Dissolved Gas Analysis: It Can Save Your

Transformer
M. Duval, Hydro-Quebec (IREQ)
Introduction
Like a blood test or a scanner examina-
T o keep their size acceptable, modern high-voltage
transformers are built with relatively tight insula-
tion tolerances compared to older equipment
and are consequently subjected to increasingly high
stresses in service. They are also quite expensive; the
tion of the human body it can warn
about an impendent problem, give an
early diagnosis, and increase the
cost of a large 765-kV unit easily exceeding $2 million. chances offinding the appropriate cure.
It is therefore very important to closely monitor their
in-service behavior to avoid catastrophic failures, costly
outages and losses of production.
Dissolved gas analysis (DGA) is a very efficient tool A single-stroke vacuum extraction technique is used
for that purpose. Like a blood test or a scanner exami- mostly in North America [ 21. It is based on the following
nation of the human body it can warn about an impen- principle, depicted in Fig. 2. The whole system is first
dent problem, give an early diagnosis, and increase the evacuated, then valve b is closed and the oil sample is
chances of finding the appropriate cure. The operating introduced through a septum and allowed to degas in the
principle is based on the slight albeit harmless deterio- large evacuated volume; valve a is then closed and the
ration of the insulation that accompanies incipient mercury level is raised from 1 to 2 to bring the extracted
faults, in the form of arcs or sparks resulting from di- gases back to atmospheric pressure. Two other tech-
electric breakdown of weak or overstressed parts of the niques are also used, mainly in Europe: the multi-stroke
insulation, or hot spots due to abnormally high current Topler pump, which allows repeated vacuum extractions
densities in conductors. Whatever the cause, these on the same oil sample by using a non-return valve c, and
stresses will result in the chemical breakdown of some the very simple stripping technique, in which the carrier
of the oil or cellulose molecules constituting the di- gas from the GC is bubbled directly through the oil
electric insulation. The main degradation products are sample. International round robin tests have shown that,
gases, which entirely or partially dissolve in the oil properly used, all three extraction methods give fairly
where they are easily detected at the ppm level by comparable results.
DGA analysis. While oil sampling is necessary for complete DGA
Actually, DGA analysis in its present form has been analysis and diagnosis purposes, monitoring gas-in-oil
used extensively for the last 20 years for the routine
monitoring of transformers in service, and new applica-
tions, e.g. monitoring of factory acceptance tests or
other power system components, are now being devel-
oped.
stopcock valves \
Gas-in-Oil Analysis and Monitoring Transformer sampling valve
with 5/32 in. hole in plug
A precise DGA analysis always requires sampling a small
volume of oil from the transformer. Sampling by syringe, Fig. 1. Oil sampling by syringe
as shown in Fig. 1,is probably the most popular technique
although others are available [ 11. Oil samples are usually
taken at the bottom of the tank, from the drain valve, but
- To GC. through
I I loop or syringe
-

[g
To GC
also, for special purposes, at the top, from the radiators,or Carrier gas
the gas relay. The filled syringe is then sent to the
laboratory for analysis. The analyticalprocedures used are
abundantly documented [ 11. They involve removing the
dissolved gases, then injecting them into a dedicated gas Oil
chromatograph (GC). Commercial equipment incorpo-
rating all the necessary columns, detectors, switching stripping Vacuum extraction
valves and gas carrier systems are available from several
manufacturers. Fig. 2. Laboratory extraction of gases from oil
0883-7554/89/0011-0022$01.00 0 1989 IEEE

