DGA analysis

Nurul Azmi bin. Nordin

[Company address]
Contents
Formation of Gases in Transformer Oil................................................................................................... 2
Types of Faults ........................................................................................................................................ 3
Fault Identification by Gas Type ............................................................................................................. 3
DGA Diagnostic Tools .............................................................................................................................. 4
DGA Diagnostic Tool Selection ................................................................................................................ 4
Key Gas Method ...................................................................................................................................... 5
Doernenburg Ratio Method (DRM) ........................................................................................................ 6
Duval Triangle Method (DTM) ................................................................................................................ 8
Diagnostic Methods Summary .............................................................................................................. 10
Conclusion ............................................................................................................................................. 12
The Future ............................................................................................................................................. 13
Formation of Gases in Transformer Oil
Thermal and electrical stresses that occur within normal operating transformers generate
hydrocarbon gases. Which can indicate potential problems within the transformer. Some gas
generation is expected as transformers age, so it is important to separate normal gassing rates from
excessive gassing rates. Since normal gas generation varies with transformer design, loading, and the
type of insulating material used, general gassing rates are used for all transformers to define abnormal
behavior.
Typical gases that appear in transformers are hydrogen (H ), methane (CH ), ethane (C H ), ethylene
2 4 2 6

(C H ), and acetylene (C H ). These gases begin to form at specific temperatures and dissolve within
2 4 2 2

the insulation oil of a power transformer, as shown in Figure 1. The types and quantities of the gases
that form will depend on the nature and intensity of the fault.

Figure 1

Hydrogen and methane begin to form in small amounts around 150 °C. The production of hydrogen
continues to increase as the temperature increases. At about 250 °C, the production of ethane starts.
At about 350 °C, the production of ethylene starts. After reaching their maximum points, methane,
ethane, and ethylene production goes down as the temperature increases.
Acetylene production begins between 500 and 700 °C. In the past, the presence of only trace amounts
of acetylene was considered to indicate a temperature of at least 700 °C had occurred, however, recent
discoveries have led to the conclusion that a thermal fault (hot spot) of 500 °C can produce trace
amounts (a few ppm). Larger amounts of acetylene can only be produced above 700 °C by internal
arcing.
Between 200 and 300 °C, the production of methane exceeds that of hydrogen. Starting at about 275
°C and up, the production of ethane exceeds that of methane. At about 450 °C, hydrogen production
exceeds all other gases until about 750 to 800 °C, at which point a larger amount of acetylene is
produced.
Thermal decomposition of cellulose materials will also begin at about 100 °C or less and produce
carbon monoxide (CO), carbon dioxide (CO ), hydrogen (H ), methane (CH ), and oxygen (O ).
2 2 4 2

Therefore, it is imperative to operate transformers below 90 °C.

Types of Faults
A proper fault diagnosis will include warnings of any gases with concentrations, increments, rates of
change, or ratios that exceed the standard limits, along with short interpretive remarks and
recommendations based on the findings. To help identify the different faults when a diagnosis is being
made, the following classes have been identified after the physical inspection of hundreds of faulty
transformers detectable by visual inspections and DGA results. They are also summarized in Figure 2.
Partial Discharge (PD) – the corona type occurs, which can result in deposits of ‘X-Wax’ on paper
insulation, or the sparking type occurs, which can induce pinholes (carbonized punctures) in paper
that may be difficult to find.
Discharges of Low Energy (D1) – occurs in oil and/or paper, as indicated by large carbonized
punctures in paper (pinholes), carbonization of the paper surface (tracking), or carbon particles in oil
(as in an LTC).
Discharges of High Energy (D2) – occurs in oil and/or paper, as indicated by extensive destruction
and carbonization of paper or metal fusion at the discharge extremes, extensive carbonization in oil,
and in some cases, tripping of the equipment confirming a large current follow-through.
Thermal Fault (T1) – occurs in oil and/or paper below 300 °C, turning the paper “brownish”.
Thermal Fault (T2) – occurs in oil and/or paper above 300 °C and below 700 °C, carbonizing the
paper.
Thermal Fault (T3) – occurs in oil and/or paper above 700 °C with strong evidence of carbonization
of the oil, metal coloration (at 800 °C), or metal fusion (below 1,000 °C).
Figure 2

