Electric Power Systems Research 144 (2017) 32–40

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Influence of tertiary stabilizing windings on zero-sequence

performance of three-phase three-legged YNynd transformers. Part I:
Equivalent circuit models
Angel Ramos a,∗ , Juan Carlos Burgos b
a
Gas Natural Fenosa, Madrid, Spain
b
Universidad Carlos III de Madrid, Leganés, Madrid, Spain

a r t i c l e i n f o a b s t r a c t

Article history: The presence of a stabilizing winding (or tertiary stabilizing winding when is used to for auxiliary
Received 4 October 2015 applications) in three-phase three-legged YNynd transformers remarkably affects the zero-sequence
Received in revised form 27 July 2016 performance of both the transformer and the network. This paper presents a detailed analysis of the
Accepted 29 October 2016
influence of the stabilizing winding on the zero-sequence behavior of three-phase three-legged YNynd
Available online 23 November 2016
transformers. Based on a complete set of onsite zero-sequence measurements taken in three power
transformers, transformer zero-sequence performance is analyzed in relation to internal design features
Keywords:
such as stabilizing winding position relative to high-voltage and low-voltage windings or the presence
Three-winding transformers
Tertiary stabilizing windings
of magnetic shunts in the tank. Based on these measurements, this paper assesses the ability of various
Transformer zero-sequence performance equivalent circuit models to reproduce zero-sequence performance accurately. A companion paper that
Onsite zero-sequence impedance complements this study evaluates the influence of stabilizing windings on tank overheating hazard and
measurements short-circuit duty in the event of asymmetrical faults.
Transformer equivalent circuits © 2016 Elsevier B.V. All rights reserved.

1. Introduction of three-phase three-legged YNynd power transformer behavior

depending on whether the stabilizing winding is closed or open.
Delta stabilizing windings have been used since early electric For all these reasons, in this paper (Part I) a complete set of zero-
power system development to avoid some of the network operation sequence measurements was performed on three YNynd power
drawbacks of using wye connections in power transformers and transformers of different internal design under different operating
autotransformers [1–3]. Utilization of stabilizing windings even conditions (with and without stabilizing winding) in order to eval-
became mandatory under some local regulations, and other exter- uate the accuracy of several equivalent circuit models in predicting
nal uses were also found for these windings [4,5]. impedances and currents. In the second part of this study (Part II),
Although the need for “critical” analysis of the decision to presented in a companion paper, these results will be used to ana-
include a stabilizing winding in Yy-connected transformers [6–8], lyze tank overheating hazard and short-circuit duty in three-phase
particularly in relation to three-phase three-legged transformers, three-legged transformers. This information will be useful to design
was pointed out many years ago, many utilities still maintain this and planning engineers when conducting thorough assessment of
practice for reasons of “tradition” without analyzing the decision. the need for stabilizing windings.
Making an appropriate decision about the need for stabilizing
windings requires deep understanding of the transformer’s zero-
sequence performance. Although many authors have studied zero- 2. Onsite zero-sequence measurements
sequence behavior and transformer modeling in detail [9–15], there
is little information in technical literature providing an overview To analyze stabilizing windings’ influence on zero-sequence
performance and its dependence on the windings’ relative position
and on the presence of magnetic shunts, a complete set of tests
∗ Corresponding author at: Architecture and Network Design Manager, Electrical
was performed on three three-phase three-legged YNynd power
Distribution Division, Gas Natural Fenosa, Avda. San Luis 77, Madrid 28033, Spain.
transformers (see Table 1 and Fig. 1 for nameplate data and inter-
E-mail addresses: nal characteristics). These internal designs cover the vast majority
[email protected], [email protected] (A. Ramos) of core-form construction types of power transformers.

http://dx.doi.org/10.1016/j.epsr.2016.10.065
0378-7796/© 2016 Elsevier B.V. All rights reserved.
 A. Ramos, J.C. Burgos / Electric Power Systems Research 144 (2017) 32–40 33

Fig. 1. Schematic description of the internal construction design of the transformers measured.

Table 1 Table 4
Characteristics of transformers subject to zero-sequence measurements. Results and calculations of zero-sequence measurements taken in transformer #2.

Data Transformer Test code V (%) I1 (%) I2 (%) I3 (%) Z0 (%)

#1 #2 #3 HOO (In Table 6) (In Table 6) – – 135.29

HOC 0.136 0.699 – 0.695 19.41
MVA 25/25/8.33 75/75/25 150/150/50 HSO 0.171 1.258 1.223 – 13.63
kV 45/16.05/10 220/71/10 230/71/20 HSC 0.110 0.821 0.931 0.164 13.45
Vector group YNyn0 + d11 YNyn0 + d11 YNyn0 + d11 LOO (In Table 6) – (In Table 6) – 142.30
Short-circuit impedance 10.8% 14.0% 14.1% LOC 0.055 – 1.171 1.160 4.68
Inner winding Stabilizing Stabilizing Low-voltage LSO 0.115 0.766 0.839 – 13.76
Intermediate winding Low-voltage Low-voltage High-voltage LSC 0.039 0.322 1.218 0.893 3.24
Outer winding High-voltage High-voltage Stabilizing
Magnetic shunts NO YES NO
Table 5
Results and calculations of zero-sequence measurements taken in transformer #3.
Table 2
Set of zero-sequence measurements to be performed. Test code V (%) I1 (%) I2 (%) I3 (%) Z0 (%)

