1
New Indices to Evaluate the Impact of
Harmonic Currents on Power Transformers
Pooya Bagheri, Wilsun Xu, Fellow, IEEE and Kiarash Shaloudegi
 regular-designed transformer could be a cheaper alternative
Abstract— This paper presents a new concept to quantify the
impact of harmonic currents on power transformers. The basic
idea is to convert the impact of harmonic currents into the impact
of an equivalent additional fundamental frequency current. Two
equivalent loading indices are developed accordingly with the
support of well-established industry standards. Several analytical
examples are presented to demonstrate the merits of the proposed
approach. Based on field data, the method has been applied to
assess the impact of nonlinear residential loads on distribution
transformers. It was found that the impact of residential load
harmonics on transformers is approximately proportional to the
square of current THD. Other applications of the proposed
method include estimating transformer overloading level due to
harmonics and determining the rating of a regular transformer
that can operate properly for a given harmonic distortion
condition.
Index Terms— Power Quality, Harmonics, Transformers
I. INTRODUCTION
A POWER transformer is an important asset of utility
companies and it directly affects the reliability, security
and efficiency of power systems. The impact of distorted or
harmonic currents on transformers has been a concern to
industry for many years. For example, k-factor has been
developed and widely used to characterize the impact [1]-[3].
Based on the fundamental frequency and harmonic currents
passing through a location, a k-factor index can be calculated
for that location. The value is then used to select a k-rated
transformer. This transformer is expected to operate properly
under the harmonic condition of that location.
The k-factor index has been found useful for selecting k-
rated transformers by owners of industrial and commercial
facilities. However, it cannot address at least two needs that
are of high interest to utility companies, as explained below
1) K-rated transformers are not normally used by utility
companies. These companies are interested in using an
oversized transformer to serve harmonic-rich feeders. At
present, there is no method or index to guide the over-
sizing of a transformer under harmonic conditions. In fact,
facility owners will also benefit from a method that can
recommend a larger transformer as oppose to a k-rated
transformer under given harmonic conditions. A larger but
The authors are with the department of Electrical and Computer
Engineering, University of Alberta, Edmonton, AB T6G 2V4, Canada
(emails: wxu@ualberta.ca, pbagheri@ualberta.ca, shaloude@ualberta.ca).
to a k-rated transformer.
2) Many utility transformers have been in operation for years
and they experience increased harmonic currents. Utility
companies are interested in knowing if the harmonic
currents are overloading the transformers and by how
much. Such information will be very useful for a company
to make an informed decision on if harmonic mitigation
measure must be taken. With the increased penetration of
nonlinear loads in homes and residential feeders [4]-[6], it
has become important to utility companies to know how
their distribution transformers are affected.
Several efforts have been made to analyze and quantify the
harmonic effects on transformers. References [7]-[9] have
tried to model the eddy current loss for a wide range including
harmonic frequencies. The work of [10] has extended the loss
studies to include the shell-type transformers as well.
References [11] and [12] presented a thorough study of
harmonics impact on a transformer performance for both of the
dry and liquid-filled types. Moreover, works such as in [13]-
[14] and [15] have conducted several experimental tests to
calculate the eddy current loss in single phase and three phase
transformers respectively. In [16], the Finite Element Method
(FEM) simulations have also been employed to model the
transformer loading abilities under harmonic conditions.
Finally, the IEEE std. C57.110 [17] can be considered as an
ultimate and comprehensive summary of all the developed
loading assessment methods in the literature such as the k-
factor, harmonic loss factor (FHL) and etc. However, the
results are not sufficient to address the two needs explained
earlier.
In response to this situation, this paper introduces a new
concept, called equivalent loading, to characterize the
harmonic impact on transformers. The basic idea is to convert
the impact of harmonic currents into the impact of an
equivalent additional fundamental frequency current. With the
proposed approach, it becomes possible to quantify the impact
of harmonics as an increase in the 60Hz current. Thus,
selecting a properly oversized transformer to handle harmonic
currents becomes possible.
The paper is organized as follows. Section II presents a
brief review on different types of transformer power loss and
the way they are influenced by harmonic currents. Section III
introduces new equivalent loading indices to quantify the
impact of distorted current on a transformer. Field
measurement data are analyzed in Section IV as an example
application of the proposed methodology. In addition, a
technique will be also presented and applied on the sample
data to estimate loading condition of residential transformers

