Renewable and Sustainable Energy Reviews 46 (2015) 201209

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

A review of monitoring methods for predictive maintenance of electric

power transformers based on dissolved gas analysis
Haroldo de Faria Jr.n, Joo Gabriel Spir Costa, Jose Luis Mejia Olivas
UFABC Federal University of ABC, CECS, Rua Santa Adlia, No. 166, Santo Andr, SP, 09.210-170 Brazil

art ic l e i nf o

a b s t r a c t

Article history:
Received 19 September 2014
Received in revised form
18 January 2015
Accepted 24 February 2015
Available online 17 March 2015

Electric power transformers are the link between the generators of a power system and the transmission
lines and between lines of different voltage levels. Power transformers undergo changes in their
operational life expectancy and reliability over the years. Currently, several tools for diagnosis and
assessment of their operational condition are available, including diagnostic techniques based on
dissolved gas analysis in the insulating oil. Through monitoring of dissolved gases in oil, it is possible to
perform detailed data analysis, seeking systemic failure prediction. The adoption of new technologies for
maintenance of power transformers can induce substantial changes in the reliability of such equipment
in view of the existence of a global trend to decrease operational costs, predict maintenances and control
substations in a centralized way. This paper describes the main factors that lead to lifetime reduction in
transformers and reviews the main methods used for predictive maintenance based on dissolved gas
analysis. The advantages and disadvantages of each one are outlined and some future directions for
research are proposed.
& 2015 Elsevier Ltd. All rights reserved.

Keywords:
Power transformers
Predictive maintenance
Dissolved gas analysis

Contents
1.
2.

3.

4.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Electric power transformers and maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
2.1.
Constructive aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
2.2.
Maintenance methods for electrical equipment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
2.2.1.
Corrective maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
2.2.2.
Preventive maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
2.2.3.
Predictive maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
2.2.4.
Proactive maintenance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Fault analysis in power transformers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
3.1.
Operational lifetime degradation factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
3.1.1.
Failures caused by the network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
3.1.2.
Failures caused by components, parts and pieces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
3.1.3.
Failures due to degradation of the insulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
Methods for diagnosis and assessment of the operational condition of power transformers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
4.1.
Dissolved Gas Analysis (DGA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
4.1.1.
IEC 60599 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
4.1.2.
Key Gas Method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
4.1.3.
Method of Duval . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
4.1.4.
Method of Doernenburg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

Corresponding author. Tel.: 55 11992558733.

E-mail address: haroldo.faria@ufabc.edu.br (H. de Faria Jr.).

http://dx.doi.org/10.1016/j.rser.2015.02.052
1364-0321/& 2015 Elsevier Ltd. All rights reserved.

202

H. de Faria Jr. et al. / Renewable and Sustainable Energy Reviews 46 (2015) 201209

4.1.5.
Method of Rogers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
5. Analysis of methods and future directions of research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208

1. Introduction
Electric power transformers are robust and efcient electric
equipment that play a fundamental role in supplying electric
energy at adequate voltage levels to consumers. Nevertheless,
power transformers undergo changes in their reliability and
operational lifetime over the years. This is mainly due to the
heavy loading of the equipment, driven by the need to achieve
increased prots and the related reluctance to invest in new
facilities by the power companies in a competitive market environment [1]. Due to improved monitoring and maintenance methods which emerged with technological advances, their lifespan
have increased.
Since power transformers have a high cost and are very
important to the availability of electrical power systems, several
tools for diagnosis and assessment of their operational condition
are available. Some diagnostic techniques can be based on the
analysis of dissolved gases in oil [26], on the monitoring of liquid
and solid insulation from the physicalchemical analysis of the
insulating oil and analysis of lifetime from the denition of the
degree of polymerization of insulating paper, among others [7].
Studies in the past decades have proved that the dissolved gases in
transformer oil are related closely to incipient faults [4]. If an
incipient failure of a transformer is detected before it leads to a
catastrophic failure, predictive maintenance can be deployed to
minimize the risk of failures and further prevent loss of services
[10]. Therefore, online monitoring and ofine testing are vital for
assessing power transformer conditions [11].
Methods of diagnosis of potential faults concealed inside power
transformers have attracted much research interest [5]. Dissolved
gas analysis (DGA) is a common practice for incipient fault
diagnosis and preventive maintenance of power transformers.
These methods test and sample the insulation oil of transformers
periodically to obtain the constituent gases in the oil due to
breakdown of the insulating materials inside the equipment
[612]. When there is any kind of fault, such as overheating or
discharge fault inside the transformer, it will produce a corresponding characteristic amount of gases in the transformer oil [9].
Through the analysis of the concentrations of dissolved gases, their
gassing rates, and the ratios of certain gases, the DGA method can
determine the fault type. As study results indicate, corona, overheating and arcing are the three main causes for insulation
degradation in power transformers [6]. In DGA, the fault related
gases commonly used are hydrogen (H2), methane (CH4), acetylene (C2H2), ethylene (C2H4), ethane (C2H6), carbon monoxide
(CO), and carbon dioxide (CO2). Therefore, if we forecast these
dissolved gases content in power transformer oil according to the
recent historical data, incipient failures of power transformer and
its development trend will be found out early, minimizing the
probability of a transformer loss [5].
It is possible to monitor various parameters of a transformer,
enabling early identication of failures, so that they can be treated
predictively. Various types of sensors can be installed on the
transformer to measure variables such as the temperature of the
oil and the windings, the dissolved gases and moisture content of
the oil, the capacitances and power factor of the bushings and the
contact wear on load tap changers. The data obtained by these

