Received: 1 March 2022 Revised: 4 October 2022 Accepted: 17 November 2022 The Journal of Engineering

DOI: 10.1049/tje2.12221

REVIEW

A review of modelling techniques of power transformers for digital

real-time simulation

Md Maidul Islam1 Matthias Musil1 Md Jamal Ahmed Shohan1

Md Omar Faruque1 Georg Lauss2 Ali Dehkordi3 Paul Forsyth3
Panos Kotsampopoulos4 Kai Strunz5 Zhihui Li5 Yi Zhang3 Peng Liu6

1
Electrical & Computer Engineering, FSU, Abstract
Tallahassee, USA
This task-force paper documents and summarizes the models of power transformers that
2
Electric Energy SystemsEnergy Department, have been proposed, used, and validated in the exercise of digital real-time simulation.
Austrian Institute of Technology, AIT, Vienna,
Austria
Power transformer is one of the most important equipment in power systems, and its mod-
3
elling for electromagnetic transient simulation has evolved over time, especially in the area
RTDS Technologies Inc., Manitoba, Canada
of real-time simulation. The focus of the paper is to document and archive the models that
4
Electrical and Computer Engineering, National
have been well accepted and used for transient analysis in digital real-time simulation so
Technical University of Athens, NTUA, Athens,
Greece that readers can use it as a master document for transformer modelling in real-time simu-
5
Sustainable Electric Networks and Sources of
lation studies. It includes both conventional and specialized models of power transformers
Energy, TUB, Berlin, Germany that have been broadly acknowledged by the power engineering community. The models
6
Department of Electrical and Computer provided here come with detailed mathematical representation and their implementation
Engineering, University of Alberta, Canada techniques. A comparative study is also performed to illustrate the differences in their per-
formances. In the end, an application guideline has been provided to guide the readers to
Correspondence select the appropriate model for their study.
Georg Lauss, Austrian Institute of Technology, AIT,
Vienna, Austria.
Email: [email protected]

Funding information
H2020 Research Infrastructures, Grant/Award
Number: 870620

1 INTRODUCTION performance parallel structured computing platform that is

mainly used to test the effect of a newly developed component
Digital real-time simulation is a powerful tool that utilizes the onto a system before actually implementing it in real field.
fast computational power of digital hardware to characterize This type of testing enables the understanding of the transient
system behaviour in real-time. It is generally used to observe behaviour of such addition or replacement of new hardware
the transient behaviour of the system on a simulation test-bed under the dynamic influences. A technology which is widely
to improve design accuracy, develop a fast prototype, avoid known as hardware-in-the-loop (HIL) simulation or testing
any catastrophic system failure, and hence minimize the risk of reduces the risk of investment through the use of a prototype
failure in the real field. development and testing that interacts with a simulated system
With increasing processing power and the capability of environment in real-time.
parallel processing, an in-depth and accurate analysis of a more A detailed real-time simulation technique is broadly explained
complicated system is now a reality. As a result, the potential in ref. [1], which requires the states of the simulated system to
of developing more sophisticated and detailed models of be solved accurately achieving resemblance to corresponding
power system components for real-time simulation is contin- physical counterpart within an acceptable range and within the
uously growing. A digital real-time simulator (DRTS) is a high length of time required for the real physical counterpart to run.

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is
properly cited.
© 2023 The Authors. The Journal of Engineering published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.

J. Eng. 2023;2023:e12221. wileyonlinelibrary.com/iet-joe 1 of 16

https://doi.org/10.1049/tje2.12221
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2 of 16 ISLAM ET AL.

with electrical configuration. There are four different techniques

of developing the electric equivalent circuit. This modelling
technique is easiest to use but lacks some of the detailed anal-
ysis capabilities like geometrical configuration effects. Magnetic
equivalent circuit takes core geometry into account and thus is
suitable for more detailed analysis including magnetic character-
istics and the effect of geometrical configuration. Finite element
solution based models are implemented on FPGA and provide
more accurate representation of non linear behaviour and mag-
netic field characteristics. In the following sections, different
real-time transformer models are discussed followed by test case
FIGURE 1 Types of transformer models for real-time simulation simulations for some of the models. An application guideline is
provided in the latter part of the paper so that the users can
choose the appropriate model, i.e. the model that fits the best
Real-time models are required to solve a whole, often grid-scale for a certain analysis. The guideline also mentions the required
model within a time-step small enough to be able to reproduce input parameters and associated modelling complexity for each
all the necessary transient effects faithfully. The models there- specific model.
fore need to be solvable as fast as possible while still being
as much accurate as necessary. A 50 𝜇s is the widely accepted
time-step to study dynamic response up to 3 kHz of power 2 ELECTRIC EQUIVALENT CIRCUIT
systems. To investigate higher frequency phenomena, smaller MODEL
time-step (<1 𝜇s) may be required [2]. Generally speaking, more
accurate models require longer computation time and more 2.1 Piecewise linear saturation model [4, 5]
powerful computing resources. With the drastic improvement in
hardware speed, real-time digital simulation can now be imple- The piecewise linear model uses parameterized curve based
mented in the range of nano-second time-step for complex ideal switches to model the magnetizing path of the trans-
cases in the fields of power electronics hardware design, pro- former. The model only requires a proper saturation curve
totyping, and testing. Summary papers published by the TF [1, which can be easily obtained from the core material charac-
3] discuss real-time simulation platforms and their applications teristics. This model is capable of simulating a magnetically
in details. decoupled, three-phase saturable transformer in real-time with-
Transformers are one of the most important parts of power out requiring any geometrical parameters. Ref. [6] describes this
systems. Switching effects or power system faults may lead to model with both saturation properties off (ideal) and on (sat-
inrush currents which may be up to the short circuit current in uration) mode. This model is easiest to implement and does
magnitude. The amplitudes of inrush current decay to steady- not require core or magnetic data. A YNd11 transformer is
state magnetizing current slowly after a few cycles. Magnetic discussed here as an example.
saturation effects and inrush currents are some of the important
phenomena of a transformer to be studied through real-time
HIL simulation. Depending on the objective of the study, mod- 2.1.1 Equivalent circuit YNd11 transformer
els with different fidelity need to be adopted to achieve accurate
results. The main differences in modelling approaches are the Figure 2 shows the equivalent circuit of the specified trans-
chosen coupling mode such as coupled or decoupled inductance former considering saturation characteristics. The magnetizing
as well as ideal, linear, saturable magnetization models. Based on inductance Lm can be varied by switching, depending on the
this concept, the existing real-time transformer models can be magnetic flux Ψ.
divided into three main categories, as shown in Figure 1.
Each real-time model has its own strengths and drawbacks
in producing the transient behaviour of transformers. The 2.1.2 Modelling the saturable magnetizing
model input parameters also vary corresponding to the adopted inductance
circuit design to produce transient behaviour suitable for dif-
ferent types of analysis. It should be noted that, in this paper Usually, the saturation characteristic of an inductor is drawn in a
“transient” analysis has been used for time domain analysis magnetic flux density (B) vs. magnetic field strength (H) curve.
perspective. Transient analysis is a vast area and careful consid- The inductors flux linkage-magnetizing current (Ψ–i) curve is
eration should be taken while selecting appropriate transformer equivalent, if its core is geometrically regular. As it can be seen
models. Some of the models presented in this paper has been from Figure 2, the magnetizing path is approximated by switch-
used for applications like inrush transients, fault analysis, and ing the two inductances in parallel. The modelled curve is shown
switching transients etc. However, not all models are capable in Figure 3. It has three sections, where Section 1 is the lin-
of such analysis, or a certain model may not be suitable for all ear part of the inductor’s magnetization, and Sections 2 and 3
types of detailed transient analysis in real-time. Electrical equiv- are the saturated ones. The transition from Section 1 to Sec-
alent circuit model approximates the saturation characteristics tions 2 and 3 is done by closing the switch and adding Lm2 in
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ISLAM ET AL. 3 of 16