in future.
II. BACKGROUND ON TRANSFORMER POWER LOSS
The main effect of a nonlinear current on a transformer is
the increased power loss. The high frequency current
components cause extra power losses which may result in a
transformer overloading. Overheating of transformers is one of
the main reasons for insulation ageing and consequently their
loss of life [18]. Fig. 1 graphically classifies the different
power losses in a transformer. This categorization is based on
IEEE Std. C57.110 [17] which divides the total transformer
loss into no load loss (excitation loss) and load loss. In this
section, a brief review on different types of transformer loss is
presented where the harmonics impact on each of them is
discussed as well.
Ohmic Loss
Load Loss
Total Stray Loss
Transformer
Loss
Hysteresis Loss
No Load Loss
Eddy Current Loss
Fig. 1 Classification diagram of different transformer losses
A. No-Load Loss
No-load losses can be subdivided into two main parts: the
core eddy current loss and hysteresis loss. The excitation
current loss also constitutes a small portion of no load losses.
Basically, the amount of no-load loss is a function of
lamination thickness, core material and the applied voltage
frequency [19]. Indeed, this loss is mostly related to the
voltage (and its associated harmonics) applied to the
transformer. Harmonic currents can theoretically affect voltage
distortion by flowing through system impedance, but the
impact is generally negligible. Besides, the no-load loss mostly
emerge at the less overheating-concerned parts of the
transformer such as core and structural body rather than the
winding portion which involve the most concerned spots
subject to the risk of overloading. Consequently, the effect of
harmonic currents on increasing the transformer no-load loss
can be neglected in this study [17].
B. Load Loss
Load loss relates to the portion of power loss that is
produced due to the load current flowing through the
transformer windings. It can be divided into two main parts:
ohmic loss (RI2 loss or resistive loss) and stray loss (eddy
current loss) [17].
Ohmic part represents the well-known resistive power loss
associated with current flow in winding conductors and it is
mainly affected by the conductor material, cross section and
length of wires in the transformer winding. Therefore, main
factors for determining the ohmic loss are the Root Mean
Square (RMS) value of current and total resistance of wires in
the windings. However, different frequency of each harmonic
component can slightly increase the ohmic loss by changing
the resistance value due to the skin effect. By neglecting the
skin effect, ohmic power loss (PC) becomes proportional to
2
square of the total RMS current even in presence of harmonic
distortions:
PC   I h2 (1)
h
Where Ih is the RMS of hth order harmonic current. In other
words, the mere fact that a current is distorted does not itself
increase ohmic loss in a transformer unless the distortions have
contributed to a higher current level.
The other part known as stray loss is due to parasitic
electromagnetic flux in the windings, core, core clamps,
magnetic shields, enclosure or tank wall of the transformer
[17]. Therefore, it highly depends on the size and design of a
transformer. Stray loss, itself, can be subdivided into two
components generally known as “winding stray loss” and
“other stray losses”. The winding stray loss emerges from
winding conductor strand eddy currents as well as circulating
Winding Stray
currents between strands or parallel winding circuits [17]. The
Loss
“other stray losses” portion, which essentially encompasses all
Other Stray the remaining parasitic losses, is mainly due to the stray
Losses
electromagnetic flux in transformer structural parts like core,
core clamp, magnetic shield enclosure and the tank walls. The
stray flux induces eddy currents in those parts that eventually
leads to this power loss.
Among different types of power loss, harmonic currents
affect the stray loss portion the most. At each harmonic
frequency, the winding stray loss (PEC) is approximately in
proportion to square of harmonic frequency order (h) and
harmonic current (Ih) as expressed below [17].
PEC   I h2 h 2 (2)
h
The other stray loss (POSL) portion is also proportional to
square of harmonic current component. However its
relationship with harmonic order is more complicated. Based
on several experimental studies, IEEE standard suggests that it
is relatively proportional to h0.8 for each harmonic component
[17].
POSL   I h2 h0.8 (3)
h
III. EQUIVALENT LOADING INDICES
In this section, the concept of representing the impact of
harmonic currents as the impact of an equivalent fundamental
frequency current is introduced. Two indices called
‘equivalent loading index’ and ‘harmonic impact on
transformer loading’ are proposed. The indices are developed
using the same theories behind the k-factor and other
transformer derating methods of the IEEE Std. C57.110 [17].
A. Index definition
The proposed concept can be explained as follows: a
transformer experiences X amperes of fundamental frequency
current and Y amperes of harmonic current. The resulting
power loss is Ploss. When the same transformer is supplied by
X+ΔX amperes of fundamental frequency current only, the
power loss also becomes Ploss. We can then consider that the
impact of the harmonic current is equivalent to increasing the
transformer fundamental frequency current from X to X+ΔX.
In other words, X+ΔX is the equivalent (60Hz) current

