IEEE Power System Relaying and Control (PSRC) Committee Working Group I26

Mathematical Models for Current, Voltage, and Capacitively Coupled Voltage Transformers
IEEE Power System Relaying and Control (PSRC) Committee

Working Group I26
November 2023
Mathematical Models for Current, Voltage, and

Capacitively Coupled Voltage Transformers
Chair: Michael Meisinger
Vice Chair: Steve Turner
Secretary: Amir Makki
Members and Contributors
Alejandro Avendano
Yue Chen
Ritwik Chowdhury
Normann Fischer
Robert Frye
Juan Gers
Ljubomir Kojovic
Yuan Liao
Federico Lopez
Peter McLaren
Sakis Meliopoulos
James Niemira
Dean Ouellette
Athula Rajapakse
Demetrios Tziouvaras
Jim van de Ligt
Jack Wilson
i
ACKNOWLEDGMENTS
Working Group I26 gratefully acknowledges Power System Relaying Committee (PSRC)
Working Group C5 that authored and published the paper “Mathematical models for
current, voltage, and coupling capacitor voltage transformers” [1] in January 2000. The
paper reviewed and presented several mathematical models of instrument transformers.
This report significantly reuses the content from the paper. This report revises the
contribution for improved technical accuracy and clarity and builds upon the paper to
capture the more recent developments on this topic.
ABSTRACT
This report reviews and presents several mathematical models of instrument

transformers, including those used to represent the nonlinear magnetic core of instrument
transformers. The transient response of the instrument transformer is compared to test
results recorded in the laboratory. The report provides practical guidelines as to which of
the physical elements of instrument transformers are important to model during transient
studies and which elements may be ignored without sacrificing the accuracy of the
simulation results. The Electromagnetic Transients Program (EMTP) data files used to
generate the models are also provided in an appendix to help new EMTP users model
instrument transformers for evaluation of high-speed protective relaying systems.
KEYWORDS
capacitive, CCVT, coupling capacitor, CT, current transformer, CVT, EMTP, simulation,
transfer function, VT
ii
CONTENTS
1. Introduction .............................................................................................................. 1
2. Mathematical Models of Magnetic Core Representation ............................................ 2
2.1 EMTP and ATP Inductor Models ...................................................................... 2
2.2 EMTDC Model .................................................................................................. 4
2.3 Seetee Model ................................................................................................... 6
2.4 Jiles-Atherton Model ......................................................................................... 8
3. CT Model ............................................................................................................ 12
3.1 Equivalent Circuit............................................................................................ 12
3.2 Current Transformer Model Data .................................................................... 12
3.3 CT Simulation Results .................................................................................... 13
4. CCVT Model ........................................................................................................... 15
5. VT Model ............................................................................................................ 22
6. Merging units and Instrumentation Error Correction ................................................ 23
7. Conclusions ............................................................................................................ 24
8. References ............................................................................................................ 25
APPENDIX A EMTP Model Setup for CT Saturation .................................................. 28
A.1 CT and Secondary Burden Data ..................................................................... 28
A.2 CT Saturation Data ......................................................................................... 28
A.3 EMTP Subroutine CONVERT ......................................................................... 28
A.4 EMTP Output from CONVERT Subroutine ..................................................... 29
A.5 EMTP Subroutine HYSDAT ............................................................................ 29
A.6 EMTP Output from Hysdat Subroutine ............................................................ 30
APPENDIX B Test system Parameters for Jiles-Atherton Model ................................ 32
B.1 Test system parameters used for estimation algorithm ................................... 32
B.2 Test system parameters used for Validation ................................................... 33
APPENDIX C Modeling CT Remanence .................................................................... 35
APPENDIX D Recent work in modeling instrument transformers ................................ 39
D.1 Summary ........................................................................................................ 39
D.2 Detailed methods from selected papers .......................................................... 42
iii
THIS PAGE LEFT BLANK INTENTIONALLY
iv
1. Introduction
Electric power systems are subjected to many types of disturbances such as lightning,
faults, or routine operations such as line energization and de-energization, opening of
disconnects, and switching of inductive or capacitive loads. When a sudden change
associated with such disturbances occurs, such as a fault, a redistribution of the electric
and magnetic energy stored in the capacitive and inductive elements of the network
occurs due to the dynamics of the power system. This redistribution of energy cannot
occur instantaneously, and the power system goes through a transient state before it
reaches a new steady state.
During the first few cycles following a power system fault, high-speed protective relays
are expected to make a correct decision as to the presence and location of the fault in
order to preserve system stability and to minimize the extent of equipment damage. The
majority of protective relays make their decisions based on fundamental frequency
(e.g., 50 Hz or 60 Hz) voltage and current signals. However, it is precisely at this moment
that the voltage and current signals can have fault induced transients with frequencies
above and below the fundamental frequency.
The dynamic performance of high-speed protective relays depends, to a large extent, on
the signals reproduced by the instrument transformers, and these signals depend on the
overall transient response of the instrument transformers and the type of transients
generated by the power system.
The transient performance of current transformers (CTs) is influenced by a number of
factors, a significant one being the exponentially decaying dc component of the primary
current. Its presence influences the build-up of core flux, a phenomenon which can cause
saturation, and subsequently results in errors in the magnitude and phase angle of the
measured signals. The core flux is composed of an alternating and a unidirectional
component corresponding to the ac and dc components of the primary fault current,
respectively. The transient flux generated by the dc component of the primary fault current
can be quite large compared to the one created by the ac component. The CT core may
also retain an unknown amount of remanent flux, because of the ferromagnetic character
and whether or not the CT has anti-remanence air gaps. This remanent flux either aids
or opposes the build-up of core flux and could contribute to CT saturation, especially
during subsequent faults in the power system such as high-speed autoreclosing into a
permanent fault, depending on the relative polarities of the primary dc component and the
remanent flux. Furthermore, after primary fault interruption, the CT could still produce a
unipolar decaying subsidence current due to the release of the stored magnetic energy.
The transient response of magnetic voltage transformers (VTs) and capacitively coupled
voltage transformers (CCVTs) depends on several distinct phenomena taking place in the
primary network such as, sudden decrease of voltage at the transformer terminals due to
a fault or sudden overvoltages on the unfaulted phases caused during line to ground faults
on the network. Sudden decrease of voltage at the primary terminals could generate
internal oscillations in the windings of magnetic VTs, which creates a high frequency on
1
the secondary side. These high-frequency oscillations are typically damped within 15 to
20 milliseconds. However, in the case of CCVTs, energy stored in the capacitive and
inductive elements of the device generate transients with a low frequency and an
aperiodic characteristic, which may last up to 90 milliseconds [2]. Sudden increase of
voltage at the primary terminals of magnetic VTs could cause saturation of the magnetic
core.
The transient errors produced by the instrument transformers can have a major impact
on the dependability and security of protective relays and cause misoperations, delayed
operations, or failures to operate. Fault induced transients and transients generated by
the measuring transformers make the study of protective system performance for high-
voltage (HV) and extra-high voltage (EHV) networks purely based on fundamental
frequency incomplete.
The use of an electromagnetic transients program (EMTP), to adequately model the
different power system components and to generate transient data to assess the
performance of modern high-speed protective relays, has become essential.
2. Mathematical Models of Magnetic Core Representation

Models for instrument transformers are especially important for studying CT saturation,
ferroresonance phenomena, harmonics, and subharmonics and their effect on the
performance of protective relaying. The major nonlinear effects in iron cores are
saturation, eddy currents, and hysteresis. The predominant effect in CTs is saturation,
whereas the predominant effects in CCVTs and VTs are saturation and ferroresonance.
The working group reviewed several iron-core models described in the literature and are
discussed in this section.
2.1 EMTP and ATP Inductor Models

The EMTP and Alternative Transients Program (ATP) support two classes of nonlinear
elements, a true nonlinear model (Type-93), and two pseudo-nonlinear models (Type-96
and Type-98) [3]. In the true nonlinear model, the nonlinearity of the element is explicitly
defined as a nonlinear function, i.e., the flux linkage  is a function of the current
(  = f (i ) ). The EMTP and ATP then solve the combination of nonlinear equations and
an appropriate system equivalent at each time step using a Newton-Raphson iterative
procedure.
In the pseudo-nonlinear model, the nonlinearity is defined as a number of piece-wise
linear segments. Such linear segments are represented by the program with a resistor in
parallel with an appropriate current source. In the particular case of a nonlinear inductor,
the flux is monitored at each time-step in order to determine which linear segment should
be used to compute the inductance at that time step. Note that this methodology does not
model the true nonlinearity, since the program relies on previous time-step results to
decide on what segment to operate next. The EMTP and ATP change segments only
after they have operated illegally outside the range of the current segment. It is therefore
2
important to use a small time-step during the simulation so that the operating point moves
up and down the nonlinearity in small increments.
EMTP and ATP support the following two additional magnetic saturation routines:
• Subroutines CONVERT (in EMTP) and SATURATION (in ATP) are designed to
convert rms v − i saturation curve data into peak  − i data with the hysteresis loop
ignored.
• Subroutines HYSDAT (in EMTP) and HYSTERESIS (in ATP) are designed to provide
the hysteresis loop data required by the Type-96 pseudo-nonlinear reactor model.
This model uses data-points from the CT secondary excitation curve, Vrms-Irms, as inputs
to subroutine CONVERT to obtain datapoints of peak flux,  , vs. peak current i. The
result is a piece-wise linear model because a small finite number of datapoints are used,
usually 10 or less.
The peak  − i data are computed from the rms saturation curve voltage and current
values [4]. The conversion of rms voltage values to flux is only a re-scaling procedure.
For each linear segment in the  − i curve,
2Vk
k = (1)

where  is the angular frequency.
For the first linear segment in the  − i curve,
I 1 = I 1− rms 2 (2)
Note that Equations (1) and (2) are only valid for a single frequency sinusoidal waveform.
Assuming that  k ( ) =  k sin  , for the following segments ( k  2 ), the peak current is
obtained by evaluating Ik − rms for each segment k , using the following Equation (3) that
provides a parabolic estimate, as given in [4].
1 2  /2
2  k sin   sin  −  1  k sin  −  k −1
I 2
k − rms = ( ( ) 2 d +  ( I 1 + k
) 2 d +...+  ( I k −1 + ) 2 d ) (3)
 0
L1 1
L2 k −1
Lk
In (3) only the last segment Lk is unknown. Equation (3) can be rewritten in the form
a lk Yk + blk Yk clk = 0
2
+ (4)
1
with coefficients a lk , blk , and clk known, and Yk = to be computed.
Lk
3
Y k can be solved from (4) and it must be positive. The peak current i lk is computed from
i k = i k −1 + Yk ( k −  k −1 ) (5)
One point from the  − i curve is inputted in the subroutine HYSDAT that generates the
major hysteresis loop data. This routine contains predefined trajectories in the  − i plane
to decide which path to follow when the flux increases or decreases.
A resistance in parallel with the inductor is used to represent core losses due to eddy
currents. Other elements also represented are the primary and secondary resistances,
the leakage reactances, and the secondary burden resistance and reactance.
These EMTP/ATP models allow the user to represent the effects of remanent flux left in
the CT following primary current interruption. They can also be easily integrated directly
into EMTP/ATP power system studies. Finally, because the model parameters are
derived from CT test data, a degree of validity is implied.
The main disadvantage is the need to preprocess the datapoints from the CT excitation
curve to generate the data for the Type-96 hysteretic reactor model.
2.2 EMTDC Model

