Professional Documents
Culture Documents
6.1 Introduction
for different fault temperatures and the ratios of certain key gas concentrations are
indicative to fault types.
The dissolved gas analysis (DGA) has been a widely utilised and powerful tool
to detect incipient faults in oil-filled power transformers [2–5]. The traditional
practice of diagnosing transformer conditions through DGA is carried out off-line
by manually extracting a sample of transformer insulation oil (by syringe), sending
it to a laboratory, and waiting for diagnosis results. By applying DGA techniques
on an oil sample, dissolved gases can be quantified. The concentration and the
relation of individual gases can predict whether a fault has occurred and what type
it is likely to be. Over the last four decades, DGA and its interpretation have
become a popular and reliable tool for assessing conditions of oil-filled trans-
formers and other oil-filled electrical equipment.
H2
CH4 C2H6 C3H8 ....
C2H4 C3H6 ...
C2H2
the oil, namely H2, CO, CO2, CH4, C2H6, C2H4, C2H2, etc. If a certain level is
exceeded, gas bubbles arise and oil-filled transformers are subject to electrical
and thermal stresses, which may break down insulation materials and release
gaseous decomposition products. Evaluation procedures for DGA have been
implemented widely based upon the guidelines recommended by IEC [2], IEEE
[3] and CIGRE [4].
The immediate effect of the breakdown of hydrocarbon molecules as a result of
the energy of a fault is to create free radicals as indicated in Fig. 6.1[1]. These
subsequently recombine to produce low molecular weight hydrocarbon gases. This
recombination process is largely determined by operation temperatures, but also
influenced by other factors. The result is that the pattern of gases appearing in the
oil has a form as shown in Fig. 6.2. For the lowest temperature faults both CH4 and
H2 are generated, with CH4 being predominant. As the temperature of a fault
increases, C2H6 starts to be evolved and CH4 is reduced, so that the C2H6/CH4
ratio becomes predominant. At still higher temperatures the rate of C2H6 evolution
is reduced and C2H4 production commences and soon outweights the proportion of
C2H6. Finally, at very high temperatures C2H4 puts in an appearance and as the
temperature increases still further it becomes the most predominant gas. It is noted
that no temperature scale is indicated along the temperature axis of Fig. 6.2, and
the graph is subdivided into types of faults. The areas include normal operating
temperatures go up to about 140°C, hot spots extend to around 250°C and high-
temperature thermal faults to about 1000°C. Peak C2H4 evolution occurs at about
700°C [1].
98 6 Transformer Condition Assessment Using Dissolved Gas Analysis
Hydrogen
Methane
Arcing/sparking
Gas Evolution
Ethane
Ethylene
High-temperature
Hot spots Acetylene
thermal faults
Temperature
The key gas method relates key gases to fault types and attempts to detect four
fault types [3], including overheating of oil, overheating of cellulose, corona
(partial discharge) and arcing, based on key gas concentrations (C2H4, CO, H2,
C2H2) expressed in ppm (part per million).
1. Overheating of oil: Decomposition products include C2H4 and CH4, together
with small quantities of H2 and C2H6. Traces of C2H2 may be formed if a fault
is severe or involves electrical contacts. The principal gas is C2H4.
2. Overheating of cellulose: Large quantities of CO2 and CO are evolved from
overheated cellulose. Hydrocarbon gases, such as CH4 and C2H4, are formed if
a fault involves an oil-immersed structure. The principal gas is CO.
3. Corona: Low-energy electrical discharges produce H2 and CH4, with small
quantities of C2H6 and C2H4. Comparable amounts of CO and CO2 may result
from discharges in cellulose. The principal gas is H2.
4. Arcing: Large amounts of H2 and C2H2 are produced, with minor quantities of
CH4 and C2H4. CO2 and CO may also be formed if a fault involves cellulose.
The insulation oil may be carbonised. The principal gas is C2H2.