experienced by the transformer and ΔX represents the impact
of harmonics. Since, in a sinusoidal condition, the transformer
power loss is proportional to the square of the total RMS
current, the ratio of the transformer total load loss in the
harmonic operation (PTL) to the rated condition (PTL-R) can be
employed as to derive the equivalent current index (Ieq):
2
I eq
P I RMS ( pu)
2
 TL (4)
IR PTL  R
Where IR is the transformer rated current. Consequently, the
per-unit value of the proposed index can be determined
through the following equation:
I eq (1) PTL
I eq ( pu)   I eq ( pu)  (5)
IR PTL  R
As discussed in the previous sections, the total load loss of
a transformer consists of the ohmic power loss (RI2), eddy
current loss (PEC) and other stray loss (POSL) as below:
PTL  PC  PEC  POSL (6)
Now by considering the ohmic power loss at the rated
condition (RIR2) as the base value, PTL can be expressed in a
per-unit form by following equations [17],
 PTL ( pu)  PC ( pu)  PEC ( pu)  POSL ( pu)
 h  hm ax
 CP ( pu )  I 2
RMS ( pu )   I h2 ( pu)
 h 1
h  hm ax
 P ( pu)  P
EC  R ( pu )  I h ( pu ) h
2 2
 EC
h 1
 h  hm ax
 POSL ( pu)  POSL R ( pu) h1 I h ( pu) h
2 0.8
where Ih(pu) is the hth harmonic order of the transformer
current in per-unit form (Ih(pu)=Ih/IR). PEC-R(pu) and POSL-R(pu)
are respectively the ratio of the eddy-current power loss and
the other stray losses to the copper loss in the transformer
rated condition (In fact, PEC-R(pu) and POSL-R(pu) are constant
coefficients different for each transformer, highly dependent
on transformer size and construction. They will be further
discussed in the next sections). Applying (7) to the rated
sinusoidal current operation, PTL-R can be obtained as below.
PTL  R ( pu)  1  PEC  R ( pu)  POSL R ( pu) (8)
Finally, by substituting (7) and (8) in (5), the proposed
index can be derived as shown in (9).
h  hmax
2
I RMS ( pu)  PEC  R ( pu)  I h2 ( pu)h 2  POSL R ( pu)  I h2 ( pu)h0.8
I eq ( pu)  h 1
1  PEC  R ( pu)  POSL R ( pu)
When the calculated Ieq(pu) exceeds one, the total produced
loss is more than the maximum permissible power loss and it
indicates overloading of the transformer. Whereas, index
values less than one confirm that the transformer is safely
operating without being prone to overheating. To achieve a
better understanding of the index application, an example of a
100 kVA transformer supplying an 80kVA load can be
considered. For instance, a calculated equivalent index value
of 1.1pu indicates that the transformer is 10% overloaded.
Moreover, this index value reveals that the loading effect of
this harmonic load is similar to that of a pure linear load of
size 1.1pu * 100kVA=110 kVA.
Based on the above definition of equivalent loading, the
3
second index can be directly defined as well to quantify exact
increase of transformer loading due to non-sinusoidal
characteristic of the flowing current. We call the index
“Harmonic Impact on Transformer Loading (HITL)” which is
defined as below,
I eq ( pu)  I RMS ( pu)
HITL(%)   100 (10)
For the 100kVA transformer example discussed above,
HITL is thus 37.5% (=(110-80)/80).
B. Determining Transformer Parameters
In addition to per-unit harmonic distribution of transformer
current, the defined indices equivalent loading indices are also
dependent on transformer parameters such as PEC-R(pu) and
POSL-R(pu). These parameters vary among transformers of
different size and structure. As a general approximate rule,
they increase as the capacity of transformer increase, however,
two transformers of the same size can still have different
parameters due to dissimilar design types [12]. Most of the
developed calculation methods for these parameters involve
conducting complex computation tasks for each individual
transformer [12]-[13], [17]. Besides, such methods usually
require very detailed information about transformer design that
is generally not made available by manufacturers. Most
importantly, in the case of distribution transformers owned by
an electrical utility, the transformers are already installed and
(7)
operated, thus, it will be even a more difficult task to obtain
required parameters of such transformers.
Based on the data and analyses available in different
references ([7]-[17]), an extended range of 0.01~0.3 can be
assumed for PEC-R(pu) to count for almost every size and type
of different transformers. This paper adopts a conservative
approach to use the maximum end of this range, i.e. to
consider PEC-R(pu)=0.3 in loading assessment of any
transformer type and size. Then, the POSL-R(pu) parameter can
be derived accordingly based on the following approximate
ratios recommended by [17].
 33%  0.5, For Dry Type
POSL R ( pu)  67%
 (11)
PEC  R ( pu) 67%
  2, For Liquid  filled Type
 33%
Since most of the comon transformers are of liquid-filled
h  hmax
type, for the illustrative and application studies in this paper,
h 1 (9)
POSL-R(pu) is assumed to be the twice value of PEC-R(pu).
Apparently, all of the analyses can be performed in a similar
fashion for dry type transformers by inversing such assumption
(i.e. to estimate POSL-R(pu) as half of PEC-R(pu)).
C. Summary of the Method
Based on the above discussions, the overall proposed
methodology to quantify loading of a transformer serving
harmonic loads can be summarized as below:
1) Measure the transformer current. Then, derive the
individual harmonic current components and express them in
the per-unit format by using the transformer rated current as
the base value (Ih(pu)=Ih/IR).
2) For the transformer parameters, assume the PEC-R(pu) to
be 0.3 as a conservative approach and determine POSL-R(pu)
 4

