As part of my equipment strategy, I am planning on doing some Partial Discharge Testing of my MV Switchgear, MV Cables and Generators. My MV system is 6900 VAC. I would like to hear comments, pro or con, on this type of testing technology, from any one who has used it in their PM programs. What was your experience with this technology, and do you feel it gave you useful data.
Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations cowski on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!
Partial Discharge Testing, Pros or Cons. 8
- Thread starter jburn
- Start date
-
2
- #2
electricpete
Electrical
I work at a plant which has PD installed on approx 30 13.8kv motors.
Pro's
- Relatively easy to take the data (takes about 15 minutes per machine... once per six months per machine).
- Does not require removing the machine from service.
- I believe in some cases the results may be more probing than simple insulation resistance tests.
- There is a lot of info available which assists you to attempt to predict where the failure might lie so you can plan repairs.
- Some of the vendors offer good training and support on interpretation of results and general advice on machines.
- There are conferences where info is exchanged on experience with this technology.
- Some vendors provide a good database of results.
Con's
- It can be somewhat time-intensive to review the data for a large number of machines. If all machines are at low levels it's not a problem. Machines which have elevated or increasing levels require study. Approx 50% of the machines at our site exceed the warning level established by the vendor.
- The coupling capacitor method may suffer from lower sensitivity to problems on coils deep in the winding (not at the line terminals). There are also the coils we expect to have fewest problems (but they still do have some problems).
The vendor will tell you that it is not a perfect technology and best when combined with other available test and inspection techniques.
Pro's
- Relatively easy to take the data (takes about 15 minutes per machine... once per six months per machine).
- Does not require removing the machine from service.
- I believe in some cases the results may be more probing than simple insulation resistance tests.
- There is a lot of info available which assists you to attempt to predict where the failure might lie so you can plan repairs.
- Some of the vendors offer good training and support on interpretation of results and general advice on machines.
- There are conferences where info is exchanged on experience with this technology.
- Some vendors provide a good database of results.
Con's
- It can be somewhat time-intensive to review the data for a large number of machines. If all machines are at low levels it's not a problem. Machines which have elevated or increasing levels require study. Approx 50% of the machines at our site exceed the warning level established by the vendor.
- The coupling capacitor method may suffer from lower sensitivity to problems on coils deep in the winding (not at the line terminals). There are also the coils we expect to have fewest problems (but they still do have some problems).
The vendor will tell you that it is not a perfect technology and best when combined with other available test and inspection techniques.
-
3
- #3
We have been using PDSL (Partial Discharge Site Location) testing as a method to help direct maintenance on our MV underground distribution cables. We haven't used it too much for switchgear diagnostics, though (because to some extent, we expect to find PD in air-insulated gear when testing above rated voltage).
Our distribution voltages are 34.5kV and 12.47kV (phase-to-phase), and we have had similar results at both voltages, although there are some additional challenges at 34.5kV.
Initially, we started to use PDSL as a means of determining the health and remaining life of our cables. We started to discover, though, that if we found any discharge generated from electrical trees in the cable insulation, we did not have much time left before the cable failed (this is specific to XLPE insulation, incidentally). Typically, the remaining life of an XLPE cable with discernible PD is in the realm of a few hours to a few days, depending on the partial discharge inception voltage (PDIV) of the defect and the frequency of transients in the ditribution system that are likely to exceed the PDIV.
My next statement is probably somewhat unique to our particular distribution system, but we have found that the majority of our problems are not in the cable, but in the accessories used to put them together (splices, terminations, elbows, etc.). Our normal failure statistics indicate a ratio of roughly 4:1 for accessories to cable. PD testing revealed a ratio of "hits" in accesories to cable much higher than that, but I suspect that is because we are intentionally testing at 1.5-2.0 PU.
We haven't made any drastic decisions to pull everything out that discharges, though. This is because the splices and accessories we are using are a little more discharge-tolerant than XLPE cable. In several instances (mostly dependent on the customer being served) we have removed some "suspect" items from service and gone through forensic analysis. In almost all of these cases, we have found workmanship errors that were the cause of the PD in the accessories, so we feel that PDSL is at least validated in that respect.
