2009 Meeting


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


Outline

• 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


CDFI Background

Rick Hartlein

2


• 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 6Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

Composition of US MV system

Inst

alle

d Ca

paci

ty (

%)

UNKNOWNTRXLPEEPRXLPEHMWPEPILC

100

90

80

70

60

50

40

30

20

10

0

25

50

75

CDFI Background/Overview


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


3


NEETRAC Members

Non NEETRACMembers Supporters

Dept of Energy

Diagnostic Providers

CDFI


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


4


CDFI Activities

CDFI

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


CDFI Activities

Lab 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


5


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


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

6


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.


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


7


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

Cable System Failure ProcessRick Hartlein


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


Major Cable Components

Jacket (Recommended)

Metallic Shield/Neutral

Insulation Shield

Insulation

Conductor or Strand ShieldConductor

Cable System Failure Process 28Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

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


8


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 growth

Electrical tree growing from water tree

Cable System Failure Process 30Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

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.


9


SAGE Approach to

Diagnostic Programs

Nigel Hampton


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


SAGE at WorkFailures [#]

Time

Selection

Action

Generation

Evaluation

Decreasing Failures

Increasing Failures

SAGE Concept 36Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

When to deploy diagnostics

Time (Years)

Cabl

e Sy

stem

Per

form

ance

403020100

Operational Stress

Condition AssessmentCommissioning

10


Global ContextComparison with many tests

DatabasesStandards

Context – is important

Local ContextComparisons within one area

DataGeneration from

Diagnostic Measurement


Case StudyRoswell, GA

November 2008 & January 2009

Nigel Hampton

TDRTan Delta

Monitored WithstandOffline PD


39

Roswell Map

Case Study: Roswell 40Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

SELECTION

Case Study: Roswell

11


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

Selecting a Diagnostic Technology


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


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.

Case Study: Roswell

12


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


ACTION

Case Study: Roswell


Initial Corrective Action Options

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

• Complete system replacement.


GENERATION

Case Study: Roswell

13


Overhead and Cabinet Terminations


Monitored Withstand

Case Study: Roswell


If this had been a Simple Withstand

Length Tested (miles)1086420

18 Segments Tested

No Failures On Test


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

14


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


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

Case Study: Roswell


Targeted Offline PD


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

Case Study: Roswell

15


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


EVALUATION

Case Study: Roswell


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 4 months from

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) 5 months from test


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.

Case Study: Roswell

16


Diagnostic Accuracies

Nigel Hampton


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.



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

Diagnostic Accuracies 64Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

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

But the diagnostic is imperfect...


17


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.”

Diagnostic Accuracies 66Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

“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 [%

]


Diagnostic Testing Technologies

Nigel Hampton

18


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)

Diagnostic Testing Technologies 70Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

Utility Use of Diagnostics



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.


No TestingTesting - one techniqueTesting - > one technique

27.8%

30.6%

41.7%

Survey of Use of Diagnostics

19




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


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




DatabasesStandards

Context


DataGeneration from



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?


20


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.


Withstand Test Process

HOLDINITIAL

Time

Voltage

t = 0 tTest

Voltages and Times for VLF covered in IEEE 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



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



21


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





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.



DatabasesStandards


22


Withstand Testing Experience

Time on Test [Minutes]

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


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

Days

637

Days

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


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.


23


Dielectric Loss (Tan δ)Dielectric losses - 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 it consists of two parameters; a resistor and a capacitor [IEEE Std. 400]

• When voltage is applied to the cable, the total current will be the contributions of the capacitor current and the resistor current


Cable System Equivalent

Cable system (cable, splices, and terminations) is reduced to simple circuit.



DataGeneration from



Tan δ 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


24




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


Lengths within a Locality

Length (ft)

Tan

Del

ta (

1e-3

)

50004000300020001500

10

1

25



DatabasesStandards


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

“Cool Wall”


Tan δ Interpretation

Tip Up

Tan

Del

ta

-1010-1-

-

150

6

0

No Action

Further Study

Action Required


Based on 258 Conductor Miles


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.


26


TDR Principles

Near End

TDREquipment

Far End

JointL

Joint


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 only be performed by a service provider.• Assessment criteria are unique to each embodiment of the

technology



DataGeneration from



Discharge Occurrence

No PD PD


27




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



DatabasesStandards


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

28


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


PD

P1 P2 P3

Time P1-P3

Time P2-P3

Am

plitu

de [m

V]

-50

10152025

5

0 2 4 6 8Time [μs]

Near End

DetectionEquipment

Far EndP1P2

P3

PD Activity



DataGeneration from



PD Pulse

140 mV

180 pC


29


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





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

30


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



119

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

PD

31


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



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.


