Content uploaded by Houman Pezeshki
Author content
All content in this area was uploaded by Houman Pezeshki on May 05, 2015
Content may be subject to copyright.
1212 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 29, NO. 3, JUNE 2014
Impact of High PV Penetration on Distribution
Transformer Insulation Life
Houman Pezeshki, Member, IEEE,PeterJ.Wolfs, Senior Member, IEEE,and Gera
rd Ledwich, Senior Member, IEEE
Abstract—The reliable operation of distribution systems is criti-
cally dependent on detailed understanding of load impacts on dis-
tribution transformer insulation systems. This paper estimates the
impact of rooftop photovoltaic (PV) generation on a typical 200-
kVA, 22/0.415-kV distribution transformer life under different op-
erating conditions. This transformer supplies a suburban area with
a high penetration of roof top photovoltaic systems. The trans-
former loads and the phase distribution of the PV systems are sig-
nificantly unbalanced. Oil and hot-spot temperature and remnant
life of distribution transformer under different PV and balance
scenarios are calculated. It is shown that PV can significantly ex-
tend the transformer life.
Index Terms—Distribution transformer, life assessment, roof top
PV, unbalanced operation.
I. INTRODUCTION
MODERN distribution systems serve a variety of diverse
customers. Three-phase four-wire systems, such as 400/
230-Vrms systems found in Europe, the U.K., and Australia,
will typically serve 60 to 120 consumers with a single trans-
former. The customers may be three or single phase. Some ef-
forts are made at construction to balance the phase loading but
significant unbalances develop during normal operation. While
the systems are robust, unbalance has undesirable effects in-
cluding reduced transformer life, increased losses and power
quality problems due to phase voltage variations and negative
sequence voltages.
Transformers operated under unbalanced conditions will
suffer more extreme stresses than under balanced conditions.
The transformer life is largely determined by the insulation
life [1]–[3]. Mechanical, electrical, and thermal stresses affect
the oil-paper insulation system [4]. The main factors that
determine the insulation life of oil-immersed transformers are
the transformer load, ambient temperature, moisture content
and the oxygen content of the oil [5]. For unbalanced loading
the resulting increased loss, and the concentration of the losses
Manuscript received February 13, 2013; revised May 21, 2013 and August
20, 2013; accepted October 16, 2013. Date of publication November 26, 2013;
date of current version May 20, 2014. This work was supported by the National
and International Research Alliances Program of the Queensland Government.
Paper no. TPWRD-00146-2013.
H. Pezeshki and G. Ledwich are with the School of Electrical Engineering
and Computer Science. Queensland University of Technology, Brisbane 4000,
Australia (e-mail: houman.pezeshki@student.qut.edu.au).
P. J. Wolfs is with the Power and Energy Centre, CQUniversity, North Rock-
hampton, Queensland, Australia (e-mail: p.wolfs@cqu.edu.au).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TPWRD.2013.2287002
in one or two phases, affects the insulation system of the
transformer and reduces its life time [6], [7].
To maximize the return on their investment, utilities will take
advantage of a transformer’s full cyclic loading capability to
achieve to financial savings and reduced operating costs. Op-
timal utilization of a transformer can be achieved by taking ad-
vantage of a transformer’s thermal time constant and the diurnal
variation of the load and ambient temperature. It is necessary to
have accurate models for predicting winding hot-spot tempera-
ture (HST) and top-oil temperature (TOT).
The development of accurate prediction models of the HST
and TOT for substation, distribution and power transformers has
been the subject of a substantial amount of research [15]–[17].
IEEE Standard C57.91-1995 [8] and IEC standard 60076-7
[9] describe in detail methods to calculate the HST and offer
guidance on temperatures that should not be exceeded at either
winding or structural hotspots to avoid undue aging failures
from gassing. These standards, and recent publications, assume
balanced loading of the transformer. Residential transformers
have a high degree of unbalance. It is practically difficult to
maintain an accurate knowledge of the street phase connections
due to network maintenance and recording errors.
