_{1}_{0}
IEEE Transactions on Power Delivery, vol. 9, No. 1, January 1994
TRANSFORMER MAGNETIZING 
CURRENT 
AND 
IRONCORE LOSSES 
IN HARMONIC 
POWER 
FLOW 
M. A. S. Masoum, Member, IEEE
_{E}_{.} _{F}_{.} _{F}_{u}_{c}_{h}_{s}_{,} _{F}_{e}_{l}_{l}_{o}_{w}_{,} _{I}_{E}_{E}_{E}
Department of Electrical and Computer Engineering, University _{o}_{f} Colorado, Boulder, Colorado 80309
Keywords: load flow, anisotropic nonlinear transformers, harmonics, ironcore losses.
Abstract  The problem of modifying the harmonic power flow analysis to permit the inclusion of nonlin ear anisotropic transformers and computation of their
ironcore (and copper) losses, so far ne ected in power flow studies, constitutes the objective ofl this pa er. The nonlinear model used for transformers is capabE of sim ulating saturation of ironcores, anisotropy of lamina tions, and the ironcore and copper losses associated with anisotropic transformers, but simple enough to be included in the harmonic power flow algorithm without deteriorating its convergence properties. The de en dency of ironcore losses on the maximum value ofthe
fact th at ironcore losses are a 1unction of the wave
total fundamental and harmonic
flux density and the
form of the induced volta e that _{i}_{s} its harmonic phase
shifts with respect to the fundamental phase angle are included. The induced voltage is transformed from fre quency domain to time domain: the instantanenous induced voltage and the computed Xi characteristics are employed to compute the instantaneous magnetiz
ing and coreloss currents.
Thereafter, _{a} transformation
is made from time domain back to frequency domain to compute the fundamental and harmonic components of the abovementioned currents. Therefore, transformer harmonic couplin s are pro erly modeled and included in the analysis. $he modifed harmonic load flow for
mulation is applied to a balanced threephase feeder consisting of a groundedwye, groundedwye nonlinear anisotropic transformer and linear and nonlinear loads.
The main contribution of this paper is the computa
tion of ironcore
used in balanced
and copper) losses of transformers th eephase and singlephase systems
at (non)sinusoidal operation.
Introduction
The nonlinearity of power system components and loads results in nonsinusoidal currents and voltages which can propagate throughout the power system. The deviation from sinusoids is generally expressed in terms of harmonics. Consequences of power system harmon ics are the degradation of telephone communications caused by electric noise, amplification of harmonic levels resulting from series and parallel resonances [l], over volta es and, therefore, failures of capacitor banks and transformers [21.
Perhaps the best avenue for the harmonic analysis is the power flow (load flow) study, which is the calculation of the line loading given the genera tion and demand levels. The work on harmonic power
flow analysis was originated
about a decade ago by Xi?,
Heydt and Grady [3,4]: the HARMFLO program [4]in its present form is capable of analyzing power systems containing nonlinear loads with known vi characteris tics such as fullwave rectifiers, nonlinear resistors [3 HVDC converters [5] and gaseous discharge lighting dl
under balanced threephase conditions. The work on unbalanced fundamental load flow analysis originated
about two decades ago [7]. The recent multiphase har
monic load flow technique of
Xu, Marti and Dommel
bhot the balanced and unbalanced (fundamental and
8,9 
is 
nonlinearity 

