Experimental Investigations On The Dissolved Gas Analysis Method (DGA) Through Simulation of Electrical and Thermal Faults in Transformer Oil
Experimental Investigations On The Dissolved Gas Analysis Method (DGA) Through Simulation of Electrical and Thermal Faults in Transformer Oil
Experimental Investigations On The Dissolved Gas Analysis Method (DGA) Through Simulation of Electrical and Thermal Faults in Transformer Oil
Dissertation
vorgelegt von
2014
Approved by the examining committee on: 27th June, 2014
I declare that this dissertation represent my own work, except where due ac-
knowledgement is made.
———————-
Jackelyn Aragón Gómez
“Reach high, for stars lie hidden in you. Dream deep, for every
dream precedes the goal.”
― Rabindranath Tagore
Acknowledgment
I am greatly thankful of Prof. Dr. Oliver Schmitz, Head Director of the Insti-
tute of the Applied Analytical Chemistry and Prof. Dr.-Ing. Holger Hirsch,
Head Director of the Institute of Electrical Power Transmission at the Uni-
versity of Duisburg-Essen, for their willingness to review and examine my
dissertation in spite of the time constraints and giving worth to my efforts
of years. I deeply appreciate their professional and constructive approach
towards my research work. I would like to mention very special thanks to
Prof. Dr. Mathias Ulbricht, Head Director of Institute of Technical Chemistry
for his encouragement, trust and support.
I express my deep gratitude and love for my mother, father, relatives and
friends, who have wished me the strength, health and courage to accomplish
the work. They have been always there to express their love and support
regardless of the distance and time.
Last but not least, very loving thanks to my wonderful son Arjun for the mo-
tivation and energy he provided me in the difficult times with his cheerful
smile. He is the most precious gift in my life.
List of Figures ix
Abstract xv
Kurzfassung xviii
vii
3.3.6 Mtronix partial discharge (PD) measurement system . . 50
3.3.7 Transformer oils used for the investigations . . . . . . . . 51
References 100
List of Figures
2.1 Core of a step-down transformer . . . . . . . . . . . . . . . . . . . 7
2.2 Equivalent circuit of an ideal transformer . . . . . . . . . . . . . . 8
2.3 Schematic representation of a high voltage power transformer . 9
2.4 Temperature dependent evolution of gases in oil [Singh et al.,2010] 12
2.5 Effect of refining process on properties of oil [Heathcote, 1998] . 14
2.6 Basic structures of hydrocarbons in transformer oil [Nynas Naph-
thenics, 2004] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.7 Typical oil molecule [Nynas Naphthenics, 2004] . . . . . . . . . . 17
2.8 Oxidation process of a hydrocarbon . . . . . . . . . . . . . . . . . 20
2.9 Decomposition of alkanes with respect to temperature[Shirai
et. al., 1997] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.10 Factors influencing the gassing process of oil . . . . . . . . . . . 24
2.11 Development of electrical breakdown of oil during fault gas
generation [Schmidt, 1982] . . . . . . . . . . . . . . . . . . . . . . . 25
2.12 Temperature dependent solubility of gases in oil [Griffin, 1988] 28
2.13 One dimensional diffusion tube . . . . . . . . . . . . . . . . . . . 31
2.14 Gas generation pattern method . . . . . . . . . . . . . . . . . . . . 34
3.1 High voltage suppply circuit for partial discharges in the oil
tank (12 liters) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Disk electrode arrangement . . . . . . . . . . . . . . . . . . . . . . 38
3.3 Hotspot setup for gas-in-oil analysis . . . . . . . . . . . . . . . . . 38
3.4 High voltage supply circuit for electric discharges in the oil
tank (600 liter) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.5 Schematic view of oil circulation system for the large setup . . . 41
3.6 Volumetric flow of a fault gases in the oil balance of the large
scale setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.7 Interaction scheme of the automation system for the large scale
setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.8 Measurement principle of Kelman Transport X,[Kelman Trans-
port X Infoblat, 2009] . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.9 Description of vacuum degassing system [ECH Elektrochemie
Halle GmbH] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.10 Shaking syringe used for manual extraction of gas-in-oil . . . . 47
3.11 Measurement principle of transformer gas monitoring system
(TGM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
ix
3.12 Partial discharge control and measurement system for the small
test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.13 Phase resolved partial discharge diagram for a PD event (PRPD) 51
xiii
4.11 Gas-in-oil concentration after oil reconditioning and air satu-
ration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.12 Total energy dissipated during simulated arcing discharge (AD)
faults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.13 Fault diagnostic by applying the CIGRE fault interpretation
scheme (arcing discharge) . . . . . . . . . . . . . . . . . . . . . . . 81
4.14 Oil pump rate for arcing discharge experiments . . . . . . . . . . 82
4.15 Overall mass transfer coefficients (KOL(i) , [m/s]) of fault gases . 86
Abstract
Gas generation in transformer oil is induced by electrical and thermal faults
resulting from unfavorable operating conditions in transformers. Along with
aged conditions of transformers, operating factors such as high temperature,
strong electrical fields, electrical discharges, mechanical stresses, insulation
damage and contaminants pose imminent risks of malfunctioning and ir-
reversible damage to the transformers. Transformer monitoring methods
based on dissolved gas analysis (DGA) have gained great significance and
attention in order to ensure timely and accurate diagnostics of the electrical
and thermal faults occurring in the transformers. Gases that act as fault indi-
cators are hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide
and dioxide.
The dissolved gas analysis (DGA) has been widely acknowledged as an ef-
fective and rather simple method for fault diagnostic of transformers. How-
ever the diagnostic of the faults by DGA directly depends on the knowledge
about gas generation patterns produced by various types of faults. More-
over, the reliability of the diagnostic depends considerably on the technics of
gas-in-oil extraction and analysis as well as the procedures for oil sampling
and storage.
xv
Abstract
The small scale setup was used for simulating the partial discharges (PD)
fault and hotspots (HS) faults of different temperatures in transformer oil.
The setup allowed the investigations of the fault gas generation process by
means of various commercially available DGA monitoring techniques.
The large scale experimental setup, which includes oil tank, conservator tank
as well as oil circulation and reconditioning system, provided an enhanced
model of an air-breathing power transformer. It was also equipped with an
automated system to control valves and pumping rates of oil circulation and
reconditioning system, as it occurs in a power transformer for cooling pur-
pose. The large scale setup was employed for simulating intense arcing dis-
charge (AD) faults at high voltage levels.
The gas concentrations obtained for the three types of faults (PD, HS and AD)
were interpreted using the fault interpretation scheme CIGRE TF 15.01.01
(CIGRE scheme), which is one of the latest DGA interpretation schemes. The
interpretations suggest that the CIGRE scheme provides inconsistent diag-
nostic of PD and HS faults, otherwise it seemed to be consistent in diagnos-
ing AD faults. Based on the overall results it was concluded that the CIGRE
scheme should be applied cautiously and additional factors must be consid-
ered for a reliable diagnostic. The graphical DGA interpretation method, so
called gas generation pattern method, resulted to be sufficiently reliable at
the diagnostics of PD, HS as well as AD faults. This method exhibits great
potential to be utilized for fault diagnostics as a separate method or in com-
xvi
Abstract
The large scale setup was also used for investigating the continuous diffusion
of fault gases from oil into the atmosphere via the air-breathing conservator
tank. The investigations confirmed that the diffusion process has a strong
influence on the diagnostics of faults using DGA. The diffusion of gases leads
to an underestimation gas-in-oil concentrations and consequently incorrect
fault diagnostics. The diffusion flux of gases tends to increase at higher oil
circulation rates, especially for highly volatile fault gases, such as hydrogen.
xvii
Kurzfassung
Die Gasentstehung in Transformatorölen wird durch elektrische und ther-
mische Fehler verursacht, welche zu ungünstigen Betriebsbedingungen in
Transformatoren führen. Mit dem Alterungsprozess des Transformators geh-
en ungünstige Betriebsfaktoren wie erhöhte Temperaturen, starke elektrische
Felder, elektrische Entladungen, mechanischer Stress und Schäden der Iso-
lation einher, welche zur Verstärkung jener fehlerhaften Betriebsbedingun-
gen und dementsprechend zu stärkerer Gasentstehung führen und somit das
akute Risiko für irreversible Schäden oder eine Explosion erhöhen. überwach-
ungs- und Diagnosemethoden, die auf der Dissolved Gas Analysis (DGA)
basieren, haben aufgrund ihres Einsatzpotentials zur kontinuierlichen und
zuverlässigen Erkennung von Transformatorenfehlern, große Bedeutung und
Aufmerksamkeit gewonnen. Einige der im Transformatoröl gelösten Gase
können als Wasserstoff, Methan, Ethan, Ethylene, Acetylen, Kohlenmonoxid
und Kohlendioxid identifiziert werden.
Die Dissolved Gas Analsysis (DGA) ist als effektive und einfache Methode
zur Fehlerdiagnose in Transformatoren anerkannt. Allerdings beruht die
Fehlerdiagnose der DGA auf dem Wissen über die Zuordung der einzelnen
Gasmuster zu den verschiedenen Fehlern in Transformatoren. Darüber hin-
aus hängt die Zuverlässigkeit der Fehlerdiagnose beträchtlich von der Tech-
nik der Gas-in-Öl Extraktion und Analyse, sowie der Probenentnahme und
Probenlagerung ab. Die durchgeführten experimentellen Untersuchungen
hatten das Ziel zu verstehen, wie die elektrischen und thermischen Fehler die
Gasentstehung verursachen und die herkömmlichen Gas-in-Öl Messungen
einzuschätzen, um die Anwendungen der DGA zu verbessern. Zu diesem
Zweck wurden Experimente durchgeführt, in denen elektrische und ther-
mische Fehler in Transformatoren simuliert worden sind. Dabei waren die
Versuchsaufbauten mit verschiedenen DGA-überwachungstechniken ausges-
tattet, so dass die Gasmuster kontinuierlich verifiziert werden konnten. Zu-
dem wurde untersucht welchen Effekt die Probenentnahme und die Art der
Durchführung der Gas-in-Öl Messung sowie der Diffusionsfluss der entste-
henden Gase auf die DGA-Ergebnisse hatten.
xviii
Kurzfassung
Ein großer Teil dieser Arbeit befasste sich mit dem Aufbau von experimente-
llen Versuchsanordnungen, um thermische und elektrische Fehler wie Teilent-
ladungen, Durchschläge und Hotspots zu simulieren. Dafür wurde ein kleiner
maßstabgerechter Versuchsaufbau (30 kV und 12 Liter Öl) und ein großer
maßstabsgerechter Versuchsaufbau (100 kV und 600 Liter Öl) konstruiert.
Diese beinhalteten einen Öltank sowie ein Ausdehnungsgefäß, einen Ölkreis-
lauf, Hochspannung und ein Steuerungssystem, die einem Leistungstrans-
formator nachempfunden waren. Mittels dieser Versuchsaufbauten war es
möglich, unter kontrollierten Parametern wie Stromstärke, Spannung, Tem-
peratur und Ölbedingungen, die elektrischen und thermischen Fehler im Öl
zu erzeugen.
xix
Kurzfassung
xx
1 Background and Objectives of
the Research
1.1 Description of the problem
Power technology has been passing through great challenges in the attempt
to cope with the ever-growing world energy demand. Nonetheless, the tech-
nology of oil-filled transformers has remained almost unchanged for about
a century. At present there is a large fleet of transformers exposed to differ-
ent cumulative stresses which eventually are approaching the end of their
expected lifespan. Consequently, many of these transformers would need to
be replaced or repaired to safeguard power supply. However many power
agencies are shifting their attention to monitoring and diagnostic methods,
that can provide a reliable assessment of the transformer condition. In that
trend, online monitoring techniques based on the dissolved gas analysis (DGA)
have gained great importance as a continuous and non-invasive method, but
still there are cost and technical limitations of these monitoring techniques,
regarding their accuracy at diagnosing faults in transformers.