As far as the different technologies used to accomplish diagnostic testing, we have not yet settled on PD alone as the answer. I tend to believe that we will end up using a combination of PD, tangent delta, and assessment of the condition of the concentric neutrals as a measure of the health of our distribution system.
Our distribution voltages are 34.5kV and 12.47kV (phase-to-phase), and we have had similar results at both voltages, although there are some additional challenges at 34.5kV.
Initially, we started to use PDSL as a means of determining the health and remaining life of our cables. We started to discover, though, that if we found any discharge generated from electrical trees in the cable insulation, we did not have much time left before the cable failed (this is specific to XLPE insulation, incidentally). Typically, the remaining life of an XLPE cable with discernible PD is in the realm of a few hours to a few days, depending on the partial discharge inception voltage (PDIV) of the defect and the frequency of transients in the ditribution system that are likely to exceed the PDIV.
My next statement is probably somewhat unique to our particular distribution system, but we have found that the majority of our problems are not in the cable, but in the accessories used to put them together (splices, terminations, elbows, etc.). Our normal failure statistics indicate a ratio of roughly 4:1 for accessories to cable. PD testing revealed a ratio of "hits" in accesories to cable much higher than that, but I suspect that is because we are intentionally testing at 1.5-2.0 PU.
We haven't made any drastic decisions to pull everything out that discharges, though. This is because the splices and accessories we are using are a little more discharge-tolerant than XLPE cable. In several instances (mostly dependent on the customer being served) we have removed some "suspect" items from service and gone through forensic analysis. In almost all of these cases, we have found workmanship errors that were the cause of the PD in the accessories, so we feel that PDSL is at least validated in that respect.
As far as the different technologies used to accomplish diagnostic testing, we have not yet settled on PD alone as the answer. I tend to believe that we will end up using a combination of PD, tangent delta, and assessment of the condition of the concentric neutrals as a measure of the health of our distribution system.
-
3
- #4
jbartos
Electrical
- Jan 15, 2000
- 6,440
electricpete
Electrical
JBARTOS - you get a star for some very good links.
I have made more than 200 tests at partial discharges with an MPS equipment, and I have some doubt on how interpret the output result (Vi, Ve, pC). I use it for evalue the insulation on dielectric heatsink, but we don't know which is the acceptable level of pC at 3kV (the operating voltage) to be confident on a 20 year life.the insulant is 280x300mm silicon sheets, thermal conductor. I have read the Konig-Rao book, but I cannot find useful data from oscilloscope on the kind of PD. Let's start talking about that, I have the impression that a carefull interpretation of that data can say much on the life, quality, assembling of the electric insulation.
How to conduct a test? I start with Vi, I write down Vi and pC, then I go to 3kV, I write down pC after waitnig 15-30 sec when is stabilzed, then lowering V until pC go under 10 pC and I write kV as Ve (estiction). Is that correct?
Thanks for the replay.
How to conduct a test? I start with Vi, I write down Vi and pC, then I go to 3kV, I write down pC after waitnig 15-30 sec when is stabilzed, then lowering V until pC go under 10 pC and I write kV as Ve (estiction). Is that correct?
Thanks for the replay.
electricpete
Electrical
I've never done the off-line test you mention... only the on-line test on a motor operating and energized from it's normal supply... so take any comments I have with a grain of salt.
One thing I understand they look at during the offline test is the inception voltage (when the PD starts on the way up) in addition to the exctinction voltage you mention. In general, a trend downward overtime in either inception or extinction voltages or a lower value compared to comparable "sister" test specimens (as well as an increase in activity at given voltage) would be an indication of a problem. This trending/comparison process is viewed by some as more valuable than a comparison to some absolute limit, because of the large number of variables between different types of test specimens.