these publications. 
a main contribution in this field. However, in
harmonic) power flow analyses, transformers are _{a}_{s}_{} sumed to be linear elements and as a result transformer
and ironcore losses have been neglected in
Transformers are one of the contributors to harmonic distortion. The subject of power quality dis tortion by nonlinear transformers has been investigated since the early 1900’s and many contributions have been made [10,11] including a nonlinear model of anisotropic
transformers capable of computing
istics and the fundamental
the Xi character
as well as the harmonic
(non)linear loads [13,14,1d _{3}_{.}
ironcore losses for (non sinusiodal excitations and
References 11 and 12 in
troduce a detailed algorithm for the inclusion of trans
harmonic power flow approximating mea sured Xi characteristics, however, transformer coreloss current and ironcore losses are neglected in their anal ysis.
In this paper, the nonlinear transformer model
of [13] will be used to modify the harmonic power flow
alogrithm of
[3,4] to permit the inclusion of transformer
magnetizing current and ironcore losses in power flow
studies. The main advanta e of the algorithm presented
in this paper compared to
ta at of references 11 and 12 is
the computation
of transformer ironcore losses within
In addition com
the harmonic load flow algorithm.
puted transformer Xi characteristics are used as com
Measurements are not always feasible and/or possibi e
pared to the measured ones (as proposed in
11 and 12 .
for large power transformers.
92 SM 4960 PWRD A paper recommended and approved by the IEEE Transmission and Distribution Committee of the IEEE Power Engineering Society for presenta tion at the IEEE/PES 1992 Summer Meeting, Seattle, WA, July 1216, 1992. Manuscript submitted January 20, 1992; made available for printing May 15, 1992.
LargeSignal Nonlinear Model for Anisotropic Transformers
Given the physical dimensions of an &SO tropic transformer and the measured with and cross grain BH characteristics of its nonlinear ironcore ma terial (as obtained from Epstein frame measurements), a quasi threedimensional magnetic field analysis is used in [13] to compute representative BH and loss density characteristics and the Xi curve of the trans
08858977/94/$0
14.00 @ 1992 IEEE
11
former. These representative characteristics which ac
count for the presence of interlamination and buttjoint
air gaps and the anisotropy of laminations are used
with phasefactor functions in the lar esignal nonlin
em equivalent circuit of transformers 141. The phase
factor functions account for the h3that ironcore losses
are a function of the waveform of the induced volt e,
that is its harmonic phase shifts with r ect to%e
fundamental phase an
e.
Therefore th3i and the
representative BH an d@lossdensity characteristics of a
nonlinear transformer must be computed in order to use
the largesignal nonlinear transformer equivalent circuit
of [13]. Note that these characteristics are to be com
puted only once for a given transformer type.
In order to include anisotropic transformers
in the harmonic power flow analysis, nonlinear three
phase power system transformers are assumed to consist
of three identical singlephase transformers. The above
assumption is due to the fact that a largesignal non
linear equivalent circuit for three and five legged three
phase transformers capable
ironcore losses has not yet been reporte8 in the liter
of computin
transformer
ature. The proposed model of [13] as shown in Figure
1 is capable of computing the magnetizing and core
loss currents for
a 'ven instantaneouslyinduced voltage
(Appendices I1 an3 111).
The 3'd order harmonic generated by trans
former saturation is of zero sequence (althou h the sys
tem is balanced). The zerosequence equident circuit
of threephase transformers depends on the transformer
connection and is discussed in reference 16 (pages 298
300) assumming zero magnetizing and coreloss cur
rents. The zerosequence equivalent circuit of three
phase transformerstaking into account transformer sat
uration (e.g., ma etizin and coreloss currents) is iden
tical to that of I?!
igure 1 or groundedwye, grounded
wye transformers. Note that primary and secondary
transformer impedances (gp)= Ilp+jXj:) and 9ik)=
R, +jXi:)) have the same values for positive, negative,
and zerosequence equivalent circuit since the three
phase transformer is assumed to consist ofthree identi
cal singlephase transformers. Zerosequence equivalent
circuits for other types of transformer connections are
presented in _{[}_{1}_{7}_{]}_{.}
*) Updated at each Newtonkaphson iteration U&
transformer induced voltage and Xi characteristic; r$
= iteration number. 

