Four-Winding Transformer and Autotransformer Modeling

For Load-Flow and Short-Circuit Analysis, Part II: Application

Carlos A. Muñoz Alberto Rojas
Universidad Nacional de Ingeniería, Lima, Perú COVIEM S.A., Lima, Perú

Abstract
Purpose- This paper presents the application of the theory for computing the positive and zero
sequence admittance matrices of three-phase four-winding transformers and autotransformers. This
theory was presented as a Part I paper. The software WinFdc of the company COVIEM S.A. has been
upgraded to handle four-winding transformers and autotransformers for load-flow, short-circuit and
harmonics analysis.
Methodology- The calculation engines of WinFdc and its interfaces with a MS-Access database and
CADOpia/AutoCAD were upgraded to model four-winding transformers and autotransformers.
Validation tests were carried out taking as a benchmark results for three-winding transformers. For
load-flow analysis the robustness of the calculation engine was tested by modeling heavily loaded
four-winding transformers. For short-circuit analysis complete models of four-winding transformers
and autotransformers were developed and tested.
Research implications- The proposed theory has been successfully validated and applied. Calculation
engines for load-flow, short-circuit and harmonics can be reliable upgraded.

Key words: Four-windings, transformers, autotransformers, sequence networks, load-flow, short-

circuit, harmonics

I. Introduction

The authors of this article presented in reference [1] a theory for computing the positive and zero
sequence admittance matrices of four-winding transformers and autotransformers that are required
for load-flow and short-circuit calculations. In reference [1] the proposed theory was validated using
MS-Excel spreadsheets. The proposed theory has now been used to upgrade a Peruvian software,
called WinFdc, and developed by the authors, for load-flow, short-circuit and harmonics analysis of
balanced three-phase power networks. The results of this research and software development effort
are presented.

Section II presents an overview of WinFdc. Section III shows the application of the proposed theory
to the load-flow module. Section IV shows the application of the proposed theory to the short-circuit
module. Section V presents the application of the proposed theory to the harmonics module.
Finally, in Section VI the main conclusions of this research effort are outlined.

For testing and validating the four-winding transformer and autotransformer models, data from the
Peruvian high voltage power grid (SEIN – Sistema Eléctrico Interconectado Nacional) has been
used.

The proposed theory can be successfully applied to upgrade software for power network analysis.
Careful software design can give robust calculation engines.

II. WinFdc’s Overview

WinFdc is a Peruvian software that can model balanced three-phase power networks for load-flow,
short-circuit and harmonics analysis. The main features of WinFdc are:

 A load-flow calculation engine called NewFdc

 A short-circuit calculation engine called NewCf
 A harmonics calculation engine called FdcHarmo
 A database engine using MS-Access
 A graphical interface using CADOpia or AutoCAD
 An interface with the PSS/E of Siemens PTI

Page 1 of 8
The calculation engine of the load-flow module (NewFdc) uses the fast decoupled method [2] to
obtain a solution. The real matrices B’ and B’’ are built taking into account the type of each bus
v
(slack, PV, or PQ). NewFdc also builds a real matrix B to use it for sensitivity analysis of changes
v
in taps positions of transformers. Matrix B has the same structure of matrix B’’ but is built taking
into account the equipments and reactance values used for matrix B’ (tap positions are 1.0 pu and
only positive sequence reactances of lines and transformers are modeled).

The calculation engine of the short-circuit module (NewCf) uses the theory given in reference [3].
NewCf can simulate faults (three-phase LLL, two-phase L-L, phase to ground L-G, or two-phase to
ground LL-G) at a bus or at a specified point in a line. Also, one-open conductor (1-OC) or two-
open conductors (2-OC) faults at a specified point in a line can be simulated. Positive and zero
sequence impedances of generators, lines, transformers, capacitors, reactors and loads are
processed by NewCf. For generators, the negative sequence impedances can be processed. Also,
zero sequence mutual impedances between lines can be modeled. NewCf can give results for a flat
profile (prefault voltages 1.000 pu and 0.00 degrees, without loads, with transformer tap positions
equal to 1.000 pu, and without capacitors and reactors) or it can use a load-flow solution given by
NewFdc. NewCf can simulate faults using a conventional method or can perform calculations
according to ANSI or IEC standards.

The calculation engine of the harmonics module (FdcHarmo) has been developed using the same
network methods used by NewCf. As a matter of fact, this has allowed reusing a lot of NewCf
source code. The equipment impedances are computed for each harmonic frequency of analysis
besides the nominal frequency. These impedances can be computed according to IEEE or CIGRE
standards. FdcHarmo can perform a frequency scan to obtain the positive and zero sequence
impedance for a range of frequencies, or it can process a set of injected harmonic currents at
several buses to obtain a solution of harmonic voltages and currents.