Fault Identification by Gas Type

Potential faults such as over-heating, partial discharge, and sustained arcing produce a range of gases.
The concentrations and composition of these gases can be used to identify the type and estimate the
severity of the fault, as shown in Figure 3. Since all types of faults create a variety of gases and not just
one, the diagnostic approaches that detect multiple gases and account for the entire gassing picture
often get more accurate results.

Figure 3
DGA Diagnostic Tools
Most of the DGA diagnostic tools in use today can be found in the IEEE C57.104 or IEC 60599 guides.
Based on these two guides, other national and international guides that include additional tools are
also available, but for our purposes we will only discuss those tools found in the IEEE and IEC guides.
Figure 4 summarizes the diagnostic tools that we will be discussing, as found in the latest IEEE guide,
a recent draft standard (IEEE C57.104- D11d; not approved) of the IEEE guide, and the IEC guide.
Figure 4

DGA Diagnostic Tool Selection

There is a wide range of diagnostic tools available for DGA. Some are simpler, using only sums or single
ratios of gases, alongside a guideline, to determine different warning levels. Others are more complex,
taking multiple gas ratios and fitting them to a precise range of values.
Total Combustible Gas (TCG) procedure and Total Dissolved Combustible Gas (TDCG) procedure are
the two diagnostic tools that are non-ratio based. TCG deals with the gases in the gas headspace and
TDCG the gases dissolved in oil. They can offer an indication that gas levels are increasing and provide
suggested guidelines for operators to perform manual DGA sampling intervals. However, since they
do not offer any significant diagnostic value regarding fault type, it is recommended by the IEEE guide
to combined them with other diagnostic tools in order to get a better understanding of what is
happening in the transformer.
There are three more diagnostic tools that may also be used as a complement to other diagnosis
methods for a more accurate assessment of transformer conditions. These are the CO /CO ratio, the
2

O /N ratio and the C H /H ratio, as described in further detail below.

2 2 2 2 2
CO /CO Ratio – this popular ratio is used to detect paper involvement in a fault. If the ratio is below
2

3, it is a strong indication of a fault in paper, either a hot spot or electrical arcing with a temperature
above 200 °C. If the ratio is above 10, it indicates a fault with a temperature below 150 °C. However,
this ratio is not very accurate because it is affected by the CO and CO coming from oil oxidation and
2

normal cellulose aging, so with a high quantity of CO , seeing a significant change in the CO /CO ratio
2 2

is nearly impossible.
O /N Ratio – a decrease of this ratio indicates excessive heating.
2 2

C H /H Ratio – a ratio between 2 and 3 in the main tank indicates contamination by the LTC
2 2 2

compartment. In these situations, the level of acetylene in the main tank can be quite high, so in order
to diagnose true main tank problems, incremental changes in acetylene must be monitored.
The remaining DGA diagnostic tools that we will discuss are the main interpretation methods used for
fault diagnosis of power transformers. This includes the Key Gas, Dornenburg Ratio, Rogers Ratio, IEC
Basic Gas Ratio, Duval Triangle, and CIGRE ratio methods. The majority of these methods are ratio-
based, meaning they use a subset of the ratios below to diagnose a fault type based on the fit of each
ratio result to a specific range of values:

 Ratio 1 (R1) = CH /H
4 2

 Ratio 2 (R2) = C H /C H
2 2 2 4

 Ratio 3 (R3) = C H /CH

2 2 4

 Ratio 4 (R4) = C H /C H
2 6 2 2

 Ratio 5 (R5) = C H /C H
2 4 2 6

It is important to remember that when using ratio-based diagnostic tools, minimum gas levels are
required as defined in the guides, for the ratio analysis to be considered valid.