Figure 
1 
former as used in Earmonic power flow analysis.
Sigle hase equivalent circuit of a trans
In Figure 1 buses i and j represent primary
and
bus i + 1 represents the node, where the trans? ormer
and secondary transformer terminals, respectivel
cross branch meets the series branch between Xi:) and
Xi:).
The voltage at
this bus is equivalent to
the
transformer induced voltage.
Also, Rr), dk),X(k)
L P
and XP) are the transformer perunit primary and sec
ondary resistances, the primary and secondary leakage
inductances (for the kth harmonic), respectively. Note
that the magnetizing and coreloss currents are modeled
as constant current sources for a given induced voltage.
These currents are to be updated (Appendices I1 and
111) at each NewtonRaphson iteration since they are
functions of the induced voltage which will be changing
with each iteration.
The equivalent circuit of Figure 1 works fine,
however, in order to save computing time the core
loss currents (fundamental and harmonics) are assumed
to be zero during NewtonRaphson iterations and are
only included in the last ste of the solution proce
dure
for the computation of tYl e transformer ironcore
losses (A pendix 111). This is a valid assumption since
the _{c}_{o}_{r}_{e}_{}_{P}
oss currents (fundamental and harmonics) are
Inclusion of Nonlinear Transformers in the
Harmonic Power Flow Algorithm
There are two _{t}
es of buses defined in the
harmonic power flow al
ori%
[3,4]: linear buses e.g.,
PQ, PV and swing busy and nonlinearload buses (e.g.,
buses connected to fullwave rectifiers). _{I}_{n} order to in
clude transformers in the harmonic power flow alge
rithm a third type of bus (for modeling transformer
induced voltages) is defined and called a "transformer
bus"
(e.g.,
bus
2 + 1 in
Figure
1).
Therefore, trans
formers are treated as nonlinear loads connected to the
newly defined transformer buses injecting current har
monics into the power system. The functional form of
a transformer bus is identid to a nonlinearload bus
formulation of [3,4], to permit the inclusion of newly
defined transformer buses, are given in Appendix I.
Note that in order to include the magnetiz
ing and coreloss currents in the load flow analysis an
additional bus (bus i + 1 of Figure 1) is required. Ad
ditional buses for the re resentation of transformer in
duced voltages are not $ways required since power sys
tem transformers are mall connected (at one side)
to transmission lines or
cab9 e s. In such cases the se
ries impedance of the transformer side connected to the
transmission line could be added to the line impedance
avoiding the addition of an extra bus.
12
The convergence properties of the har
monic load flow in the presence of nonlinear transform
ers are similar to those of the harmonic load flow algo
rithm with linear transformers, however, the presence of
triplen current harmonics (in the case of oundedwye
transformers) and the addltion of nodes Er the repre
sentation of transformer induced voltages (if transmis
sion line and transformer impedances are not combined)
‘increase the required CPU time of one NewtonRaphson
iteration. The increase in computing time depends on
the number of transformer buses and the highest har
monic order considered.
An Illustrative Example
sixpulse rectifier
_{[}_{T}_{I} _{T}_{h}_{e} _{2}_{4}_{0}
(F=0.07 per unit reactance at 60 Hz
as ex
V, lkVA anisotropic
has the following per
and R is com uted by the load flow algorithm
plained in
transformer ppendix A
unit impedance values at
Rp)=0.020 pu;
Xi;’ =0.007 pu;
R(,1)=0.027 pu;
AY!’) =0.016 pu.
The equivalent circuit of the thr%ephasesystem of Fig
ure 2 (as used in harmonic power flow analysis) is shown
in Figure 3. Note that the grounded wyedelta trans
former connecting the sixpulse bridge rectifier to bus
#2 is assumed to be
linear as done in [3,4] and all
impedances are in perunit with a voltampere base of
1.0 kVA and a voltage base of 240 V.
It
is
desired to
keep the
volta es
at
the
nonlinearload bus (bus #2)
put
(bus #? about 1 perunit for a rated transformer
and at the transformer out
resistive loa . To establish this consider the following
cases:
Case 1: The
eneration bus voltage is regulated at 1.0
perunit and tie transformer is loaded with a rated re
sistive load. The rectifier injects about 19 and 9% of the
5th and 7th current harmonics, respectively, resulting in
a nonsinusoidal voltage waveshape
transformer losses are aL out
nals of the transformer
at the primary termi
first row of Table la). However,
45% below (first row of Ta
ble lb) the rated losses (e.g., transformer losses under
sinusoidal input and rated output conditions) which is
due to the fact that the fundamental transformer volt
age magnitude is only about 73%.
Case 2: In order to achieve rated fundamental trans
former voltage magnitude, the voltage at the genera
tion bus is increased to 127%. The injected current
harmonics by the rectifier and the harmonic volta es at
the transformer input are about the same as for 8ase 1
Iormer output volta e and current are rated. This re
second row of Table la). However, fundamental trans
sults in about a 1.4% increase above the rated trans
former losses (second row of Table lb), which is due to.
the harmonic voltages at the primary terminals of the
transformer.
This configuration has the disadvantage
of overvoltage operation of the rectifier _{a}_{t} _{a} voltage of
106%, and an overvoltage of
127% of the generation bus.
Case 3: Another method for achieving rated trans
former output voltage is by introducing
a capacitor at
and increasing the volta e at the eneration bus (thirh
the secondary terminals of the transformer (bus #5
row of Tables la and lby to llO#. This configuration
results in rated transformer output apparent power and
losses however, transformer real output current drops
to 50% of the rated current resulting in a low output
power factor.
Cases 4 and 5: These cases are examples of over
voltage operations of transformers and are identical to
Case 3 with the exception that the voltage at the gen
eration bus is increased to 115 and
120%, respectively.
As shown in rows 4 and 5 of Tables la and lb, respec
tively overvoltage operations of about 6 (Case 4) and
12% (Case 5) result in transformer losses of about 11
and 23% above the rated losses, respectively, and pos
sibly transformer failures.
Case 6: In order to improve the transformer secondary
power factor of Case 3 the capacitor bank is placed
at the primary terminals of the transformer (bus #3
As shown in the last rows of Tables la and lb, wit
X, = 61%, transformer secondary apparent power and
losses are rated and the power factor
at the transformer
secondary has improved.
Case 7: This case illustrates a transformer terminat
ing an open line and is identical to Case 6 with the ex
ception that transformer load is removed. Transformer
unloading results in considerable increase in volta es at
busses 2, 3, and 4 and a substantial increase ( 205) in
transformer ironcore losses.
BUS 1 
BUS 2 
BUS 3  

7 

LONG LINE 
SHORT LINE IN 
NONLINEAR 

TRANSFORMER 

x, =7.5% 

LINEAR 

=‘ ’ 
LOAD 
BUS
’ Q=O
Figure 2 The threephase system of the illustrative ex
ample.
BUS 1
*) Updated at each NewtonRaphson iteration using
transformer induced voltage and Ai characteristic; N
= iteration number. 

