IEEE Transactions on Energy Conversion, Vol. 14, No.

1, March 1999

44

Saturation Modelling in D-Q Axis Models of

Salient Pole Synchronous Machines

E.Levi, Member IEEE

Liverpool John Moores University
School of El. Eng., Elec. & Physics
Liverpool L3 3 A F , UK
Abstract - Presently available transient d-q axis models of
saturated salient pole synchronous machines are derived by
selecting either all the winding currents o r all the winding flux
linkages as state-space variables. This paper presents a number
of novel models where state-space variables are selected in
various different ways. The idea originates in the generalised
flux and generalised inductance concept, which has recently
been successfully developed and applied in derivation of
numerous models of saturated smooth air-gap a.c. machines.
The concept is here extended to salient pole synchronous
machines. Saturation itself is accounted for by means of single
saturation factor approach and conversion of anisotropic to
isotropic machine is executed in the usual way. The new models,
presented in the paper, yield the same accuracy as the existing
models that are based on single saturation factor approach. This
statement is confirmed by a simulation study and a sample of
simulation results is included. However, a number of new
models are considerably simpler than the existing ones and thus
more convenient for application in simulations.

I. INTRODUCTION
Magnetic saturation in orthogonal d-q axis models of
salient pole synchronous machines is traditionally modelled
by selecting either all the winding currents as state-space
variables [l-71, or all the winding flux linkages as state-space
variables [1,8-111. Most of the research in conjunction with
saturation representation in salient pole synchronous machine
models has been directed towards investigation of the crosssaturation impact on accuracy of the simulation results and
towards accurate representation of the q-axis saturation. The
trend in the past was to ignore completely cross-saturation
and account for the saturation at first in the d-axis only [8],
PE-095-EC-1-09-1997 A paper recommended and approved by the
IEEE Electric Machinery Committee of the IEEE Power Engineering
Society for publication in the IEEE Transactions on Energy
Conversion. Manuscript submitted May 21, 1997; made available for
printing October 3, 1997.

and then in both d- and q-axes [3,12]. The importance of

accounting for cross-saturation was recognised later and the
current state-space models that are used nowadays [1,2 4-71
include both the q-axis saturation and the cross-saturation
representation in an explicit form. The current state-space
model appears to be more popular for two reasons: winding
currents are of interest rather than flux linkages, and it
explicitly accounts for cross-saturation so that the effects of
self- and cross-saturation can be separated.
An intrinsic difficulty in saturation representation in d-q
axis models of salient pole synchronous machines appears to
be determination of the q-axis saturation curve. Magnetising
curve for the d-axis is always available as it is given with
open-circuit (no-load) d-axis characteristic. If the machine is
such that the q-axis magnetising curve can be determined
experimentally or if it is determined by means of finite
element analysis, then two-saturation factor approach to
saturation representation may be utilised [ 1-4,6,7,13].
However, in most practical cases q-axis magnetising curve is
not available and one then assumes that the degree of q-axis
and d-axis saturation is the same. As discussed in [ 121, q-axis
saturation effects are adequately accounted for with the use of
open circuit d-axis saturation curve. This approach leads to
single saturation factor representation of saturation
[3 ,5,10,11,13,14].
In order to represent saturation by means of a single
saturation factor, an anisotropic salient pole machine is
converted into an equivalent isotropic machine [5,7,10,14].
Single saturation factor method is adopted here and
conversion of anisotropic to isotropic machine is performed as
defined in [14].
The aim of this paper is to extend recently introduced idea
of arbitrary selection of state-space variables in saturated a.c.
machine modelling, based on the generalised flux and
generalised inductance concept [ 151, to the modelling of
saturated salient pole synchronous machines. The procedure,
originally developed for single-cage induction machines [ 151,
has already been successfully applied in modelling of
saturated double-cage and deep-bar induction machines
[16,17], and smooth air-gap synchronous machines [ 181. As
shown in [15-171, it is possible to derive an appropriate
model for single-cage and double-cage induction machines,