Test number Test code HV winding LV winding Stabilizing winding HOO (In Table 6) (In Table 6) – – 87.98
HOC 0.045 0.265 – 0.257 16.85
1 HOO Energized Open circuit Delta open HSO 0.109 0.896 0.782 – 12.17
2 HOC Energized Open circuit Delta close HSC 0.062 0.897 0.487 0.396 6.90
3 HSO Energized Short circuit Delta open LOO (In Table 6) – (In Table 6) – 104.27
4 HSC Energized Short circuit Delta close LOC 0.123 – 0.359 0.324 34.20
5 LOO Open circuit Energized Delta open LSO 0.053 0.366 0.375 – 14.06
6 LOC Open circuit Energized Delta close LSC 0.065 0.502 0.470 0.061 13.88
7 LSO Short circuit Energized Delta open
8 LSC Short circuit Energized Delta close
(I3 ). Voltage and current values are shown in percentage of base
quantities (rated voltage and rated apparent power). Calculation of
Table 3
Results and calculations of zero-sequence measurements taken in transformer #1.
zero-sequence impedance is indicated in Eq. (1),
V
Test code V (%) I1 (%) I2 (%) I3 (%) Z0 (%) Z0(ABC) = (1)
I/3
HOO (In Table 6) (In Table 6) – – 80.14
HOC 0.249 1.581 – 1.243 15.76 where (ABC) is the test code (e.g., HOO in test 1 in Table 2) and V
HSO 0.150 1.590 1.386 – 9.43 and I are the measurements taken in the energized winding.
HSC 0.145 1.555 1.659 0.387 9.35
Zero-sequence impedance modules are presented in percent-
LOO (In Table 6) – (In Table 6) – 86.09
LOC 0.033 – 0.599 0.545 5.58 ages in Tables 3–5, calculated as indicated in Eq. (2),
LSO 0.054 0.503 0.507 – 10.76 Z0(ABC) V (%)
LSC 0.019 0.225 0.587 0.360 3.29 Z0(ABC) (%) = = (2)
Zbase I (%)
Relation between Eqs. (1) and (2) could be easily derived from
The internal design of transformer #1 (with inner stabilizing well-known relations between rated and base quantities.
winding and no magnetic shunts) is typically employed in low- and As illustrated in Ref. [16], non-linear behavior of HOO and
medium-power YNynd transformers. When rated power is above LOO measurements does not produce a major problem of model
30–40 MVA, transformer designers include magnetic shields, as in accuracy when predicting short-circuit currents. Consequently, to
transformer #2. For high-power transformers (above 100 MVA), facilitate understanding of Tables 3–5 these tests consider average
three-winding YNynd transformers may include outer stabilizing impedance values. All values measured in these tests are shown in
windings, as in transformer #3 (with or without magnetic shields). Table 6. Some additional considerations about “no-load” tests (i.e.
As demonstrated in Ref. [16], accurate representation of the HOO and LOO measurements) can be found after Table 6.
zero-sequence performance of three-phase three-legged YNynd As indicated previously, an accurate representation of the
transformers can be achieved by taking onsite low-voltage mea- zero-sequence performance of three-phase three-legged YNynd
surements, requiring 8 tests as indicated in Table 2. transformers can be achieved by means of low-voltage onsite
The results and calculations deriving from the tests performed measurements. With the exception of no-load tests, the rest of zero-
on the three transformers in Table 1 are shown in Tables 3–5, which sequence impedances (i.e. HOC, HSO, HSC, LOC, LSO and LSC) show
present measurements of voltage in the energized winding and of a good consistency between low-voltage onsite measurements and
currents in the HV side (I1 ), LV side (I2 ) and stabilizing winding high-voltage factory tests [16,17]. This is an expected result as these
34 A. Ramos, J.C. Burgos / Electric Power Systems Research 144 (2017) 32–40

Fig. 2. Nonlinear performance of ferromagnetic materials [19].

