M. Pompili
Department of Electrical Engineering of University of Roma La Sapienza
00184 Roma (Italy)
and R. Bartnikas
Institut de Recherche dHydro Quebec,
Varennes, Quebec, J3X 1S1, Canada
ABSTRACT
The behavior of corrosive sulfur in mineral oils is examined in terms of the failures
observed in transformers, the surfaces of the copper sulfide covered conductors and
degraded paper insulating tapes. The role of Dissolved Gas Analysis (DGA) in the
evaluation of the risk of copper sulfide formation is described. The degree of
corrosiveness of some sulfur compounds is examined and compared using a Kraft
paper wrapped-copper test (Standard IEC 62535). The occurrence of DBDS as the
most relevant corrosive compound is compared with the presence of other corrosive
species in insulating mineral oils. A number of mitigation techniques for corrosive
sulfur are described and evaluated.
Index Terms Corrosive sulfur, transformer failures, mineral oils, corrosion
mechanism, dielectric losses, corrosive sulfur mitigation.
forming the walls of the occluded cavities would tend decomposition point. Moreover, low temperature thermal
to reduce the voltage gradients across the cavities that faults usually affect a larger portion of the power apparatus,
are necessary to initiate and sustain the partial thereby leading to a more diffused deposition of conducting
discharge process. copper sulfide within the overall insulating system
structure.
2 TRANSFORMER FAILURES
One of the most common types of corrosive sulfur
related transformer failure in the field are found to be
due to high temperatures in the windings, commonly Evidence of general
resulting from inadequate cooling of the windings or overheating in the
design defects. The occurrence of copper sulfide on the faulty winding.
copper conductors and the Kraft paper tapes adjacent to
them represents possible sites of overheating in either an
otherwise healthy winding or an additional source of
overheating in an already faulty winding. Examination of
failed transformer windings provides evidence that many
failures occur either in close vicinity to the top portion of
the winding where the temperature is highest, or in zones
of the winding where insufficient cooling results in a
higher degree of cellulose chain scission than what can
be anticipated on the basis of longevity of the unit in
service. These high temperature hot spots frequently
reveal copper sulfide deposits (cf. Figures 1 and 2),
strongly suggesting that in these failures, corrosive sulfur Copper sulfide found
is either the main cause of the fault or in the least an on conductors
important contributing factor.
Tables 1 and 2 refer respectively to DGA (Dissolved
Gas Analysis) data obtained prior to failure for the units,
whose failures are depicted in Figures 1 and 2. The
Figure 1. Copper sulfide deposit on a copper conductor at a failure site on
dissolved gas analysis data clearly implies overheating the winding of an industrial distribution transformer (23 kV / 4 MVA)
(fault type T1-T2 according to Guidance IEC 60599); this
would suggest that DGA, even if it is not capable to
intercept the electrical fault that leads to the failure of the
unit (probably due to the rapid development of the fault in
its final stages), can be an effective technique to assess
the risk of dangerous formation of copper sulfide in units
impregnated with corrosive sulfur containing oils. A
condition of overheating detected in transformers or shunt
reactors in service, whose oil has been found to be
corrosive according to Method IEC 62535 or Method
ASTM D 1275 B, should thus be considered as a
symptom of possible copper sulfide formation either on
bare copper surfaces or on the paper tapes wrapped onto
the conductors.
Thermal faults at lower temperatures (T1 according to
Guidance IEC 60599, i.e. T< 300 C), that are usually
deemed less dangerous than thermal faults at medium and Conductors meltdown in a
high temperatures for the safety of the unit, should be taken failed winding
Table 2. DGA data obtained prior to a corrosive sulfur related failure (see The accepted scale of corrosiveness places elemental sulfur
Figure 2); the ethane, methane and O2/N2 ratios indicate an overheating as being the most corrosive element against copper; this is
condition (fault type: T1, T < 300 C according to IEC 60599). followed in descending order by hydrogen sulfide,
4 months before the 10 months before the mercaptans, sulfides and disulfides. Whilst sulfur organic
Gas
failure (l/l) failure (l/l)
compounds, wherein the sulfur atom is part of the aromatic
Oxygen 384 468
Nitrogen 67293 67440 structure (e.g. thiophenes and related compounds) are
Carbon dioxide 4993 4223 normally considered as non-reactive. Standard IEC 62535 [5],
Carbon monoxide 704 479 which is intended to test for corrosive sulfur and take into
Hydrogen 154 205 account the effect of the paper tapes (as opposed to ASTM
Methane 89 103
Method D 1275B [6] that excludes the involvement of the
Ethane 216 204
Ethylene 95 131
paper insulating tapes), has been found not to respect the
Acetylene <1 <1 conventionally accepted scale of corrosive sulfur reactivity.
