See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/261923919

A review of dissolved gas analysis measurement and interpretation

techniques
Article in IEEE Electrical Insulation Magazine · May 2014
DOI: 10.1109/MEI.2014.6804740
CITATIONS
198
READS
13,558
3 authors:
Some of the authors of this publication are also working on these related projects:
Hybrid solar-wind-grid power generation system; Modeling, simulation and MPPT” View project
Transformer Asset Management View project
Norazhar Abu Bakar
Technical University of Malaysia Malacca
33 PUBLICATIONS 568 CITATIONS
SEE PROFILE
A. Abu-Siada
Curtin University
279 PUBLICATIONS 4,203 CITATIONS
SEE PROFILE
Sheikh Islam
Ludwig-Maximilians-University of Munich
517 PUBLICATIONS 7,100 CITATIONS
SEE PROFILE
All content following this page was uploaded by A. Abu-Siada on 14 May 2016.
The user has requested enhancement of the downloaded file.

May/June — Vol. 30, No. 3 39

FEATUREARTICLE
0883-7554/12/$31/©2014/IEEE

A Review of Dissolved Gas

Analysis Measurement
and Interpretation Techniques
Key words: dissolved gas analysis, gas chromatography, hydrogen on-line
monitoring,
photo-acoustic spectroscopy
Norazhar Abu Bakar
Electrical and Computer Engineering
Department, Curtin University,
Western Australia; and Faculty
of Electrical Engineering,
Universiti Teknikal Malaysia Melaka,
Malaysia
A. Abu-Siada and S. Islam
Electrical and Computer Engineering
Department, Curtin University,
Western Australia
A review of several techniques for
measurement of the concentrations
of dissolved gases in transformer oil,
and of several interpretations of dissolved gas analysis (DGA) data, is
presented.
Introduction
Dissolved gas analysis (DGA) is used to assess the condition
of power transformers. It uses the concentrations of various gases dissolved in the transformer oil due to
decomposition of the
oil and paper insulation. DGA has gained worldwide acceptance
as a method for the detection of incipient faults in transformers.
Due to the thermal and electrical stresses that the insulation
of operating transformers experience, paper and oil decomposition occurs, generating gases that dissolve in the oil
and reduce
its dielectric strength [1]–[6]. Gases generated through oil decomposition include hydrogen (H 2), methane (CH4),
acetylene
(C2H2), ethylene (C2H4), and ethane (C2H6). On the other hand,
carbon monoxide (CO) and carbon dioxide (CO2) are generated
as a result of paper decomposition [7], [8]. Faults such as overheating, partial discharge, and sustained arcing
produce a range
of gases, the concentrations of which can be used to identify
faults and estimate their severity.
In 1978 IEEE published guidelines for the detection of gases in
oil-immersed transformers, known as ANSI/IEEEC57.104-1978
[9]. It covers instrumentation, sampling procedures, methods for
extracting and analyzing gases, and data interpretation. In 1992
IEEE published further guidelines (IEEE Std C57.104-1991) [8]
that deal mainly with interpretation of DGA data. This standard
was revised in 2008 (IEEE Std C57.104-2008) [10]. In 1977
IEC published guidelines for the sampling of gases and oil in
oil-filled electrical equipment and the analysis of free and dissolved gases. In 2009 ASTM issued ASTM Standard
D3612-02
[2], which deals with analysis of gases dissolved in electrical
insulating oil using gas chromatography (GC).
GC measurements are always conducted in a laboratory environment because of the complexity of the equipment
required;
oil samples are collected from operating transformers and transported to the laboratory for gas extraction and
analysis. Vacuum
extraction, stripper extraction, and headspace sampling are currently used to extract gases from the oil [2]; the shake
test can
also be used [11]. After extraction the gases are analyzed using
GC. Due to the time and costs involved, DGA using GC is usually performed only once a year on operating
transformers. More
frequent testing is performed only if significant concentrations
of fault gases are detected during routine tests [12].
Various techniques for interpretation of DGA data have been
developed, e.g., the key gas, Doernenburg ratio, Rogers ratio,
IEC ratio, and Duval triangle methods. Each of these techniques
relies on the accumulated knowledge and experience of vari
40 IEEE Electrical Insulation Magazine
ous experts rather than rigorous quantitative scientific models
[1], [7], and therefore they may yield different diagnoses for the
same oil sample. These interpretation techniques are discussed
below.
On account of the limitations of the GC approach, several
new analytical techniques have been developed, e.g., hydrogen
on-line monitoring [13], [14] and photo-acoustic spectroscopy
(PAS) [12], [15]. Both require less time than GC analysis. These
three techniques are also discussed below.
DGA Methods
GC
GC has been used to analyze gases dissolved in insulating
oil for the last 60 years [16], [17]. It has been used regularly
by the UK Central Electricity Generating Board for monitoring
and routine assessment since 1968. It became more popular after
IEEE, IEC, and ASTM published relevant guidelines. Currently,
GC is accepted as the best of the three DGA techniques for measuring the concentrations of gases dissolved in
transformer oil,
e.g., combustible gases (hydrogen, carbon monoxide, methane,
ethane, ethylene, and acetylene) and noncombustible gases (carbon dioxide, nitrogen, and oxygen). It can also be
used to measure the concentrations of gases released into the space above
the oil, i.e., into the gas blanket. As stated above, GC analysis is
conducted only in the laboratory environment, following several
standards, e.g., those dealing with extraction of the oil sample
from the operating transformer, transporting it to the laboratory,
and extracting the dissolved gases [2], [9], [18]–[20].
According to ANSI/IEEEC57.104-1978 [9], oil samples can
be stored and transported to the laboratory in calibrated stainless
steel cylinders, flexible metal cans, syringes, or glass bottles.
All the containers must meet stated leak criteria. According to
ASTM D923 [20], amber or clear glass bottles may be used, fitted with glass stoppers or with screw caps
