1

Hysteresis Modeling in the Matlab/Power

System Blockset
Silvano Casoria, Member IEEE, Patrice Brunelle, Gilbert Sybille

the hysteresis representation can be of importance in the

implementation of instrument transformers models such as
Abstarct--This paper presents a static hysteresis model current transformers (CT) and capacitor coupled voltage
for the saturable transformer blocks in the Power System transformers (CCVT) [3].
Blockset (PSB), a MATLAB/Simulink-based simulation Saturation and hysteresis modeling have been the subject
tool. The model defines a relation between the flux (Φ) and of many papers with different approaches depending on the
the excitation current measured in dc (when the eddy application. Emphasis, may be put on the empirical or on the
current losses are not present). It is useful under transient analytical nature of the expressions (e.g. exponential,
conditions and can represent minor loops. The model hyperbolae, polynomials, arctangent) used for the operating
exhibits all the main features of hysteresis and is extended point (Φ, Ι) trajectory representation.
to represent also the saturation characteristic. A hysteresis Importance can also be given to the algorithm (e.g.
design tool consisting of graphical user interface (GUI )
iterative for more precision, non iterative for speed of
allows precise adjustment to any major loop or saturation
execution) that will dictate the behavior of the operating point
characteristic. An example case is described to illustrate
trajectory during transients (e.g. for the formation of the minor
the application of the model.
loops).
The proposed model uses a semi-empirical characteristic,
Index Terms-- Hysteresis, Modeling, Graphical user interface. by having an arctangent analytical expression representing the
operating point trajectories [4]. The expression parameters are
deduced by curve fitting empirical data defining the major
I. INTRODUCTION loop or the single-valued saturation characteristic. A
hysteresis design tool consisting of a graphical user interface
T HE MATLAB/Power System Blockset (PSB) is a
electromagnetic transient program capable of representing
typical power equipment such as transformers, lines,
(GUI) allows precise adjustment to any hysteresis major loop
or saturation characteristic. The algorithm selected for
modeling the trajectories behavior and the formation of the
machines, and power electronics to simulate power systems
hysteresis minor loops is based on the one used in the
[1,2]. The blockset uses the Simulink® graphical
Electromagnetic Transients Program (EMTP) [5]. We have
environment, allowing a model to be built using simple click
improved the algorithm by adding the possibility of unlimited
and drag procedures.
number of embedded minor loops formation.
The PSB saturable transformer model, has the magnetizing
characteristics of the core modeled by a resistance (Rm)
simulating the core active losses (eddy currents and hysteresis
II. HYSTERESIS MODEL
losses) and a saturable inductance in parallel. The saturation
characteristic is specified as a piecewise linear characteristic The static model of hysteresis defines the relation between
of a nonlinear relation between the flux and the magnetization the flux (Φ) and the magnetization current (i.e. the current
current. For increased accuracy, modeling of the hysteresis through the nonlinear inductance in the model) that is equal to
phenomenon is introduced in version 2.3. the total excitation current measured in DC (when the eddy
The effect of hysteresis and eddy losses are of significance current losses are not present).
in the studies of: subsynchronous oscillations where system The fundamental characteristics of the model [6] are:
losses appears to be determinant; residual flux effects on the 1. The symmetric variation of the flux produces a
transformer inrush currents; ferroresonance phenomena; symmetrical current variation between –Imax and Imax,
harmonics generated by half-cycle saturation. Furthermore, resulting in a symmetrical hysteresis loop whose shape
and surface depend on the value of Φmax as shown in
S.Casoria, G. Sybille are with Institut de recherche d’Hydro-Québec, Fig. 1. The major loop is produced when Φmax is equal
1800, boul. Lionel-Boulet Varennes (Québec) J3X 1S1, Canada (e-mail: to the saturation flux (Φs). Beyond that point the
casoria.silvano@ireq.ca, sybille.gilbert@ireq.ca). characteristic reduces to a single-valued saturation
P.Brunelle is withTransÉnergie Technologies Inc, 740, Notre-Dame
Ouest Montréal (Québec) H3C 3X6, Canada (e-mail: characteristic, which is asymptotic to the air core
p.brunelle@transenergie-tech.com ). Tel. : (514) 282-8401 p. 203 inductance (Ls). The remanent flux (Φr), the coercive
Fax. : (514) 282-8402
 2