Before including models of four-winding transformers and autotransformers, WinFdc had models of
two-winding and three-winding transformers and autotransformers. These models had been fully
tested. They are used as benchmark models for testing and validating the four-winding models
when a winding of a four-winding transformer or autotransformer has no impact in the calculations
because that winding has no load or it is a start winding that is not connected to other equipments
of the network.

WinFdc software is in Spanish. The figures shown below corresponding to WinFdc are therefore in
Spanish, but the authors believe that they are clear enough for readers not familiar with Spanish.
Nevertheless, the authors are ready to answer to any questions concerning these figures.

III Load-flow module

A four-winding transformer model has been included in the NewFdc calculation engine of WinFdc.
This model takes only into account the positive sequence admittance matrix since the three-phase
network is assumed perfectly balanced. The method for building the admittance matrix is the same
in the case of an autotransformer. Since NewFdc uses the fast decoupled method, two admittance
matrices are built: a real matrix with only the transformer reactances and a complex matrix by
processing all the impedances and the magnetizing admittance. The real matrix is used for building
v
matrices B’ and B ; the complex matrix is using for building matrix B’’ and computing the current and
power (active and reactive) injected to each winding.

Figures 1a and 1b show the data of the Juliaca four-winding transformer. This is the same data that
was used in reference [1]. Factor 2.00 in Figure 1a indicates that base power is 2x30 MVA. The
impedances in Figure 1b are at a common base power of 30 MVA (Pbase).

Page 2 of 8
 Figure 1a: Juliaca Network Transformer Data

Figure 1b: Juliaca Equipment Transformer Data

Page 3 of 8
Figure 1b shows that the magnetizing admittance is equal to zero. WinFdc models of two-winding
and three-winding transformers have the magnetizing admittance modeled connected to the
primary bus, in the four-winding model the magnetizing admittance is modeled connected to the
primary input node an ideal four-winding transformer (see reference [1]). Neglecting the
magnetizing admittance allows comparing load-flow simulation results between four-winding and
three winding transformers.

For the first load-flow test, a four-bus network was set up in WinFdc. The bus names and base
voltages are those shown in Figure 1a. Equivalent generators were modeled to be connected one
at a time to 138 kV, 60 kV and 22.9 kV buses. The following loads were set up, as shown in Figure
2. P is in MW and Q is in MVAr.

Figure 2: Loads for Juliaca Test Case

Table 1 below shows the load flow comparison between four-winding and three-winding models:

Results for Juliaca Test Case

Energized at 138 kV – Loads at 60 kV and 22.9 kV connected – Maximum Power Mismatch 0.001 MW
Four-winding model Three-winding model
V pu degrees V pu Degrees
Juliaca 138 kV 1.0000 0.000 1.0000 0.000
Juliaca 60 kV 0.8997 -7.456 0.8997 -7.456
Juliaca 22.9 kV 0.8792 -9.773 0.8792 -9.773
Juliaca 10 kV 0.8901 -8.752 Not modeled
Number of iterations Converged in 7 iterations Converged in 8 iterations
Energized at 138 kV – Loads at 22.9 kV and 10 kV connected – Maximum Power Mismatch 0.001 MW
Four-winding model Three-winding model
V pu degrees V pu Degrees
Juliaca 138 kV 1.0000 0.000 1.0000 0.000
Juliaca 60 kV 0.9350 -5.593 Not modeled
Juliaca 22.9 kV 0.9005 -9.188 0.9005 -9.188
Juliaca 10 kV 0.9000 -9.370 0.9000 -9.370
Number of iterations Converged in 7 iterations Converged in 7 iterations
Energized at 60 kV – Loads at 22.9 kV and 10 kV connected – Maximum Power Mismatch 0.001 MW
Four-winding model Three-winding model
V pu degrees V pu Degrees
Juliaca 138 kV 1.0029 0.482 Not modeled
Juliaca 60 kV 1.0000 0.000 1.0000 0.000
Juliaca 22.9 kV 0.9710 -2.641 0.9710 -2.641
Juliaca 10 kV 0.9707 -2.778 0.9707 -2.777
Number of iterations Converged in 4 iterations Converged in 4 iterations
Energized at 138 kV – All loads connected – Maximum Power Mismatch 0.001 MW
Four-winding model Three-winding model
V pu degrees V pu Degrees
Juliaca 138 kV 1.0000 0.000 Not applicable
Juliaca 60 kV 0.8066 -13.579 Very heavy load for the
Juliaca 22.9 kV 0.7597 -19.434 four-winding txfo. Capacity 2x30 MVA
Juliaca 10 kV 0.7587 -19.727 Total load equal to
Number of iterations Converged in 15 iterations 110 MW + j 65 MVAr

Table 1: Load Flow Results for Four and Three Winding Models

Page 4 of 8
The load-flow results for the Juliaca test case shown above indicate an exact agreement between
the four-winding model and the three-winding model. The last simulation of this case corresponds to
a very high load for the four-winding transformer; convergence is obtained with voltages well below
1.000 pu. This indicates that the algorithm of NewFdc is robust for this kind of simulations. For the
case of energizing at 60 kV, the 138 kV busbar voltage is 1.0029 pu; the voltage drop at 60 kV is
negative and the model behaves like a three-winding transformer with a negative reactance at the
60 kV side.