sensors can be analyzed and the results used to indicate if the

equipment is under some kind of fault or close to one. A risk
analysis can also be performed on the data to calculate important
indicators such as the probability of failure of a transformer
operating at a certain condition.
2. Electric power transformers and maintenance
2.1. Constructive aspects
Transformers applications are diverse and they are widely used
by the industry and distribution and transmission systems. In
these installations the transformer is commonly used to lower or
raise the system voltage level [13].
The lifetime of a transformer is mainly determined by its
insulation system, such as the type of material used and how it
was manufactured [14].
The insulation system is designed based on factors determined by the shape and characteristics of the active part (set
comprised of core and windings of a transformer) and the
gradients of temperature specied. Among the various materials
that make up the cooling system we can highlight the radiators
(the main component of this system), fans, pumps and insulating
oil, used as a refrigerant. The heat load to which the internal
components of transformers are exposed to is severe so that the
cooling system is essential for its proper functioning, emphasizing the fact that these systems vary according to the operational
context (conditions and environment) and the power capacity of
the transformer.
2.2. Maintenance methods for electrical equipment
Maintenance is considered a strategic activity that ensures
operation reliability of equipments and industrial processes. Maintenance should seek the intervention in equipment through a
strategy of reducing the intervention time, leaving the system
unavailable for the shortest time possible [7].
Among the various forms of maintenance we can highlight four
main types which are:

Corrective maintenance;
preventive maintenance;
predictive maintenance; and
proactive maintenance.

2.2.1. Corrective maintenance

Corrective maintenance xes aws and performance indexes
through system restoration. This form of maintenance can be
divided into two types, namely:
Unplanned corrective maintenance: failure is corrected randomly, without an intervention plan.
Planned corrective maintenance: correction of fault and/or
performance occurs in a planned manner, due to aw detection
through preventive and/or predictive maintenance, ensuring
reduced costs and implementation time.

H. de Faria Jr. et al. / Renewable and Sustainable Energy Reviews 46 (2015) 201209

2.2.2. Preventive maintenance

This type of maintenance seeks to increase the reliability and
availability of equipment by reducing failures, avoiding the need
for unplanned corrective maintenance. The objective of this type
of maintenance is the prevention of problems and failures, seeking
correction of faults before they occur.
Due to preventive action, the replacement of equipment parts
may occur prematurely. Nevertheless, equipment failures can
occur, since preventive maintenance should be performed at
specic time intervals [16].
The intervals between the interventions are based on the experience of the maintenance team, the managers of the company,
statistical studies or according to equipment manuals recommendations, while always considering the needs of the company.
Table 1 shows a comparison between the types of corrective and
preventive maintenance.

2.2.3. Predictive maintenance

Predictive maintenance means maintenance with a focus on
failure prediction, occurring through follow-ups with a specic
systematic on parameters and equipment conditions. This type of
maintenance did not emerge as a replacement for corrective and
preventive maintenance, but as an additional tool, which seeks to
minimize, through the monitoring of specic parameters, maintenance costs and losses in equipment.
Its main function is to collect data with the equipment under
operation, generating minimal interference in its operation.
Through the data collected, a diagnosis and trend analysis should
be done, identifying potential problems through historical analysis
of similar equipment and knowledge acquired over time.
Several variables can be monitored, such as density, ow, pressure,
vibration, temperature, voltage, current, electrical resistance, capacitance, inductance, among others.

203

3. Fault analysis in power transformers

Through the experience with the use of transformers, together
with data acquired over the years, it was realized that failures in
these equipment follow a pattern which is known as the Bathtub
Curve [19]. The bathtub curve can be divided into three main
periods, as shown in Fig. 1, which are as follows:
First period: represented by failures due to infant mortality of
equipment originated from errors in design or assembly. This
denes the period that includes all faults, before the stable
behavior of the curve.
Second period: represents the rate of continuous failures over
time, indicating failure rate between 1% and 2% per year.
Third Period: an increase in the failure rate occurs, caused by
aging of the equipment, which leads to degradation of the
insulation (solid and liquid).