Table 6 affect zero-sequence magnetizing impedance, whose importance

Results and calculations of zero-sequence measurements taken in no-load tests.
when analyzing unbalanced loads and faults is expected to be much
Transformer Test code V (%) I (%) Z0 (%) lower than that of the other zero-sequence impedances. In Ref.
#1 HOO 0.397 0.526 75.45 [17], a complete sensitivity analysis is carried out in order to eval-
0.801 0.980 81.74 uate actual implications of the uncertainty of the zero-sequence
0.950 1.141 83.28 magnetizing impedance, concluding that on-site low-voltage mea-
LOO 0.149 0.186 80.09 surements can be used to obtain a faithful representation of
0.649 0.737 88.11
transformer zero-sequence behavior for steady-state short-circuit
0.805 0.893 90.17
calculations.
#2 HOO 0.100 0.075 133.62
In some other calculations (i.e. tank losses due to unbalanced
0.161 0.117 136.96
LOO 0.154 0.110 139.62
currents or inrush current calculations) a more detailed model tak-
0.255 0.180 141.70 ing into account non-linear behavior should be used [21], as linear
0.314 0.220 142.64 equivalent circuits (as those proposed in Section 3) could not pro-
0.522 0.360 145.23 vide enough accuracy.
#3 HOO 0.041 0.048 85.89
0.075 0.085 88.04
3. Calculation of zero-sequence equivalent circuit model
0.102 0.115 88.80
0.122 0.137 89.18 parameters
LOO 0.108 0.105 103.11
0.167 0.160 104.18 Once a complete set of measurements is available, the next
0.219 0.209 104.53
step is to ascertain whether these measurements are suitable
0.274 0.261 105.27
for establishing zero-sequence equivalent circuits that repre-
sent transformer performance when delta stabilizing windings
impedances are largely dominated by leakage fluxes with linear are closed or open. Two approaches were considered: classic T-
behavior [18]. type equivalent circuits whose simplicity usually makes this the
In the case of “no-load” tests (i.e. HOO and LOO) of three-legged option selected by engineers, and the more complex six-impedance
transformers, zero-sequence flux closes its path from the magnetic equivalent circuit described in Ref. [10] as necessary for accurate
core to the tank through oil gaps and structural parts. This flux determination of internal currents in three-winding YNynd trans-
circulation path is significantly linearized due to the presence of formers.
non-ferromagnetic elements, but the non-linear effect of the tank
steel magnetic permeability is still present as can be deduced from 3.1. Three-impedance (T-type) equivalent circuit models
the dependence of impedance with voltage as shown in Table 6.
Under no-load conditions (in the absence of delta winding), the Utilization of T-type equivalent circuits in transformers is
transformer zero-sequence response is markedly influenced by the mandatory in commercial power system analysis tools used by pro-
magnetic permeability of the tank steel. Actually, the measure- tection, operation and planning engineers. For this reason, several
ments in Table 6 are fragments of a continuous dependence of types of T-type equivalent circuits for YNynd transformers will be
no-load zero-sequence impedances versus applied voltage, which analyzed in order to ascertain the best alternative.
is qualitatively similar to that in Fig. 2 [19]. The first alternative to be considered for the zero-sequence
As this slightly non-linear behavior of the transformer sup- equivalent circuit of the transformer (model A) is based on the
pose a source of error in HOO and LOO measurements, it must be model proposed by IEC standard 60076-8 [22], and considers the
checked if this error is a problem or not in order to obtain a suf- equivalent circuit shown in Fig. 3. This circuit has a branch in par-
ficiently accurate representation of zero-sequence performance of allel with the magnetizing branch, so when the delta winding is
the transformers in the calculation of short circuit currents (such as closed, current flows in both impedances. If the delta winding is
those in the companion paper [20]). These measurements mainly open, current flows through the magnetizing impedance only.
 A. Ramos, J.C. Burgos / Electric Power Systems Research 144 (2017) 32–40 35

Table 7
Zero-sequence impedance (T-type) calculation results for transformer #1.

Zero-sequence equiv. circuit (model A) Zero-sequence equiv. circuit (model B1) Zero-sequence equiv. circuit (model B2)

Impedance Z0 (%) Impedance Z0 (%) Impedance Z0 (%)

Z01 10.08 Z 01 9.79 Z”01 1.08

Z02 −0.20 Z’02 −0.42 Z”02 9.53
Z03 6.07 Z’03 5.99 – –
Z0M 78.38 – – Z”0M 77.84
QME = 8.82% QME = 0.16% QME = 1.51%

Fig. 3. T-type zero-sequence equivalent circuit model A (switch S is open when

delta is open).

Fig. 5. Six-impedance zero-sequence separate equivalent circuit (model C).