The load-flow four-winding model was tested in the Peruvian power network (SEIN). This network
modeled in WinFdc has 1155 buses. A 220/138/22.9/10 kV four-winding transformer at
Carhuaquero substation was added to the database. Figure 3 below shows the parameters of the
Carhuaquero transformer.

Figure 3: Carhuaquero transformer equipment data

Load flow simulations for the SEIN test case, including the Carhuaquero transformer, converge
even with a power mismatch as low as 0.001 MW. A test with a very high load of 290% gives
convergence with a power mismatch of 0.001 MW; the total generation of the SEIN is 4874 MW.
These results indicate again a robust algorithm for NewFdc.

IV Short-circuit module

A four-winding transformer model has been included in the NewCf calculation engine of WinFdc.
This model takes into account the positive sequence admittance matrix as well as the zero
sequence admittance matrix. The method for building the positive sequence admittance matrix is
the same in the case of an autotransformer; the winding connections (Ag: grounded
autotransformer, A: ungrounded autotransformer, Yg: start grounded, Y: start ungrounded, or D:

Page 5 of 8
Delta) are not considered for positive sequence. The method for building the zero sequence
admittance matrix takes into account winding connections and grounding resistances. There are 99
combinations of winding connections for zero sequence.

Although in practice only one or two combinations will be used (AgAgYgD or YgYgYgD), the 99
combinations have been included in NewCf; for example, an AADD or YgDDD transformer can be
modeled by NewCf. These 99 combinations have been fully tested and debugged. The positive and
zero sequence admittance matrices are built for a conventional kind of short circuit analysis. NewCf
also builds another set of positive and zero sequence admittance matrices according to ANSI and
IEC standards.

Table 2 below shows a short circuit comparison for the Juliaca test case:

Results for Juliaca Test Case

Energized at 138 kV – Winding Connections AgAgYgD – Rg(p) = 2 Ohm and Rg(t) = 4 Ohm
Short-circuit power LLL and Four-winding model Three-winding model
L-G 10000 MVA R/X=0 LLL (kA) L-G (kA) LLL (kA) L-G (kA)
Juliaca 138 kV 41.837 42.281 41.837 42.281
Juliaca 60 kV 5.366 6.750 5.366 6.750
Juliaca 22.9 kV 5.919 4.718 Not modeled
Juliaca 10 kV 18.282 0.000 18.282 0.000

Energized at 138 kV – Winding Connections YgYgYgD – Rg(p) = 2 Ohm, Rg(s) = 2 Ohm and Rg(t) = 4 Ohm
Short-circuit power LLL and Four-winding model Three-winding model
L-G 10000 MVA R/X=0 LLL (kA) L-G (kA) LLL (kA) L-G (kA)
Juliaca 138 kV 41.837 42.281 41.837 42.281
Juliaca 60 kV 5.366 6.701 5.366 6.701
Juliaca 22.9 kV 5.919 4.718 Not modeled
Juliaca 10 kV 18.282 0.000 18.282 0.000

Energized at 60 kV – Winding Connections AAYgD – Rg(t) = 4 Ohm

Short-circuit power LLL and Four-winding model Three-winding model
L-G 10000 MVA R/X=0 LLL (kA) L-G (kA) LLL (kA) L-G (kA)
Juliaca 138 kV 2.333 2.160 2.333 2.160
Juliaca 60 kV 96.225 96.225 96.225 96.225
Juliaca 22.9 kV 11.061 5.484 Not modeled
Juliaca 10 kV 49.780 0.000 49.780 0.000

Energized at 22.9 kV – Winding Connections YgYgDD – Rg(p) = 2 Ohm and Rg(s) = 3 Ohm
Short-circuit power LLL and Four-winding model Three-winding model
L-G 10000 MVA R/X=0 LLL (kA) L-G (kA) LLL (kA) L-G (kA)
Juliaca 138 kV 0.982 1.093 0.982 1.093
Juliaca 60 kV 4.221 4.993 Not modeled
Juliaca 22.9 kV 252.118 252.118 252.118 252.118
Juliaca 10 kV 27.634 0.000 27.634 0.000

Notes:
A) The currents LLL and L-G in kA are the total fault currents at each busbar. A prefault voltage of 1.000 is
used.
B) Rg(p), Rg(s) and Rg(t) are the grounding resistances in ohms.