22 IEEE Electrical Insulation Magazine November/December 1989-Vol. 5, No. 6

Authorized licensed use limited to: INDIAN INSTITUTE OF SCIENCE. Downloaded on January 8, 2010 at 02:02 from IEEE Xplore. Restrictions apply.
 formation in service is more reliable with hydrogen-in-oil
detectors, which are not yet used extensively but most
likely will be so in the near future. These devices are
based on the use of semi-permeable tubes or membranes
permanently installed on the transformers, allowing the
hydrogen in the oil to diffuse into a small gas space where
it is easier to detect and measure. Two main commercial Cnrrier gas
models are available. The first [ 31 allows the hydrogen
concentration to be measured at intervals of a few hours H2
(dissolved m the Oil)
by a portable GC that can be connected to semi-perme-
able tubes (Fig. 3). Continuous remote monitoring from Fig. 3. Hydrogen-in-oil detector 1
the substation is possible with the second model [ 4 ] ,
which uses a fuel-cell type detector (Fig. 4). Systems
based on similar principles have also been proposed [ 51.
Such devices allow the main drawback of oil sampling, i.e.,
development of a fault in the period of time between ectrodes
samplings (which can run from 6 months to 2 years for
routine maintenance, depending on the type of trans-
former) to be considerably reduced, even eliminated in
the case of the continuous-monitoringmodel. When an
abnormal condition is detected by the hydrogen-in-oil
monitor, a conventional oil sampling it still needed for
diagnosis and will be, until reliable monitors of the other
specific dissolved gases become available, particularly for
acetylene and ethylene. Although devices indicating the
presence of acetylene have been proposed recently [61,
they are still unable to measure its concentration.
Fig. 4. Hydrogen-in-oil detector 2. Continuous monitor-
Fault Diagnosis ing
The main gases formed as a result of electrical and
thermal faults in transformers and evaluated by DGA are
H,, CH4, C2H2,CzH4, C,H,, CO, and CO,, whose relative
concentrations depend on the fault type. A general rule, and hydrogen (H,) with the “cold gas plasma of corona
based on thermodynamic considerations, is that the discharges, although mixtures of all gases, including the
degree of chemical unsaturation of the gases formed is other saturated hydrocarbons (CH4, C,H6), are usually
related to the energy density of the fault. Acetylene obtained in most cases of faults. Their relative proportions
(CH = CH) is thus mainly associated with arcing, where have been correlated, through empirical observations and
temperatures reach several thousand degrees, ethylene laboratory simulations, with the various types of fault
(CH, = CH,) with hot spots between 150°C and 1000°C, commonly encountered in transformers in service. The

-~~

a. High-energy arcing (b20 In)

b. Low-energy arcing, tracking
c. Corona discharges
d. Hot spots, T < 2OO0C
e. Hot spots, 200 < T < 400’C
f. Hot spots, T > 4OO0C

Triangle coordinates:

Fig. 5. Duval‘s triangle. Ref. 8, revised 1980.

IEEE Electrical Insulation Magazine NovemberDecember 1989-Vol. 5, No. 6 23

Authorized licensed use limited to: INDIAN INSTITUTE OF SCIENCE. Downloaded on January 8, 2010 at 02:02 from IEEE Xplore. Restrictions apply.
 Ratios of characteristic
t I

IC,H,]
I

Arcing
/ '
I

gases

-CZHZ ~ _cZH4 _
Code of range of ratios cf14 HZ ' . 6

< 0.1 0 1 0
0.1-1 1 0 0

>3
1-3 1
2
2
2
1
2
0' '
10
I
100
1
lo00
I
10000
Gas concentration, ppm
Characteristic fault I I l l
No fault 0 0 0 Fig. 6. Probability of transformer failure in service as a
function of dissolved C,H4 concentration. Influence of
Partial discharges of 0 1 0 fault type.
low energy density
partial discharges of
high energy density