PV at the distribution level has become widespread. Previous
studies [24]–[26] have identified many impacts that roof top PV
mayhaveonalocaldistributionnetwork including changes in
voltage profile and network power flows [24]. The problem of
voltage fluctuations resulting from the passage of clouds is also
addressed in [27], [28]. In particular, variations of nodal volt-
ages in small or weak electrical grids (e.g., SWER systems) have
been reported to cause system instability. Studies have also been
conducted to explore the extent to which the geographical diver-
sity of distributed PV mitigates the short term output variability
caused by rapidly changing weather conditions. Spatial distri-
bution significantly reduces transients caused by clouds.
Distribution systems are typically designed for specificload
profile based on consumption patterns. When roof top PVs are
deployed, the pattern of electric power demand will change.
Australian residential consumption has an early evening peak.
The addition of PV does not strongly reduce the peak load but
will reduce the energy served. As a result the load factor, the
ratio of average to peak load, is reduced. This paper studies the
impact of roof top PV on the transformer insulation life. A dy-
namic thermal model was used for the prediction of the hot-spot
temperature. The insulation aging impact was analysed using
one year of residential electric power load data, drawn from the
Perth Solar City High Penetration PV Trial, [10]. One year of
ambient temperature data is integrated into the model to esti-
mate the life impact.
0885-8977 © 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
PEZESHKI et al.: IMPACT OF HIGH PV PENETRATION ON DISTRIBUTION TRANSFORMER INSULATION LIFE 1213
The work can be separated into two main steps. The first step
is to identify the consumer phase connection and to process
smart meter data to allow two data sets to be established. These
data sets are the actual transformer phase loading and the
loading that would have resulted in the absence of the installed
PV systems. The second step is to use these two data sets to
calculate the transformer hot-spot and oil temperatures under
the different scenarios. The addition of PV is shown to be
beneficial with regard to hot-spot temperatures and reduces the
transformer loss of life (LOL).
II. DATA ACQUISITION AND PHASE ALLOCATION
The 400/230 V feeder, shown in Fig. 1, is supplied from a 200
kVA Dyn 22 kV/400 V distribution transformer and includes 77
residential consumers. Of these, 34 consumers have roof top PV
systems which have average ratings of 1.88 kW. The total in-
stalled PV capacity is 64 kW representing a penetration of 32%.
Load data, including energy consumption solar power gener-
ation, voltage and current is recorded by smart meters on the
Western Power network at the point of connection to each con-
sumer switchboard at 15-min intervals. Smart meter data has
been collected since July 2011. At the time of the recording there
were two three phase meters (meter number 49 and 55) that were
not active and no recording available for these meters.
To determine the loading of the transformer the authors have
previously published a method using cross correlation of con-
sumer voltage profiles to identify their phase connection [11].
Using the known phase connections of the residential loads, the
data collected from the smart meters was aggregated to deter-
mine the phase loading on the transformer. Fig. 2 shows the
predicted transformer loading (kW) during the 7-day window
that includes the annual peak day. The sampling rate is 15 min-
utes. The network under study is significantly unbalanced but
reflective of normal network conditions. The unbalance results
from the poor allocation of customer loading among the three
phases. For instance, the loading of phase A is much less than
phase B and C during day time peak hours.
III. THERMAL AGING FORMULATION
A. Loss of Life of Distribution Transformer
Several models have been introduced to assess life estima-
tion of insulation in transformers [1]–[4], [12], [13]. A wide va-
riety of methods has been presented for loss-of-life inference
for power and distribution transformers, such as those proposed
in [14], Clause 7 and updated in [15]. When inferring the trans-
former LOL acceleration rate using these methods, the calcu-
lation of the winding hot-spot temperature (HST) is the most
critical issue [3], [4]. The methods proposed in [1], [2] were
followed by a series of papers [5]–[7], [12], [13], [15] dealing
with more accurate calculations of HST.