that the completed equivalent circuit (six branches) is required to

calculate the current circulating in the delta stabilizing winding.
To verify that statement and ascertain the veracity of the cur-
rents predicted by the T-type and six-branch equivalent circuit
models, a third alternative is proposed, as indicated in Fig. 5. In
Fig. 4. T-type zero-sequence separate equivalent circuits (model B).
this equivalent circuit, the stabilizing winding is represented by a
branch with a switch S that is closed when the delta winding is
closed.
The second alternative (model B) is based on IEEE standard In the same way as stated above for T-type equivalent circuits,
C57.12.90 [23] and considers two different T-type equivalent cir- the impedances of the zero-sequence equivalent circuit shown in
cuits for the YNynd transformer depending on whether the delta Fig. 5 were calculated from zero-sequence impedance measure-
stabilizing winding is closed or open (Fig. 4). ments. Optimization based on the Levenberg–Marquardt algorithm
The impedances of the zero-sequence equivalent circuits in [24] was applied to Eqs. (A.17)–(A.24) shown in the Appendix A.
Figs. 3 and 4 were calculated from the zero-sequence impedance Table 10 shows the results of the zero-sequence impedance calcu-
measurements presented in Tables 3–5. To obtain the circuit lations and the QME.
parameters in model B1 (Fig. 4—B1), measurements HOC, HSC, LOC
and LSC were used, while measurements HOO, HSO, LOO and LSO 4. Analysis of results
were used for the circuit in model B2 (Fig. 4—B2).
As the number of measurements is higher than that This section compares the ability of the different zero-sequence
of the impedances, an optimization process based on equivalent circuits to reproduce transformer behavior accurately.
the Levenberg–Marquardt algorithm [24] was performed. For this purpose, and for each test configuration, the predicted
Levenberg–Marquardt algorithm mixes the Gauss–Newton impedances and currents obtained from the various equivalent cir-
method and the gradient descendent method to improve the con- cuit models will be compared with real measurements. It should
vergence of the problem. In the optimization process, the objective be noted in Tables 7–10 that some impedances could be negative
was to minimize the quadratic mean error (QME) between the as a result of the mathematical transformations used to obtain the
measured impedances (shown in Tables 3–5) and those obtained models, without representing capacitive effects.
from the equivalent circuit. The equations used in the optimization
process for each model are detailed in the Appendix A. The results
of calculation of the zero-sequence impedances of the equivalent 4.1. Models’ accuracy in predicting impedance
circuits are shown in Tables 7–9. The QME is also indicated in the
tables. Resistive components of the measurements were ignored, In Section 3, the overall appropriateness of the different
as their influence on the results is not relevant [16]. equivalent circuits was assessed by means of the QME obtained
when performing the optimization process to obtain the models’
impedances. Table 11 summarizes the QME values. For model B,
3.2. Six-impedance equivalent circuit model the average value of models B1 and B2 is shown.
The results summarized in Table 11 show that model B offers
As indicated previously, Garin states [10] that an abbreviated greater accuracy (QME below 1%). Model C also presents quite low
equivalent circuit (T-type) is sufficient for system calculations, but QME values (in the 1–3% range). Conversely, model A presents
36 A. Ramos, J.C. Burgos / Electric Power Systems Research 144 (2017) 32–40

Table 8
Zero-sequence impedance (T-type) calculation results for transformer #2.

Zero-sequence equiv. circuit (model A) Zero-sequence equiv. circuit (model B1) Zero-sequence equiv. circuit (model B2)

Impedance Z0 (%) Impedance Z0 (%) Impedance Z0 (%)

Z01 14.39 Z’01 14.13 Z”01 4.96

Z02 −0.45 Z’02 −0.61 Z”02 9.12
Z03 5.26 Z’03 5.28 – –
Z0M 131.87 – – Z”0M 131.70
QME = 5.32% QME = 0.15% QME = 1.03%

Table 9
Zero-sequence impedance (T-type) calculation results for transformer #3.

Zero-sequence equiv. circuit (model A) Zero-sequence equiv. circuit (model B1) Zero-sequence equiv. circuit (model B2)

Impedance Z0 (%) Impedance Z0 (%) Impedance Z0 (%)

Z01 −1.54 Z’01 −1.59 Z”01 −0.46

Z02 15.73 Z’02 15.65 Z”02 14.61
Z03 23.12 Z’03 18.48 – –
Z0M 89.34 – – Z”0M 88.99
QME = 1.15% QME = 0.22% QME = 0.64%

Table 10 different to Z 01 and Z 02 is so different to Z 02 . As model A is not

Zero-sequence impedance (6-impedances) calculation results for transformers #1,
capable to represent accurately both situations, QME in model A
#2 and #3.
is significantly higher than in model B. On the other hand, in the
Zero-sequence equiv. circuit (6-impedances) case of transformer #3, due to its internal design, both the stabiliz-
Impedance Transformer #1 Transformer #2 Transformer #3 ing winding and the tank when acts as a virtual delta winding are
Z0 (%) Z0 (%) Z0 (%) positioned as the outermost winding. In this case, leakage fluxes
Z01 73.35 177.07 834.42 trajectories when stabilizing winding is closed or open are more
Z02 −204.26 −82.58 579.08 similar and the differences in QME of model A and B are not too
Z03 229.94 67.49 102.77 high.
Z012 8.35 11.90 12.64 To provide more detailed information about impedance pre-
Z013 −39.07 −67.78 15.09
Z023 5.19 4.31 −107.95
diction errors in each test configuration, Table 12 shows the
QME 2.54% 1.17% 1.26% impedance values obtained in each measurement and the relative
error of the predicted impedances of the different equivalent cir-
cuit models. As revealed in Ref. [16], the errors are normally higher
Table 11 in no-load configurations due to transformers’ non-linear behavior,
Comparison of the different models’ quadratic mean error.
without significantly affecting zero-sequence performance analy-
Model Quadratic mean error (QME) sis. The conclusions drawn about the different models from Table 11
Transformer #1 Transformer #2 Transformer #3 can be extrapolated to the results in Table 12.