This model represents the magnetizing branch of the transformer equivalent circuit as a
non-linear inductor in parallel with a nonlinear resistor [5], [6]. The current in this branch
has two components—the magnetizing current (im) and the hysteresis current (ic). The
magnetizing current is described by a power series equation. The hysteresis current is
described by an exponential equation using the Steinmetz coefficient of the core material.
In a variant of this model, terms are added to ic to represent eddy current and other effects.
The coefficients of the power series for im are derived from the saturation curves for the
transformer by an iterative calculation that produces a curve relating instantaneous flux
and current values.
The resulting equations produce a smooth and continuous B-H loop representation similar
to EMTP/ATP. This model also requires preliminary calculations using the CT saturation
curve data to determine the power series coefficients.
Figure 1 shows the equivalent circuit for the magnetizing branch of the transformer
consisting of a nonlinear inductor in parallel with a nonlinear resistor.
4
i
i
im ic
V
Vc
a b
Figure 1: Equivalent Circuit and Current Source for the Magnetizing Branch.
The majority of electromagnetic transient programs use the Dommel algorithm [7], where
the inductances and capacitances are represented by a conductance in parallel with a
current source. Nonlinear elements are, however, represented by piece-wise linear
conductances, which has the disadvantage of requiring re-evaluation of the overall
system conductance matrix at any time step when the solution calls for a change from
one section to the next in the piece-wise linear conductance. The re-evaluated
conductance matrix has to be inverted, adding yet further computational penalties.
Avoiding matrix inversion is a worthwhile practice. In EMTDC the nonlinearity of both
elements in Figure 1a is combined into a voltage/current relationship as shown in Figure
1b. The eddy current loss resistor is constant for the frequency range of interest and is
not included in the current source. Therefore, the current source contains the magnetizing
current component, im, and a hysteresis component, ic.
The magnetizing current component, im, is related to the core flux  by the equation
aj
i m =  k j (6)
where kj and aj are both positive for all j [5]. Typically, three terms of the power series
give a very good fit to the measured curve over a wide range of values. The summation
is performed over the fundamental period.
The hysteresis component, ic, is of the form
ic = k h v  −1 (7)
where kh is a constant and  is the Steinmetz coefficient for the material. Constant kh is,
in general, frequency dependent but has been set for these studies to a value appropriate
for the rated frequency (50 Hz or 60 Hz). The total current source is therefore given by
i = im + ic (8)
The voltage v across the current source i is computed from the flux  using the number
of secondary turns N2 as
5
𝑑𝜓
𝑣 = 𝑁2 (9)
𝑑𝑡
It is also convenient to express im in terms the reluctance R and N2 as
R
im =  (10)
N2
R
And can, therefore, be put in the form of (11) that performs the summation over the
N2
fundamental period.
R a
=  k j j −1 (11)
N2
In order to use (6) and (7) it is necessary to know v and . The value of the current source,
i , injected into the network will in turn affect the value of v, depending on the network
source impedance as seen from the terminals of this current source and the magnitude
of i in comparison to the other currents flowing in the network. Voltage v and current i
cannot be calculated independently of each other, and an exact solution will require an
iterative technique which adjusts v until the calculated value of i injected into the network,
along with currents from all other sources, results in the same value of v. An iterative
technique slows the simulation and, in the case of real-time digital simulators, other
algorithms have to be used to reduce the time of the simulation to meet the constraints of
the specified time-step. [6], [8].
2.3 Seetee Model
In this model, the incremental permeability of the core material,  0  r = dB / dH , is

obtained by taking the derivative of the Frolich equation [9]. The permeability is calculated
at each time step to vary the nonlinear magnetizing inductance, Lm, in response to the
core flux density. When the core flux density reaches a value that causes the  r = 1 , the
permeability is fixed at this level until the flux density is again reduced. The result is a
smooth single-valued anhysteretic curve relating the core flux density, B, to the
magnetizing force, H.
Although this model does not simulate a multi-valued B-H loop, comparison with full scale
test data on saturating CTs have shown excellent agreement in the waveforms produced
[10]. This suggests that the multi-valued aspect of the hysteresis loop may have only a
minor influence on the distortion of CT secondary currents.
Only two datapoints on the core are needed for the Frolich equation coefficients. These
are,  m , the maximum core relative permeability, and Bsat , the saturation flux density.
The nonlinear inductance, Lm, represents the varying magnetizing inductance of the CT
core during large current excursions. Lm is calculated from the relation [9]:
6
1
Lm =  0  r N 2 Ac
2
(12)
Lc
Where:  r = relative maximum incremental permeability of the core material. For silicon
steel it ranges from 2000 to 5000.
 0 = permeability of free space 4E-7 H/m.

N 2 = number of CT secondary turns.
Ac = cross sectional area of the CT core in m2.
Lc = the mean length of the flux path in m.
The variable core permeability is the slope of the B/H curve at any specific point:
dB
 0 r = (13)
dH
The CT model needs a defined relationship between B and H so that the permeability can
be calculated. The Frolich equation can be used. One form of the equation is:
H
B= (14)
c+bH
where b and c are constants determined from the core material. The core permeability
is expressed by:
dB (1 − b B )
2
 0 r = = (15)
dH c
When the core flux density reaches the saturation flux density, Bsat, the relative
permeability of the core approaches unity. This provides one condition for defining how
the constants b and c are related. For silicon steel, Bsat is about 1.8 Tesla.
The maximum slope of the B/H curve occurs at B=0 where the curve crosses the H axis.
This is the maximum permeability of the core material, m, and is a known constant.
Therefore, setting B=0 in (15) and solving for c :
1
c= (16)
 0 m
Using this value of c and letting B=Bsat, the constant b is obtained by:
1−1/ m
b= (17)
Bsat
7
Using these constants, the Frolich equation defines the B/H curve between B=0 and
B=Bsat. The model disregards the equation for flux densities greater than Bsat and sets
the permeability to  0 .
The model is based on several assumptions listed below. Due to extremely large
excursions of the core flux, it turns out these assumptions have a negligible effect on the
accuracy of the results. The assumptions of the model are:
• The B/H curve is assumed to be single valued. This is the same as assuming that
the B/H curve has no area.
• The maximum permeability of the core occurs at B=0. For most magnetic
materials, the permeability is low at low values of H, rises quickly to a maximum
as H increases, and then gradually decreases at large values of H approaching
that of free space at extreme values of H.
These assumptions limit the application of this model to large flux excursions caused by
high currents. The results of this model may not be very accurate for very low currents.
2.4 Jiles-Atherton Model

This model is based on current physical theories of magnetic domains in ferromagnetic
materials [11], [12]. It uses a Langevin expression to generate an anhysteretic M-H curve.
This expression is used in a state equation to simulate the effects of domain wall bending
and translation, i.e., reversible and irreversible domain wall motion, thereby modeling M-
H curves. The expression used for the anhysteretic M-H curve, Man(H), is:
𝐻 + 𝑀 𝑎
𝑀𝑎𝑛 (𝐻) = 𝑀𝑠 {𝑐𝑜𝑡ℎ ( )− } (18)
𝑎 𝐻 + 𝑀
Where Ms = saturation magnetic moment of the core material,
H = magnetization force in the core,
 = interdomain coupling factor,
𝑎 = domain wall density,

M = magnetic moment.
These parameters are dependent on the type of core material. The equation for the
contribution of the reversible domain wall motion to the core magnetization is:
dM rev  dM dM 
= c an −  (19)
dH  dH dH 
where: Man = the anhysteretic magnetic moment, and

c = the domain flexing parameter.
8
For the irreversible wall motion the equation is:

−1
dM irr  k 
=   −  ( M an − M irr ) ( M an − M irr ) (20)
dH   0 
dH dH
where:  = directional parameter, +1 for 0 and -1 for 0 ,
dt dt
0 = permeability of free space,
k = domain pinning parameter.

Combining equations (18), (19), and (20) yields the total equation:
−1
dM 1  k  dM an
=  −  ( M an − M )  ( M an − M ) + c (21)
dH 1 + c   0  1 + c dH
Using the expression for the anhysteretic moment given in (18) for Man, the equation can
be solved for the core magnetic moment as a function of H. However, this is a static
solution and does not model the dynamic losses of the core material. Therefore the model
includes a damping influence of the domain wall motion given by:
dM
=  (b( H ) − M (t )) (22)
dt
where:  = the domain damping parameter.

Given the magnetizing force as a function of time, H(t), numerical integration techniques
are used to solve (18), (21), and (22) simultaneously to obtain the core flux values at
specified times.
Other core parameters that are needed to model a transformer core using this method
are the magnetic cross section of the core, the magnetic path length, and the core
stacking factor for laminated cores.
The various core hysteresis parameters required in this model, , a, c, k, and  can be
calculated from experimental measurements of the coercivity, remanence, saturation flux
density, initial anhysteretic susceptibility, initial normal susceptibility, and the maximum
differential susceptibility. The domain damping parameter, , is derived from widening of
the dynamic B-H loop with frequency.
Some commercially available programs use the Jiles-Atherton model to simulate the
dynamic behavior of magnetic devices. An example of a B-H loop for a commercial 1200/5
current transformer core calculated by using one of these programs is shown in Figure 2.
In this simulation, a sinusoidal current was used to excite the secondary winding with the
primary open circuited. The current starts at zero and continues for 2.5 cycles. The initial
9
magnetization curve can be seen to progress into a cyclical B-H loop. The secondary
excitation current was 3.53 amperes rms.
B - kG
H - Oersted
Figure 2: B-H Loop for the core of a 1200/5 CT Simulated by the Jiles-Atherton
Method.
The following parameters were used in the simulation shown in Figure 2.