current (Ic) and its slope (dΦ/dI) depend on the core Where sgn = 1 for an ascending trajectory and –1 for a
material. descending trajectory.
The parameters α, c, b are related to the residual
(remanent) flux Φr, coercive current Ic, and slope dΦ/dI at the
coercive current as shown in Fig. 1. The parameters a and e
are related to the saturation point coordinates Is and Φs. These
parameters are calculated by solving a set of non-linear
equations during the parameterization step, presented in the
next section, using the hysteresis design tool.
During simulation, the model uses the inverse relation I(Φ)
since the independent or input variable is the flux. The inverse
relation is defined as a series of N equidistant points
connected by line segments. A table containing the
coordinates of these points (In, Φn) and the segments slope is
defined by solving iteratively a set of non-linear equations.
B. Minor loops formation
To generate the closed minor loops we apply the
fundamental characteristic 3 given before. It stipulates that
Fig. 1. Hysteresis loops. each curve tends to return to the previous-to-last point of
reversal, e.g. reversal point 1 on the ascending half of the
2. In transient conditions an oscillating magnetizing current
major loop in Fig. 3. Effectively, after detecting a reversal at
will produce minor asymmetrical loops, as shown in Fig.
point 2, a new trajectory 2-1 is calculated. First, Dmax, which
2 and all points of operation are assumed to be within the
major loop. Loops once closed have no more influence on is the distance between the reversal point 2 and the
the subsequent evolution. appropriate half of the major loop is calculated. In addition,
3. Inside a minor loop, the magnetizing curve depends only the vertical distance Dmin between 1 and the major loop is
on the last two reversal points, and each curve tends to defined (0 in this example). By assuming, that the vertical
return to the reversal point previous-to-the last (e.g. the distance D decreases linearly with Φ, that is the independent
evolution 1-2-3-4-3-5-2-6-1) in Fig. 2. variable, all the points of the trajectory are then defined. A
linear relation D(Φ) of D as function of the flux is found at
each reversal point and is stored in memory along with the
coordinates of this point. Memorization is important in order
to represent properly the behavior of embedded loops as the
evolution on Fig. 2 illustrates.

Fig. 2. Internal minor loops.

A. Arctangent basic characteristic

The minor loops are derived from the major loop ascending
(i.e. dΦ/dt > 0) or descending (i.e. dΦ/dt < 0) trajectories Fig.3. Formation of minor loop trajectory.
applying simple rules adapted from [4]. The basic analytical
When a reversal point is overtaken, like point 3 in the
expression Φ(I) representing a major loop half cycle is given
evolution 3-4-3-5, the parameters of the trajectories 3-4-3 are
by (1).
discarded and the curve continues on the trajectory 2-1,
Φ = − sgn[a ∗ arctan(− sgn∗ b ∗ Ι + c ) − sgn∗ α ∗ Ι + e] (1) already in memory. Point 1 will not be reached since the flux
 3

reverses at point 5. Next, the reversal point 2 is overtaken in value and consequently the condition is not detected. In series
the evolution 5-2-6, and the parameters of the trajectories A, the reference is such that the condition is detected and the
constituting the loop 2-5-2 are discarded. Finally, the curve procedure applied.
continues on the trajectory 6-1.
C. Special conditions
The assumption of linear dependence of D(Φ) on Φ can lead
to the difficulty where for certain values of F, the minor loop
trajectory would lie outside the major loop, as in Fig. 4. To
avoid this, the trajectory is tested and constrained to follow
the major loop. This situation arises only near the saturation
points and the small correction introduced to the trajectory
would not significantly affect the model accuracy.