The useful lifetime of a transformer is related to its use and to

the faults to which the equipment is subjected through time,
which accelerates aging, increasing the likelihood of failure and
reducing the period of stability of the product.

3.1. Operational lifetime degradation factors

It is possible to relate the main faults in power transformers
with their main causes, which are:
Failures caused by the network.
Failures caused by components, parts and pieces.
Failures due to degradation of the insulation.

2.2.4. Proactive maintenance

Proactive maintenance focuses primarily on determining the
root causes of machine failure and dealing with those issues before
problems occur. Anticipation is key in proactive maintenance
programs since it commissions corrective actions aimed at the
sources of failure. Proactive maintenance is geared to detect
contamination of uids and lubricants used in machines. In the
case of power transformers, strict monitoring of oil condition
would prevent the occurrence of problems that would require
component replacement (preventive maintenance) or monitoring
of established conditions for potential breakdown like vibration
and heat (predictive maintenance).

Fig. 1. Bathtub curve [19].

Table 1
Types of maintenance.
Maintenance
Preventive

Corrective

Routine

Inspection

Systemic

Selective

Planned

Performed with the

equipment in
operation.

Monitoring of the condition of the

equipment using human senses.

Based on the duration of

operation time calendar.

Performed after completion Performed after aw and/or failure

of operational lifetime.
of the equipment or component.

Nonplanned

204

H. de Faria Jr. et al. / Renewable and Sustainable Energy Reviews 46 (2015) 201209

3.1.1. Failures caused by the network

3.1.1.1. Mechanical deformation of the windings due to shortcircuits. Short-circuits in transformers result in high intensity
mechanical stresses in the active part of the equipment since the
resultant forces of these efforts are related to the current squared.
Two main types of mechanical deformations in the windings
due to the action of short-circuits can be identied:
Elastic deformation: reversible deformation not involving
structural changes in the active part.
Plastic deformation: irreversible deformation, which permanently changes the structure of the active part, generating the
need for intervention through corrective maintenance.
3.1.1.2. Defects originated from overvoltage. The increase of the
dielectric insulation request from the windings and/or
temperature rise is related to damage caused by overvoltage.
Overvoltages of short duration, which occur in periods of the
order of microseconds, are difcult to detect and its damages are
caused by internal electrical arcs. Long duration overvoltage result
in temperature increases.
The overvoltages in transformers can be divided into:

Temporary overvoltage.
Switching overvoltage.
Very fast transient overvoltage.
Atmospheric discharge overvoltage [15].

3.1.2. Failures caused by components, parts and pieces

3.1.2.1. Accessories and components. Failures in accessories and
components can cause damage only to these parts and pieces or
may cause serious damage to the transformer itself. Among the
failures that do not cause serious damage to the equipment are
functional problems such as improper operation of protection
systems, erroneous variable readings, small amounts of oil leakage,
among other faults that can be corrected by replacing parts and pieces.
Serious damage to the equipment can be caused due to the
explosion of bushings, switches and other failures that result in
substantial losses to the transformer. In this case, it is necessary to
put the equipment out of service for repairs in factory or eld [15].
3.1.2.2. Tap changers. Defects at On Load Tap Changers (OLTCs)
normally cause serious damage to transformers and its main
causes are:

Mechanical problems.
Wear of contacts.
Inadequate maintenance.
Deterioration of insulating oil of load keys.

Whereas defects at No Load Tap Changers (NLTCs) commonly

occur due to the following factors:
Maneuvers on energized transformers, generating electric arcs.
Incorrect maneuver, causing incorrect closing of the contacts so
that the transformer energization produces electrical arcs or
overheating.
Pressure loss on the springs responsible for tightening the
moving contacts, generating increased temperature which,
over the time, causes degradation of the contacts and could
generate electrical arcing [15].
3.1.2.3. Bushings. Loss of dielectric properties, degradation of seals
and aging of the insulating material are among the major causes of
internal faults in bushings. Among the external causes of failures