Secondary turns = 240
Core area = 2 sq. inches
Magnetic flux path length = 39.4 inches
Core stacking factor = 0.95
 Interdomain coupling = 9.5E-5

a Thermal energy factor = 100 A/m
Bsat Saturation flux density = 1.8 T
c Domain flexing factor = 0.07
 Domain damping factor = 1E+7

k Domain pinning parameter = 150 A/m
0 permeability of free space = 4E-7 H/m
10
Equations other than Langevin’s have been used for the anhysteretic B-H curve in the
Jiles-Atherton model. Two examples are:
M ( H)
M an = M s (23)
M ( H) + 1
and,
𝑀𝑎𝑛 = 𝑀𝑠 (tanh(𝑀(𝐻))) (24)
Jiles-Atherton models based on (18) are sometimes referred to as Level 1 and those
based on (23) are called Level 2. One commercially available simulation program allows
a choice of Level 1 or Level 2. However, the model using Level 2 does not include the
interdomain coupling and damping effects. It is better suited to model ferrite and
powdered iron cores than Level 1 and is more efficient in the calculation.
Reference [13] provides an update to the Jiles-Atherton model. The derivative of the
total magnetization with respect to the magnetic field is given by
𝑑𝑀 1 (𝑀𝑎𝑛 − 𝑀) 𝑑𝑀𝑎𝑛
= ( +𝑐 ) (25)
𝑑𝐻 (1 − 𝑐𝛼 𝑑𝑀𝑎𝑛 ) 𝛿𝑘 𝛼(𝑀𝑎𝑛 − 𝑀) 𝑑𝐻𝑒
𝑑𝐻𝑒 𝜇0 − 1−𝑐
𝐻𝑒 = 𝐻 + 𝛼𝑀 (26)
where:  = +1 or –1, based on the sign of dH/dt,
𝐻𝑒 = effective field intensity.

An improved Anhysteretic function for the Jiles-Atherton model is proposed as
𝑎1 𝐻𝑒 + 𝐻𝑒𝑏
𝑀𝑎𝑛 = 𝑀𝑠 (27)
𝑎3 + 𝑎2 𝐻𝑒 + 𝐻𝑒𝑏
where, 𝑎1 , 𝑎2 , 𝑎3 , and 𝑏 are appropriately chosen constants [13].
The derivative of 𝑀𝑎𝑛 with respect to 𝐻𝑒 is given by

(𝑏−1)
𝑑𝑀𝑎𝑛 𝑎1 𝑎3 + 𝑏𝑎3 𝐻𝑒 + (𝑏 − 1)(𝑎2 − 𝑎1 )𝐻𝑒𝑏
= 𝑀𝑠 (28)
𝑑𝐻𝑒 (𝑎3 + 𝑎2 𝐻𝑒 + 𝐻𝑒𝑏 )2
In addition, it is proposed that an adjustable 𝑘 be used in (25):
𝑀 2
𝑘modified = 𝑘0 [1 − 𝛽 ( ) ] (29)
𝑀𝑠
11
where 𝑘0 is the base value for 𝑘 and 𝛽 is a parameter controlling the degree of adjustment.
The improved method was incorporated in a CT model in [13]. The improved model is
able to more accurately model the shoulder region of the 𝐵 − 𝐻 loop and thus is able to
get more accurate simulation results in both heavy and light burden conditions.
3. CT Model
The transient response of CTs and the correct CT models in an EMTP simulation are very
important in the evaluation of high-speed protective relaying systems.
3.1 Equivalent Circuit

A reasonably accurate equivalent circuit of a CT up to a few kHz is shown in Figure 3 with
all quantities referred to the secondary side. The primary branch is negligible and could
be ignored without sacrificing the accuracy of the results.
(Rp+ jwLp)/N2 R2+ jwL2
Rb
Rc
jwLm
jwLb
Figure 3: Equivalent Circuit of CT.
3.2 Current Transformer Model Data

The following summarizes the procedure recommended to generate the necessary data
to simulate the CT saturation cases using the EPRI/DCG EMTP (see APPENDIX A for
EMTP files and example model data):
1. Use the subroutine CONVERT(for EMTP) or SATURATION(for ATP) to convert
the CT saturation curve Vrms vs. Irms data to peak flux vs. peak current.
2. Use the subroutine HYSDAT(for EMTP) or HYSTERESIS(for ATP) to generate the
hysteretic reactor { type 96 } data. The above subroutine requires the peak flux
and peak current at the point where the hysteresis curve becomes single valued.
This point was selected as the point where Ie = 10.0 Amperes rms with a
corresponding voltage of 175.0 Volts rms.
3. Modify the last datapoint of the hysteresis curve generated by EMTP or ATP
subroutine to correct for the CT air-core reactance which can be calculated from
12
the CT dimensions. The calculated air-core inductance from the CT dimensions is

approximately 156.4 microhenries. The air-core inductance from the hysteresis
curve is about 155.3 microhenries.
4. The current and voltage below which the magnetizing inductance was assumed to
be linear were selected from the V-I curve as Vrms = 105.0 Volts and Irms = 0.05
Amperes.
3.3 CT Simulation Results

All four different core models investigated by the working group produced nearly identical
results and the simulation results compared very well to the actual laboratory test data.
In the following example, the EMTP hysteretic reactor (type 96) model was used since it
allows inclusion of remanent flux in the CT core. Figure 4a and Figure 4b show the primary
and secondary CT currents scaled to the primary side, obtained in the laboratory, while
Figure 5a and Figure 5b show a comparison of the recorded laboratory secondary CT
waveforms and the simulated secondary CT waveforms using the EPRI EMTP for two
cases with different burden values. The results shown in Figure 5a and Figure 5b can be
regarded as typical output. The first case results are quite satisfactory, with the exception
of slightly larger magnitude negative loops. The simulation results for the second case
are nearly identical to the laboratory tests. The EMTP models used are shown in more
detail in APPENDIX A.
20
10
Current [kA]
-10
-20
25 50 75 100
Time [ms]
a)
30
20
10
Current [kA]
-10
-20
0 25 50 75 100
Time [ms]
b)
Figure 4: Laboratory Test Data of CT Saturation with Remanence in the Core and
a) Burden of 1.3 + j 0.175 Ohms and b) Burden of 8.33 + j0.175 ohms.
13
100
80
60
40
Current [A]
20
0
-20
-40
-60
-80
25 50 75 100
Time [ms]
a)
80
60
40
20
Current [A]
-20
-40
-60
-80
25 50 75 100
Time [ms]
b)
Figure 5: Comparison of EMTP Simulations (solid red line) and

Laboratory Test Data (dashed blue line) with Remanence in CT Core for
a) Burden of 1.30 + j0.175 ohms and b) Burden of 8.33 + j0.175 ohms.
The significant features to be included in a CT model are:
• A non-linear core representation with ability of the model to accept initial core
remanence. Representation of hysteresis is not as critical.
• The final slope of the hysteresis curve vs. current is compared with the air- core
reactance of the CT and adjusted accordingly. This helps provide a good transient
performance of the model when compared to actual laboratory tests.
The PSRC CT saturation calculator includes many of the aspects discussed in this section
and has been used quite extensively for protection applications in the industry [14].
14
4. CCVT Model
The equivalent circuit of a CCVT is shown in Figure 6. In order to avoid some of the
insulation problems in a wound VT at high voltage levels, some of the step down ratio is
achieved in a capacitive divider prior to applying the voltage to the voltage transformer.
The capacitive voltage divider C1 and C2 brings the primary voltage down to C1V/(C1+ C2)
where V is the voltage applied to the divider. This voltage is then applied to the wound
VT via a tuning reactor Lt and the output of the VT is connected to the burden. In many
designs the tuning reactor is in the ground line rather than the high voltage lead of the VT
but this does not affect the analysis and simulation results. The function of the tuning
reactor is to minimize the equivalent source impedance at the system nominal frequency
(50 or 60 Hz) by tuning it to the capacitance C1 || C2 in order to allow the burden to draw
its current without causing significant voltage drop and thus affecting the accuracy of the
measured voltage.
Line Voltage
C1
Lt r1 L1 r2 L2
C2
Rm Lm Rb
n:1
Voltage
Transformer
Figure 6: Basic CCVT Model with Equivalent Circuit of VT.
This feature was necessary when the instruments comprising the burden drew significant
currents from the CCVT. The interaction of the source capacitance with the combined
inductance of the tuning reactor Lt and the magnetizing inductance Lm of the voltage
transformer is responsible for the subsidence transient problem and could lead to
ferroresonance under particular excitation conditions. The elements shown in Figure 6
are sufficient to describe the CCVT for frequencies up to 500 Hz. The major simulation
challenge is the representation of the non-linear elements in the magnetizing branch of
the VT (Lm and its parallel resistance) since both vary with the excitation of the core.
Some authors have used non-linear elements available in EMTP [15], [16], while others
have used custom current-source circuits (see Section 2) which fit well into the Dommel
algorithms [6]. The latter are particularly suitable for real-time simulation when there may
be no time for iterations. Reclose operations can cause ferroresonance because of the
accumulated flux. Figure 7a shows simulation of a reclose operation from [6] for a case
that gives rise to a subsidence transient. (Note the absence of any high frequency
oscillation at the voltage collapse due to the absence of stray capacitances in the model).
The burden in this case is the rated value of 200VA. The transient can be seen much
more clearly on the core flux waveform shown in Figure 7b. No ferroresonance occurs.
15
a)
b)
Figure 7: For a reclose case, a) Secondary Voltage and b) Core Flux Density.
This low frequency phenomenon can also give rise to ferroresonant conditions if the
reclose occurs when the flux density is at or near saturation and the burden is very small
(Figure 8).
This case is for a light burden (20VA for a 200VA CCVT) condition [8]. Another condition
known to give rise to ferroresonance is saturation of the auxiliary transformer in a relay
input circuit that forms the burden of the CCVT [17]. As the auxiliary transformer comes
out of saturation, the sudden transition from a very low to a high impedance can give rise
to ferroresonance, depending on the flux condition in the VT of the CCVT. One of the
early techniques used to avoid ferroresonance was to add sufficient damping to the
burden. A more sophisticated technique uses ferroresonance suppression circuits (FSCs)
with a low impedance around the subsidence transient frequency to increase the damping
at this frequency.
Figure 8: Secondary Voltage with a Small Burden (20 VA for a 200 VA CCVT).
16
FSC can be designed in two operational modes: active and passive.