characteristics of an ideal switch. In the “on” state, the switch-

ing curve “S” is equal to the current is and in the “off” state, “S”
is equal to the voltage us over the switch.

2.1.3 Modelling of the YNd11 transformer

In this section, a short summary of the equations used for devel-

oping piecewise linear model has been presented. The full list
of formulas, steps and theory behind the model can be found in
ref. [4].

1. Assumptions: For the sake of simplification, the following

assumptions are made for the modelling process. All three
phase impedances of the primary winding and the secondary
winding are identical, respectively, resulting in: RA = RB =
FIGURE 2 Equivalent circuit of the three-phase YNd11 transformer RC = R1 , LA = LB = LC = L1 , Ra = Rb = Rc = R2 , La =
from ref. [4] Lb = Lc = L2 . It is assumed that the core is symmetrical and
core losses are distributed symmetrically for all three phases.
Furthermore, all three phases use the same saturation curve,
so all magnetizing resistances (RM ) and inductances (LM )
satisfy the following relations: Lm1A = Lm1B = Lm1C =
Lm1 , Lm2A = Lm2B = Lm2C = Lm2 , RmA = RmB = RmC =
Rm . In this model, for subsystem coupling purpose, the
core-losses are modelled outside of the transformer block.
However the core loss can be modelled inside the trans-
former block by including the core loss resistance Rm and
core loss current ic in the model.
2. Circuit part of the system: Using circuit analysis techniques
(nodal and mesh analysis), the node and mesh equations are
obtained for each phase. For simplicity, the t symbol repre-
senting time variable is omitted and only the equations of one
FIGURE 3 Approximated Ψ–i curve from ref. [4]
phase are derived, as they are the same for all phases except
for the subscript.

L1 i̇A = −R1 iA − umA + uA − uN (1)

L2 i̇ac = −R2 iac − umA + uac (2)

Ψ̇ A = umA (3)

Lm1A i̇m1A = umA (4)

Lm2A i̇sA = umA − usA (5)

im1A + isA − imA = 0 (6)

FIGURE 4 i–u curve of an ideal switch [4]

iA + iac − imA = 0 (7)

parallel to Lm1 . By increasing the number of parallel switched
inductances, the error of the saturation characteristic could be iA + iB + iC = 0 (8)
decreased. However, a higher number of switched inductors
also leads to a higher computation time, which is not favourable 3. Switching part of the system: The switching part of the system
for real-time simulations. This type of saturation curve mod- consists of six algebraic equations formed by the parame-
elling is also shown in refs. [5, 7]. The switches are presented terized curves with variables s, us , and is as shown in Table 1.
by an ideal switch model as proposed in ref. [4]. Figure 4 shows Here, for each phase, when the switch is “on”, s is defined by
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 of 16 ISLAM ET AL.

TABLE 1 Switching part of the system description [4]

Description

Switch “on” Switch “off”

Phase A usA = 0, sA − isA = 0 isA = 0, sA − usA = 0

Phase B usB = 0, sB − isB = 0 isB = 0, sB − usB = 0
Phase C usC = 0, sC − isC = 0 isC = 0, sC − usC = 0

FIGURE 6 Mutually coupled windings [6]

FIGURE 7 Equivalent circuit for two mutually coupled windings [6]

for a single-phase representation in the following subsections,

but can easily be extended to three phases.

2.2.1 Mutual coupling and model equations

The electrical relationship of this circuit is given by:
[ ] [ ] [ ]
E1 L11 M12 d i1
= (9)
E2 M12 L22 dt i2

with the coupling coefficient,

FIGURE 5 Transformer model representation [6] M12
K12 = √ (10)
L11 L22
the current is through the switch, and when switch is “off”,
s becomes the voltage us over the switch. Considering the turns ratio,
√
E L11
a= 1 = , (11)
2.2 Look-up and interpolation model [6, 8] E2 L22

Equation (9) can be rewritten as:

This is one of the standard models used in ref. [6] for real-time
digital simulation. Based on equivalent circuit representation, [ ] [ ] [ ]
E1 L11 aM12 d i1
the model can be ideal, linear or saturated. Figure 5 presents = i2 (12)
equivalent circuit representation of these three models. The aE2 aM12 a2 L22 dt
a
ideal transformer model does not have any magnetizing induc-
tance and represented by only leakage reactance. The linear
model includes a magnetizing branch modelled with inductive 2.2.2 Equivalent circuit for two mutually
branch inductance LMAG . The sum of inductances L1 and L2 coupled windings
represents the leakage reactance. In the saturation model the
magnetizing branch is placed at one end of the leakage reac- Equation (12) leads to the equivalent circuit used in the trans-
tance and the changeable current source in parallel represents former models in ref. [6]. An important factor to be considered
core effects. here is that winding losses are not included as they are assumed
Compared to the piecewise linear model, it is based on the to be zero. For Figures 6 and 7, the following definitions are
mutual coupling of windings. The model is further explained derived:
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ISLAM ET AL. 5 of 16

FIGURE 8 Magnetizing branch representation [6]

FIGURE 9 The source current (left) and hysteresis loop characteristics

L1 = L11 (1 − K12 ) (13) (right) [6]