parameter of the transformer according to (11) (Unless exact
value of PEC-R(pu) & POSL-R(pu) are available for the
transformer, which would be a rare scenario as discussed).
3) Finally, calculate the equivalent loading by using (9).
The maximum permissible value of this index to avoid an
overloading condition is 1pu. In order to quantify the
increased amount of loading due to the distorted nature of the
current, determine harmonic impact index as defined in (10).
D. Illustrative Examples
As an illustrative example, the proposed indices are
employed in this section for loading assessment of a
still less than THD in the percentage form. For example,
current of load #1 with a THD equal to 109.35% gives HITL
of 77.5%. In addition, the results once again show how k-
factor might be misleading in loading evaluation, e.g., the k-
factor calculated for load #1 is 8th times higher compared to
the corresponding one of load #3, however, actual ratio
between loading levels of the two cases is much less than 8
according to the proposed equivalent index (indeed, the index
shows that the current of load #1 increases the transformer
loading 73% more than the one of load #3).

H-Impact on Transformer Loading(%)

1.25 13
transformer supplying a sample nonlinear load. The example

Equivalent Loading Index (pu)

Equivalent Loading Index(pu)
load is a six-pulse converter (its harmonic spectrum is shown 1 Harmonic Impact
in Fig. 3). The equivalent loading index is calculated for the 12
RMS load current varying from 0 to 1 per-unit of the 0.75
transformer rated current. The result is shown in Fig. 2(a). In
addition, k-factor is also derived and presented in Fig. 2(b) to 0.5
11
provide a comparative analysis with the proposed index. The
0.25
following standard definition of k-factor is used [17].
h  hm ax
0 10
K  factor   I h ( pu) h 2 (12) 0 0.2 0.4 0.6
Total RMS Current(pu)
0.8 1
h  h1
(a)Equivalent Loading Index and HITL
As intuitively expected, both k-factor and the equivalent
3
loading indices become larger as the current increases. At the
0.9pu current, the equivalent loading reaches the 1.0pu value.
It states that the transformer in this example can supply the 2
K-factor