A 8.82% 5.32% 1.15%

B 0.83% 0.59% 0.43%
4.2. Accuracy of the different models’ current prediction
C 2.54% 1.17% 1.26%
As explained previously, a T-type equivalent circuit model is
expected to be sufficiently accurate for system calculations but not
greater dispersion in the different transformers, producing values for internal current prediction. However, a more complex model
above 5% in transformer #1 and #2. (with at least six branches) is expected to offer more accurate cur-
Despite being topologically quite similar, the accuracy of models rent prediction.
A and B are very different. As it was indicated, the main differ- The aim of this section is to verify the current prediction accu-
ence between models A and B is that while in model A the same racy of the different equivalent circuits for both external (primary
impedances Z01 and Z02 represent the behavior of a YNynd trans- and secondary windings) and internal (stabilizing winding) cur-
former in all conditions, model B uses different impedances in the rents.
series branches (Z 01 different to Z 01 and Z 02 different to Z 02 ) Table 13 shows each model’s average error in primary and
to represent the behavior of a YNynd transformer whether stabi- secondary current prediction for the different test configurations,
lizing winding is open or closed. The underlying physical reason and Table 14 shows the mean error in stabilizing winding cur-
behind the difference in accuracy of models A and B is that in rent prediction. Tables 15–17 show detailed winding current values
a three-legged core-type wye–wye transformer, when calculating obtained in each measurement and the relative error (in relation to
open-circuit zero-sequence impedances, the tank acts as if it were measurement) of the predicted currents for the different equivalent
a high-impedance delta-connected outermost winding [10]. On the circuit models.
contrary, in a three-winding YNynd transformer this effect is not As expected, in model C current prediction accuracy remains
present due to the delta stabilizing winding. good (error less than 3–5%) for both internal and external currents.
In the case of transformers #1 and #2, model B1 represents the Conversely, equivalent circuit A made poor predictions not only for
zero-sequence equivalent circuit of a wye-wye transformer with internal but also for external currents, especially in one of the trans-
an innermost delta (real) winding and model B2 represents the formers. Meanwhile, model B offers good predictions for external
zero-sequence equivalent circuit of a wye-wye transformer with currents and good predictions for delta winding currents in about
an outermost delta (virtual) winding. In this case, leakage fluxes 75% of cases. The reason why model B has less accuracy in pre-
trajectories are quite different and this is the reason why Z 01 is so dicting delta winding currents than model C is related with tank
 A. Ramos, J.C. Burgos / Electric Power Systems Research 144 (2017) 32–40 37

Table 12
Comparison of the different models’ impedance prediction.

Test code Model Transformer #1 Transformer #2 Transformer #3

Z0 (%) Error (%) Z0 (%) Error (%) Z0 (%) Error (%)

