Electric Power Systems Research 83 (2012) 255261

Contents lists available at SciVerse ScienceDirect

Electric Power Systems Research
j our nal homepage: www. el sevi er . com/ l ocat e/ epsr
Short survey
Method for determining optimal power transformers exploitation strategy
Olga Ristic, Vladica Mijailovic

Technical Faculty, University of Kragujevac, Svetog Save 65, 32000 Cacak, Serbia
a r t i c l e i n f o
Article history:
Received 21 January 2011
Received in revised form12 August 2011
Accepted 10 September 2011
Available online 11 October 2011
Keywords:
Power transformer
Preventive maintenance
Refurbishment
Spare parts
Condition monitoring
Costs
a b s t r a c t
The paper suggests a procedure for determining an optimal combination of activities and measures to
minimize the expected total costs during the planned exploitation period of power transformers.
Combinations of different types of preventive maintenance, spare parts keeping and installation of
condition monitoring systems of power transformer individual components are taken into consideration.
Power transformer is a complex system, consisting of six components (functional parts). It is assumed
that each component has two independent, competing failure modes: a wear-out failure mode, modeled
by two-parameter Weibull distribution, and a chance failure mode, characterized by an exponential
distribution.The superposition of failures and planned maintenance are taken into account in the analysis
of some forms of preventive maintenance.
The application of the method suggested and the benets it provides are demonstrated for one trans-
former station (TS) 110/10 kV/kV with 2 31500 kVA-110 kV Star-10 kV Delta transformers in case of
radial supplying of customers as well as in case when outage of one power transformer does not affect
the power supply to customers.
2011 Elsevier B.V. All rights reserved.
1. Introduction
Equipment failures in distribution substations may cause inter-
ruptions in the power supply, which result in cost increase on both
power distribution company and customers. The most severe con-
sequences arise after power transformer failures, because the time
for renewal of a damagedtransformer canbe very long andrenewal
process can be expensive.
Power transformer failures are classied as repairable or unre-
pairable. Renewal time of repairable failures is not usually long and
spare parts are not necessary. Unrepairable failures request using
of spare parts, so the renewal time depends on necessary spare
parts availability. Purchase of spare parts will substantially reduce
the renewal time of unrepairable failures on the one hand, but, it
implies considerable investment cost on the other.
Power transformer failure rate can be inuenced by perform-
ing preventive maintenance of certain scope and/or by installing
condition monitoring systems of power transformer individual
components.
During the preventive maintenance of the rst transformer fail-
ure of the second transformer is possible, which, depending of the
concept of power supply, may result in the total interruption or
reduction in power supply to customers.

Corresponding author. Tel.: +381 62 283 557; fax: +381 32 342 101.
E-mail address: miltea@tfc.kg.ac.rs (V. Mijailovic).
The following sections will present a precise procedure for
determining optimal combination of activities and measures that
can ensure a minimum of expected total cost during the planned
power transformers exploitation period.
2. Basic assumptions
At any point of time, the status of power transformer can be
classied as either operating or failed. Failed status is a result of
major failures and/or minor failures. According to CIGRE, major
failure is a failure of a power transformer which causes the ces-
sation of one or more of its fundamental functions. A major failure
will result in an immediate change in the systemoperating condi-
tions. A minor failure is a failure of a power transformer other than
a major failure or any failure, even complete, of a constructional
element or a sub-assembly which does not cause a major failure of
the equipment.
Major failures can be repairable or unrepairable. Minor failures
are repairable and can be repaired for t 24h.
Hence, probability that the component k of power transformer
is in the operating status equals
R
k
(t) = exp((
k,mf
+
k,MF
) t) exp
_

_
t

k
_

k
_
(1)
i.e. it is adopted that the component has two independent failure
modes: a chance failure mode and a wear-out mode [1]. Chance
failures are those failures that cannot be improved with mainte-
nance, such as human error, manufacture and design defects, and
0378-7796/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2011.09.007
256 O. Ristic, V. Mijailovic / Electric Power Systems Research 83 (2012) 255261
Nomenclature
t time
C
new
purchase cost of newpower transformer
C
new,k
purchase cost of power transformer component k
C
new,oil
purchase cost of newoil
C
u
cost of oil ltration and drying
C
pm
cost of material for performing 1-day preventive
maintenance
C
reg.-oil
cost of oil regeneration
C
reg.-ins
cost of insulation systemregeneration
b number of functional parts-components of power
transformer
f
k
number of failureclasses of power transformer com-
ponent k with regard to the failure repair time
p
k
probability that the failure occurs on power trans-
former component k
p
k,i
probability that the failure of class i occurs on
component k
C(t,t +1) average yearly failure renewal cost per transformer
during time interval (t,t +1)
C

