Lifetime Estimation and Monitoring of Power Transformer Considering Annual Load Factors (2013)
Lifetime Estimation and Monitoring of Power Transformer Considering Annual Load Factors (2013)
Lifetime Estimation and Monitoring of Power Transformer Considering Annual Load Factors (2013)
: Lifetime Estimation and Monitoring of Power Transformer Considering Annual Load Factors
ABSTRACT
This paper focuses on the realistic remaining service life prediction and monitoring
system of power transformers, using an algorithm developed by the authors. The
power transformer is one of the vital components of power transmission and
distribution substations. It is important to know the loading capacity and
remaining service life of a power transformer to ensure its economical and secure
operation. This information is directly related to the condition of cellulosic
insulation. According to classical method of prediction, the power transformer is
considered to be operated under constant load for the remainder of its lifespan.
However, in reality the load factor varies continuously, and this is a key factor
affecting the service life of power transformers. In this study, a realistic method for
monitoring and predicting the service lifetime of power transformers is presented
by considering annual load factors. The proposed method is applicable to any
power transformer and can be updated by considering relevant standards for
different liquid and solid insulation materials. As a case study, life-loss calculation
and the remaining service life predictions are carried out using a thermal model
that has been created considering a 250 MVA mineral oil–immersed power
transformer.
Index Terms – Lifetime, loss of life, power transformer, prediction, remaining
service life.
DOI 10.1109/TDEI.2014.004155
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 21, No. 3; June 2014 1361
problem with the insulating paper [6]. For example, DP tests TOP-OIL AND HOT-SPOT THERMAL MODELS
determine the mechanical strength of paper by measuring
The loadability and usable life of a power transformer
the cellulose level [6, 7]. But DP tests, along with other
depend primarily on the internally generated heat. It is
tests, do not provide how long the service life of the power
therefore essential to predict thermal behaviors of a
transformer would be, because there are many other factors
transformer during normal cyclic loadings and
affecting the insulation aging rate.
particularly in the presence of overload conditions or
Aging or deterioration of insulation is time dependent and unplanned outages [15]. For this reason, hot-spot
a function of the temperature, and moisture and oxygen temperature rise is the most important parameter in
content. Moisture and oxygen contributions to the insulation determining the service lifetime of a transformer. The
deterioration can be minimized in modern oil preservation hot-spot temperature is the sum of the ambient
systems. That leaves the insulation temperature as the only temperature, the top-oil temperature rise, and the hot-spot
controlling parameter of aging [8]. Hence, the degree of temperature rise, as shown in Eq. (1).
insulating paper aging is determined according to the hot-
spot temperature of the transformer in practice [8-12].
THS TH S TTO TA (1)
Power transformers have different loading values during
different times of the day. However, aging calculation is where T A is the ambient temperature, oC, ∆T TO is the top-
carried out according to the average value of the loading and oil temperature rise over ambient temperature, oK, and
temperature values are recorded on an hourly basis. ∆T HS is the hot-spot temperature rise over top-oil
Therefore the results will not be realistic because the effects temperature, oK.
of short-term changes on aging acceleration will not be
considered. Nevertheless, if the time resolution of The hot-spot and top-oil temperatures vary with the
calculation is increased, the effects on aging of peaks and loading of the power transformer in transient conditions.
transient situations can be taken into consideration. The However, their response rates are different due to the
current hardware and software technology makes this thermal capacities and thermal resistances of the winding
possible. Some methods given in the literature perform the and the oil. Thermal capacity and thermal resistance
aging calculations using high-resolution load and directly determine the time constant of the related
temperature data of the transformers. Although their aging parameters. Therefore, the hot-spot temperature calculation
calculations are correct, predictions of the transformer’s is repeated for each step of the load period. Furthermore,
remaining service life based on this method may be the loss-of-life calculations are performed at each step
deceptive because, in situation of this type, the power simultaneously. Based on these definitions, heat transfer
demanded from the transformer increases over the years. can be defined by differential equations, applicable for
Furthermore, the temperatures also increase during operation arbitrarily time-varying load factor (per-unit) and ambient
period. For this reason, the calculations and predictions temperature (TA) [14]. This model is based on the
should be made for the remaining service life while equivalent thermal circuit theory proposed by Swift et. al.
