Dissolved Gas Analysis - Past, Present and Future

Fredi Jakob
Weidmann-ACTI Inc.
I. INTRODUCTION
It has been well over thirty years since dissolved gas analysis, DGA, was introduced as
a diagnostic tool for monitoring mineral oil filled transformers. It is now universally accepted as
the method of choice to locate incipient thermal and electrical faults. DGA methodology and
applicability have evolved significantly since its inception. The evolutionary development
includes new laboratory methods, on-line DGA, application to additional types of fluid filled
equipment, application to dielectric fluids other than mineral oil and new diagnostic
interpretation protocols.
At this time the ASTM has approved two procedures for laboratory DGA and a third
method is on the verge of approval. Devices, which periodically or continuously monitor one
or more gas in oil concentrations, are available for on-line analysis. DGA was originally
developed for transformers, but it is now being applied to load tap changers, oil filled circuit
breakers, oil filled bushings and cable oil. Data interpretation methods have been extensively
developed for transformers filled with mineral oil and IEEE is currently developing a guide for
silicone fluid filled transformers.
II. LABORATORY METHODS
1
ASTM procedure, D 3612 , provides two different methods for the extraction of
dissolved gases from dielectric fluid. Method D 3612-A involves an extraction process that
precedes the subsequent gas analysis. The dielectric fluid is delivered into a previously
evacuated glass apparatus and is stirred for a sufficient time to extract most of the dissolved
gases. The volume of the evacuated glassware must be large in comparison to the volume of
dielectric fluid so that most of the dissolved gases leave the fluid phase. The evolved gas is
compressed and the volume and temperature are measured. Then, one or more aliquots of
the gas are separated and quantified with a gas chromatograph. Each gas concentration in
the oil is reported in ppm by volume. ASTM 3612-A is time consuming and labor intensive, but
it has low detection limits and generates reproducible data. Mercury is used in the extraction
apparatus so precautions must be taken to minimize worker exposure to mercury vapors.

ASTM D 3612-B uses direct injection of an oil sample onto a heated gas stripping
column that is in series with a separation column, both of which may be contained within a
single gas chromatograph. The advantages of this method are speed and elimination of
potential mercury health hazards. One disadvantage is that method B cannot be used with
Silicone fluids because excessive foaming will damage the separation columns. Also, since
fault gases are very soluble in mineral oil the extent of gas extraction may not be as complete
as in the two stage method. The ASTM document states, The limit of detection for hydrogen
specified in Method B is higher than that specified for Method A. This could effect the
interpretation of results when low levels of gases are present.1 ASTM D 3612, which contains
a comparison of the detection limits of methods A and B, is summarized in Table 1.

Table 1
1
Minimum Gas Detection Limits, ppm vol
Gases
Method D 3612-A
Method D 3612-B
Hydrogen
5
20
Hydrocarbons
1
1
Carbon Oxides
25
2
Atmospheric Gases
50
500
New DGA methodology that involves a two stage process2 has been developed and is
being reviewed by ASTM committee D-27 (Proposed Designation D 3612-C). The first stage
of the proposed headspace method involves a partitioning of dissolved gases between oil
and an inert gas, argon. Oil samples are introduced into a sealed container that has been
purged with argon. Samples are then heated and vigorously agitated for an extended period
of time. Ostwald partition coefficients are used to correlate gas concentrations in the
headspace of the sample container with the initial gas concentrations in the oil. The second
stage involves analysis of the evolved gases with a gas chromatograph. The advantages of
this headspace method include automation of the extraction and analysis process and
elimination of mercury. Detection limit comparisons between the headspace method and D
3612-A are being developed. One area of concern is potential variability of Ostwald
coefficients with oil type, though most believe the difference of results due to this factor is not
significant.
III. ON-LINE DGA
In-time maintenance of electrical equipment, performing maintenance on only those
units that require it, is a desirable goal. In order to fully implement this objective, equipment
would have to be effectively monitored on an ongoing basis. Since DGA is accepted as an
excellent diagnostic monitoring tool a great deal of effort has been expended to develop online DGA. The two steps required to implement this are automated separation of the fault
gases from the oil followed by quantitative analysis. One or more gases can be extracted from
the oil and subsequently analyzed with a chromatograph, an infra-red spectrometer or a mass
spectrometer. Separation of the gases from the oil can be achieved with a semi-permeable
membrane or a mechanical device. The small mobile hydrogen atoms, and to a lesser extent
other small molecules such as CO, are most readily separated. The hydrogen can then be
measured. Fortunately, hydrogen is formed with any type of fault, partial discharge, heating or
arcing and the measurement of this gas can be used to continuously monitor the condition of a
transformer. Hydrogen detection methods are best used as trigger devices to indicate when a
laboratory DGA is appropriate.
Mechanical extraction devices for separation of gases from oil have also been
developed. Oil is fed into a cylinder fitted with a piston that is moved up and down. On the
down stroke headspace is created in the cylinder and the gases partition between the oil and
the headspace. During the compression stoke the gas is pushed into a reservoir and the cycle
is repeated. This multiple stage extraction procedure is potentially more effective than a single
stage extraction such as that used in ASTM D 3612-A . The extracted gases can then be