Measured 80.14 – 135.29 – 87.98 –

A 88.46 10.38 146.26 8.11 87.80 0.20
HOO
B 78.92 1.52 136.66 1.01 88.53 0.63
C 77.25 3.60 136.85 1.15 89.01 1.17

Measured 15.76 – 19.41 – 16.85 –

A 15.71 0.29 19.45 0.20 16.83 0.14
HOC
B 15.78 0.13 19.41 0.00 16.89 0.24
C 16.36 3.80 19.34 0.34 16.91 0.35

Measured 9.43 – 13.63 – 12.17 –

A 9.88 4.77 13.94 2.26 11.84 2.75
HSO
B 9.57 1.49 13.49 1.03 12.09 0.66
C 9.62 2.02 13.52 0.84 12.38 1.72

Measured 9.35 – 13.45 – 6.90 –

A 9.87 5.59 13.90 3.32 6.93 0.48
HSC
B 9.34 0.12 13.44 0.07 6.88 0.23
C 9.28 0.76 13.35 0.77 6.82 1.13

Measured 86.09 – 142.30 – 104.27 –

A 78.18 9.19 131.42 7.65 105.07 0.77
LOO
B 87.37 1.49 140.82 1.04 103.60 0.64
C 88.26 2.52 140.65 1.16 102.28 1.90

Measured 5.58 – 4.68 – 34.20 –

A 5.43 2.62 4.61 1.53 34.10 0.30
LOC
B 5.57 0.18 4.67 0.21 34.13 0.20
C 5.73 2.73 4.77 1.90 34.63 1.26

Measured 10.76 – 13.76 – 14.06 –

A 8.73 18.85 12.52 8.98 14.16 0.73
LSO
B 10.60 1.53 13.90 1.02 14.15 0.62
C 10.99 2.14 13.89 0.95 14.23 1.18

Measured 3.29 – 3.24 – 13.88 –

A 3.41 3.77 3.29 1.63 14.05 1.22
LSC
B 3.30 0.19 3.23 0.19 13.91 0.22
C 3.25 1.17 3.29 1.55 13.97 0.66

Table 13 5. Conclusions
Comparison of the different models’ mean errors in primary and secondary current
prediction.
Differences in the zero-sequence performance of three-phase
Model Quadratic mean error (QME) three-legged YNynd power transformers when the stabilizing
Transformer #1 Transformer #2 Transformer #3 winding is closed and when it is open is a key factor in assess-
ing the need for this winding. This performance is determined by
A 7.7% 3.8% 1.2%
B 0.8% 1.3% 0.8% zero-sequence flux circulation inside the transformer under dif-
C 2.3% 1.7% 1.3% ferent network operating conditions, meaning that internal design
features such as the relative position of the windings (HV, LV and
stabilizing) or the presence of magnetic shields in the tank are of
Table 14 great importance.
Comparison of the different models’ mean errors in stabilizing winding current This paper proposed taking a complete set of onsite low-voltage
prediction. zero-sequence measurements to obtain overall information about
Model Quadratic mean error (QME)
the zero-sequence performance of the three most frequent design
types found in core-form power transformers, complementing the
Transformer #1 Transformer #2 Transformer #3
results published by the authors in Ref. [16].
A 29.2% 16.0% 23.0% Based on these measurements, the parameters of different
B 27.0% 9.3% 11.4% equivalent circuits were calculated using a set of equations and an
C 2.3% 2.2% 2.0%
optimization process. These circuit models comprised two T-type
equivalent circuits (the simpler option commonly used in represen-
tation of transformers in power system analysis tools) and a more
complex six-branch equivalent circuit.
current circulation when zero-sequence voltages are applied to the For the T-type models, the model whose parameters are differ-
transformer [10]. For example, in case of zero-sequence voltages ent depending on whether the stabilizing winding is closed or open
applied to HV winding, HV ampere-turns should be balanced with was more accurate than the alternative with common parameters
LV ampere-turns, stabilizing winding ampere-turns and ampere- in all cases. Separate T-type zero-sequence equivalent circuits usu-
turns from tank currents. This last term (ampere-turns from tank ally produce a complete and sufficiently accurate description of the
currents) is not taken into account by model B, as this is a T-type zero-sequence behavior of YNynd transformers for all stabilizing
model (see Fig. 4—B1).
38 A. Ramos, J.C. Burgos / Electric Power Systems Research 144 (2017) 32–40

Table 15
Comparison of the different models’ current prediction for transformer #1.

Test Code Model Current prediction for transformer #1

I1 (%) Error (%) I2 (%) Error (%) I3 (%) Error (%)

Measured 1.581 – – – 1.243 –

A 1.586 0.33 – – 1.472 18.39
HOC
B 1.579 0.09 – – 1.579 27.02
C 1.523 3.62 – – 1.232 0.91

Measured 1.590 – 1.386 – – –

A 1.517 4.58 1.521 9.74 – –
HSO
B 1.566 1.49 1.395 0.67 – –
C 1.558 2.01 1.364 1.60 – –

Measured 1.555 – 1.659 – 0.387 –

A 1.473 5.27 1.527 7.93 0.050 87.00
HSC
B 1.557 0.15 1.674 0.96 0.117 69.66
C 1.567 0.79 1.741 4.99 0.372 3.84

Measured – – 0.599 – 0.545 –

A – – 0.617 2.86 0.572 4.99
LOC
B – – 0.601 0.35 0.601 10.36
C – – 0.584 2.49 0.550 0.95

Measured 0.503 – 0.507 – – –

A 0.553 10.06 0.624 23.19 – –
LSO
B 0.507 0.96 0.514 1.52 – –
C 0.496 1.31 0.496 2.14 – –

Measured 0.225 – 0.587 – 0.360 –

A 0.203 9.92 0.565 3.65 0.337 6.45
LSC
B 0.222 1.22 0.586 0.21 0.363 0.96
C 0.231 2.73 0.594 1.16 0.372 3.35

Table 16
Comparison of the different models’ current prediction for transformer #2.

Test code Model Current prediction for transformer #2

I1 (%) Error (%) I2 (%) Error (%) I3 (%) Error (%)

Measured 0.699 – – – 0.695 –

A 0.698 0.19 – – 0.671 3.48
HOC
B 0.699 0.00 – – 0.699 0.56
C 0.701 0.34 – – 0.670 3.58

Measured 1.258 – 1.223 – – –

A 1.230 2.23 1.235 0.96 – –
HSO
B 1.271 1.02 1.189 2.77 – –
C 1.269 0.83 1.188 2.84 – –

Measured 0.821 – 0.931 – 0.164 –

A 0.794 3.24 0.872 6.32 0.075 54.56
HSC
B 0.821 0.05 0.929 0.24 0.107 34.66
C 0.827 0.75 0.928 0.34 0.163 0.80

Measured – – 1.171 – 1.160 –

A – – 1.189 1.58 1.144 1.38
LOC
B – – 1.173 0.23 1.173 1.19
C – – 1.149 1.84 1.181 1.83

Measured 0.766 – 0.839 – – –

A 0.831 8.48 0.921 9.83 – –
LSO
B 0.800 4.48 0.830 1.04 – –
C 0.800 4.41 0.831 0.98 – –

Measured 0.322 – 1.218 – 0.893 –

A 0.311 3.47 1.197 1.78 0.851 4.60
LSC
B 0.331 2.80 1.218 0.01 0.887 0.62
C 0.331 2.70 1.198 1.70 0.914 2.43

winding operation options. Nevertheless, in some cases, stabilizing a precise evaluation of internal transformer currents (especially
winding current prediction showed errors, though of slight signif- circulating current inside the delta stabilizing winding) is required.
icance. It should be noted that in case of no load condition and These experimental data and results will be of great use in analy-
stabilizing winding open, non-linearity of magnetizing impedances sis of tank overheating hazard and short-circuit duty in three-phase
can introduce a certain error in the use of the models. three-legged YNynd power transformers, which will be addressed
The six-impedance equivalent circuit supposes greater com- in the companion paper (Part II) to this study.
plexity both in obtaining the parameters and in calculating circuit
currents. Nevertheless, this option offers greater accuracy when
 A. Ramos, J.C. Burgos / Electric Power Systems Research 144 (2017) 32–40 39