Figure 3 
Equivalent circuit of Figure 2 as used in the 
modified harmonic load flow analysis.
13
Voltage at NonlinearLoad Bus harmonic order k
Transformer Secondary Voltage harmonic order k
Case 
R 
X, 

[%I 
I%] 

1 
lOOLO 
100 
 
2 
127LO 100 
 

3 
llOLO 
200 
116 
4 
115LO 
200 
116 
5 
12OLO 
200 
116 
6***llOLO 100 
61 

7*** 
llOLO 
 
 
[WI 

1 
1648 

2 
30.64 

3 
28.67 

4 
32.68 

5 
36.99 

68’8 
30.13 

7**.* 
43.79 

Case 
Vz’ 
R X, 

1 
[%I lOOLO 100 
IRI  

2 
127LO 100 
 

3 
llOLO 200 
116 

4 
115LO 200 
116 

5 
120LO 
200 
116 
_{6}_{*}_{*}_{*} 
_{l}_{l}_{O}_{L}_{O} _{1}_{0}_{0} 
61 

_{7}_{*}_{*}_{*} 
_{l}_{l}_{O}_{L}_{O} 
_{} 
 
Case 

[WI 

1 
16.48 

2 
30.64 

3 
28.67 

4 
3268 

5 
36.99 

;;’ 6 
30.13 

7”* 
4379 
N* Time** 

[sec] 
k=l 
k=3 
k=5 
k=7 
k=l 

25 
59 
77.12L7.96 
0.12L 128. 8.68L26.4 
5.25L53.772.89L9.70 

15 
40 

14 
38 
101.0L8.76 
1.57L 47.3 
2.79L53.0 
0.65L89.5100.6L12.4 

29 
67 
106.7L8.54 
2.08L 48.7 
2.74L50.0 
0.62L92.2106.3L12.2 

19 
49 
112.4L8.33 
2.68L 50.0 
2.69L47.2 
0.59L94.9111.9L11.9 

12 
35 
104.4L14.9 
1.06L15.6 
2.09L124. 

26 
63 
118.1L9.59 
2.71 L15.3 
1.94L103. 0.14L144.119.8L10.3 

[w] 
[%]I 
k=l 
k=3 
k=5 
k=7 
I 
k=l 
42.19 
45.7 b.30L98.1 
0.43L111 
0.28L114 
0.05L94.8)72.89L9.70 

78.84 
_{1}_{.}_{4}_{2} 
7.75L96.9 
3.15L108 
0.98L106 0.20L58. 
100.1L8.38 

77.55 
0.25 
7.14L100. 
2.91 L121 0.89L144 
0.05L73. 
100.3L47.5 

8626 
11.0 
9.07L100. 
3.87L119 
1.09L141 0.05L 72. 
106.0L47.7 

95.36 
_{2}_{2}_{.}_{7} 
11.3L99.6 4.97L118 
1.28L138 0.67L 51. 
111.5L48.0 

78.05 
0.40 
7.85L107. 
3.24L140 
0.91L178 

44.39 
42.9 
15.5L100. 
7.101121 1.72L145 
0.12L 75. 
0 

N* Time** 
_{V}_{o}_{l}_{t}_{a}_{g}_{e} _{a}_{t} _{N}_{o}_{n}_{l}_{i}_{n}_{e}_{a}_{r}_{}_{L}_{o}_{a}_{d} _{B}_{u}_{s} harmonic order k 

[min] 
k=l 
k=3 
k=5 
k=7 
k=l 

13 
20 
77.12L7.96 
0.12L 128. 8.68L26.4 
5.25L53.7 72.89L9.70 

24 
37 
106.0L6.72 0.91L 130. 9.12f43.8 
4.25L49.9 100.I.L8.38 

13 
20 
101,OL8.76 
1.57L 47.4 
2.79L53.0 
0.65L89.5100.6L12.4 

16 
24 
106.7L8.54 
2.08L 48.7 
2.74L50.0 
0.62L92.2106.3L12.2 

13 
19 
112.4L8.33 
2.68L 50.0 2.69L47.2 
0.59L94.9111.9L11.9 

24 
37 
104.4L14.9 
1.06L15.6 
2.09L124. 

20 
30 
118.1L9.59 
2.71 L15.3 
1.94L103. 0.14L144. 119.8L10.3 
k=3 
k=5 
k=7 
0.14L 123. 8.11L17.7 
4.85L66.1 

3.92L62.5 

2.04L 38.8 
5.94L98.4 
l.llL143. 
2.72L 40.2 
5.79L95.2 
1.06L140. 
3.49h 41.5 
5.64L92.3 
1.02L137. 
3.11L143. 
0.58L828 

2.86L12.2 
3.07L113. 
0.55L113 
k=3 
k=5 
k=7 
0.14L 123 8.11L17.7 
4.85L66.1 

1.04L 125 851L35.5 
3.92L62.5 

5.38L 118 25.8L15.0 
6.74L132 

7.16L 119 25.1L11.9 
6.39L135 

9.19L 121 24.5L8.92 
6.17L138 

3.11L143. 
0.58L828 
_{0}
_{0}
_{0}
106.0L6.72 0.91L 130. 9.12L43.8 4.25L49.8 100.1L8.38 1.04L 125. 8.51L35.5
0.18~130.100.2L17.5 1.08L15.2
0.10L 13. 100.2L17.5 1.08L15.2
Transformer Secondary Voltage harmonic order k
k=3 
k=5 
k=7 
0.14L 123. 8.11L17.7 
4.85L661 

1.04L 125. 8.51L35.5 
3.92L62.5 

2.04L 38.8 
5.941198.4 
1.11L143. 
2.72L 40.2 
5.79L95.2 
1.06L140. 
3.49L 41.5 
5.64L92.3 
1.02L137. 
3.11L143. 
0.58L82.8 