Also the pulses can be categorized as "positive" or "negative", according to the polarity of the current in the test capacitor which is actually opposite the sign of the voltage-to-ground at the time the pulse occurs. Positive pulses occur during negative half-cycle of the voltage waveform and are centered at approx 225 degrees into a sin(theta) voltage waveform. Negative pulses occur during the positive half cycle and are centered approx theta = 45 degrees into the "sin(theta=2Pift)" waveform.
Positive pulses imply voids on the outer surface of the insulation (between insulation and some ground plane), while negative pulses imply voids on the inner surface of the insulation (between insulation and energized condutor). If positive and negative pulses are approx equal, it is often assumed that the voids are in the bulk insulation(these assumptions are from the motor world).
Also if phase-to-phase arcing is possible in your application, that would likely show up with pulses shifted 30 degrees from the above-mentioned 45 degree and 225 degree positions.
At a given voltage level, the magntitude of individual pulses is somtimes taken as an indication of the size of the associated voids, while the pulse count rate is taken as an indication of the number of voids present in the insulation.
One thing I understand they look at during the offline test is the inception voltage (when the PD starts on the way up) in addition to the exctinction voltage you mention. In general, a trend downward overtime in either inception or extinction voltages or a lower value compared to comparable "sister" test specimens (as well as an increase in activity at given voltage) would be an indication of a problem. This trending/comparison process is viewed by some as more valuable than a comparison to some absolute limit, because of the large number of variables between different types of test specimens.
Also the pulses can be categorized as "positive" or "negative", according to the polarity of the current in the test capacitor which is actually opposite the sign of the voltage-to-ground at the time the pulse occurs. Positive pulses occur during negative half-cycle of the voltage waveform and are centered at approx 225 degrees into a sin(theta) voltage waveform. Negative pulses occur during the positive half cycle and are centered approx theta = 45 degrees into the "sin(theta=2Pift)" waveform.
Positive pulses imply voids on the outer surface of the insulation (between insulation and some ground plane), while negative pulses imply voids on the inner surface of the insulation (between insulation and energized condutor). If positive and negative pulses are approx equal, it is often assumed that the voids are in the bulk insulation(these assumptions are from the motor world).
Also if phase-to-phase arcing is possible in your application, that would likely show up with pulses shifted 30 degrees from the above-mentioned 45 degree and 225 degree positions.
At a given voltage level, the magntitude of individual pulses is somtimes taken as an indication of the size of the associated voids, while the pulse count rate is taken as an indication of the number of voids present in the insulation.
electricpete
Electrical
I'll assume that when you say rising edge of absolute value, you mean 0-90 degrees and 180-270 degrees into a sin wave representing voltage to ground.
These are the two typical locations for pulses to occur.
The reason appears to be: #1 - A given void will only discharge once per ac cycle: #2 - the discharge occurs very "soon" (in comparison to ac cycle time) after the threshhold is reached.
IF you see equal amounts in the 0-90 degree period as comopared to the 180-270 period, that means you have comparable number of positive and negative pulses, which as stated above generally means voids within the bulk. If one or the other show significantly more, that would indicate the voids are more on the outer or inner surface of the insulation (see above).
These are the two typical locations for pulses to occur.
The reason appears to be: #1 - A given void will only discharge once per ac cycle: #2 - the discharge occurs very "soon" (in comparison to ac cycle time) after the threshhold is reached.
IF you see equal amounts in the 0-90 degree period as comopared to the 180-270 period, that means you have comparable number of positive and negative pulses, which as stated above generally means voids within the bulk. If one or the other show significantly more, that would indicate the voids are more on the outer or inner surface of the insulation (see above).
electricpete
Electrical
jbartos - I have to question the context (not content) of your last message (8/4/01).
You started with "suggestion to the previous 2 messages". Those two messages were mine and Doge's dated 7/24/01. They dealt exclusively with trying to understand why the pulses occur predominantly when d|V|/dt >0.
This has nothing to do with evaluating dV/dt>0 vs dV/dt<0). Also has nothing to do with a lightning pulse since the test is done under sinusoidal excitation.