The suggested relationships between key gases and fault types are summarised
as follows:
1. O2 and N2: non-faults.
2. H2: corona.
6.2 Fundamental of Dissolved Gas Analysis 99
A detected gas volume and its distribution may be generated over a long time
period by a relatively insignificant fault or in a very short period of time by a more
severe fault. Hence, one measurement does not indicate the rate of gas generation
and may provide very little information about the severity of a fault. Once a
suspicious gas presence is detected, it is important to sample again and calculate
the gassing rate of a gas, which can indicate whether the fault that generated the
gas is active or not [3]. The equation for computing gassing rate is as below [5]:
Ci2 Ci1 G
Ri ¼ ; ð6:1Þ
Dt q
where Ri is the gassing rate (ml/h or day), Ci1 the first sample concentration (ppm),
Ci2 the second sample concentration (ppm), G the total oil weight (ton), q the density
of oil (ton/m3) and Dt the actual operating time of a sampling interval (hour or day).
As a convenient basis for fault diagnosis, the gas ratio methods are coding schemes
that assign certain combinations of codes to specific fault types. The codes are
generated by calculating ratios of gas concentrations and comparing the ratios with
predefined values, which have been derived from experience and are modified
continually. A fault condition is detected when a gas combination fits the code
pattern of a particular fault. The most commonly used gas ratio method is the
Rogers ratio method [6] as listed in Table 6.1, which is able to distinguish more
types of thermal faults than that of the Dörnenberg ratio method [3].
Additional attention should be paid to the following conditions while applying
the gas ratio method according to Table 6.1:
1. Significant values quoted for ratio calculations should be only regarded as
typical.
2. Transformers fitted with an in-tank OLTC may indicate faults of code 202/102,
depending on seepage or transmission of arc decomposition products in the
diverter switch tank into the transformer tank oil.
100 6 Transformer Condition Assessment Using Dissolved Gas Analysis
Ratios of gases
\0.1 0 1 0
0.1–1 1 0 0
1–3 1 2 1
[3 2 2 2
Cases Characteristic fault Typical examples
0 No fault 0 0 0 Normal aging
1 Partial discharges of low 0 1 0 Discharges in gas-filled cavities resulting
energy density from incomplete impregnation, or
supersaturation or cavitation or high
humidity
2 Partial discharges of high 1 1 0 As above, but leading to tracking or
energy density perforation of solid insulation
3 Discharge of low energy 1 ? 2 0 1 ? 2 Continuous sparking in oil between bad
connections of different potential or
to floating potential. Breakdown of
oil between solid materials
4 Discharge of high energy 1 0 2 Discharges with power follow-through.
Arcing - breakdown of oil between
windings or coils or between coils to
earth. Selector breaking current
5 Thermal fault of low 0 0 1 General insulated conductor overheating
temperature \150°C
6 Thermal fault of low 0 2 0 Local overheating of the core due to
temperature range concentrations of flux. Increasing hot
150°C–300°C spot temperatures; varying from
7 Thermal fault of medium 0 2 1 small hot spots in core, shorting links
temperature range in core, overheating of copper due to
300°C–700°C eddy currents, bad contacts/joints
8 Thermal fault of high 0 2 2 (pyrolytic carbon formation) up to
temperature [700°C core and tank circulating currents
3. The ratios for a combination of multiple faults may not fit the predefined codes
in Table 6.1.
4. Combinations of ratios not included in Table 6.1 may occur in practice. A great
amount of consideration is being given to the interpretation of such combinations.
The majority of faults are slow to develop, which can be detected by DGA
monitoring. Locations of faults detectable using DGA as reported by CIGRE [4]
are listed below:
6.2 Fundamental of Dissolved Gas Analysis 101
1. Within a winding.
2. Cleats and leads.
3. In a tank.
4. A selector switch.
5. A core.
However, instantaneous faults are rapid and sometimes cannot be predicted by
DGA. Instantaneous failures that cannot be prevented by DGA are [4]:
1. Flash over with power follow-through.
2. Serious failures, developing within seconds to minutes and therefore not pos-
sible to be detected using DGA.
Table 6.2 Typical values of key gases for generation and transmission transformers
Key gas Key gas concentration (ppm) Suspect of indication
C2H2 [20 Power discharge
H2 [100 Partial discharge
P
CxHy Thermal fault
P
[1000 if up to C1, C2, C3-Hydrocarbons
P
[500 if up to C1,C2-Hydrocarbons
COx = 1,2 [10000 Cellulose degradation
electric charge carried by the oil-flow and so forth. To deal with the uncertainties
arising from fault diagnosis, based on gas contents extracted from transformer oil
samples, various techniques have been attempted by many researchers, including
EPS, FL, ANNs and so on. Firstly, fault types are classified based on on-site
experience, according to the combined criteria of total combustible gases, gas
generation rates, the key gas method and on-site inspections. Subsequently, vari-
ous CI methods are employed to reveal the relationships between fault symptoms
and malfunction types founded on gas-fault mapping schemes.