Although deterioration of insulation is a function of temper-
ature, moisture content, oxygen content and acid content, the
model presented in this paper is based only on the insulation
temperature [9]. Since the temperature distribution is not uni-
form, the part that is operating at the highest temperature will
normally undergo the greatest deterioration. Therefore, the rate
Fig. 1. Perth Solar City High Penetration Feeder Site, image courtesy of
Western Power.
of aging is referred to the winding hot-spot temperature. Equa-
tions (1) and (2) describe, respectively, the relative aging rate
VT for a thermally upgraded paper (reference temperature of
110 ) and non-thermally upgraded paper (reference tempera-
ture of 98 C) [9]
(1)
(2)
Temperature is of importance since chemical reactions such
as the deterioration of cellulose in paper is accelerated at el-
evated temperatures. In Table II, the thermal model parame-
ters are presented. The equivalent life at the reference tempera-
ture that will be consumed in a given time period for an actual
temperature cycle can be calculated by (3) [8], where is
equivalent aging factor for the total time period, n is index for
the time interval t, N is total number of time intervals, is
the time interval and is aging acceleration factor for the time
interval
(3)
When a normal insulation life for a well-dried oxygen-free
transformer system is defined, percent loss of insulation life can
be calculated in (4) [8]. In this paper, we choose the normal life
as 180,000 hours (20.55 years). Under this normal life value,
normal percent loss of life for operation at a rated hot-spot tem-
perature of 110 for24his0.0133%
Loss of Life Normal insulation life (4)
The normal life expectancy is a conventional reference basis
for continuous duty under normal ambient temperature. and
rated operating conditions.
B. Hot-Spot Temperature Model
In [14], a transformer thermal model was developed as a se-
ries of algebraic difference equations. In [16], [17], Swift et al.
1214 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 29, NO. 3, JUNE 2014
Fig. 2. Pavetta transformer power output, January 21–27, 2012.
TAB L E I
TOP OIL AND HST COMPARISON UNDER DIFFERENT LOADING CONDITION
proposed a basic approach basedonheattransferbasedonthe
application of the lumped capacitance, thermal resistance elec-
trical analogy. The transformer heating model used in this anal-
ysis is based on [9] (Fig. 2) IEC 60076 develops the hotspot
temperature equations in the following way:
(5)
where is the HST in degrees Celsius, is the top-oil tem-
perature at the current load, and is the total HST rise at the
nth time step, where is calculated in (6)
(6)
and are derived from the difference equations
for HST rise, and can be calculated
(7)
where Dt is the time step in minutes, and are experimen-
tally-derived constants related to the thermal recovery of the
transformer, is the winding time constant in minutes,
is hotspot-to-top-oil gradient at rated current in Kelvin, Ka is
the load factor (current load/rated load), and y is the exponen-
tial power of current versus winding temperature rise (winding
exponent). Similarly, can be evaluated
(8)
where is the average oil time constant in minutes. The top-oil
temperature must be calculated and substituted back into (5)
(9)
Equations (7) and (8) would be accurate if all phases of a
three phase transformer are loaded identically or for single
phase transformers which are commonly used in North America
or in rural areas of Australia (e.g. SWER systems). However,
the loads on the phases of the typical three phase distribution
transformer are not balanced. It is possible to derive an expres-
sion analogous to (7) and (8) for each phase if the time varying
loads on each phase are known.
The phase currents of the transformer would determine the
winding to oil temperature differential of that phase so (8) could
be rewritten for each individual phase
(10)
(11)
(12)
The current load that would impact on the top oil tem-
perature would be the rms value of each individual phase current
at that given time
(13)
(14)
(15)
PEZESHKI et al.: IMPACT OF HIGH PV PENETRATION ON DISTRIBUTION TRANSFORMER INSULATION LIFE 1215
Fig. 3. Ambient temperature from June 2011 to July 2012.
C. Ambient Temperature and Roof top PV Generation
As described in the thermal model (9) and in publications
[12], [18], the ambient temperature affects the hot spot tem-
perature and impacts the life duration and the aging rate of
transformer. Therefore as one of the input to the thermal model,
one year ambient temperature data of July 2011–2012 were
collected from Australian Bureau of Metrology Perth Airport
weather station which is close to the high PV penetration
trial [19]. The ambient temperature and solar irradiance was
obtained at a 15-min rate to be consistent with the smart meter
load data sampling times (Fig. 3).
D. Household Load Profiles
The transformer daily load curve is determined by the aggre-
gated demand measured by the smart meters connected to in-
dividual consumers. In this work 15-min intervals are used, so
a daily load curve is made up of 96 pairs of time and demand
values. In order to guarantee a representative set of field data,
a total of 365 days of measurements were collected from op-
erating smart meters at the high PV penetration trial in Perth.