Key Gas Method

The Key Gas method is based on the quantity of fault gases that are released from the insulating oil as
the chemical structure breaks at varying temperatures in the transformer. This method uses
individual gas levels, or ‘key gases,’ for detecting faults. Figure 5 summarizes the key gases and their
fault indications.
Figure 5

This method offers diagnostics by calculating the relative proportions (in percent) of the key gases
within the transformer. Figure 6 indicates these key gases and their relative proportions to indicate
the four general types of faults.
 Figure 6

Doernenburg Ratio Method (DRM)

The Doernenburg method can be found in the IEEE C57.104-1991 guide. It has fallen out of favor in
some parts of the world due to its complexity and the evolution of it into the Rogers Ratio and Basic
Gas Ratios approaches, but compared to other diagnostic methods, the DRM still holds its value as one
of the more effective diagnostic tools available.
In order to use the DRM, the concentration of one of the key gases (H , C H , C H , C H , or CH ) needs to
2 2 2 2 4 2 6 4

be at least double the relevant L1 concentrations, as shown in Figure 7. When this criterion is met,
there are four possible ratios that can be calculated if they contain the key gas of concern. Figure 8
shows the proposed fault diagnostics is based on the ranges of the four ratios. The IEEE standard (IEEE
C57.104-1991) also has an illustrated step-by-step application of this ratio method for gases extracted
from the transformer oil only.
Figure 7
 Figure 8

Rogers Ratio Method (RRM)

The Rogers Ratio method evolved from the Doernenburg method and is used exactly the same way,
but instead of needing significant concentrations of the key gases, the RRM can be used when the
concentrations exceed the values listed in Figure 7 (rather than double).
Values for the three gas ratios, corresponding to suggested diagnostic cases, are shown in Figure 9.
The fault types (cases) that are provided have been chosen by combining some cases from the number
of fault types originally suggested by Rogers.
Figure 9

Despite having better accuracy, the Doernenburg Ratio, Rogers Ratio, and Basic Gas Ratio approaches
have one drawback where some combinations of gases do not fit into the specified range of values
when calculated and the fault type cannot be determined. To further demonstrate this, Figure 10
shows a three-dimensional view of the Basic Gas Ratio (IEC 60599-2007-05).
Figure 10
Duval Triangle Method (DTM)
The Duval Triangle method was developed from IEC TC10 databases and an existing IEC 60599 Ratio
method. Approximately 200 plus inspected fault cases in service were used to develop the Triangle.
Within the triangle there are six (6) potential fault zones covering partial discharges, electrical faults
(arcing high and low energy), and thermal faults (over various temperature ranges), plus a DT zone
(mixture of thermal and electrical faults).
The use of the DTM is based upon three key gases (CH , C H , and C H ) that correspond to the increasing
4 2 4 2 2

energy levels of gas formation, as shown in Figure 11. These gas concentrations are calculated and
then plotted along the three sides of a triangle diagram using the following ratios:

 %CH = (CH /CH +C H ) x 100

4 4 4 2 2

 %C H = (C H /CH +C H +C H ) x 100
2 4 2 4 4 2 4 2 2

 %C H = (C H /CH +C H +C H ) x 100
2 2 2 2 4 2 4 2 2

Figure 11

As previously stated, one drawback of the gas ratio methods is that some results can fall outside the
codes and no diagnostics can be given (unresolved diagnostics). This does not occur with the Duval
Triangle method because it is a ‘closed system.’ It always provides a diagnosis, with a low percentage
of wrong diagnoses. In fact, according to some published reports, the DTM provides more accurate
and consistent diagnoses than any other method available currently.
The Triangle method is indicated in Figure 11. In addition to the 6 zones of individual faults
mentioned in Table 2 (PD, D1, D2, T1, T2 or T3), an intermediate zone DT has been attributed to
mixtures of electrical and thermal faults in the transformer.
2H2 and C2H4 are used in all interpretation methods to represent high energy faults (such as arcs)
and high temperature faults. H2 is preferred in several of these methods to represent very low
energy faults such as PDs, where it is produced in large quantities. CH4, however, is also
representative of such faults and always formed in addition to H2 in these faults, in smaller but still
large enough amounts to be quantified. CH4 has been chosen for the
6
Triangle because it not only allows to identify these faults, but provides better overall diagnosis
results for all the other types of faults than when using H2.
This good performance of the Triangle with CH4 might be related to the fact that H2 diffuses much
more rapidly than the hydrocarbon gases from the oil through gaskets and even metal welds.
Therefore, gas ratios using H2 are probably more affected by the loss of this gas than those using
hydrocarbons gases only, which have much lower and comparable diffusion rates.
The three sides of the Triangle are expressed in triangular coordinates (X,Y,Z) representing the
relative proportions of CH4, C2H4 and C2H2, from 0% to 100% for each gas.
To display a DGA result in the Triangle, one must start with the concentrations of the three gases,
(CH4) = A, (C2H4) = B and (C2H2) = C, in ppm. First calculate the sum of these three values: (CH4 +
C2H4 + C2H2) = S, in ppm, then calculate the relative proportion of the three gases, in %: X = % CH4
= 100 (A/S), Y = % C2H4 = 100 (B/S), Z = % C2H2 = 100 (C/S). X, Y and Z are necessarily between 0
and 100%, and (X + Y + Z) should always = 100 %. Plotting X, Y and Z in the Triangle provides only
one point in the Triangle.
For example, if the DGA results are A = B = C = 100 ppm, X = Y = Z = 33.3%, which corresponds to
only one point in the centre of the Triangle, as indicated in Figure 11a.
Figure 11a

The zone in which the (X,Y,Z) point falls in the Triangle in Figure 1 allows to identify the fault
responsible for the DGA results. The example of Figure 11a would indicate a fault. The X, Y and Z
values can as well easily be calculated manually.
Faults in paper vs. faults in oil
Faults in paper are generally considered as more serious than faults in oil, because paper is often
located in areas of high electric field (in the windings, or as voltage barriers), and the destruction of
paper insulation may lead to short circuits or severe arcing.
Faults in paper, fortunately, are much less frequent than faults in oil (typically, in 10 % of cases
only), however, because of the more serious consequences, their detection by DGA or other means is
of great interest.
A popular way of detecting faults in paper by DGA is by looking at the CO2 to CO ratio. Values < 3 are
a good indication of faults in paper of a temperature > 200 to 300 °C (including arcing), where paper
degrades very rapidly or even carbonizes. However, there is always a large background of CO and
CO2 in oil (except in the first years of operation of the transformers), so that caution should be
exercised when interpreting the value of this ratio. Using increment values of CO and CO2 over the
last analysis is preferable, but the uncertainty on the incremented ratio is high and should be
calculated to determine its reliability.
Values of the CO2 to CO ratio > 10 are also an indication of thermal faults in paper at temperatures <
150 °C, but such temperatures have only a long-term aging effect on paper and on the reduction of
transformer life, which can be more precisely evaluated by furans formation, (when regular Kraft,
not thermally-upgraded paper, is used).
DGA results appearing in the T1 and T2 zones may also be an indication of paper involvement, since
most inspected cases of thermal faults in paper have been observed in these zones, as shown in
Figure 5. One should verify, however, that the oil used is not stray gassing, since stray gassing also
produces gases in these zones (see below). Thermal faults in oil, by comparison, are observed mostly
in the T3 zone.
A sharp increase in the formation of furans may in some cases be a confirmation of faults in paper at
temperatures > 250 °C.
The most severe faults, in terms of type and location, are generally considered as : - high-energy
arcing D2 in paper (and in oil). - medium-to-high temperature faults T2-T3 in paper (> 250 °C) - low
energy arcing D1 in paper (tracking, arcing) - high temperature faults T3 in oil (> 700 °C)
The less severe faults, which can often be tolerated for relatively long periods of time as long as they
don’t evolve into a more severe one are : - low-energy discharges PD/D1 in oil (corona, sparking) -
low temperature faults T1 in paper (< 150 °C) - medium temperature faults in oil (< 500 °C). - these
faults are difficult to find by visual inspection.
In breathing transformers, the normal O2 to N2 ratio is around 0.5. In sealed and nitrogen blanketed
ones, this ratio should be zero but in reality it often has a significant value because of leaks in
gaskets, tank covers, etc. A reduction in the value of the O2 to N2 ratio, below 0.3 in the case of
breathing transformers, is usually an indication of excessive heating inside the transformer.