For instance, ANNs are the most widely used fault classifiers in DGA. In
[7, 8], an ANN was utilised to detect faults based merely on previous diagnostic
results. Theoretically, an ANN can be trained to represent any observable phe-
nomenon if there are sufficient data available. The experience obtained not only
encompasses the existing human diagnostic knowledge, but also explores the
unknown relationship between fault conditions and gas data. However, a disad-
vantage of ANNs in practice is that such relationships are embedded in an ANN
structure. It is not easy for a non-expert user to employ these relationships to
explain the conclusion of an inference. On the other hand, FL was employed to
improve the assessment capability of DGA in [9, 10], which can convert DGA
interpretation standards and other human expertise into ‘‘if-then’’ rules to form a
decision making system. Moreover, EPSs combined with other CI techniques, e.g.
fuzzy models and EAs, have been developed for DGA, which can evaluate
ongoing conditions and also suggest proper maintenance actions [11, 12]. The
fuzzy set concept can be integrated into an EPS to handle uncertain thresholds, gas
ratio boundaries and key gas analysis. EP has also been employed to automatically
modify the fuzzy ‘‘if–then’’ rules and simultaneously adjust the corresponding
membership functions. Other statistical methods like principal component analysis
and correlation analysis were proposed in [13], which are mainly concerned with
identifying the key variables and the key gas interdependence of a transformer
feature. In summary, these CI techniques could provide a firm heuristic basis for
future DGA research.
6.5 Summary
References
1. Heathcote MJ (1998) The J&P transformer book, 12th edn. First published by Johnson &
Phillips Ltd, Newnes imprint, UK
2. International Electrotechnical Commission (1978) IEC60559: interpretation of the analysis of
gases in transformers and other oil-filled electrical equipment in service. International
Electrotechnical Commission Standard, Geneva, Switzerland
3. The Institute of Electrical and Electronics Engineers (1994) Transformers Committee of the
IEEE Power Engineering Society, IEEE guide for the interpretation of gases generated in oil
immersed transformers, IEEE Std. C57.104–1991. The Institute of Electrical and Electronics
Engineers, Inc., New York
4. Mollmann A, Pahlavanpour B (1999) New guidelines for interpretation of dissolved gas
analysis in oil-filled transformers. Electra CIGRE France 186:30–51
5. Bureau of Standards for the P.R.China (1987) GB7252-87: Guide for the analysis and
diagnosis of gases dissolved in transformer oil, National Technical Committee 44 on
Transformer of Standardization Administration of China
6. Rogers RR (1978) IEEE and IEC codes to interpret incipient faults in transformers using gas
in oil analysis. IEEE Trans Electr Insul 13(5):348–354
7. Liu YL, Griffin PJ, Zhang Y, Ding X (1996) An artificial neural network approach to
transformer fault diagnosis. IEEE Trans Power Deliv 11(4):1838–1841
8. Griffin PJ, Wang ZY, Liu YL (1998) A combined ANN and expert system tool for
transformer fault diagnosis. IEEE Trans Power Deliv 13(4):1224–1229
9. Islam SM, Wu T, Ledwich G (2000) A novel fuzzy logic approach to transformer fault
diagnosis. IEEE Trans Dielectr Electr Insul 7(2):177–186
10. Yang HT, Liao CC, Chou JH (2001) Fuzzy learning vector quantization networks for power
transformer condition assessment. IEEE Trans Dielectr Electr Insul 8(1):143–149
11. Lin CE, Ling JM, Huang CL (1993) An expert system for transformer fault diagnosis and
maintenance using dissolved gas analysis. IEEE Trans Power Deliv 8(1):231–238
12. Huang YC, Yang HZ, Huang CL (1997) Developing a new transformer fault diagnosis
system through evolutionary fuzzy logic. IEEE Trans Power Deliv 12(2):761–767
13. Mori E et al (1999) Latest diagnostic methods of gas-in-oil analysis for oil-filled transformer
in Japan. In: Proceedings of 13th international conference on dielectric liquids (ICDL’99),
Nara, Japan, 20–25 July 1999