Fig. 5. Very small minor loops (series B) are approximated by line segments
(series A).

D. Initial trajectory
At the beginning of the simulation the initial trajectory
must be specified. The trajectory depends on the operating
point past evolution, i.e. by the last two reversal points.
Assuming that the last reversal point is situated on the major
loop, the initial trajectory would be determined by the
Fig. 4. Solution to a probable illegal operating condition.
residual flux value and on the reversal point position on either
Two other operating conditions should be identified in the ascending or descending half. An increasing flux initially
order to avoid generating superfluous internal loops or indicates that the reversal point is on the descending half and
trajectories. In the hysteresis design tool, two tolerances can vice versa for a decreasing flux. Figure 6, illustrates the two
be adjusted: TOL_F (% of Fs) and TOL_I (%of Ic). These possibilities (trajectory A and B) for a residual flux of 0.1 pu.
parameters trigger detection of one of the following two
conditions:
1. Closed minor loops in steady state.
2. Very small minor loops, whose internal surface may be
neglected.

The first condition will be assumed if the Φ value of the

actual reversal point is in close proximity to the previous-to-
last reversal point Φ value, i.e. within an adjusted tolerance
TOL_F (default = 0.1% of Φs). In that case, no new trajectory
is calculated and the operating point will remain on the same
loop. This may be illustrated by the potential evolution 3-4-3-
4 in Fig. 2, forming a closed loop in steady state.
The second condition will be assumed when the distance
between the Ι coordinate of the actual point of reversal and the
previous-to-last is less then a threshold value TOL_I (default
= 0% of Ic or null). If such a condition is detected, the
evolution within theses two points will follow a trajectory Fig.6. Choice of the initial trajectory (A or B) is based on the
defined by a line-segment. This is illustrated in Fig. 5 by the residual flux and the last reversal point on the major loop.
two series, A and B, of imbedded minor loops, with similar The direction of the flux can be determined during the
current evolution, only shifted vertically. In series B steady state solution when the transformer is connected to the
(evolution 1-2-3-4-3-1), the reference is adjusted to a high network, otherwise an increasing flux is assumed. The user
 4

may specify the residual flux value or it is automatically IV. MODEL IMPLEMENTATION
adjusted so that the simulation starts in steady state. The hysteresis and saturation characteristic of the Saturable
Transformer of PSB is built with Simulink blocks. The Fig. 8
III. MODEL PARAMETERIZATION illustrates the corresponding electrical model for one phase
The parameters of the hysteresis major loop and the single- implemented into the Power System Blockset. The winding
valued saturation characteristic are specified in the hysteresis resistances and leakage inductances as well as the core active
design tool user interface available with the Power System losses resistance Rm of the saturable transformer are modeled
Blockset. The remanent flux, the saturation flux and current by state variables. This state-space model is incorporated into
points, the coercive current, and the slope of the flux at the the global linear state-space system of the rest of the network
coercive current point can be entered in pu or in SI units. In connected to the transformer terminals.
addition, two vectors of current-flux pair of points specify the
saturation characteristic. The graphical user interface tool is
illustrated in Fig. 7.

Fig.8. Electrical model of the Power System Blockset saturable transformer.

The magnetizing branch is modeled as a current source

where the current is computed from the flux obtained by
integrating the voltage across the magnetization branch
(Vmag). The block diagram of Fig. 9 shows the Simulink
implementation of the model.

Fig.7. Graphical user interface for the design of the hysteresis and saturation
characteristic.