considering the increasing demand from the power [16]. The top-oil temperature, TTO, can be determined as
transformer over time. follows [16]:
In the present study, a new method has been proposed to
I 2 1 dT
predict the service life of transformers, considering two .TTOR n oil TO TTO TA n
1 1
(2)
factors. These are (1) high measurement-time resolution and 1 dt
functional algorithm, and (2) equations relating to
Where,
predictions of the loading demand and a realistic life-loss. In
addition, only the effect of temperature on aging is I is the load current, pu.
considered. TA is the ambient temperature, oC.
TTO is the top-oil temperature, oC.
2 TRANSIENT THERMAL MODEL AND β is the ratio of load to no-load losses,
FUNCTIONAL ALGORITHM oil is the top-oil time constant, min.
Thermal aging of insulation materials in mineral oil– ∆T TOR is the rated top-oil temp-rise over-ambient, oK.
immersed transformers has been formulated in n is the exponent that defines non-linearity.
ANSI/IEEE-C57.91 and IEC-60354/1991 [8, 12]. There are
some differences in the IEEE and IEC standards with Furthermore, the following equation can be used to
regard to the definitions of thermal aging methods for calculate the hot-spot temperature:
mineral oil–immersed transformers. According to the IEEE dTHS
I 2 THSR THS TTO m
1 1
standard, the normal lifetime of a power transformer is m H (3)
20.55 years. On the other hand, IEC standard does not dt
define the total life span but it usually is assumed to be 30 Where,
years, depending on the aging rate determined by the hot-
spot temperature [13, 14]. THS is the hot-spot temperature, oC.
1362 Y. Biçen et al.: Lifetime Estimation and Monitoring of Power Transformer Considering Annual Load Factors
∆THSR is the rated hot-spot temp-rise over top-oil, oK. a power transformer [8]. It is assumed that the insulation
H is the hot-spot time constant, min. deterioration can be modeled as a per-unit quantity for a
reference temperature of 110°C, and the following equation
m is the exponent defining non-linearity.
can be used for accelerated aging predictions [13, 14]:
Figure 1 shows the differential equations used in the
15000 15000
calculations in a flow-chart block diagram form.
383 H 273
F AA e pu ( 4)
dL FAA dt (5)
Thus, the life loss over a given load cycle can be calculated
by:
T
1
T 0
L FAA dt (6)
140
120
100
Temperature oC
80
60
40
Top-oil temperature
20 Hot-spot temperature
0 1 2 3 4 5 6
x 104 sec
1000 min
Figure 5. Temperature responses of power transformer.
10
8
FAA
Load changing
0
0 1 2 3 4 5 6
1.5
1.3 x 104 sec
Load factor (pu)
1000 min
1
Figure 6. Aging acceleration factor.
1
0.5
0.5 3 REALISTIC REMAINING SERVICE
LIFE PREDICTION
0
0 1 2 3 4 5 6
In the literature and the standards, the remaining service
4
x 10 sec life calculations of transformers are made using equations
1000 min considering the temperature effect on the degradation of the
Figure 4. Load step changes. paper material only [12]. Although these equations
determine the life loss of the transformers accurately, the
temp-rise over-ambient (∆TTOR) is 38.3 oK, the rated hot-spot remaining service life of transformers is not realistic. This is
temp-rise over top-oil (∆THSR) is 20.3 oK, the hot-spot time because the calculations are extrapolated and carried out
constant ( H ) is 6 min, the non-linearity exponents (n, m) according to the nominal load operation principle. The
demand of loading in a day is not constant but changes
are 0.9 and 0.8, respectively.