Table 17
Comparison of the different models’ current prediction for transformer #3.

Test code Model Current prediction for transformer #3

I1 (%) Error (%) I2 (%) Error (%) I3 (%) Error (%)

Measured 0.265 – – – 0.257 –

A 0.265 0.13 – – 0.211 18.07
HOC
B 0.264 0.24 – – 0.264 2.75
C 0.264 0.36 – – 0.250 2.76

Measured 0.896 – 0.782 – – –

A 0.921 2.80 0.783 0.18 – –
HSO
B 0.902 0.64 0.774 0.93 – –
C 0.880 1.72 0.762 2.51 – –

Measured 0.897 – 0.487 – 0.396 –

A 0.893 0.44 0.481 1.21 0.327 17.39
HSC
B 0.899 0.27 0.487 0.02 0.412 4.10
C 0.907 1.18 0.490 0.60 0.410 3.56

Measured – – 0.359 – 0.324 –

A – – 0.360 0.28 0.286 11.76
LOC
B – – 0.359 0.19 0.359 10.97
C – – 0.354 1.26 0.325 0.38

Measured 0.366 – 0.375 – – –

A 0.379 3.70 0.373 0.64 – –
LSO
B 0.375 2.56 0.373 0.53 – –
C 0.369 0.92 0.371 1.09 – –

Measured 0.502 – 0.470 – 0.061 –

A 0.507 1.01 0.464 1.21 0.034 44.87
LSC
B 0.513 2.26 0.469 0.22 0.044 27.90
C 0.516 2.86 0.467 0.66 0.060 1.31

Acknowledgement Z0(HOO) = Z  01 + Z  0M (A.13)

  
Z0(HSO) = Z 01 +Z 02 //Z 0M (A.14)
The authors would like to thank Miguel Angel Cardiel, from
Universidad Carlos III de Madrid, for performing equivalent circuit  
Z0(LOO) = Z 02 +Z 0M (A.15)
calculations.
  
Z0(LSO) = Z 02 +Z 01 //Z 0M (A.16)
Appendix A.
- Alternative C equivalent circuit (Fig. 5):
In this section, the equations that allow transition from mea-  
sured quantities to equivalent circuit impedances are presented for Z0(HOO) = Z01 // Z01Y + (Z02Y + Z02 ) // (Z03Y ) (A.17)
each model. For simplicity, the development of shunt impedances  
is not shown. Z0(HOC) = Z01 // Z01Y + (Z02Y + Z02 ) // (Z03Y + Z03 ) (A.18)
 
Z0(HSO) = Z01 // Z01Y + (Z02Y ) // (Z03Y ) (A.19)
- Alternative A equivalent circuit (Fig. 3):
 
Z0(HSC) = Z01 // Z01Y + (Z02Y ) // (Z03Y + Z03 ) (A.20)
Z0(HOO) = Z01 + Z0M (A.1)
 
Z0(HOC) = Z01 + Z0M //Z03 (A.2) Z0(LOO) = Z02 // Z02Y + (Z01Y + Z01 ) // (Z03Y ) (A.21)
 
Z0(HSO) = Z01 + Z0M //Z02 (A.3) Z0(LOC) = Z02 // Z02Y + (Z01Y + Z01 ) // (Z03Y + Z03 ) (A.22)
Z0(HSC) = Z01 + Z0M //Z02 //Z03 (A.4)  
Z0(LSO) = Z02 // Z02Y + (Z01Y ) // (Z03Y ) (A.23)
Z0(LOO) = Z02 + Z0M (A.5)  
Z0(LSC) = Z02 // Z02Y + (Z01Y ) // (Z03Y + Z03 ) (A.24)
Z0(LSO) = Z02 + Z0M //Z01 (A.6)
where impedances Z01Y − Z02Y − Z03Y are corresponding star-
Z0(LSO) = Z02 + Z0M //Z01 (A.7) triangle transformation from Z012 − Z013 − Z023 :
Z0(LSC) = Z02 + Z0M //Z01 //Z03 (A.8) Z012 · Z013
Z01Y = (A.25)
Z012 + Z013 + Z023
- Alternative B1 equivalent circuit (Fig. 4—B1):
Z012 · Z023
Z02Y = (A.26)
Z0(HOC) = Z01 + Z03 (A.9) Z012 + Z013 + Z023
Z013 · Z023
Z0(HSC) = Z01 + Z02 //Z03 (A.10) Z03Y = (A.27)
Z012 + Z013 + Z023
Z0(LOC) = Z02 + Z03 (A.11)
References
Z0(LSC) = Z02 + Z01 //Z03 (A.12)
[1] L.F. Blume, Influence of transformers connections on operation, AIEE Trans. 33
- Alternative B2 equivalent circuit (Fig. 4—B2): (May) (1914) 753–770.
40 A. Ramos, J.C. Burgos / Electric Power Systems Research 144 (2017) 32–40