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.


32


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%



Multiple Diagnostics

Tan Delta / PDVLF / Tan Delta

Category

75.0%

25.0%


DataGeneration from



33


Monitored Withstand - Data

Time [min]

Tan-

delt

a [1

e-3]

20151050

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

vs. Voltagegives Tan-deltaOptional Ramp Up

through the 15 minute withstandStability of Tan-delta monitored


Monitored Withstand Data - Elbow

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

Tan-delta vs Voltage

Tan-delta Stability

Before Failure

Failure

After Failure

Segment HL_23_22Elbow Failure Hampton Leas

“Cool Wall”



DatabasesStandards


Monitored Withstand

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

Teste

34


Accuracies Revisited

Why do they matter?

Josh Perkel


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 136Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

Incorrect Diagnosis

Future service failures Unneeded Corrective Actions

No Action Required Action Required


35


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



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


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


36


Probabilistic Approach – Tan δ

100101FOT

5

2

1

0.01

Elasped Time at Oct 08 (Month)

Failu

res

in S

ervi

ce [

% o

f Te

sted

Seg

men

ts]

ActNo Action

Action_1

100101FOT

5

2

1

0.01


Failu

res

in S

ervi

ce [

% o

f Te

sted

Seg

men

ts]

ActNo Action

Action_1

100101FOT

5

2

1

0.01


Failu

res

in S

ervi

ce [

% o

f Te

sted

Seg

men

ts]

ActNo Action

Action_1


100101FOT

5

2

1

0.01


Failu

res

in S

ervi

ce [

% o

f Te

sted

Seg

men

ts]

12

7.4%

0.9%

ActNo Act

Action_1


Probabilistic Approach - PD

Days Between Test & Failure

Perc

ent

1000100101

99

9080706050403020

10

532

1

0.01

No PDPD

PD Class



The Things We Know Now That We Did Not Know Before


By Diagnostic TechniqueVLF DC Tan Delta

PD On PD Off TDR

IRC DAC

No UseOccasionalStandardTesting

Category


37


CDFIDielectric Withstand

Josh Perkel

CDFI Research 146Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

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.

CDFI Research


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


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

length adjustments.

2000 ft.

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

Censored

Failure

CDFI Research

38



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%


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

CDFI Research


VLF Lab Program

Josh Perkel


152

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

CDFI Research

39


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

CDFI Research

Failures are the primary metricfor evaluation


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 Fails

14 Survive

No Fails14 Survive

3 VLF Fails11 Survive

2 VLF Fails12 Survive

No Fails14 Survive

2 60 Hz. Fails12 Survive

CDFI Research

No aging failures for any condition


Time of Failure on Test

Time on VFL Test (min)

Perc

ent

of 2

0 ft

Sam

ples

Fai

ling

100101

99

90

8070605040

30

20

10

5

3

2

1

15 30 60

63

253

C orrelation 0.997

Shape 1.46825Scale 254.187Failure 5C ensor 51

Max Time120 Mins

No Failures•Before 15 mins•After 60 mins


Testing Voltage (Uo)

Perc

ent

7.06.05.04.03.02.01.5

99

90

8070605040

30

20

10

5

3

2

1

6.438

25.257

2.5

3.6

C orrelation 1.000

Shape 4.04729Scale 4.88340Failure 5C ensor 51

156

Voltage of Failure on Test

More failures occur at higher test voltages

CDFI Research

40


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


158

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 @ Uo & 2Uo.

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

• All failures occurred at the appropriate time. i.e. within the VLF testing periods.

• 80% (4 out of 5) of VLF failures between 15 and 60 mins

CDFI Research


Selecting a Diagnostic TechnologyKnowledge-Based System

Nigel Hampton

Selecting a Diagnostic Technology 160Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

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


41


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



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


42


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


Summary

Rick Hartlein


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

43


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 very difficult to predict whether or not the problems/defects detected by PD and Tan δ will lead to failure in the short/medium term.

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 170Presented at the Spring 2009 ICC Education Session Copyright GTRC 2009

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


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

cdfi icc education spring 2009 minutes - used slides & … · 2009. 7. 16. · 6 21 presented...

Documents