A snapshot of all household load profiles (current) is shown in
Fig. 4.
E. Distribution Transformers
The 102 node 400/230 V distribution network is connected
to the high voltage 22 kV Western Australia’s South-West In-
terconnected System (SWIS) through a 200 kVA distribution
transformer that complies with the prevailing Australian Stan-
dard AS2374. Within the Western Power service area, approx-
imately 17 000 distribution transformers are in service. More
than 3,000 of these are 200 kVA units. These transformers are
non-thermally upgraded paper and its life duration is 30 years.
The loading patterns of the distribution transformer shown in
Fig. 1 without and with rooftop PV generation is of interest
in this study. The transformer ratings and impedance values
are representative of current in-service distribution transformer
types used in Western Australia. Transformer data are listed in
Tables V and VII.
IV. RESULTS AND DISCUSSION
A generalized analysis framework was developed to inves-
tigate the distribution transformer loss of life under proposed
scenarios. In each scenario, the annual loss of life rate and the
expected lifetime of the transformers were determined. These
scenarios are:
1) unbalanced operating conditions (with solar input);
2) unbalanced operating conditions (no solar input);
3) balanced operating conditions (with solar input);
4) balanced operating conditions (no solar input).
A. Unbalanced Operating Conditions (With Solar Input)
To investigate the impact of PV onthelifeofthetransformer,
a one year set of 15-min measurements of transformer load
and ambient temperature was assembled. Equations (1)–(5), to-
gether with transformer thermal parameters, were used to de-
termine the transformer thermal response (Fig. 5). Equations
(10)–(14) of Section III-D, werethenusedtoobtainanLOL
rate, and total LOL accumulated by the transformer over the
given year.
The distribution transformer under study is substantially un-
balanced. Out of 77 connected residential consumers, 13 are
connected to phase A, 17 connected to phase while phase C
is serving 21 customers and the 26 of the premises have three
phase connection. Fig. 5 shows the temperature profiles corre-
sponding to one summer week during the trial that includes the
annual peak day for the transformer. The peak demand day oc-
curred on the second day of a heat wave1and immediately pre-
ceded the Australia Day public holiday.
It is evident that phase C is heavily loaded. At the peak time
the loading on phase C is 360 A/90 kVA or 1.34 p.u. and this
value is close to the allowable maximum cyclic loading. In Aus-
tralia it is acceptable practice to load a transformer up to 1.4 its
ratingforshortperiodoftimeinagivenyear[20].Inthisin-
stance the utility company would not notice this overloading
incident as the total energy sales from the transformer are used
to predict peak loads. The energy sales are aggregated over the
three phases, which at peak time was 196 kVA, and not the in-
dividual phase loading. Based on this approach the transformer
will be kept in service until the total loading on it would reach
1.4 p.u. or 280 kVA, this assumption has been used as a basis
to create four test cases. These investigate the LOL of the trans-
former if the loading on the transformer increases in 10% incre-
ments until it reaches the set value of 280 kVA (1.4 p.u.).
Case 1 illustrates the transformer HST and LOL quantities
that correspond to the current unbalanced state of the trans-
former with 64 kW of PV. The results are presented for the peak
day in the summer, Fig. 5(a) and Fig. 6(a) as well as for the day
with lowest load in the winter, Fig. 5(b) and Fig. 6(b), for each
phase of the transformer. The horizontal axis is the time of day
in 15-min intervals. In Fig. 5 the vertical axis is HST, in Fig. 6
the vertical axis is LOL.
From Fig. 5 it is clear that the unbalance has caused different
hotspot temperatures in each leg of the transformer. For example
on the peak day the phase A winding would reach 90 C whereas
phase C winding exceeds 130 C. This 40 temperature dif-
ference drives the rapid degradation of the phase C insulation.
This temperature difference is much less at lighter loads (Fig. 6).
As can be seen in Fig. 6 in the summer, the LOL is dominated
by the higher transformer temperatures during the late afternoon
and evening peak. Mention should be made of the high LOL rate
of the phase C, in fact, it exceeds the design rate of 1-day per 24
h,bylosingmorethan3daysin24h.Onthecontrary,inwinter,
1This discussion is based on the Bureau of Meteorology’s definition of a heat
wave as three or more consecutive days with daily maximum temperatures ex-
ceeding 35 C.