incorporating a pulpboard liner faced with tin or aluminum foil, or a suitable oilresistant plastic such as
polyethylene, polytetrafluoroethylene,
or a fluoro-elastomer. ASTM D3612 [2] states that gases in the
oil can be separated using vacuum extraction, stripper column
extraction, or headspace sampling methods. Vacuum extraction
(Figure 1) is suitable for extraction of a fraction of each of the
dissolved gases, whereas stripper column extraction will remove
nearly all of each gas. Headspace sampling (Figure 2) will also
extract a fraction of each headspace gas. Another extraction
method, known as the Shake Test, was introduced by Morgan–
Shaffer in 1993 [11]. It can extract dissolved gases quickly, even
in the field.
A basic gas chromatograph setup is shown in Figure 3. The
oil sample is volatilized in the injection port, and its gaseous
components separated in the column [17]. A gas entering the
column may remain in the gaseous state, i.e., continue to travel
with the carrier gas, or dissolve in the liquid on the surface of the
stationary phase, or condense on the stationary phase. In the latter two cases its progress through the column will be
much slower than if it continued to travel with the carrier gas. Argon, helium, nitrogen, and hydrogen are normally
used as carrier gases
to transfer the sample from the injector through the columns into
the detector [2], [17], [21]. Some of the gaseous components
that travel with the carrier gas collide and re-enter the stationary
phase, which is a thin film of thermally stable polymer coated
on the inner wall of the column. Simultaneously, the gaseous
Figure 1. Extraction of dissolved gases from insulating oil using
the vacuum extraction method [2].
Figure 2. Extraction of dissolved gases from insulating oil using the headspace method [2].
May/June — Vol. 30, No. 3 41
components immobilized in stationary the phase travel with the
carrier gas. This occurs thousands of times for each gaseous
component during its travel along the column. The time taken by
each gaseous component to pass through the column (the retention time) varies from component to component, and
depends on
the temperature of the column and the chemical structure of the
stationary phase. The separation characteristics of the columns
and the detection sensitivity may vary with the carrier gas [21].
There are two main types of column, namely packed and
capillary. Most users prefer capillary columns with a stationary phase coated on the inner wall [17]. Capillary
columns have
smaller diameters but are longer than packed columns and, therefore, have substantially higher separation capacities.
Separation
capacity is defined as the number of gaseous components that
can be separated between the carrier gas and the stationary phase
within the column [17]. Smaller diameter and lengthier columns
will slow the travel rates of gaseous components and hence slow
the separation process. It is essential that the temperature of the
column be kept constant, in order to achieve known and constant
separation capacity.
The gaseous components are washed out of the column one
by one, depending on their retention time, and interact with the
heated detectors, generating electrical signals [21]. These signals are recorded by a data-collection system and
plotted against
elapsed time, producing a chromatogram. Gases can be identified using the retention time, and the concentration of a
given
gas is determined by the magnitude of the associated electrical
signal.
Several different types of detector are available, e.g., thermal conductivity, flame ionization, nitrogen–phosphorus,
flame
photometry, electron capture, atomic emission, and electrolytic
conductivity [17], [21].
Flame ionization detectors are normally used to detect hydrocarbon and carbon oxide gases, due to their high
sensitivity,
and thermal conductivity detectors are used to detect permanent
gases such as H2, O2, and N2 [2]. Mass spectrometer detection
using electron ionization techniques is very sensitive and is normally used for measurements on complex gaseous
mixtures.
Molecules such as methyl acetate, 2-methylfuran, phenol,
methyl formate, furan, methanol, ethanol, acetone, isopropyl alcohol, and methyl ethyl ketone, present in the oil as
solids, can
also be detected [22]–[25].
Hydrogen On-Line Monitoring
The hydrogen on-line monitor [16], [26] is a rugged low-cost
device introduced by Syprotec and further developed by the Institut de Recherche d’Hydro Quebec. It is widely
accepted that
a majority of faults in oil-filled electrical equipment lead to the
generation of hydrogen [8]; the hydrogen on-line monitor therefore focuses on key gases such as hydrogen and
carbon monoxide [14]. In this way early detection of faults, especially hot
spots, partial discharges, and arcing, can be achieved. Ethylene
and acetylene can also be detected but with reduced sensitivity.
A hydrogen on-line monitor consists essentially of a sensor in
contact with the oil, and an electronic unit. The sensor is placed
in a rugged brass housing containing a fuel cell, a temperature
sensor, and a Teflon membrane (Figure 4). The sensor can be
installed on a flange or valve on the transformer pipe work between the cooling bank and the main tank [16], or on
the upper part of the transformer [27]. Hydrogen in the oil permeates
through the Teflon membrane, along with atmospheric oxygen,
and is chemically burnt in the electrolyte cell, thus generating a
small current proportional to the hydrogen gas concentration in
the oil. (Simultaneously, water is formed as a result of the reaction between hydrogen and atmospheric oxygen.) The
current is
amplified and measured as a voltage drop across the load resistor
R connected between two porous electrodes, the voltage drop
yielding the hydrogen gas concentration in the oil. An alarm is
activated if the hydrogen concentration reaches a predetermined
level.
Figure 3. A basic gas chromatograph.
42 IEEE Electrical Insulation Magazine
The detection sensitivity of the monitor depends on the percentage of the individual combustible gases passing
through the
membrane. The detection percentages are currently 100 for hydrogen, about 15 for carbon monoxide, 8 for
acetylene, and 1 for
ethylene [16], [27].