The hysteresis characteristic is displayed in the interface

and it can be updated at any time during the specification of Fig.9. Simulink model of the hysteresis implemented in the Power System
the parameters. The flux and currents parameters are also Blockset.
displayed on the graph for a fast and convenient design. The The Power System Blockset provides the magnetization
characteristics are saved in MATLAB data files for a later use voltage of the magnetizing branch to the hysteresis model at
with the hysteresis transformer model. As many characteristics each simulation time step and the model returns the
as necessary can be generated and saved in order to be used corresponding current that is injected in the rest of the
by different transformers in the simulation model. network system.
A special interface option allows the user to view an The major hysteresis loop and the saturation characteristic
interactive simulation of the minor loop formation. The are implemented with look-up tables for a fast an efficient
hysteresis design tool performs an animation of the flux- simulation. The upper and lower major trajectories are
current evolution based on the trajectory initial and final flux modeled separately and are controlled by a Simulink switch
as specified by the user. based on the sign of the flux. A second switch allows the
The two tolerances TOL_F and TOL_I can also be model to pass from the major loop calculation to the inner or
specified in the interface and they are saved in the data file minor hysteresis loops calculation.
with the other hysteresis parameters. The Simulink S-function is in the heart of this process. Its
algorithm controls the transition between the major and the
minor hysteresis loops, calculates the minor loops trajectories,
 5

and outputs the magnetization current. 3. At first zero crossing after the breaker opening order, the
The parameters that define the hysteresis model and also current is interrupted and a flux of 0.83 pu stays trapped
the saturation region are passed to the S-Function via a in the transformer core.
MATLAB data file. When the simulation is started in 4. The breaker is reclosed at t = 9 cycles, at a zero crossing
Simulink, the S-function first calculates the initial trajectory of source voltage, producing an additional flux offset of
based on the initial flux and on the sign of the initial approximately 1.02 pu. The peak flux now reaches 1.85
magnetization voltage across the magnetizing branch. pu, driving the transformer into the saturated region. Peak
The S-function continuously monitors the flux and controls magnetization current is 0.81 pu.
the Simulink switches when the flux leaves the minor loop
region to enter in the saturation region or when the flux leaves
the major hysteresis loop to enter in the minor loop zone. The
function also calculates the trajectories and evolution of the
internal minor loops and outputs the current as a function of
the input flux.

V. APPLICATIONS EXAMPLE
Fig.10. illustrates the simulation of the hysteresis in a
saturable transformer block of the PSB. Phase A of the
primary winding is connected on a 500 kV network, and the
secondary is not connected to any load. The transformer is
rated 500 kV/230 kV, 450 MVA (150 MVA per phase) and
the flux-current saturation characteristic of the transformer
was modeled with the hysteresis design tool. The air core
inductance, i.e. the last segment of the saturation
characteristic, is 0.4 pu.
A programmable source is used to vary the internal voltage Fig.11. Simulation results for the flux and magnetization current of the
of the equivalent 500 kV network. During the first 3 cycles saturable transformer.
the source voltage is programmed at 0.8 pu. Then at 50ms, the
voltage is increased to 1.1 pu. In order to illustrate remanent
VI. CONCLUSIONS
flux and inrush current at transformer energization, the circuit
breaker initially closed is first opened at 0.1 s, and then it is The paper presents a model of the static hysteresis effect
reclosed at 0.15 s. of the ferro-magnetic core in power transformers. This model
is suitable for simulation of switching transients and has the
capability of representing minor loops. It is conceived
according to principal characteristics of transformer core
obtained experimentally and given in the literature. The
model, that can also represent the saturation characteristic, has
been implemented in the latest Matlab/Power system Blockset,
and included as an option in the Saturable Transformer model.
Fig.10. Power System Blockset diagram of the application example.
The model is based on the characteristics of the major
The initial flux in the transformer is set at zero and the hysteresis loop out of which the internal trajectories are
source phase angle is adjusted at 90 degrees so that the flux defined using the translation principal and a linear
remains symmetrical around zero when simulation is started. compensation to generate closed loops. An arctangent relation
The following observations can be made on the simulation between the flux and the exciting current is defined. The
results of Fig.11: expression parameters are deduced by curve fitting empirical
1. From 0 to 0.05 s, the voltage and flux peak values are at 0.8 data defining the major loop or the single-valued saturation
pu. Typical square wave of magnetization current can be characteristic. A hysteresis design tool consisting of a
observed. As no remanent flux was specified, the graphical user interface (GUI) allows precise adjustment to
magnetization current and the flux are symmetrical. The any hysteresis major loop or saturation characteristic.
flux travels on minor loops. The major hysteresis loop and the saturation characteristic
2. From 0.05 to 0.1 s, the voltage is raised to 1.1 pu. The are implemented with look-up tables for a fast an efficient
flux now reaches 1.1 pu and it travels on the main simulation. A Simulink S-function contains the algorithm
hysteresis loop. Current pulses appear on the controlling the transition between the major loop and the
magnetization current indicating beginning of saturation.
internal minor loops, calculates the minor loops trajectories,
and outputs the magnetization current.
 6