sample six-pulse converter type of load up to 90% of its rated

capacity. The results also show that equivalent loading is
linearly correlated with the total RMS current. HITL index is
also observed to be a constant value of 11% for different
current levels. It suggests the interesting fact that impact of
harmonics on transformer loading is merely dependent on
harmonic current distribution rather than the amount of load.
In other words, the percentage increase in transformer loading
due to harmonics is influenced by the load nature rather than
load size. Therefore, based on the proposed methodology, one
has to consider at least 11% capacity margin when choosing a
transformer size to serve this type of load.
Apparently, the k-factor plot (Fig. 2(b)) does not provide
any basis for such observations. Besides, it appears as a
nonlinear function of current. This is in contrast with the linear
relationship observed between equivalent loading and the
current. In this example, an evaluation based on k-factor might
lead to the inaccurate conclusion that for the loads larger than
60%, the transformer will be overheated and require a “k-
factor transformer”, which is not true according to the
equivalent loading index. In general, such comparative cases
clarify why traditional indices such as k-factor are not
sufficient to evaluate the equivalent loading of transformers
under non-sinusoidal conditions.
The indices are also calculated for other sample types of
nonlinear loads and the results are presented in Table I (The
harmonic current spectrums of the sample loads are shown on
Fig. 3). For all of the cases, the load size is assumed equal to
transformer size (i.e. IRMS=1pu). The results indicate different
transformer loading levels for different types of loads.
Apparently, the currents with larger THD values lead to higher
equivalent loading indices, however, in general HITL index is
1
0
0 0.2 0.4 0.6 0.8 1
Total RMS Current(pu)
(b)K-factor
Fig. 2 Loading assessment of a transformer supplying a sample 6-pulse
converter at different load RMS current levels
IV. FIELD MEASUREMENTS
In this section, the proposed methodology is applied to real
field measured data associated with a number of transformers
serving residential loads. The data belongs to 6 different
transformers located in Edmonton, Canada. The nominal size
and voltage rating of the transformers are 37.5kVA and
25kV:120V:120V (primary: secondary: secondary)
respectively. Current of each transformer is measured for
different period lengths varying from one to four days (the
whole set of data collectively includes the full-24 hour data for
23 days).
A. Method Application
As an application example, the introduced method is
deployed to obtain one-day equivalent loading profile for one
of the measured transformers. Figures 4 and 5 show the
harmonic content and average IDD & TDD (Based on [20]
definition) for the measured current of the chosen transformer.
The transformer nominal full-load current is selected as the
base value for the per-unit quantities of Fig. 4.
 5

Individual Harmonic Distortion (%)

100

80

60

40

20

0
2nd 3rd 4th 5th 7th 9th 11th 13th 15th THD
Harmonic Order
Switch Mode power supply 6-pulses Converter
12-pulses Converter 18-pulses Converter
24-pulses Converter Thyristor-controlled reactor (balanced)
Thyristor-controlled reactor (unbalanced) Arc furnace (active arc)
Arc furnace (stable arc)
Fig. 3 Harmonic spectrum of sample nonlinear load currents ([12] & [20])
TABLE I: The equivalent loading, Harmonic impact and k-factor indices for different sample nonlinear currents
Load Current The Equivalent Harmonic Impact on
Load Type K-factor
# THD (%) Loading Index(pu) Transformer Loading (%)
1 Switch Mode power supply 109.35 1.776 77.55 12.27
2 6-pulses Converter 21.46 1.153 15.28 2.80
3 12-pulses Converter 6.28 1.042 4.17 1.50
4 18-pulses Converter 3.64 1.012 1.18 1.14
5 24-pulses Converter 3.39 1.010 1.03 1.12
6 Thyristor-controlled reactor (balanced) 5.86 1.012 1.20 1.13
7 Thyristor-controlled reactor (unbalanced) 15.07 1.030 2.98 1.31
8 Arc furnace (active arc) 11.24 1.014 1.39 1.14
9 Arc furnace (stable arc) 2.90 1.001 0.14 1.01

The equivalent loading index calculated for different times

0.2 of the day is shown in Fig. 6. As observed, the index is below
Fundamental Component (divided by 5)
Harmonic current component (pu)