Approximately 18,000 hydrogen on-line monitor systems
were in operation globally in 2003 [11]. The hydrogen concentration measurement accuracy is ±10% over the
temperature
range 20 to 40°C [27]. The accuracy diminishes if the monitor
temperature varies outside this range, say as a result of variation
of the oil temperature [26]. Although the monitor can detect incipient faults in a power transformer, it cannot
identify the fault
types.
Photo-Acoustic Spectroscopy (PAS)
In 1880 Alexander Graham Bell observed that a sound was
emitted when a thin disk was exposed to mechanically chopped
sunlight [28]; this may be the first recorded observation of the
photo-acoustic effect. Subsequently, photo-acoustic spectroscopy has been used in various applications such as
monitoring
of ambient air quality and air pollution from car exhausts. However, its use to monitor the condition of power
transformers is
relatively new [12].
The basic principle is that fault gases absorb infrared light
energy and convert it into kinetic energy in the form of pressure
(sound) waves, which are detected by a microphone [12]. This
microphone converts the pressure waves into electrical signals
[29]. The photo-acoustic spectrum of the fault gases is obtained
 by measuring the intensity of the sound waves produced by various infrared wavelengths, the latter selected using
optical filters.
In this way the concentrations of various fault gases, e.g., H2,
CO, CO2, CH4, C2H4, C2H2, and C2H6, can be obtained [12], [29].
Figure 4. Schematic of a hydrogen on-line monitor [16].
Figure 5. Basic photo-acoustic spectroscopy [30].
Figure 6. Schematic of a photo-acoustic spectroscopy–based dissolved-gas-analysis system [29].
May/June — Vol. 30, No. 3 43
Figure 5 shows the basic operating principle, and Figure 6 shows
some construction details [29].
PAS is a very stable diagnostic tool, suitable for monitoring
the condition of power transformers [12]. However, since each
fault gas absorbs infrared light at a specific wavelength, selecting the center wavelength for each gas is a critical
process [29].
Figure 7 shows the characteristic absorption of various gases,
and water vapor. The detection sensitivity of the equipment varies from gas to gas, and its detection accuracy is
influenced by
the external environment, e.g., temperature and pressure, and by
vibration [31]. The advantages and disadvantages of GC, hydrogen on-line monitoring, and PAS are compared in
Table 1.
Interpretation of DGA Data
Key Gas Method (KGM)
The generation of fault gases within transformer oil requires
sufficiently high temperatures (energy) to break chemical bonds
in the oil, which then reduces its dielectric insulation strength
[28]. The KGM uses the individual concentrations of the six
Figure 7. Characteristic absorption of fault gases [29].
Table 1. Comparison of Gas
Chromatography (GC), Hydrogen On-Line
Monitor, and Photo-Acoustic Spectroscopy
(PAS).
Method Advantage Disadvantage
• Able to detect and analyze every • Can be used only in the laboratory,
GC gas dissolved due to the complexity
in transformer oil of the equipment
• Long time required to complete a
• Provides highest accuracy and repeatability
test on a transformer
• Results can be used to identify the fault type • Expensive
• An expert is needed to conduct the test and interpret the data
• Rugged, low-cost, and continuous • Can detect only H2, CO, C2H2, and
Hydrogen on-line monitor
on-line monitoring C2H4
• Best concentration accuracy only
• Can detect incipient faults within the monitor
temperature range 20–40°C
• Results cannot be used to identify the fault type
• Results are sensitive to the wave
number range of the optical
PAS • Continuous on-line monitoring
filters and their absorption
characteristics
• Concentration accuracy
• Can detect and measure the concentration of a wide range influenced by the external
of fault gases temperature and pressure, and by
vibration
• Results can be used to identify the fault type • Still undergoing development
 44 IEEE Electrical Insulation Magazine
fault gases shown in Figure 8. The four common fault conditions are distinguished by the quoted percentage
concentrations
of the six gases. These percentages are based on the practical
experience of various experts.
These KGM charts look simple, but they are not widely accepted as reliable diagnostic tools for power transformers
[32].
Studies based on an IEC data bank of inspected transformers
show that only 42% of KGM diagnoses are correct [33].
Doernenburg Ratio Method (DRM)
This method [34], [35] uses the ratio of gas concentrations
to indicate fault types. Predefined limits for the CH4/H2, C2H2/
C2H4, C2H2/CH4, and C2H6/C2H2 ratios are used to interpret the
DGA results (Table 2) [34].
DRM diagnosis cannot be applied unless the concentration of
at least one of the key gases (H2, C2H4, CH4, and C2H2) exceeds
twice the relevant L1 concentration (Table 3) and the concentration of at least one of the two gases appearing in any
one of
the four ratios exceeds the relevant L1 concentration [8]. The
proposed fault diagnosis is based on the ranges of the four ratios
shown in Table 2.
Rogers Ratio Method (RRM)
This method is similar to the DRM. However, whereas the
DRM requires significant concentrations of the fault gases, the
RRM can be used provided the concentrations exceed the L1
values in Table 3 [8].
The RRM originally used four concentration ratios, namely
C2H6/CH4, C2H2/C2H4, CH4/H2, and C2H4/C2H6, leading to 12
proposed diagnoses [9]. However, the condition C2H6/CH4 <1
held for 10 of the 12 suggested diagnoses, i.e., the ratio C2H6/
CH4 was of little diagnostic value [32], [35]. This ratio was
therefore omitted in the revised IEEE Standard C57.104-1991
[8], and the original 12 suggested diagnoses were replaced by
six (including the normal state), as shown in Table 4 [34]. However, inconsistencies have been reported [1], [7], the
success rate
for correct fault type identification being 58.9% [7].
IEC Ratio Method (IRM)
This method uses the same three ratios as the revised RRM but
suggests different ratio ranges and interpretations, as shown in
Table 5 [36]. A new gas ratio has been introduced, namely C2H2/
H
2 to detect possible contamination from on-load tap-changer