2.86L12.2 
3.07L113. 0.55L113 
0.18~130.100.2L17.5 1.08L15.2
Transformer Magnetizing Current harmonic order k
Transformer Secondary Current harmonic order _{k}
Core Losses Total Losses At
[W] 
[%] 
k=l 
k=3 
k=5 
k=7 
k=l 
k=3 
k=5 
k=7 

42.19 45.7 2.30198.3 0.43L111 0.28L114 0.05L94.872.89L9.70 0.14L 123 8.11L17.7 4.85L66.1 

78.84 
1.42 
7.75L96.9 
3.15L108 0.98L106 0.20L58.6 100.1L8.38 
1.04L 125 8.51L35.5 
3.92L62.5 

77.55 0.25 7.14L100. 2.91L121 0.89L144 0.05L73.8 100.3L47.5 5.38L 118 25.8L15.0 
6.74L132 

86.26 
11.0 
3.07L100. 
3.87L119 
1.09L141 
0.05L 
72.7 
106.0L47.7 
7.16L 119 
25.1 L11.9 6.39L135 

95.36 
22.7 11.3L99.6 4.97L118 1.28L138 0.67L 51.8 111.5L48.0 9.19L 121 24.5f8.92 
6.17L138 

78.05 
0.40 7.85L107. 
3.24L140 0.91L178 0.10L 13.2 100.2,~17.5 1.08L15.2 
3.11 L143. 
0.58L82.8 

44.39 
42.9 
15.5L100. 
7.101121 1.72L145 
0.12L 75.5 
0 
0 
0 
0 
14
Case yi) 
R 
X, 
N* Time** 
Voltage at NonlinearLoad Bus harmonic order k 
Transformer Secondary Voltage harmonic order _{k} 

[%I 
[%I 
[sec] 
k=l 
k=3 
k=5 
k=7 
k=l 
k=3 k=5 
k=7 

1 
lOOLO 100 
 
21 
11 
77.40L8.07  8.721 25.4 5.28154.4 73.17L9.86  8.15L 16.5 4.88L66.7 

2 
12710 100 
 
17 
9 
106.8L7.01  9.07L 40.5 4.29L51.8 101.OL8.80  8.48L 31.6 3.96L64.1 

3 
11OLO 200 
116 
20 
11 
101.7L9.11  2.65L56.3 0.64L92.4 101.5L12.9  5.80L102. 
1.09L 139. 

4 
11510 
200 
116 
13 
8 
107.7L8.96  2.57L54.2 0.61L95.8 107.5L12.7  5.62L100. 
1.03L 136. 

5 
120LO 
200 
116 
13 
8 
113.6L8.84  2.49L52.2 0.58L99.2 113.3L12.6  5.44L98.2 0.99L 132. 105.1115.3  2.00/129. 0.17L 127.101.OL18.1  120.1L10.4  1.77L113. 0.12L 135.122.1L11.3  

6*** 
llOLO 100 
61 
10 
7 
3.07L 147. 0.571 78.8 

V*** 
llOLO 
 
 
14 
8 
2.97L 121. 0.54L 105 
As explained before, in order to save com
puting time transformer coreloss currents (fundamental
and harmonics) are assumed to be zero during Newton
Raphson iterations and are only included in the last
step of the solution procedure for the com utation of
the transformer ironcore losses (Appendix EI). To 'us
tify this assumption, the above seven cases are studied
(Table 2) with transformer coreloss currents included
in the entire Newton Raphson iterations (as ex
plained in the last paragraph of Appendix 11). Com
parison of Tables 1 and 2 indicates that elimination of
transformer coreloss currents during NewtonRaphson
iterations intro duces considerable saving in computing
in transformer ironcore losses (CO 1umn 2 of Fables lb
time (column 6 of Tables la and 2a and negli 'ble error
and 2b) and power system voltages (columns 7 to 14 of.
Tables la and 2a and columns 5 to 12 of Tables lb and
2b).
In order to compare the convergence behav
ior of the modified harmonic load flow algorithms (with
nonlinear transformers) with the harmonic load flow al
gorithms with linear transformers, the above seven cases
are studied without the nonlinear model of the trans
former and the addition of the transformer
bus as shown
in Table 3. Although, the number of required iterations
for the convergence of the two algorithms (column 5 of
Tables la and 3) are about the same, the required CPU
computing times (column 6 of Tables la and 3) are in
creased due to the fact that nonlinear transformers are
treated
as nonlinear loads for which additional Jacobian
entries must be computed.
and output (secondary) powers. However, this method
results in large errors since transformer losses are only
a fraction of the input and output real powers [17]. A
better approach for measuring transformer losses (for
transformers with transformation ratios close to one)
was employed, where [vd(t) vp(t)] and [id(t) +ip(t)] as
well as [vd(t)+vp(t)] and [is(t)  ip(t)]are measured si
multaneously [17] and transformer losses are computed
as 

where Pl(t) = [U$) 

