Degradation of cellulosic insulation in power

transformers. Part 3: Effects of oxygen and water on

ageing in oil
A.M.Emsley,X.Xiao, R.J.Heywood and M.AIi

Abstract: Heat, water and oxygen accclcrate the degradation 01' cellulose insulation in eleclrical
tl.ansrormers. Their effects on insulation have bccn studied over a number of years using degrcc or
polymcrisation (DP) of tlie paper a s a iiicasurc of agcing. Results arc reported of a systemic study to
nieastirc the relative cl'fcctivemss of Cach component individually and in combination, on ageing i n oil
in a three-piiramcter, two-lcvcl, partial factorial experiment. Ageing was measured in terms of change
of degree of polymerisation of the paper and analyscd according to recently developed models. In
addition, Lhc concenlrations of furlhral-hised degradation products in tlie oil wcrc incastired Water
and Lcmpertiturc arc most effective in accelelating ageing, with oxygen about one-third as elficicnt.
There is a strong synergistic effcct between tenipcrature and water, a weaker synergism between
temperature and oxygen, but, importantly, a11 apparent anlagonistic effect bctwccn water and oxygcn,
at low water levels. I T correct, the aiitagonisni implies hiit the effectiveness of watcr, as an acmlcrator
of ageing, decreases with incrcasing levels oroxygeii in the oil at low watcr levels, which could explain
why the insulation in scrapped transformers is sometimes found to be iii very good condition.

1 Introduction exponcnlially to a inaxiiiiuin valuc and then decrcascd,

when the paper was fully degraded. It was therefore pro-
A high-voltage power tl.ansfornier. is cxpected to operate posed that fiirfural analysis or oil WAS likely to bc parlicu-
reliably for up to 40 years. The onset of dccrcasing reliabil- larly uscrul as a trouble-shooting tool Ibr the detection of
ity is associated with a variety ofcauscs, many of which are high mtcs or ageing of papcr. In this paper, the cll'ects arc
design specific. However, the underlying, life-limiting Tal- investigated of temperature, water in the paper, and oxygen
Lure of all designs is related to the thermal degriidiition oi in the oil, on the kinetics of ageing (as ineasured by change
paper insulation on the windings [I,21. The paper contains of DP) and on the Tormation of furan-based ageing prod-
about 90% oT ccllulose, 6 7 ' X of lienii-cellulose and 3 4 % ucts.
of lignin, and it deteriolalcs slowly as the constituent largc
niolecukir chains break down during service. The process is 1. I Effects of temperature, water and oxygen
accelerated by heat, moisture and oxygen, which arc all on insulation life
present in an operating transformer. Thcrc is no iiiformatioii in the literature on the cll'ects of
I n a previous paper [3], studies have been reporled 01' the moisturc or oxygen on furfural formation, but Einsley and
[ormation oT rurfiiral products during the ageing of dried Slcvcns [I] rcvicwcd the literature of tlie effects of tempera-
Kraft and cotton insulation paper in dry iiisulating oil ture, oxygen and water on the ageing oT translbrmer insula-
under acceleralcd laboratory conditions. It was dernon- tion generally. The following provides B short resume with
strated that 2-furaldehyde was the mas1 significant product key references.
or paper ageing detcctable in the oil and that the concentra-
tions of all fllran-rebated products inilially iiicreascd to a 1. I . I Effects of temperature: Degradation is a chcmi-
low plateau concentration in the oil. As ageing proceeded cal reaction and as such should obey the Arrlicnius theory
and the DP of the paper dropped below about 400, the of reaction kinetics, which relates tlie log or tlie reaction
con~ntrationof ageing products i n the oil began to rise rate constinit to the rcciprocal oT tlie absolute temperature.
Values Tor the activation energy of degradation of ccllulose
0WE, 20(X) in oil, quoted in the literature, range, from of the order of
IEE Pmw~v/ingsoiiliiic 110. 20000021 85kJ/niolc for the oxidalivc degradation rcaction, to IZOltJi
D M IO. 1C49/ip~sml:2tMooo2 I molc for the Iiydrolylic degradation rcaction 1661. Above
1':ipcs ilreival 14111 Scptcnclnkr I999 about 140"C, the rate or dcgt.adation 1x1sbccii reported to
A.M. Emslcy is will? thc School of Pliysia ;and Chcmislty, Univcrsily of increase significantly implying a possible change in mccha-
Surccy, Guildford. Sunzy, C U 2 5x11,U K nisni [7-12].
Ernslcy and Stcvcns [I21 showed that most of the
reported data can he approximately represented by the
Elmistam equation [ 131