Discharges of low energy type of gas considered, the fault type (arcing or hot spot),
the equipment (power transformers ET, without or with
tap changers communicating with the main tank TC, and
Hot spots T < 150°C 1 sealed instrument transformers IT), the age of the equip-
ment, and whether or not cellulose was involved in the
Hot spots 1 5 0 < T < 300°C 0 2 0 fault. As a general trend, safe levels appear to be lower for
Hotspots 300<T<700"C 0 2 1 arcing (as compared to undiagnosed faults and hot spots),
for sealed ITS and PTs (as compared to PTs and PT-TCs,
Hot spots T<700"C I 0 I 2 I 2 I respectively), in the early and late years of the equipment,
and when cellulose is involved (effect on C,H4 level only,
however). Examples of acceptable and dangerous levels

most widely used tool for that purpose is the IEC-IEEE

ratio method depicted in Table I [ 71. One drawback of this
method in its present form is that a sigtllficant number of 601 I I 1 I
DGA results in service fall outside the proposed codes,
and cannot be diagnosed. Other methods that overcome
this limitation have therefore been developed, such as
that shown in Fig. 5 , based on a triangular representation
and Hydro-Quebec experience [81, or using nomographs
PI.

Acceptable Levels
The gas-in-oil concentration levels found by DGA
analysis may be so small (a few ppm) that the correspond-
ing transformer fault cannot be considered signrfcant or
harmful to equipment. The levels considered as unsafe or
unacceptable in service, however, are still a controversial
issue and are usually determined by norms that arbitrarily
consider the upper 5 or 10 percent of a reference I / I
population as unacceptable or dangerous [ 1 ~ 3 1Re- .
cently, a more factual approach has been proposed, based "0 10 100 1000 10000
on the actual probability of a fault or failure occurring in Gas concentration,ppm
service, as calculated from the large DGA computer data Fig. 7. Probability of transformer Mure in service as a
bank at Hydro-Quebec[ 14-16]. Examples of the probabil- function of dissolved C,H, concentration. Influence of
ities of failure thus obtained are indicated in Figs. 6-9, equipment type. PT:power transformer; TC: tap changer
from which values of acceptable and dangerous levels can communicating with the main tank; IT: instrument
be deduced. They have been shown to depend on the transformer.

24 IEEE Electrical Insulation Magazine November/December 1989-Vol. 5, No. 6

Authorized licensed use limited to: INDIAN INSTITUTE OF SCIENCE. Downloaded on January 8, 2010 at 02:02 from IEEE Xplore. Restrictions apply.
 TABLE II:
ACCEPTABLE LEVELS VS EQUIPMENT AGE
(INSTRUMENT TRANspoRMERs I T), IN PPM
50
Years C A czH4 %
n <2 20 10 200
340- 2-5 40 40 5000

I- 30
>
5-10
10-20
20
40
10
10
100
40
-
500
200
40

j 20 -
TABLE III:
>20
10 - 10-20 ACCEPTABLE LEVELS VS EQUIPMENT TYPE, IN PPM
2-5
5-10
3
I I 1 I I
0: 10 100 two loo00 1OOOOO
Gas concentration (ppm)

Fig. 8. Probability of transformer failute in service as a

IT:Power transformer
function of dissolved C,H, concentration. Influence of IT : Instrument transformer
equipment age. PT:power transformer TC: Tap changex communicating with the main tank

in ppm are given in Tables 11-V. In some cases, even more

specific values may have to be defined, for example,lower
dangerous levels in 765-kV overhead current transform-
ers affected by thermal runaway problems [ 171.
Fault CZHZ -4 HZ
Undqposed 200 500 200
Hot spot 600 600 400
Arcing 100 100 100

Expert Systems
Data such as that given in Fig. 5 and Tables I-VI can be
I 1 I I incorporated into expert systems to facilitate decision
10 100 1000 10000 making by operating and maintenance personnel, even at
the substation level, by screening rapidly the equipment
Gas concentration,ppm
needing special supervision from that operating normally.
Fig. 9. Probability of transformer Mure in service as a A number of expert systems are already available [ 18-20]
function of dissolved C,H4 concentration. Increasing pa- and will undoubtedly mushroom in the future. The quality
per participation in the fault when CO,/CO ratio de- of any expert system,however, will always depend on the
creases. quality of the underlying rules and field experience.