compartments [38]. Another improvement is 3-D graphical representation of ratio ranges, which yields more reliable
diagnoses, and diagnoses of faults associated with ratios outside the
ranges quoted in Table 5 [1].
Duval Triangle Method (DTM)
This method was developed from an existing IEC 60599 Ratio method and IEC TC10 databases [36]. It interprets
DGA data
Figure 8. Key gases method chart [16].
Table 2. Doernenburg
Ratio Method
Concentration Ratios
[34].
Suggested fault diagnosis CH4/H2 C2H2/C2H4 C2H2/CH4 C2H6/C2H2
Gas Gas
Oil Gas space Oil Gas space Oil Oil
space space
Thermal fault >1 >0.1 <0.75 <1 <0.3 <0.1 >0.4 >0.2
Not
PD <0.1 <0.01 <0.3 <0.1 >0.4 >0.2
significant
Arcing >0.1 to <1 >0.01 to <0.1 >0.75 >1 >0.3 >0.1 <0.4 <0.2

May/June — Vol. 30, No. 3 45

using graphical presentation [39], [40]. It uses the concentrations of CH4, C2H2, and C2H4, which are plotted along
three sides
of a triangle [39], as shown in Figure 9. Within the triangle there
are seven fault zones, covering partial discharge, thermal faults
at various temperatures, and electrical arcing. According to [7],
[35], and [37], the DTM provides more accurate and consistent diagnoses than the other ratio methods. However,
careless
implementation leads to incorrect diagnoses [40]. Moreover,
since the triangle does not include the fault-free condition, the
method cannot be used to detect incipient faults. Duval [36] proposed other versions of DTM, e.g., DTM 2 for oil-
type load tap
changers, DTM 3 for equipment filled with nonmineral oils, and
DTM 4 and DTM 5 as complements of the original DTM for
low-temperature faults in transformers. H2, CH4, and C2H6 are
used in DTM 4 (Figure 10), and C2H4, CH4, and C2H6 in DTM 5
(Figure 11). DTM 4 and DTM 5 are applicable only when PD,
T1, or T2 faults have been identified by the original DTM. The
various DGA interpretation methods discussed above are summarized in Table 6.
Artificial Intelligence and DGA
As stated above, fault diagnosis techniques using fault gas
concentrations or ratios thereof are based on the practical experience of various experts, rather than on quantitative
evidencebased scientific theory, and may therefore lead to different conclusion for the same oil sample methods
[41]. Ratio methods
may also fail to diagnose multiple simultaneous fault conditions.
The availability of extensive DGA data has therefore motivated researchers to develop an alternative approach to
DGA data
interpretation, based on artificial intelligence (AI) techniques
[37], [41]–[49]. Fuzzy logic [41], [44], [45] and neural network
[37], [42], [46], [48] techniques have been popular for this pur
Table 4. Revised Suggested Rogers Ratio Method Diagnoses.
C2H2/ C2H4/ Suggested fault
Case CH4/H2
C2H4 C2H6 diagnosis
0 <0.1 >0.1 to <1 <1 Normal
Low energy
1 <0.1 <0.1 <1
density PD
2 0.1 to 3 0.1 to 1 >3 Arcing
Low temperature
3 <0.1 >0.1 to <1 1 to 3
thermal fault
Thermal fault
4 <0.1 >1 1 to 3
<700°C
Thermal fault
5 <0.1 >1 >3
>700°C