Iv&) 
+'Up(t)l 
mary and secondar 
Transformer loss =
 vp(t)] [ia(t) + ip(t)] and p2(t) =
are transformer primary and seconA)
currents, respective1
[is@)  ip(t)l. vp(t), vs t),ip(t , and is(t)
ary vo tages, pri
For trans
formers with transgrmation ratios very zfferent from
one, the above method
will give accurate losses if two
identical transformers are connected backtoback. As
shown in Table 4, the measured and computed losses
indicate good agreement.
Table 4:
Measured and computed (usin
the proposed
modified harmonic power flow algorithm! losses for the
_{a}_{n}_{f} parallel RC loads.
sin
ephase transformer of [13] subjected to resistive
Measurement Procedure In order to show the accuracy of the modified
harmonic power flow algorithm, transformer (ironcore
and copper) losses are measured and computed
for the
sin ephase transformer of 131 subjected to a resistive
an%l
a
arallel RC loads. 4ransformer losses are com
puted PTable 4) using the proposed modified
harmonic
load flow algorithm with transformer primary terminals
modeled as the
minals modeled as
a parallel RC branch tied to
an additional bus (a
tation of transformer induced voltage.
Transformer losses could be measured as the
difference between the measured real input (primary)
*)Percentage of error in computed losses (using the pro
posed modified harmonic load flow) with respect to
measured losses.
Discussion
The modified harmonic load flow algorithm
resented above is an extension of the harmonic load
gow algorithm with linear transformers [3,4]. Nonlinear
transformers are treated as nonlinear loads, which are
connected to the newly defined transformer buses and
are injecting harmonics into the power system.
Some of the shortcomings of the present har
monic load flow approach taking into account nonlinear
transformers are:
1. Additional buses are introduced for the represen
Using a stepsize of about 5% (e.g., I Av(j)I =
0.05. V(j)lold), we have not experienced any numerical
instabilities by introducing the finite difference e ua
tions of Step
2 in the harmonic power flow algorism.
However, the required CPU time for a given iteration
increases considerably
(see Tables la and 3) since ad
ditional Jacobian entries are introduced by the trans
former equivalent circuit.
In the above formulation, the transformer
coreloss current is assumed to be zero. However, for
a given voltage V, at transformer bus m, the trans
former coreloss currents can be included in each step
of the NewtonFbphson iterative procedure, if the real
and imaginary components of the exciting transformer
current are computed using
(k)
Ee,m + fiAg,m = Iezei(r),m + j
(k)
Iezci(i),m
for k = 1,3,   ,L, where j = J7 and
is trans
former coreloss current as computed in Appendix 111.
nary magnetizing currents the corresponding Jacob&
matrix entries are computed using the same above
mentioned fourstep procedure.
APPENDIX
I11
Computation of Transformer Fundamental and Harmonic IronCore Losses
In order to save computing time, transformer
fundamental and harmonic coreloss currents are as
sumed to be zero throughout the iterative Newton
Raphson solution procedure. After the solution has con
verged, the proposed method of [14] is used to com
pute the ironcore losses and the associated coreloss
currents.
For given fundamental and harmonic induced
voltage amplitudes and phase angles at transformer bus
m the procedure is as follows:
Step 1: Repeat Step 1 of Appendix 11.
Step 2: Synthesize the instantaneous fluxlinkage time
functions and use the computed flux link 7density curves (e.g., Figure 4 of each mesh of the transformer gri6 systemO 141) t e versus flux (e.g., Figure etermine 
for 
16 of [13]) the fluxdensity time functions.
Step 4: The kth harmonic total ironcore losses
Pcore,total
(k)
are the sum of all local kth harmonic iron
core losses. The total transformer ironcore losses are
k=1,3, ...
The harmonic coreloss resistance and coreloss currents
The phase shift of ,,&f
E%)).
17
is identical to that of
References
C .K. Duffey and R. P. Stratford, ”Update of Har;
monic Standard IEEE519: IEEE Recommended
Practices and Requirements
for Harmonic Control in
Electric Power Systems, IBEE Trans. on Industry
Applications, Vol”. 25, No. 6, November/December
1989.
V. 
M. Montsinger,“Loading Transformer by Tem 

perature”, Trans. AIEE., Vol. 49, 1930, p. 776. 

D. 
Xia and G. Heydt, “Harmonic Power Flow Stud 

ies, Part1 and Part II”,IEEE Trans. on Power 

Apparatus and Systems, June 1982, Vol. PAS101, 

No.6 , pp. 12571265 and pp. 12661270. 