are occurrence of vandalism, contamination of porcelain and

damage by mechanical shock.
As a result of these failures catastrophic damages occur such as
explosions and res. These events can permanently damage the
transformer or even cause damage to equipment and people
nearby, such as when shards of porcelain are launched due to
explosions.
3.1.2.4. Connections. Among the components of transformers there
are several pieces and parts that are interconnected through
connections such as connectors and screws.
When faults occur in these components there is an increase in
electric current density, causing localized temperature increase,
which through time cause a deterioration of these parts of the
equipment and leads to failures.
3.1.3. Failures due to degradation of the insulation
A reduction in the capacity of the insulation system of a transformer as the equipment is subjected to faults and its operational lifetime
increases is normal. Depending on the severity of the aws, the
insulation capacity is reduced in a shorter time.
Among the many factors that cause degradation of the insulating material, highlights go to the action of high temperatures,
humidity in oil and insulating paper and the presence of oxygen in
the internal environment of the transformer [18].
An important point to note is the fact that the insulation
system of transformers has elements that end up collaborating
with the degradation of one another, so that the insulation ends up
destroying itself over time. This process is accelerated by failures
and overload of the equipment. An example of this interaction is
that of the insulating paper acting as a catalyst to the formation of
acids in the oil, which reacts with the paper, causing degradation
[7].

4. Methods for diagnosis and assessment of the operational

condition of power transformers
Utilities and maintenance companies have been using predictive maintenance techniques in order to identify any faults or
defects evolving in equipment. These methods measure different intrinsic characteristics of the insulation system such as
insulation degradation compounds and physical or chemical
parameters.
The main techniques available are listed below:
Physicalchemical analysis: characteristics of insulating oil are
evaluated such as dielectric strength, moisture content, color,
acidity, interfacial tension, among others.
Furfural: verication of furfuraldehyde (2FAL) content dissolved in the oil, with the aim of detecting the aging of
insulating paper.
Particle analysis: SUSPENDED particles in insulating oil are
identied, compared and classied according to size.
Vibration: sensors are attached to the transformer to measure
vibration during operation.
Thermographic inspection: checks the status of connections
and internal overheating detection by comparing the temperature of the hot spots, room temperature and normal operating
temperature.
Method of acoustic emission: through a set of high-frequency
acoustic sensors, one can locate and triangulate pulses to
identify and locate faults in transformers [20].

H. de Faria Jr. et al. / Renewable and Sustainable Energy Reviews 46 (2015) 201209

Dissolved Gas Analysis (DGA): this method can continuously

evaluate the condition of the transformer insulation system,
identifying possible failures.
4.1. Dissolved Gas Analysis (DGA)
With the DGA method it is possible to evaluate continuously the
operating condition of the transformer, identifying potential aws and
causes to the formation of gases through the use of various criteria.
Among the gases dissolved in oil, combustible and noncombustible gases can be found, as described in Table 2.
The distribution of gases takes place according to the transformer
failure and the components involved therein. These gases can also be
classied according to the type of failure which originated them and
the material involved in the process [7] as indicated in Table 3.
These gases can be identied by the gas chromatography
technique, which performs a physicalchemical analysis through
separation of chemical compounds [21].
There are several methods for failure diagnosis of transformers
through identication of dissolved gases. The most used are listed
below:
IEC 60599 Method;
Key Gas Method;

Table 2
Dissolved gases in insulating oil.
Gases
Combustile

Non-combustile

Carbon monoxide (CO) Oxygen (O2)

Hidrogen (H2)
Methane (CH4)
Nitrogen (N2)
Ethane (C2H6)
Ethylene(C2H4)
Carbon dioxide (CO2)
Acetylene (C2H2)

Table 3
Classication of dissolved gases in accordance with
the type of failure and material involved.
Corona
Oil
H2
Cellulose
H2, CO, CO2
Pyrolysis
Oil
Low temperatures
CH4, C2H6
High temperatures
C2H4 H2 (CH4, C2H6)
Cellulose
Low temperatures
CO2 (CO)
High temperatures
Electric arc
H2, C2H2 (CH4, C2H6, C2H4)

205

Method of Duval;
Method of Doernenburg; and
Method of Rogers.
All methods are empirical and their results are based on
correlations between the gases found in gas analysis. The ratios
used in these methods are listed below.
R1:
R2:
R3:
R4:
R5:

(CH4/H2);
(C2H2/C2H4);
(C2H2/CH4);
(C2H6/C2H2); and
(C2H4/C2H6).