Table 3. Dissolved Gas Concentrations (L1) for the Doernenburg Ratio

Method [8].
Key gas L1 concentrations (ppm)
Hydrogen (H2) 100
Methane (CH4) 120
Carbon monoxide (CO) 350
Acetylene (C2H2) 35
Ethylene (C2H4) 50
Ethane (C2H6) 65
Figure 9. Coordinates and fault zones in the Duval triangle method [39].
46 IEEE Electrical Insulation Magazine
Figure 10. Coordinates and fault zones in the DTM 4 [36].
Figure 11. Coordinates and fault zones in the DTM 5 [36].
May/June — Vol. 30, No. 3 47
pose, followed by support vector machine [43], [50], [51] and
particle swarm optimization [49] techniques. Abu-Siada et al.
[41] developed an AI model based on a combination of KGM,
DRM, RRM, IRM, and DTM (Figure 12). KGM is first applied.
If the key gas concentrations are less than the L1 concentrations
quoted in Table 3, the transformer is assumed to be fault free,
and no further analysis is performed. If the key gas concentrations exceed the L1 concentrations quoted in Table 3,
the ratio
methods and DTM are applied in order to diagnose the fault
type. The fault type indicator D is calculated as
D
AD
A
iii
i
ii
=i
==
==

∑∑
51
51
, (1)
where D
i (a number in the range 0–10) is the decision value derived from fuzzy logic models based on the ith method (KGM,

DRM, RRM, IRM, and DTM) and Ai is the accuracy of the ith
method obtained from an analysis of 2,000 DGA samples with
preknown faults [41]. If any of the ratio methods involves a ratio
that is not consistent with the relevant proposed codes, the corresponding Di is set to zero. D takes a numerical value
in the range
0 to 10, where 0 ≤ D < 2 indicates no fault, 2 ≤ D < 4 indicates
overheated cellulose, 4 ≤ D < 6 indicates overheated oil, 6 ≤ D
< 8 indicates corona in the oil, and 8 ≤ D < 10 indicates arcing
in the oil.
As reported in [37], [41], [49], [51], AI approaches provide
more accurate and reliable transformer diagnoses than KGM,
DRM, RRM, IRM, and DTM. However, although a majority of
the AI approaches diagnose faults with high accuracy, some of
them fail to distinguish between thermal faults in oil and the
same faults in cellulose, and engineering judgment is then required [41], [42].
Conclusion
Three methods of measuring the concentrations of fault gases
 dissolved in transformer oil, namely gas chromatography, hydrogen on-line monitoring, and photo-acoustic
spectroscopy, are
compared and discussed in this article. The high accuracy of gas
chromatography is widely acknowledged. However, gas chromatography measurements are expensive and time
consuming,
and industry therefore tends to favor hydrogen on-line monitoring and photo-acoustic spectroscopy. Hydrogen on-
line monitoring can detect incipient faults but cannot provide detailed
fault diagnosis. Photo-acoustic spectroscopy provides more accurate gas concentration data than hydrogen on-line
monitoring,
but its accuracy may be affected by external gas pressure and
Table 5. Suggested IEC Ratio Method Diagnoses [37].1
Case Characteristic fault C2H2/C2H4 CH4/H2 C2H4/C2H6
PD Partial discharges NS <0.1 <0.2
Discharges of
D1 >1 0.1–0.5 >1
low energy
D2 High energy discharges 0.6–2.5 0.1–1 >2
T1 Thermal faults <300°C NS >1 but NS <1
Thermal faults >300°C
T2 <0.1 >1 1–4
and <700°C
T3 Thermal faults >700°C <0.2 >1 >4
1NS = not significant whatever the value.