L2 = L11 (1 − K12 ) (14) representing the major loop of 𝜙 vs. I characteristics are gen-
erated by solving Equation (16) and stored offline in the simula-
aM12 = L11 K12 (15) tion processor’s RAM. A second set of curves representing flux
differences are then used along with these curves to calculate
the required current injection for a given amount of flux. In this
2.2.3 Modelling the saturable magnetizing model, eddy losses have not been modelled separately. However,
branch the core loss component includes both hysteresis loss and eddy
loss. The user can add eddy loss component by defining the
Opposed to the previous model, non-linear magnetizing effects loop as a percentage of the known 1 p.u magnetizing current.
can also be modelled by adding a variable current source IHS In ref. [8], a parameter estimation approach to model the satu-
in parallel to a fixed inductance LMAG . Using the Dommel solu- ration characteristics of the transformer saturation model based
tion technique shown in ref. [9], which replaces an inductance by on experimental data is presented. Such technique can also be
an equivalent resistance and a parallel current source, the mag- extended to include hysteresis and magnetization loss.
netizing branch can be represented as shown in Figure 8. As
a simplification, the two current sources are added to a single
one. The current source IS now represents the core saturation 2.3 Multi-scale transformer modelling
and losses. Also, the stray inductances L1 and L2 are added and [11–14]
inserted after the magnetizing branch for this model to work.
To model the hysteresis loop and saturation effects, an algo- Multi-scale modelling is aimed at the adaptive tracking of both
rithm to calculate IS from the magnetic flux has been developed natural waveforms, as they are observed in reality, and the
in ref. [6] and is based on ref. [10]. envelopes of AC waveforms. The decision on which waveform
to track depends on the level of accuracy needed at a certain
√
[ (𝜙s − 𝜙K )2 + 4DLA + (𝜙s − 𝜙K )] D moment in time. As such, instead of just processing real instan-
IS = − (16) taneous signals which is common in digital real-time simulation,
2LA 𝜙K
multi-scale modelling integrates the advantages of instantaneous
where, and phasor signals by formulating all models in terms of analytic
signals. Adding the quadrature component [s(t )] as the imag-
√
−B − B 2 − 4AC inary part to a real instantaneous signal s(t ), the corresponding
D= analytic signal, marked by underscore is obtained [11]:
2A
LA LA IM − 𝜙M s(t ) = s(t ) + j [s(t )] (18)
A= B=
𝜙K
2 𝜙K (17)
The Fourier spectrum of a corresponding analytic signal s(t )
C = IM (IM LA − 𝜙M + 𝜙K ) does not extend toward negative frequency ranges. Therefore,
for analytic signals s(t ) of limited bandwidth, the maximum
VM frequency observed can be reduced by shifting the Fourier
𝜙M = 𝜙K = K 𝜙M
2𝜋 f spectrum of s(t ) by the so-called shift fs , toward negative
frequencies as follows:
where, 𝜙s is the winding flux obtained from the integration of
the winding voltage, LA is the air core inductance, IM is the mag- [s(t )] = s(t )e− j 𝜔s t (19)
netizing current at 1 p.u. voltage and K is the p.u. knee point
value [6] as shown in Figure 9. where the angular frequency is 𝜔s = 2𝜋 fs . Of particular interest
The characteristic drawn by Equation (16) is now also eval- is the case where the shift frequency is made equal to the car-
uated for the negative side and shifted horizontally dependent rier frequency fc of an AC waveform: fs = fc . In this case, the
on the desired loop width. A lookup table and interpolation complex envelope is obtained by Equation (19). The AC carrier
method is applied in this model. At first, a set of curves is eliminated, and transients are emulated efficiently as it is for
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6 of 16 ISLAM ET AL.

dynamic phasor signals. If the shift frequency is set to zero, then

the carrier is visible as it is the case in simulators of the EMT-
type. In frequency-adaptive simulation of transients (FAST),
the shift frequency appears as a variable simulation parame-
ter to support multi-scale modelling. Here analytic signals are
processed to allow for both the simulation of natural and enve-
lope waveforms as well as the smooth transition between them
[11–14]. Multi-scale modelling is also well suited for the rep-
resentation of transformers. As mentioned in earlier parts of
Section 2, the transformer saturation is modelled by insert-
ing saturable magnetizing branches. Considering that analytic
signals are marked by an underline, (9) can be rewritten as:

di(t ) FIGURE 10 Transformer model based on terminal duality

= L −1 [v(t )] (20)
dt
with, 2.4 Transformer model based on terminal
( ) duality principle [15, 16]
L11 M12
i = (i1 , i2 )T , L = , v = (E1 , E2 )T (21)
M21 L22 In this section, a transformer model based on the principle
of duality is presented. The equivalent circuit based on the
Inserting i(t ) = [i(t )]e j 𝜔s t according to Equation (19), Equa- principle of duality has a closer correlation with the physi-
tion (20) results in: cal interpretation of short circuit leakage reactances. Duality
approach takes into account the fact that transformer windings
d[i(t )] of different levels are basically concentric hollow cylinders. In
= e− j 𝜔s t (− j 𝜔s i(t ) + L −1 v(t )) (22)
dt practice, windings with lower rated voltages are closer to the
magnetic core and the windings with higher voltage ratings are
Applying the trapezoidal method of integration, in con- further from the core. In contrast, the star equivalent circuit,
tinuous time domain, input analytical signal u(t) and output popular amongst power system engineers, does not represent
analytical signal o(t) can be expressed as: the physical leakages of a practical transformer. Although, from
terminal point of view, it produces correct results. Such a rep-
do(t ) resentation may closely describe a primitive transformer with
= k𝜏 u(t ) (23)
dt loose magnetic coupling. The electric equivalent circuit of a
transformer based on duality principal can be achieved in
Here, 𝜏 is time-step size and k denotes the time-step counter.
three steps. Firstly, extract magnetic reluctance circuit, draw the
Comparing u(t)=o(t), o(t)=v(t), and L −1 = k from ref. [12]:
magnetic circuit’s graph and finally substitute reluctance with
inductance and MMF source with current source. Figure 10
i(k) = G v(k) + 𝜂(k) (24)
shows the terminal duality based transformer model of a three
winding core type transformer.
where, Ls120 and Ls230 are short circuit inductances between wind-
𝜏 ings 1–2 and 2–3 and M120 is the mutual coupling. In the TDM
G= 𝜏 (25) branches:
2L(1 + j 𝜔s )
2

⎛ 1 − j 𝜔s 𝜏 𝜏v(k − 1) ⎞ vb = Zb ⋅ ib (27)
𝜂(k) = e j 𝜔s t ⎜ 2
𝜏 i(k − 1) + 𝜏
⎟ (26)
⎜ 1 + j 𝜔s 2L(1 + j 𝜔s ) ⎟⎠
⎝ 2 2 Here, vb , ib are voltages and currents in TDM branch, and
Zb = 𝜔 ⋅ Lb is the TDM admittance matrix. To produce the
where G is a conductance matrix, and 𝜂 is a source vec-
admittance matrix seen from the primary side of the windings,
tor collecting information from time-step (k − 1). For fs = 0
incidence matrix (A) is formed which depicts the connections
Hz, G and 𝜂 are suitable for representing natural waveforms
among nodes and branches [15].
including the carrier as it is the case in simulators of the EMT-
type [9]. For fs = fc , the envelopes of the waveforms can be
represented very efficiently as also used in the phasor-type sim- vb = AT ⋅ vws , iws = A ⋅ ib , Yws = A ⋅ Zb ⋅ AT (28)
ulation. By changing the shift frequency, the simulator switches
between a detailed EMT solution with a small time-step (<50 Here, vws , iws are transformer secondary voltages, Yws is the
𝜇s) suitable for real-time simulation and phasor solution with a winding admittance matrix from secondary. Finally, the admit-
larger time-step. tance matrix seen from the primary nodes is formed by
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ISLAM ET AL. 7 of 16

incorporating ideal transformer influences [6]:

[ ] [ ]
inp [N ]−1 ⋅ Yws ⋅ [N ]−1 −[N ]−1 ⋅ Yws ⋅ [N ]−1
=
inp′ −[N ]−1 ⋅ Yws ⋅ [N ]−1 [N ]−1 ⋅ Yws ⋅ [N ]−1

[ ]
vnp
⋅
vnp′
(29)
⎡n1 0 … 0⎤
⎢0 n2 … 0⎥
where, [N ] = ⎢
… … … …⎥
⎢ ⎥
⎣0 0 … nn ⎦
Following the same process, inverse of inductance matrix [L]−1
can be calculated, and series connected [L]−1 and winding resis-
tance r are used to incorporate the model for electromagnetic
FIGURE 11 Faulted transformer model based on terminal duality
transient analysis. The current, voltages, and history terms are
represented by:

3 MAGNETIC EQUIVALENT CIRCUIT

i(t ) = [Geq ] ⋅ i(t ) + Ih (TOPOLOGY BASED)
( )−1
Δt 3.1 Unified magnetic equivalent circuit
Ih = [Geq ] ⋅ v(t − Δt ) + I + [L]−1 ⋅ [r] + (30) model [6, 17–19]
2
( )
Δt −1 So far, the proposed models either have used decoupled or
I − [L] ⋅ [r] ⋅ i(t − Δt )
2 only the same phase coupled windings and are based on the
electrical equivalent circuit of the transformer. The unified mag-
netic equivalent circuit (UMEC) model however is based on the
Δt Δt
where, [Geq ] = (I + [L]−1 ⋅ [r])−1 ⋅ ( [L −1 ]) magnetic equivalent circuit (and therefore on the core geome-
2 2
try) and also considers coupling between windings of different
phases in a three limb three-phase transformer. It is therefore
2.4.1 Faulted transformer model based on more accurate and realistic than conventional models, but also
terminal duality method requires more calculation steps in the simulation. The model
and derivation of the equations was introduced in ref. [17]. In
Analytical methods are used to calculate parameters of a termi- ref. [18], a simplification of the matrix operations has been
nal duality equivalent circuit for a faulted transformer. In this developed to increase performance. In ref. [19], authors pre-
analytical approach, in addition to the axial component of the sented a demonstration with UMEC based transformer model
flux, the effects of transverse component of the flux in asym- to simulate open phase conditions, DC Bias and internal fault
metric conditions are taken into account. It depicts magnetic cases in a real-time simulator.
circuit behaviour more accurately compared to star equivalent
models which are good for terminal condition representation
but lack in simulating real physical leakages. A leakage branch 3.1.1 Magnetic equivalent circuit of a three limb
is included in the model which enables it to predict the mag- three-phase transformer
netic inrush current and mutual couplings between branches.
Figure 11 shows the configuration of a faulted transformer Figure 12 shows the core flux paths and the unified magnetic
model based on terminal duality. Here, Ls12 is the short cir- equivalent circuit of a three limb three phase transformer. It
cuit inductance between two pieces of winding 1, Ls13 and considers the windings’ stray fluxes and fluxes between the
Ls23 are the short circuit inductances between winding 1 and 2, single-phase limbs. The P ′ s are the permeances of the magnetic
M12 , M23 , andM34 are the coupling between two branches. The circuit. As mentioned, geometrical parameters such as the wind-
faulted transformer model based on the terminal duality prin- ing numbers, as well as the permeance at every time-step need
ciple has been included in the RTDS [6] simulator. This model to be known. The permeance is a function of core geometry.
is suitable for better inrush current and internal fault analysis as
well as for a more detailed representation of mutual coupling 𝜇 0 𝜇R A
between branches. P= (31)
l
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 of 16 ISLAM ET AL.

FIGURE 13 Saturation curve [18]

Combining Equations (32) and (34) gives:

𝜙 = [M ][P][N ] (35)

where,

[M ] = [I ] − [P][A]([A]T [P][A])−1 [A]T (36)

The branch fluxes 𝜙 are related to the winding voltages.

Applying Faraday’s Law of magnetic flux, these values can be
calculated by numeric integration (here trapezoidal):

Δt
𝜙s (t ) = 𝜙s (t − Δt ) + (V (t ) + Vs (t − Δt )) (37)
2 s
FIGURE 12 Three limb three phase transformer [17] where s denotes the subset of magnetic branches on which
windings are mounted. Solving Equations (35) and (36) for
where 𝜇0 𝜇R is the core permeability taken from the B–H curve, i gives the standard trapezoidal discrete format of the trans-
and A and l are the core cross-section and magnetic circuit former equation:
branch length respectively. Geometry dependency is eliminated
is (t ) = [Yss ]Vs (t ) + Ihist (38)
by applying the normalized core concept introduced in ref. [17].
with the equivalent admittance matrix:
3.1.2 Model formulation Δt
[Yss ] = [Pss ]−1 [N ]−1 (39)
2 ss
The detailed model formation and corresponding equations in
this section can be found in ref. [18]. A brief summary of this and the history current injection vector:
modelling technique is discussed here. ( )
−1 Δt −1
First, the relationship between the flux in every branch and Ihist = [Pss ] [N ] Vs (t − Δt ) + 𝜙s (t − Δt ) (40)
2 ss
the magnetomotive force (MMF) is described by:

𝜙k = [Pk ]([N ]i − Θ) (32)

3.1.3 Saturation modelling
where 𝜙k is the flux vector, Pk is the diagonal permeance matrix,
i is the winding current vector, N is the diagonal matrix for the For this model, the saturation is represented by updating
number of windings, and Θ is the MMF branch vector. At each the permeance matrix [P] after every step. This is done by
magnetic circuit, the sum of the fluxes must be zero. evaluating:

[A]T 𝜙 = 0 (33) 𝜙i
Pi = (41)
Θi
Applying the branch-node connection matrix A to the vector
of the node MMF gives the branch MMF: from the pre-calculated (and stored on the processor) saturation
(flux vs. MMF) curve as shown in Figure 13. It is important to
[A]T Θnode = Θ (34) note that, the non linearity of the saturation curve in Figure 13
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ISLAM ET AL. 9 of 16

simulation. The geometry-based model can describe the flux

paths and sufficiently solve the non-linearity, and thus pro-
vide relatively accurate results compared with the finite element
tools.