4.1.1. IEC 60599 Method

The IEC 60599 Method classies and identies transformer
failures according to the relationships shown in Table 4.
Two important remarks may be made about the method.
There are variations in the ratios used in some countries and the
C2H2/C2H6 ratio can be used to replace the CH4/H2 ratio. These
ratios are to be observed only in cases where one of the
dissolved gases has a high concentration value and/or a high
growth rate [7].
4.1.2. Key Gas Method
In this method, the aws are associated with the gas composition prole as shown in Table 5. Key gases are dened in the IEEE
guide as gases generated in oil-lled transformers that can be
used for qualitative determination of fault types, based on which
gases are typical or predominant at various temperatures.
If historical data of dissolved gases for diagnosis is not available,
risks present in the equipment can be identied and classied
according to the criteria depicted in Table 6. Condition 1 represents
normal operation. Condition 2 indicates that the equipment may
be operating under fault with total gases above normal values.
Condition 3 indicates an elevated level of decomposition. Condition 4 indicates excessive degeneration and continuous operation
can result in failure.
4.1.3. Method of Duval
This method interprets dissolved gas data through the use of a
triangle of relative percentages of CH4, C2H2 and C2H4, where each
vertex considers that 100% of the analyzed gases are comprised of
one of these compounds [17].
4.1.4. Method of Doernenburg
In this method, relationships R1R4 are used and a signicant
amount of gas is needed to validate its use. According to the
method, there are three main types of failures, which are:
Thermal decomposition;
low intensity partial discharges (corona); and
partial discharges of high intensity (electric arc).

Table 4
Interpretation of data obtained by DGA and classication of aw types according to IEC 60599.
Abreviation

Description

C2H2/C2H4

CH4/H2

C2H4/C2H6

PD
D1
D2
T1
T2
T3

Partial discharges
Low energy discharges
High energy discharges
Thermal aw, T o 300 1C
Thermal aw, 300 1C o To 700 1C
Thermal aw, T o 700 1C

Non-signicant value
41.0
0.62.5
Non-signicant value
o 0.1
o 0.2

o 0.1
0.10.5
0.11.0
Non-signicant value
4 1.0
4 1.0

o 0.2
4 1.0
4 2.0
o 1.0
1.04.0
4 4.0

206

H. de Faria Jr. et al. / Renewable and Sustainable Energy Reviews 46 (2015) 201209

Table 5
Gas composition prole for the Key Gas Method.
Key Gas Method
Key gas

Fault type

Typical proportions of generated combustile gases

H2 e C2H2

Electrical high energy

(arcing)
Electrical low energy partial
discharge
Thermal oil
Thermal oil and cellulose
Electrolysis

High quantities of H2 e C2H2 are produced. Smaller proportions of CH4 and C2H4. CO2 e CO can be formed if cellulose is
involved. Oil may be carbonized.
Mainly H2. Small quantities of CH4. Traces of C2H4 and ethylene.

H2
C2H4
CO
H2

Mainly C2H4. Smaller proportions of C2H6, CH4 and H2. Traces of C2H2 at very high fault temperatures.
Mainly CO. Hydrocarbon gases such as CH4 and C2H4 if fault involves structures in oil.
Mainly H2.

Table 6
Risk diagnosis in transformers in accordance with the concentrations of dissolved gases in ppm (part per million).
Conditions

H2

CH4

C2H2

C2H4

C2H6

CO

CO2

Total gas

Condition
Condition
Condition
Condition

100
101700
7011800
4 1800

120
121400
4011000
4 1000

1
29
1035
435

50
51100
101200
4200

65
66100
101150
4150

350
351570
5711400
41400

2500
25004000
400110000
4 10000

720
7211920
19214630
44630

1
2
3
4

No
Yes

Yes

H2CH4 C2H2
C2H4 > 2L1

Gas
input

Normal
condition

No

C2H6 CO
> L1

No

Relation test
OK?

Relations
analysis not
applicable
Resample

Yes
R1 = CH4 / H2
No

R2 = C2H2 / C2H4

Yes
R1 < 0.1

Yes
R3 < 0.3

Yes
R4 > 0.4

R3 = C2H2 / CH4
R4 = C2H6 / C2H2
No

No

R3 > 0.3

R2 > 0.75

0.1 < R1 < 1

Non identified fault

Resample

No
Yes

Yes

Yes

Yes
Electric arc
discharge

R4 < 0.4

No
No
Yes

No
Yes

R1 > 1

R2 < 0.75

Non identified fault

Resample

No
Yes

R3 < 0.3

Partial discharges
Radio interference
voltage (RIV)

Yes
R4 > 0.4

Thermal fault

Fig. 2. Gas relations analysis according to the method of Doernenburg.

Table 7
Limit values for gases dissolved in oil.
Key gas

Concentrations
(ppm)

Hydrogen (H2)
100
Methane (CH4)
120
Carbon monoxide (CO) 350
Acetylene (C2H2)
35
Ethylene (C2H4)
50
Ethane (C2H6)
65

Fig. 2 shows the step by step analysis of relationships by the

Method of Doernenburg.
Initially, the values of gases obtained by DGA are compared to
threshold values, thus identifying the existence of problems with
the equipment and if the concentration of gases is sufcient for
the application of the method [17]. Limit values for the ratio of
dissolved gases in oil can be found in Table 7.
The same method can be used to analyze free gases; however,
the limits should follow a different correlation. In Table 8 one can
identify the threshold values for the correlations for dissolved
gases and free gases.