FSCs in an active operational mode consist of capacitors and iron core inductors
connected in parallel and tuned to the fundamental frequency. They are permanently
connected on the secondary side and affect the CCVT transient response.
FSCs in a passive operational mode consist of a resistor connected on the secondary
side. This resistor can be permanently connected. Another option is to have a gap or an
electronic circuit connected in series with the resistor, which are activated whenever an
overvoltage occurs. This design avoids the loading effect of the FSC under normal
conditions. These FSCs do not affect CCVT transient response unless an overvoltage
occurs.
An active operational mode FSC design is shown in Figure 9a. Capacitor Cf is connected
in parallel with an iron core inductor Lf tuned to the fundamental frequency. Resistor Rf is
a damping resistor designed to damp ferroresonance oscillations within one cycle. The
circuit is tuned with a high Q factor in order to attenuate ferroresonance oscillations at
any frequency except the fundamental (Figure 9d, Curve A).
The FSC can be modeled using two different Lf representations: FSC model with Lf
represented as an air core inductance, as shown in Figure 9b. Components Cf and Lf are
tuned to 60 Hz. This circuit has a low Q factor as shown in Figure 9d, Curve B. The FSC
can also be modeled as shown in Figure 9c. Lf represented as a non-saturable
transformer. Primary and secondary windings are connected with polarities as presented
in Figure 9c. The calculated Lf value is incorporated in the transformer model as a self-
inductance so that parallel resonance occurs at the fundamental frequency. At other
frequencies, only the leakage inductance is involved and the damping resistor is the one
which attenuates ferroresonance oscillations. This circuit has a high Q factor, very close
to the actual design as shown in Figure 9d, Curve C. FSC simulation using the transformer
representation of Lf is more accurate [16]. Figure 10 shows the effect this filter has on
damping of the ferroresonance.
17
Figure 9: Ferroresonance filter circuit: a) Actual Design, b) and c) Digital Models,

and d) Frequency Response of 9a (Curve A), 9b (Curve B), and 9c (Curve C).
The equivalent circuit shown in Figure 6 is not adequate for predicting the high frequency
response of the CCVT. At frequencies around 1kHz [17] it is necessary to consider the
stray capacitances in the various circuit elements in order to explain the measured
frequency responses in the linear region of operation [15]. Figure 11 shows the influence
of the coupling capacitor and stray capacitances on the frequency response for a typical
CCVT. Figure 12 shows the equivalent circuit of a particular CCVT including stray
capacitances.
18
Figure 10: Burden Voltage for the case shown in Figure 8 with Ferroresonance
Filter in the Circuit.
Figure 11: Frequency response of CCVT in Linear Region.
Figure 13 shows the frequency response for the simulation model and the actual CCVT.
The frequency response tests were conducted at low voltage levels and do not reveal any
of the effects which result from VT core saturation, e.g. ferroresonance. The response
does however reveal a rejection “cusp” around 1 kHz where the response falls off rapidly.
The rejection cusp is almost entirely due to the parallel resonance of the compensating
inductor and its stray capacitance. The voltage across the compensating inductor at this
frequency can be seven to eight times the voltage at the nominal system frequency. If the
19
line voltage input contains frequencies around this value, e.g. capacitor bank
energization, it is important to model any spark gap across the tuning inductor which can
trigger and alter the frequency response under this resonant condition [17].
H Lcl Magnetizing Inductance = 43.15 H

1 Lm Air Core Inductance = 91.1 H
Active Inductance = 39.7 KH
Lh Air-Core Inductance = 5.5 mH
C1 7910 pf Active Inductance = 5.5 H
0.001 0.001
296 Ohm 6.5 H 155:1 X
Ce=C1 || C2 1
Cc 1158 pf
1.5 Ohm
C2 5 MOhm 256 pf 8 MOhm Rc Lm 200
Vpk
145040 pf Cp 100 Ohm
Y
Lcl 1
127 Ohm 5.66
Ohm
100 ohm Lh
25 Ohm X
2
Y
2
Vflash=350 Vpk
Figure 12: Complete Equivalent Circuit of a TEHM 345 kV CCVT for a 60 Hz power
system.
Figure 13: Frequency response of CCVT in Linear Region.

The significant features which must be included in a CCVT model are therefore:
• A good nonlinear core representation for any elements containing iron. The core
model is expected to at least be capable of accepting starting core remanence if it
does not include hysteresis.
20
• Stray-capacitance of the compensating inductor and the primary capacitance across

the VT when the bandwidth of the relay under test extends beyond the rejection cusp
frequency (typically greater than 500Hz).
• Proper modeling of any voltage limiting elements, e.g., spark gaps, saturating reactors
or metal oxide varistors across the compensating inductor or ferroresonant filter.
• Proper modeling of the drain coil if it is part of the CCVT when the CCVT is used to
couple a carrier signal on the power line in addition to providing voltage signals to
protective relaying systems.
For evaluating CCVT transients during power system faults, a simple approach of
representing the CCVT using a transfer function has been used [18], [19]. The advantage
of using transfer functions to model a CCVT is its ease-of-use with different EMTP
software without requiring knowledge of the different parameters. The disadvantage is
that the model does not consider the effects of any nonlinear elements. Consequently,
the model is only suitable for applications that do not require consideration of nonlinear
elements, such as evaluation of security and speed of an underreaching distance zone
or determining accuracy of impedance-based fault locating functions [18], [20].
The transfer functions for an active FSC CCVT and a passive FSC CCVT are represented
by (30) and (31), respectively, for 60 Hz systems. The transient response of these transfer
functions for a close-in bolted fault at a voltage peak and a close-in fault at a voltage zero
are shown in Figure 14. The steady-state gains of the transfer functions are unity and a
subsequent division by the turns ratio for a specific system provides the desired scaling.
587.7 • 𝑠 3 + 2.642𝐸5 • 𝑠 2 + 6.485𝐸7 • 𝑠

𝐻𝐴 (𝑠) = (30)
𝑠 4 + 819 • 𝑠 3 + 5.27𝐸5 • 𝑠 2 + 9.74𝐸7 • 𝑠 + 1.6𝐸10
2.923𝐸4 • 𝑠 4 + 3.67𝐸4 • 𝑠 3 + 2.035𝐸11 • 𝑠 2
𝐻𝑃 (𝑠) = 5 (31)
𝑠 + 2.93𝐸4 • 𝑠 4 + 1.17𝐸8 • 𝑠 3 + 2.05𝐸11 • 𝑠 2 + 1.54𝐸13 • 𝑠 + 6.5𝐸13
Figure 14: Transient Response of CCVT Transfer Functions [19].
21
The same CCVTs represented in (30) and (31) for 60 Hz systems have the transfer
functions shown in (32) and (33) for 50 Hz systems.
401.1 • 𝑠 3 + 1.819𝐸5 • 𝑠 2 + 4.464𝐸7 • 𝑠

𝐻𝐴 (𝑠) = (32)
𝑠 4 + 624 • 𝑠 3 + 3.976𝐸5 • 𝑠 2 + 6.704𝐸7 • 𝑠 + 1.098𝐸10
2.923𝐸4 • 𝑠 4 + 2.546𝐸4 • 𝑠 3 + 1.413𝐸11 • 𝑠 2
𝐻𝑃 (𝑠) = 5 (33)
𝑠 + 2.93𝐸4 • 𝑠 4 + 1.147𝐸8 • 𝑠 3 + 1.421𝐸11 • 𝑠 2 + 1.066𝐸13 • 𝑠 + 4.509𝐸13
5. VT Model
Modeling of magnetic voltage transformers is, in principle, similar to modeling any other
instrument transformer. However, the large inductance of the primary winding and the
importance of heavy saturation and hysteresis loop require special attention. Figure 15
shows the model parameters needed to accurately simulate the transient response of
VTs.
The model parameters can be determined from tests and manufacturers’ datasheets as
follows:
• Rout is the output load resistance that can be measured with a conventional, high
accuracy Ohm-meter.
• Cin is the input capacitance that can be determined from the manufacturer’s data on
capacitive current versus input voltage.
• Rcore is the core resistance that can be determined from the manufacturer’s data on
iron losses versus applied voltage.
• Rext is the external source resistance that can be measured using a conventional Ohm-
meter.
• Lout is the output load inductance.
• Lext is the external source inductance.
• Lm is the magnetizing inductance.

The most difficult parameters to be determined are Lout, Lext, and Lm of the hysteretic
element. In particular, it is very difficult to separate the value of L ext from the hysteretic
element Lm and a special measurement may be required to accomplish this task. This
special test is described below to help users obtain these parameters.
The primary of the VT is energized with a dc source of ~125 volts to excite the core. This
can take up to a few seconds. Then the primary of the VT is shorted out and the flux is
allowed to decay. This can take up to tens of seconds. Next, the polarity of the dc source
is reversed and the sequence is repeated. After several such sequences the core has
reached a repeatable pattern of energizations and de-energizations. The waveshapes of
primary voltage, primary current, and secondary voltage are recorded during each
22
sequence. From these waveshapes a flux-current hysteretic function can be generated.