This test, in which the specimen consists of a piece of copper
conductor covered with a single turn of paper tape insulation,
3 CORROSIVE SULFUR EFFECTS would indicate, as is substantiated in Figure 5, that alkyl
Corrosive sulfur leads to the formation of substances that mercaptans produce either no or negligible indication for
are insoluble in oil, precipitate as sludge or form copper corrosive sulfur.
sulfide deposits on the copper conductors and paper insulating
tapes. In transformer and shunt reactor windings, copper
sulfide growth is accompanied by its migration within the
layers of the insulating paper. In most of the observed cases,
the copper sulfide concentration tends to be highest within the
inner tape layers adjacent to the copper conductors, with a
tendency to diffuse outwards; this process is strikingly in
evidence in Figure 3, which was obtained on Kraft paper tapes
retrieved from an actual transformer.
A plausible explanation for the observed decrease in the
concentration of copper sulfide towards the outer layers is that
it is due to the higher concentration gradient of copper sulfide
at the surface of the copper conductor. In addition, there is
also an increased temperature gradient effect at the copper
conductor surface or in its vicinity, as a result of the high
currents carried by the copper conductor, which will augment
the reaction rate of the corrosive sulfur with the copper to
form copper sulfide. It can be normally anticipated that the
Kraft paper will tend to be darkened more at elevated
temperatures as a consequence of the oxidation of cellulose
papers (c.f. Figure 4). Yet, it is also frequently observed that Figure 4. Copper sulfide deposits on a conductor and the adjacent paper
the outer layer tapes may became even more darkened due to tape of a winding in a 130 kV/130 MVA transformer, where symptoms of
the oxidation of the oil itself. overheating are also substantiated by DGA measurements.
1658 R. Maina et al.: Corrosive Sulfur Effects in Transformer Oilsand Remedial Procedures
4 CORROSION MECHANISM
In contradistinction, dibenzothiophene, which is normally
taken as being inert, produces a corrosive indication. As to be Among corrosive sulfur compounds, dibenzyl disulfide
anticipated, the IEC test is an effective test for sulfides and (DBDS) appears to play a predominant role in the corrosion
disulfides for which it was intentionally devised. Evidently, process. Identified as a major sulfur compound in several
the apparent and partially unexpected changes in the scale of insulting oils [11], it is present in approximately 90% of the
reactivity obtained with the IEC test are to be attributed to its insulating oils employed within the European market.
intended scope, i.e., to identify oils that are capable to transfer Considering 20 mg/kg of DBDS as the threshold to render an
copper sulfide onto the insulating papers and not to assess the oil corrosive, 89.7% of the oils in this population derive their
reactivity of the oil on copper. This provides important test corrosiveness from DBDS (cf. Figure 6). Hence it is not
data, because the deposited semiconducting copper sulfide surprising, that most of the investigative efforts on the
enhances the conductivity and dielectric loss of the paper corrosiveness of the mineral insulating oils have been carried
tapes, thereby creating the possibility of thermally unstable out on DBDS.
regions and hot spots along the paper tapes that eventually It has been postulated that a more reactive corrosive species
may lead to thermally induced dielectric breakdown of the can result when DBDS is reduced to benzyl mercaptan [11]
insulating systems in transformers and shunt reactors [3, 4]. with a consumption of a stoichiometric hydrogen.