The actual average age of transformer fleets in developed countries have sur-
passed 30 years, which was the limiting life-span designed for these trans-
formers in order to operate safely and reliably [TD-World, 2007]. Moreover
the deregulation of the energy market has forced power agencies to avoide
investments on new transformers and to utilize their currently installed elec-
trical components at their full capacity, over longer distances and periods
in order to minimize service costs. Thus, an increased concern about the
current condition assessment of transformers has arisen implementation of
maintenance strategies and condition monitoring techniques. An effective
detection of the faulty conditions in transformers through condition strate-
gies can prolong their technical lifespan and prevent costly breakdowns [Ny-
nas Naphthenics, 2004].
The transformers operating under the present power loads are highly prone
to develop the type of faults that can lead to high outage costs as well as fire
or explosions. Unfavorable operating conditions due to high temperature,
strong electrical fields, electrical discharges, mechanical stresses, insulation
1
1 Background and Objectives of the Research
Until the date, the combination of transformer oil and cellulose paper has
remained the most effective insulating system due to their excellent dielec-
tric strength and relatively low cost. However, the transformer oil undergoes
the oxidation process due to free radical reactions between unstable hydro-
carbon molecules and oxygen. This reactions are catalyzed in presence of
copper and iron in transformer, and further accelerated due to the heat dis-
sipated from windings and the core. The oxidation of the oil leads to loss of
its dielectric strength and generate wide range of solid, liquid and gaseous
compounds. These compounds can further trigger the development of faults
in the transformer.
The essential part of DGA lays on the gas-in-oil analysis and interpretations
for a reliable fault diagnostic. Though the mechanisms of gas generation are
2
1.1 Description of the problem
very complex and not yet fully understood, the patterns of gas generation
corresponding to the types of faults are established on the basis of practi-
cal experiences [Sanghi, 2003]. The patterns of gas generation are usually
represented in terms of thresholds of gas concentrations or key gas ratios
pertaining to the types faults. Thus, the comparison of the concentration of
the characteristics gases in the oil sample with the established thresholds and
key gas ratios may help to detect whether a certain type of fault has occurred
in the transformer, or any incipient fault is developing, or the transformer is
working under normal condition.
Over the last three decades, a great deal of research has been carried out
to improve the DGA and its interpretation schemes. However, the relia-
bility of the DGA methods is limited due to several factors related to: lack
of knowledge about diffusion process of gases, the sampling of service oil,
stray gassing behavior, techniques of dissolved gas extraction, techniques of
gas measurement and the lack of sufficient knowledge about patterns of gas
generation due to various types of faults.
Due to free interface between transformer oil and the surrounding atmo-
sphere, located in the conservator of an air-breathing transformer, there is
a continuous diffusion of fault gases from oil into the air. Depending on the
interval of sampling, this process may have certain impact on DGA result.
However, due to lack sufficient knowledge none of the present DGA meth-
ods address the impact of the diffusion process.
3
1 Background and Objectives of the Research
Several methods such as CIGRE interpretation scheme, key gas analysis, IEC
ratio etc. have been proposed to interpret the DGA results to identify the
types of faults and the condition of a transformer. However, the interpreta-
tions based on these schemes are often inconsistent and highly dependent
on the additional information such as the history of transformer operation,
experience of the interpreter etc. The experience gained by using these meth-
ods indicates that these methods are characterized by certain limitations, e.g.
none of the methods is able to recognize slowly developing low energy faults
or simultaneously occurring faults. Moreover, the presently used interpreta-
tion schemes do not provide any information regarding the intensity of the
faults [Sun et al., 2012].
• Investigation of the diffusion process of fault gases from oil into the
atmosphere through the open conservator tank, and the effect of circu-
lation rate on the diffusion process.
4
2 Power Transformers and
Dissolved Gas Analysis of Oils
The electrical power has been essential for the development of the modern
society. The reliable supply of electricity has become not only essential for in-
dustrial, commercial and residential sectors but also a duty of governments.
In the alternating current (AC) electrical supply systems transformer is an
indispensable component. In a normal power generation plant electricity
is generated at about 11000-15000 volts (11-15 kV). In distribution network
it is passed to the first transformers to step up the voltage to the transmis-
sion level at approximately 220 to 500 kV. At the end of transmission route
the voltage is stepped down to a sub-transmission level at about 33 kV to
132 kV. Then it is sent to distribution utility substations where the voltage is
reduced to approximately 11 kV to 22 kV. Finally, the electric power with re-
duced voltage is sent to local transformers where it is further stepped down
to the consumer voltage level of 415 V to 240 V [Harlow, 2004]. Several trans-
formers are placed between power plant and consumers, which perform the
important role of transforming the generated power to higher and/or lower
voltage levels according the requirements.
5
2 Power Transformers and Dissolved Gas Analysis of Oils
The voltage induced in the secondary winding follows the Faraday’s law of
induction, which states, ”the electromotive force in any circuit is directly pro-
portional to the time rate of change of magnetic flux through the circuit” [Sadiku,
2006].
dΦ
Vs = Ns (2.1)
dt
where,
Moreover, the cross section area ’A’ is constant and the magnetic field changes
with the time depending on the excitation of the primary winding. The
same magnetic flux Φ passes through the primary and secondary windings
6
2.1 Power transformers
where,
7
2 Power Transformers and Dissolved Gas Analysis of Oils
Transformers can function as very efficient machines, but in the practice there
are considerable amounts of energy losses. Modern power transformers have
been designed to exceed 95 % efficiency, yet most of the power loss occure
due to electrical resistance of the iron core and the windings (eddy-current
losses). The lost energy dissipates in the form of heat. [Heathcote, 1998]. The
equation 2.4 represents a transformer that functions at a real time efficiency
[Heathcote, 1998].
Vs Ns Ip
= = (2.4)
Vp Np Is
where,
Ip primary current
Is secondary current
• Oil tank and its cooling system: The oil tank is the outer part of a trans-
former, in which the core, windings and transformer oil are contained.
Cooling collars are attached to the outer surface of the tank to achieve
heat dissipation. For small transformers cooling is carried out through
air circulation and heat radiation, however large transformers require
adequate cooling system coolant substance. This is usually achieved
8
2.1 Power transformers
9
2 Power Transformers and Dissolved Gas Analysis of Oils
expansion. The tank has a breather, which is a glass vessel filled with
silica gel to protect the oil from moisture and to allow the gases in oil to
diffuse in to the atmosphere. Due to this system transformer behaves as
an open-breathing system that allows gas exchange between the main
tank and the surrounding environment [Nynas Naphthenics, 2004].
• The Buchholz relay: Also known as a gas relay, is a safety device installed
on the pipe connecting the main tank and the conservator. Under nor-
mal conditions the relay is completely filled with oil and it activates
when the floats switch is displaced by certain accumulation of gas. This
relay can switch off the transformer when a strong surge of oil flows
to the conservator, or when the oil level falls down to dangerous levels
[Nynas Napthenics, 2004].
10
2.2 Incipient faults in transformers
Sometimes a very high voltages can cause formation of plasma in oil through
which electric current can flow freely as an arcing discharge [Arvidsson,
2005]. Arcing discharges generate very high temperatures (above 5000 ◦ C)
and large amount of gases, mainly acetylene and hydrogen. This type of
faults are very dangerous and if not controlled, can cause excessive pressure
in the transformer tank, causing even explosion.
Depending on the type of fault specific patterns of gases (the gas compounds
and their ranks in terms of concentrations) are generated in oil. The amount
11
2 Power Transformers and Dissolved Gas Analysis of Oils
12
2.3 Dissolved gas analysis of transformer oils
The techniques of refining crude oil have a great impact on the required
dielectric properties, oxidation stability and gas absorbing properties of a
transformer oil (figure 2.5). Refining of crude oil consists of a several phys-
ical and chemical treatments that remove unstable aromatics, harmful polar
molecules and ions.
Since 1990 hydrogenation has been applied as a more efficient and environ-
mental friendly extraction technique. Hydrogenation process converts po-
lar compounds, aromatic and heteroatomic molecules into desired saturated
13
2 Power Transformers and Dissolved Gas Analysis of Oils
ring and chain molecules through a controlled catalytic reaction using hy-
drogen under high pressure and temperature. The level of hydrogenation
has significant impact on dielectric properties and thermal oxidation stabil-
ity of the oil. Therefore, antioxidants or inhibitors are added to highly hydro-
genated oils to enhance their resistance to thermal oxidation [Gradnik, 2007].
14
2.3 Dissolved gas analysis of transformer oils
Among the other additives in transformer oil are metal deactivators and pas-
sivators. Metal deactivators perform their function by decreasing the cata-
lyst effect of metals. The passivators which are acid compounds that bind to
metal surfaces by mean of the acid groups, forming a protective molecular
layer that avoid the catalytic effect of metals in contact with the oil [Arvids-
son, 2001].
Chemical composition
The basic chemical composition of transformer oils includes a mixture of dif-
ferent hydrocarbons structures, such as paraffinic, naphthenics, and aromat-
ics (figure 2.6). The content of these hydrocarbons varies according to source
of the crude oil and the refining process. Depending on the content of paraf-
finic structures (Cp), the oils can be classified as naphthenic (Cp < 50), inter-
mediate (50% < Cp < 56%), and paraffinic (Cp > 56%) [Nynas Naphthenics,
2004]. This classification is aimed at providing a guideline for the application
of oil.
Paraffinic transformer oil offers low solvency for oxidation products and
15
2 Power Transformers and Dissolved Gas Analysis of Oils
moisture. They exhibit low thermal stability, when they are cooled below
their pour point. Paraffin structures consist of saturated hydrocarbons form-
ing straight type chains (N-alkanes) or branched type chains (isoparaffins).
On the other hand, transformer oil with higher content of naphthenic molecules
exhibit excellent properties at low temperature and high solvency for oxida-
tion products and moisture. This type of oils are commonly preferred for
application in transformers.
16
2.3 Dissolved gas analysis of transformer oils
vice span. These specifications are defined on the basis of the required phys-
ical, dielectric and chemical properties of the oil. Generally, a transformer oil
should have following properties:
• low viscosity and low pour point for adequate heat exchange and flow
circulation during operation at low temperature,
• high flash point to avoid risks of oil vaporization and thus risks of fire
or explosions,
Oil with good oxidation stability can restrain the generation of decompo-
sition products such as inorganic acids, alkali, corrosive sulfur, and sludge
17
2 Power Transformers and Dissolved Gas Analysis of Oils
under normal and faulty operations, which in consequence can protect the
dielectric property of the oil and prolong the service span of the transformer.
18
2.4 Oxidation of transformer oils
The schematic outline in figure 2.8 presents the oxidation process of a hydro-
carbons. The oxidation process of hydrocarbons involves three basic stages:
initiation, propagation and termination. In the initiation stage oxygen reacts
spontaneously to breakdown a hydrogen molecule and form hydroperox-
ides, which again dissociate into free radicals [Sanghi, 2003]. The propaga-
tion is the rapid progression of the branched chain reactions, and the ter-
mination occurs with the formation of stable intermediate radicals and non-
reactive compounds that donate hydrogen molecules to free radicals [Lip-
shtein et al., 1970].
The table 2.5 presents a simplified general model of possible reactions which
occur during the oxidation of transformer oils [Nynas Naphthenics, 2004].
Reactions (1), (2) and (3) correspond to the initiation stage, in which perox-
ides and free radicals are produced under the influence of heat, high elec-
trical stress and metals. From reactions (4) the propagation and branching
takes place, resulting in the production of some stable and soluble byprod-
ucts, such as alcohols, aldehydes, carboxilic acids, ketones, ester and water
[Arvidsson, 2001]. The hydrocarbon molecules follow the reaction path and
lead to accumulation of the byproducts until the action of certain inhibitors
form termination products suppressing further oxidation.