IIS
where DP = degree of polymerisation at time t, DPo is the response variable to assess the effectiveness of individual
initial DP,k is a constant and t the ageing time. parameters and synergistic and antagonistic interactions
The activation energy for k was shown to be 113kJ/mole, between them. The basic statistical principles, based on so-
irrespective of the environmental conditions (oxygen, called Yates tables, can be found in the literature [19]. The
water) or type of cellulose. However, Emsley, Heywood three variables were heat, oxygen level in the oil and water
etal. [14] have since shown that k is not constant, but content of the paper; the response variable used was the k2
decreases with ageing time and that eqn. 2 represents the parameter of eqn. 2. In addition, some experiments were
data more accurately: carried out with intermediate values of oxygen and temper-
1 -- kl ( 1 - exp(-kat)}
1 =O ature (the latter enabled activation energies to be calcu-
- (2) lated). Table 1 defines how the parameter values were
DPt DPO k2 achieved experimentally.
where kloand k2 are constants. If k2 is small, eqn. 2 reduces
to eqn. 1 with k = klo. 2.2 Ageing conditions
Kraft and cotton paper, impregnated with mineral oil, were
1.1.2 Effects of oxygen and water: According to aged isothermally in temperature-controlled ovens in the
Fabre and Pichon [15], oxygen increases the rate of degra- laboratory; no electrical stress was applied. Six experiments
dation of paper in oil containing 0.3-5% moisture by a fac- were set up as indicated in Table 2 with controlled levels of
tor of 2.5 and water increases the rate in simple proportion oxygen and water added. The experimental method was
to its concentration. Reducing the oxygen from saturation described in a previous paper [3]. The experiments were
level in the oil (30000 ppm) to less than 300 ppm, reduces repeated at temperatures of 120 2 2, 140 2 2 and 160 2 3°C.
the ageing by a factor of 16. Copper promotes the oxida- Water content was measured using a Karl Fischer potenti-
tion of the oil and thus preferentially getters the oxygen ometric titration procedure (Mitsubishi moisture meter
D61. model CA-05) and the oxygen content was measured with
Water accumulates in an ageing transformer as a direct a potentiometric oxygen meter (Systech EC50). Samples
result of insulation ageing, increasing by about 0.5% every were removed periodically and the oil analysed for furan
time the DP of the cellulose is halved, which can result in content and the DP of the paper was measured as
decreased electrical strength [16, 171. The predicted water described previously [3, 20-221.
level in transformer paper at the end of life at 80°C is 5%,
with about 0.1% in the oil [15-171. The rate of degradation, 14001
at normal service temperatures and 4'30 moisture in the
paper, has been shown to be 20 times greater than at 0.5%
(as-dried condition), so moisture accumulation has a sig- 1000

nificant effect on insulation life [17]. 800

T h s paper reports the results of laboratory experiments n
600
under a range of conditions of temperature, water and oxy-
gen, using a partial factorial approach. The relative effec- 400
tiveness of each component individually is assessed and 200
synergistic/antagonistic interactions between the compo-
nents are determined. 0 500 1000 1500 2000
time, h
Table 1: Experimentaltreatmentsto define parameter levels
Fig. I, Eflects o$ water and oxygen on DP c h g e during ugehg of B a j i
puper m oil ut 140 C
Low Medium High 0low water and low oxygen
A low water and high oxygen
-Oil- degassed in nitrogen vacuum degassed as-received high water and low oxygen
for 20 min at 60°C overnight oil A high water and high oxygen
Water 16 h in vacuo at 115°C as-received
DaDer 3 Results and discussion

2 Experiment design 3. I Factorial analysis from DP kinetics

Fig. 1 shows the effects of water and oxygen on DP during
2. I Partial factorial experiments ageing in oil at 140°C. The lines through the points were
A two-level partial factorial experiment uses the effects of fitted using a commercial numerical-analysis package
hgh and low values of relevant parameters on a common (ModelMaker from Chenvell Electronics) and eqn. 2 (see

Table 2: Summary of the ageing experiments

~~ ~

Ageing parameter

Water content Oxygen content

Experiment Paper (%) oil (ppm) in the oil

1 Low oxygen and low water c 0.1% (low) 8 c 400 ppm (low)
2 Medium oxygen and low water c 0.1% (low) 10 -1% (medium)
3 High oxygen and low water c 0.1% (low) 20 2.9-3.3% (high)
4 Low oxygen and high water 5 5 7 . 5 % (high) 8 c 400 ppm (low)
5 Medium oxygen and high water 5.5-7.5% (high) 10 -1% (medium)
6 High oxygen and high water 557.5% (high) 20 2.9-3.3%