0885-8969/99/$10.00 0 1997 IEEE

45

for any set of selected state-space variables, using the

generalised flux concept. In the case of a smooth air-gap
synchronous machine, the majority of possible models can be
obtained as well, using the same procedure: however, three
models cannot be derived due to the absence of the q-axis
excitation winding [18], so that the total number of obtainable
models equals 37. In the case of a salient pole synchronous
machine, the total number of available models reduces further
due to saliency, so that 34 models can be derived (including
the two existing models).
The derivation procedure is described in the paper and
three novel models are presented in developed form. Some of
the models are considerably simpler than the existing two
models. A simulation study is performed and it is verified that
the novel models yield the same results as the existing ones.
A sample of simulation results is included. The accuracy of
all the possible models of a salient pole saturated synchronous
machine is thus the same, as long as the single saturation
factor approach is used.
The paper is organised as follows. Salient pole synchronous
machine model, single saturation factor approach and an
overview of the concept of the generalised flux and
generalised inductance are provided in Section 11. The
extension of the procedure to the saturated salient pole
synchronous machine modelling and the classification of the
models are contained in Section 111. Derivation procedure is
discussed and three complete models are given in Section IV.
Section V presents simulation results, Section VI discusses
properties and applicability of the novel models, while
Section VI1 summarises conclusions.
11. SALIENT POLE SYNCHRONOUSMACHINE
MODELAND THE
APPROACH
TO MAINFLUXSATURATION MODELLING

A. Basic Salient Pole Synchronous Machine Model

The model of a salient pole synchronous machine may be
given in the rotor reference frame (rotor electrical speed is o)
with the following set of equations:

where v, i and denote voltage, current and flux linkage,

respectively. Indices s, f and r stand for stator, excitation and
damper winding. Torque and mechanical equations are of no
interest and are omitted. There is one damper winding in qaxis and leakage flux between the two rotor windings in daxis is zero. It is assumed that leakage flux saturation and
main flux saturation may be treated independently, as
indicated by (4). Index m denotes parameters and variables
associated with the magnetising flux, while index CT describes
constant leakage inductances. As the machine is of salient
structure, damper winding parameters hffer for the two axes.
For the same reason, space vectors of magnetising flux and
magnetising current are spatially displaced. The open circuit
d-axis magnetising curve of the machine is assumed to be a
known non-linear function: generator open circuit test is
performed and dmrete data are approximated with a
continuous function using, say, least-squares fitting.
Inductance terms Ldm and L, denote saturated values of
mutual inductances along d- and q-axis.
B. Single Saturation Factor Approach
Let the unsaturated and saturated values of mutual d- and
q-axis inductances be Ldmo,L,,
and Ldm , L,, , respectively.
Saliency factor is defined as

-/,

= JL9,IL.,= const.

m=

(5)

and is, according to (3, assumed to be constant at all

saturation levels. It is hence assumed that d- and q-axis
saturate to the same degree.
The anisotropic salient pole machine is further converted
into an equivalent isotropic machine, in which newly defined
values of magnetising flux and magnetising current space
vectors are aligned. Magnetising flux and magnetising
current equivalents in an isotropic machine are defined by
means of saliency ratio as

~,=,/w
i p = , / M ,

(6)

so that it becomes possible to define a unique mutual

(magnetising) inductance,

L, =w,/i,
vf = Rfif + d y f / d t
0 = Rdridv+ d V d r / d t
0 = Rqriq,+ d V q y l d t

Vds = Ld'ds

'Vdm
+ Vdm

(7)

(2)

The open circuit d-axis magnetising curve is used as function

y , = f(iJ and magnetising flux components of (4) are now

(3)

Vdm = Ldmidm = L p

(ip)idm

Vqm = Lqmiqm = ~ , (i,u)mziqm

u

Vqs = Ldqs + Vqm

V f = Lofif 'Vdm

Vdv = LDdridr

C. Generalised Flux and Generalised Inductance Concepts

Vqv =

Lcrqjqv + Vqm

Vdm = L d J d m

Vqm = LqJqm

idm= ids + i f +idr

iqm= i,, +iqr

(4)

Generalised flux space vector is defined for a smooth airgap a x . machine as a flux space vector that has to satisfy two
criteria [15-171. It is required to be aligned with magnetising

46

flux space vector (or, which is the same in a smooth air-gap

machine, with magnetising current space vector) and it must
be possible to express generalised flux as a linear combination
of the selected state-space variables. Correlation between
generalised flux space vector and the magnetising current
space vector is established by means of the generalised
inductance A, that depends on the level of main flux
saturation. Generalised inductance is always a physically
definable inductance of the machine for a single-cage
induction machine, while generalised flux may or may not
exist physically in the machine [15]. For a double-cage
induction machine and a smooth air-gap synchronous
machine, there are cases when neither generalised inductance
nor the generalised flux physically exist in the machine [ 16181. However, classification of all the models and their
derivation can always be based on the need to find the time
derivative of the generalised inductance, the time derivative
of the inverse of the generalised inductance or neither of the
two [15-181. In the case of the smooth air-gap synchronous
machine three models cannot be derived because of the
absence of the q-axis excitation windmg.
111. MAINFLUXSATURATION MODELLING
IN SALIENT POLE
SYNCHRONOUS
MACHINES
A . Application of the Generalised Flux Concept