19
2 Power Transformers and Dissolved Gas Analysis of Oils
20
2.4 Oxidation of transformer oils
The Arrhenius law (equation 2.5) states that at higher temperatures the re-
action rate increases due to the collisions between the molecules caused by
higher activation energy [Arvidsson, 2001]. This equation can be used to
represent the oxidation process of transformer oil undergoing a thermal de-
composition of its hydrocarbon molecules due to a fault of certain intensity.
dC −∆E
= Ak · e RT (2.5)
dt
where,
dC
dt = change in concentration with the time [mol/s]
Ak = reaction constant according to kinetics [-]
T = absolute temperature [K ]
The thermal decomposition of hydrocarbons occurs as primary and secondary
decomposition. The products of a primary decomposition are in equilibrium
with their initial hydrocarbon, furthermore the products of primary decom-
position decompose by heat as secondary decomposition products [Shirai et.
al., 1997]. The thermal decomposition of an alkane is useful to represent the
decomposition of new transformer oil due to carbon-to-carbon bonds split-
ting and dehydrogenation mechanisms. Usually carbon-carbon splitting de-
composition of an alkane results into lower molecular weight alkane and
alkene, in addition the alkane dehydrogentaion generates alkanes and hy-
drogen [Shirai et. al., 1997].
21
2 Power Transformers and Dissolved Gas Analysis of Oils
Cn H2n+2 ↔ H2 + Cn H2n
Cn H2n+2 ↔ CH4 + C(n−1) H2(n−1)
(2.6)
...
Cn H2n+2 ↔ C(n−2) H(2n−2) + C2 H4
Thus, the total mole number of products in the system is made of x moles of
an alkane decomposed from 1 mole (1 − x) at certain temperature in equilib-
rium and the products Ai x [Shirai et. al., 1997]:
(1 − x) + 2ΣAi x = (1 + x)
(2.8)
The figure 2.9 shows some alkanes and their respective temperatures of de-
composition.
Figure 2.9: Decomposition of alkanes with respect to temperature[Shirai et. al., 1997]
22
2.4 Oxidation of transformer oils
The transformer oil has hydrocarbon molecules with an average carbon num-
ber of 20 and average molecular weight of approximately 280 Da. Thus, the
thermal decomposition of eicosane (C20 H42 )aims at representing the process
gas generation in transformer oil [Shirai et. al., 1997]. The thermal decompo-
sition of eicoisane at above 300 ◦ C can be defined by the following equation
2.9 [Shirai et. al., 1997].
with,
x1 + 2(x2 + x2 + x4 ) + 3(x5 + x6 ) = 20
4x1 + 6x2 + 4x3 = 2x4 + 8x5 + 6x6 + 2x7 = 42
23
2 Power Transformers and Dissolved Gas Analysis of Oils
The breakdown of transformer oil due to electrical stress has been explained
by the theory of avalanche ionization of atoms dissociated by electron col-
lisions [Wadhwa, 2012]. Figure 2.11 represents the simplified theory of oil
breakdown based on avalanche ionization of atoms. During discharge events
electrons are introduced in the oil by the cathode, creating an electro-thermal
condition that produces localized high temperatures. In this condition small
gas bubbles are generated as charge carriers which propagate further ioniza-
tion in oil, that in turn cause further generation of fault gases. This theory
applies to pure transformer oil free from any particles and moisture.
24
2.4 Oxidation of transformer oils
Figure 2.11: Development of electrical breakdown of oil during fault gas generation
[Schmidt, 1982]
decomposition of n-octane (C8 H18 ). Table 2.6 presents the relative enthalpies
required for the fault gas formation.
Thus, the generation of methane, ethylene and ethylene requires the least
generation energy, while hydrogen and acetylene need much larger genera-
tion energy. Therefore, generation of acetylene in large amount is indication
of a severe fault, such as arcing discharge which can result in serious damage
to a transformer. Also hydrogen, which is generated in different amounts by
almost all faults, provides certain indication of type and intensity of the re-
lated fault.
25
2 Power Transformers and Dissolved Gas Analysis of Oils
R• + AH ↔ R − H + A•
RO• + AH ↔ R − O − H + A•
(2.10)
R − O − O• + AH ↔ R − O − OH + A•
OH • + AH ↔ H2 O + A•
where,
The molar Gibbs free energy expression can be written in terms of chemical
potentials, as following [Atkins et al., 2006]:
∑
∆Gm,mix = xi ∆µi,mix (2.12)
i
26
2.5 The gas-in-oil as ideal solution
with,
∆µi,mix = RT ln xi ; which is defined as change in the chemical potential of
component ’i’ on mixing.
When there is an equilibrium between liquid and gas phases, then the chem-
ical potentials of ’i’ are equivalent (µi,liq = µi,gas ) [Atkins et al., 2006].
Hence, for a given component ’i’ the Raoult’s law can be applied over the
whole range of concentrations as following [Atkins et al., 2006]:
Pi = xi ∗ Pi∗ (2.14)
where,
xi = mole fraction
The ability of fault gases to dissolve in oil is highly dependent on the factors
such as temperature, pressure as well as chemical composition. Figure 2.12
represents the solubility of hydrocarbon gases in oil as a function of temper-
ature. The solubility of hydrocarbon gases in oil is higher than the solubility
of hydrogen and atmospheric gases in oil. Furthermore, the solubility of hy-
drogen, nitrogen, carbon monoxide and oxygen increases with increase in
temperature, while the solubility of carbon dioxide, acetylene, ethylene and
ethane reduces with the increase in temperature [Griffin, 1988].
One of the most common methods applied for quantification of gas solubility
in oil is the Henry’s law. The principle of this law is limited to dilute ideal
27
2 Power Transformers and Dissolved Gas Analysis of Oils
Pgas = kH C (2.16)
Where,
Pgas = partial pressure [atm]
The Bunsen solubility coefficient, α, is defined as the volume of gas that can
be absorbed by the unit volume of solvent at a specified temperature, under
the partial pressure of the gas of 101.3 kPa [Fogg et al., 1991]. For ideal solu-
tions the equation is written as equation 2.17 [Battino et al., 1966].
Vgas 273.15
α= · (2.17)
Vsol T
28
2.5 The gas-in-oil as ideal solution
Where,
T = absolute temperature, [K ]
29
2 Power Transformers and Dissolved Gas Analysis of Oils
The second Fick’s law (equation 2.20) is used in non-steady condition and
time dependent diffusion. It relates the rate of change of concentration at
30
2.6 Diffusion of gases
∂C ∂2C
=D· (2.20)
∂t ∂x2
Where, t is time in seconds.
Thus the injected mass of concentration C(x, t) is given by the following ex-
pression [Scott et al., 2002].
∫
C(x, t) = C(x, t)∂V (2.21)
v
M
e−x /4Dt
2
C(x, t) = √ (2.22)
A 4πDt
In the case of a pure diffusion process, the instantaneous point source diffu-
sion model could be applied to estimate diffusion coefficient of compounds
within a certain volume. For instance, diffusion of gases dissolved in the oil
volume of a transformer tank, however gases generated in transformers are
31
2 Power Transformers and Dissolved Gas Analysis of Oils
32
2.7 The dissolved gas analysis methods
Among interpretation schemes, the most popular and frequently used scheme
is the key gas method, which consist of correlating a fault type with the
concentration of one of the key gases (C2 H4 , CO, H2 or C2 H2 ), for instance
high concentrations of hydrogen predicts partial discharge, ethylene indi-
cates overheating above 500 ◦ C and acetylene is characteristic for a thermal
fault above 1000 ◦ C. The key ratios consist of CH4 /H2 for partial discharge,
C2 H2 /C2 H4 for arcing discharge, C2 H2 /C2 H6 for discharges, and C2 H4 /C2 H6
for hotspots. These ratios of gas concentrations are compared with certain
threshold values corresponding to the types of faults [Sun et al., 2010].
33
2 Power Transformers and Dissolved Gas Analysis of Oils
Table 2.8: Fault diagnostic by applying the CIGRE fault interpretation scheme (par-
tial discharge)
Part I: Key gas ratios
Key gas Ratio [-] Fault type
(C2 H2 /C2 H6 ) >1 Arcing discharge (AD)
(H2 /CH4 ) > 10 Partial discharge (PD)
(C2 H4 /C2 H6 ) >1 Oil overheating (OH)
Part II: Thresholds [ppm]
Key gas Thrsh. [ppm] Fault type
C 2 H2 > 20 Arcing discharge (AD)
∑ H2 > 100 Partial discharge (PD)
∑Cx Hy > 1000 Oil overheating (OH)
COx > 10000 Cellulose degradation (C)
34
2.7 The dissolved gas analysis methods
During this study the data obtained from experimentally simulated faults
and the analysis of corresponding fault gas generation is used to evaluate
the accuracy and reliability of the CIGRE interpretation scheme and the gas
generation pattern method.
35
3 Experimental Setups and
Measurement Equipments
The design and construction of several experimental setups were required to
investigate the gassing behavior of the transformer oils. These setups were
aimed at simulating the gas-in-oil generation as it occurs in power transform-
ers operating with certain thermal or electrical faults.
3.1.1 Oil tank (12 liter) and high voltage supply system
The figure 3.1 shows a schematic representation of oil tank and the high volt-
age supply system. The oil tank of 12 liter capacity was made of a plexiglass
cylinder and sealed at the top and bottom using steel plates. A smaller tank
of 6 liter capacity was placed on the top of the oil tank to simulate the func-
tion of open-breathing conservator tank to contain the oil in case of thermal
expansion.
36
3.1 Small scale setup
Figure 3.1: High voltage suppply circuit for partial discharges in the oil tank (12
liters)
37
3 Experimental Setups and Measurement Equipments
38
3.2 Large scale setup system
the bottom covers of the oil tank. Due to high resistance the semi-conductor
materials like graphite induce high temperature in short time when they are
subjected to electric current. This allows to create hotspot in the oil tank to
simulate thermal faults in the transformer. Apart from the graphite other
materials were tested to create hotspots, nevertheless only graphite induced
the temperatures high enough to trigger oxidation of the transformer oil.
A needle-plate electrodes were fixed on the invert side of the top cover on
the main tank, it was connected to the high voltage circuit through the bush-
ing. The distance between the needle and the plate was adjustable. The high
voltage circuit consisted of an autotransformer (0-400 V AC), a high volt-
age transformers (0-100 kV), an earthing motor, security fences with warning
lamps, electronic circuits and a control panel.
39
3 Experimental Setups and Measurement Equipments
Figure 3.4: High voltage supply circuit for electric discharges in the oil tank (600
liter)
40
41
3.2 Large scale setup system
Figure 3.5: Schematic view of oil circulation system for the large setup
3 Experimental Setups and Measurement Equipments
Figure 3.6: Volumetric flow of a fault gases in the oil balance of the large scale setup
The volumetric flow of a fault gases in the oil balance of the large scale setup
is demonstrated in the figure 3.6. The volume of a given fault gas (VT (i) ) dis-
solved in oil can be expressed by its concentration gradient (CT (i) ), which
42
3.2 Large scale setup system
varies with time, mixing velocity and solubility of that particular gas. The
volumetric flow of the fault gas is transferred to the conservator (VC(i) ) at a
specific pump flow rate (F R) for oil circulation aimed at mixing of oil balance.
The concentration gradient of the fault gas (CC(i) ) in the conservator tank
decreases as the gas diffuses to the atmosphere (CD(i) ). At the same time,
the oil balance keeps on circulating, mixing and transferring certain amount
of gases out of the oil balance, which results in an ongoing decrease of the
fault gas concentration in the oil balance until the diffusion process reaches
its equilibrium.