1 for the whole day, indicating that there is no instance of

3rd Harmonic
7th Harmonic overloading.
0.15
9th Harmonic Effect of harmonics is better observed in Fig. 8, where the
5th Harmonic
HITL index is plotted besides the current THD. In this case,
0.1
the impact of harmonics on transformer loading is not
significant by being less than 10% for most of the day. A
0.05 correlation between current THD and the Harmonic impact
index is clearly observed in this figure. The more distorted the
0 current (higher THD value), the more the impact of harmonics
0 6 12 18 24
Time (hour) is on transformer loading. Once again, it is worth to state that
Fig. 4 Daily harmonic currents profile for the sample transformer (The such observations would be absent if we limited our analysis to
fundamental component is presented by division on 5 to fit in the plot) usage of a traditional index such as k-factor (see Fig. 7).
8
B. Relationship with Current THD
Daily Average Current IDD(%)

Results of Fig. 8 indicated a strong correlation between

6
current THD and harmonics impact on loading of the
measured sample transformer. In this section, such relationship
4 is examined by analyzing the whole set of measurement data.
All of the measured data (whole 23 days associated with all of
2 the 6 transformers) is gathered to conduct this study. In order
to reduce complexity of this analysis, only 24 snapshots are
0 chosen from data of each day where each snapshot represents
3rd 5th 7th 9th 11th 13th 15th 17th 19th 21st 23rd 25th TDD
Harmonic Order
one hour (1AM, 2Am, 3AM, … ,11PM). So, in overall,
Fig. 5 Daily average harmonic current spectrum of the sample measured 552(=24×23) cases are present. For each individual case,
service transformer current THD and HITL indices are determined and the result is
 6

shown by the scatter plot of Fig. 9, where each of 552 points is n

presenting one snapshot of the data set.  [( yi  y )( xi  x )]
i 1
a n
(14)
1
 ( xi  x ) 2
i 1
0.8 Where n is total number of cases (=552), xi and yi reperesent
THDI(%)2 and HITL(%) of each cases respectively and
Loading(pu)

0.6 x represents average of all sample xi values [21].

The slope parameter (a) obtained by the explained method
0.4 is 0.0151. Therefore, one can estimate harmonic impact on
transformer loading for the measured service transformers by
0.2 Total RMS Current the following expression:
Equivalent Loading Index 2
HITL(%)  0.0151 [THD I (%)] (15)
0
0 6 12 18 24 The mean Standard Error (SE) of the above approximation
Time (hour) in the sample data can be also obtained as follows.
Fig. 6 Daily profile of total RMS current and equivalent loading index for the
n
measured service transformer
 [ HITL (%) Estimated  HITL (%) Actual ]2
i 1
1
Mean SE  (16)
n
0.8 The calculated error is 0.55%. Such negligible error
confirms that approximate equation (15) is capable to
K-factor

0.6
represent the actual relationship of HITL and THD in the field
0.4 measured sample data very accurately (This approximating
relationship is plotted as a solid green curve in Fig. 9).
0.2
20
H-Impact on Transformer Loading(%)

0
0 6 12 18 24
Time (hour) Approximate Curve
Fig. 7 Daily profile of k-factor for the measured service transformer 15

20 40
H-Impact on Transformer Loading(%)

10
H-Impact THD
Current THD(%)

10 20
0
0 10 20 30
THDI(%)
Fig. 9 HITL index and current THD for every measured hourly snapshots

0 0 20
H-Impact on Transformer Loading(%)

0 6 12 18 24 Regression Line
Time (hour)
Fig. 8 Daily profile of Harmonic Impact (HITL index) and current THD for
15
the measured service transformer

The strong positive correlation is obvious in the figure.

However, the relationship seems different from a simple linear 10
function, rather, it appears to have the form of a quadratic
correlation. Such hypothesis can be better evaluated by 5
plotting the HITL index versus square of current THD as
separately shown in Fig. 10. This figure clearly reveals that
HITL index follows a linear trend with square of current THD. 0
0 200 400 600 800 1000 1200
The relationship can be expressed as below. 2
[THDI(%)]
2
HITL(%)  a  [THD I (%)] (13) Fig. 10 HITL index versus square of current THD for every measured hourly
Where a is the slope of this linear relationship. We can use a snapshots
classic Least-Square Regression method to fit an approximate
line (solid green line in Fig. 10) to this set of data and estimate C. Future Loading Forecast
the slope parameter (a) [21]. This section presents a technique of utilizing the developed
theories of this paper in order to predict loading situation of a