The magnetic flux linkage can be obtained by numerically integrating the input voltage
wave as m = Vdt = Vt, where t is the sampling time of the waveform. The parameter
Lext can be determined as the slope of m curve at the maximum current. If the input
current is not enough to drive the VT into heavy saturation this test could be repeated
with higher dc voltage with the help of a capacitor, rather than a permanent dc source.
The capacitor is first charged and then discharged through the primary of the VT. This is
a transient way of testing which eliminates the steady state heating of the VT primary
winding, which could be substantial if a permanent dc source is used for testing. As an
example, the value of Lext was determined to be around 38.5 Henry for a commercial
34.5 kV/115 V VT.
n:1
Ideal
Rext Lext Transformer Rout Lout
Cin Rc Lm
Figure 15: Equivalent Circuit of a VT with Hysteretic Core
6. Merging units and Instrumentation Error Correction

The introduction of merging units provide an excellent opportunity to integrate the error
correction algorithms within the merging unit and stream the corrected data to the
destination which is normally a process bus. Merging units have powerful computing
devices to accommodate this approach. Alternatively, the error correction process can be
performed at the process bus [21].
One important issue is that the error correction algorithms may introduce appreciable
time-delays in delivering the data. Regarding this issue, it is recognized that there are two
types of error correction algorithms: (a) algorithms that operate on one sample at the time,
and (b) algorithms that require data over a time interval, for example half-cycle data (using
the IEC 61869-9 preferred sample rate of 4800 Hz that will be 40 samples on a 60 Hz
system). These two methods will introduce different latencies. For example, algorithms of
the (a) variety will introduce minimal latency. Specifically, each sample is corrected and
then placed into the stream. In this case the latency will be affected only by the time
required to perform error correction on one sample, practically negligible. Algorithms of
type (b) first require collection of the data over the specified time interval, then perform
the error correction, and finally place the corrected data into the stream. Assuming that
23
the algorithm is really fast, the latency is at least equal to the time interval required to
collect the data.
7. Conclusions
This report has investigated a number of mathematical models that could be used to
represent the nonlinear behavior of the magnetic core of instrument transformers.
The results of four mathematical models of CTs were compared with actual laboratory
saturation test data. The results from all four models were almost identical and in
extremely good agreement with the laboratory test data.
An approach to determine the nonlinear magnetizing branch parameters of VTs was
presented. The parameters for CCVT models were described, including their active and
passive ferroresonance suppression circuits. For certain protective relaying applications,
modeling the nonlinear elements of a CCVT may not be required, for which simplified
transfer function representations were discussed.
The most significant features and parameters that allow accurate modeling of an
instrument transformer have been described in detail. Modeling instrument transformers
and considering their transient response helps provide a high degree of security and
dependability for high-speed protective relays in their design and applications. The use
of electromagnetic transient programs to adequately model different power system
components and to generate transient data to assess the performance and verify correct
application of microprocessor high-speed protective relays has become essential.
24
8. References
[1] D. A. Tziouvaras, P. McLaren, G. Alexander, D. Dawson, J. Esztergalyos, C. Fromen,
M. Glinkowski, I. Hasenwinkle, M. Kezunovic, L. Kojovic, B. Kotheimer, R. Kuffel, J.
Nordstrom, and S. Zocholl, “Mathematical models for current, voltage, and coupling
capacitor voltage transformers,” in IEEE Transactions on Power Delivery, vol. 15,
no. 1, pp. 62–72, January 2000, © 2000 IEEE doi: 10.1109/61.847230.
[2] IEC 61869-5:2011, Instrument Transformers – Part 5: Additional Requirements for
Capacitor Voltage Transformers.
[3] M. Kezunovic, L. Kojovic, A. Abur, C. W. Fromen, D. R. Sevcik, F. Phillips,
“Experimental Evaluation of EMTP-Based Current Transformer Models for Protective
Relay Transient study,” in IEEE Transactions on Power Delivery, vol. 9, no. 1,
pp. 405-412, January 1994.
[4] W. L. A. Neves and H.W. Dommel, “On modeling iron core nonlinearities,” IEEE Trans.
on Power Systems, vol. 8, no. 2, pp. 417-423, May 1993.
[5] J. R. Lucas, “Representation of magnetization curves over a wide region using non-
integer power series,” IJEEE, vol. 25, no. 4, 1988, Manchester U.P., U.K., p335.
[6] J. R. Lucas, P. G. McLaren, W. W. L. Keerthipala, and R. P. Jayasinghe, “Improved
simulation models for current and voltage transformers in relay studies,” IEEE Trans.
on Power Delivery, vol. 7, no. 1, pp. 152-159, January 1992.
[7] H. W. Dommel, “Digital computer solution of electromagnetic transients in single and
multiphase networks,” IEEE Trans. on PAS, PAS 88, no. 4, pp. 388-399, April 1969.
[8] P. G. McLaren, J. R. Lucas, and W. W. L. Keerthipala, “A digital simulation model for
a CCVT in relay studies,” IPEC ‘93, Singapore, March 1993.
[9] R. Garrett, W. C. Kotheimer, and S. E. Zocholl, “Computer simulation of current
transformers and relays for Performance Analysis,” proceedings of the 14th Annual
Western Protective Relay Conference, Spokane, WA, October 1987.
[10] R. Chowdhury, D. Finney, N. Fischer, and D. Taylor, “Determining CT Requirements
for Generator and Transformer Protective Relays,” proceedings of the 46th Annual
Western Protective Relay Conference, Spokane, WA, October 2019.
[11] D. C. Jiles, D. L. Atherton, “Theory of ferromagnetic hysteresis,” Journal of Magnetism
and magnetic Materials, vol. 61, pp. 48-60, 1986.
[12] D. C. Jiles, J. B. Thoelke, M. K. Devine, “Numerical determination of hysteresis
parameters for modeling of magnetic properties using the theory of ferromagnetic
hysteresis,” IEEE Transactions on Magnetics, vol. 28, no. 1, pp. 27-34, January 1992.
[13] U. D. Annakkage, P. G. McLaren, E. Dirks, R. P. Jayasinghe, and A. D. Parker, “A
current transformer model based on the Jiles-Atherton theory of Ferromagnetic
hysteresis,” IEEE Tran PWRD, vol. 15, no. 1, January 2000, pp. 57-61.
25
[14] G. Swift, “Theory for CT SAT Calculator (PSRC),” IEEE PSRC Committee
Spreadsheet, September 2003.
[15] M. Kezunovic, C. W. Fromen, S. L. Nilsson, L. Kojovic, V. Skendzic, and D. R. Sevcik,
“Digital models of coupling capacitor voltage transformers for protective relay transient
studies,” IEEE Trans. on Power Delivery, vol. 7, no. 4, October 1992, p1927.
[16] L. A. Kojovic, M. Kezunovic, and S. L. Nilsson, “Computer simulation of a
ferroresonance suppression circuit for digital modeling of coupling capacitor voltage
transformers,” ISMM International Conference, Orlando, Florida, 1992.
[17] M. J. Wiseman, “CVT transient behavior during shunt capacitor switching,” Ontario
Hydro Study No. W401, April 15, 1993.
[18] B. Kasztenny, D. Sharples, V. Asaro, and M. Pozzuoli, “Distance Relays and
Capacitive Voltage Transformers – Balancing Speed and Transient Overreach,”
proceedings of the 54th Annual Georgia Tech Protective Relaying Conference,
Atlanta, GA, May 2000.
[19] R. Chowdhury and N. Fischer, “Transmission Line Protection for Systems With
Inverter-Based Resources – Part I: Problems,” in IEEE Transactions on Power
Delivery, IEEE, vol. 36, no. 4, August 2021, pp. 2416-2425.
[20] B. Kasztenny and R. Chowdhury, “Security Criterion for Zone 1 Applications in High
SIR Systems with CCVTs,” proceedings of the 76th Annual Georgia Tech Protective
Relaying Conference, Atlanta, GA, May 2023.
[21] P. E. Emeka Obikwelu and A. P. S. Meliopoulos, “CT Saturation Error Correction
Within Merging Units Using Dynamic State Estimation,” 2020 IEEE 3rd International
Conference on Renewable Energy and Power Engineering (REPE).
[22] A. H. A. Bakar, C. H. Lim, and S. Mekhilef, “Investigation of Transient Performance of
Capacitor Voltage Transformer,” IEEE International Power and Energy Conference,
Putra Jaya, Malaysia, pp. 509-515, November 2006.
[23] D. Fernandes, W. L. A. Neves, and J. C. A. Vasconcelos, “Identification of parameters
for coupling capacitor voltage transformers,” International Conference on Power
Systems Transients (IPST), 2001.
[24] D. Fernandes, W. L. A. Neves, and J. C. A. Vasconcelos, “A coupling capacitor voltage
transformer representation for electromagnetic transient studies,” International
Conference on Power Systems Transients (IPST), 2003, New Orleans, USA.
[25] D. Fernandes, W. L. A. Neves, and J. C. A. Vasconcelos, “Coupling Capacitor Voltage
Transformer: Laboratory Test and digital Simulations,” International Conference on
Power Systems Transients (IPST), 2005.
[26] D. Fernandes, W. L. A. Neves, and J. C. A. Vasconcelos, “Comparisons between lab
measurements and digital simulations for a coupling capacitor voltage transformer,”
IEEE Transmission & Distribution Conference and Exposition: Latin America, 2006.
26
[27] D. Fernandes, W. L. A. Neves, and J. C. A. Vasconcelos, “Coupling capacitor voltage

transformer: A model for electromagnetic transient studies,” Electric Power Systems
Research (EPSR), 2007.
[28] C. A. Silva, D. Fernandes Jr, and W. L. A. Neves, “Correction of the secondary voltage
of coupling capacitor voltage transformers in real time,” International Conference on
Power Systems Transients (IPST), 2011.
[29] N. Villamagna and P. A. Crossley, “A CT saturation detection algorithm using
symmetrical components for current differential protection,” in IEEE Transactions on
Power Delivery, vol. 21, no. 1, January 2006, pp. 38-45.
[30] F. C. F. Guerra and W. S. Mota, “Current transformer model,” in IEEE Transactions
on Power Delivery, vol. 22, no. 1, January 2007, pp. 187-194.
[31] F. C. F. Guerra and W. S. Mota, “Magnetic core model,” IET science, measurement &
technology, 2007.
[32] R. M. Rifaat, “Considerations in applying EMTP to evaluate current transformer
performance under transient and high current fault conditions,” International
Conference on Power systems Transients, 2005.
[33] F. M. Uriarte, V. A. Centeno, J. D. L. Lopez, and J. Depablos, “Continuous vs.
piecewise hysterisis model of a current transformer”, Research in Microelectronics
and Electronics, 2006.
[34] F. M. Uriarte, “Hysterisis Modeling by Inspection,” 38th IEEE NAPS 2006.
[35] P. Rafajdus, P. Bracinik, and V. Hrabovcova, “Current transformer analysis under
transient conditions,” 2010 XIX International Conference on Electrical Machines
(ICEM), 2010.
[36] W. Zang, G. Shu, Z. Feng, and R. Unbehauen, “Digital simulation models of a
capacitor voltage transformer,” Electrical Engineering, vol. 87, 2005.
[37] S. Liang, Q. Hu, and W.J. Lee, “CCVT calibration for harmonic voltage measurement
by digital simulation and field test,” IEEE North American Power Symposium, 2009.
[38] Y. Wang, T. Liu, and X. Hu, “Study of current transformer calibrating system based on
equivalent model,” 10th IEEE International Conference on Industrial Informatics
(INDIN), 2012.
27
APPENDIX A EMTP Model Setup for CT Saturation