Table 6. Comparison Among Key Gas Method (KGM), Doernenburg Ratio Method (DRM), Rogers Ratio Method
(RRM), IEC Ratio Method (IRM), and Duval
Triangle Method (DTM).
Type Method
Uses individual gas concentrations, easy t
KGM
very conservative
Uses four gas concentration ratios (CH4/H
C2H2/CH4, and C2H6/C2H2) to indicate thr
DRM
uses specified concentration limits to diffe
between faults
Uses three gas concentration ratios (C2/H2
and
RRM
C2H4/C2H6) to indicate five fault types, us
concentration limits to differentiate betwe
Similar to RRM but excludes the C2H6/CH
IRM six fault types, uses specified concentratio
to differentiate between faults
Uses triangular map to indicate six faults,
DTM
normal state

48 IEEE Electrical Insulation Magazine

vibration. Several methods are available for the interpretation of
DGA data, the Doernenburg, Rogers, IEC ratio, Duval triangle,
and key gas methods being widely used by utilities. Unfortunately, these methods may yield conflicting diagnoses
for the
same oil sample. Several artificial intelligence methods have
been developed in response to this situation. However, in some
cases a certain amount of engineering judgment may be required
to obtain a credible diagnosis.
References
[1] H.-C. Sun, Y.-C. Huang, and C.-M. Huang, “A review of dissolved gas
analysis in power transformers,” Energy Procedia, vol. 14, pp. 1220–
1225, 2012.
[2] Standard Test Method for Analysis of Gases Dissolved in Electrical Insulating Oil by Gas Chromatography, ASTM D3612-02 (Reapproved
 2009),
2009.
[3] T. Suwanasri, E. Chaidee, and C. Adsoongnoen, “Failure statistics
and power transformer condition evaluation by dissolved gas analysis
technique,” in International Conference on Condition Monitoring and
Diagnosis, 2008. CMD 2008, 2008, pp. 492–496.
[ H.-T. Yang and C. C. Liao, “Adaptive fuzzy diagnosis system for
4 dis
] solved gas analysis,” IEEE Trans. Power Del., vol. 14, p. 1342,
1999.
[5] K. Spurgeon, W. H. Tang, Q. H. Wu, Z. J. Richardson, and G. Moss,
“Dissolved gas analysis using evidential reasoning,” in IEE Proceedings—Science, Measurement and Technology, vol. 152, 2005, p. 110.
[6] A. Emsley and G. Stevens, “Kinetics and mechanisms of the low-temperature degradation of cellulose,” Cellulose, vol. 1, pp. 26–56, 1994.
[7] A. Abu-Siada and S. Islam, “A new approach to identify power transformer criticality and asset management decision based on dissolved
gas-in-oil analysis,” IEEE Trans. Dielectr. Electr. Insulation, vol. 19, pp.
1007–1012, 2012.
[8] IEEE Guide for the Interpretation of Gases Generated in Oil-Immersed
Transformers, IEEE Std C57.104-1991, p. 0_1, 1992.
Figure 12. Artificial intelligence interpretation method proposed in [41]. DGA = dissolved gas analysis, KGM = key
gas
method, DTM = Duval triangle method.
[9] IEEE Guide for the Detection and Determination of Generated Gases in
Oil-Immersed Transformers and Their Relation to the Serviceability of
the Equipment, ANSI/IEEE Std C57.104-1978, p. 0_1, 1978.
[10] IEEE Guide for the Interpretation of Gases Generated in Oil-Immersed
Transformers, IEEE Std C57.104-2008 (Revision of IEEE Std C57.104-
1991), pp. C1–27, 2009.
[11] M. Duval, “New techniques for dissolved gas-in-oil analysis,” IEEE
Electr. Insul. Mag., vol. 19, pp. 6–15, 2003.
[12] D. Skelly, “Photo-acoustic spectroscopy for dissolved gas analysis:
Benefits and experience,” in 2012 International Conference on Condition
Monitoring and Diagnosis (CMD), 2012, pp. 29–43.
[13] A. de Pablo, W. Ferguson, A. Mudryk, and D. Golovan, “On-line condition monitoring of power transformers: A case history,” in Electrical
Insulation Conference (EIC), 2011, 2011, pp. 285–288.
[14] G. L. Martin, “The Hydran(R)—A system for the detection and monitoring
of failure conditions in power transformers,” in IEE Colloquium on Monitors and Condition Assessment Equipment (Digest No. 1996/186), 1996,
pp. 3/1–3/5.
[15] A. Stadler, “Analyzing UV/Vis/NIR and photo-acoustic spectra: A note to
the band gap of Tix Si1 - xO2,” IEEE Trans. Semicond. Manuf., vol. 26,
pp. 156–161, 2013.
[16] M. Webb, “Anticipating failures by dissolved-gas monitoring,” Power
Eng. J., vol. 1, pp. 295–298, 1987.
[17] O. D. Sparkman, Z. E. Penton, and F. G. Kitson, “Chapter 2—Gas chromatography,” in Gas Chromatography and Mass Spectrometry, 2nd