1 I6 IEE Proc -Scr Meus Technol, Vol 147, No 3, May 2000

 Table 3: Arrhenius parameters of ageing constants calculatedfrom eqn. 2

Experiment Activation
Pre-exponential Pre-exponential Activation
energy energy
Low oxygen and low water 5.74x 108 113.00 0.343 22.72
Medium oxygen and low water 3.25x IO8 113.00 0.124 19.82
High oxygen and low water 8.01 x IO8 113.00 1.63~
10l2 122.72
Low oxygen and high water 6.18 109 113.00 1.69x IOl3 122.78
Medium oxygen and high water 9.58x IO9 113.00 6.67x IOl3 125.54
High oxygen and high water 1.10 109 113.00 1.52x IOl3 119.87

1221). k l , and k2 were assumed to vary with temperature tion of the k2 values indicates that the effect originates
according to the Arrhenius law and the data were fitted by under low-temperatureflow-water conditions, where the
optimising the Arrhenius parameters, using data from all low-oxygen value is higher than the medium-oxygen value.
temperatures. The activation energy for k,, was fixed at
113kJ/mole, based on previous experience [l, 21, 221, to 1400
reduce the number of variables, and Table 3 gives the cal- 1200[
culated values for the remaining parameters. The interest-
ing features of this Table are
(i) the low values of the k, parameters at low water and
low to medium oxygen; and
(ii) both the k l , and the k2 pre-exponential factors decrease,
at low water level, as the level of oxygen increases from low
to medium concentration in the oil.
Fig. 2 illustrates both effects graphically, plotting k, against
0 2000 4000 6000 8000 10000 12000
temperature in standard Arrhenius format. As a result of
time, h
the low k2 value, the DP in experiments at 120°C and low
water levels levelled off at between 400 and 500 and never Fig.3 DP churge during ageing of puper in oil ut Imocund low-wuter/low
oxygen levels
reached the critical 200 value where the paper loses all 0 120°C
mechanical strength (Fig. 3). A 160°C
140°C

The antagonistic effect at low water levels needs further

-2 confirmation but, if real, could explain why insulation
removed from some long-serving, scrapped transformers
can still be in very good condition. It is suggested that it
might be explained in terms of hydrogen bonding between
the oxygen and the free water, causing a 'shielding effect',
whch reduces the reaction between water and cellulose.
The implication is that oxygen has a greater affinity to free
water than to cellulose at low temperatures (120°C and pre-
-10L sumably below). Therefore, at transformer-operating tem-
2.25 2.30 2.35 2.40 2.45 2.50 2.55 2.60
reciprocaltemperature, 1000IK peratures, the inhbiting effect of oxygen on the hydrolytic
degradation could be stronger and increasing the oxygen
Fig.2 Efsects of water und oxygen on the uctivation energy of the k2 p a r m - content of the oil could actually reduce the rate of ageing
eter ofagemg offiu$puper in oil
A high water and high oxygen by making the water inaccessible to the cellulose. (Note
0 high water and medium oxygen that, to be strictly accurate, k2 defines the rate at which the
high water and low oxygen
A low water and high oxygen rate of ageing decreases with time. If t h s is greater at
0 low water and medium oxygen
0 low water and low oxygen higher oxygen levels, then the rate of ageing decreases to a
near-zero value sooner and the DP levels off earlier, at a
higher value of -400-500, compared with < 200 normally).
Measured values of k2 were used as the response variable
in a partial factorial analysis of the results, using the Yates 3.2 Furan analysis
table in Table 4. The relative effectiveness of each parame- Each data measurement was taken from a dlferent sample
ter individually and in combination is calculated in the final at a different exposure time, which results in some variabil-
column. The more positive the number, the more effective ity in the results; nevertheless it is clear that the effects of
is the combination in accelerating ageing and a negative oxygen and water on furfural formation from Kraft (wood)
number indicates an antagonistic effect. Thus it is clear that and cotton papers were s i d a r at all temperatures, water
heat and moisture are about equally effective individually, and oxygen conditions. The production of furfural always
with a significant synergistic effect in combination. Oxygen outweighed that of the other three furans (Figs. 4a and b at
alone and in combination with heat is about one thrd as 140°C) and the main production occurred when the DP
effective, but, interestingly, the water-oxygen combination dropped below about 400, as reported previously [3].
is negative indicating an antagonistic effect. Close examina- Cotton and Kraft papers also showed similar trends, but
IEE Proc - S a MeaA Technol, Vol 147, No 3 May 2000 117
 Table 4 Yates analysis of ageing using 120°C and 160°C as low and high temperatures