4 FINITE ELEMENT SOLUTION

BASED TRANSFORMER MODEL [26, 27]

Recently, the real-time finite element (FE) transformer model

was also proposed and implemented on FPGA [26]. Com-
pared with other models, a finite element model can solve
the non-linearity more accurately and provide comprehen-
sive information such as the detailed distribution of magnetic
field, saturation, eddy current loss, and hysteresis loss. The
transmission line modelling (TLM) method is employed in
FEM to perform real-time electromagnetic transient simula-
tion for 2D analysis of single-phase transformer coupled to
external networks. TLM-FE method has been implemented
in parallel hardware architectures to achieve required com-
putational efficiency for real-time finite element simulation.
The real-time finite element transformer model starts by
solving the magnetic vector potential A. In a 2D prob-
lem, magnetic vector potential A, and the impressed current
FIGURE 14 High-fidelity MEC transformer model [21] density has only z− components and so the magneto-
dynamic problem is governed by the following Ampere’s
law,
may induce numerical oscillations. Special treatment should be
done to keep the solution stable.
𝜕A
∇ ⋅ (𝜐∇A) = 𝜎 − Jz (42)
𝜕t
3.2 High fidelity MEC transformer models
[20–22] where 𝜐 is the field-dependent reluctivity, 𝜎 is the conduc-
tivity, Jz is the impressed current density. using the Galerkin
The high-fidelity magnetic equivalent circuit-based (HF-MEC) finite element scheme [28] the magnetodynamic problem is
transformer model, which is capable of calculating the non- solved through following steps: the domain is discretized
linear saturation effects, hysteresis loss, and eddy current loss with triangular domain and nodes, elemental equations are
has been proposed and implemented on FPGA for real-time formed, global matrix is assembled and finally the system
execution [20, 21]. To obtain the detailed representation of of equation is solved. The classical weighted-integral equa-
the flux path, the developed HF-MEC model is shown in tion of element Ωe is found with applying natural boundary
Figure 14, where each coil has been divided into several sub conditions:
coils with a smaller number of turns. In the MEC network,
this sub coil is represented by a group consisting of differ- ( )
𝜕Ae 𝜕W e 𝜕Ae 𝜕W e
ent values of nonlinear permeances and magnetomotive forces 𝜐e + dxdy
∫ ∫Ωe 𝜕x 𝜕x 𝜕y 𝜕y
in the winding branches. A real-time Sen transformer has
been also implemented on FPGA using the detailed MEC (43)
model [22]. 𝜕Ae
The hysteresis in the transformer core is modelled using 𝜎 W e dxdy = JzeW e dxdy
∫ ∫Ωe 𝜕t ∫ ∫Ωe
Preisach theory and the eddy currents are incorporated using
a frequency-dependent network [23–25]. The Ampere’s law is
utilized for the MEC calculation and the Faraday’s law is used where Ae is the magnetic vector potential over Ωe , and W e
to obtain the induced voltage of the windings. The Newton– is the weighted function. According to Galerkin method,
Raphson iteration has been involved to handle the nonlinearity after further processing, a set of three equations with three
and increasing computational burden. The data parallelism and unknown vertex magnetic potential values are obtained. The
pipelining of the hardware blocks such as the matrix–matrix development of these equations can be found in ref. [26].
multiplication module and the sparse matrix solver module In this process, three equations with three unknown vertex
have been fully explored to optimize the latency for real-time magnetic potential values are obtained after performing the
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
10 of 16 ISLAM ET AL.

integral:

⎡b1 b1 + c1 c1 b1 b2 + c1 c2 b1 b3 + c1 c3 ⎤ ⎡A1 ⎤
𝜐e ⎢ ⎥⎢ ⎥
4Δe ⎢b1 b2 + c1 c2 b2 b2 + c2 c2 b2 b3 + c2 c3 ⎥ ⎢A2 ⎥
⎢b b + c c b2 b3 + c2 c3 b3 b3 + c3 c3 ⎥⎦ ⎢⎣A3 ⎥⎦
⎣ 1 3 13

⎡ 𝛿A1 ⎤ (44)
⎡2 1 1⎤ ⎢ 𝛿t ⎥ ⎡ ⎤
e 1
𝜎e Δe ⎢ ⎥ ⎢⎢ 𝛿A2 ⎥⎥ Jz ⎢ ⎥
+ = ⎢1⎥
12 ⎢⎢
1 2 1⎥
⎥ ⎢ 𝛿t ⎥ 3 ⎢ ⎥
⎣1 1 2⎦ ⎢ 𝛿A ⎥ ⎣1⎦
⎢ 3⎥
⎣ 𝛿t ⎦

Here, reluctivity 𝜐 e , conductivity 𝜎e , and current density Jze are

constant over finite element,Δe is the area of the triangular ele-
ment, ai , bi , ci (i = 1, 2, 3) are the coordinates of vertices. In a
2D problem, A is defined by B = Δ × A and in one element, B
and A are expressed as:
( )2 ( e )2
𝛿Ae 𝛿A 1
B2 = + = ((b b + c c )(A − A2 )2
𝛿x 𝛿y 4((Δe )2 1 2 1 2 1

+ b2 b3 + c2 c3 )(A2 − A3 )2 + b1 b3 + c1 c3 )(A1 − A3 )2 )
(45)

This nonlinear 𝜐 e − B 2 relation can be found from B–H curve

representation. B–H relation can be represented by the Preisach
hysteresis model [29]. The magnetization moment M and mag-
netic field H can be expressed as an integration over the
Preisach triangle:

M (t ) = 𝜇(𝛼, 𝛽)𝛾𝛼𝛽 H (t )d𝛼d𝛽

∫ ∫𝛼≥𝛽

= 𝜇(𝛼, 𝛽)H (t )d𝛼d𝛽 − 𝜇(𝛼, 𝛽)H (t )d𝛼d𝛽

∫ ∫S + ∫ ∫S −
(46)
Here, 𝜇(𝛼, 𝛽) is the Preisach distribution function that can
be obtained from experimental results. The B–H relation can
be found with Equation (47):

B = 𝜇0 (M + H ) (47)

The Presiach distribution function is obtained from experi-

mental results. The experimental results are first approximated
using a 2-D Cauchy distribution function, and after time dis- FIGURE 15 TLM technique applied in the FEM solution [27]
cretization, the nonlinear elemental equations are solved with
Newton–Raphson algorithm. Instead of assembling the ele-
mental equations to a global nonlinear system, the magnetic
vector potential has been solved using the transmission line allelism and deep data pipelining for real-time emulation
modelling (TLM) technique, which is a decentralized algo- on FPGA.
rithm in nature, and suitable for parallel processing. Figure 15 To interface the finite element transformer with exter-
represents different types of TLM models applied in the nal networks, an indirect field-circuit coupling scheme has
FEM solution. The admittance memory matrices were applied been proposed, which is very suitable for multi-rate sim-
to efficiently reduce the required TLM iterations, and all ulation. According to the Faraday’s law, the following
the hardware blocks were optimized to achieve massive par- equation can be used to calculate the induced winding
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ISLAM ET AL. 11 of 16

FIGURE 17 Circuit model for inrush current demonstration [30]