1216 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 29, NO. 3, JUNE 2014
Fig. 4. Typical daily load profile of each of the 75 customers on Pavetta. (Vertical axis: Time (Hour); horizontal axis: Current (A)).
Fig. 5. Comparison of the daily evolution of the hotspot, top oil, and ambient
temperature peak day (a) low load day (b).
the peak of the LOL rate is well below the designed value. This
is due to the moderate loads combined with relatively low am-
bient temperature.
To see how future load growth would affect the transformer
hot-spot temperature, simulations were carried out and com-
pared together with the base load case (cases 2–4). For case 5, a
worst case scenario is investigated by increasing the load to 1.4
p.u. (280 kVA).
Fig. 7 shows the top oil temperature and HST on the peak
summer day when the transformer is loaded 40% above the
nameplate 200 kVA rating and compares them with reference
case. The oil reached 97 and hot-spot temperatures for each
phase reached 126 C, 162 C, and 185 C, respectively. The
HST limit of 160 C was thus violated for both of the phases B
and C, and rapid degradation is expected.
Fig. 6. Daily evolution of the loss of life on each phase of transformer, peak
day (a) low load day (b).
B. Unbalanced Operating Conditions (No Solar Input)
In order to demonstrate the benefit that roof top PV could pro-
vide in reducing the transformer loss of life, the solar generation
was removed the system. The production pattern of PV units
was obtained from the calculated and collected values using
the solar irradiance measurements during the first three months
of the trial (July–September 2011) and the smart meter data in
15-min interval during the rest of the period of the trial (October
2011–June 2012). For the first three months of the trial, only
net household consumption data was available. In the last nine
months of the trial, a two channel record of household load and
solar generation was available for all single phase customers.
The solar generation profile of the first three months was es-
timated using solar irradiance, ambient temperature and rating
information for the PV modules and inverter. The method was
confirmed by correlating with generation pattern of the last nine
months of the trial. Of the 34 premises with PV, 12 houses had
dual reading meters that captured both the PV generation and
consumption of the houses. The PV generation for the other 22
PEZESHKI et al.: IMPACT OF HIGH PV PENETRATION ON DISTRIBUTION TRANSFORMER INSULATION LIFE 1217
Fig. 7. Effect of possible load growth on TOT and HST.
Fig. 8. Improvement in top oil temperature and HST in the presence of PV.
houses which only had net meter recording could be calculated
from these observations.
The results for five operating conditions are shown in Fig. 8.
In each graph the dotted line represents the system with the PV
and the solid line the system without PV. The first row is the
reference case (current state of the transformer), cases for addi-
tional loadings to 1.3 p.u. are shown in this figure. The final 1.4
Fig. 9. Temperature difference in hot spot and oil of transformer and reduction
in LOL as a result of PV generation.
p.u. loading case will be considered separately. Fig. 9 shows the
temperature profiles corresponding to a peak transformer over-
load of 1.4 p.u.
Without PV generation, the oil and hot-spot temperatures
reached 100 Cand190 C, respectively. The addition of 64
kW of PV generation lowered this to 180 CfortheHSTand
95 C for the top oil. These values are still extremely high.
The PV benefit occurs during the time leading up to the peak.
Lower loadings in the afternoon allow the transformer to enter
the peak period with lower oil temperatures. In this example
the LOL saving for Phase A, B, and C is 0.2,14, and 160 days
for each phase, respectively. Except for cases where the PV
installations are larger than the peak load, PV will decrease the
daily top oil temperature and HST and extend transformer life.
The extent of the improvement depends on the loading ratio of
the transformer and the PV penetration level.
Table II provides a summary on each phase of transformer
LOL and the benefit that roof top PV could provide to improve
the transformer aging process based on its current and future
loading. The first conclusion from Table II is that regardless
of the operation scenario, the LOL rate of the transformer is
far higher in summer. This may be due to the combined ef-
fect of higher ambient temperature and electricity use driven
by cooling loads in this season. It further implies that roof top
1218 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 29, NO. 3, JUNE 2014
TAB L E II
SEASONAL VARIATION IN LOL OF TRANSFORMER
TAB L E III
LOSS OF LIFE IMPROVEMENT WITH PV GROWTH
PV could provide a higher LOL reduction in the summer and
is a suitable option for distribution transformer-life extension.