A C2H2 / H2 ratio > 3 in the main tank is a probable indication of contamination from
currentbreaking activity in the LTC compartment.

Diagnostic Methods Summary

The most important aspect of DGA diagnostic methods is being able to correctly diagnosis recent or
potential faults within a monitored transformer. With that in mind, the main DGA interpretation
methods discussed in this paper are summarized in Figure 12 and their accuracy is shown in Figure
13, as evaluated using the IEC data bank of inspected transformer failures and various other reports.
 Figure 12

Figure 13

Conclusion
In this paper we presented the DGA diagnostic tools that have the ability to provide results indicating
a specific type of fault that is either present or pending to occur in a transformer. This information can
then be used to determine the appropriate course of action for that specific transformer. For example,
some faults can be considered less severe than others and the transformer can be allowed to continue
operating without immediate action to shut down and remove power.
Understanding the available DGA diagnostic methods can be overwhelming, especially when the
various types of techniques are taken into perspective. If, for instance, the only technique used, is to
take manual oil samples, send them to an oil laboratory, and then wait for the results and direction
from the lab, the various diagnostic tools presented here are not applicable because everything is
being determined by the oil laboratory used. However, if the end user wants to analyze the oil samples
and get results on their own, then the techniques and methods discussed will need to be fully
understood so that they are properly used.
The key objective in DGA of fault gases is to correctly diagnose the fault that is potentially generated.
Some diagnostic tools can perform better than others, so it’s important to review the most recent
information when incorporating them into DGA procedures.
We have presented and discussed the most common methods for analyzing dissolved gas data to
interpret potential fault types, including the Key Gas, Dornenburg Ratio, Rogers Ratio, IEC Ratio, Duval
Triangle, and CIGRE methods. Keep in mind that many of these methods may yield conflicting
diagnoses for the same oil sample and some results from the ratio methods (Dornenburg, Rogers, and
IEC) may be unresolved (no diagnosis). With the Duval Triangle Method, this is not a concern. The two
independent studies, as previously described, have shown that the Duval Triangle Method resulted in
a 96 percent correct rate of the determination of a transformer fault and although it had a higher
percentage of wrong diagnostics than the DRM, it had a lower rate of 0 percent for ‘unresolved’
diagnostics.
The Future
DGA diagnosis techniques presented thus far use fault gas concentrations or ratios based on the
practical experience of various experts, rather than on quantitative evidence. Now, with the
availability of extensive DGA data, researchers are motivated to develop an alternative approach to
DGA data interpretation alone. These different or alternate approaches include artificial intelligence
(AI) techniques, Fuzzy logic, and neural networks techniques.
Recent development of the artificial intelligent (AI) model based on a combination of KGM, DRM, RRM,
and DTM techniques shows some future insight, as shown in Figure 14. AI approaches provide more
accurate and reliable transformer diagnoses than KGM, DRM, RRM, and DTM alone. However, even
though most of the AI approaches can diagnose faults with high accuracy, some of them fail to
distinguish between thermal faults in oil and the same faults in cellulose, so engineering judgment is
still required.
Figure 14