k,i
renewal cost of class i failure on component k if
spare component k is not available
C
CMS-b
k,i
renewal cost of class i failure on component k if
condition monitoring system is installed and spare
component k is not available
r

k,i
renewal time of class i failure on component k if
spare component k is not available
C

k,i
renewal cost of class i failure on component k if
spare component k is available
r
CMS-b
k,i
renewal time of class i failure on component k if
condition monitoring system is installed and spare
component k is not available
r

k,i
renewal time of class i failure on component k if
spare component k is available
C
s,k
purchase cost of condition monitoring system of
power transformer component k
p
CMS
probability that the failure of class i occurs on
power transformer component k after installing
of condition monitoring system
C
m-cond.mon.sys.
yearly cost of condition monitoring system
preventive maintenance

av
average failure rate of power transformer

k,MF
,
k,mf
major failure rate and minor failure rate of com-
ponent k, respectively

0
t,MF
initial (statistical) data for major failure rate at time
t
, the Weibull scale parameter and the Weibull shape
parameter, respectively
R(t), U(t) reliability and unreliability, respectively
W
1
(t,t +1), W
2
(t,t +1), and W
dod
(t,t +1) expected yearly
energy not delivered during time interval (t,t +1)
due to outage of one transformer, due to outage of
both transformers and due to the superposition of
failures and planned maintenance, respectively
C
EN
loss of revenue and load curtailment cost per kWh
not delivered
N duration of the planned period, expressed in years
P
inst
substation installed capacity
t
pm
durationof 1-day preventive maintenance perform-
ing
t
oil
, t
ins
, and t
ref
duration of oil-regeneration, insulation
regeneration and power transformer refurbishment
performing, respectively
C
ref
(T
ref
) cost of refurbishment performing at the point of
time T
ref
d
k
degree of major failures detected by the condition
monitoring system on power transformer compo-
nent k
Z(t,t +1) total expectedcost duringtimeinterval (t,t +1) (sum
of investment cost, failures renewal cost and load
curtailment cost)
bad weather such as lightning or ice storms. These failures gener-
ally have a constant failure rate over the transformer exploitation
period and maintenance cannot reduce the failure rate. Wear-out
failures are those failures whose probability increases with the
service age or operations, so that maintenance can renew the cor-
responding conditions and thus reduce the failure rate. A chance
failure mode is characterized by the exponential distribution and a
wear-out mode is modeledby two-parameter Weibull distribution.
Power transformer consists of six component-functional parts
[2,3]: 1. Windings +oil, 2. Core; 3. Bushings; 4. Tank; 5. On-load
tap-changer; 6. Other accessories.
The main failures occurring on each functional part of power
transformer are as follows:
- Windings, oil, core
Partial discharges, abnormal oil and cellulose aging, loose con-
nections, oil contamination, excessive water content, overheating
of laminations, overheating due to circulating currents, turn-to-
turn failures, phase-to-phase failures, mechanical failures, open
winding, external faults.
- Bushings
Moisture contaminationdue to deteriorationof gasket material
or cracks in terminal connections, partial discharges.
- Tank
Poor tank weld, corrosion, external damages.
- On-load tap-changer
Local hotspots due to contact overheating, signicant increase
in required torque, sparking, oil leaks, partial discharges.
- Other accessories
Arcing, local overheating, electrical failures of pumps and fans,
internal or external blocking of radiators resulting in poor heat
exchange.
With regard to the failure repair time, there are three failure
classes [2,3]:
i =1) failures which can be repaired for t 1 day,
i =2) failures which can be repaired for 1 day<t <30 days,
i =3) failures which can be repaired for t 30 days.
Relevant data for power transformer components (p
k
, p
k,i
, r