FIGURE 16 Equivalent FE transformer model coupled to external

networks [26] TABLE 2 Transformer data

Transformer data
voltage:
Configuration Yg − Yg

Nl 𝜕A Rating 100 MVA

U = rI + dS (48)
ΔS ∫S 𝜕t Base frequency 60 Hz
Leakage inductance 0.1 pu
where I is the winding current, r is the winding resistance, N No load losses 0.01 pu
is the number of turns, l is the average circumferential length
Primary voltage 13.8 kV
of each turn, S is the winding zone, and ΔS is the area of the
winding zone. Secondary voltage 230 kV
According to the partial differential chain rule, the above
equation can be rewritten as:
models, and UMEC models with core saturation on (satura-
Nl 𝜕A 𝜕ip Nl 𝜕A 𝜕is
U = rI + dS + dS (49) tion) and off (ideal) mode. The circuit [30] has been modified
ΔS ∫S 𝜕ip 𝜕t ΔS ∫S 𝜕is 𝜕t to show the steady state, inrush current and fault conditions.
Figure 17 shows the circuit diagram used to demonstrate the
Applying Equation (49) to the primary and secondary model behaviour. A three-phase voltage source with an inter-
windings respectively, the self and mutual inductance of the nal impedance of 0.1 Ω is connected to the transformer as
transformer can be extracted as: shown in Figure 17. The transformer data are provided in
Np lp 𝜕A Np lp 𝜕A Table 2.
Lp = dS , Mps = dS The transformer secondary is connected to an R–L load con-
ΔSp ∫S 𝜕ip ΔSp ∫S 𝜕is
p p sisting of 529 Ω and 0.45 H per phase. Saturation and hysteresis
(50) data are not applicable to all models. For including saturation
Ns ls 𝜕A Nl 𝜕A and hysteresis effects, saturation, magnetization, and core data
Msp = dS , Ls = s s dS
ΔSs ∫S 𝜕ip ΔSs ∫S 𝜕is are required for corresponding models.
s s

The finite element modelling technique is used to calculate

the self and mutual inductances, shown in Figure 16, to interface 5.1 Case study 1: steady-state analysis
with external networks, and these nonlinear inductance values
can be updated by the finite element computation for different Steady-state analysis is the simulation during normal operating
winding currents. conditions. Steady-state voltage and current mostly depend on
The results of the real-time models have been validated the source behaviour and types of load connected. The differ-
against the commercial FE software, and the finite element ent transformer models have been simulated with a three-phase
transformer models showed good performance in terms of source and constant R–L load. Saturation and hysteresis mode
providing detailed field-transient information and time-varying was enabled for piecewise linear and UMEC transformer model
total loss in real-time with saturation. The steady-state performance is almost identi-
cal for all models. There is a slight phase difference for models
with saturation ON.
5 MODEL DEMONSTRATION IN To illustrate the differences in steady-state performances
REAL-TIME SIMULATION of different models, Fourier transformations have been per-
formed. Figure 18(top) and 18(bottom) shows the steady-state
In this section, four case studies have been performed on a current and magnitude spectrum of steady-state FFT output,
real-time simulation platform to demonstrate the suitability of respectively. As can be seen from both Figures 18(top) and
some of the models studied above. Both steady-state and tran- 18(bottom), the magnitudes are almost the same for all the
sient behaviour have been observed using ideal and saturation models with a slight phase difference. Figure 18(bottom) shows
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
12 of 16 ISLAM ET AL.

FIGURE 19 Transient inrush current

TABLE 3 Current magnitude for different models under inrush

Transformer model Total current (kA)

Ideal 5.9973
Saturation 18.3796
UMEC without saturation 5.9674
UMEC with saturation 9.7351

FIGURE 18 Steady state analysis of the steady-state current (top) and the
magnitude spectrum of steady state FFT (bottom)

The UMEC model with saturation also shows a spike

in the current magnitude due to switching. The current
the fundamental component of steady-state current for differ- settles in approximate 10–12 cycles. Inrush current is a
ent models and they are very similar in magnitude. The other phenomenon related to magnetic saturation and the trans-
frequency components were insignificant and hence not shown. former models without saturation properties do not reflect
these transient behaviours.

5.2 Case study 2: inrush current

5.3 Case study 3: fault simulation
Inrush current is produced due to switching effect, i.e. while
transformer starts up. This is a very important phenomenon Fault analysis is another crucial aspect of transient analysis of
as the rating of the system components depends on this type transformers. Fault current can be several times the steady-state
of analysis. To simulate the inrush effect, at first, the system is current, and protection systems must be designed according to
operated under steady-state conditions. A three-phase breaker such analysis. Here, a line-to-line (L-L) fault has been placed
is used to disconnect and then reconnect the circuit to pro- between lines A and B at the load side of the circuit. The fault
duce the inrush current. Theoretically, maximum inrush current has been configured with 0.01 Ω resistance. The fault occurs at
is produced when the breaker is closed on the zero crossing 0.04 th second of the simulation and lasts for two cycles. After
of the flux. For the saturation-enabled models, the flux mag- two cycles, the fault is cleared, and regular operation continues.
nitude that produces a maximum inrush current can also be The linear and UMEC models are tested with both saturation
observed. ON and OFF configuration.
As shown in Figure 19, the ideal transformer model and Figure 20 shows fault current for L-L fault between Line A to
UMEC model without saturation show no impact of inrush B, and a magnified view of the peak points for different models.
effects, however, the other two models (saturation model and We can observe that all models have slightly different fault cur-
UMEC model with saturation) show significant inrush current. rent output with the ideal model showing the minimum. Table 4
Table 3 summarizes the current magnitude for different mod- summarizes the peak fault current magnitude for different mod-
els during an inrush scenario. The saturation model shows a els. The magnitude does not vary much with saturation on or off
big inrush current of almost three times of its steady-state mode but from the zoomed in view, there is a phase difference
value. observable when saturation is on.
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ISLAM ET AL. 13 of 16

FIGURE 20 Line-to-line (A–B) fault current with magnified view of the

peak current

TABLE 4 Peak fault current for different models

Peak fault
Transformer model current (kA)

Ideal 13.00
Saturation 13.20
UMEC without saturation 13.23
UMEC with saturation 13.32