G. 
T. Heydt 
and 
W. M. 
Grady, 
“User Manual, 
HARMF’LO Version 3.1”, Electric Power Research 

Institute, Publication EL  4366  CCM, Palo Alto, 

CA, 1985. 
W. Song, G. T. He dt, and W. M. Grady “The In
tegration of HVD6Subsystems into the Harmonic
Power Flow AI orithm”, IEEE Trans. on Power
Apparatus an8 Systems, August 1984, Vol. PAS
103, No.$, pp. 19531961.
W. M. Grady and G. T. Heydt, “Prediction of Power
System Harmonics Due to Gaseous Discharge
Lighting”, w, March 1985, Vol. PAS104, pp. 554
561.
A. H. ElAbiad and D. C. Tarsi, “Load Flow Solu
tion
of
Untransposed
EHV
5th Power Industry Computer
Networks”, Proc. of
Applications Confe
rence, Pittsburgh, pp. 377384, 1967.
W. Xu, J. Marti, and H. W. Dommel, “A Mul
tiphase Harmonic Load Flow Solution Technique”,
IEEE Trans. on Power Systems, February
1991, Vol.
6, No. 2, pp. 174182.
W. Xu, J. Marti, and H. W. Dommel, “Harmonic
Analysis of Systems with Static Compensators” ,&d,
February 1991, Vol. 6, No. 2, pp. 183190.
[lo] H. W. Dommel, A. Yan, and W. Shi, “Harmonics
from 
Transformer 
Saturation”, IEEE Trans. on 
Power 
Delivery, April 1986, Vol. PWRD1, pp. 209 
215.
1111 A. Semlyen, E. Acha, and J. Arrillaga,“Harmonic Norton Equivalent for the Magnetising Branch of
a
Transformer”, IEE Proc., March 1987, Vol. 134, Pt.
C, No.2, pp. 162169.
[12] A. Semlyen, E. Acha, and J. Arrillaga, “Newton
Type Algorithms for the Harmonic Phasor Analy
sis of Nonlinear Power Circuits in Periodical Steady
State with Special Reference to Magnetic Nonlinear
ities”, IEEE Trans. on Power Delivery, July 1988,
Vol. 3, No.3, pp. 10901098.
[13] E. F. Fuchs, M. A. S. Masoum, and D.
J. Roesler,
“Nonlinear
Model
of
Anisotropic Power Trans
Part I: Ai
formers for Nonsinusoidal Operations.
Characteristic”, w, January 1991 , Vol. 6, No.1 ,
pp. 174185.
[14]
M. A. S. Masoum, E. F. Fuchs, and D. J. Roesler,
“Nonlinear
Model
of
Anisotropic Power Trans
formers for Nonsinusoidal ODerations. Part 11: Mae
netizin
1991, $01.
and CoreLoss CGrrents”, U, Octobzr
6, No. 4, pp. 15091516.
M. A. S. Masoum, E. F. Fuchs, and D. J. Roesler,
“Impact of Nonlinear Loads on Anisotropic Trans
formers”, U, October 1991, Vol. 6, No. 4, pp.
17811 788.
W.
D. Stevenson, “Elements of Power System Anal
ysis”, McGrawHill Book Company, Fourth Edition,
1982.
hl. A. S. Masoum, “Generation and Propagation of
Harmonics in Power Systems Containing Nonlinear
Transformers and Loads”, Ph.D. Thesis, University
of Colorado at Boulder, May 1991.
W.
M. Grady, Private communication, January 22,
1991.
Mohammad A. S. Masoum was
born in Isfahan, Iran, on August
23, 1959. He received his Bache
lor’s, Master’s and Ph.D. de ees
in Electrical and Computer Kgi
neering in 1983, 1985 and 1991,
respectively, from the Univer
sity of Colorado. His main in
terests include numerical analy
sis as applied to harmonic power
flow studies, finitedifference and
finiteelement formulati on _{a}_{s} well
as optimization of electrical devices and solar power
plants. He has been on the staff of the Electrical En
gineering Department at University of Colorado as a
Teaching Assistant, Research Assistant, and instructor
since 1983. Presently Dr. Masoum is a postdoctoral
fellow at the University of Colorado at Boulder.
Ewald F. Fuchs received his
Dip1.Ing. degree in Electrical
En ineering from the University
of Btuttgart, Germany, and the
Ph.D. degree in Electrical En
neering from the University of
f3 olorado at Boulder in 1967 and
1970, respectively. From 1972 to
1978, he was employed as an en
gineer in the Lar e Steam Tur
bine Division of &e Kraftwerk
Union AG in Miilheim, Germany.
Presently Dr. Fuchs is a Professor of Electrical Engi
neering at the University of Colorado, Boulder, Col 