The targeted installation of roof top PVs along the feeder, and
even on a specific phase, could be considered as a life exten-
sion strategy. There are voltage rise limitations on the number
and location of the installed roof top PVs. Considering this 7 ad-
ditional PVs (with average rating of 1.88 kW) were randomly
allocated to consumers on Phase C. Table III shows the corre-
sponding life improvement.
C. Balanced Operating Conditions (With Solar Input)
In the first two scenarios the transformer was significantly
unbalanced. In the last two scenarios examine the benefitof
balanced operation with PV generation. Load balance can be
achieved using a distribution STATCOM or optimal rephasing
strategies with laterals or individual loads [21]–[23]. Phase
identification systems introduced in [11], can be combined
with rephrasing to improve balance. Fig. 10 compares the
transformer peak day when the transformer is balanced to the
current unbalanced case. The lower phase C current reduces
the peak time HST to 115 Cfrom130 C. A reduction of 15
C in HST translates into reduction in LOL of approximately 2
days. It should be noted in Fig. 10 phase A and B would have
experience higher temperatures and age faster. The benefitis
that whole transformer will age at the same rate.
D. Balanced Operating Conditions (No Solar Input)
To conduct a comprehensive comparison, the daily HST and
annual LOL rates were calculated when the transformer was
balanced and had no PV connection. Table IV shows that the
Fig. 10. Balanced versus unbalanced loading: HST, TOT and LOL.
transformer will suffer from rapid loss of life when the winding
is under excessive stress. Balancing the phases will assure ex-
cessive LOL does not occur in one phase. Important benefits
may be realized due to the life extension of distribution trans-
formers brought about by customer-owned PV units even when
the transformer is balanced.
It should be noted that under scenario 2, loading of 1.4 p.u.,
the transformer would lose more than 12 years of its life in one
year. If the transformer was not upgraded it could reach its end
of life within 1–2 years of operation.
V. C ONCLUSION AND FUTURE WORK
PV generation will extend the life of oil-immersed distribu-
tion transformers even when the peak demand occurs well after
sunset. The presented results correspond to a three phase res-
idential transformer, but the result of this study could also be
PEZESHKI et al.: IMPACT OF HIGH PV PENETRATION ON DISTRIBUTION TRANSFORMER INSULATION LIFE 1219
TAB L E IV
LOL UNDER DIFFERENT LOADING SCENARIOS
(BALANCED VERSUS UNBALANCED)
TAB L E V
ELECTRIC CHARACTERISTICS OF THE TRANSFORMER UNDER STUDY
TAB L E VI
GEOMETRIC CHARACTERISTICS OF THE TRANSFORMER UNDER STUDY
TAB L E VII
PARAMETERS FOR THE THERMAL MODEL OF THE 200-kVA TRANSFORMER
applied to single phase distribution transformers. The impact
of PV on single phase distribution transformer (in the US, or
SWER in Australia) is similar to the case when the three phase
transformer is balanced (scenarios C and D). It is expected as the
coincidence of PV with commercial load is higher better life ex-
tension will result for commercial load transformers.
The main focus of this paper is on the impact of PV on three
phase transformer, but the method explored in this paper is ap-
plicable to any other form of single phase generator such as
combined heat and power fuel cell modules, which are not dis-
patchable and are driven by the demand (i.e. hot water) of the
household.
A thermal model was developed to assess the transformer
temperatures over a 12 month cycle allowing a cumulative mea-
sure of loss of life to be determined for various scenarios. This
paper is based on 15-min field data and captures the impact of
solar variability at these time scales. The variations in irradiance
produced by changes in cloud cover can cause faster fluctuations
in the power generated by roof top PV. The short fluctuations
(less than 15 min) would not have a significant effect on oil tem-
perature (with time constant of 180 min) but could change the
winding temperature in a magnitude of 2–3 (the winding time
constant is 10 min). This will not significantly contribute to the
aggregated loss of life given the short duration.