Figure 3.7: Interaction scheme of the automation system for the large scale setup
43
3 Experimental Setups and Measurement Equipments
This is a portable gas analysis device (figure 3.8) based on the principle of
photo-acoustic spectroscopy (PAS) which can detect gas-in-oil concentrations
of characteristic gases such as methane, ethane, ethylene, acetylene, carbon
dioxide and carbon monoxide, and moisture within a range between 1 to
50000 ppm with accuracy of approximately ± 2 ppm. Since the PAS can not
detect the atomic structure of hydrogen, the hydrogen concentration is deter-
mined by a semiconductor sensor (SnO2 ) which can measure up to 5000 ppm
of hydrogen concentration. The measurements are conducted at 50 ◦ C using
50 ml of oil sample and approximately 5 ml of gas sample [Kelman Transport
X Infoblat, 2009].
The PAS measurements are based on the facts that absorption of infrared
light cause heat-expansion of a gas and that each gas absorbs only a specific
infrared light spectrum. Therefore, a particular sound wave produced by
expansion of a gas can indicates the type of the gas and the intensity corre-
sponding to the sound wave indicates the concentration of the gas [Cargol,
2005].
44
3.3 The DGA equipments
To perform the measurements oil samples are injected into a sample bottle
using a glass syringe. After an assessment of temperature the oil is stirred
and circulated for certain time in order to induce partitioning of dissolved
gases and oil. Once the equilibrium is achieved, the gases accumulated in
the headspace are analyzed by the photo-acoustic spectrometer (PAS). The
results are displayed on in-built screen and recorded in the device memory.
In case of a manual operating mode, the oil samples are first injected into
a vacuum degassing device using a 100 ml glass syringe. The vacuum de-
gassing device performs the extraction of dissolved gases by means of di-
aphragm pumps. The process of degassing occurs in a pressure range be-
tween 2 to 200 mbar in several cycles and it last for approximately 2 min. Its
principle is based on the standard norms IEC 60567.
45
3 Experimental Setups and Measurement Equipments
The figure 3.10 presents the details of the syringe used for the system. The
syringe has a metal plunger that can be locked in two different positions: first
one is at 40 ml volume, and second one is to produce the vacuum headspace.
After filling the syringe with an oil sample, the plunger is locked at the sec-
ond position. Then the oil filled syringe is shaken manually for a certain time
46
3.3 The DGA equipments
to induce partitioning of the oil sample in oil and gas phases. The syringe is
then placed standing vertical in order to allow equilibrium conditions be-
tween the two phases and let the gases to accumulate in the headspace. Af-
ter the equilibrium is reached, the plunger of the syringe is unlocked and left
free in order to let internal pressure equalize with the atmospheric pressure.
In the following step the gases accumulated in the headspace are extracted
with another syringe and analyzed by means of a gas chromatographer.
Cl0 Vl = Cl Vl + Cg Vg (3.1)
Where, Vl = 40 mL (volume of the oil sample) Vg = volume of the gas (mea-
sured with the shaking syringe after gas extraction and compression)
47
3 Experimental Setups and Measurement Equipments
After rearranging the equation 3.1 the concentration of gas-in-oil can be de-
termined by the equation 3.3.
Vg
Cl0 = Cg K (3.3)
Vl
The solubility factors presented in the table 3.1 were provided by the technol-
ogy supplier (Energy Support GmbH). These coefficients are generally deter-
mined experimentally based on oil standards [ASTM Test Method D2780] or
calibrated according to a particular method.
Table 3.1: Solubility factors for gases, calibrated for the shaking syringe extraction
method
Gas component H2 CH4 C2 H6 C2 H4 C2 H2 CO CO2 O2 N2
Solubility factors 1.6 2.5 9.0 6.1 4.7 1.7 4.1 1.5 1.4
The TGM can determine gas saturation contents for hydrogen, oxygen, ni-
trogen, carbon dioxide, carbon monoxide as well as the sum of hydrocar-
bons. The gas extraction is based on the natural internal standard (NIS)
which specifies that pressure and nitrogen concentration in an equilibrium
gas headspace are comparable to those in the atmosphere [Aragón-Patil et
48
3.3 The DGA equipments
When this system is set for a given measurement, the equilibrium gas head
space is produced and the related parameters are measured. Further the gas
samples from the equilibrium headspace are extracted and analyzed using
chromatography. The resulting gas concentrations are then given in to a
software that determines the final gas-in-oil concentrations according to the
measured parameters, such as pressure of oil solution in the gas provisioning
column, oil temperature as well as certain calibrated solubility factors.
49
3 Experimental Setups and Measurement Equipments
Figure 3.12: Partial discharge control and measurement system for the small test
setup
50
3.3 The DGA equipments
the acquisition units. The acquisition unit in connected to the fiber optic con-
troller via fiber optic cables and coaxial connectors. The monitoring system
can record and store data from PD events in real time at 25 frames per second.
The figure 3.13 presents a PRPD diagram that can be interpreted according
to the distribution of PD pulses of a specific intensity (the dot colors corre-
spond to an intensity indicated on the gamut scale), the amplitude of phase
voltage ( green curve), and the average magnitude of charge (QIEC ) and ap-
plied voltage.
Figure 3.13: Phase resolved partial discharge diagram for a PD event (PRPD)
51
3 Experimental Setups and Measurement Equipments
Table 3.2: Main characteristics of the transformer oils used for the experiments
Name Nynas Nytro Nynas Nytro Nynas Shell Diala
Lyra 3000X 10 GBN DX
Type High grade High grade High grade
Inhibited Inhibited Uninhibited Inhibited
Kinematic viscosity 9.4 9.4 8.9 17.0
at 40 ◦ C [mm2 /s]
Flash point [◦ C] 138 135 148 136
Density [kg/m3 ] 871 871 886 877
(at 20 ◦ C) (at 20 ◦ C) (at 20 ◦ C) (at 15 ◦ C)
Pourpoint [◦ C] - 60 - 40 - 57 < - 60
Breakddown 40-60 40-60 40-60 >60
Voltage [kV]
cal composition and additives in the transformer oils are not known due to
confidentiality of the product information.
52
4 Results and Discussions
4.1 Investigations of factors affecting the
dissolved gas analysis (DGA)
Dissolved gas analysis (DGA) of representative oil samples play key role in
diagnostics of condition of a power transformer. Therefore quality of the oil
samples is crucial for precise diagnostics. Quality of oil samples can be sig-
nificantly influenced by the practices used for oil sampling and storage. The
oil properties such as stray gassing and the techniques used to extract the
gases dissolved in oil can also affect DGA.
53
4 Results and Discussions
• The gas concentrations for all the extraction techniques were measured
using the gas chromatography from Mobil GC (Energy Support) with
same calibration parameters.
The figure 4.1 and table 4.1 present the concentration and relative standard
deviations for the critical gases obtained using the extraction techniques un-
der investigations. The results demonstrate that the gas extraction technique
vacuum (VE) leads to the highest gas-in-oil concentrations, which means
maximum gas extraction and least loss of gases during the extraction. Thus,
vacuum (VE) can be considered as a most efficient gas-in-oil extraction tech-
nique, followed by equilibrium gas head escape (EHS) and dynamic head
space (DHS). The gas extraction using shaking syringe (SS) leads to the low-
est gas-in-oil concentrations. The concentrations of the critical fault gases
obtained using shaking syringe (SS) are in general 50 % to 60 % lower than
that obtained using vacuum extraction (VE), which can be attributed to low
gas extraction efficiency of the shaking syringe technique.
Figure 4.1: Gas-in-oil concentrations obtained using different gas extraction technics
54
4.1 Investigations of factors affecting the dissolved gas analysis (DGA)
Table 4.1: Mean gas-in-oil concentrations and relative standard deviations (RSTD)
obtained for multiple measurements using different gas extraction meth-
ods
Component Gas-in-oil concentration [ppm]
H2 CH4 C 2 H6 C2 H4 C2 H2
Manual VE 155.8 21.8 2.1 25.4 138.5
RSTD [ %] 2.3 5.3 10.2 4.5 6.3
Online VE 153.4 21.8 2.4 28.0 143.7
RSTD [ %] 2.6 6.1 8.1 7.2 7.4
SS 78.9 8.86 2.4 9.84 53.60
RSTD [ %] 5.0 7.8 13.6 9.8 8.9
EHS 144.8 17.0 0.0 23.6 122.5
RSTD [ %] 2.1 0.0 0.0 2.1 1.6
DHS 144.4 22.4 3.2 27.2 115.2
RSTD [ %] 5.5 8.1 15.2 6.6 1.6
The relative standard deviations (RSTD) indicate that equilibrium head space
(EHS), with a RSTD below 2.1 %, produced the best measurement consis-
tency for all the gases. The gas extraction carried out using syringe (SS) and
dynamic head space (DHS) leads to highly inconsistent measurements, espe-
cially in the case of hydrogen (RST D > 5%). The manual and online vaccum
(VE) technique show small RSTD (< 2%) for hydrogen, however it is incon-
sistant in case of other gases.
• Aged Nynas Nytro 10 GBN transformer oil (not-inhibited and air satu-
rated) and
55
4 Results and Discussions
Air-trapping
• Oil aging: achieved by heating up the oil at 120 ◦ C for 4 days continu-
ously.
• The samples were kept under dark and ambient conditions for 14 days
before carrying out DGA.
Figure 4.2: Influence of air bubbles on the concentration of stored oil samples
56
4.1 Investigations of factors affecting the dissolved gas analysis (DGA)
These gas components are among the least soluble in oil thus their concen-
trations presented the largest deviations due to the effect of air trapped in
the sample. In general, the gas-in-oil concentrations tend to decrease as the
volume of the air bubbles increases. In the case of air bubble larger than 8
% of the oil volume, the concentration of hydrogen reduces approximately
by 35 %. The deviations become already significant when volume of an air
bubble exceeds 2 % of the oil sample volume.
Table 4.2 shows the deviations in the concentrations of selected gases for the
oil sample with air bubble of 1 ml volume. It can be observed that a 1 ml air
bubble which occupies 2 % of the oil volume, causes significant reduction in
the gas-in-oil concentrations in the oil sample. The concentrations of critical
fault gases such as ethylene, acetylene and hydrogen showed the largest de-
viations.
Table 4.2: Effect of 1 ml air bubble on gas concentrations in oil stored for 14 days
Gas Concentration of gas-in-oil [ppm] Deviations [ %]
0 days 14 days
CO2 1022 940 8.7
C 2 H4 4152 2850 45.7
C 2 H2 693 605 14.5
C 2 H6 455 398 5.3
CH4 1701 1625 4.7
CO 106 94 12.8
H2 1187 904 31.3
57
4 Results and Discussions
Table 4.3: Gas-in-oil concentrations of samples stored in gas syringes and aluminum
bottles
Gas Concentration of gas-in-oil [ppm] Deviations [ %]
Glass syringe Aluminum bottle
C2 H4 21.6 20.1 7.4
C2 H2 80.2 78.7 1.2
C2 H6 6.1 5.8 5.2
CH4 15.9 15.4 3.2
CO 16.5 15.0 10.0
H2 89.0 83.6 6.5
Daylight/darkness
The effect daylight or darkness on the oil samples was investigated under
following considerations:
• Reference oil samples: aged oil stored in dark without the influence of
air bubbles.
Figure 4.3 and table 4.4 shows the concentrations of hydrogen and carbon
dioxide for the oil samples stored under light and darkness. The sample of
new oil stored in dark indicates a decrease in the concentration of carbon
dioxide by approximately 12 %. This change can be caused by a chemical
quilibrium among certain carbonic acids in the oil composition. However,
the concentration of hydrogen does not show any change.