separated and analyzed by gas chromatography or quantified without prior separation by infrared or mass spectroscopy .
On-line DGA is theoretically feasible but reliability, detection limits and economic issues
must still be resolved before the method is widely implemented. Detection of a limited number
rather than all of the gases may be a sufficiently viable compromise.
IV. APPLICABILITY OF DGA
A. Fluid Types
Thermal or electrical faults release energy that will result in the partial molecular
destruction of dielectric fluids. The extent of the molecular rearrangement depends on the
available energy and the nature of the dielectric fluid. Faults will produce gases in mineral oils,
high molecular weight hydrocarbons, PCBs, silicone fluids, perchlorethylene and other
dielectric fluids. Since DGA interpretation is empirical in nature it has only been effectively
3
applied to dielectric fluids that have been extensively studied. The current IEEE guide ,
57.104, covers the application of DGA to mineral oils only. High molecular weight
hydrocarbons, also known as less flammable hydrocarbons, contain similar molecules to those
present in conventional mineral oil and produce the same gases under similar fault conditions.
The IEEE guide is applicable, without modification, for these fluids. Silicone fluids produce
the same fault gases as mineral oils when they are thermally or electrically stressed, but
relative concentrations and fault gas concentration ratios are different. A separate IEEE trial4
use guide, P1258, has been developed for these fluids . IEEE guides for the other fluids are
not currently available.
B. Equipment Types
DGA was initially developed to monitor transformers and the success of this method is
well known. Fault conditions can exist in any type of electrical equipment and if this equipment
contains dielectric fluids, fault gases will be produced. Table 2 shows the correlation between
fault types and the various fault gases.

Gases
Hydrogen
Methane, Ethane, Ethylene
Acetylene
Carbon Oxides

Table 2
Fault Gases
Indication
Partial discharge, heating, arcing
Hot Metal gases
Arcing
Cellulose insulation degradation

1. Load Tap Changers

When a load tap changer, LTC, operates arcing occurs and the expected fault gases,
acetylene and hydrogen, are produced. One might initially assume that the presence of these
gases masks the gases produced by other faults. Coking and misalignment of contacts are
the most common problems that occur in LTCs. Coking is a cumulative problem that starts
with an initial deposition on the contact surfaces, which results in increased contact resistance,
followed by additional carbon build up on the contacts. This process leads to exponentially
5
increased heating or thermal runaway and carbon build up. Youngblood was one of the
first investigators to realize that the coking problem would result in the production of the hot
metal gases, methane, ethane and especially ethylene. The concentration of these gases
depends on a number of variables including breathing type, manufacturer, model type, etc.
Generic fault gas threshold values, similar to those in the IEEE guide for transformers, have
been developed by Youngblood and are given in Table 3.
Table 3
5
LTC Monthly Watch Criteria
LTC Type

Hydrogen

Acetylene

Ethylene

Free or Desiccant
Breather

>1500 ppm

>1000 ppm

>1000 ppm

Sealed

>5000 ppm

>9000 ppm

>1200 ppm

Vacuum

>10 ppm

>5 ppm

> 100 ppm

6
Charles Baker and others have developed manufacturer specific flag points and this
approach, which is illustrated in Table 4, is probably the most promising.

Table 4
Equipment Specific Action Levels
(McGraw Edison LTC 550)
Hydrogen

Methane

Ethane

Ethylene

Acetylene

LT1

100

100

100

500

100

100

150

LT2

250

200

200

1200

200

500

300

1000

3000

LT3 500
LT1 = Abnormal

400
400
LT2 = High

2000
400
LT3 = Very High

CO

CO2

A typical LTC case history is documented below:

AC TLH-21 138KV x 12KV 50 MVA Free Breather
Date: February 25, 1993
Date
02/25/93

Mfr.
AC

Serial Number
018226580301

C2H2

CH4

C2H6

C2H4

H2

CO

CO2

34

71

350

This unit was determined to be operating properly. The low concentrations of hydrogen and
acetylene are considered normal for a free breathing unit. The unit was scheduled for annual
testing.
Date: February 25, 1994
Date

Mfr.

02/25/94 AC

Serial Number
018226580301

C2H2
44

CH4
1812

C2H6

C2H4

576

3143

H2

CO CO2

149 33

645

This unit was in thermal runaway when tested. Notice the high level of ethylene, which is the
key gas for overheating. This unit was already heavily coked when the DGA test was
conducted. The unit was removed from service and repaired. The reversing switch and some
moveable dial contacts were replaced.
Date: February 27, 1995

Date
02/27/95

Mfr.
AC

Serial Number
018226580301

C2H2

CH4

C2H6

C2H4

H2

CO

CO2

55

11

22

33

440

The unit is operating normally after completion of the repairs. The LTC was placed on a six
month test interval based on its previous failure history.
2. Oil Filled Bushings
John Stead presented a paper at the 1996 Doble conference7 showing how DGA could
resolve conflicting Power Factor results for two bushings from the same manufacturer. The
DGA data for these two bushings is given in Table 5. This data clearly indicates that partial
discharge is occurring in bushing 2. The low ratio of CO2 / CO indicates severe overheating of
the paper in the bushing. Physical inspection confirmed the interpretation of the DGA results.