One can hypothesize that most likely the chemical reaction Alternatively, it has been suggested the copper sulfide may be
of sulfur with copper cannot per se account entirely for the formed directly upon the surface of the Kraft paper, whereby
capability of the oils to transfer copper sulfide into the the copper is transported as a complex of DBDS, resulting in
cellulose fibers of the paper tapes. There have been by-products such as bis-benzyl and dibenzyl sulfide [7].
suggestions that certain specific corrosive sulfur compounds Traces of the by-products identified in [7] and [11] have been
may play a significant role in the transfer of copper in the form detected in mineral oils that are currently in service.
of complex from the copper conductor surface unto the paper It is interesting that a minor portion of the insulating oils are
[7, 8]. Our own studies have indicated that the movement of found to be corrosive independently of the presence of DBDS.
copper particles in this elemental form can occur within the These oils exhibit a propensity to form the copper sulfide
oil. This would infer that copper particles may enter and observed on the surface of the Kraft paper tapes; yet the
become trapped between the interstices of the paper fibers and compounds responsible for this corrosion process remain to be
at its surface. When in contact with corrosive sulfur oil, these identified, though some studies indicate that they should be
particles would combine with the corrosive sulfur to form sought among the disulfides [12]. Figure 7 delineates the total
copper sulfide [9]. These copper particles may occur in disulfide content in given corrosive oil observed in our
mineral oils in neutral or ionic state in the form of large laboratory in which the oil has been exposed to a wrapped
micelle-like structures [10]. In neutral form, movement of copper conductor (conditions are the same as those in Method
copper particles through the relatively open fibre structure of IEC 62535) at 150 C for increasing amounts of time. The
Kraft paper could result from Brownian movements and from resulting deposition of copper sulfide is readily discernible
their concentration gradient at the copper surface. upon the Kraft paper tapes as portrayed in Figure 8. The
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 16, No. 6; December 2009 1659
nature of the shiny surface deposits was identified by means of well as the shape of the tank. In addition, prior to changing
SEM/EDX (Scanning Emission Microscopy/Energy the oil, a test of corrosive sulfur on a suitable mixture of the
Dispersion X-Ray) technique; the results are shown in Figure old and new oils should be made to ascertain whether the
9, where the peaks of copper and sulfur can be readily mitigation effort is adequate to provide a non corrosive oil.
recognized. As we have pointed out earlier, it is these deposits Corrosive sulfur may be removed from the oil by means of
of semiconducting copper sulfide that augment significantly chemical treatment. There are four common methods
the dielectric losses of the tapes [4]. available, namely: oil depolarization, traditional oil
reclamation, solid/liquid (S/L) extraction, and liquid/liquid
(L/L) extraction. Oil depolarization has been employed
5 CORROSIVE SULFUR MITIGATION both off-load and on-load to reduce the corrosive sulfur
Several mitigation techniques may be employed to reduce content in transformers. It uses a combination of solid
the impact of corrosive sulfur in transformers and reactors. reagents, chemicals and sorbents and is capable of reducing
The most obvious approach is to use metal surface passivators the DBDS content down to a level of < 5 mg/kg. Figure 12
to provide a film cover over the copper conductor to protect its depicts the rate of decrease of corrosive sulfur as a function
surface. A common metal passivator in use is that of of the number of the treatment cycles.
toluitrazole-dialkyl-ammine (commercially available as
Irgamet39) that is usually added to new and used transformer
oils in concentration levels of approximately 100 mg/kg. The
purpose of the metal passivator is to protect the copper
conductor by forming a molecular layer onto the copper
surface; however, it does not remove the corrosive compounds
from the oil and it evinces some unexpected and undesired
collateral effects, such as the release of hydrogen and carbon
oxides. This often leads to erroneous gas evolution test results.
Figure 10 portrays hydrogen gas evolution in a transformer
prior to and following metal passivation; the introduction of a
metal passivator is seen to lead to a very marked increase in
gas generation.
Perhaps one of the most adverse effects of thermal stress
and oxidation by-products of the oil is the ensuing
instability of the metal passivator. Metal passivator
depletion during service is observed in highly loaded power
transformer units, containing oils where the oxidation
process has already commenced. An initial depletion of the
metal passivator is to be anticipated in 10-15% of the units Figure 6. Occurrence of DBDS in 242 corrosive oils.
passivated during the service; on the other hand,
transformers initially impregnated with a passivated oil
appears not to be affected by problems of abnormal
passivator consumption. A disconcerting effect of metal
passivator depletion is illustrated in Figure 11, where it can
be perceived that a reduction below a concentration of 25-
30 mg/kg can cause the insulating oil to revert back to a
corrosive state; surface discoloration of the copper
specimens demonstrates the growth of the copper sulfide
films on the copper conductor specimens.