58
4.1 Investigations of factors affecting the dissolved gas analysis (DGA)
Figure 4.3: Effect of light on the concentrations of hydrogen and carbon dioxide
Table 4.4: Concentrations of hydrogen and carbon dioxide stored under light and
dark
Gas-in-oil concentration [ppm]
Oil New oil Aged oil
Condition H2 CO2 H2 CO2
Reference 5 514 72 544
Light 5 504 90 547
Darkness 5 453 78 543
In the case of aged oil samples stored in light, there was a clear increase in
the concentration of hydrogen by approximately 18 %. This increase can be
attributed to photochemical reactions that involves decomposition of aging
compounds in oil, such as water and acids. For the aged samples stored in
dark, the concentration of hydrogen increased by 8 %, possibly as a conse-
quence of ongoing aging and chemical reactions.
Temperature
59
4 Results and Discussions
Storage duration
60
4.1 Investigations of factors affecting the dissolved gas analysis (DGA)
for 16 days and monitored during the storage duration. The experiments
were carried out under following conditions:
61
4 Results and Discussions
The tests were carried out according to the GIGRE guidelines for stray gassing
oils [Griffin et al., 2005]. The methodology of CIGRE recommends prepara-
tion of samples of new (air saturated) oil with and without a copper strip
which represents the copper windings inside a transformer. These samples
were then placed in an oven at 90 ◦ C for 12 days duration, during which the
gas concentrations in the samples were continuously monitored.
Results obtained for the oil Nynas Nytro 10GBN are ploted in figures 4.6,
4.7 and 4.8. Rapid increase in the concentrations of hydrogen, methane and
ethane in the sample with copper strip can be observed in the plots. These
high gas generation rates correspond to stray gassing effect of the trans-
former oil. The plot in figure 4.6 is characterized by a quick increase of hy-
drogen concentration before reaching a plateau from which the concentration
does not change further, which is the typical tendency of the oils with stray
gassing behavior.
Figure 4.6: Stray gassing behavior of hydrogen in oil with and without copper
62
4.1 Investigations of factors affecting the dissolved gas analysis (DGA)
Figure 4.7: Stray gassing behavior of hydrocarbon gases in oil with copper
Figure 4.8: Stray gassing behavior of hydrocarbon gases in oil without copper
63
4 Results and Discussions
The data obtained from the experiments was used to derive correlation be-
tween fault energy and concentrations of fault gases. Furthermore, in order
to estimate diffusion rates of critical fault gases, diffusive behavior of the
generated fault gases was also analyzed. The experiments and the following
analysis of results include consideration of the know-how of measurement
accuracy of the equipments and oil sampling etc., acquired during the exper-
iments described in section 4.1.
• Small scale setup along with its partial discharge system (see sections
3.1.1 and 3.1.2)
• Transformer oil Shell Diala DX (12 liters); new, air saturated and at nor-
mal ambient conditions (see table 3.2)
For the adequate application of partial discharges, the following factors re-
quired to be predefined and verified before the experiments:
64
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
4. The oil was kept in quiescent condition for 12 hours after filling up the
test tank. In this way, it was reduced the possibility of PD created in air
bubbles.
5. The sampling point for gas-in-oil analysis was located near the elec-
trode, see figure 4.9.
Figure 4.9: Test tank (12 liter) with the electrode setup for producing partial dis-
charges
The table 4.5 shows gas-in-oil concentrations of fault gases measured after
reconditioning of the transformer oil Shell Diala DX aided by treatment using
the VOT-3 system followed by air saturation with normal atmospheric air
bubbled for several hours in the oil barrel.
65
4 Results and Discussions
Table 4.5: Gas-in-oil concentration of oil after reconditioning and air saturation
H2 CH4 C2 H6 C2 H2 C2 H4 CO N2 O2
Gas-in-oil [ppm] 5.7 1.5 8.3 7.6 5.2 9.3 56320 35330
Methodology
Further, the voltage was increased until a maximum level where the PD
pulses became more continuous and evenly distributed along the sinusoidal
phase voltage. Generally, the frequency of PD pulses varies randomly in suc-
cessive cycles, characterized by certain increasing or decreasing frequency
with respect to time and applied voltage.
Results
Figure 4.10 and figure 4.11 present some examples of PRPD diagrams ob-
tained for each PD pattern. Both of the PD patterns exhibit a mean apparent
charge (QIEC ) of approximately 1000 pC that remained almost constant dur-
ing the test period.
The pulses and discharges within a PD event exhibit stochastic character that
make them differ among themselves at instant values of voltage. Thus, the
estimation of total energy involved in a partial discharge event requires a sta-
tistical assessment of the total number of discharges, including their individ-
ual pulse apparent charge and instantaneous voltage during an AC voltage
period [Florkowska et al., 1999].
66
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
The figure 4.12 present the concentration-time profile of fault gases generated
during the application of the PD-1. The generation of gas-in-oil was moni-
tored using online DGA system by taking the samples at 1 hour interval from
the neighbourhood of the electrode.
67
4 Results and Discussions
The PD-1 was carried out for approximately 58 hours before stopping it man-
ually. The PD pulses were generated in a homogeneous distribution scattered
in a broad manner along the phase voltage curve at a constant applied volt-
age of approximately 11 kV. Most of the discharge pulses occurred without
production of sparks or gas bubbles in a continuous way throughout the test
period. This PD-pattern presented relative lower energy content and the gen-
eration of fault gases occurred at slow rate which allowed high dissolution of
gases in oil. The highest concentration was obtained for hydrogen at about
172 ppm. Other fault gases reached maximum concentrations below 10 ppm.
The PD-2 (figure 4.13) was conducted for approximately 115 hours at approx-
imately 19 kV of applied voltage until it was terminated by a spontaneously
occurred arcing discharge. This PD pattern was characterized by the discon-
tinuity of discharge pulses scattered in narrow shape along the phase voltage
curve. Most of the discharge pulses presented higher energy content in com-
pared to the first pattern. The PD pulses occurred erratically accompanied
by powerful sparks and abundant formation of gas bubbles throughout the
test period. During the PD-2 the highest measured concentration was of hy-
drogen, about 905 ppm. It was followed by a high concentration of acetylene
68
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
in the range of approximately 304 ppm, along with methane and ethylene
above 150 ppm. Additionally, perforations on the pressboard surface were
observed as a result of the intensity of discharges.
Table 4.6 summarizes the most important parameters of the two PD patterns
which are differed in their intensities. These two patterns correspond to PD
experiments that presented the best stability during the test periods.
Verification of the course of these PD faults was carried out by means of a gas
generation pattern method, which analyses the patterns obtained by plotting
normalized gas-in-oil concentrations of H2 , CH4 , C2 H6 , C2 H4 and C2 H2 in oil
[Okubo, et al., 1999 and Tsukioka et al., 1978]. In this method the gas concen-
trations are first normalized by dividing each of them by the concentration of
the dominant gas (with maximum concentration) among the selected gases.
Thus, the dominant gas is always represented by normalized concentration
of 1.0 [-]. Then these normalized concentrations are plotted in the sequence:
H2 , CH4 , C2 H6 , C2 H4 and C2 H2 , as shown in Figure 4.14.
69
4 Results and Discussions
Figure 4.14: Gas concentration patterns obtained for the partial discharge experi-
ments PD-1 and PD-2
70
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
Consequently, the pattern displayed for PD-1 indicates the typical distribu-
tion of gas-in-oil concentrations corresponding to a partial discharge fault
(see figure 2.14), whereas the pattern for PD-2 shows a characteristic gas-in-
oil concentrations for a discharge fault. The comparison of the gas patterns
for PD-1 and PD-2 implies that these faults occurred with different levels of
energy, the partial discharge fault dissipated lower energy than the discharge
fault. However, due to technical limitations of the setup, it was not possible
to determine the energy dissipated during these electric faults.
The fault diagnostic was also carried out by applying the CIGRE fault inter-
pretation scheme (see table 2.8). The CIGRE scheme provides the interpre-
tations of the fault in transformer oil based on the gas-in-oil concentrations.
With already known character of the fault, the diagnostic was aimed at cross-
checking the fault interpretations provided in the CIGRE scheme.
The table 4.7 shows the comparison of the key ratios and threshold values
obtained for the PD experiments with those provided in the CIGRE scheme.
Table 4.7: Fault diagnostic by applying the CIGRE fault interpretation scheme (par-
tial discharge)
Key gas CIGRE scheme PD Experiments
Part I: Key gas ratios
Key gas Ratio Fault Ratio Ratio
Key gas [-] type PD-1 [-] PD-2 [-]
(C2 H2 /C2 H6 ) >1 AD 1.35 4.49
(H2 /CH4 ) > 10 PD 19.32 4.9
(C2 H4 /C2 H6 ) >1 OH 0.76 2.39
Part II: Thresholds [ppm]
Key gas Threshold Fault Threshold Threshold
Key gas [ppm] type PD-1 [ppm] PD-2 [ppm]
C2 H2 > 20 AD 4 314
∑ H2 > 100 PD 172 905
∑Cx Hy > 1000 OH 190 1515
COx > 10000 CD 470 446
A D: arcing discharge, PD: partial discharge, OH: oil overheating,
C D: cellulose degradation
Interpretation of the key gas ratios and thresholds for PD-1 experiment ac-
cording to the CIGRE scheme suggests that PD-1 occurred as a partial dis-
71
4 Results and Discussions
charge event without oil overheating and damage to cellulose, which is al-
ready known from the discharge pattern monitored during the experiment.
Thus, the CIGRE scheme correctly interpret the type of fault in the case of
PD-1.
The interpretation for PD-2 indicates that, according the CIGRE scheme this
fault occurred as a high intensity partial discharge. However, in reality the
PD-2 occurred as a low intensity partial discharge followed by arcing dis-
charge which resulted in oil overheating. Thus, in the case of PD-2 CIGRE
scheme fails to interpret the fault correctly.
The purpose of these experiments was to generate the characteristic fault gas
patterns and concentrations pertaining to local overheating due to thermal
fault. The hotspot experiments were carried out under following considera-
tions:
• Transformer oil Nynas Nytro Lyra (12 liters); air saturated and at nor-
mal ambient conditions (see table 3.2)
Before the tests, the oil was reconditioned. The table 4.8 presents the gas-in-
oil condition of transformer oil Nynas Nytro Lyra after reconditioning and
air saturation. The figure 4.15 presents an image of the test setup during the
conduction of a hotspot fault using graphite conductor.
Table 4.8: Gas-in-oil concentration of oil after reconditioning and air saturation
H2 CH4 C2 H6 C2 H2 C2 H4 CO N2 O2
Gas-in-oil [ppm] 5.2 0.8 12.2 10.5 4.5 9.3 57597 24407
72
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
Methodology
Rearranging the equation 4.1 gives equation 4.2, which can be used for calcu-
73
4 Results and Discussions
lation the temperature from a measured resistance. The graphite lead used
in pencils has certain amount of impurities, therefore resistivity coefficient
of 3.75−5 Ohm.m (at 20 ◦ C) was adopted for the estimations. Experimental
parameters for the hotspot faults at different temperatures can be seen table
4.9.
ρ·L
RT = (1 + α · (T − 20)) (4.1)
A
1 A
T = · (RT · − 1) + 20 (4.2)
α ρ·L
Where,
RT = Resistance of graphite at a specific temperature, Ohm
ρ = Resistivity of material at the reference temperature 20 ◦ C, (3.75−5 Ohm.m)
L = Length of material, (0.01 m)
A = Cross sectional area, (4.65−7 m2 )
T = Temperature, ◦ C
α = Temperature coefficient of graphite (4.84−4 1/◦ C )
Results
The hotspot faults were analyzed using the gas generation pattern method
(see figure 2.14). The normalization of gas-in-oil concentrations for a hotpot
of temperature 500 ◦ C or above was based on the concentration of ethylene,
while for the hotspot of lower temperature it was based the concentration of
acetylene. The figure 4.17 present a comparison of maximum gas-in-oil con-
74
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
75
4 Results and Discussions
Figures 4.18 and 4.19 show the patterns of gas-in-oil concentrations obtained
for the hotspots faults. It can be observed that the gas-in-oil concentrations
for the hotspots of temperature 500 ◦ C and above exhibit a very similar pat-
terns, which can identified as a typical pattern for a thermal fault. However
the hotspot at 300 ◦ C exhibits a pattern of a fault-free condition of oil or, that
of a slow developing thermal fault due to unusual generation of acetylene.