H. de Faria Jr. et al. / Renewable and Sustainable Energy Reviews 46 (2015) 201209

207

Table 8
Relationships for dissolved gas in oil and free gas suggested by the method of Doernenburg.
Suggested diagnosis for fault

Relation 1 (R1) CH4/H2 gas

origin

Relation 2 (R2) C2H2/C2H4

gas origin

Relation 3 (R3) C2H2/CH4 gas Relation 4 (R4) C2H6/C2H2

origin
gas origin

Dissolved in oil

Free gas

Dissolved in oil

Free gas

Dissolved in oil

Thermal decomposition

41.0

40.1

o 0.75

o 1.0

Corona (low intensity partial discharges)

o 0.1

o0.01

Non-signicant

Electric arc (high intensity partial discharges) 40.1

o 1.0

40.01
o0.1

40.75

41.0

Yes
Gas
input

R2 < 1

No

Dissolved in oil

Free gas

o 0.3

o 0.1

4 0.4

40.2

o 0.3

o 0.1

4 0.4

40.2

4 0.3

40.1

o 0.4

o 0.2

Yes
0.1 < R1 < 1

Free gas

Yes
R5 < 1

Normal
condition

No

No

R1 = CH4 / H2

Yes

R2 = C2H2 / C2H4

Low temperature
thermal fault
Overload

1 < R5 < 3

R5 = C2H2 / C2H6
Yes

Yes
R1 > 1

Thermal fault
< 7000 C

1 < R5 < 3
No
Yes

Thermal fault
> 7000 C

R5 > 3

Yes

Yes

Yes
R1 < 1

R2 < 1

Partial discharges
Radio interference
voltage (RIV)

R5 < 1

No
Yes

Yes
1 < R2 < 3

Yes
0.1 < R1 < 1

R5 > 3

High energy
electric arc

Fig. 3. Gas relations analysis according to the method of Rogers.

Table 9
Relationships for dissolved gas in oil and free gas and failure diagnosis suggested by
the Method of Rogers.
Case Relation 2
Relation 1
(R2) C2H2/C2H4 (R1) CH4/H2
0

o 0.1

Relation 5
Failure diagnosis
(R5) C2H4/C2H6 suggested
o1.0

Normal unit

o 0.1

40.1
o1.0
o0.1

o1.0

1.03.0

0.11.0

43.0

o 0.1

1.03.0

o 0.1

40.1
o1.0
41.0

1.03.0

o 0.1

41.0

43.0

Partial discharge low

energy density arc
Electric arc high
energy discharge
Low temperature
thermal failure
Thermal failure
o 700 1C
Thermal failure
4700 1C

4.1.5. Method of Rogers

This method follows the same procedure as the Method of
Doernenburg, however, only three ratios are used and there is no
dependence on the gas concentration level for validation of the
method [17].
Fig. 3 shows the step by step analysis of gas ratios by the
Method of Rogers.
The ratios for dissolved gases in the oil and free gases, together
with suggested failure diagnosis according to the Method of
Rogers, are shown in Table 9.

5. Analysis of methods and future directions of research

The ve DGA methods presented in Section 4 are used to
analyze and interpret the signicance of gases present in oilimmersed power transformers.