Variables Response Treatment Effectiveness

Yates analysis
Temperature Water Oxygen
parameter k2 combination fact01
__
low IOW low 0.00038243 1 I+a+b+ab+c+ac+hc+abc= Total 0.080783
high low low 0.00073761 a +l+ab-Lnac-c+abc-bc= effA 0.018731
IOW high low 0.00080564 b &ia&I-a+bc+abc-c-ac = eff8 0.018220
high high low 0.02579900 ab ab-&a+l+ahc-bcac+c = effA8 0.017223
low IOW high 0.00008530 C c+ac+bc+abc-l-a-&ab i
i effC 0.006333
high low high 0.00274710 ac ac-c+abcbc-a+l-aLnb = effAC 0.006057
low high high 0.00165600 hc bcwbc-c-ac&ab+l+a = effEC -o.oia8io
high high high 0.04857000 abc a b c b c a c + c a b + b + + l = effAEC 0.0004904
effA = effectiveness of parameter a, effAE = combined effectiveness of parameters a and h, etc.

the fosmation of furfiisal occursed more rapidly honi K r d i

papcr, which also produced about 25% inore product than
cotton papcr (Figs, So and h). In general, the ftirfi~ralcon-
centration initially increased slowly, lcvcllcd off, then
incscased exponentially to a niaximum and finally
decreased, when the papcs was fully degraded, a s rcposted
previously [D]. The initial increase can he scen most clearly
ill the figures in [he low-water expel-inicnls and is too short
a1 high-wales and high-tempelstusc conditions to he clearly
delincatd. The decrcase at the end m a y be duc to the cvap-
oration oftlic fuvans at high ageing conditions [3].
1600
lime, h

1000

A \
.
~

3
in
200
loool
E o
r: 0 1000 2000 3000 4000 5000 6000 7000 500
.-0 time. h
F
I

3000i
0
0 1000 2000 3000 4000 5000

2500 h

0
0 500 1000 1500
ems,h
2000 2500
.
3000 3500
4 Conclusions

4. I Degree of polymerisation data

h (i) Ileet and water accclcsate the ageing of transformer
paper in oil equally cffectivcly and are three times more
cll'cctive than oxygen. Oxygen imd water arc both syncrgis-
tic with tcmperaturc; the synergism is stronger with water
than with oxygen.
(ii) Thcrc appears to bc an mitagonistic interaction between
oxygen and w ~ k r which
, rosults in a reduclioii in aging
As dctnonstrated above for change of DP, wates was rate at low walcr levels and low ~cnipeniturcs(I 20°C) wlicn
inorc effectivc at ageing than oxygen and produced fur- the oxygen concentixtion in the oil incseases from <
furals inore rapidly hut, unfostunatcly, there arc not yet 400ppni to -1%. However, high levels, of watcr (> 4%)
adcquaw kinetic models o f fiirfural production to permit outwcigh any ameliorative effects due to the presence of
the data to be analysed niorc liilly. oxygcn
I18 / / < I <I'm.~.Sci. rMmii liklirml., IVd 147. rVo, ~ 7 ,rlliii, 2ilOii
(iii) The antagonism bctwccn water and oxygen l i m y 5
explain why thc insulation in sonic tl.ansforniers remains in
good condition even after long service. 6
7
4.2 Furan data
(iv) Both water and oxygen increase tlic rate 01' l'orination
or ageing products, but water is more effective. 2-rllraldc-
hydc is always tlic predominant product. Thc concentla-
lions of all products increase to a niaxiniiiiii then decrcasc
under a11 ageing conditions and the inaximuni rates of pro-
duction occur a l Dl's below ahout 400.
(v) Both the initial iiicrcasc in hiran concentrations in tlic
oil ;ind the total liroduciion of fiirans is approximately
1.2-1.5 titiics grcalcr kom Kraft piiper than Tram cotton
]xiper, probably due lo the presencc 01' hcmielluloses iii the
wood-based Iclart papcr.

5 Acknowledgments

The work for this papcr was sponsored by and is publishcd

by pcrmission or National Power plc and the National
Grid Company plc. The coiitacl at National Power is Mr
Allan McIntosh.

6 References