5.4 Case study 4: multi-scale simulation of

the open circuit

The case of a three-phase transformer consisting of three

single-phase units connected in a Yg –Yg configuration with a
secondary-side open circuit is considered here. The circuit is FIGURE 21 Open-circuit multi-scale simulation; envelope and natural
waveforms of primary-side (top) and details of natural waveform during open
as shown in Figure 17. The transformer parameters are from circuit around 1.8 s (bottom)
Table 2. The piecewise linear saturation curve is derived from
Figure 3.
At time t =1.7 s, the secondary side of the transformer turns ate model selection should be based mainly on the intended
into a state of open circuit for a period of 0.3 s. From 1.7 to study purpose and the availability of input data required for
2.0 s, the small time-step size of 50 𝜇s is used and the shift specific models. Some models have been designed to observe
frequency is set to 𝜔s = 0 while natural waveforms are being specific features only and may not be helpful for simulat-
tracked. For other times, the step size is extended to 10 ms and ing other features. The electric equivalent circuit models may
the shift frequency is 𝜔s = 2𝜋 fc for tracking the envelope of be used for a basic analysis of the system, which does not
the AC waveform. require detailed magnetic or geometric analysis. For such cases,
The simulation result for the line current on the primary side depending on available data, either piecewise linear or lookup
over the whole interval of 3.5 s is shown in Figure 21(top). and interpolation model may be used. For high-frequency
As shown in Figure 21(bottom), the envelope of the current AC transient analysis, a multi-scale transformer model would
is being tracked during steady state. From 1.7 to 2.0 s during be suitable. Faulted terminal duality based model is ideal for
the open circuit, the details of the transients and saturation are detailed fault analysis and performance analysis of transformers
of interest and shown as natural waveforms. Thus, multi-scale before installing in a real system. UMEC and high fidelity MEC
modelling is able to adapt to the situation observed and has transformer models would be suitable for detailed magnetic
shown to offer good speed-ups in simulation, as demonstrated behaviour analysis. However, these models are computation-
in studies [31]. ally more demanding and require detailed core geometry and
magnetization data. Finite element based models also require
high computational resources and are ideal for detailed mag-
6 APPLICATION GUIDELINE netic field distribution and saturation characteristics study. The
discussed models however are not suitable for producing tran-
The simulation results show that, models with different fidelities sient surges along the windings which would require layer by
produce results with different accuracy levels. The appropri- layer modelling. Apart from the case studies provided in this
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
14 of 16 ISLAM ET AL.

TABLE 5 Application guideline

Observable
Transformer model Inputs needed features Advantages Limitations

Piecewise linear Primary and secondary side Saturation effects, Useful for reduced Oversimplified model,
saturation model voltage, Transformer rating, inrush current computational unable to represent
[4–6] leakage inductance, no load compensation, complexity, can be magnetically
loss, copper losses, relay protection used with saturation coupled transformer
magnetizing current, air core strategies curve details
reactance, leaky integrator without geometric
time constant, knee voltage, parameters
loop width as percentage of
magnetizing current, eddy
current loss
Lookup and Primary and secondary voltage, Saturation Effects, Easy to use and set up, Not suitable for fast
interpolation Transformer rating, no load Inrush current can be used for a transient analysis
model [6, 8] loss, base frequency quick simulation
Multi-scale Primary and secondary side High frequency AC Suitable for electro Time-step size and
transformer model voltage, transformer MVA, transients, fault mechanical shift frequency
[11–14] leakage inductance, saturation studies oscillation in power adjustments, and
and magnetization data electric system modelling of
non-linearity can be
challenging
Terminal duality Primary and secondary voltage, Magnetic inrush Can be used to Detailed core
based transformer rating, leakage inductance, current, hysteresis, accurately depict information needed
model [15, 16] mutual inductance, core data Fault simulation magnetic circuit
behaviour during
fault
Unified magnetic Primary and secondary side Over excitation Suitable for observing High computational
equivalent circuit voltage, transformer MVA responses, inrush detailed magnetizing effort
model [6, 17–19] Leakage inductance,core and current characteristics with
copper loss, base frequency, geometrical
leakage reactance, copper loss, configuration effects
no load loss, magnetizing
current, winding-limb length
and area
Detailed MEC based Primary and secondary voltage, Hysteresis analysis, Ideal for detailed Complex architecture,
real-time rating, core loss, leakage Eddy current analysis of magnetic parallelism, and
transformer model inductance behaviour characteristics pipelining required
[20–22]
Real-time finite Primary and secondary voltage, Detailed distribution Suitable for magnetic High computational
element rating, core loss, leakage of magnetic field, characteristics effort, interfacing
transformer model inductance saturation,eddy analysis and parallel with external circuit
[26, 27] current loss, and processing capable is challenging
hysteresis loss system

paper, readers can also find some practical implementation and real-time simulation, different modelling techniques are suitable.
real-time simulation examples in corresponding papers. For In this paper, established models for the real-time simula-
example, experimental validation of the ideal transformer model tion of transformers have been reviewed, summarized, and
[32], Lookup and interpolation model [6, 8], UMEC [6, 18, 19], documented. Case studies have been presented to provide
Real-time finite element transformer models [26] have been dis- insight into the implementation and functioning of different
cussed in the corresponding references. A summary of these key models. Finally, an application guideline has been developed
information for selecting the appropriate model is provided in so that readers can select appropriate transformer real-time
Table 5. models for their intended study. The paper discussed dif-
ferent transformer modelling techniques, many of which are
applied in different real-time digital simulators including the
7 CONCLUSIONS commercial ones. The paper is compiled with the objective
that the reader can get a comparative idea of different mod-
Transformers are one of the most critical and expensive com- els and then choose the appropriate one for the intended
ponents in the power system. For different kinds of analysis in study.
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ISLAM ET AL. 15 of 16