continuously. Moreover, the amount of loading for a day,
According to these parameters, the model is run. The month, and year can be different for different periods. The
loading variation used in first stage is shown in Figure 4. In service life of the power transformer depends on the loading
this stage, the ambient temperature is assumed to be constant conditions and the loading factor usually shows an
(30 oC), to observe the effects of the loading variation. The increasing trend as a function of use in service. Therefore,
hot-spot and top-oil temperatures of the power transformer are two different loading characteristics during the operation
obtained for a 1000-minute period and shown in Figure 5. It period can be considered. Although the loading
can be observed from the figure that the hot-spot temperature characteristics are usually given in daily periods, the yearly
is more rapidly affected by the loading changes than the top- loading characteristics are shown in Figures 7 and 8, in
oil temperature due to the much lower time constant of order to show the long-term aging effects. While forming
1364 Y. Biçen et al.: Lifetime Estimation and Monitoring of Power Transformer Considering Annual Load Factors
this characteristic, the average has been obtained by II-C programming sense, to provide high resolution and
grouping the daily loading levels in terms of years and memory optimization, equations (6), (7) and (8) can be
arranging them according to their sizes. However, if the rearranged as follows:
resolutions of the grouping are increased, then the loading
characteristics turn into an “s” shape. The life loss of a N12 N30 N24 N60
transformer operated under nominal load for most of its FAA t n T12
1
physical life is shown in Figure 7. It can be stated that a L y n1 n1 n1 n1 T Lm dt (9)
N12 N30 N 24 N60
transformer operated under such conditions completes its t n 12 0
physical life much later than that of a transformer under n1 n1 n1 n1
nominal operation. Figure 8 shows the life loss of a
transformer operated above the nominal load for most of its
physical life. A transformer that operated under such L y T12
conditions completes its physical life earlier than a LOL y % 100 (10)
Normal Insulation Life
transformed under nominal operation. In Figure 7, “LL”
shows that the physical life ends late and in Figure 8, “EL”
shows that the physical life ends early. RT y Normal Insulation Life L y T12 (11)
Load factor(pu) However, the value of the term “Normal Insulation Life” in
Lf2 equation (10) is 20.55 years (based on the relevant standards)
[8, 12]. This is based on the allowable hot-spot temperature of
1
110 oC that is determined according to standard life-time of
Lf1 20.55 years of insulation (cellulose) given on the nameplate
0.5 ratings. This value does not give realistic result when
compared with the calculation of the remaining service life
because the loading in the period until the calculation time is
20.55 LL 40 Years often not nominal, nor will it be nominal after this point.
Accordingly, the life loss of the insulation will not be at the
Figure 7. Grouped average load factor, over the life of the transformer. nominal value at all. For example, if we consider a newer
power transformer operated under the nominal loading level
during a period of 1 year, the life loss of the insulation will be
Load factor(pu) less than 1 year. If we assume this value to be around 0.55
Lf2
years, then according to the classical approach given in [8],
1 the remaining life is calculated to be: 20.55-0.55=20 years.
Continuing with the classical approach, during the remaining
Lf1 years, this calculation is regarded as correct for the situation
in which the transformer is operated in the nominal load.
0.5 However, this situation cannot be realized in the future.
Therefore, the calculation should be performed considering
Years the insulation aging data of the previous periods, while
10 EL 20.55
predicting the remaining service life of power transformer. If
Figure 8. Grouped average load factor, over the life of the transformer. the assumption that the same loss of life will be experienced
in the next periods as in the previous ones is accepted, then
equations (9)–(11) can be rearranged according to the years as
The changes in LL and EL depend on the loading shown in equations (12)–(14).
performance of the transformer. The normal insulation life is
the values stated in the standards (ANSI/IEEE-C57.91), for
the nominal loading is given as 20.55 years [8, 12], and is Ly1 T12 Ly 2 T12 Ly3 T12 Lyk T12
Lyk (12)
included in Figures 7 and 8. The certain period percentage k T12
life loss, LOL%, and remaining life, RT, can be found as
follows:
k
L yk T12 n
L T n 1
LOL % 100 (7 ) LOL yk % 100 (13)
Normal Insulation Life Normal Insulation Life
o
This calculation is based on the simple moving average Temperature C Load (pu)
method (SMA). According to this method, all data of the
past years has the same significance. Therefore, SMA has a 90 Load changing 1.8
Ambient temp.
major drawback in terms of lifetime estimation varying with 80
Hot-spot temp.