Figure 4.18: Gas concentration patterns obtained for hotspots of different tempera-
tures
The fault diagnostic for hotspots was carried out by applying the CIGRE
fault interpretation scheme, in order to verify the fault interpretations pro-
vided in the CIGRE scheme. The table 4.10 shows the comparison of the key
ratios and threshold values obtained for the hotspot experiments with those
(thermal faults) provided in the CIGRE scheme. Based on the both criterion
of the CIGRE scheme (ratios and threshold) the hotspot fault of temperature
300 ◦ C (HS-4), 500 ◦ C (HS-3) and 1000 ◦ C (HS-1) can be interpreted as arc-
ing discharge fault, while hotspot fault of temperature 1000 ◦ C can also be
76
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
Figure 4.19: Gas concentration patterns obtained for hotspots of different tempera-
tures
Table 4.10: Fault diagnostic by applying the CIGRE fault interpretation scheme
(hotspot)
Key gas CIGRE scheme Hotspots
Part I: Key gas ratios
Key gas Ratio Fault Ratio [-]
[-] type HS-1 HS-2 HS-3 HS-4
(C2 H2 /C2 H6 ) >1 AD 3.9 0.3 3.4 11
(H2 /CH4 ) > 10 PD 0.3 0.3 0.8 0.8
(C2 H4 /C2 H6 ) >1 OH 22.6 5.1 7.71 3.5
Part II: Thresholds [ppm]
Key gas Threshold Fault Threshold [ppm]
[ppm] type HS-1 HS-2 HS-3 HS-4
C2 H2 > 20 AD 25 25 125 110
∑ H2 > 100 PD 189 100 177 15
∑Cx Hy > 1000 OH 173 595 823 2066
COx > 10000 CD 897 645 560 475
A D: arcing discharge, PD: partial discharge, OH: oil overheating,
C D: cellulose degradation
77
4 Results and Discussions
ria. Thus, it can be suggested that the CIGRE scheme provides ambiguous
interpretations of the hotspot faults.
This type of fault involves high amount of energy and it can result in se-
vere failures of transformers. The simulation of arcing discharges under con-
trolled parameters (voltage, current and duration time) was carried out in
order to analyze gas-in-oil generation and gas behavior in oil induced due to
arcing discharge fault. Very repetitive experiments were carried out under
defined parameters in order to generate the faults of different intensities. In
all cases the initial conditions of oil were maintained constant by means of
oil reconditioning.
The results of these experiment were further used to investigate the process
of diffusion of gases through oil before escaping to atmosphere via conserva-
tor tank.
The arcing discharge experiments were carried out using following equip-
ments:
• Transformer oil Nynas Nytro Lyra, new air saturated oil at normal am-
bient conditions (see table 3.2)
Methodology
All experiments were carried out with oil at normal ambient conditions. Pre-
vious to every experiment, the oil volume (approximately 600 litters) was re-
conditioned by means of several recirculation cycles between the main tank
and VOT-3 system (see figure 3.4). Afterwards, the air saturation of oil was
achieved by flow circulation between the main tank and conservator tank for
approximately 12 hours. The table 4.11 displays concentrations of gases after
oil reconditioning and air saturation.
78
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
Table 4.11: Gas-in-oil concentration after oil reconditioning and air saturation
H2 CH4 C2 H6 C2 H2 C2 H4 CO N2 O2
Gas-in-oil [ppm] 5.0 1.0 12.0 10.0 5.0 9.3 57597 24407
The energy delivered by the arcing discharge can be estimated using equa-
tion 4.3.
EAD = Us · Is · t (4.3)
Where,
EAD = total energy dissipated by the arcing discharge [kJ]
Us = secondary voltage [kV]
Is = secondary current [mA]
t = time period of fault application [s]
Immediately after conducting the arcing discharge faults in the oil tank, gas-
in-oil concentrations were analyzed automatically at 1 hour interval at differ-
ent sampling points (40, 90 and 160 cm) along the main tank (see figure 3.4).
During the generation phase, the main tank was kept disconnected from the
conservator tank, which remained closed until the gas-in-oil concentrations
reached their maximum saturation in oil.
Results
The total energy dissipated during the three arcing discharge faults are pre-
sented in Table 4.12. Figure 4.20 shows the maximum concentrations of the
selected gases measured during these experiments. It can be seen that hy-
drogen and acetylene are the dominant gases in all the cases. In the case
of AD-3 the concentrations are slightly higher than that for the other cases,
which can be associated with the higher amount energy dissipated during
the experiment AD-3.
79
4 Results and Discussions
Table 4.12: Total energy dissipated during simulated arcing discharge (AD) faults
High voltage parameters AD-1 AD-2 AD-3
Secondary current [mA] 156 160 164
Total energy [kJ] 456 456 467
The gas-in-oil concentrations obtained for the arcing discharge faults were
also analyzed using the gas generation pattern method. In all the three arc-
ing discharge experiments the gas concentrations were normalized on basis
of concentrations of hydrogen. The figure 4.21 displays the graphical pat-
terns of gas-in-oil concentrations. It can be observed that the resulting pat-
terns for the arcing discharge experiments are very similar. Thus, it can be
suggested that the arcing discharge faults can be represented by this typical
pattern of the gas-in-oil concentrations.
The comparison of the CIGRE interpretation scheme (table 4.13) with the gas-
in-oil concentrations obtained for the arcing discharge experiments shows
that the key gas ratios and thresholds for all the three arcing discharge ex-
periments indeed fit with the CIGRE criterion. From the CIGRE scheme the
three faults can be interpreted as arcing discharge followed by overheating
of the oil.
80
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
Figure 4.21: Gas concentration patterns obtained for the arcing discharge faults
Table 4.13: Fault diagnostic by applying the CIGRE fault interpretation scheme (arc-
ing discharge)
Key gas CIGRE scheme Arcing discharge
Part I: Key gas ratios
Key gas Ratio Fault Ratio [-]
[-] type AD-1 AD-2 AD-3
(C2 H2 /C2 H6 ) >1 AD 79.0 89.2 87.1
(H2 /CH4 ) > 10 PD 7.3 7.0 6.4
(C2 H4 /C2 H6 ) >1 OH 13.8 13.8 18.1
Part II: Thresholds [ppm]
Key gas Thrsh. [ppm] Fault type Threshold [ppm]
[ppm] type AD-1 AD-2 AD-3
C2 H2 > 20 AD 395 446 455
∑ H2 > 100 PD 475 490 515
∑Cx Hy > 1000 OH 1003 1081 1140
COx > 10000 CD 150 148 156
A D: arcing discharge, PD: partial discharge, OH: oil overheating,
C D: cellulose degradation
81
4 Results and Discussions
Hence, the CIGRE scheme indicates that these faults occurred certainly as
arcing discharge of high energy which produced oil overheating. Though
partial discharge could be also included in this fault interpretation it is dis-
carded due to indication the key gas ratio.
Methodology
In order study the diffusion of the gases from oil to atmosphere four differ-
ent arcing discharge experiments were conducted under identical conditions
as those of the three arcing discharge experiments mentioned in the previous
section 4.2.3. However, these experiments differ with respect to the rate of oil
circulation (mixing) inside the oil tank during the generation and diffusion of
fault gases. Table 4.14 presents the experiment number with its correspond-
ing oil pump rate.
Right after applying the arcing discharge in the oil tank, gas-in-oil concentra-
tions were monitored by DGA online system at 1 hour intervals. The sam-
pling points were located along a vertical distance at 40 cm, 60 cm and 160
82
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
cm measured from the tank cover (figure 3.4). The continuous monitoring
of gas-in-oil concentrations at the specific locations provided concentration-
time profiles which allowed to identify gas generation phase and diffusion
phase.
During the phase of gas generation the main tank remained closed and dis-
connected from the conservator tank. This condition was maintained until
the gas concentrations reached their maximum saturation levels in oil and
quasi-homogeneous concentration in the whole oil volume. In the second
phase the main tank was connected to the conservator. In this phase a pro-
gressive decrease of gas-in-oil concentrations occurs as a result of diffusion
of gases from the oil balance to the atmosphere.
Results
Figure 4.22: Gas-in-oil concentration profile at different locations in the oil balance
(with zero oil circulation)
83
4 Results and Discussions
Figure 4.23: Gas-in-oil concentration profile at different locations in the oil balance
with oil pump circulation 3 l/min
The diffusion or mass transfer rate of gases diffusing from the oil balance to
atmosphere can be associated with the mass transfer coefficient (KOL ). The
84
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
The mass balance of a fault gas-in-oil diffusing into the atmosphere can be
expressed as equation 4.4 [Guo et al., 2002].
( )
dWL CG
= −A · KOL · CL − (4.4)
dt H
Where,
dWL
dt = Mass gradient of fault gas [g/h]
A = surface area of diffusion in the conservator tank [m2 ]
KOL = overall mass transfer coefficient in the liquid phase [m/h]
CL = concentration of fault gas in the oil [µg/m3 ]
CG = concentration of fault gas in the air [µg/m3 ]
H = Henry’s constant [µg/m3 ]
The overall mass transfer coefficient of gases dissolved in oil (KOL ) was de-
termined by means of non-linear regression fitting of time-concentration data
obtained in each experiment of gas generation. This statistical method al-
lowed to model the experimental data by a nonlinear diffusion function based
on a combination of different independent variables. The procedure was ap-
plied to obtain the mass transfer coefficient of critical fault gases such as hy-
drogen, acetylene, and methane, which are produced at the highest concen-
trations.
In order to plot the time-concentration curve for the diffusion phase, the max-
imum concentrations (saturation concentration) were assumed to be the ini-
tial concentrations at start of the diffusion phase. The concentration of the
fault gases in the air was assumed to be zero due to their infinite dilution.
The fitting curve (M Tca ) obtained for the diffusion phase of hydrogen at an
oil pump rate of 3 l/min is presented in figure 4.24. The results for the overall
mass transfer coefficients (KOL(i) ) of the key fault gases i, are summarized in
table 4.15.
The diffusive flux of these gases was determined according to the gradient
flux law which states that the diffusive flux of a given component is a func-
tion of its concentration gradient and mass transfer rate, see equation 4.5
[Scott et. al, 2002].
85
4 Results and Discussions
Figure 4.24: Estimation of mass transfer coefficient (KOL ) for hydrogen (oil pump
rate: 3 l/min)
Table 4.15: Overall mass transfer coefficients (KOL(i) , [m/s]) of fault gases
oil pump rate [l/min]
3.0 1.5 1.0 0
H2 2.4 x 10−2 2.1 x 10−2 1.9 x 10−2 6.9 x 10−3
C2 H2 2.3 x 10−3 3.1 x 10−3 1.3 x 10−3 4.2 x 10−4
CH4 2.1 x 10−2 3.2 x 10−2 1.8 x 10−2 1.9 x 10−3
The diffusive flux of the investigated fault gases for different oil pump rates
is presented in figure 4.25. It can be observed that the diffusion rate of the
gases increases as the the oil pump rate is higher. Irrespective of the oil pump
rates, the diffusion flux of hydrogen is always higher than the diffusion flux
of other gases, followed by the fluxes of methane and acetylene. These ranks
of diffusive fluxes also comply with the solubility of these gases in oil at
normal ambient conditions.