AUTHOR CONTRIBUTIONS 7. Dommel, H.: Electro-Magnetic Transients Program Theory Book. Bon-
Md Maidul Islam: Conceptualization, formal analysis, investiga- neville Power Administration, Portland, OR (1986)
8. Peacock, B., Steurer, M., Langston, J., Baldwin, T., Henry, S.: Estimation of
tion, methodology, supervision, validation, visualization, writing
parameters for modeling of saturation effects for a transformer model for
- original draft, writing - review and editing. Matthias Musil: For- a real-time digital simulator. In: 2009 41st Southeastern Symposium On
mal analysis, validation, writing - original draft. Jamal Ahmed System Theory, pp. 306–310. IEEE, Piscataway, NJ (2009)
Shohan: Formal analysis, investigation, methodology. Omar 9. Dommel, H.: Digital computer solution of electromagnetic transients
Faruque: Formal analysis, supervision, writing - original draft, in single- and multiphase networks. IEEE Trans. Power Apparatus
PAS-88(4), 388–399 (1969)
writing - review and editing. Georg Lauss: Formal analysis,
10. Talukdar, S., Bailey, J.: Hysteresis models for system studies. IEEE Trans.
methodology, supervision, validation, visualization, writing - Power Apparatus 95(4), 1429–1434 (1976)
original draft, writing - review and editing. Ali Dehkordi: For- 11. Strunz, K., Shintaku, R., Gao, F.: Frequency-adaptive network modeling
mal analysis, methodology, visualization. Paul Forsyth: Formal for integrative simulation of natural and envelope waveforms in power sys-
analysis, methodology, visualization. Panos Kotsampopoulos: tems and circuits. IEEE Trans. Circuits Syst. I Regul. Pap. 53, 2788–2803
(2006)
Formal analysis, methodology, validation, writing - original
12. Gao, F., Strunz, K.: Frequency-adaptive power system modeling for mul-
draft, writing - review and editing. Kai Strunz: Formal analy- tiscale simulation of transients. IEEE Trans. Power Syst. 24, 561–571
sis, methodology, supervision, writing - original draft, writing (2009)
- review and editing. Zihui Li: Formal analysis, investigation, 13. Rupasinghe, J., Filizadeh, S., Gole, A., Strunz, K.: Multi-rate co-simulation
methodology. Yi Zhang: Formal analysis, investigation, method- of power system transients using dynamic phasor and EMT solvers. J. Eng.
2020, 854–862 (2020)
ology, visualization. Peng Liu: Formal analysis, investigation,
14. Fan, S., Ding, H.: Time domain transformation method for accelerating
methodology, visualization. EMTP simulation of power system dynamics. IEEE Trans. Power Syst.
27, 1778–1787 (2012)
ACKNOWLEDGEMENT 15. Alvarez-Marino, C., Leon, F., Lopez-Fernandez, X.: Equivalent circuit
This work has received funding from the European Union’s for the leakage inductance of multiwinding transformers: unification of
terminal and duality models. IEEE Trans. Power Delivery 27, 353–361
Horizon 2020 research and innovation program under grant
(2012)
agreement No. 870620 in the ERIGrid 2.0 project. 16. New feature: faulted transformer model, https://www.rtds.com/faulted-
transformer-model. Accessed 23 December 2021
CONFLICT OF INTEREST 17. Enright, W., Nayak, O., Irwin, G., Arrillaga, J.: An electromagnetic tran-
Prof. Kai Strunz may be considered both author of this paper sients model of multi-limb transformers using normalized core concept.
Paper presented at the International Conference On Power System
and editor related to IET at the same time.
Transients, Seattle, 22–26 June 1997
18. Zhang, Y., Maguire, T., Forsyth, P.: UMEC transformer model for the real
DATA AVAILABILITY STATEMENT time digital simulator. Paper presented at the International Conference On
Data are available on request. Power Systems Transients, Montreal, 19–23 June 2005
19. Liang, Y., Zhang, Y., Dehkordi, A.: Key aspects in modelling transform-
ers for real time simulation: open phase, DC bias and internal faults. In:
ORCID
2019 IEEE 8th International Conference On Advanced Power System
Md Maidul Islam https://orcid.org/0000-0002-6245-3609 Automation And Protection (APAP), pp. 1891–1895. IEEE, Piscataway,
Md Jamal Ahmed Shohan https://orcid.org/0000-0002-7563- NJ (2019)
3086 20. Liu, J., Dinavahi, V.: A real-time nonlinear hysteretic power transformer
Georg Lauss https://orcid.org/0000-0002-8397-2213 transient model on FPGA. IEEE Trans. Ind. Electron. 61, 3587–3597
(2014)
Kai Strunz https://orcid.org/0000-0002-2043-4549
21. Liu, J., Dinavahi, V.: Detailed magnetic equivalent circuit based real-
time nonlinear power transformer model on FPGA for electro-
REFERENCES magnetic transient studies. IEEE Trans. Ind. Electron. 63, 1191–
1. Omar Faruque, M.D., Strasser, T., Lauss, G., et al.: Real-time simulation 1202 (2016)
technologies for power systems design, testing, and analysis. IEEE Power 22. Liu, J., Dinavahi, V.: Nonlinear magnetic equivalent circuit-based real-
Energy Technol. Syst. J. 2(2), 63–73 (2015) time sen transformer electromagnetic transient model on FPGA for HIL
2. Kotsampopoulos, P., Lagos, D., Hatziargyriou, N., Faruque, M., Lauss, G., emulation. IEEE Trans. Power Delivery 31, 2483–2493 (2016)
Nzimako, O., Forsyth, P., Steurer, M., Ponci, F., Monti, A., Dinavahi, V., 23. Chen, Y., Dinavahi, V.: FPGA-based real-time EMTP. IEEE Trans. Power
Strunz, K.: A benchmark system for hardware-in-the-loop testing of dis- Delivery 24, 892–902 (2009)
tributed energy resources. IEEE Power Energy Technol. Syst. J. 5, 94–103 24. Chen, Y., Dinavahi, V.: An iterative real-time nonlinear electromagnetic
(2018) transient solver on FPGA. IEEE Trans. Ind. Electron. 58, 2547–2555
3. Guillaud, X., Faruque, M., Teninge, A., Hariri, A., Vanfretti, L., Paolone, (2011)
M., Dinavahi, V., Mitra, P., Lauss, G., Dufour, C.: Others applications 25. Chen, Y., Dinavahi, V.: Hardware emulation building blocks for real-time
of real-time simulation technologies in power and energy systems. IEEE simulation of large-scale power grids. IEEE Trans. Ind. Inf. 10, 373–381
Power Energy Technol. Syst. J. 2, 103–115 (2015) (2014)
4. Ke, X., Ionutiu, R.: Modeling three-phase saturable transformers for real- 26. Liu, P., Dinavahi, V.: Real-time finite-element simulation of electromag-
time simulation. In: IECON-th Annual Conference of the IEEE Industrial netic transients of transformer on FPGA. IEEE Trans. Power Delivery
Electronics Society, pp. 3783–3789. IEEE, Piscataway, NJ (2014) 33, 1991–2001 (2018)
5. Marti, J., Linares, L., Dommel, H.: Current transformers and coupling- 27. Liu, P., Dinavahi, V.: Finite-difference relaxation for parallel computation
capacitor voltage transformers in real-time simulations. IEEE Trans. of ionized field of HVDC lines. IEEE Trans. Power Delivery 33, 119–129
Power Delivery 12, 164–168 (1997) (2018)
6. RTDS manual: RSCAD Documentation (2003) https://www.rtds.com/. 28. Zienkiewicz, O., Taylor, R., Zhu, J.: The Finite Element Method: Its Basis
Accessed 10 October 2021 and Fundamentals. Elsevier, New York (2005)
 20513305, 2023, 1, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12221 by CochraneChina, Wiley Online Library on [31/08/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
16 of 16 ISLAM ET AL.

29. Mayergoyz, I.: Mathematical models of hysteresis and their applications.

Academic Press, San Diego, CA (2003) How to cite this article: Islam, M.M., Musil, M.,
30. Brunke, J., Frohlich, K.: Elimination of transformer inrush currents
Shohan, M.A., Faruque, M.O., Lauss, G., Dehkordi, A.,
by controlled switching II. application and performance considerations.
IEEE Trans. Power Delivery 16, 281–285 (2001) Forsyth, P., Kotsampopoulos, P., Strunz, K., Li, Z.,
31. Xia, Y., Strunz, K.: Multi-scale induction machine model in the phase Zhang, Y., Liu, P.: A review of modelling techniques of
domain with constant inner impedance. IEEE Trans. Power Syst. 35, power transformers for digital real-time simulation. J.
2120–2132 (2019) Eng. 2023, 1–16 (2023).
32. Meghwani, A., Srivastava, S., Srivastava, A.: Development of real-time dis-
https://doi.org/10.1049/tje2.12221
tribution system testbed using co-simulation. In: 2020 21st National Power
Systems Conference (NPSC), pp. 1–6. IEEE, Piscataway, NJ (2020)