k,i
,
r

k,i
) are presented in Table 1 [25].
3. Model development [214]
In this section, the following individual cases and their combi-
nations will be analyzed:
g
1
) Operation without performing preventive maintenance (run to
failure) and without keeping of spare components
g
2
) One-daypreventivemaintenance(practically, t
pm
=8h)visual
examination, checkingthe state of the transformer andreplace-
ment of worn-out parts
O. Ristic, V. Mijailovic / Electric Power Systems Research 83 (2012) 255261 257
Table 1
Power transformer components reliability data [25].
k p
k
, % Failure
class i
p
k,i
, % r

k,i
, day r

k,i
, day
1 Windings 26.4 2 14.54 30 15
3 85.46 250 15
2 Core 2.4 2 50 30 15
3 50 180 15
3 Bushings 12 1 14.82 1 1
2 51.85 40 3
3 33.33 40 15
4 Tank 7.9 1 58.82 1 1
2 23.53 3 3
3 17.65 90 15
5 On-load tap-changer 40.7 1 25.61 1 1
2 52.44 3 3
3 21.95 40 3
6 Other accessories 10.6 1 65.22 1 1
2 17.39 15 15
3 17.39 40 15
g
3
) Oil regeneration+g
2
)
g
4
) Insulation systemregeneration+g
2
)
g
5
) Power transformer refurbishment. After this, transformer will
be as good as new
Optimal point of timeT
ref
for performingrefurbishment is deter-
mined according to the criterion that during the rst year after
power transformer refurbishment expectedtotal costs are less than
in case when refurbishment is not performed.
For cases g
1
g
5
equations for calculation of expected yearly
cost per transformer during time interval (t,t +1), expected yearly
unavailability and expected yearly down-time of one transformer
and of both transformers during time interval (t,t +1) are given in
[4,5].
g
6
) Installing the condition monitoring system of power
transformer individual components. Expected life-time of
commercially available systems is 10 years. Purchase price is
about 10% of the corresponding component purchase price,
C
s,k
=0.1C
new,k
. Values for d
k
are presented in Table 2 [14].
Detecting of failures in the early stage of development prevents
an outage of power transformer and reduces cost and time of
Table 2
Values for d
k
[14].
Component Detection rate d
k
, %
(1) Windings 70
(2) Core 70
(3) Bushings 80
(4) Tank
(5) On-load tap-changer 75
(6) Other accessories 100
repair. Neglecting of detected failures repair cost seems to be
too optimistic. Because of reasons of certainty, an assumption
is adopted: all detected major failures on windings and oil will
be treated as faults, which can be repaired in 5 days and for
5000EUR. Detected major failures on other components will
be treated as minor failures.
The system for condition monitoring of certain power trans-
former component enables detection of failures in the early stage
of development. It leads to the reduction of major failure rate and
the increase in minor failure rate. As a result of major failure rate
reduction, the Weibull scale parameter will be raised.
For example, let us denote with

k
, k = 1, 2 scale parameters
for windings and core after installing of the condition monitor-
ing system. The condition monitoring systemwill be installed after
operation period of length T
s
without failures.
During next 10 years of operation reliability of transformer will
be
R

tot
(t) R
1
(T
s
) R
2
(T
s
) R

1
(t T
s
) R

2
(t T
s
)
b

k=3
R
k
(t) (2)
R

k
(t T
s
) = exp[(

k,MF
+

k,mf
) (t T
s
)] exp
_

_
t T
s

k
_

k
_
,
k = 1, 2 (3)
Expected yearly cost and transformer unavailability during the
rst year after installing of a conditionmonitoring systemfor wind-
ings and core are as follows, respectively:
C
ET,no-sp.-cond.mon.sys.
(T
s
, T
s
+1) =
(R