Table 2: Short Circuit Results for Four and Three Winding Models

The short-circuit results for the Juliaca test case shown above indicate an exact agreement
between the four-winding model and the three-winding model. Short-circuit simulations were carried
out with the SEIN test case that includes the Carhuaquero four-winding transformer. NewCf gave
and exact agreement taking as a benchmark the three-winding model of transformers and
autotransformers.

A special test case was set up for validating the impact of the zero sequence magnetizing
admittance of the four-winding model. A four-winding connection AAYY was specified in the Juliaca

Page 6 of 8
test case, and B0 was given a value of 10% (0.10 pu at 2x30 MVA). Juliaca four-winding
autotransformer was energized at 60 kV with a power source that has a LLL short-circuit power of
10000 MVA and R/X = 0, and a L-G short-circuit power of 10000 MVA and R/X = 0. A L-G fault was
simulated at Juliaca 138 kV. The zero sequence Thevenin impedance computed by NewCf is:
0.00000 + j 5.32640 pu at 100 MVA and 138 kV. According to the classic T&D book (reference [4]),
2
the zero sequence impedance in ohms is: 1) power source 0.01*60 /100 = 0.36 ohms, and 2)
2
autotransformer 10*(138-60) /60 = 1014 ohms. The total zero sequence impedance is j 1014.36
ohms; the value in pu at 100 MVA and 138 kV is j 5.326402016 pu. There is an exact agreement of
the zero sequence impedance value computed by NewCf with the value computed using the
formula of the T&D. Also, the L-G fault current at 138 kV is 0.221 kA.

V Harmonics module

A four-winding transformer model has been included in the FdcHarmo calculation engine of WinFdc.
This model takes into account the positive sequence admittance matrix as well as the zero
sequence admittance matrix. The methods for building the positive sequence admittance matrix and
the zero sequence admittance matrix are the same that the methods used for the short-circuit
calculation engine. The only difference is that the transformer impedances are computed to the
frequencies of interest from the 60 Hz values. As a matter of fact, almost all the source code of
NewCf for modeling four-winding transformers has been reused in FdcHarmo.

The tests and validation of FdcHarmo are similar to the tests and validation of NewCf. A software
error in the four-winding models of NewCf was corrected also in FdcHarmo, and vice versa.
Comparison tests were carried out between four-winding models and three-winding models. Tests
were performed for comparing results of impedance values from frequency scans. Then tests were
performed for comparing results of harmonic voltages and current due to a set of harmonic injected
currents. The tests were successful and gave as a result and exact agreement between the four-
winding model and the three-winding model.

VI Conclusion

The theoretical approach, presented by the authors, for modeling three-phase four-winding
transformers and autotransformers for load-flow and short-circuit analysis, has been applied and
validated. A Peruvian software for load-flow, short-circuit and harmonics analysis has been
upgraded to handle four-winding transformers and autotransformers. The tests and validations
carried out were successful. A careful software design and programming gives as a result robust
calculation engines.

As a result of this research and software development effort, the software WinFdc [5] was upgraded
from version 2.02 to version 2.03.

References:

[1] C. A. Muñoz, A. Rojas, "Four-Winding Transformers and Autotransformers Modeling for Load-Flow
and Short-Circuit Analysis, Part I: Theory", available at www.researchgate.net.
[2] B. Stott, O. Alsac, "Fast Decoupled Load Flow”, IEEE PAS 93, No. 3 pp. 859-867, May/June 1974.
[3] Paul M. Anderson, "Analysis of Faulted Power Systems”, The Iowa State University Press, USA, Third
printing 1978.
[4] Central Station Engineers of the Westinghouse Electric Corporation, "Electrical Transmission and
Distribution Reference Book", East Pittsburgh, Pennsylvania, USA, September 1950.
[5] A. Rojas, C. A. Muñoz, “WinFdc 2.03 User Manual”, in Spanish, COVIEM S.A., Lima, Peru.

About the authors:

Carlos A. Muñoz is associate professor at Universidad Nacional de Ingeniería at Lima, Peru. He is also the
general manager of COVIEM S.A. COVIEM designs, builds and carries out the commissioning of electrification
projects. His research interests are load-flow, short-circuit, and electrical protections. He can be contacted at
[email protected].

Page 7 of 8
Alberto Rojas is a consultant engineer. He works at COVIEM S.A. in Lima, Peru, His research interests are
load-flow, short-circuit, harmonics and grounding systems analysis; and also software development for power
systems studies. He can be contacted at [email protected].

Manuscript finished on October 19, 2016.

Page 8 of 8