My proposed explanation was: These [d|V|/dt>0] are the two typical locations for pulses to occur.
The reason appears to be: #1 - A given void will only discharge once per ac [half] cycle: #2 - the discharge occurs very "soon" (in comparison to ac cycle time) after the threshhold is reached..
If you accept these two statements, then it should be apparent that partial discharge will occur only when |V| is rising.
I believe your statement is definititely worth considering in another context... namely that in evaluating the concern-level which should be afforded to pulses measured at different polarities, we should consider both the polarity and magnitude of the V and dV/dt they might be exposed to during some future surge transient. That's a good point and I don't argue with it. I just don't see that it has anything to do with the two messages of 7/24/01.
You started with "suggestion to the previous 2 messages". Those two messages were mine and Doge's dated 7/24/01. They dealt exclusively with trying to understand why the pulses occur predominantly when d|V|/dt >0.
This has nothing to do with evaluating dV/dt>0 vs dV/dt<0). Also has nothing to do with a lightning pulse since the test is done under sinusoidal excitation.
My proposed explanation was: These [d|V|/dt>0] are the two typical locations for pulses to occur.
The reason appears to be: #1 - A given void will only discharge once per ac [half] cycle: #2 - the discharge occurs very "soon" (in comparison to ac cycle time) after the threshhold is reached..
If you accept these two statements, then it should be apparent that partial discharge will occur only when |V| is rising.
I believe your statement is definititely worth considering in another context... namely that in evaluating the concern-level which should be afforded to pulses measured at different polarities, we should consider both the polarity and magnitude of the V and dV/dt they might be exposed to during some future surge transient. That's a good point and I don't argue with it. I just don't see that it has anything to do with the two messages of 7/24/01.
electricpete
Electrical
jbartos - After I posted and re-read my message I can see now your context was a response to the third paragraph of my 7/24 message.
As stated above, I agree with you that it's worthwhile to consider the characteristics of voltages surges that might occur in the future.
Sorry for a lot of ado about nothing. Sometimes it's tricky to catch a person's meaning in these forums, so I watch the context closely.
As stated above, I agree with you that it's worthwhile to consider the characteristics of voltages surges that might occur in the future.
Sorry for a lot of ado about nothing. Sometimes it's tricky to catch a person's meaning in these forums, so I watch the context closely.
- Thread starter
- #13
A lot of good discussion on this topic. I appreciate all of the responses. Just a follow up note. I had a company come in for three days, to do partial discharge testing. They were able to test 6 sections of switchgear (15KV class), four (4) generators, and about 12 potential transformers. All testing is accomplished without a shutdown. They found some partial discharge activity on one feeder. Suspect the partial discharge is coming from a cable splice about 50 ft from the switchgear. The splice is in a cable tray. I was impressed with the ability of this technology to pinpoint the location of the partial discharge activity. It does take a highly skilled engineer/technician to interpret the data as it is being taken.
We have installed PD sensors on a 34,5kV switchgear, and do some measurements (UPDA Cutler-Hammer), but we need more information about the analisys and diagnostics of the results, we are not sure about levels of PD (hi-med-low). Could you please give some tips or tutorial about it.?
Thanks,
Eduardo O.
Thanks,
Eduardo O.
- Thread starter
- #15
eot
Sorry, I am afraid I can not provide you with much help. I had them (CH) come in for three days of testing. My impression is that it takes a highly skilled engineer or technician to analyze the data. They only found PD activity on one feeder. Looked like he spent a lot of time ruleing out other "Noise" as not being PD.
Sorry, I am afraid I can not provide you with much help. I had them (CH) come in for three days of testing. My impression is that it takes a highly skilled engineer or technician to analyze the data. They only found PD activity on one feeder. Looked like he spent a lot of time ruleing out other "Noise" as not being PD.
jbartos
Electrical
- Jan 15, 2000
- 6,440
Suggestion: Visit
instead of
which appears to be down
and References:
1. Gabe Paoletti, P.E., Alex Golubev, PhD, "Partial Discharge Theory and Applications to Electrical Systems," (Cutler-Hammer Engineering Services), TAPPI Conference in March 1999
instead of
which appears to be down
and References:
1. Gabe Paoletti, P.E., Alex Golubev, PhD, "Partial Discharge Theory and Applications to Electrical Systems," (Cutler-Hammer Engineering Services), TAPPI Conference in March 1999
benlanz
Electrical
- Mar 26, 2004
- 113
I would like to dispel few misconceptions about PD testing in MV cables.