tot
(T
s
) R

tot
(T
s
+1))
_
_
2

j=1
p
j
f
j

i=1
p
CMS
j,i
C
CMS-b
j,i
+
b

k=3
p
k
f
k

i=1
p
k,i
C

k,i
_
_
Im
CMS
(T
s
, T
s
+1)
+
R
tot
(T
s
) C
s,12
_
Ts+10
Ts
R

tot
(t) dt +(R

tot
(T
s
) R

tot
(T
s
+10))
_
_
2

j=1
p
j
f
j

i=1
p
CMS
j,i
r
CMS-b
j,i
+
b

k=3
p
k
f
k

i=1
p
k,i
r

k,i
_
_
+C
m-cond.mon.sys.
(4)
U
ET,no-sp.-cond.mon.sys.
(T
s
, T
s
+1) =
(R

tot
(T
s
) R

tot
(T
s
+1))
_
_
2

j=1
p
j
f
j

i=1
p
CMS
j,i
r
CMS-b
j,i
+
b

k=3
p
k
f
j

i=1
p
k,i
r

k,i
_
_
Im
CMS
(T
s
, T
s
+1)
(5)
Im
CMS
(T
s
, T
s
+1) =
_
Ts+1
Ts
R

tot
(t) dt +(R

tot
(T
s
) R

tot
(T
s
+1))
_
_
2

j=1
p
j
f
j

i=1
p
CMS
r
CMS-b
j,i
+
b

k=3
p
k
f
k

i=1
p
k,i
r

k,i
_
_
(6)
g
7
) Determination of optimal amount of spare power transformer
components. For this case equations for calculation of invest-
ment cost, expected yearly cost per transformer during time
interval (t,t +1), expected yearly unavailability and expected
258 O. Ristic, V. Mijailovic / Electric Power Systems Research 83 (2012) 255261
2
W
1
W
dod
W
dod

max
inst
P
P
1
a
1
b
2

8760 ] [h t
max
inst
P
P
2
inst
P
1

8760
] [h t
a b
Fig. 1. Annual hourly load-duration diagram(a
1
=0.8, b
1
=0.4).
yearly down-time of one transformer and of both transformers
during a time interval (t,t +1) are given in [5].
g
8
) Determination of optimal combination of activities and mea-
sures. For each activity (measure) listed above an economic
justication of implementation will be determined, and then
an optimal combination that provides minimum total annual
costs will be established.
(1) Let us calculate and denote with Z
g
1
(t, t +1), t = 1, N 1
total expected cost during time interval (t,t +1) in case of
power transformer operation without keeping of spare com-
ponents, without performing preventive maintenance and
without installing of condition monitoring systems. These are
reference values, based on which preliminary analysis will be
performed in step (2).
(2) For each activity g
k
, k = 2, 7, values Z
g
k
(t, t +1) will be calcu-
lated. The comparison of these values with values Z
g
1
(t, t +1),
t = 1, N 1 will provide the determining of the point of time
t
1k
for performing of activity g
k
, k = 2, 7. Performing of activity
is justied if Z
g
k
(t
1k
, t
1k
+1) <Z
g
1
(t
1k
, t
1k
+1). Values Z
g
k
(t, t +
1), corresponding to the activity whose performing is justied
before all other activities, will be adopted as newreference val-
ues. If it is found that performing of several individual activities
is justiedat the same point of time t
a
, values Z
g
k
(t, t +1) which
are minimal during time interval (t
a
, N) will be adoptedfor new
reference values. Thus, activity A
1
which is to be performed
rst is determined, as well as the activities whose performing
is potentially justied.
(3) Assuming that activity A
1
is performed at the point of time t
a
,
next activity A
2
whose performing is justied at point of time
t
b
(t
b
t
a
) will be determined.
(4) Assuming that activity A
1
is performed at the point of time t
a
and activity A
2
is performed at the point of time t
b
, the third
activity A
3
whose performing is justied at the point of time t
c
(t
c
t
b
) will be determined, etc.
4. Application
An application of the model will be demonstrated for
one TS 110/10kV/kV, with two power transformers installed,
P
inst
=231.5MVA =63MVA. Calculations are made for planned
period of N=40 years.
The annual hourly load-duration diagram for the substation is
displayed in Fig. 1. Peak load is 80% and minimum load is 40%,
with regard to the installed capacity. It is assumed that the load
has reached its saturation and that there are no annual load level
increments during the planned period.
Calculations are made under assumption that failures occur
during the peak load period, which is the most critical case. The
loss of revenue and load curtailment cost per kWh not delivered
is C
EN
=0.10EUR/kWh (average electricity price for medium size
Table 3
Power transformer components costs [25].
k Transformer component C
new,k
, EUR
1 Windings +oil 250000+40000
2 Core 80000
3 Bushing 800
4 Tank 28000
5 On-load tap-changer 42000
6 Other accessories 22000
Table 4
Costs of failures renewal on power transformer components [25].
k p
k
, % Failure
class i
p
k,i
, % C

k,i
C

k,i
1 Windings 26.4 2 14.54 0.2C
new,1
+Cu Cu
3 85.46 0.5C
new,1
+Cu Cu
2 Core 2.4 2 50 0.2C
new,2
+Cu Cu
3 50 0.5C
new,2
+Cu Cu
3 Bushings 12 1 14.82 0.4C
new,3
0.4C
new,3
2 51.85 C
new,3