86
4.2 Experimental Investigations of Fault Gas Generation and Diffusion
Figure 4.25: Diffusion flux of fault gases escaping the transformer oil for different oil
pump rates
87
5 Summary and Outlook
5.1 Summary
Efficient and safe functioning of a transformer requires effective and timely
assessment of its condition. A very dangerous condition of the transformer
arise from the fault gas generation as a result of incipient faults and aging
of the insulation materials. Application of the dissolved gas analysis (DGA)
method has been widely recognized as a useful technique to assess fault con-
ditions of a transformer. The transformer oil is a vital information carrier
that provides insight into any malfunctioning of a transformer and it allows
to apply timely any the most suitable corrective measurement. However, an
effective gas-in-oil analysis is essential and its effectiveness is subjected to
the adequate conduction of the DGA method.
Since about four decades, the DGA method has being gaining great attention
as one of the most efficient and convenient methods for fault diagnostic in
transformers. Thus, power agencies has joined forces with the oil diagnos-
tic field in order to develop advanced measurement techniques for gas-in-oil
analysis, as well as to improve statistical interpretation schemes used for the
fault diagnostic.
The main objective of this research project was to evaluate the factors affect-
ing the results of DGA method including the commercially available gas-
in-oil measurement techniques. Additionally, this work focused on investi-
gating the generation of fault gases by subjecting transformer oil to typical
electrical and thermal faults as they occurred in transformers and verifying
those faults by means of the DGA method. The knowledge acquired from
these experimental investigations is deemed to contribute at a reliable fault
interpretation and diagnostic according to the DGA method.
An extensive part of this research work was focused on designing and con-
structing experimental setups to conduct simulations of thermal and electri-
cal faults as they occur in power transformers. The construction of a small
scale setup (30 kV and 12 liter oil tank) and a large scale setup (100 kV and
600 liter oil tank) including oil tanks, oil circulation system, high voltage sys-
tem, and control panel was accomplished with necessary details.
88
5.1 Summary
The small scale setup, consisting of a 12 liter test vessel and a high voltage
supply system of 30 kV, was used to simulate partial discharges fault and
hotspots of different temperatures. Experiments carried out in the small test
setup were aimed at gaining insight about the fault gas generation process
occurring in transformer oil when the oil is subjected to electrical or ther-
mal stresses of different intensities. Moreover, this setup was also useful at
evaluating the generation of gas-in-oil and measurement efficiency of vari-
ous commercially available DGA techniques.
The design of the large scale experimental setup provided an enhanced model
of an air-breathing power transformer and its oil circulation system. This
large scale setup allowed investigations by subjecting a larger volume of oil
(600 liter) to electrical stresses such as arcing discharges of high intensity. The
oil circulation system controlled by an automated system allowed to con-
duct mixing or regeneration of the whole oil volume, resembling the slow
flow rate of oil in a power transformer for cooling purpose. This setup was
used to investigate the fault gas generation in oil due to high intensity arcing
discharge faults. The diffusion flux of gases from oil to atmosphere via con-
servator tank of a transformer was also investigated by means of this setup.
89
5 Summary and Outlook
The analysis revealed that the measurements using gas extraction technique
VE (both online and manual modes) leads to the highest concentrations of
gases, which implies that VE technique has the highest efficiency to extract
gases, followed by EHS and DHS techniques which also lead to relatively
high gas concentrations. However, concentrations achieved by the manual
extraction method, using SS, demonstrated the lowest efficiency for gas ex-
traction. This fact is attributed to the manual generation of the vacuum ex-
traction and several man induced deviations involved in the conduction of
the gas extraction. The evaluation of relative standard deviations (RSTD) of
the measurements suggests that EHS technique has the highest consistency
(RST D < 2.1%) of achieving gas extraction from oil sample. VE technique
also achieves sufficiently high consistency. On the other hand, SS and DHS
techniques are relatively inconsistent (RST D > 5%) in performing gas extrac-
tion.
Sampling and storage of oil: Dissolved gas analysis should be carried out
according to the method proposed in the international standards guidelines
IEC 60567 [IEC 60567, 2005]. However these guidelines are sometimes ne-
glected in the practice, which often leads to exposure of oil sample to air-
trapping, light, high temperature and prolonged storage duration, conse-
quently affecting the quality of DGA results and fault diagnostics.
90
5.1 Summary
Light exposure and temperature variations of oil samples can directly influ-
ence oxidation reactions in the oil. Exposure to sunlight can trigger photo-
chemical oxidation reactions among hydrocarbon molecules and decompo-
sition products, which create further generation of molecular hydrogen. An
exposure to daylight for 8 days at normal ambient conditions causes up to 18
% increase in the concentration of hydrogen.
Stray gassing behavior: Some types of transformer oils, when they are ex-
posed to copper windings in transformer at temperature above 60 ◦ C, exhibit
abrupt increase in concentrations of hydrogen and methane. This behav-
ior was demonstrated though experiments that confirmed the stray gassing
tendency of non-inhibited oils such as Nynas Nytro 10GBN. Based on this
results, it is recommended to consider a possible stray gassing behavior of
an oil during any fault interpretations by DGA.
The partial discharge (PD) faults were generated in the small scale setup.
Since partial discharge is a very instable phenomenon, after carrying out nu-
merous attempts only two consistent PD events of duration 58 hours and
115 hours, at a mean apparent charge of 1000 pC were achieved. The statis-
tical assessment of total number of instantaneous discharges during the PD
events, required for estimation of total fault energy, was not possible due to
technical limitations. Nonetheless, the PD events were assessed on the basis
of phase resolved partial discharge (PRPD). The PD event of 58 hour dura-
tion occurred with high pulse frequency and mean voltage of 11 kV. The PD
event of 115 hour occurred with low pulse frequency and mean voltage of 19
kV, it was characterized by an arcing discharge fault at the end.
91
5 Summary and Outlook
Furthermore, the assessment of the gas generation process using gas genera-
tion pattern method (see section 2.14) suggests that the PD event of 58 hour
duration is indeed a partial discharge fault while the PD event of 115 hour
duration may have been combined with an arcing discharge fault. Thus, it
was demonstrated that partial discharge faults can be detected and distin-
guished using gas generation patterns. On the other hand, the interpretation
of the gas-in-oil concentration using CIGRE fault interpretation scheme led
to uncertain diagnostics of the PD event of 115 hour duration.
The hotspot (HS) faults were investigated using the small test setup, the
hotspots were achieved by powering a semi-conductor material (graphite
pencil lead) at different current levels. Due to technical limitations it was
impossible to measure the actual oil temperature near a hotspot. Therefore
the temperature was estimated as a function of time, current, voltage and
resistance of the semi-conductor material. Thus, it was possible to achieve
hotspots of temperatures approximately 300 ◦ C, 750 ◦ C, 500 ◦ C and 1000 ◦ C
by controlling the current, voltage and resistance.
The assessment of the thermal faults using gas generation pattern method
indicates that the hotspots of above 500 ◦ C temperature are represented by
an identical gas generation pattern, which is identified as a typical gas gener-
ation pattern for a thermal fault. This pattern is characterized by the concen-
tration of ethylene being highest among all the critical gases. On the other
hand, thermal fault at 300 ◦ C was characterized by the concentration of acety-
lene being highest, which is associated with slow oil overheating. The gas
generation pattern produced by this fault indicates an incipient fault evolv-
ing slowly and unnoticeable to become a threatening overheating fault.
The assessment of the hotspots of 300 ◦ C, 500 ◦ C and 1000 ◦ C using CIGRE
fault interpretation scheme lead to misinterpretation of the fault as arcing dis-
charge faults, while the hotspot 750 ◦ C could not be interpreted due to lack
of sufficient criteria. This is caused due to limitations of the CIGRE scheme
92
5.1 Summary
to distinguish the fault on the basis of its dissipated energy. However, the
gas generation patterns obtained for the hotspots show that a reliable diag-
nostic of these faults can be achieved using the typical gas generation pattern.
The arcing discharge (AD) fault was investigated using the large scale setup.
Three discharge fault experiments were conducted by applying primary volt-
age of 380 V and secondary voltage of 95 kV to needle-plate electrodes lo-
cated in the oil tank. During the experiments large concentrations of hy-
drogen and acetylene, along with minor quantities of ethylene and methane
were generated due to breakdown of hydrocarbon molecules under high in-
tensity current and temperature of the arcing discharge fault.
The assessment of the arcing discharge faults using gas generation pattern
method showed that the three faults produce very similar gas generation
pattern. This pattern can be recognized as a typical gas generation pattern
for arcing discharge faults. The CIGRE fault interpretation scheme also pro-
vided a coherent interpretation of the faults as arcing discharge faults with
subsequent oil overheating due to high amount of dissipated energy.
Diffusion of fault gases to atmosphere via conservator is one of the most crit-
ical factor that can cause deviations in DGA. The diffusion of the gases can
result in reduced gas-in-oil concentrations, consequently leading to false in-
terpretation of the faults.The investigations were focused on the diffusion
flux of the three critical fault gases: hydrogen, methane and acetylene. These
gases were chosen because they are usually found in largest concentrations
in oil, as well as they are distinguished by their solubility in oil. Four dif-
ferent experiments were conducted using different oil circulation rates. The
concentrations of the gases were monitored for a certain period after the con-
ductions of the faults. The mass transfer coefficients for the gases were ob-
tained from the concentration-time profiles and the corresponding diffusion
flux was estimated using the gradient flux law.
Highest diffusion flux was obtained for hydrogen, followed by methane and
then acetylene, which implies that the diffusion flux depends on the solubil-
ity of the gas. Furthermore, it was demonstrated that the diffusion flux of
gases is strongly dependent on the rate of oil circulation between the main
tank and the conservator, higher the circulation rate stronger the diffusion
flux of the gases.
93
5 Summary and Outlook
• The factors related to sampling and storage of oil samples, namely air-
trapping, light exposure and temperature, have significant influence on
gas-in-oil concentrations. Therefore extreme care should be taken in re-
gard with the procedures of oil sampling and storage of samples. How-
ever, samples stored under adequate conditions can guarantee reliable
measurements of gas-in-oil concentrations up to 16 days.
• Partial discharge faults are a low intensity faults which occur at low
temperatures, however in some cases they can be accompanied by arc-
ing discharges producing large concentrations of hydrogen, methane
and acetylene. The thermal faults due to hotspots above 500 ◦ C gener-
ate ethylene, methane, hydrogen and acetylene in large amounts. The
amount of generated gases increases rapidly for temperatures above
1000 ◦ C. Arcing discharge faults involve much higher amount of en-
ergy dissipation, producing the largest concentrations of hydrogen and
acetylene.
94
5.2 Outlook
5.2 Outlook
Based on the experience gained during these investigations, following sug-
gestions can be pointed out for future research work and practical application
of DGA.
• For the further investigations related to the fault gas generation and
their evaluation by DGA, it is recommended to add other factors affect-
ing the gas generation process, such as the presence of insulation paper,
fluctuation of transformer voltage load and oil temperature.
95
5 Summary and Outlook
96
List of Symbols and
Abbreviations
List of Symbols
97
List of Symbols and Abbreviations
List of Abbreviations
AC Alternating current
DC Direct current
HV High voltage
PD Partial discharge fault
AD Arcing discharge fault
HS Hotspot fault
OH Oil overheating fault
CD Cellulose degradation fault
FR Oil flow rate
ASTM American Society for Testing and Materials
IEC International Electro-technical Commission
CIGRE Conseil international des grands réseaux électriques
(The International Council on Large Electric Systems)
TF Taks Force of CIGRE
DGA Dissolved gas analysis
VE Vacuum extraction
SS Shaking syringe
EHS Equilibrium head space
DHS Dynamic head space
RSTD Relative standard deviations
MODBUS Automation application for communication between devices and
data exchange
98
List of Symbols and Abbreviations
99
References
Acheson, 1990: Acheson, D. J., 1990. Elementary Fluid Dynamics. Oxford Ap-
plied Mathematics and Computing Science Series, Clarendon Press, England.