transformer in future. It is of interest to determine the
maximum increase rate of harmonics that might put a specific
transformer in an undesirable overloading status at the future
time. For this purpose, the maximum allowed harmonic impact
index (HITL) can be determined as follows based on its
definition at (10).
1
HITL(%) max  100  [  1] (17)
(1  r )  I RMS ( pu) present
Where r is the expected increase rate of the total service load
amount. Then, according to the established relationship of
(15), one can approximatively estimate the maximum current
THD that avoids transformer overloading as below.
HITL (%) m ax
THD (%) m ax  THD (%) present 
HITL (%) present
As an illustrative example, this method is applied to the
data of sample measured transformer. It is assumed that size of
transformer loads do not change significantly in future, i.e.
r=0. Fig. 11 shows the present and calculated maximum
allowed levels of current THD associated with the whole
sample day.
By interpretation of Fig. 11 results, one can state that by a
safe margin, even if in future, the current THD increases up to
two times of its present level, the transformer of this example
can still operate without overloading.
150
Maximum Allowed Level
125
Current THD(%)
100
75
50
Present Level
25
0 [11]
0 6 12 18
Time (hour)
Fig. 11 Present and maximum allowed (to avoid overloading) current THD
profiles for the sample measured transformer
V. CONCLUSIONS
This paper presented a new concept and associated indices
to quantify the impact of harmonics on transformers. The
indices are developed based on well-established industry
standards. Different from existing indices such as the k-factors,
the application of the proposed indices is not limited to
selection of specially designed transformers. The indices can
reveal the “true” loading of a transformer under harmonic
conditions and they can be understood intuitively by people
with limited power quality knowledge. Potential use of the
indices includes estimating transformer overloading level due
to harmonics and determining the MVA size of a regular
transformer that can operate properly for given harmonic rich
conditions.
Studies on field measured data showed that the impact of
harmonics generated by residential loads on a transformer is
approximately proportional to square of transformer current
THD. For example, if a distorted current with THD of 8% is
7
currently increasing power loss in a transformer by 5%, one
can predict that a twice value of current THD (i.e. 16%) will
introduce four times amount of extra equivalent loading (i.e.
20%) in the same transformer. Such approximate relationship
can serve as a useful guideline for utilities to plan capacity
margin for new transformers serving residential loads.
VI. REFERENCES
[1] J. M. Frank, "Origin, development, and design of K-factor
transformers," IEEE Industry Applications Magazine, vol.3, no.5,
pp.67-69, Sep/Oct 1997.
[2] G. W. Massey, "Estimation methods for power system harmonic effects
on power distribution transformers," IEEE Trans. Industry Applications,
vol.30, no.2, pp.485-489, Mar/Apr 1994.
(18) [3] C. H. Kung, M.J. Devaney, and C.M. Huang, "The design of a hand-
held k-factor meter using an air-core current transformer," IEEE Trans.
Instrumentation and Measurement, vol.46, no.4, pp.811-816, Aug
1997.
[4] K. D. McBee and M. G. Simoes, "Evaluating the long-term impact of a
continuously increasing harmonic demand on feeder-level voltage
distortion," IEEE Trans. Industry Applications, vol. 50, no. 3, pp.
2142-2149, May./Jun. 2014.
[5] N. R. Watson, T. L. Scott, and S. Hirsch, "Implications for distribution
networks of high penetration of compact fluorescent lamps," IEEE
Trans. Power Delivery, vol. 24, no. 4, pp. 1521-1528, Jul. 2009.
[6] J. Chen, D. Salles, W. Xu, and W. Freitas, "Assessing the collective
harmonic impact of modern residential loads - Part II: Application,"
IEEE Trans. Power Delivery, vol. 27, no. 4, pp. 1947-1955, Oct. 2012.
[7] F. De Leon and A. Semlyen, "Time domain modeling of eddy current
effects for transformer transients," IEEE Trans. Power Delivery, vol.8,
no.1, pp.271-280, Jan 1993.
[8] F. De Leon and A. Semlyen, "Efficient calculation of elementary
parameters of transformers," IEEE Trans. Power Delivery, vol.7, no.1,
pp.376-383, Jan 1992.
[9] F. De Leon and A. Semlyen, "Detailed modeling of eddy current effects
for transformer transients," IEEE Trans. Power Delivery, vol.9, no.2,
pp.1143-1150, Apr 1994.
[10] M. W. Thomas, "Mathematical Model of Load Loss in Shell Form
Transformers," IEEE Trans. Power Apparatus and Systems, vol.PAS-
98, no.1, pp.174-180, Jan. 1979.
L. W. Pierce, "Transformer design and application considerations for
24
nonsinusoidal load currents," IEEE Trans. Industry Applications,
vol.32, no.3, pp.633-645, May/Jun 1996.
[12] D. E. Rice, "Adjustable Speed Drive and Power Rectifier Harmonics-
Their Effect on Power Systems Components," IEEE Trans. Industry
Applications, vol.IA-22, no.1, pp.161-177, Jan. 1986.
[13] E. F. Fuchs, D. Yildirim, and W. M. Grady, "Measurement of eddy-
current loss coefficient PEC-R, derating of single-phase transformers,
and comparison with k-factor approach," IEEE Trans. Power Delivery,
vol.15, no.1, pp.148-154, Jan 2000.
[14] D. Yildirim and E. F. Fuchs, "Measured transformer derating and
comparison with harmonic loss factor (FHL) approach," IEEE Trans.
Power Delivery, vol.15, no.1, pp.186-191, Jan 2000.
[15] E. F. Fuchs, L. Dingsheng and J. Martynaitis, "Measurement of three-
phase transformer derating and reactive power demand under nonlinear
loading conditions," IEEE Trans. Power Delivery, vol.21, no.2,
pp.665,672, April 2006.
[16] S. Taheri, A. Gholami, I. Fofana, and H. Taheri, “Modeling and
simulation of transformer loading capability and hot spot temperature
under harmonic conditions,” Electric Power Systems Research, Volume
86, May 2012, Pages 68-75.
[17] IEEE Recommended Practice for Establishing Liquid-Filled and Dry-
Type Power and Distribution Transformer Capability When Supplying
Nonsinusoidal Load Currents, IEEE Std C57.110-2008, Aug. 2008.
[18] K. T. Muthanna, A. Sarkar, K. Das, and K. Waldner, "Transformer
insulation life assessment," IEEE Trans. Power Delivery, vol.21, no.1,
pp. 150- 156, Jan. 2006.
[19] G. R. Slemon and A. Straughen, Electrical Machines, Addison-Wesley
Publishing Company, 1981.
 8