3 33.33 C
new,3
+Cu Cu
4 Tank 7.9 1 58.82 0.1C
new,4
0.1C
new,4
2 23.53 0.2C
new,4
0.2C
new,4
3 17.65 C
new,4
+Cu Cu
5 On-load tap-changer 40.7 1 25.61 0.1Cnew,5 0.1Cnew,5
2 52.44 0.2Cnew,5 0.2Cnew,5
3 21.95 0.4Cnew,5
6 Other accessories 10.6 1 65.22 0.1C
new,6
0.1C
new,6
2 17.39 0.5C
new,6
0.5C
new,6
3 17.39 C
new,6

households in EU). Adopted value for yearly cost of refurbishment
is 0.015C
new
/year. After insulation systemregeneration, technical
age of paper insulation will be reduced for x =0.3.
The other data used for calculations are the following:
t
oil
=5 days, t
ins
=10 days, C
u
=0.2C
new,oil
, t
ref
=28 days.
C
pm
=100EUR, C
reg.-oil
=8000EUR, C
reg.-ins.
=10000EUR, t
ins
=10
days, t
ref
=28 days, and C
s,k
=0.1C
new,k
. Annual cost of the preven-
tive maintenance of the condition monitoring systemis 0.01C
s,k
.
Purchase prices for components of power transformer 110kV,
31.5MVA arepresentedinTable3. Dataabout failures renewal costs
on power transformer components are presented in Table 4.
Parameters in (1) are determined based on exploitation data
about major failure rate of power transformers in the USA, Fig. 2
[15]. It is adopted that the average failure rate of power trans-
formers during 30 years of exploitation is
av
=0.0151/year [2].
Based on Fig. 2 and data presented in Table 1, by applying the
least-squares method, parameters of Weibull distribution are eval-
uated for each power transformer component, Table 5. The same
table shows the data about scale parameters when condition mon-
itoring system is installed on each component individually, based
O. Ristic, V. Mijailovic / Electric Power Systems Research 83 (2012) 255261 259
Fig. 2. Statistical data about major failure rate
0
t,MF
during time interval
(19732003) [15].
on Fig. 2, Tables 1 and 2. Because of clarity of exposure, the fol-
lowing remark must be made: probability that the population of
power transformers have similar characteristics is very small, even
in the case when units with equal design and manufacturing pro-
cesses can be selected. Also, exploitation conditions (temperature,
loading, maintenance practices, etc.) are different fromunit to unit.
This limits the accuracy of the Weibull distribution parameters. For
example:
In[16] a total populationof almost 500transformers is analyzed.
The failure data are tted with a two-parameter Weibull distribu-
tion. The estimated Weibull distribution parameters are =3.3 and
=70. Then, the total populationis dividedintwo sub-populations:
a population of Group 1 transformers with a voltage level above
50kVand a populationof Group 2 transformers with a voltage level
equal to50kV. Results are the following: for Group1: =2.8, =36;
for Group 2: =2.8, =80.
In [17] (page 78), exploitation data about bushing failures are,
also, ttedwitha two-parameter Weibull distribution. For different
data sources, Weibull distribution parameters are =(1.33.3) and
=(702418).
The results of the analysis are the following:
- Radial power supply to customers
Applying the procedure previously described, it is calculated
that installing systems for condition monitoring of windings and
core is justied from the rst year. Bearing in mind the expected
exploitation period of these systems, it is obvious that these sys-
tems will be installed at the beginning of the eleventh, twenty-rst
and thirty-rst year of power transformer exploitation period. This
activity will be marked as A
1
. In the second step the analyses will
be performedtodetermine justiability of implementationof other
measures and activities (summary activity A
2
), under the assump-
tion that activity A
1
is implemented. Hence, activity A
2
includes
installing of the condition monitoring system for bushings from
the rst year of power transformer exploitation period.
For the sake of completeness, it is important to note that after
performing of activities A
1
and A
2
, during the planned period of
Table 5
Weibull distribution parameters.
k Transformer component
k