Aragón et. al., 2007: J. Aragón P., M. Fischer, S. Tenbohlen, 2007. Improvement
of dissolved gas analyses (DGA) by means of experimental investigation of generated
fault gases and a fuzzy logic based interpretation scheme. Proceedings of the 15th
International Symposium on High Voltage Engineering, Slovenia, paper no.
T7-157.
Aragón et. al., 2007: Aragón P. J., Tenbohlen S., Bräsel E., Sasum U., 2007.
Improved Monitoring of Dissolved Transformer Gases on the Basis of a Natural In-
ternal Standard (NIS). Cigre A2/D1 Colloquium, Belgium.
ASTM D 3612, 2004: ASTM International Standard, 2004. Test Method for
Analysis of Gases Dissolved in Electrical Insulating Oil by Gas Chromatography,
Annual Book of ASTM Standard, Vol. 10.03.
ATKINS, et. al., 2006: Atkins P. and Paula J. De., 2006. Atkins’ Physical Chem-
istry, Oxford University Press, Oxford, UK.
Battino et. al., 1965: Rubin Battino, H. Lawrence Clever. The Solubility of
Gases in Liquids. Illinois Institute of Technology, Chicago, Illinois, Received
October 25.
Benner et. al., 1982: Benner C.O., Myers J.E., 1982. Momentum, heat and mass
transfer. 3rd Edition. MacGraw-Hill, New York.
100
References
Borsi et. al., 1999: Borsi H., Dumke K., Gockenbach E., 1999. Relationship
between faults and generated gases in transformer liquids. Proceedings of 13th
International Conference on Dielectric Liquids, Japan.
Borsi et. al., 1999: Borsi, H., K. Dumke, E. Gockenbach, 1999. Relationship be-
tween faults and generated gases in transformer liquids. Proceedings on the 13th
International Conference on Dielectric Liquids, Nara, Japan.
Cargol, 2005: Tim Cargol, 2005. An overview of online oil monitoring technolo-
gies. Fourth Annual Weidmann-ACTI Technical Conference, San Antonio,
U.S.A..
Coates et. al., 1986: John P. Coates Lynn C. Setti, 1986. Infrared Spectroscopic
Methods for the Study of Lubricant Oxidation Products, A S L E Transactions,
29:3, pp. 394-401
Coates et. al., 2008: Coates John P. and Lynn C. Setti, Infrared spectroscopic
methods for the study of lubricant oxidation products. ASLE Transactions, 29:3,
pp 394-401.
Danikas, 1993: Danikas M., 1993. The Definitions used for Partial Discharge
Phenomena. IEEE Transactions on Electrical Insulation Vol. 28, No. 6, pp.
1075-1080
101
References
Duval, 1989: Duval, M., 1989. Dissolved Gas Analysis: It can save your trans-
former. IEEE Electrical Insulation Magazine, Vol., 5 (6).
Duval, 2002: Duval M., 2002. A Review of Faults Detectable by Gas-in-Oil Anal-
ysis in Transformers. IEEE Electrical Insulation Magazine, Vol. 18 (3).
Duval, 2005: Duval M., 2005. Improving the Reliability of Transformer Gas-in-
Oil Diagnosis. IEEE Electrical Insulation Magazine, Vol. 21 (4).
Drews, 1998: Drews A. W., 1998. Manual on hydrocarbon analysis. ASTM In-
ternational, 6th edition, pp. 20.
Fischer and Tenbohlen, 2009: Fischer M. and Tenbohlen S., 2009. Improved
condition assessement by fuzzy-modelling, adjustment and merging of DGAs Inter-
pretation methods. Proceedings of the 16th International Symposium on High
Voltage Engineering, Johannesburg, 2009.
Florkowska et. al., 1999: Florkowska B. and Zydron P., 1999. The Correlation
of PD-Charge and Energy Patterns in the Study of Partial Discharge Mechanism in
Thermostatting Insulation. High Voltage Engineering Symposium, 22-27, Con-
ference Publication No 467.
Fofana et. al., 2007: Fofana, I., Borsi, H., Gockenbach, E. and Farzaneh, M.,
2007. Aging of transformer insulating materials under selective conditions. Euro-
pean Transactions in Electrical Power, Vol. 17, pp. 450470.
Fofana et. al., 2008: Fofana, I.; Sabau, J.; Bussieres, D.; Robertson, E.B., 2008.
The mechanism of gassing in power transformers. IEEE International Conference
on Dielectric Liquids, 2008. pp.1-4.
102
References
Fruth et. al., 1992: Fruth, B.; Niemeyer, L., 1992. The importance of statistical
characteristics of partial discharge data. IEEE Transactions on Electrical Insula-
tion, Vol.27, no.1, pp.60-69.
Gaffuri et. al., 1997: Gaffuri, P., Faravelli, T., Ranzi, E., Cernansky, N. P.,
Miller, D., d’Anna, A. and Ciajolo, A. (1997). Comprehensive kinetic model for
the low temperature oxidation of hydrocarbons. American Institute of chemical
engineers (AIChE) Journal, Vol. 43, pp. 12781286.
Gibeault, 1997: Gibeault J.P., 1997. On-line monitoring of key fault gases in
transformer oil: an operational experience accumulated over the years. 10th Inter-
national Symposium on High Voltage Engineering (I.H.S.), Canada.
Griffin, P. J., 1998: H.G. Erdman, 1988. Criteria for the Interpretation of data for
dissolved gases in oil from transformers. Electrical Insulating oils, STP 998, pp.
89-106
Guo et. al., 2003: Guo Z.; N. F. Roache, 2003. Overall mass transfer coefficient
for pollutant emissions from small water pool under simulated environmental condi-
tions. Oxford Journals, Life Sciences Medicine, The Annals of Occupational
Hygiene, Vol. 47, Issue 4, pp. 279-286.
103
References
Ioffe et. al., 1984: B. V. Ioffe; A. G. Vitenberg, 1984. Head-Space Analysis and
Related Methods in gas Chromatography. Wiley-Interscience publication, U.S.A.
ISBN 0-471-06507-2.
Jakob et. al., 2012: Jakob, F.; Noble, P.; Dukarm, J.J., 2012. A Thermodynamic
Approach to Evaluation of the Severity of Transformer Faults. IEEE Transactions
on Power Delivery, Vol.27, no.2, pp.554-559
Khaled et. al., 2009: Khaled Shaban, Ayman EL-Hag and Andrei Matveev,
2009. Predicting Transformers Oil Parameters. IEEE Electrical Insulation Con-
ference, Montreal, Canada.
Khaled, 2009: Khaled Shaban, Ayman EL-Hag and Andrei Matveev, 2009.
Predicting Transformers Oil Parameters. IEEE Electrical Insulation Conference,
Montreal, Canada, pp. 196-199.
Krawiec, 2009: Steve Krawiec, 2009. Review of recent changes to mineral insulat-
ing oil specifications. Electrical Insulation Conference, 2009. IEEE, pp. 363 367
Leblanc et. al., 1993: Y. Leblanc, R. Gilbert, M. Duval, and J. Hubert, 1993.
Static headspace gas chromatographic determination of fault gases dissolved in trans-
former insulating oils. Journal of Chromatography, Vol. 633, pp.185-193.
Lipshtein et. al., 1970: Lipshtein R.A. and M. Shakhnovich, 1970. Transformer
oil. 2nd edition, Israel Program for Scientific Translation, Jerusalem.
104
References
Muhr et. al., 2006: M. Muhr, R. Schwarz, 2006. Partial discharge measurement
as a Diagnostic Tool for HV-Equipments. Properties and applications of Dielec-
tric Materials, 2006. 8th International Conference on, IEEE, pp.195-198.
Murphy, 1970: Murphy E. J., 1970. The generation of gases in the dielectric break-
down of oil. J. Phys. D: Appl. Phys. Vol. 3, pp. 917-923.
NCho et. al., 2011: NCho J. S., I. Fofana, A.Beroual, T. Aka-Ngnui and J.
Sabau, 2011. The gassing tendency of various insulating fluids under electrical dis-
charge. IEEE Trans. Elec. Insul., Vol. 18, pp 1616-1625
Narang et. al., 2012: Ena Narang, Er. Shivani Sehgal, Er. Dimpy Singh, 2012.
Fault detection techniques for transformer maintenance using dissolved gas analy-
sis. International Journal of Engineering Research Technology (IJERT).Vol. 1
Issue 6.
Sabau et. al., 2000: John Sabau and Rolf Stokhuyzen, 2000. The Side Effects
of Gassing in Transmission Power Transformers. IEEE Conference on Electrical
Insulation and Dielectric Phenomena, pp. 264-267.
105
References
Sangui, 2003: Sanghi, R., 2003. Chemistry behind the life of a transformer. Reso-
nance Magazine, Vol. 8(6), pp. 17-23
Scott et. al., 2002: Scott A. S. and G. H. Jirka, 2002. Environmental fluid mechan-
ics Part I: Mass transfer and diffusion. 2nd Edition, Institut für Hydromechanik,
Universität Karlsruhe, Germany.
Sherif, et. al, 2012: Sherif S. M. Ghoneim, Sayed A. Ward, 2012. Dissolved
Gas Analysis as a Diagnostic Tools for Early Detection of Transformer Faults. Ad-
vances in Electrical Engineering Systems (AEES) Vol. 1, No. 3, pp. 152-156.
Shirai et. al., 1997: Shirai, M.; Ishii, T.; Makino, Y., 1977. Evolution of Hydrogen
from Insulating Oil in Transformers. IEEE Transactions on Electrical Insulation,
Vol.EI-12, no.4, pp. 266-272
Shirai et. al., 1997: Shirai, M.; Shimoji, S.; Ishii, T., 1977. Thermodynamic Study
on the Thermal Decomposition of Insulating Oil. IEEE Transactions on Electrical
Insulation, Vol. EI-12, no. 4, pp. 272-280.
Sing et. al., 2010: Sukhbir Sing, M.N. Bandyopadhyay, 2010. Duval Trian-
gle: A Noble Technique for DGA in Power Transformers. International Journal of
Electrical and Power Engineering, Vol. 4, pp. 193-197
Sokolov, et. Al., 2001: Sokolov, V.; Bulgakova, V.; Berler, Z., 2001. Assess-
ment of power transformer insulation condition,” Proceedings Conference on
Electrical Insulation Conference and Electrical Manufacturing Coil Wind-
ing, pp.605-613
Sokolov, et. Al., 2001: Sokolov V., Berler Z. and Rashkes V., 1999. Effective
methods of assessment of insulation system condition in power transformers: A view
based on practical experience. Proceedings Conference on Electrical Insulation
Conference and Electrical Manufacturing Coil Winding, pp.659 -667
Sun et. al., 2012: Huo-Ching Sun, Yann-Chang Huanga, Chao-Ming Huangb,
2012. A Review of Dissolved Gas Analysis in Power Transformers. Energy Proce-
dia 14, pp. 1220 1225
106
References
Van Brunt, 1994: Van Brunt, R.J., 1994. Physics and Chemistry of Partial Dis-
charge and Corona, Recent Advances and Future Challenges. IEEE Trans. on
Dielectrics and Electrical Insulation, Vol. 1(5).
Wang et. al., 2012: Zhongdong Wang; Xiao Yi; Jinping Huang; Hinshaw; J.V.,
Noakhes, J., 2012. Fault gas generation in natural-ester fluid under localized ther-
mal faults. IEEE Electrical Insulation Magazine, Vol. 28, no. 6, pp. 45-56.
107
Erklärung
Hiermit versichere ich, dass ich die vorliegende Arbeit mit dem Titel:
„Experimental
Investigations on the Dissolved Gas
Analysis Method (DGA) through Simulation of Electrical and Thermal Faults
in Transformer Oil”
selbst verfasst und keine außer den angegebenen Hilfsmitteln und Quellen benutzt
habe, und dass die Arbeit in dieser oder ähnlicher Form noch bei keiner anderen
Universität eingereicht wurde.
"The CV is not included in the online version for reasons of data protection."