Partial Discharge in the Context of Distribution Cable Testing

1
Partial Discharge in the Context of Distribution
Cable Testing

• Steven Boggs

2
What is PD?

• A gas discharge which does not bridge the system
  electrodes
• Discharge in a cavity
• Corona off an electrode
• Tracking discharge along an interface
• Discharge from electrical tree growth

3
Not All PD is Dangerous

• PD between floating metallic components
• PD between a metallic component and semicon
• PD between concentric neutral wires and semicon
  rendered nonconducting by solvents
• PD involving PD-resistant insulation

4
What Do We Detect?

• The discharge causes a transient increase in the
  capacitance between the electrodes by shorting
  out some of the field
• The increase in capacitance results in a
  reduction in the voltage across the electrodes
• The power supply recharges the electrodes

5
Circuit Prior to Discharge
6
Circuit During Discharge
During Discharge to Discharge
7
Cavity Before and After Discharge
8
PD Inception and Extinction

• Can have peak field in cavity prior to inception
• Can have peak-to-peak field in cavity after
  inception
• Thus PD inception can require up to twice the
  voltage of PD extinction although 1.5 to 1.7 pu
  is more common in practice
• This is the basis for PD testing to 2 to 2.5 pu

9
Typical Discharge Pattern
Applied Voltage
Field in Cavity
10
Phase-Resolved PD Analysis Provides Greater
Specificity
11
Discharge Magnitudes for Spherical Cavities
12
Discharge Magnitudes for Long, Thin Structures
PD Magnitude Depends Mainly on Length in
the Direction of the Field
13
PD Inception Depends on a free Electron

• Most free electrons come from cosmic rays
• 3 e-/cm3-s in air, 5 minutes waiting for a 1
  mm3 cavity
• After the first PD, charge is stored on the
  cavity wall
• XLPE has a negative electron affinity
• Free electrons will go from the polymer to a
  vacuum
• For small cavities much below 1 mm, PD is often
  not self-sustaining
• Charge may disappear by surface conduction
  between PDs

14
PD Detection Sensitivity

• Can detect cavities in the range of 1 mm dia
• Can detect electrical trees in the range of 1 mm
  long
• Can detect tracking along dielectric interfaces
  of some mm.
• Cannot detect water trees unless they convert to
  electrical trees

15
Tracking Along an Interface
16
Tracking Along an Interface
17
Time Domain Cable PD Testing

• Off-Line PD Detection
• Testing at 1.5 to 2.5 pu to initiate PD which can
  sustain at normal operating voltage
• Locate PD by reflected pulse
• Must trigger on first pulse above noise
• Sensitivity could be increased by about 10 if
  could implement on-line DSP treatment of data
  stream
• Can use DSP to enhance detection of 2nd pulse
• Energize with AC, LF, damped oscillation, etc.

18
Frequency Domain PD Testing

• Use spectrum analyzer to find region where noise
  is small and compare background to signal during
  test
• Can use spectrum analyzer in zero span mode (as
  bandpass filter) to correlate signal with
  operating voltage
• Sensitive enough for in-service PD testing
• Spectrum analyzer averages to improve S/N

19
Considerations for Time Domain Field PD Testing
of Cable?

• Sufficient Bandwidth for PD Location
• Sufficient Dynamic Range for PD Detection in
  Noise
• Effect of Far End Termination on Reflected Pulse
• Detection of Second Pulse in Noise

20
Considerations for Frequency Domain PD Testing

• PD vs other power frequency correlated noise
  (SCRs, etc.)
• Ability to locate PD sources mostly based on
  frequency content which depends on cable
  attenuation and varies with cable type
• Calibration of PD measurement is problematic

21
Assumptions for In-Service Testing

• The system has been operating for a long time
• If PD could initiate, it would initiate from
  surges on the system, etc.
• Once initiated, PD is likely to continue as the
  PDEV is much less (50 to 70) of the PDIV
• Thus if PD can occur on the system, it is likely
  to be present during an in-service test

22
PD Propagation in a Cable
Unlimited Bandwidth
With Effect of 5-pole, 15 MHz Input Filter
23
Frequency Domain Detection
24
What Do You Want From PD Testing?

• An indication of where to spend your money
• Location of defective components where systematic
  installation errors have been made
• Location of defective installed components
• Location of cable nearing end of life

25
But What Do You Do When Someone Recommends You
Spend 1M?

• How certain do you have to be?
• What level of proof do you demand in terms of
  evidence on your system and evidence of past
  success?
• What kind of data do you want out of your PD
  tests?

26
Assumptions for In-Service Testing

• The system has been operating for a long time
  if PD could incept, it has incepted
• Testing will take place without outages which
  extinguish PD
• A correlation with power frequency in the zero
  span mode indicates PD
• Tests can be carried out at frequent intervals
  along the cable, if necessary, to locate PD sites

27
Can Water Trees Be Detected?

• Not directly Water trees do not generate PD
• Water trees can be detected if they cause
  inception of an electrical tree
• But will the electrical tree grow to failure?

28
So What Makes Sense?

• Specify detailed report with data and written
  analysis thereof
• Require a detailed explanation of what the test
  crew will do and on what they base their analysis
• Require advanced knowledge of what the test crew
  needs to know about your system
• Location and, if possible, model of splices
• Type of cable, XLPE, TR-XLPE, EPR (brand), PILC

29
Do Routine PD Tests Make Sense?

• If we had routine pregnancy testing of the
  population, we would probably have 100,000
  pregnant males false positives!
• Many tests only make sense when applied to a
  suspect population depends on rate of false
  positives and false negatives

30
What are the Risks Off-Line?

• Polymer high field degradation has a distinct
  threshold above which the rate of degradation
  increases rapidly
• A few minutes at 2.5 pu could be equal to a long
  time at normal operating voltage
• Risk of causing a water tree to convert to an
  electrical tree which grows to failure either
  during ac or during the impulse which results
  from a breakdown
• Cant really quantify risks only time will tell
• But risks can be minimized by testing at no more
  than 2 to 2.5 pu. The rational for testing at gt2
  pu is not clear

31
What are the Risks In-Service

• No risk from elevated voltage
• Risk of misdiagnosing power frequency correlated
  interference from power electronics (e.g., SCR
  switching)
• Risk that PD source cannot be located with
  sufficient accuracy or PD from accessories cannot
  be distinguished from cable PD

32
Conclusion

• The technology is not yet mature
• Primary value today is probably investigating
  known problem areas rather than routine testing
• Time-domain off-line testing provides good PD
  location accuracy, but testing at elevated
  voltage is required as are outages.
• In-service testing has no risk of causing
  failures, but little has been published which
  quantifies efficacy