Identifying and Analyzing Quick Developing Faults With DGA: Michel Duval Hydro Quebec Ireq Research Scientist

Identifying and Analyzing Quick
Developing Faults with DGA
Sponsored by:
Michel Duval
Hydro Quebec Ireq
Research Scientist
Michel Duval, creator of the Duval Triangle
and internationally renowned leader in the
field of dissolved gas analysis (DGA) for
transformer condition monitoring will offer
new insights into the use of DGA.
Catastrophic Failures
Material of Serveron DO NOT REPRODUCE WITHOUT PERMISSION 3

Failures in Service
-The failure rate of power transformers in service (internal

failures needing repairs) typically is 0.3% per year.
-For a population of 2000 transformers, this means 6

transformers will fail in the next year.
-Less than 1 will fail catastrophically.
-1994 will not fail.
-200 (i.e., 10% of the population at or above IEEE/IEC

condition 1) will form abnormal amounts of gases because
of faults.
Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION 4
The Monitoring Dilemma
- Nobody knows which 6 of the 2000 transformers will fail

next year.
- To identify them, all the transformers need to be monitored,

including the 1800 operating normally, just for the purpose of
detecting the 6 that will fail and need repairs and the less
than 1 that may eventually fail catastrophically.
- In economic terms, the cost of monitoring is justified as long

as it does not exceed the cost of not detecting the 6 failures
and the catastrophic one (typically, >20M$).

Monitoring Tools
- General tools for monitoring oil temperature, pressure,

partial discharges, etc are available, e.g., from Qualitrol.
- However, for the early detection of faults and failures, the

main monitoring tool is dissolved gas analysis (DGA).
- More than 1 million DGA analyses are performed by ~600

laboratories and ~ 40,000 on-line gas monitors each year
worldwide.

DGA Monitoring Techniques
DGA Monitoring Off-Line:

-also called manual DGA or laboratory DGA
-consists in taking oil samples from transformers and

sending them to the laboratory for DGA analysis
-”normal” sampling frequency is typically one year -

every month or week in case of abnormal gassing.
7 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

DGA Monitoring Techniques
DGA Monitoring with Portable Monitors:

-also requires oil sampling, but DGA results are available
more frequently, e.g., every day.
DGA Monitoring with On-Line Monitors:

-does no require oil sampling
-provides a DGA analysis every 1 or 4 hours.

Advantages and Limitations of
Laboratory DGA
-less expensive than on-line monitoring

-uses IEC/ASTM standardized techniques
-data comparable to those in existing DGA databases.
-will miss faults occurring between two oil samplings

-some laboratories are not accurate and reliable because of
sampling and laboratory errors (“bad” laboratories).

Portable Monitors
-a bit less expensive than laboratory DGA

-allow slightly more frequent DGA analysis (daily vs. weekly)
-will also miss faults occurring between two oil samplings

-are less accurate for some or all gases than “good”
laboratories because of calibration errors

On-Line Gas Monitors
-will detect fast-, hourly-developing faults

-not affected by sampling errors
-more reliable for evaluating rates of change.
-more expensive than laboratory DGA

-some on-line monitors are less accurate than “good”
laboratories for some gases
-some on-line monitors are not calibration-free and
maintenance-free as claimed by their manufacturers.

Examples of On-Line Gas Monitors

Basic Principles of Gas Monitors
-based on headspace principle for the extraction of gases

from oil (partition of dissolved gases between oil and gas
phase)
-partition coefficients must be known exactly at all

temperatures of extraction
-extracted gases are analysed by different types of

detectors (GC, IR, etc)
-monitors available in 2008 have been tested in CIGRE TB

# 409, those available since will be tested by CIGRE WG47
Types of Gas Monitors
-”multi-gas” monitors will detect from 3 to 9 different gases

(H2, CH4, C2H4, C2H6, C2H2, CO, CO2, O2, N2) depending on
the type of monitor.
-”hydrogen” monitors will detect mostly H2, sometimes CO.

Multi-Gas Monitors
Monitors of the chromatographic type:
-after gas extraction, will separate individual gases on a GC

column, then measure them with GC detectors.
-TM8, TM3 (Serveron)

-Calisto 9, portable Myrkos (Morgan Schaffer)

Monitors of the Chromatographic-Type:
-use the same standardized, NIST-traceable techniques

as laboratories (except Myrkos).
-provide automatic recalibration at fixed intervals as

laboratories do.
-require some maintenance (change of carrier gas,

calibration gas mixture, GC columns every 3 to 5 years),
but this is easy and relatively inexpensive.

Monitors of the Infrared-Type:
-after gas extraction, will measure directly individual gases with

an infrared detector, and H2 with a solid state sensor.
-Transfix 8, portable Transport-X 7 (GE-Kelman) use a photo-

acoustic (PAS) detector.
-LumaSense 9 uses a non-dispersive IR detector.
-MTE Hydrocals (4 to 9) also use an IR detector.