IEEE Electrical Insulation Magazine November/December 1989-Vol. 5, No. 6 25

Authorized licensed use limited to: INDIAN INSTITUTE OF SCIENCE. Downloaded on January 8, 2010 at 02:02 from IEEE Xplore. Restrictions apply.
 TABLEVI: DGA interpretation methods are also being revised by
OVERALL ACCURACY OF DGA RESULTS CIGRE for eventual incorporation into a new IEC docu-
(FROM IEC ROUND ROBIN m) ment [22], and by the IEEE, fkom its current guide [ 131.
Deviation from true Improvements and consensus are sought on diagnoses,
value, in % safe levels and rates of concentration increases.
Sample A Sample B New Applications
Vacuum (single stroke) 13 40 The main application of DGA analysis so far has been
Vacuum (multiple strokes) 23 35 the monitoring of transformers in service. New challenges
stripping 22 27
include monitoring of factory acceptance tests, increas-
Best Laboratory 7 14 ingly used as a quality control technique, of other power
worst Laboratory 39 70 system components such as bushings, tap changers and
oil-filled cables, and of equipment filled with insulating
Worst deviation on one gas 150 400
fluids other than mineral oils, such as silicone oils or
A = medium concentration levels alkylbenzenes.
(hydrocarbons: 9-60ppm CO, CO2 : 100-500 ppm)
B = low concentration levels
(hydrocarbons 1-10 ppm CO, COz : 30-100 ppm)
Acknowledgment
The manuscript was revised by L. Kelley-Regnier.

References
International Activities [ l ] ZEC Publication 567, “Guide for the Sampling of Gases
and of Oil from Oil-Filled Electrical Equipment and for
International cooperation is very important for sharing the Analysis of Free and Dissolved Gases,” 1977.
data and experience and reaching agreement on methods [2] ASTM Method D 3612, “Analysis of Gases Dissolved in
of analysis and interpretation that will yield reliable DGA Electrical Insulating Oil by Gas Chromatography,” 1979.
results and diagnoses. Round robin tests using standard oil [ 31 J.E. Morgan, “Transformer Fault Detection,” “Transfo-
samples [ 2 1] have been performed by the IEC and ASTM Tester” available from Morgan SchafTer Corp., Montreal,
to evaluate the performance of the various gas extraction Canada, US Patent 4,112,737, 1978.
and analytical methods used and eventually recommend [4] G. Belanger, M. Duval, “Monitor for Hydrogen Dissolved
in Transformer Oil,” “Hydran” available from Syprotech
improved standard procedures. If agreement is to be
Inc., Pointe-Claire, Canada. ZEEE Trans. Electr ZmuL,
reached between utilities and manufacturers, especially Vol. El-12, NO. 5, pp. 335-340, 1977.
on factory acceptance tests where gas concentration [5] H. Tsukioka, IC Sugawara, E. Mori, S. Hukumori, S. Sakai,
levels are much lower ( 1-10 ppm) than in service, “New Apparatus for Detecting H,, CO and CH, Dis-
laboratories will have to meet higher precision and solved in Transformer Oil,” Z E E E Trans. Electr Znsul.,
accuracy requirements (Fig. 10). Examples of accuracies Vol. EI-18, NO. 4, pp. 409-419, 1983.
obtained during the IEC round robin tests are given in [6] E. Juelke, “Method and Device for Continuous Determi-
Table VI. nation of Acetylene Produced by Glow Discharge,” Swiss
Patent CH 677740 A5, 1988.
[7] R.R. Roger, “IEEE and IEC Codes to Interpret Incipient
Faults in Transformers, Using Gas-In-Oil Analysis,” Z E E E
Tram. Elect. Znsul., Vol. EI-13, No. 5, pp. 348-354,
1978.
[8] M. Duval, “Fault Gases Formed in Oil-Filled Breathing
EHV Power Transformers. The Interpretation of Gas
Analysis Data,” ZEEE-PES Con$ Paper C 74-476-8,1974.
[9] J.O. Church et al. “Analyze Incipient Faults with Dis-
solved-Gas Nomograph,” Electrical World pp. 40-44,
Oct. 1987.
101 E. Dornenburg, W. Strittmatter, “Monitoring Oil-Cooled
Transformers by Gas Analysis,” Brown Bourn‘ Review,
Vol. 61, No. 5, 1970, p. 238.
111 R. Wilputte, M. Random, “Enseignements des Contrdles
Programmes des Huiles Isolantes des Transformateurs de
Puissance du Reseau Belge,” CZGRE Paper 12-07, 1986.
121 T. Kawamura et al, “Dissolved Gas Analysis. Its Use for
the Maintenance of Transformers”, CZGRE Paper 12-05,
Fig. 10. Precision and accuracy. A: precise and accurate; 1986.
B: precise but not accurate; C: not precise and not accu- [ 131 ANSZ-ZEEE C 57, 104, “American National Standard Guide
rate. for the Detection and Determination of Generated Gases