[2] J.F. Peters, Harmonics in transformer magnetizing currents, AIEE Trans. 34 [15] T. Ngnegueu, M. Mailhot, A. Munar, Zero-phase impedance and tank heating
(September) (1915) 2157–2182. model for three-phase three-leg core type transformers coupling magnetic
[3] L.N. Robinson, Phenomena accompanying transmission with some types of field and electric circuit equations in a finite element software, IEEE Trans.
star transformer connections, AIEE Trans. 34 (September) (1915) 2183–2195. Magn. 31 (May (3)) (1995) 2068–2071.
[4] J.F. Peters, M.E. Skinner, Transformers for interconnecting high-voltage [16] A. Ramos, J.C. Burgos, et al., Determination of parameters of zero-sequence
transmission systems for feeding synchronous condensers from a tertiary equivalent circuits for three-phase three-legged YNynd transformers based
winding, AIEE Trans. 40 (June) (1921) 1181–1199. on onsite low-voltage tests, IEEE Trans. Power Deliv. 28 (July (3)) (2013)
[5] J. Mini Jr, L.J. Moore, R. Wilkins, Performance of auto transformers with 1618–1625.
tertiaries under short-circuit conditions, AIEE Trans. 42 (October) (1923) [17] A. Ramos, Consideraciones acerca de la Utilización de Arrollamientos de
1060–1068. Estabilización en Transformadores Estrella-Estrella (in Spanish). Pd. D.
[6] R.O. Kapp, A.R. Pearson, The performance of star–star transformers, J. Inst. Dissertation, Universidad Carlos III de Madrid, 2016.
Electr. Eng. 1 (January (6)) (1955) 9–12. [18] M.F. Lachman, Y.N. Shafir, Influence of single-phase excitation and
[7] B.A. Cogbill, Are stabilizing windings necessary in all Y-connected magnetizing reactance on transformer leakage reactance measurement, IEEE
transformers? AIEE Trans. Power Appar. Syst. 78 (October (3)) (1959) Trans. Power Deliv. 12 (October (4)) (1997) 1538–1546.
963–970. [19] S.V. Kulkarni, S.A. Khaparde, Transformer engineering, in: Design and Practice,
[8] O.T. Farry, Tertiary windings in autotransformers, AIEE Trans. Power Appar. Marcel Dekker, 2004.
Syst. 80 (April (3)) (1961) 78–82. [20] A. Ramos, J.C. Burgos, Influence of stabilizing windings on zero-sequence
[9] A.N. Garin, Zero-phase sequence characteristics of transformers. Part I: performance of three-phase three-legged YNynd transformers. Part II: tank
sequence impedances of a static symmetrical three-phase circuit and of overheating hazard and short-circuit duty, Electr. Power Syst. Res. XXXX. (in
transformers, Gen. Electr. Rev. 43 (March (3)) (1940) 131–136. revision).
[10] A.N. Garin, Zero-phase sequence characteristics of transformers. Part II: [21] N. Chiesa, B.A. Mork, H.K. Høidalen, Transformer model for inrush current
equivalent circuits for transformers, Gen. Electr. Rev. 43 (April (4)) (1940) calculations: simulations, measurements and sensitivity analysis, IEEE Trans.
174–179. Power Deliv. 25 (October (4)) (2010) 2599–2608.
[11] K. Schlosser, Die Nullimpedanzen des Voll- und des Spartransformatoren [22] IEC Std. 60076-8, Power Transformers, Part 8: Application Guide. 1997.
(Zero-sequence impedance in transformers and autotransformers), [23] IEEE Std. C57.12.90, IEEE Standard Test Code for Liquid-Immersed
Brown-Boveri-Nachr. 44 (February) (1962) 78–83 (in German). Distribution, Power and Regulating Transformers, 2010.
[12] M. Christoffel, Zero-sequence reactances of transformers and reactors, Brown [24] D. Marquardt, An algorithm for least-squares estimation of nonlinear
Boveri Rev. 52 (November/December (11/12)) (1965) 837–842. parameters, J. Soc. Ind. Appl. Math. 11 (2) (1963) 431–441.
[13] K.P. Oels, Ersatzschaltungen des Transformators für das Nullsystem
(Zero-Sequence Equivalent Circuits for Transformers) ETZ-A Bd, 89, 1968, pp.
59–62 (in German).
[14] R. Allcock, S. Holland, L. Haydock, Calculation of zero-phase sequence
impedance for power transformers using numerical methods, IEEE Trans.
Magn. 31 (May (3)) (1995) 2048–2051.