Академический Документы
Профессиональный Документы
Культура Документы
I. INTRODUCTION
OWER transformers are an important and universal component of power systems, for which various modeling schemes
have been developed for many years. Leibfried and Feser [1] introduced the transfer function concept for power transformers.
This method is well known as an additional method of evaluating the impulse test of power transformers. Wilcox introduced a
time-domain modeling and modal analysis which described how
a new form of transformer model can be converted from the frequency domain into the time domain for Electromagnetic Transients Program (EMTP) implementation [2]. To date none of the
models include an accurate representation of hysteresis as given
by the JilesAtherton (JA) model [3]. The J-A model is an intermediate solution of ferromagnetic hysteresis with some physical
basis and is computationally tractable for practical problems. The
J-A model is well suited to transmission-line modeling (TLM)
for lumped components because it is formulated in terms of a
first-order differential equation.
This paper demonstrates a novel simulation program based
on the TLM method for simulating a power transformer which
include nonlinear hysteresis and eddy-current loss effects [4].
The TLM algorithm is discrete in nature and is compatible with
other time-domain simulation programs such as ATP, PSCADEMTDC, or MATLAB [5]. This is an extension on a previous
published work on modeling of nonlinear inductors displaying
hysteresis [6].
II. TLM MODELING AND APPLICATION TO TRANSFORMERS
The basic TLM technique models linear reactive components
as transmission line segments (called stubs). The stub model
representing the inductor is terminated by a short circuit, to emphasize inductive behavior, current, and hence storage in the
magnetic field must be maximized. The TLM model for a capacitor is a stub with its far end open circuit. It emphasizes voltage
differences, storage in the electric field, and hence mainly capacitive behavior.
Fig. 1. (a) Inductance, (b) equivalent circuit, (c) stub model of inductance,
TLM equivalent circuit
In many applications, it is necessary to account for the nonlinear behavior of inductors and capacitors. These can be modeled as nonlinear stubs. Magnetic coupling between components
may also be described using TLM. Mutual coupling is modeled
by a current-controlled voltage source [7], [8].
A. Linear Inductor Model
An inductor modeled using a transmission line stub is shown
in Fig. 1. The stub model representing the inductor is terminated
across it is
by a short circuit. For an inductor the voltage
. The differential term
is then
equal to
,
replaced in TLM by the discrete transform
is the time step, and
is the incident
where
pulse at the stub terminals. It is assumed that it takes one time step
for the pulse to make a round-trip to travel to the end and be
reflected back as the incident pulse in the next time step . Thus,
is
at any time step the voltage across the inductor
(1)
Also, if the reflected amplitude at the inductor terminals is
then
(2)
1821
C. JilesAtherton Model
The classical J-A model [3], [6] is described in the following
subsections.
1) Weighting Coefficient: The magnetization is split into two
parts, the anhysteretic magnetization and the irreversible magnetization. In normalized form, this is expressed by
where the subscript donates the th time step. For a shortcircuited transmission line stub, the incident pulse in the next
time step becomes
(3)
In a similar manner, capacitance can be simulated. The algorithm can be shown to be equivalent to the trapezoidal integration method adopted by ATP and PSCAD [8].
(9)
is the weighting coefficient with
is
where
the normalized anhysteretic magnetization, and
is the normalized irreversible magnetization.
2) Modified Langevin Function: The anhysteretic magnetization dependence is given by a modified Langevin function,
i.e.,
(10)
where
is the normalized saturation magnetization,
is
is the normalthe interdomain coupling coefficient, and
ized anhysteretic magnetization form factor. The coefficients
are positive constants. Also,
denotes the mod.
ified Langevin function with argument
To avoid difficulties with the modified Langevin function for
small arguments, a linear approximation is used where for
we put
.
3) Differential Equation for the Irreversible Magnetization:
In the J-A model, the derivitive of the normalized irreversible
magnetization w.r.t. the inductor current is
(4)
(11)
Thus, in TLM we can model the nonlinear inductance voltage
as the sum of two voltages given by
(5)
(6)
At each time step , the incident waves are updated for the next
with
time step
is given by
if
if
otherwise.
and
and
(12)
D. Modeling of Transformers
(7)
(8)
The equivalent circuit is then as given in Fig. 2.
1822
Fig. 4. A general ideal transformer equivalent circuits for TLM modeling. (a) Linear transformer. (b) Nonlinear hysteretic transformer core.
The primary and secondary voltages ( and ) can be derived in terms of the primary and secondary currents ( and )
and the magnetization intensity as for the lumped inductance
to give
(13)
where
and
.
1) Ideal Linear and Lossless Transformers: For linear loss, (14) then reduces to
less transformers
and
(17)
(18)
,
where
is the source voltage. The incident TLM stub voltages
and
are calculated as for the linear inductor.
2) Ideal Transformer With Nonlinear Hysteretic Core: For an
ideal transformer with nonlinear and hysteretic core inductance,
the TLM model is as given in Fig. 4(b). There is an extra source
and is given by
term representing the magnetization
(19)
(14)
The magnetization
is nonlinear so that an iterative solution
for the following simultaneous equations has to be found:
where
, and
.