ed.
Amsterdam, the Netherlands: Academic Press, 2011, pp. 15–83.
[18] Standard Guide for Sampling, Test Methods, and Specifications for Electrical Insulating Oils of Petroleum Origin, ASTM D117-10, 2010.
[19] Standard Practice for Sampling Gas from a Transformer Under Positive
Pressure, ASTM D2759-00, 2010.
[20] Standard Practices for Sampling Electrical Insulating Liquids, ASTM
D923-07, 2007.
[21] E. Forgács and T. Cserháti, “CHROMATOGRAPHY | Principles,” in Encyclopedia of Food Sciences and Nutrition, 2nd ed., L. Trugo, P.
Finglas,
B. Caballero, Ed. Oxford: Academic Press, 2003, pp. 1259–1267.
[22] S. Okabe, S. Kaneko, M. Kohtoh, and T. Amimoto, “Analysis results for
insulating oil components in field transformers,” IEEE Trans. Dielectr.
Electr. Insul., vol. 17, pp. 302–311, 2010.
[23] J. Jalbert, R. Gilbert, Y. Denos, and P. Gervais, “Methanol: A novel
approach to power transformer asset management,” IEEE Trans. Power
Del., vol. 27, pp. 514–520, 2012.
[24] J. Jalbert, S. Duchesne, E. Rodriguez-Celis, P. Tetreault, and P. Collin,
“Robust and sensitive analysis of methanol and ethanol from cellulose
degradation in mineral oils,” J. Chromatogr. A, vol. 1256, pp. 240–245,
Sep. 21, 2012.
[25] A. Schaut, S. Autru, and S. Eeckhoudt, “Applicability of methanol as
new marker for paper degradation in power transformers,” IEEE Trans.
Dielectr. Electr. Insul., vol. 18, pp. 533–540, 2011.
[26] J. L. Kirtley, W. H. Hagman, B. C. Lesieutre, M. J. Boyd, E. P. Warren,
H. P. Chou, and R. D. Tabors, “Monitoring the health of power transformers,” IEEE Comp. Appl. Power, vol. 9, pp. 18–23, 1996.
[27] D. Jie, I. Khan, Z. D. Wang, and I. Cotton, “Comparison of HYDRAN
and laboratory DGA results for electric faults in ester transformer fluids,”
in Annual Report—Conference on Electrical Insulation and Dielectric
Phenomena, 2007. CEIDP 2007, 2007, pp. 731–734.
[28] N. A. Muhamad, B. T. Phung, T. R. Blackburn, and K. X. Lai, “Comparative study and analysis of DGA methods for transformer mineral
oil,” in
2007 IEEE Lausanne Power Tech, 2007, pp. 45–50.
[29] Q. Zhu, Y. Yin, Q. Wang, Z. Wang, and Z. Li, “Study on the online dissolved gas analysis monitor based on the photoacoustic
spectroscopy,” in
2012 International Conference on Condition Monitoring and Diagnosis
(CMD), 2012, pp. 433–436.
[30] C. Haisch and R. Niessner, “Light and sound-photoacoustic spectroscopy,” Spectroscopy Eur., vol. 14/5, 2002.
[31] W. Fu, C. Weigen, P. Xiaojuan, and S. Jing, “Study on the gas pressure
characteristics of photoacoustic spectroscopy detection for dissolved
gases in transformer oil,” in 2012 International Conference on High Voltage Engineering and Application (ICHVE), 2012, pp. 286–289.
[32] S. A. Ward, “Evaluating transformer condition using DGA oil analysis,”
in Annual Report. Conference on Electrical Insulation and Dielectric
Phenomena, 2003, pp. 463–468.
May/June — Vol. 30, No. 3 49
[33] S. Corporation, “Serveron White Paper: DGA Diagnostic Methods,”
2007.
[34] IEEE Guide for the Interpretation of Gases Generated in Oil-Immersed
Transformers—Redline, IEEE Std C57.104-2008 (Revision of IEEE Std
C57.104-1991)—Redline, pp. 1–45, 2009.
[35] R. R. Rogers, “IEEE and IEC codes to interpret incipient faults in transformers, using gas in oil analysis,” IEEE Trans. Electr. Insul., vol. EI-
13,
pp. 349–354, 1978.
[36] M. Duval, “The duval triangle for load tap changers, non-mineral oils and
low temperature faults in transformers,” IEEE Electr. Insul. Mag., vol.
24, pp. 22–29, 2008.
[37] V. Miranda and A. R. G. Castro, “Improving the IEC table for transformer failure diagnosis with knowledge extraction from neural
networks,”
IEEE Trans. Power Del., vol. 20, pp. 2509–2516, 2005.
[38] M. Duval and A. dePabla, “Interpretation of gas-in-oil analysis using new
IEC publication 60599 and IEC TC 10 databases,” IEEE Electr. Insul.
Mag., vol. 17, pp. 31–41, 2001.
[39] M. Duval, “A review of faults detectable by gas-in-oil analysis in transformers,” IEEE Electr. Insul. Mag., vol. 18, pp. 8–17, 2002.
[40] M. Duval and J. Dukarm, “Improving the reliability of transformer gasin-oil diagnosis,” IEEE Electr. Insul. Mag., vol. 21, pp. 21–27, 2005.
[41] A. Abu-Siada, S. Hmood, and S. Islam, “A new fuzzy logic approach for
consistent interpretation of dissolved gas-in-oil analysis,” IEEE Trans.