An application example illustrates the simulation of

hysteresis in a transformer with no load by showing the
remanent flux at isolation and the inrush current at
energization.

VII. REFERENCES

[1] G. Sybille, P. Brunelle, H. Le-Huy, L.A. Dessaint, K. Al-Haddad,

“Theory and Application of Power System Blockset, a
MATLAB/Simulink-Based Simulation Tool for Power Systems,” IEEE
PES Winter Meeting Conference Proceedings, 2000.
[2] Power System Blockset, User’s guide, Version 2. The MathWorks Inc.,
February 2000
[3] Nickie Menemenlis, Yassine H.-Maharsi, “Real-Time Implementation of
a CCVT Including a Noniterative Dynamic Hysteresis Model,”
ICDS’97, Montreal, Quebec, Canada, May 28-30, 1997, pp.111-116.
[4] M. M. Hassani, "Simulation numérique en temps réel des
transformateurs de puissance," Ph.D. dissertation, Dept. Electrical Eng.,
Univ. Sherbrooke (Québec), Canada, 1992.
[5] J. G. Frame, N. Mohan, Tsu-huei Liu, “Hysteresis modeling in an
Electro-Magnetic Transients Program,” presented at the IEEE PES
winter meeting, New York, New York, January 31-February 5, 1982.
[6] S. Casoria, M.M. Gavrilovic, Xuan-Dai Do, “A Model of the
Transformer Core Hysteresis for Digital Simulation of Electromagnetic
Transients in Power Systems,” IMACS TC1-IEEE International
Symposium, Laval University, Quebec City,1987.

VIII. BIOGRAPHIES

Silvano Casoria (M’1978) was born in Cairo,

Egypt in 1953. He received the B.Ing. and M.Sc.A
degrees in Electrical Engineering from École
Polytechnique de Montréal in 1977 and 1981
respectively. Since 1981, he is with Institut de
recherche (IREQ) at Hydro-Québec as a research
engineer in the Laboratoire Simulation de réseaux.
From 1988 to 1989, he was a trainee at ASEA
Brown Boveri (ABB) plant in Ludvika, Sweden.
During this period, he developed software
subsystems implemented in the control and protection system for the Québec-
New England HVDC multi-terminal transmission system. His area of
professional interests include simulation and study of HVDC transmission and
control systems, model development in transient analysis programs like
EMTP, EMTDC, HYPERSIM, PSB.
Patrice Brunelle was born in Quebec City, Canada, in 1968. He received
the B.Sc. degree in Genie Unifié in 1992 from the Université du Québec a
Chicoutimi, Chicoutimi, Canada, and the M.Sc. degree in electrical
engineering in 1994 from University Laval, Ste-Foy, Canada. Since 1997, he
is with TransEnergie Technologies, a subsidiary of Hydro-Québec Trans
Energie.
Gilbert Sybille obtained his electrical engineering degree in France in
1970 and his M.Sc. from Laval university (Quebec) in 1978. In 1978 he joined
the Power System Simulation department of Institut de recherche d'Hydro-
Québec (IREQ) as a research engineer. He has been project leader in many
simulation studies where he developed an expertise in real time testing of
FACTS controllers. He has also developed various models and softwares for
the IREQ's real time simulator. He is the technical leader and one of the main
contributors in Power System Blockset which is jointly developed by The
Mathworks and Hydro-Québec.