A linear lossless transformer equivalent circuit is then as
given in Fig. 4(a). The controlled sources representing mutual
are as follows:
terms of the type
(15)
(16)
where
(20)
(21)
and
. We have chosen the
where
NewtonRaphson technique for its efficiency and stability so a
1823
(22)
where is the iteration number.
by
(23)
and
is
is
(24)
The iteration is started with initial values taken from the previous time step and continued until suitable convergence criteria
are met. In this work, this is set as
and
and
(25)
(26)
where
is the copper conductivity,
is the mean turn
is the bare wire area, and
is the number of turns
length,
for winding .
The winding leakage inductance for each winding
of a
scrapless lamination core and assuming concentric secondary
and primary windings as shown in Fig. 5 is given by [9]
(28)
where is the primary rms input voltage,
is the lamination
is the volume of the iron core,
is
factor of the iron core,
the iron core area,
is the iron core conductivity, is the lamis the number of turns in the primary
ination thickness, and
winding. It is normal to then simply represent the eddy-current
loss as due to a shunt impedance across the primary winding
given by
(29)
A more complete and representative model of eddy-current loss
is obtained by adding a third single-turn winding loaded by a
given by
resistance
(30)
The full TLM equivalent circuit for a nonlinear transformer including leakage reactance, iron, and copper loss is as shown in
Fig. 6.
There are now three nonlinear simultaneous equations to
and which have the form
solve for
(31)
(27)
The eddy-current losses are due to circulating currents in the
iron core and behave as a third winding the resistance of which
can be calculated from the fundamental of power frequency .
(32)
(33)
1824
Fig. 6. Full transformer TLM model equivalent circuit with nonlinear hysteretic core, leakage inductance, conductor losses, eddy-current losses, and source and
load impedances.
where
and the Jacobian matrix is now
(34)
III. RESULTS
To demonstrate the modeling procedure, a small 25 kVA,
11 kV/220 V power transformer with the parameters as given
in the Appendix was modeled using the TLM method described
in Section II. The JilesAtherton parameters were typical of a
core made of FeSi sheets [10]. The transformer was loaded with
a series resistor and inductor which would provide a 25 kW load
of 0.9 p.f. at 220 V.
As a comparison, the same transformer parameters and
electrical supply and load were modeled using the MATLAB
power system block set with a transformer hysteretic core. The
Fig. 7. TLM simulated B H curve under full load and the supply voltage increasing to rated voltage = 0:0434 s.
Fig. 8. MATLAB simulink simulated B H curve under full load and the
supply voltage increasing to rated voltage = 0:0434 s.
1825
Fig. 10. Comparison of the simulated secondary current using the TLM model
or the MATLAB simulink model for rated load as supply voltage increases to
rated value = 0:0434 s.
Fig. 11. Comparison of the simulated primary current using the TLM model or
the MATLAB simulink model for no load as supply voltage increases to rated
value = 0:0434 s.
TABLE I
RESULTS FOR POWER LOSSES, POWER TRANSFERRED, AND EFFICIENCY
(P INPUT/P OUTPUT %) AT RATED LOAD WITH 0.9 P.F
AND RATED SUPPLY VOLTAGE
Fig. 9. Comparison of the simulated primary current using the TLM model or
the MATLAB simulink model for rated load as supply voltage increases to rated
value = 0:0434 s.
1826
APPENDIX A
Data for 11 kV/220 V, 25 kVA power transformer
m
m
S m
S m
Fig. 12. TLM simulated B H curve minor loops for an initial 0.3 p.u. supply
with a dc pulse is in Fig. 14.
H
H
A m
A m
A m
Fig. 13. MATLAB simulink simulated B H curve minor loops for an initial
0.3 p.u. supply with a dc pulse as in Fig. 14.
Core dimension.
Magnetic path length.
Core area.
Lamination thickness.
Iron conductivity.
Copper conductivity.
Number of turns in primary.
Number of turn in secondary.
Lamination factor.
Window utilization factor.
Mean turn length.
Primary copper loss resistance.
Secondary copper loss
resistance.
Primary leakage inductance.
Secondary leakage inductance.
Saturation magnetization.
H m
Magnetization weighting
factor.
Permeability of the free space.
ACKNOWLEDGMENT
This work was supported in part by the UK EPSRC research
council.
REFERENCES
Fig. 14. Supply voltage waveform used to reproduce hysteretic curves given in
Figs. 12 and 13.
John Paul was born in Peterborough U.K., in 1960. He received the M.Eng. and
the Ph.D. degrees in electrical and electronic engineering from the University
of Nottingham, Nottingham, U.K., in 1994 and 1999 respectively. His Ph.D.
dissertation involved the application of signal processing and control system
techniques to the simulation of general material properties in time-domain TLM.
He is currently a Research Associate with the George Green Institute for Electromagnetics Research at the University of Nottingham. His research interests
are in the application of signal processing techniques for material modeling in
1827
time-domain computational electromagnetics, the simulation of complete systems for electromagnetic compatibility studies, and the interaction of electromagnetic waves with biological tissues.
Okan Ozgonenel was born in Samsun, Turkey. He received the M.Sc. degree
in electrical education from Marmara University in 1992 and the Ph.D. degree
in electrical engineering from Sakarya University in 2001.
He has been with Ondokuz Mayis University, Samsun, Turkey, since 1991,
where he is a Lecturer in the Electrical and Electronics Engineering Department.
His main research interests are digital algorithms, digital signal processing, simulation methods for power transformers, power system control and protection,
and wavelet techniques.