208

H. de Faria Jr. et al. / Renewable and Sustainable Energy Reviews 46 (2015) 201209

Another class of methods can be used to forecast the future gas

contents of a power transformer according to historical data.
Neural network models have been extensively used, but these
rely on a large amount of historical data while the data in practice
may not be available. Back propagation neural network is widely
used for its simplicity, but does not present a high correct
diagnosis rate. On the other hand, when the training samples
and input information dimension numbers are large, the training
time is long [6]. In this regard, the work presented in [5] develops
a rough set, fuzzy wavelet neural network and least squares
weighted fusion algorithm which simplies the input dimension
and training sample number, so that the network structure is
simplied, and the training time is improved. Under the circumstances of limited samples, several models have been proposed [4].
Many effective attempts have been developed such as fuzzy linear
regression model, grey model (GM), grey-extension model and
support vector machine (SVM) model. The least squares support
vector machine (LS-SVM) is introduced as reformulations to
standard SVM which simplies the model of standard SVM in a
great extent by applying linear least squares criteria to the loss
function instead of the traditional quadratic programming
method. LS-SVM used in conjunction with optimization methods
such as genetic algorithms and particle swarm optimization to
determine optimum free parameters of support vector machines
has achieved promising results.
Since forecasting methods are not the main focus of this work,
this section presents a comparison between the ve analysis
methods used for predictive maintenance based on dissolved gas
analysis and points out some future directions of research.
Comparisons were based on a literature review [2,3,16,18,2027],
and aspects considered important for the analysis of trends in the
electric power industry were highlighted.
One point of agreement among different authors is the fact that
the methods of dissolved gas analysis present serious limitations
when used separately for the analysis of incipient failures in power
transformers. The main limitations can be highlighted:
The criteria listed in the methods are not able to cover all
situations and the implementation of more than one method is
sometimes necessary.
The same concentration of gases analyzed with different
methods may lead to different results.
Presentation of erroneous results, i.e. when the gas concentration is very small, the methods can indicate failures in equipment that are in normal condition.
These methods are easily implemented and present good
results for analysis after the occurrence of faults, however, are
less sensitive in detecting faults in a predictive manner.
A good fault analysis in transformers using a dissolved gas
analysis depends a lot on the accuracy and reliability of the
measured concentrations of gases.
An important point is the fact that each method of gas analysis
has its own advantages and limitations, not always leading to the
same results [27]. The accuracy of the analysis depends intrinsically on the specialist responsible for reading the data and
showing the results.
During the analysis, it is necessary to take into account factors
such as differences in transformers, maintenance conditions,
operation, voltage levels, oil volume, among others, making it
difcult to obtain a tool with 100% accuracy and reliability.
A successful methodology was developed in [20] using a
sequential combination of methods to initially determine if the
equipment is in a state of normality or failure. If a failure was
detected, more than one method was applied to identify the type
of anomaly. This combination of methods enabled the suppression

of individual weaknesses of each methodology by the other

methods, providing a very high level of accuracy.
There is a trend in the adoption of computational methods for
analysis of failures in transformers. They must be used together or
based on standard methodologies in order to obtain the highest
possible degree of reliability and accuracy.
Risk analysis tools are currently being used in a wide range of
industry applications. A monitoring system for predictive maintenance of power transformers can also be built to reap the
benets of risk analysis. Through the employment of sensors, the
gases dissolved in transformer oil can be constantly analyzed using
a gas chromatography system. An extension of the main tank can
be used to sample the circulating oil. The state of operation of the
equipment can be detected using a combination of methods, as
described in [20]. The information obtained from the equipment
current operating state can be compared with information from a
reliable database, indicating the state in which the equipment is
operating, with the risk of failure. This type of analysis would
allow the removal of the equipment for maintenance before the
occurrence of serious shortages, increasing the reliability of the
power grid. However, as highlighted in [3], even the use of
software for DGA analysis using a combination of methods may
not have a high accuracy and sometimes fail to distinguish some
types of faults. It is recommended that the fault diagnosis be
validated by an expert, avoiding with this any mistake in the
diagnosis, giving a better credibility to the results.
The work presented in [2] points out that a diagnosis totally
dependent on an expert may leave unnoticed some types of faults
that are in development. A solution to that could be the use of
different types of tests and diagnosis like DGA, frequency response
and thermal analysis. However, this solution may be difcult to
implement as an online monitoring tool.

6. Conclusions
This paper presented a review of predictive maintenance
methods and current methodologies for fault diagnosis, seeking
to present the key concepts related to transformer maintenance
with the use of dissolved gas analysis in insulating oil.
There is a trend toward the adoption of computational techniques
as a means of implementing and/or combining methods of gas
analysis such as Rogers, Doernenburg, Duval and Key Gas. These
implementations and combinations seek to increase the reliability of
the analysis, since these methods alone are not 100% reliable. A
sequential combination of methods is indicated to rst analyze
whether the equipment is in a state of normality or abnormality.
Abnormal operation can be studied with more than one technique
together to identify the type of anomaly that is occurring.
The main challenges found in the literature review are related
to the need for condence in the data obtained from the gas
analysis. All methods investigated also present difculty in obtaining a high level of reliability and accuracy when subjected to
analysis of low amounts of dissolved gases.
A state of the art predictive maintenance tool should employ a
real time DGA monitoring system that uses a set of DGA methods
for the analysis of gases and compares measured data with historic
database to perform risk analysis. A prediction failure indicator
could use results from Monte Carlo simulation, allowing the
shutdown of the equipment before it is committed to signicant
damage.
References
[1] Fu W, McCalley JD, Vittal V. Risk assessment for transformer loading. IEEE
Trans Power Syst 2001;16:34653.