k

k
1 Windings +oil 3.58 57.0131 79.805
2 Coil 3.58 111.395 155.927
3 Bushing 3.58 74.3159 116.5
4 Tank 3.58 102.321 102.321
5 On-load tap-changer 3.58 54.8723 80.213
6 Other accessories 3.58 98.8079
0
200000
400000
600000
800000
1000000
1200000
37 35 33 31 29 27 25 23 21 19 17 15 13 11 9 7 5 3 1
Time t (year)
T
o
t
a
l

e
x
p
e
c
t
e
d

y
e
a
r
l
y

c
o
s
t
,

E
U
R
a)
b)
c)
d)
Fig. 3. Total expected yearly cost for case g
1
, Zg
1
(t, t +1) (curve a), for case when
activity A
1
is performed, Z
A
1
(t, t +1) (curve b), for case when activities A
1
and A
2
are
performed (curve c) and for case when activities A
1
, A
2
and A
3
are performed (curve
d).
40 years, it is neither justied to perform 1-day preventive main-
tenance, nor oil regeneration, nor insulation system regeneration.
It remains to analyze the justication of purchasing spare power
transformer components and justication of performing power
transformer refurbishment (summary activity A
3
).
By gradual analysis, it is calculated that the summary activity A
3
consists of:
- Storing of spare on-load tap-changer from the beginning of the
ninth year,
- Storing of spare windings fromthe beginning of the tenth year,
- Storing of spare tank fromthe fourteenth year,
- Storing of spare other accessories fromthe fteenth year and
- Performing of power transformer refurbishment at the beginning
of the twenty-fth year. Hence, it is obvious that purchasing of
spare core is not necessary.
Comparative overviewof values Z
g
1
(t, t +1) (curve a), Z
A
1
(t, t +
1) (curve b), Z
A
1
+A
2
(t, t +1) (curve c) and Z
A
1
+A
2
+A
3
(t, t +1) (curve
d) is given in Fig. 3.
- Outage of one power transformer does not affect the customers sup-
plying
Denote with Z
g(n1)
1
(t, t +1) total expected cost for the case
g
1
. Neither purchasing of spare components nor performing of any
forms of preventive maintenance is justied. It is only justied to
install the system for condition monitoring of bushings from the
rst year. This activity will be marked as A
4
. Fig. 4 shows com-
parative overview of values Z
g(n1)
1
(t, t +1) (curve e) and values
Z
A
4
(t, t +1) (curve f).
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
39 37 35 33 31 29 27 25 23 21 19 17 15 13 11 9 7 5 3 1
Time t (year)
T
o
t
a
l

e
x
p
e
c
t
e
d

y
e
a
r
l
y

c
o
s
t
,

E
U
R
e)
f)
Fig. 4. Comparative overview of values Z
g(n1)
1
(t, t +1) (curve e) and values
Z
A
4
(t, t +1) (curve f).
260 O. Ristic, V. Mijailovic / Electric Power Systems Research 83 (2012) 255261
0
100000
200000
300000
400000
500000
600000
700000
800000
900000
39 37 35 33 31 29 27 25 23 21 19 17 15 13 11 9 7 5 3 1
Time t (year)
T
o
t
a
l