Dielectr. Electr. Insul., vol. 20, pp. 2343–2349, 2013.
[42] K. F. Thang, R. K. Aggarwal, D. G. Esp, and A. J. McGrail, “Statistical
and neural network analysis of dissolved gases in power transformers,” in
Eighth International Conference on Dielectric Materials, Measurements
and Applications, 2000, IEE Conf. Publ. No. 473, 2000, pp. 324–329.
[43] A. K. Mehta, R. N. Sharma, S. Chauhan, and S. Saho, “Transformer diagnostics under dissolved gas analysis using Support Vector
Machine,” in
2013 International Conference on Power, Energy and Control (ICPEC),
2013, pp. 181–186.
[44] N. A. Muhamad, B. T. Phung, and T. R. Blackburn, “Comparative study
and analysis of DGA methods for mineral oil using fuzzy logic,” in International Power Engineering Conference, 2007. IPEC 2007, 2007, pp.
1301–1306.
[45] D. R. Morais and J. G. Rolim, “A hybrid tool for detection of incipient
faults in transformers based on the dissolved gas analysis of insulating
oil,” IEEE Trans. Power Del., vol. 21, pp. 673–680, 2006.
[46] C. Pan, W. Chen, and Y. Yun, “Fault diagnostic method of power
transformers based on hybrid genetic algorithm evolving wavelet neural
network,” IET Electr. Power Appl., vol. 2, pp. 71–76, 2008.
[47] A. Shintemirov, W. Tang, and Q. H. Wu, “Power transformer fault classification based on dissolved gas analysis by implementing bootstrap
and
genetic programming,” IEEE Trans. Syst., Man, Cybern., Part C: Appl.
Rev., vol. 39, pp. 69–79, 2009.
[48] H.-X. Wang, Q.-P. Yang, and Q.-M. Zheng, “Artificial neural network for
transformer insulation aging diagnosis,” in Third International Conference on Electric Utility Deregulation and Restructuring and Power
Technologies, 2008. DRPT 2008, 2008, pp. 2233–2238.
[49] Z. J. Richardson, J. Fitch, W. H. Tang, J. Y. Goulermas, and Q. H. Wu, “A
probabilistic classifier for transformer dissolved gas analysis with a particle swarm optimizer,” IEEE Trans. Power Del., vol. 23, pp. 751–759,
2008.
[50] W. Tang, S. Almas, and Q. H. Wu, “Transformer dissolved gas analysis
using least square support vector machine and bootstrap,” in Chinese
Control Conference, 2007. CCC 2007, 2007, pp. 482–486.
[51] K. Bacha, S. Souahlia, and M. Gossa, “Power transformer fault diagnosis
based on dissolved gas analysis by support vector machine,” Electr.
Power Syst. Res., vol. 83, pp. 73–79, 2012.
Norazhar Abu Bakar received the BEng
(Hons) degree in electronics and electrical
from Leeds University, UK, and the MSc
(Eng) degree in control systems from Sheffield University, UK. He served as a lecturer
at Universiti Teknikal Malaysia Melaka,
Malaysia. Currently, he is a PhD student at
Curtin University. His research interests are
in condition monitoring and advanced control systems.
Ahmed Abu-Siada (M ’07, SM ’12) received his BSc and MSc degrees from Ain
Shams University, Egypt, and the PhD degree from Curtin University, Australia, all
in electrical engineering. Currently, he is a
senior lecturer in the Department of Electrical and Computer Engineering at Curtin
University. His research interests include
power system stability, condition monitoring, power electronics, and power quality.
He is editor-in-chief of the international journal Electrical and
Electronic Engineering and a regular reviewer for IEEE Transactions on Dielectrics and Electrical Insulation, IEEE
Transactions on Power Electronics, and IEEE Transactions on Sustainable Energy. He is the vice-chair of the IEEE
Computation
Intellegence Society, WA Chapter.
Syed Islam (M ’83, SM ’93) received the
BSc degree from Bangladesh University of
Engineering and Technology, Bangladesh,
and the MSc and PhD degrees from King
Fahd University of Petroleum and Minerals, Saudi Arabia, all in electrical power engineering in 1979, 1983, and
1988 respectively. He is currently the chair professor of
Electrical Power Engineering at Curtin University, Australia. He received the IEEE T Burke Haye Faculty
Recognition award in 2000. His research interests are in condition monitoring of transformers, wind energy
conversion, and
power systems. He is a regular reviewer for IEEE Transactions
on Energy Conversion, IEEE Transactions on Power Systems,
and IEEE Transactions on Power Delivery. Islam is an editor of
the IEEE Transactions on Sustainable Energy.
View publication stats