A.1 CT and Secondary Burden Data
Turns Ratio 900/5
Mean Core Length 0.4987 meters
Area 1.91532E-3 square meters
Winding Resistance 0.253 Ohms
Remanent Flux 0.4645 Volt-sec
Burden (case 1) 1.30 + j0.175 Ohms
Burden (case 2) 8.33 + j0.175 Ohms
A.2 CT Saturation Data

Irms (Amps) Vrms (Volts)
40.0E-3 85.0
50.0E-3 105.0
100.0E-3 140.0
200.0E-3 150.0
10000.0E-3 175.0
9999
A.3 EMTP Subroutine CONVERT

C =============================================
C WRITTEN BY Demetrios A. Tziouvaras
C =============================================
C
BEGIN NEW DATA CASE
SATURATION
C
C CONVERSION OF V-I CHARACTERISTIC
C TO PEAK FLUX VS PEAK CURRENT 900/5
C C-100 BUSHING CURRENT TRANSFORMER
C SQUARE D ARMCO M4 GRAIN ORIENTED
C
C FOR THE EMTP TO ACCEPT VOLTAGE IN VOLTS AND
28
C CURRENT IN AMPS
C WE MUST SET VBASE = 1.0*E-3 AND PBASE = 1.0*E-6
C
C FREQU][ VBASE][PBASE ][IPUNCH][KTHIRD]
60.0 1.0E-3 1.0E-6 0 0
C
C IRMS AMPS ] VRMS ]
C
40.0E-3 85.0
50.0E-3 105.0
100.0E-3 140.0
200.0E-3 150.0
C 1000.0E-3 160.0
10000.0E-3 175.0
9999
BLANK CARD ENDING CASE
BLANK
A.4 EMTP Output from CONVERT Subroutine

Ipeak (Amps) Peak Flux(Volts-sec)
.56568542E-01 .31886203E+00
.71508353E-01 .39388839E+00
.17030469E+00 .52518452E+00
.47571570E+00 .56269770E+00
.23289864E+02 .65648065E+00
9999
A.5 EMTP Subroutine HYSDAT

C =============================================
C WRITTEN BY Demetrios A. Tziouvaras
C =============================================
BEGIN NEW DATA CASE
SATURATION
C
C V-I CONVERSION TO TYPE-96 HYSTERETIC INDUCTOR
C USE THE LAST ROW OF DATA FROM THE CONVERT
C ROUTINE PUNCHED OUTPUT
C
C FREQ ]
88.
29
C ITYPE][ LEVEL][IPUNCH]
1 4 0
C
C I SAT][F SAT ]
C
23.28986.6564807
BLANK CARD ENDING CASE
BLANK
BEGIN NEW DATA CASE
A.6 EMTP Output from Hysdat Subroutine

Ipeak (Amps) Flux peak (Volt-sec)
-.87336977E+01 -.64103411E+00
-.58224650E+01 -.63717244E+00
-.26201093E+01 -.62365665E+00
-.11644930E+01 -.61014089E+00
-.43668486E+00 -.59855593E+00
.14556162E+00 -.57538601E+00
.50946568E+00 -.54835446E+00
.84425739E+00 -.50201465E+00
.10189314E+01 -.42478163E+00
.11644930E+01 -.30893209E+00
.14556163E+01 .20659833E+00
.16011778E+01 .28576219E+00
.20378627E+01 .38616512E+00
.26201093E+01 .46339814E+00
.31732433E+01 .50201465E+00
.41485062E+01 .54063116E+00
.56769031E+01 .57538601E+00
.77875472E+01 .60241760E+00
.10189313E+02 .62172585E+00
30
.14556162E+02 .64103411E+00
.23289860E+02 .65648070E+00
C .32023558E+02 .66034237E+00
.39762571E+03 .71462608E+00
.99990000E+04
Notes:
1. The final slope of the data from the subroutine HYSDAT was modified to the slope
of the calculated air-core inductance of the CT. The commented line with a C in
front of the data, is the actual HYSDAT data, compared with the modified data from
the line that follows which reflects the calculated CT air-core inductance.
2. The residual flux was calculated from the final hysteresis datapoints:
Remanent flux = 65.0%*.71462608 Volt-sec
31
APPENDIX B Test system Parameters for Jiles-Atherton Model

B.1 Test system parameters used for estimation algorithm
Fig. B.1 shows the test case used for Estimation of Jiles-Atherton parameters consisting
of a source, bus, breaker, load, a generic CT and fault. The generic CT is used with the
manufacture CT data to produce the B-H loop. The 3 phase breaker current is used as
the CT primary current.
The test system parameters, i.e., source impedance, load impedance and fault point on
wave are arbitrarily set as long as B-H loop can be obtained. Table B.I shows a set of
parameters used for the testing.
Fig. B.1. Test case for Jiles-Atherton parameter estimation
32
Table B.I. Test system parameters used for Estimation algorithm

C400 C400_600/5 C100_900/5 C100_1000/5 10L800_1200/5 C400
150/5 1080/1
Source magnitude 300 500 600 700 800 5000
(kV)
Source frequency 60 60 60 60 60 60
(HZ)
source impedance 10@89° 10@89° 10@89° 10@89° 10@89° 10@89°
(RRL)
Load1 L (H) 0.45 0.45 0.45 0.45 0.45 0.45
(LR) R (ohm) 180 180 180 180 180 180
Load 2- R (ohm) 100 100 100 100 100 100
Point on wave arbitrarily arbitrarily arbitrarily arbitrarily arbitrarily Arbitrarily
B.2 Test system parameters used for Validation

Fig. B.2 shows the test case used for validation consisting of a source, bus, breaker, load,
a generic CT and a J-A CT models used for comparison. The generic CT is used with the
manufacture CT data and the J-A CT model is used with the estimated data. The 3 phase
breaker current is used as the CT primary current.
Fig. B.2. Test case for parameter Comparison with Generic CT

In Fig B.3, the black curve is the generic CT model and the red curve is the test J-A CT
model. The secondary current (Is), B(t) and H(t) are almost identical in comparison in
Fig.B.3.
33
Fig. B.3. Waveform Comparison of J-A CT with estimated parameters and

Generic CT
The test system parameters, i.e., source impedance, load impedance and fault point on
wave are arbitrarily set as long as B-H loop can be obtained. Table B.II shows a set of
parameters used for validation testing.
TABLE B.II Test system parameters used for Validation

C400 C400_600/5 C100_900/5 C100_1000/5 10L800_1200/5 C400
150/5 1080/1
Source magnitude 150 200 150 150 300 600
(kV)
Source frequency 60 60 60 60 60 60
(HZ)
source impedance 10@89° 10@89° 10@89° 10@89° 10@89° 10@89°
(RRL)
Load1 L (H) 0.45 0.45 0.45 0.45 0.45 0.45
(LR) R (ohm) 180 180 180 180 180 180
Load 2- R (ohm) 100 100 100 100 100 100
Point on wave arbitrarily arbitrarily arbitrarily arbitrarily arbitrarily arbitrarily
Fault Duration (sec) 0.1 0.1 0.1 0.1 0.1 0.1
34
APPENDIX C Modeling CT Remanence

The “generic” piecewise linear curve fitting models allow the user to specify an initial value
of remanence but do not explicitly model remanence for ongoing situations such as a
reclose operation or a GIC event. This may lead to errors in such situations. To illustrate
the reclose situation using the same CT as in Figure C.1 shows the input current referred
to the secondary for a single test “shot” [13]. The initial remanence is zero.
Figure C.1. A single test shot referred to the secondary
35
Figure C.2. Secondary current and core flux for the input current of Figure C.1
Figure C.2 shows the model waveforms for the secondary current and core flux. JA is the
Jiles Atherton model (proved against recorded waveforms in [13]) and “Lucas” is a generic
piece wise linear model [6]. The total burden on the CT secondary is a resistance of
15  and inductance of 0.8 mH. The 2 models agree well during the shot but thereafter
the flux wave forms diverge. When the primary current goes to zero there is a small
current (magnetizing component) left in the secondary. The flux in the Lucas model
decays in accordance with the secondary time constant L/R where L is largely the
magnetizing inductance referred to the secondary (primary on open circuit). The JA model
remains high.
36
Figure C.3. Secondary current and core flux for a second shot applied after 5
seconds. Note initial flux
If a second shot is applied after 5 seconds the initial values of core flux will be different
leading to differences in the current waveform for the first half cycle. This is shown in
Figure C.3 where the first half cycle of current is much larger in the Lucas model than in
the JA model. (This is difficult to see because of the scales being used to demonstrate
the flux decay.)
If the secondary burden is reduced to a resistance of 1.5 ohms and inductance of 8mH
then Figure C.4 shows the results for the first shot.
37
Figure C.4. Flux and current waveforms for a light burden case.
It takes much longer to reach saturation and the Lucas model flux decays much more
slowly due to the longer secondary time constant. For a high speed reclose in say 1
second the difference between the two models will likely not be significant. Note the
regular waveshape of the current after saturation and the removal of the offset.
38
APPENDIX D Recent work in modeling instrument transformers