Space vector notation (underlined variables) is used in

what follows for all the quantities that possess both d- and qaxis components, for the sake of abbreviated writing. The set
of space vectors that may be selected as state-space variables
is given with

It

:
(9)
Three space vectors are to be selected from this set of eight.
Let the selected ones be denoted as x1,x2,x3.The
generalised flux space vector definition, used in [15-181 for
smooth air-gap machines, has to be modified for a salient
pole machine. Generalised flux is now required to be aligned
with magnetising current space vector equivalent
in the
[XI

=[is

if

I,

!r

ys

Yff

y r y,

f,

isotropic machine (or, which is the same, with magnetising

flux equivalent I,U ) and can be expressed as a linear

Generalised inductance A, defined in (1 l), depends on the

main flux saturation level (i.e., it is function of L,). As shown
in [15-181, all the models of single-cage and double-cage
induction machines and all except three models of smooth
air-gap synchronous machines, obtainable using the
generalised flux concept, can be classified into three groups.
The groups encompass models where dA/dt, d(l/A)/dt or none
of the two, respectively, is required in the model derivation.
B. Classi'cation

The total number of models, obtainable using the

generalised flux concept, is in the case of a smooth air-gap
synchronous machine equal to 37, [NI. Generalised flux and
generalised inductance can be made equal to magnetising flux
and magnetising inductance in 34 models while in the three
remaining either generalised flux or both generalised flux and
generalised inductance do not exist physically in the machine.
In the case of salient pole synchronous machine it turns out
that the only obtainable models are those for which
generalised flux Y and generalised inductance A can be made
and magnetising
equal to magnetising flux equivalent ~f,,
inductance L, , respectively, defined in (G)-(7). Hence the
total number of models that can be derived utilising the
generalised flux concept reduces from 37 for smooth air-gap
to 34 for salient pole synchronous machine.
A consequence of non-existence of the q-axis excitation
winding is that a number of state-space variable selections in
terms of space vectors lead to, apparently, selection of six real
state-space variables. However, for each of these cases only
two state-space variables for the q-axis are required, so that
one has to be omitted. The issue is discussed in detail in [ 181.
The available 34 models can be classified into three groups,
encompassing all the models that require d N d t , d(l/A)/dt or
neither of the two, respectively, during the model derivation.
(As already pointed out, A = L,, for all the models.) Content
of each of the three groups is as follows.
A Models (dMdt required). The group contains 13 models:
0

--P

combination of the selected state-space variables in the form:

Y = (axdl +bx,,

+ c x d 3 ) + j ( l / m ) ( a x q+bxq2
l
+cxq3) . (10)

One or more of the coefficients a, b, c in (10) may take

different values for d- and q-axis in the salient pole
synchronous machme, as discussed later on. As the
generalised flux space vector is aligned with magnetising
current space vector equivalent , then

of the Models

pure current models (four models);

models with magnetising current and any two flux
linkage space vectors (three models);
models with two currents (one of which is magnetising
current) and one flux linkage space vector (six models).

1/A Models (d(l/A)/dt required). There are 20 models:

models with two currents (none of which is magnetising
current) and one flux linkage space vector, excluding the
. .
.
.
L s > L r , Y f f; V / s > ' f
,if ,is
models
(seven

v,

models);
models with two flux linkages and one current which is
y s , i r ,y f ,
not magnetising current, excluding ~f

-SI-Y

IY
,

,if

and

~f
-r

,is,iyf

models (ten models).

47

models with magnetising flux and any other two flux

linkages (three models);
For all the models of this group, the coefficients a, b, c of (IO)
must be saturation independent, as explained in [ 15-18].