One obvious mitigation technique, that is commonly
used, consists of simply replacing the existing oil in a
transformer or shunt reactor unit by a new, non corrosive
oil. Evidently in this approach, the corrosive compound
content is reduced by diluting the existing oil according to
the ratio between the volume of the new oil added and the
remnant volume of the old oil (i.e., the existing remaining
oil at the bottom of the tank and windings of the specimen
unit). The latter ratio may vary from 5 to 12% depending Figure 7. Decay of total disulfides during exposure of a corrosive oil to
on the amount of paper and other absorbing materials as copper and paper at 150 C.
1660 R. Maina et al.: Corrosive Sulfur Effects in Transformer Oilsand Remedial Procedures
140
80
60 Corrosive
Non Corrosive
40
20
Non Corrosive
0
24/03/06
13/05/06
02/07/06
21/08/06
10/10/06
29/11/06
18/01/07
Sampling date
Figure 11. Decay of the metal passivator and its effect on corrosiveness of
Figure 8. Increasing deposition of copper sulfide on copper conductors
the oil.
(left side specimen) and kraft paper tapes (right side specimen) as a
function of time.
160
Concentration (mg/kg)
140 DBDS
Irgamet 39
Copper 120
100
80
Sulfur
60
40
20
0
0 1 2 3 4 5 6 7 8 9 10
Number of treatment cycles
Figure 9. SEM/EDX spectrum of deposits formed on Kraft paper by a
corrosive oil not containing DBDS. Figure 12. Removal of DBDS by depolarization of the oil
Gas evolution (ml/l)
(l/l)
Hydrogen As concerns the reduction of corrosive sulfur compounds
Passivation
trend
in oils, it should be borne in mind, that in the case of DBDS
containing oils, the reduction also involves necessarily the
residual oxidation stability of the oil. It has been
demonstrated that DBDS acts as an oxidation inhibitor, i.e.,
as a secondary antioxidant, with the consequence that its
Date of measurement (month/year)
removal may require the necessity to re-inhibit the oil with
Figure 10. Example of stray gassing following the passivation with
Irgamet39. a classical cresolic antioxidant, such as di-tert-butyl-para-
cresol (DBPC). Table 3 provides some typical values of
oxidation stability prior to and following depolarization as
The traditional on-line method has been originally well as those obtained after re-inhibition.
developed to reduce the polar compound concentration in
mineral oils, using fuller earths and clays. However, it was
also found to be effective in reducing the corrosive sulfur Table 3. Oxidation stability of depolarised oil before and after the re-
content. The solid/liquid method involves the forcing of inhibition with di-tert-Butyl-para-Cresol (DBPC)
oils through a suitable sorbent to reduce the corrosive
Oxidation After re-
sulfur content, whilst the liquid/liquid corrosive sulfur After on-Load IEC 60296
stability inhibition with
depolarization reference
extraction method is based on the selectivity of certain oil- parameters,
(test duration:
DBPC 0.3%
values (for
immiscible solvents, which are placed in contact with the according to (test duration:
164 h) unused oil)
IEC 61125 500 h)
oil in a separate cistern to extract the corrosive sulfur
sludge (%) 0,91% 0,76 < 0,80
compounds. However, this system of corrosive sulfur
extraction has not been validated in industry. TAN (mg/g) 1,55 0,78 < 1,2
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 16, No. 6; December 2009 1661
Corrosive
Corrosive
system employed for the removal of DBDS from Cigre's CCD
4c
transformer specimens under on-load conditions. It can be 4b
also employed to remove the metal passivator added to the 4a
3b
oil to protect the copper conductors surface against 3a
2e
corrosive sulfur. 2d
2c
Max 800 lt/h 2b
Non Corrosive
Non Corrosive
2a
1b
1a
0,14 tan
tan delta
that becomes insufficient to protect the copper, the TAN
0,12
corrosiveness is augmented.
0,1
The data in Figure 14 would suggest that the ASTM
0,08
method is more sensitive to corrosive sulfur than the IEC
method, which is directly attributed to the two different 0,06
15
effective corrosiveness of the oil. In order to take
10 preventive measures against the corrosive sulfur in the
IFT (10