Quite a number of papers have been found related to modeling instrument transformers
since the publication of the PSRC working group paper. The following section first
presents a summary of surveyed papers, and then provides more details of the modeling
methods of several selected papers.
D.1 Summary
In Reference [13], Jiles-Atherton theory is applied in simulating the hysteresis loop of the
core material of the CT. An alternative function is proposed to replace the Langevin
function to achieve good agreement between test and simulated waveforms in the light
burden case typical of digital relay loading.
Reference [22] presents simulation studies of a typical 132 kV CCVT model using ATP-
EMTP software.
In Reference [23], a method to obtain the CCVT model parameters from frequency
response curves is presented. Frequency response measurements of magnitude and
phase, in the range from 10 Hz to 10 kHz, were carried out for a 230 kV CCVT at a high
voltage laboratory and used as input data to a full Newton-type fitting routine to estimate
the CCVT parameters. Analytical CCVT functions were fitted to the measured data. The
magnitude and phase errors, in the whole frequency range, are fairly small. The nonlinear
behavior of the voltage transformer (VT) magnetic core is also taken into account. The
obtained CCVT model may easily be used in connection with the Electromagnetic
Transients Program (EMTP).
In Reference [24], a CCVT model to be used in connection with the EMTP is presented.
A support routine was developed to compute the linear 230 kV CCVT parameters
(resistances, inductances and capacitances) from the magnitude and phase of the CCVT
voltage ratio measured in laboratory, from 10 Hz to 10 kHz. The errors were fairly small
in the whole frequency range. The potential transformer (PT) multiple windings, its
magnetic core and the silicon carbide surge arrester nonlinear characteristics were taken
into account in the model in order to improve transient responses due to overvoltages.
The surge arrester gap sparkover voltage was measured and the silicon carbide nonlinear
characteristic was estimated. It is shown that careful attention must be taken when
computing the ferroresonance suppression circuit (FSC) parameters to avoid numerical
instabilities in time-domain simulations. The protection circuit is very effective in damping
out transient voltages when a short circuit is cleared at the CCVT secondary side.
In Reference [25], laboratory tests of ferroresonance and circuit breaker switching were
carried out for a 230 kV CCVT. The magnetic core and surge arrester nonlinear
characteristics were taken into account in the model in order to improve the transient
response to overvoltages. Digital simulations were performed using a CCVT model with
linear parameters obtained from frequency response measurements. It is shown that the
CCVT model is fairly accurate in reproducing ferroresonance and low-frequency switching
operations measured during the laboratory tests. Simulations had shown that transient
overvoltages produced inside the CCVT, when a short circuit is cleared at the CCVT
39
secondary side, are effectively damped out by the ferroresonance suppression circuit and
the protection circuit. Comparisons of CCVT transient performance considering two types
of surge arresters, used as protection circuits, were carried out. Voltages are damped out
faster for the zinc oxide (ZnO) surge arrester as compared to silicon carbide (SiC) arrester
type. Simulation of a current chopping case study is also presented.
In Reference [26], measurements were carried out for a 230 kV CCVT in a high voltage
laboratory. In order to obtain the frequency response data to be used as input data to a
routine developed to compute the CCVT parameters, frequency response
measurements, for magnitude and phase, were carried out for the CCVT. These
measurements were made directly, without the need of using external resistors in the
CCVT terminals. It is shown that the use of external resistors, as proposed in some works
in literature, may cause distortions in the frequency response curves. In order to validate
the CCVT model in time-domain studies, ferro-resonance and circuit breaker switching
operations tests were carried out for the CCVT, and the errors were fairly small for both
tests.
In Reference [27], an accurate CCVT model for electromagnetic transient studies is
presented. The model takes into account linear and nonlinear elements. A support routine
was developed to compute the linear 230 kV CCVT parameters (resistances, inductances
and capacitances) from frequency response data. The magnetic core and surge arrester
nonlinear characteristics were estimated from laboratory measurements as well. The
model is used in connection with the electromagnetic transients program (EMTP) to
predict the CCVT performance when it is submitted to transient overvoltages, as are the
cases of voltages due to the ferroresonance phenomenon and circuit breaker switching.
The difference between simulated and measured results is fairly small. Simulations had
shown that transient overvoltages produced inside the CCVT, when a short circuit is
cleared at the CCVT secondary side, are effectively damped out by the ferroresonance
suppression circuit and the protection circuit.
In Reference [28], a compensator device capable of performing the correction of the
CCVT model secondary voltage in real time, is presented. The methodology is based on
a technique for recursive digital filtering of the CCVT secondary voltage. The parameters
of the compensator are obtained from the transfer function of the CCVT model and from
the ideal frequency response of the compensated CCVT. To evaluate the compensation
technique two case studies are analyzed: harmonic distortion at the 230 kV system and
a phase to ground fault at the 230 kV system. Real time simulations attest that the
compensator acts to bring the secondary voltage waveform close to the primary voltage
signal. This may be important for the correct indication of measuring devices and proper
operation of protection systems.
In Reference [29], A method of symmetrical component analysis for the detection of
current-transformer (CT) saturation in a numerical current differential feeder protection
relay is presented in this paper. The performance of the differential relay is investigated
for various faults on a typical Electromagnetic Transients Program/Alternative Transients
Program (EMTP/ATP) simulated transmission feeder. The simulator includes the effects
of CT saturation. A comparison between simulation and tests conducted on an analog
model testbench are also evaluated. The results show a high degree of similarity and
40
illustrate the effect that CT saturation imposes on the sensitivity and stability of the
protection scheme. An algorithm is presented that shows significant improvement in
sensitivity on internal faults while still maintaining a high level of stability on external faults
and nonfault events.
In Reference [30], a current-transformer (CT) model is presented that is useful for low-
frequency applications. To describe the iron-core magnetic behavior, a hysteresis model
is proposed, which is able to generate minor asymmetric loops and remanent flux. The
effects of classic eddy current losses and anomalous losses are represented by linear
and nonlinear resistors, respectively. The obtained results are compared with those
calculated by the Preisach’s model and measured in the laboratory. This model may be
applied in power system protection studies, as it is the case of numeric correction of
distorted secondary currents in current transformers (CTs).
In Reference [31], a nonlinear electric circuit is presented to describe the behaviour of
magnetic cores in low-frequency applications. A method of hysteresis modelling is
proposed, which is able to reproduce minor loops and remanent flux. Classic eddy-current
losses and anomalous losses are represented by a linear resistor and a non linear
resistor, respectively. The obtained results are compared with those calculated by the
models of Hodgdon and Preisach. Comparisons with experimental results are also
accomplished. This model is useful for transformer transient phenomena studies, such as
inrush currents and ferroresonance.
Reference [32] advances the applications of EMTP for studying current transformer (CT)
performance under faults. CT’s models in EMTP may require hard-to-get information at
early stages of power system projects. The paper examines approaches to make
appropriate assumptions for missing model information and tests the sensitivity of EMTP
results to such assumptions. By presenting case studies the paper facilitates such
applications of EMTP and supplements an earlier IEEE Work Group on the EMTP
modeling of protection CTs.
In Reference [33], a quicker method to implement the standard CT model in circuit
simulation is introduced. Using the introduced non-linear magnetizing equation with
adjustable parameters, the CT's saturation curve can be instantaneously reproduced
avoiding the process of defining linear segments and determining how many are
necessary. An actual CT is modeled using the suggested method and results from
excitation transients are compared with the traditional piecewise linear model as well as
with experimental results for verification.
In Reference [34], a method is proposed to model the non-linear magnetizing branch of
a CT by visual inspection of the hysteresis loop of a core. Three constants on the
hysteresis loop reveal enough information about the core loss and non-linear inductance
that power measurements and data-pairs can be avoided. An experimental hysteresis
loop is taken from a CT and its model put together by visual inspection. The visual method
presented is simulated and verified against an experimental hysteresis loop using
SimPowerSystems as the circuit simulator.
41
In Reference [35], detailed analysis of a CT under transient conditions are performed.

The equivalent circuit parameters of a real CT are determined by means of finite element
method (FEM) and verified by measurement for measuring and protective part of CT. A
developed mathematical model of CT with nonlinear B-H curve is then used in a circuit
simulation to study different transient conditions under normal and fault operation. A
comparison to the measured signal on a real power line during fault condition is also
given.
Reference [36] presents mathematical models for two kinds of CCVTs with various states,
i.e., zero load state, normal load state, short-circuit state of the primary side, and short-
circuit state of the secondary side. Digital simulation models of the CCVT corresponding
to the various states by using an adequate digital integration method are developed.
Using the basis of these models, a digital simulation of the CCVT on a PC is conveniently
realized and transient characteristics of the CCVT are researched on a PC according to
the national lab standard, GB4703, of PR China. Details are demonstrated with the aid of
a numerical example. The results indicate that the mathematical models and the digital
simulation models are in good agreement.
Reference [37] reports a CCVT calibration effort for harmonic voltage measurement
purpose. A set of frequency response studies are conducted on two types of 345 kV
CCVT. The Electromagnetic Transients Program (EMTP) is used to develop the CCVT
model and conduct this digital simulation. The accuracy of the CCVT frequency response
study result is verified through comparison of the EMTP digital simulation results with
those obtained from field measurements. The comparison results demonstrate that the
developed model can accurately predict CCVT frequency response, and can provide
good adjustment values for harmonic measurement purpose.
Reference [38] discusses a method for CT calibration.
D.2 Detailed methods from selected papers

The model in [13] is based on the Jiles–Atherton (JA) phenomenological model of a
ferromagnetic core. Presently used models use mathematical techniques which
approximate the behavior of the hysteresis loop for recoil loops and remanence effects.
For many studies these models are quite adequate except where accurate values of
remanence are required following a circuit interruption with a subsequent reclose into a
live circuit. This feature is particularly important in relay studies where it is necessary to
simulate successive faults due to reclosing onto a permanent fault. The present methods
can adequately handle such situations when the time delay between the faults is very
short (<1 s), but for longer durations the remanent flux is not modeled properly.
The Jiles-Atherton model
The Jiles-Atherton (JA) model describes the relationship between the magnetic moment
M and the magnetic field intensity H:
42
The equation to be solved to construct M/He and B/H loops is
Where,
𝛼,𝑐,and 𝑘are constants for the material being used and𝛿takes the value 1 or -1 based on
the sign of 𝑑𝐻/𝑑𝑡.
Current transformer model
In modeling current transformers in an electromagnetic transient simulation program