orado 80309. 
His main interests are design and opti 
mization techniques of drive systems, harmonic power
flow studies, and finitedifference and finiteelement for
mulations. Dr. Fuchs is a Fellow of IEEE, a member of
VDE, a member of Eta Kappa Nu, and of Sigma Xi. At
the annual meeting of the VDE in 1972 he was the re
cipient of the annual best paper prize. At the Summer
Power Meetin 1989 in Long Beach, CA, he received
the PES/IEEl$ Prize Paper Award 1989 and the IEEE
1989 Power System Relaying Committee Award.
19
Discussion
ALTERNATOR
D. J. Roesler (U.S. Department of Energy, Washington, DC):
PRIMARY
SECONDARY
The implementation of a nonlinear transformer model in har monic load flow programs (e.g., HARMFLO) is an important contribution, since it makes an assessment of the energy savings _{,}_{\} of emerging nonlinear power electronic loads easier. For exam ple, it is important to study the potential effects of singlephase adjustablespeed drive heat pumps on power system distortion and losses. Such heat pumps can have a high seasonal energy efficiency ratio (SEER) of about _{1}_{5} but a high total harmonic distortion (THD) of the current of more than 75%, or a low SEER of about 10 with a low current THD of less than 20%. To assess the benefits of such newly emerging highefficiency heat pumps one has to weigh the energy savings for the consumer in comparison to the energy loss within the power system caused by the injected harmonic currents originating in nonlinear con sumers such as variablespeed heat pumps. In reading this paper one notes that the authors have only presented a zerosequence equivalent circuit for wye/wye trans formers with both neutrals grounded. In view of the abovemen tioned need to evaluate the potential effects of highefficiency heat pumps on power system distortion and losses, one has to know as well the zerosequence equivalent circuits for other
threephase transformer connections (e.g., delta/wye). Are these zerosequence equivalent circuits identical to those listed in _{[} _{1}_{6}_{,} page 2991 or are there any differences, if magnetizing currents are not neglected? What difficulties and adjustments are antici pated in using the transformer model of this paper for unbal anced harmonic power flow analysis?
Figure D1 Groundedwye i"' ungroundedwye threephase
transformer connected to a balanced groundeduye al
ternator (a). and corresponding measured current and
voltage waveforms (b).
ALTERNATOR
SECONDARY
VSpH
M.A.S. Masoum and E.F. Fuchs: The authors
thank Mr. Roesler for his interest in this paper and for
providing an opportunity to address important ques
tions. There are two types of zerosequence currents
in power systems: type I consists of zerosequence sig
nals (e.g., 3rd, _{g}_{t}_{h}_{,}
Gth,...)
due
to system imbalance
as discussed in pages 298300 of reference 16; type I1
consists of zerosequence currents injected by nonlin
ear loads (e.g., saturated transformers). Therefore, sat
urated transformers inject zerosequence harmonics of
type I1 in balanced as well as unbalanced power sys
tems.
The injection of triplen (3rd7gth:
15th. ...
)
harmonics in
power systems by saturated transformers is greatly in
fluenced by transformer connections. Transformer sat
uration usually results in generation of zerosequence
harmonic currents as shown by the measured waveforms
of Figure D1. However, if a zerosequence path is not
provided in either the primary side or the secondary
_{s}_{i}_{d}_{e} _{(}_{e}_{.}_{g}_{.}_{,} _{i}_{n} _{c}_{a}_{s}_{e} _{o}_{f} ungroundedyve/ungroundedwye
transformers , predominantly 3rd, gth, isth,
harmonic vo1 tages are generated as indicated in Figure
...
order
D2.
Figure D2
(b)
Ungroundedwyejungroundedwpe three
phase transformer connected to a balanced ungrounded
wye alternator (a). and corresponding measured current
and voltage wayeforms (b).
In Figure D1 a zerosequence path is provided in the pri
mary, therefore. transformer saturation results in dom
inant triplen (including positive and negative sequence)
harmonics in the primary phase current (IFH 1,
ligible zero. positive and negatile sequence armomcs
in the line and phase voltages _{\}_{7}_{f}_{}_{L}_{.} In
Figure D2 no zerusequence current path is provided,
20
therefore, transformer saturation results in triplen har
monics in the phase voltages _{(}_{V}_{;}_{,}_{,} _{V}_{S}_{H}_{)}_{a}_{n}_{d} no zero
sequence harmonics in the line current (Iz) contain
ing negative and positive sequence current components.
These nonlinear phenomena associated with saturated
transformers must be properly included in transformer
zerosequence equivalent circuits as illustrated in Fig
ures D3 to D6 neglecting ironcore loss currents.
*) Updated at each NewtonRaphson iteration using
traasformer induced voltage and Ai characteristic; N
= iteration number.
**) Contains 3rd, gth, 15th,
...
order harmonic voltages
ca,used by transformer saturation.
Figure D3 Zerosequence equivalent circuit of ungroun
dedwye/ungroundedwye connected threephase trans
formers neglecting ironcore loss currents.
BUS i
M
Fi re D4 Zerosequence equivalent circuit of delta/
de!% connected threephase transformers neglecting
ironcore loss currents.
I
Figure D5 Zerosequence equivalent circuit of ground
edwye/delta connected threephase transformers ne
glecting ironcore loss currents.
Figure D6 Zerosequence equivalent circuit of ungroun
dedwye/delta connected threephase transformers ne
glecting ironcore loss currents.
If ironcore loss currents are not neglected Figure D3
becomes Figure D7 and Figures D4 to D6 can be aug
mented similarly.

Figure D7 Zerosequence equivalent circuit of ungroun
dedwyelungroundeduye connected threephase trans
formers including ironcore loss currents.
In above figures (Figures D3 to D6) threephase trans
formers are assumed to consist of three identical single
phase transformers, transformer core loss currents are
neglected, and balanced threephase systems are consid
ered. Therefore, the generated zerosequence harmonics
are caused by transformer saturation and are not due
to system imbalance.
These equivalent circuits could be used for three and
fivelegged threephase transformers under unbalanced
conditions, if
1. the fact that transformer legs are asymmetric is in
cluded in the computation of the transformer mag
netizing current (e.g., different Ai characteristics are
used for different phases
2. the existence of mutual A ux linkages between trans
,
former phases are taken into account in the compu
tation of the transformer magnetizing current.
Manuscript received December 21, 1992.