e
x
p
e
c
t
e
d

y
e
a
r
l
y

c
o
s
t
,

E
U
R
h)
c)
d)
Fig. 5. Overviewof values Z
OASC
(t, t +1) (curve h), Z
A
1
+A
2
(t, t +1) (curve c) andvalues
Z
A
1
+A
2
+A
3
(t, t +1) (curve d).
5. Conclusions
The paper suggests a method that allows determination of the
optimal combination of activities and measures, which provide
the lowest costs during the planned exploitation period of power
transformers. An optimization criterion is the sum of total cost,
comprising yearly load curtailment cost, yearly failures renewal
cost and yearly investment cost (purchasing of spare components,
purchasing and installing of conditionmonitoring systems and cost
of performing power transformer preventive maintenance). It is
shown that a relatively small investment can bring about large
savings.
The application of condition monitoring systems enables spare
components to be purchased considerably later, without need to
purchase spare core. Denote with Z
OASC
(t, t +1) total expected cost
during time interval (t, t +1) in case of keeping optimal amount
of spare components (without application of condition monitoring
systems). Overviewof values Z
OASC
(t, t +1) (curve h), Z
A
1
+A
2
(t, t +1)
(curve c) and values Z
A
1
+A
2
+A
3
(t, t +1) (curve d) is given in Fig. 5.
Appendix A. Appendix [1825]
A.1. On condition monitoring of oil-lled power transformer
Reliable and uninterrupted functioning of power transformer is
a key factor in protable generation, transmission and distribution.
To increase availability and optimize operating management on-
line condition monitoring of power transformers and diagnosis can
be used.
The successful utilities implement condition monitoring as part
of a reliability driven maintenance strategy that maximizes the
performance of power transformers at the lowest life-cycle cost,
optimizing power transformer replacement strategy and eliminat-
ing or reducing catastrophic failures. Despite numerous studies and
many publications on the subject, there have been difcult chal-
lenges ondecidingwhentoutilizeconditionmonitoringonapower
transformer and how to use condition monitoring to its potential
for early fault detection.
For correct selection of transformer on-line monitoring, it is
required to distinguish between fault detection and fault diag-
nostic. In the recent Cigre guide on transformer management this
distinction is skilfully summarized with an analogy to the proper
approach to human health problems:
Symptoms Diagnosis Cure
The rst step, which may be described as monitoring focuses
on detection of symptoms or evidence of abnormal condition. The
fundamental question to be answered is whether the situation
is normal or not. The techniques must be cost efcient and yet
sufciently sensitive and broadband to detect any potential prob-
lemat anearlystage. Suchmonitoringtechniques shouldbeapplied
regularly and preferably continuously on-line.
The second step is the diagnostic where an abnormal situa-
tion is investigated to establish the type of fault, the severity of
the problemand the corrective actions to be taken. The diagnostic
step needs to be carried out only on those units that are deemed
abnormal. Tests that wouldbe appliedcanbe more expensive and
off-line, and they need to be focused on individual attributes in an
attempt to arrive at an unambiguous diagnosis.
The distinction between monitoring and diagnostic allows
dening more accurately requirements of the ideal monitoring sys-
tem: it must provide a wide coverage of faults detection, but it does
not have to cover the full range of diagnostic and condition assess-
ment. It would be impractical to deploy on each power transformer
all the methods of interest for condition assessment.
The transformer is a very complex structure with basically 5 dif-
ferent mainmaterials: copper, coresteel, tanksteel andmain/minor
insulation and insulating oil. All metallic materials are aging very
slowly, therefore all functional life considerations/aging impacts
are related to insulated material and oil.
Selection of the most efcient monitoring system requires a
review of the insulation deteriorating process and the best meth-
ods for early detection. Excessive aging or moisture ingress may
develop problems, and those problems should be detected at an
early stage to allowfor orderly removal fromservice and repair.
Any problems developing in the winding insulation, in the con-
nections, in the core or in the shields will generate a localized high
temperature or electric discharges, resulting in decomposition of
oil and/or paper. The minute amount of gas dissolved in the oil can
alert the operator about a problemin development and the relative
proportion of each gasses can also provides indication on the type
of fault.
The aging process in the oil/cellulose insulation system under
thermal stress and their measurable effects are due to chemical
reactions in the dielectric. The temperature of the oil/paper dielec-
tric is the critical aging parameter to cause enough change in the