[20] IEEE Recommended Practices and Requirements for Harmonic Control

in Electrical Power Systems, IEEE Std. 519, 1992.
[21] D. S. Moore, Basic Practice of Statistics, 6th Edition, W. H. Freeman
and Company, 1995.

BIOGRAPHIES
Pooya Bagheri obtained the B.Sc. degree in electrical engineering from
Sharif University of Technology, Tehran, Iran, in 2010 and the M.Sc. degree
in electrical engineering from the University of Alberta, Edmonton, AB,
Canada, in 2013. He was an electrical engineer at Oil & Gas department of
Stantec Ltd., Edmonton, AB, Canada from 2013 to 2015. Currently, He is
pursuing his PhD degree in Electrical and Computer Engineering at the
University of Alberta. His research interests are power systems data analytics,
power quality and power distribution systems.

Wilsun Xu (M’90-SM’95-F’05) obtained the Ph.D. degree from the

University of British Columbia, Vancouver, in 1989. Currently, he is a
NSERC/iCORE Industrial Research Chair Professor at the University of
Alberta. His current main research interests are power quality and power
disturbance analytics.

Kiarash Shaloudegi received his B.Sc. degree in Electrical Engineering from

K.N.Toosi University of Tech. and M.Sc. degree in Electrical Engineering
from Amirkabir University of Tech. (Tehran Polytechnic), Tehran, Iran in
2009 and 2011, respectively. He is currently pursuing his Ph.D. program in
Electrical and Computer Engineering at University of Alberta, Canada. His
main research interests are Machine Learning, Data Mining, Power Quality
and Power Market.