1.6
the load projections in the future. Because the power 70 1.4
demand from the power transformers may increase over the
60 1.2
years, insulation aging may be accelerated. For example, if 50 1
the loss of life in a period of 1 year, when a power
40 0.8
transformer is used initially, is Ly1 = 0.60 years, then the loss
30 0.6
of life in the 6th year can be estimated to be Ly6 = 0.85 years.
Accordingly, in this context, an increase in the levels of 20 0.4
k 0.04
wi Lyi
FAA
3
x 10 values become stable and the accuracy of the calculated
Loss of life remaining service life estimation increases due to having
2.5 more data in hand. Therefore, this drawback will be solved
spontaneously by increasing the number of years in the long
2
run. This fact is also taken into account while calculating the
remaining service life of the transformer in this work.
LOL
1.5
0.5
0
0 1 2 3 4 5 6 7 8 9
4
x 10 sec
24 hours
Figure 13. Service life span variations (for the first few years).
5 CONCLUSIONS
In the literature, it is reported many of the power transformer
that were installed in the 1980s are still in operation, and some
of them are still in good condition according to results of tests
such as DP, Furan, CO2, and others. The service lifespan of
these transformers are more than 20.55 years whereas, the
normal expected lifespan is 20.55 years according to the
classical approach. This continued performance is related to
operating conditions. These transformers are likely to be
Figure 12. (a) Yearly load rising factor K, (b) Yearly load change pu, (c) operated below the limit values determined by the
Daily life loss change in different years.
manufacturers. In addition, the remaining life of the insulation
depends on the transformer loading conditions, and it can be
The calculation procedures and the results are presented in
estimated using the program iterations, considering the
Table 1 for both the classical method and the method proposed
variation of loading and life loss in the previous years. In this
in this paper. According to the classical method, the remaining
paper, we developed a method that can easily determine the
service life of a power transformer is predicted to be about 16.5
life-span shortening and extension of power transformers.
years for the mineral oil–immersed transformer. However,
Knowledge regarding the trend factors of the data for the
using our proposed method, the remaining life is predicted to be
previous years can help to estimate more realistically the
approximately 12 years, as shown in Table I. A difference of
demand on the transformer and the change in the life loss of the
nearly 4.5 years was observed between the two methods.
insulation in future years. As the trend factor is calculated for
In the early years when the power transformer is first in seasonal and monthly trends, prediction of the remaining life
service, an accurate estimation of its service life is not critical, can be deceptive. To obtain more stable results, the use of a
but the importance of these calculations increases as time year-based trend, in which all the months are taken into
progresses, especially towards the end of its service life. The consideration, is more reasonable. The proposed method can
lifetime estimation can be performed using the classical easily determine the degree of lifespan shortening and
method at any time. In the proposed method, the loss of life extension for power transformers, allowing remaining service
calculation is made for the first three years only, but the life to be predicted more realistically. Also, this method can be
estimation of service lifetime cannot be done due to the adapted on top the existing monitoring systems. This is
absence of required historical data as shown in Figure 13. In important information for the electricity market and will bring
addition, for the first few years, the remaining lifetime important benefits when planning for the future. Strategic plans
estimation results vary depending on the load rising factor regarding electric power transmission and distribution systems
(K), according to our approach. However, in later years, these can therefore be made more accurately.
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 21, No. 3; June 2014 1367
Table 1. Remaining service life comparison for moth classical and proposed methods