H. de Faria Jr. et al. / Renewable and Sustainable Energy Reviews 46 (2015) 201209

[2] Sun HC, Huang YC, Huang CM. A review of dissolved gas analysis in power
transformers. Energy Proc 2012;14:12205.
[3] Bakar NA, Abu-Siad A, Islam S. A review of dissolved gas analysis measurement
and interpretation techniques. Electr Insul Mag 2014;30(3):3949.
[4] Liao R, Zheng H, Grzybowski S. Particle swarm optimization least-squares
support vector regression based forecasting model on dissolved gases in oillled power transformers. Electr Power Syst Res 2011;81:207480.
[5] Dong L, Xiao D, Liang Y. Rough set and fuzzy wavelet neural network
integrated with least square weighted fusion algorithm based fault diagnosis
research for power transformers. Electr Power Syst Res 2008:12936.
[6] Hung CP, Wang MH. Diagnosis of incipient faults in power transformers using
CMAC neural network approach. Electr Power Syst Res 2004;71:23544.
[7] Ostermann R. Development of a management system for power transformers
installed in the transmission system of a state power company. [dissertation].
Porto Alegre (RS): Catholic University of Rio Grande do Sul; 2006.
[8] Chen W, Pan Ch, Yun Y. Wavelet networks in power transformers diagnosis
using dissolved gas analysis. IEEE Trans Power Deliv 2009;24:18794.
[9] Naresh R, Sharma V, Vashisth M. An integrated neural fuzzy approach for fault
diagnosis of transformers. IEEE Trans Power Deliv 2008;23:201724.
[10] Tang WH, Goulermas JY, Wu QH. A probabilistic classier for transformer
dissolved gas analysis with a particle swarm optimizer. IEEE Trans Power Deliv
2008;23:7519.
[11] Hao X, Cai-xin S. Articial immune network classication algorithm for fault
diagnosis of power transformer. IEEE Trans Power Deliv 2007;22:9305.
[12] Morais DR, Rolim JG. A hybrid tool for detection of incipient faults in
transformers based on the dissolved gas analysis of insulating oil. IEEE Trans
Power Deliv 2006;21:67380.
[13] Geromel LH. Application of intelligent systems to power transformer projects.
[dissertation]. Campinas (SP): State University of Campinas; 2003.
[14] Nogueira DS, Alves DP. Power transformers theory and essential application
topics. [dissertation]. Rio de Janeiro (RJ): Federal University of Rio de Janeiro;
2009.
[15] Bechara R. Fault analysis of power transformers. [dissertation]. Sao Paulo (SP):
So Paulo University Polytechnic School; 2010.
[16] Possi M. Maintenance and operation. Electr Sect Mag 2011;66:3240.

209

[17] Landucci Jr. A. Pre-corrective power transformers maintenance: a new maintenance concept. [dissertation]. Sao Carlos (SP): So Carlos Engineering
School; 2009.
[18] Zirbes R. Methodologies for evaluation and diagnosis of paper impregnated
with mineral oil insulation. [dissertation]. Florianpolis (SC): Federal University of Santa Catarina; 2003.
[19] Precision Micro Drives Tech Blog [internet]. London: Doug Roberts. Available from:
http://www.precisionmicrodrives.com/tech-blog/2012/08/02/lifetime-of-dc-vibra
tion-motors-mttf-t-ab-019-released; 2012 August [cited 09.05.14].
[20] Lupi Filho G. Comparison of diagnostic criteria for chromatographic analysis of
dissolved gas in power transformer insulating oil. [dissertation]. So Carlos
(SP): So Carlos Engineering School; 2012.
[21] Leopoldino R. Studies on dissolved gas in insulating vegetable oil Inuence
of electric arc and overheating. [dissertation]. So Carlos (SP): So Carlos
Engineering School; 2011.
[22] Amora MAB, Braga APS, Almeida OM, Barbosa FR, Pontes RS. Extracted rules
from a high dimension input space by attribute addition for obtaining a new
method of analysis of incipient faults in transformers. Brazilian Intelligent
Automation Symposium. Brazil; 2013.
[23] Arantes JG. Power transformers diagnosis by dissolved gas analysis in
insulating oil using neural networks. [dissertation]. Itajub (MG): Federal
University of Itajub; 2005.
[24] Carita AJQ, Leite LC, Godoy RB,A.P.P.de M. Jr. Power transformer diagnosis
using Bayesian network. In: Proceedings of the XIX Brazilian automatic
congress. Brazil; 2012.
[25] Morais DR. Intelligent tool for detecting incipient faults in transformers based
on dissolved gas analysis in insulating oil. [dissertation]. Florianpolis (SC):
Federal University of Santa Catarina; 2004.
[26] Neto AD, Assuno TCBN, Assuno JT. Classication of power transformers
based on analysis of dissolved gases in the insulating oil. XIV ITA Scientic
Meeting. Brazil; 2008.
[27] Kan H, Miyamoto T. Proposals for an improvement in transformer diagnosis
using dissolved gas analysis (DGA). Australasian Universities Power Engineering Conference; 2004.