VOIDS
Most of you are talking about voids. We have tested over 10000km of mv cable throughout the world. I can count the number of voids we have found one hand. You see academics love to theorize about them because they are mathematically beautiful, but in real life they are a low probability event. In almost every aged cable PD case we have studied an electrical tree was located near a water tree. In a few new cable installations we have found voids that the factory test missed.
Detectable PD fails rapidly? -false
One of our customers had 151 3-phase feeder cable which they were going to replace because they had all failed at least twice. We performed offline PD tests (up to 2.5Uo) on each of these cable every year for 3 years. All of the cables were left in service. After 3 years ~40% of the cables with PD site had failures. The failures were where we located the PD sites. What about the 60% that didn't fail after 3 years of service? We dissected 12 of these sites. In every case there was large water tree with electrical tree.
You see, if you have a PD system that is sensitive enough you and make accurate field measurements! Just because someone has cancer does mean they are going to die tommorrow, but you'd better do something about it or they will die eventually
Testing at Operating Voltage
3% or less PD sites will show up at operation voltage. According to our data basis about 3% of defects in cable insulation appear at operating voltage. This is because once their is PD activity at the operating voltage stess level the defect in cable insulation (not accessories)will go to rapid failure. You must go to higher voltages to have a meaningful test.
PD systems are so advanced now that utilities have faultl location crews in the field performing PD tests with great success.
VOIDS
Most of you are talking about voids. We have tested over 10000km of mv cable throughout the world. I can count the number of voids we have found one hand. You see academics love to theorize about them because they are mathematically beautiful, but in real life they are a low probability event. In almost every aged cable PD case we have studied an electrical tree was located near a water tree. In a few new cable installations we have found voids that the factory test missed.
Detectable PD fails rapidly? -false
One of our customers had 151 3-phase feeder cable which they were going to replace because they had all failed at least twice. We performed offline PD tests (up to 2.5Uo) on each of these cable every year for 3 years. All of the cables were left in service. After 3 years ~40% of the cables with PD site had failures. The failures were where we located the PD sites. What about the 60% that didn't fail after 3 years of service? We dissected 12 of these sites. In every case there was large water tree with electrical tree.
You see, if you have a PD system that is sensitive enough you and make accurate field measurements! Just because someone has cancer does mean they are going to die tommorrow, but you'd better do something about it or they will die eventually
Testing at Operating Voltage
3% or less PD sites will show up at operation voltage. According to our data basis about 3% of defects in cable insulation appear at operating voltage. This is because once their is PD activity at the operating voltage stess level the defect in cable insulation (not accessories)will go to rapid failure. You must go to higher voltages to have a meaningful test.
PD systems are so advanced now that utilities have faultl location crews in the field performing PD tests with great success.
electricpete
Electrical
benlanz
That' some good info. May I suggest you repost it as a separate item.
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
That' some good info. May I suggest you repost it as a separate item.
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
electricpete
Electrical
My suggestion was to post it as a separate item. To do that go to the bottom of the forum screen
Step 1 - select helpful tip.
Step 2 - type subject
Step 3 - type message
To post as a FAQ got to bottom of FAQ screen
Step 1 - select category
Step 2 - type subject
Step 3 - type message
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
Step 1 - select helpful tip.
Step 2 - type subject
Step 3 - type message
To post as a FAQ got to bottom of FAQ screen
Step 1 - select category
Step 2 - type subject
Step 3 - type message
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
- Status
Similar threads
- Question
- Locked
- Question
- Locked
- Question
- Question