For the remaining eight models of this group it is again

possible to obtain generalised inductance and generalised flux
as equal to L, and K, , by proper selection of the coefficients
. .
in (10). For z s , L s ,V+,!~,L,;
i~;
y s , L s , v ~ ;, , f s ,

Neither dMdt nor d(l/A)/dt required. This 'group' contains

once more a single, already available, model:
0
all the flux linkages of d-q axis windings.

models coefficients a, 6, c are to be set to 1, -L, and 0,

respectively.
For
the
remaining
four
models,
. .
v r 7 ! , , i J i y r $ ~ r ~, / ~
, , L s, , Y;J ; y r , ! r , y s> a = 1, c = O
and b takes the values of -Ldr and -LTr for d- and q-axis,
respectively (for all the models with, apparently, three statespace variables for the q-axis, the q-axis components of the
first two space vectors are selected) .
For all the 1/A models, time derivative d(l/L,J/dt is
required in the process of the model derivation. This time
derivative now equals

In all except one of the A models, magnetising current

space vector is one of the three space vector state-space
variables. For all these models, coefficient (say, ad) that
multiplies magnetising current d-axis component in (10) is to
be set to the saturated d-axis mutual inductance,
correspondmg coefficient that multiplies q-axis magnetising
current component (a4)to saturated q-axis mutual inductance,
while the other two coefficients (b,c) in (IO) are to be set to
zero (it is assumed that both d- and q-axis components of the
magnetising current are selected as state-space variables). For
the remaining A model with all the winding currents as statespace variables, all the three coefficients of (10) are to be set
to the saturated d- and q-axis mutual inductances for d- and
q-axis variables. By performing such a choice of coefficients
in (lo), the generalised inductance and generalised flux of
(11) become equal to saturated magnetising inductance L,
and magnetising flux equivalent v,,respectively. During the
process of model derivation for any of the A models, time
derivative of L , will be required. This time derivative equals
dL
dt =

- L p ) ;(idm

L = d y, /dip

1-

diqm

+m2j

qm

L,=v,/i,

dt
2

and the resulting saturation dependent inductances in system

matrix for all the models will be found in the form
Ldd = L cos2 p + L, sin2 L,d

L,, = m2 ( L , cos2 p + L sin2 p )

Ld, = L,d = m L - L

$1cospsinp

COS^ = idm/ip

s i n p = miqmJi, .

The inductances of (13) are identical to those that are present

in the available pure winding current state-space model based
on single saturation factor approach [5].
As far as the l/A models are concerned, twelve out of 20
models contain magnetising flux space vector as state-space
variable and, assuming that both d- and q-axis component of
this vector are selected as state-space variables, the coefficient
in (10) that multiples magnetising flux space vector is to be
set to 1 while the other two coefficients are to be set to zero.
Thus for all the 12 models generalised inductance and
generalised flux in (IO)-( 11) become equal to the magnetising
inductance L,, and the magnetising flux equivalent y , .

1/L = d i p / d v p

1/L,=ip/vjl

(14)
The saturation dependent terms (inverse inductances) in the
system matrix for all the 1/12 models are obtained in the form
1

-=

Ldd

1 =

L,,

1
1
-cos2 p+-sin2
LP

-(

1 -cos2
1
m 2 L,

p +-sin2
1

1 - 1
1 1
- -= -Ldq

Lqd

$1

sin p cos p

cosp=v d m / v p
Sinp=(l/m)(vqm/vp)
.
Derivation procedure is described and three models in full
developed form are given in the next Section.

Iv. DERIVATION
AND DEVELOPED
FORM OF THE MODELS
A . Model with f , , i J , f m as State-Space Variables
This model, according to the classification given in the
previous Section, is the A type model. It is derived from basic
model (1)-(4) by eliminating stator, excitation and damper
flux linkages, and damper currents from (1)-(3), using (4)
and (12)-(13). The resulting model of a saturated salient pole
synchronous machine can be expressed in matrix form as

PI = [Ald[xl/dt

b]=['ds

'qs

[x]=[ids

iqs

[BI[xl

'J
idm

1' '
Zqm]'

(16)
7

where matrices [A] and [B] are given with:

yr

48

I
-

L,
0

L,

0
0
L@

Rs

-@L,

-W~'L,

wL,

R,
0

0
Rf

COL,

0
0

Rdr

[B]=

-Rdr

-Rdr

-Rqr

PI =

-Rdr

Saturation dependent coefficients in [A] are given with (13).