(EMTDC, EMTP), the numerical problem to be solved is to determine the change in
secondary current for a given change in primary current for each time interval, 𝛥𝑡. The
following equations are provided.
where 𝑁𝑝 is the number of turns in the primary; 𝑁𝑠 is the number of turns in the secondary;
𝑙is the length of the flux path, 𝛥𝐻is the change in magnetic field strength, 𝛥𝐼𝑝 is the change
in primary current and 𝛥𝐼𝑠 is the change in secondary current. Area is the cross section
area of the core of the CT, 𝛥𝑡is the integration time step, 𝛥𝐵is the change in flux density
in the core, 𝑅𝑠 is the sum of secondary winding and burden resistances (including the lead
resistance), 𝐿𝑠 is the sum of secondary winding leakage and burden inductance (including
the lead inductance), 𝐼𝑠𝑜𝑙𝑑 is the secondary current during the previous time step.
To better model the shoulder region of the B-H loop, a modified anhysteretic function
was introduced as follows:
The equations are iteratively solved for the unknows in each time step.
43
The proposed CT equivalent circuit in [30] is shown in Fig. D.1.
Fig. D.1. The proposed CT equivalent circuit
R and L represent the resistance and the inductance of the secondary (winding and load).
The nonlinear inductor is associated with the saturation and hysteresis effects of the
magnetic core. The effects of classic eddy current losses and anomalous losses are
represented by 𝑅𝑒 (linear resistor) and 𝑅𝑎 (nonlinear), respectively. The following equation
can be obtained:
Based on integration, it is obtained
Using the trapezoidal integration rule and considering a time step h, the discretized
equations of the circuit are written as follows:
44
The nonlinear system is solved by use of the Newton-Raphson method using a specified
time step, say 1 microsecond in this work.
In Reference [27], the basic electrical diagram of a typical CCVT is shown in Fig. D.2.
Fig. D.2. Basic electrical diagram for a typical CCVT
The primary side consists of two capacitive elements C1 and C2 connected in series. The
potential transformer provides a secondary voltage vo for protective relays and measuring
instruments. The inductance Lc is chosen to avoid phase shifts between vi and vo at
power frequency. However, small errors may occur due to the exciting current and the
CCVT burden (Zb). Ferroresonance oscillations may take place if the circuit capacitances
resonate with the iron core nonlinear inductance. These oscillations cause undesired
information transferred to relays and measuring instruments. Ferroresonance
suppression circuits (FSC) tuned at power frequency (L in parallel with C) and a
45
resistance to ground have been used to damp out transient oscillations requiring a small
amount of energy during steady state.
The diagram shown in Fig. D.2 is valid only near power frequency. A model to be
applicable for typical frequencies of ferroresonance and frequencies up to a few kilohertz
needs to take into account the PT primary winding and compensating inductor stray
capacitances at least. In this work, the circuit shown in Fig. D.3 was used to estimate the
linear CCVT parameters. It comprises a capacitive column (C1, C2), a compensating
inductor (Rc, Lc, Cc), a potential transformer (Rp, Lp, Cp, Lm, Rm) and a ferroresonance
suppression circuit (Rf, Lf1, Lf2, −M, Cf).
Fig. D.3. CCVT model for estimation of parameters
Fig. D.4. (a) FSC design and (b) FSC digital model.
The FSC design is shown in Fig. D. 4a. A nonsaturable iron core inductor Lf is connected
in parallel with a capacitor Cf so that the circuit is tuned to the fundamental frequency with
a high Q factor. The FSC digital model is shown in Fig. D.4b. The damping resistor Rf is
used to attenuate ferroresonance oscillations.
46
Linear CCVT parameters estimation
The circuit shown in Fig. D.3 can be replaced by Fig. D.5 with impedances Z1, Z2, Z3,
Z4, Z5 and all elements referred to the PT secondary side. The expressions for the
referred impedances in the s domain, with s = jω, are:
Z1 = [(Rc + sLc)/r2]//(1/r2sCc);
Z2 = (Rp + sLp)/r2;
Z3 = (Rm/r2)//(sLm/r2); (1)
Z4 = (sLf1 + 1/sCf)//(sLf2);
Z5 = Rf − sM.
where r is the PT ratio and the symbol // denotes that elements are in parallel. The linear
parameters R, L, C were obtained from curve fitting algorithm based on Newton’s method
to match the transfer functions represented by the ratio vo/vi. This fitting technique is an
improvement over the one presented in earlier models because here both magnitude and
phase curves are fitted simultaneously. The technique is described below.
Fig. D.5. CCVT model for development of the analytical expressions.
Minimization technique of nonlinear functions
Methods for minimizing nonlinear functions are usually iterative, that is, given an
approximate solution, an estimate of the solution is obtained. The technique used here is
based on Newton’s method which uses a second-order Taylor series expansion to the
error function F(x) in the matrix/vector form, as follows:
F(x + p) ≈ F(x) + pT∇F(x) + pT∇2F(x)p/2 (2)
where the vector x = [x1, x2, . . ., xm] and p is the increment vector in the direction of the
estimate solution; p is obtained by minimizing the function F setting its gradient to zero
which results in:
∇2F(x)p = −∇F(x). (3)
The approximate solution xk+1 is given by:

xk+1 = xk + p = xk − [∇2F(xk)]−1∇F(xk). (4)
47
A computer routine was developed to minimize the error function F(x) using the method
described above. Besides the function y(ωi; x), it is necessary to know its first and second
derivatives with respect to each parameter of the vector x. The algorithm is described as
follows:
(1) Supply the CCVT frequency response values yi for each frequency ωi and enter with
a guess for the parameters R, L, C (vector x).
(2) Determine F(x) and evaluate F(x + p).
(3) Save the value of F(x + p) and for a user-defined number of iterations t, compare the
actual value of the error function to its old value t iterations before.
(4) If the difference is greater than a user-defined tolerance, go back to step 2. Otherwise,
stop the iterative process.
In Reference [34], an equation based on the atan function can be used to model CTs
allowing for experimental saturation curves to be reproduced with one trace. In doing so
it avoids piecewise linear segments to model the non-linear characteristic of magnetizing
branches. Such equation is given by
where 𝑑 is a flux scaling coefficient, 𝜌 controls the slope of the saturation curve, and ±𝑏
are the current offsets due to the core losses (width of hysteresis loop).
In Reference [36], to restrain ferroresonance, a damper is connected to the secondary

side of the CCVT. There are three kinds of dampers: the resonant type, the fast-saturation
reactor type, and the electronic type, categorized according to different principles. Fast-
saturation reactor type and resonant type dampers are currently dominating the market.
The equivalent circuit is shown in Fig. D.6 and Fig. D.7. Note that the capacitance voltage
divider is not shown in the figures.
Fig. D.6. the fast-saturation reactor type CCVT’s equivalent in the simulation
48
Fig. D.7. The resonance type CCVT’s equivalent circuit in the simulation
49

IEEE Power System Relaying and Control (PSRC) Committee Working Group I26

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

IEEE Power System Relaying and Control (PSRC) Committee Working Group I26

Uploaded by

Copyright:

Available Formats

Mathematical Models for Current, Voltage, and Capacitively Coupled Voltage Transformers

IEEE Power System Relaying and Control (PSRC) Committee

Mathematical Models for Current, Voltage, and

Chair: Michael Meisinger

Vice Chair: Steve Turner

Secretary: Amir Makki

Members and Contributors

This report reviews and presents several mathematical models of instrument

THIS PAGE LEFT BLANK INTENTIONALLY

2. Mathematical Models of Magnetic Core Representation

2.1 EMTP and ATP Inductor Models

For the first linear segment in the  − i curve,

2.2 EMTDC Model

2.3 Seetee Model

In this model, the incremental permeability of the core material,  0  r = dB / dH , is

 0 = permeability of free space 4E-7 H/m.

Ac = cross sectional area of the CT core in m2.

Lc = the mean length of the flux path in m.

2.4 Jiles-Atherton Model

 = interdomain coupling factor,

𝑎 = domain wall density,

where: Man = the anhysteretic magnetic moment, and

For the irreversible wall motion the equation is:

0 = permeability of free space,

k = domain pinning parameter.

where:  = the domain damping parameter.

The following parameters were used in the simulation shown in Figure 2.

 Interdomain coupling = 9.5E-5

 Domain damping factor = 1E+7

0 permeability of free space = 4E-7 H/m

𝑀𝑎𝑛 = 𝑀𝑠 (tanh(𝑀(𝐻))) (24)

where:  = +1 or –1, based on the sign of dH/dt,

𝐻𝑒 = effective field intensity.

where, 𝑎1 , 𝑎2 , 𝑎3 , and 𝑏 are appropriately chosen constants [13].

The derivative of 𝑀𝑎𝑛 with respect to 𝐻𝑒 is given by

In addition, it is proposed that an adjustable 𝑘 be used in (25):

3.1 Equivalent Circuit

Figure 3: Equivalent Circuit of CT.

3.2 Current Transformer Model Data

the CT dimensions. The calculated air-core inductance from the CT dimensions is

3.3 CT Simulation Results

Figure 5: Comparison of EMTP Simulations (solid red line) and

The significant features to be included in a CT model are:

Figure 6: Basic CCVT Model with Equivalent Circuit of VT.

FSC can be designed in two operational modes: active and passive.

Figure 9: Ferroresonance filter circuit: a) Actual Design, b) and c) Digital Models,

Figure 11: Frequency response of CCVT in Linear Region.

H Lcl Magnetizing Inductance = 43.15 H

Figure 13: Frequency response of CCVT in Linear Region.

• Stray-capacitance of the compensating inductor and the primary capacitance across

587.7 • 𝑠 3 + 2.642𝐸5 • 𝑠 2 + 6.485𝐸7 • 𝑠

Figure 14: Transient Response of CCVT Transfer Functions [19].

401.1 • 𝑠 3 + 1.819𝐸5 • 𝑠 2 + 4.464𝐸7 • 𝑠

• Lout is the output load inductance.

• Lext is the external source inductance.

• Lm is the magnetizing inductance.

sequence. From these waveshapes a flux-current hysteretic function can be generated.

Figure 15: Equivalent Circuit of a VT with Hysteretic Core

6. Merging units and Instrumentation Error Correction

[27] D. Fernandes, W. L. A. Neves, and J. C. A. Vasconcelos, “Coupling capacitor voltage

APPENDIX A EMTP Model Setup for CT Saturation

A.2 CT Saturation Data