Finally the general trend of life improvement will increase
with PV penetration until power flow reversals, comparable to
the peak demand, occur. At this point the additional winding
losses become significant.
ACKNOWLEDGMENT
The authors acknowledge the supply of consumption data
collected under the Perth Solar City trial which is a part of the
Australian Government’s U.S.$94 million Solar Cities Program.
The authors also acknowledge the support of Western Power in
supplying network data, models and technical reports.
REFERENCES
[1] G. C. Montanari, “Aging and life models for insulation systems based
on PD detection,” IEEE Trans. Dielectr. Electr. Insul., vol. 2, no. 4, pp.
667–675, Aug. 1995.
[2] G. C. Montanari and L. Simoni, “Aging phenomenology and mod-
eling,” IEEE Trans. Electrical Insulation, vol. 28, no. 5, pp. 755–776,
Oct. 1993.
[3] L. Simoni, “A general phenomenological life model for insulating ma-
terials under combined stresses,” IEEE Trans. Dielectr. Electr. Insul.,
vol. 6, no. 2, pp. 250–258, Apr. 1999.
[4] C. Dervos, P. D. Bourkas, E. A. Kayafas, and I. A. Stathopulos, “En-
hanced partial discharges due to temperature increase in the combined
system of a solid-liquid dielectric,” IEEE Trans. Elect. Insul., vol. 25,
no. 3, pp. 469–474, Jun. 1990.
[5] J.P.Crine,“Ontheinterpretation of some electrical aging and relax-
ation phenomena in solid dielectrics,” IEEE Trans. Dielectr. Electr.
Insul., vol. 12, no. 6, pp. 1089–1107, Dec. 2005.
[6] T. K. Saha and P. Purkait, “Investigation of polarization and depolar-
ization current measurements for the assessment of oil-paper insulation
of aged transformers,” IEEE Trans. Dielectr. Electr. Insul., vol. 11, no.
1, pp. 144–154, Feb. 2004.
[7] T. K. Saha and P. Purkait, “Investigations of temperature effects on the
dielectric response measurements of transformer oil-paper insulation
system,” IEEE Trans. Power Del., vol. 23, no. 1, pp. 252–260, Jan.
2008.
[8] IEEE Guide for Loading Mineral-Oil-Immersed Transformers and
Step-Voltage Regulators, IEEE Standard C57.91-201, 2012.
[9] IEC Loading Guide for Oil Immersed Power Transformers,IECStan-
dard 60076-7, 2005.
[10] Solar City High Penetration PV Trial. 2012. [Online]. Available: www.
perthsolarcity.com.au/solarcollective
[11] H. Pezeshki and P. Wolfs, “Consumer phase identification in a three
phase unbalanced LV distribution network,” in Proc. IEEE Power En-
ergy Soc. ISGT Eur., 2012, pp. 1–7.
[12] Q. He, J. Si, and D. J. Tylavsky, “Prediction of top-oil temperature for
transformers using neural networks,” IEEE Trans. Power Del., vol. 15,
no. 4, pp. 1205–1211, Oct. 2000.
[13] D. J. Tylavsky, H. Qing, G. A. McCulla, and J. R. Hunt, “Sources
of error in substation distribution transformer dynamic thermal mod-
eling,” IEEE Trans. Power Del., vol. 15, no. 1, pp. 178–185, Jan. 2000.
[14] IEEE Standard Test Procedure for Thermal Evaluation of Insulation
Systems for Liquid-Immersed Distribution and Power Transformers,
IEEE Standard C57.100-2011, 2012, pp. 1–37.
[15] L. W. Pierce, “An investigation of the thermal performance of an oil
filled transformer winding,” IEEE Trans. Power Del., vol. 7, no. 3, pp.
1347–1358, Jul. 1992.
[16] G. Swift, T. S. Molinski, and W. Lehn, “A fundamental approach to
transformer thermal modeling. I. Theory and equivalent circuit,” IEEE
Trans. Power Del., vol. 16, no. 1, pp. 171–175, Jan. 2001.
[17] G. Swift, T. S. Molinski, R. Bray, and R. Menzies, “A fundamental
approach to transformer thermal modeling . II. Field verification,” IEEE
Trans. Power Del., vol. 16, no. 1, pp. 176–180, Jan. 2001.