26 IEEE Electrical Insulation Magazine November/December 1989-Vol. 5, No. 6

__ ~ .-

Authorized licensed use limited to: INDIAN INSTITUTE OF SCIENCE. Downloaded on January 8, 2010 at 02:02 from IEEE Xplore. Restrictions apply.
 in Oil-Immersed Transformers and Their Relation to the Transformer Fault Diagnosis,” Artificial Intelligence Ap-
Serviceability of the Equipment,” 1978. plications in Chemistry, American Chemical Society Ed.,
[ 141 M. Duval, F. Langdeau, P. Gervais, G. Belanger, “Interpre- pp. 25-30, 1986.
tation of Dissolved Gas-in-Oil Levels in Power Trans- [ 191 N. Recrosio, Y. Jegou, M. Carballeira, “Dyonysos: Aide au
formers,” Minutes of the 55th Annual International Con- Diagnostic de Defauts Dans les Transformateurs,” SEE
ference of Doble Clients, Sec. 10-7.1110-7.10, 1988. Club 1l 4 t u d e s Gh&rales--Diagnostic et Maintenance
[ 151 M. Duval, F. Langdeau, P. Gervais, G. Belanger, “Accept- Par Systdme Expert Paris, Oct. 8, 1987.
able Dissolved Gas-in-oil Concentration Levels Versus [20] M.A. Marin, J.L. Jasmin, “Cooperating Expert Systems for
Age in Power and Instrument Transformers,”Minutes of Diagnoses of Electrical Apparatus,” Con$ on Expert Sys-
the 56th Annual International Conference of Doble Cli- tems Applications for the Electric Power Industry, EPRI,
ents, Sec. 10-3.1110-3.9,1989. Orlando, Florida, June 5-8, 1989.
[ 161 M. Duval, F. Langdeau, P. Gervais, G. Belanger, ‘‘Muence [21] M, Duval, Y. Giguere, “Preparation of Standard Samples
of Paper Insulation on Acceptable Gas-in-Oil Levels in of Dissolved Gases in Insulating Oils,” Minutes of the
Transformers,” Annual Report of the CELDe 1989. 51st Annual International Conference of Doble Clients,
[17] M. Duval, J:P. Crine, “Dielectric Behavior and Stabiliza- Sec. lOC-Ol/lOC-O7, 1984.
tion of Insulating Oils in EHV Current Transformers,” [22] IEC Publication 599, “Interpretation of the Analysis of
IEEE Tram Elect. InsuL, Vol. EI-20, No. 2, pp. 437441, the Gases in Transformers and Other Oil-Filled Electical
1985. Equipment in Service,” 1978.
[18] C.E. Riese, J.D. Stuart, “TOGA-An Expert System for

Circle Reader Service Number 1 0

IEEE Electrical Insulation Magazine November/December 1989-Vol. 5, No. 6 27

Authorized licensed use limited to: INDIAN INSTITUTE OF SCIENCE. Downloaded on January 8, 2010 at 02:02 from IEEE Xplore. Restrictions apply.