Table 5
7
DGA Data for Two Bushings, ppm
Gas
Bushing 1
Bushing 2
Hydrogen
1705
19132
Oxygen
5546
4041
Nitrogen
68216
50767
Carbon Monoxide
441
537
Methane
146
1256
Carbon Dioxide
710
1459
Ethylene
1.8
11
Ethane
71
409
Acetylene
<0.1
0.2
Removal of oil from a bushing is not encouraged by most manufacturers, but the
practice is common in Europe.
3. Oil Filled Circuit Breakers
Application of DGA to oil circuit breakers, OCBs, is under investigation by ACTI in
cooperation with several utilities. Most OCBs are free breathing and fault gas concentrations
could be very dependent on the sampling time. Other tests such as particle size distribution,
particle types and metals in the oil are being evaluated for the identification of problem units.
V. SUPPLEMENTARY TESTS
Carbon oxides are produced by the degradation of cellulose insulation or when oil is
heated in the presence of oxygen. The extent of cellulose degradation is a critical factor in
estimating the condition and life expectancy of solid insulation. Levels of CO and CO2 and the
ratio CO2 / CO can be used to indicate when further investigation of the cellulose condition is
warranted. The two available methods are Degree of Polymerization, DP, and Furanic
compound analysis. DP is more definitive, but this invasive procedure requires removal of
paper samples from the equipment. Furan analysis is a non-invasive procedure and is
recommended if the DGA results indicate cellulose decomposition.
A very interesting example of the effect of transformer overheating was provided to
ACTI by Mr. Charles Baker, South Carolina Electric and Gas Co8. The transformer in question
33/13.8 kV was run for 4 days without operation of the fans and pumps. The main winding
temperature of this three and a half year old transformer reached 1500 C. Rates of cellulose
degradation double for each 6-8 degree rise in temperature. If we assume a normal winding
temperature of 900, the sixty degree rise in temperature would correspond in a 210 (1024)
increase in the rate of cellulose decomposition. Four days at the elevated temperature would
correspond to 4096 days, 11.2 years of normal operation. The transformer insulation would
then be 14.5 years old. The laboratory analysis showed 0.45 ppm of 2-furfuraldehyde which
correspond, using the Chendong equation, to a DP of 530 and an apparent operating time of

25 years. There was no additional data to indicate if the transformer was overheating during
the first three and a half years of operation. Baker has recently reported that the transformer
failed.
VI. FUTURE DEVELOPMENTS
Proposed changes in DGA laboratory procedures have been noted above. IEEE Guide
57.104 was last revised in 1991 and is currently being revised once more. The committee is
proposing a two step process for utilization of DGA laboratory data. The current thinking is that
there should be one set of criteria for interpretation of the first DGA result (proposed Table 1)
and a second set of criteria for subsequent tests (proposed Table 2). Current deliberations
include the concentrations of fault gases that will be classified as normal and levels that lead
to a recommendation to monitor the transformer at shorter intervals. Once it is determined that
one or more gases have exceeded these normal levels then the rates of fault gas generation
should be determined. Table 2 will then tabulate IEEE recommendations for subsequent test
intervals and operational procedures.

REFERENCES

1.

ASTM, Annual Book of ASTM Standards, Volume 10.03, D 3612-96.

2.

Jalbert, J. and Gilbert, R. Comparison Between Headspace and Vacuum Gas

Extraction Techniques for the Gas Chromatographic Determination of Dissolved Gases
from Transformer Insulating Oils", 1994 International Symposium on Electrical
Insulation, Pittsburgh, PA, 1994.

3.

IEEE C57.104-1991, IEEE Guide for the Interpretation of Gases Generated in OilImmersed Transformers, 1991.

4.

IEEE P1258, IEEE Trial-Use Guide for the Interpretation of Gases Generated in
Silicone-Immersed Transformers, 1995.

5.

Youngblood, Rick, et. al, Application of DGA to Detection of Hot Spots in Load Tap
Changers, Minutes of the Sixtieth Annual International Conference of Doble Clients,
1993, Sec. 6-4.

6.

Charles Baker, Private Communication.

7.

Stead, J. and Jakob, F., Use of DGA to Confirm Unsatisfactory Doble Test Resluts of
115 KV Bushings, Minutes of the Sixty-third Annual International Conference of Doble
Clients, 1996, Sec. 3-4.

8.

Charles Baker, Private Communication.