1220 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 29, NO. 3, JUNE 2014
[18] H. A. Gil and G. Joos, “Models for quantifying the economic benefits
of distributed generation,” IEEE Trans. Power Syst., vol. 23, no. 2, pp.
327–335, May 2008.
[19] Australian Bureau of Metrology, Recent Weather Patterns and Aus-
tralian Bureau of Meterology. Recent weather patterns. Nov. 5, 2012.
[Online]. Available: http://www.bom.gov.au/wa/?ref=hdr
[20] Australian Standard AS2374 Distribution Transformers Overview and
Test Procedures, Australian Standard AS2374, 2007.
[21] M.Y.Huang,C.S.Chen,C.H.Lin,M.S.Kang,H.J.Chuang,and
C. W. Huang, “Three-phase balancing of distribution feeders using
immune algorithm,” IET Gen., Transm. Distrib., vol. 2, pp. 383–392,
2008.
[22] L. Chia-Hung, C. Chao-Shun, C. Hui-Jen, H. Ming-Yang, and H.
Chia-Wen, “An expert system for three-phase balancing of distribution
feeders,” IEEE Trans. Power Syst., vol. 23, no. 3, pp. 1488–1496,
Aug. 2008.
[23] A. Ghosh and G. F. Ledwich, Power Quality Enhancement Using
Custom Power Devices. Norwell, MA, USA: Kluwer, 2002.
[24] R. A. Walling, R. Saint, R. C. Dugan, J. Burke, and L. A. Kojovic,
“Summary of distributed resources impact on power delivery systems,”
IEEE Trans. Power Del., vol. 23, no. 3, pp. 1636–1644, Jul. 2008.
[25] K. M. N. Syafii and M. Abdel-Akher, “Analysis of three phase distri-
bution networks with distributed generation,” in Proc. IEEE PECon,
2008, pp. 1563–1568.
[26] F.Shahnia,R.Majumder,A.Ghosh,G.Ledwich,andF.Zare,“Sen-
sitivity analysis of voltage imbalance in distribution networks with
rooftop PVs,” in Proc. IEEE Power Energy Soc. Gen. Meeting, 2010,
pp. 1–8.
[27] G. K. Ari and Y. Baghzouz, “Impact of high PV penetration on voltage
regulation in electrical distribution systems,” in Proc. Clean Elect.
Power, Jun. 14–16, 2011, pp. 744, 748.
[28] A. Woyte, V. V. Thong, R. Belmans, and J. Nijs, “Voltage fluctuations
on distribution level introduced by photovoltaic systems,” IEEE Trans.
Energy Convers., vol. 21, no. 1, pp. 202–209, Mar. 2006.
Houman Pezeshki (S’07–M’09) received the
M.Eng. degree in electrical and power engineering
from Murdoch University, Perth Australia, in 2009
and is currently pursuing the Ph.D. degree in
electrical and computer engineering at Queensland
University of Technology, Brisbane, Queensland,
Australia.
His current research interests include smart-grid
technology especially in reference to distributed
renewable resources, power-electronics applications,
and energy-management systems.
Peter J. Wolfs (S’79–M’80–SM’97) is the Di-
rector of the Power and Energy Centre at Central
Queensland University, Rockhampton, Australia.
His research interests include smart-grid technology,
distributed renewable resources, and energy storage
and their impact on system capacity and power
quality, the support of weak rural feeders, and the
remote-area power supply.
Gerard Ledwich (SM’89) is a Professor of
Electrical Power Engineering at the Queensland
University of Technology, Brisbane, Queensland,
Australia, and Fellow of the Institution of Engineers
Australia. His current projects are in the implemen-
tation of the microgrid laboratory, wide-area control
of transmission systems, optimized investment in
distribution systems to cover new technologies
and long-term planning, demand management for
distribution peak demand, and condition monitoring
techniques for large transformers with a particular
interest in online tools. He has published one book, 3 chapters, 133 journal
papers, and more than 231 refereed conference papers. His research interests
include control systems, power electronics, power systems condition moni-
toring, and distributed generation.

