Monitors of the Infrared Type:
-do not require change of carrier gas and gas mixture.
-however, cannot measure H2 , O2 by infrared, requiring

the use of relatively inaccurate solid state sensors for that
purpose.
-some may need recalibration, typically every year,

because of contaminants in ambient air (SF6, oil vapours,
solvents) and lamp fade with time. This recalibration
cannot be done in the field and is lengthy and expensive.
-several do not meet accuracy requirements.

Hydrogen Monitors
-Hydran (GE), Hydrocal 2 (MTE) measure 100% of H2 + 18%
of CO present in oil with a membrane and fuel cell detector.
-Calisto 2 (Morgan Shaffer): measures H2 only with a

membrane, GC and TCD detector.
-Serveron, Qualitrol, Weidmann: measure H2 only with an

H2Scan Pd solid state sensor covered with an inorganic
coating (no membrane).
-TM1 (Serveron): “improved” version of H2Scan with patent

applications for temperature control and oil circulation.

The Importance of DGA Accuracy
-accurate concentration values from both laboratories and
gas monitors (15% accurate or better) are needed for
reliable DGA diagnosis, and for comparison with
concentration limits.
-an accuracy of 15% means that if 100 ppm is measured, the

actual value may be anywhere between 85 and 115 ppm.
-low concentration values (< 5 or 10 times the analytical

detection limit of the laboratory or gas monitor) are usually
quite inaccurate and unreliable and should not be used for
DGA diagnosis.
20
Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION
Accuracy of Gas Monitors
-the accuracy of gas monitors measured by manufacturers in
the factory and reported in 2008 was between ± 5% and
10% depending on the type of monitor.
-actual inaccuracies measured in transformers by CIGRE

and reported in TB 409 (2010) were significantly higher
(between ± 8% and 35% on average at routine levels, up to
50% for some gases and monitors, and up to 100% for H2
with some IR monitors).
-an accuracy of ± 20% was reported in 2014 for Hydrocals in

the factory.

Accuracy of Gas Monitors
-in case of large differences between monitor readings and

laboratory results, which lead to significantly different DGA
diagnosis (e.g., arcing or thermal problem) and different
actions on the equipment, it is recommended by the IEC and
CIGRE to verify the accuracy of both monitor readings and
laboratory results, using the procedure described in
Appendix B of CIGRE TB 409 (2010)
22
Slow- and Fast-Developing Faults
-slow-developing faults occur in ~90% of cases over periods
of months or years. Laboratory DGA and portable monitors
are adequate to detect such faults.
-fast-developing faults occur in 1 to 2% of cases over periods

of days or hours and may result into failures. On-line
monitors can detect most of them, although some cannot.
-catastrophic faults occur in less than 0.2% of cases, within

minutes or seconds, and can be detected by Buccholz relay
and sudden pressure valve. Those occurring within fractions
of a second cannot be detected by any device.
23
Slow- and Fast-Developing Faults
-Examples of gassing rates corresponding to slow- and fast-

developing faults (conditions 1 and 4):
(CIGRE Technical Brochure # 443, 2010)

DGA Diagnosis On-Line with
Multi-Gas Monitors
- DGA diagnosis methods available on-line in multi-gas

monitors include the Key Gas, Rogers and IEC methods.
- Also the Duval Triangles 1, 4 and 5, allowing to identify

the most dangerous faults (arcing D1/D2 and high
temperature hot spots T3/T2, involving carbonization of
paper C).
- And the faults of lesser concern (hot spots T3/T2 in oil

only, overheating O < 250ºC, stray gassing of oil S <
200ºC and corona PDs).
Duval Triangles 1, 4 and 5
Triangle 1
Triangle 4 Triangle 5

Detection of Quick-Developing Faults with a Multi-
Gas Monitor in a 3-Phase GSU Transformer
Day 2 – 16:00
Day 3 – 12:00
Day 2 – 12:00
Day 3 – 04:00
Day 3 – 00:00
Day 2 – 20:00
Day 3 – 16:00
Day 3 – 08:00
Day 23 – 04:00 to
Day 24 – 08:00
Followed by
transformer failure
C2H2 = 800 ppm/day!

700 MVA Transformer
C2H2 = 45 ppm/day!
28
Material of Serveron DO NOT REPRODUCE WITHOUT PERMISSION
336 MVA Transformer
(Placed in Service -1969)
C2H4 = 300 ppm/day!

1100 MVA Transformer
C2H4 = 300 ppm/day!

Reviewed Transformer Failures
-Gassing rates were all significantly above condition 4

values.
-The corresponding transformers were removed from
service 1 to 3 days after looking at monitor readings, before
potential catastrophic failure.
-However, it would have been better to remove them from
service earlier.
-Without an on-line monitor, these transformers would likely
have suffered unplanned severe damage.