mechanical and electrical properties of the material. Apart from
high temperatures, other important parameters affecting the aging
of the solid and liquid insulation include the presence of water and
oxygen in the system.
On-line monitoring system serves to detect changes in insu-
lation system in the form of an early warning system, due to
thermal/dielectric/chemical or mechanical impact, i.e. to detect:
- Hotspots,
- Degradation of the insulation,
- Excess moisture in insulation,
- Localized faults/defects,
- Partial discharges,
- Partial rupture/mechanical defects/brittleness,
- Chemical or thermal aging.
None of these defects can be detected by a singular diagnostic
procedure. Therefore, it is necessary to apply a multitude of differ-
ent methods. Furthermore, to enable a trend analysis and condition
assessment it is necessary quite frequently a sequence of samples
and measurements to identify the velocity of change, the rate of
progressiveness, the quantity of indicated signatures, etc.
The list of measurements and procedures that might be used
includes:
- Hotspot measurement,
- On-line monitoring of oil/copper temperatures, cooling plant and
ambient temperatures,
O. Ristic, V. Mijailovic / Electric Power Systems Research 83 (2012) 255261 261
- Oil-analysis,
- Furane components analysis,
- Dielectric response measurement,
- Depolymerisation index of solid insulation,
- KarlFischer test,
- Power factor of oil and insulation,
- Oil aging/oxidation tests,
- Partial discharge measurements,
- H
2
O content in oil,
- Moisture in solid insulation,
- Impedance measurements,
- Frequency response analysis,
- Transfer function analysis, etc.
References
[1] Integrationof ConditionMonitoringwithanOptimizedMaintenanceScheduler
for Circuit Breakers and Power Transformers, PSERC Publication 06-04, January
2006.
[2] Design of Reliable Industrial and Commercial Power Systems, IEEE Std 493-
2007.
[3] An international survey on failures in large power transformers in service,
ELECTRA No.88, 1983.
[4] V. Mijailovic, Method for effects evaluation of some forms of power trans-
formers preventive maintenance, Electric Power Systems Research 78 (2008)
765776.
[5] V. Mijailovic, Optimal spares availability strategy for power transformer com-
ponents, Electric Power Systems Research 80 (2010) 987992.
[6] R. Brown, G. Frimpong, H. Lee Willis, Failure rate modeling using equipment
inspection data, IEEE Transactions on Power Systems 19 (May (2)) (2004).
[7] W. Bartley, An Analysis of International Transformer Failures, Part 1 & Part 2,
2004, http://www.hsb.com/thelocomotive.
[8] G. Anders, A. Vaccaro, Innovations in Power Systems Reliability, Springer, Lon-
don, 2011.
[9] E. Bauer, Practical SystemReliability, John Wiley & Sons, 2009.
[10] P. Gill, Electrical Power Equipment Maintenance and Testing, CRC Press, 2009.
[11] A.A. Chowdhury, D.O. Koval, Power Distribution System ReliabilityPractical
Methods and Applications, John Wiley & Sons, 2009.
[12] T. Nakagawa, Advanced Reliability Models and Maintenance Policies, Springer,
2008.
[13] D.N.P. Murthy, M. Xie, R. Jiang, Weibull Models, John Wiley & Sons, 2004.
[14] S. Tenbohlen, T. Stirl, G. Bastos, J. Baldauf, P. Mayer, M. Stach, B. Breitenbauch,
R. Huber, Experience-based Evaluation of Economic Benets of On-line Moni-
toring Systems for Power Transformers, CIGRE Session, Paris, 2002, pp. 12110.
[15] B. Augenstein, Outsourced Monitoring and Reliability of Critical Assets, Dis-
tribuTech, Las Vegas, February 2003.
[16] R. Jongen, P. Morshuis, J. Smit, A. Janssen, E. Gulski, A Statistical Approach to
Processing Power Transformer Failure Data, CIRED, Vienna, 2007, p. 546.
[17] A. van Schijndel, Power transformer reliability modelling, Ph.D. Thesis, Eind-
hoven University, 2010.
[18] T. Stirl, R. Skrzypek, S. Tenbohlen, R. Vilaithong, On-line Condition Monitoring
and Diagnosis for Power Transformers their Bushings, Tap Changer and
Insulation System, http://www.unistuttgart.de/ieh/forschung/./CMD2006-
Tobias.pdf.
[19] R. Heywood, M. Boltze, S. Markalous, Power Transformer Condition Monitor-
ing and Life-Cycle ManagementExperience froma Detailed Case Study, Euro
TechCon, Stretton, United Kingdom, 2009.
[20] I. Hhlein, A.J. Kachler, S. Tenbohlen, Transformer Life ManagementGerman
Experience With Condition Assessment, CIGRE SC12/A2, Merida Kolloquium,
2003.
[21] B.D. Sparling, J. Aubin, Power Transformer Life Extension Through Better Moni-
toring, http://www.gepower.com/prod serv/plants./powergrid europe07.pdf.
[22] W.H. Tang, Q.H. Wu, Condition Monitoring and Assessment of Power Trans-
formers Using Computational Intelligence, Springer, 2001.
[23] M. Arshad, S.M. Islam, Power Transformer Condition Monitoring and Assess-
ment for Strategic Benets, http://www.itee.uq.edu.au/aupec/aupec03/./064
%20Arshad%20full%20paper.pdf.
[24] K. Nousiainen, P. Verho, MonitoringtheTemperatureAndageingof Distribution
Transformers with Distribution Automation, INSUCON, Berlin, 2002.
[25] T. Pink, P. Stewart, Power Transformer Control SystemDevelopments, TechCon,
North America, 2004.