Condition 1 Limits for Multi-Gas Monitors
Ref: CIGRE TB 443 (2010)

and WG47 (2014)

Condition 4 Limits for Multi-Gas Monitors
Ref: CIGRE TB 443 (2010)

and WG47 (2014)

On-Line Monitoring with Multi-Gas Monitors
- Multi-gas monitors will detect all types of faults, even in

their early stages at condition 1, and without false alarms.
However, they are more expensive than hydrogen only
monitors.
- The recommendation of CIGRE (TB # 409, 2010) is

therefore to use multi-gas monitors in critical transformers
(GSU, nuclear) and in abnormally gassing transformers.

Fault Detection with Hydrogen Monitors
Note: for faults T3 in paper (C), curve for H2 is a bit higher.

Ref: Duval, TechCon North America 2014.

On-Line Monitoring with Hydrogen Monitors
-hydrogen monitors are sensitive mostly to faults S and corona
PD of lesser concern for the equipment
-hydrogen is much less sensitive than hydrocarbons gases to

detect thermal faults.
-hydrogen monitors are not sensitive enough to detect arcing

faults in their early stages because of the very low acceptable
limit of C2H2

-a compromise limit of 40 ppm of H2 will not detect arcing faults D in

condition 1, and will generate false alarms in case of faults S and PD
-a limit of 2 ppm/ month cannot be detected in practice by H 2 monitors.

-a more realistic limit of 10 ppm/ month will not detect arcing and
thermal faults in condition 1.

-a limit of 40 ppm of H2 will detect all faults in condition 4 but will

generate many false alarms in case of faults S and PD.
-a limit of 10 ppm/ month may still not detect dangerous arcing in

paper in condition 4.

-because of these limitations, the recommendation of CIGRE

(in TB # 409, 2010) is to use hydrogen monitors in non-
critical transmission and distribution transformers, and in
transformers with no previous gassing history.

Identifying and Analyzing Quick
Developing Faults with DGA
QUESTIONS?
Sponsored by:
40

Identifying and Analyzing Quick Developing Faults With DGA: Michel Duval Hydro Quebec Ireq Research Scientist

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Identifying and Analyzing Quick Developing Faults With DGA: Michel Duval Hydro Quebec Ireq Research Scientist

Uploaded by

Copyright:

Available Formats

Identifying and Analyzing Quick

Developing Faults with DGA

Material of Serveron DO NOT REPRODUCE WITHOUT PERMISSION 3

-The failure rate of power transformers in service (internal

-For a population of 2000 transformers, this means 6

-Less than 1 will fail catastrophically.

-1994 will not fail.

-200 (i.e., 10% of the population at or above IEEE/IEC

- Nobody knows which 6 of the 2000 transformers will fail

- To identify them, all the transformers need to be monitored,

- In economic terms, the cost of monitoring is justified as long

Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION 5

- General tools for monitoring oil temperature, pressure,

- However, for the early detection of faults and failures, the

- More than 1 million DGA analyses are performed by ~600

Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION 6

DGA Monitoring Off-Line:

-consists in taking oil samples from transformers and

-”normal” sampling frequency is typically one year -

7 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

DGA Monitoring with Portable Monitors:

DGA Monitoring with On-Line Monitors:

8 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

-less expensive than on-line monitoring

-will miss faults occurring between two oil samplings

9 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

-a bit less expensive than laboratory DGA

-will also miss faults occurring between two oil samplings

10 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

-will detect fast-, hourly-developing faults

-more expensive than laboratory DGA

11 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

12 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

-based on headspace principle for the extraction of gases

-partition coefficients must be known exactly at all

-extracted gases are analysed by different types of

-monitors available in 2008 have been tested in CIGRE TB

-”multi-gas” monitors will detect from 3 to 9 different gases

-”hydrogen” monitors will detect mostly H2, sometimes CO.

14 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

-after gas extraction, will separate individual gases on a GC

-TM8, TM3 (Serveron)

15 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

-use the same standardized, NIST-traceable techniques

-provide automatic recalibration at fixed intervals as

-require some maintenance (change of carrier gas,

16 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

-after gas extraction, will measure directly individual gases with

-Transfix 8, portable Transport-X 7 (GE-Kelman) use a photo-

17 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

-do not require change of carrier gas and gas mixture.

-however, cannot measure H2 , O2 by infrared, requiring

-some may need recalibration, typically every year,

-several do not meet accuracy requirements.

-Calisto 2 (Morgan Shaffer): measures H2 only with a

-Serveron, Qualitrol, Weidmann: measure H2 only with an

-TM1 (Serveron): “improved” version of H2Scan with patent

19 Material of Dr.Duval DO NOT REPRODUCE WITHOUT PERMISSION

-an accuracy of 15% means that if 100 ppm is measured, the

-low concentration values (< 5 or 10 times the analytical

-actual inaccuracies measured in transformers by CIGRE

-an accuracy of ± 20% was reported in 2014 for Hydrocals in