-

B. Model with f ,, i f ,

-m

V d s = Lmids + V d m
+Vdm

Vdr =

(ids + i f

) (

Vqs = Lmiqs +V q m

/Lp)Vdm

)vqm

vqr=-L
aqr i 4s + l + L q r m L,
id,. = -ids - i f

0
0

--w

Rf

-Rdr

-Rqr

Rdr/Lp

R q r /(m"Lp

as State-Space Variables

Derivation procedure of this model, which is of 1/A type,

closely parallels the procedure presented in detail (for the
same model) for smooth air-gap synchronous machine in
1181. It is nevertheless presented here again, in order to
emphasise differences introduced by saliency. All the flux
linkages and damper currents are at first expressed in terms
of state-space variables, using (4):

Vf = L o f i f

R,
-uL~
COL,
R,
0
0

+y d m / ~ ,

iqr = -iqs

(19)

+ y q m / m2 L

This particular model is selected for presentation as an

example of an 1/A model for which, in order to make
generalised inductance and generalised flux equal to L, and
y,, respectively, coefficient b of (10) has to be equated to the
negative value of the stator leakage inductance, as explained
in Section 111. Derivation procedure closely follows the one of
the precedmg subsection. By selecting vector of state-space
variables as
[x]=[Vds

Vqs

ids

iqs

(24)

if]f

~ .

Time derivatives of damper flux linkages of (19) are found

next,

(20)

Substitution of (14) into (20), followed by substitution of (19)

and (20) into (1)-(3) enables formulation of the model in the
matrix form of (16). As the model is of 1/A type, saturation
dependent coefficients of matrix [A] become now as given in
(15). Vector of state-space variables and matrices [A] and [B]
are now

Rdr/Lp

where 1, = L ,
[x]=[i&

iqs

'f

Vdm

Vqm]'

(2 1)

Rf

- Rdr

-Rdr(l+Lm/Lp)

+ L,d,. , 1,

= L,

+ L,

and saturation

dependent coefficients in [A] are again given with (15 ) .

49

V. SIMULATION
RESULTS
The following simulation study is performed in order to
verify the derived salient pole synchronous machine models.
A generator, whose per-unit parameters and the d-axis
magnetising curve are given in Appendix, is initially operated
in steady-state with terminal stator current and voltage equal
to 1 P.u., with turbine torque of 0.9 p.u. and with power factor
of 0.9 in overexcited mode. Power invariant transformation is
applied and the equations are formed using motoring
convention for positive powers. Machine torque, load angle
and stator q-axis current are therefore all negative. The
transient selected for investigation is a step change of
terminal voltage from 1 p.u. to 1.1 P.u.. Turbine governor and
excitation voltage controller are not included. This is a
transient during which impact of main flux saturation is
undoubtedly pronounced. Simulation is at first performed
using the well-known winding current state-space model of a
saturated synchronous machine [5], which is obtainable in
identical form using the procedure outlined in this paper.
Next, the same transient is simulated using constant
parameter synchronous machine model (unsaturated model),
with magnetising inductance set to the constant value that
corresponds to the previous steady-state (0.638 P.u.). All the
variables in the two models have therefore identical values in
the previous steady-state. A sample of results is given in Fig.
1, where variations of load angle and stator d- and q-axis
currents are shown (transient is initiated at t = 12.5 s).
As the next step, the same transient is simulated using the
models of Section IV instead of the winding current statespace model. It is found that all the saturated machine models
yield the same results. As an example, Fig. 2 shows
comparison of traces obtained with winding current statespace model and with the mixed current-flux state space
model given in subsection B of Section IV. No observable
difference is present in the results. The derivation procedure
and the presented models are thus verified.

VI.DISCUSSION
Saturation modelling procedure, described in the paper,
relies on the standard d-q axis model of a salient pole
synchronous machine, requires knowledge of the machines
open circuit d-axis magnetising curve and is based around the
single saturation factor method. The two already available
models are derived using the same assumptions and the same
initial model; moreover, these two models can be derived by
means of the procedure elaborated here. If all the winding
currents are selected as state-space variables, the model that is
derived is identical to the model given in [ 5 ] . Hence the
accuracy of saturation representation in all the models
presented in the paper, as well as in all the other models that
can be derived (including the two already available models),
is the same as long as single saturation factor approach is

used. Simulation study, whose results are presented in Section

V, confirms this statement. The novel models are thus neither
less nor more accurate than the already existing two models.
The advantage of a number of novel models is reduced
complexity of the system matrix, enabling faster simulations,
when compared with the available current state-space model.
For example, in system matrix of the model given in subsection A of Section IV, 9 out of 25 elements are zero and
only 10 elements depend on saturation (in the current statespace model all 25 elements have non-zero values and depend
on saturation). The situation is even more favourable in the
model presented in subsection B of Section IV where 12

-20.5
12

12.5

(-Saturated

SG model

0.2

13.5
14
14.5
15
Time (s)
-Unsaturated SG model

12

12.5

-Saturated

13

13

SG model

13.5
14
14.5
15
Time (s)
-Unsaturated SG model

-0.8,

8 -1.8 -L

-2-l

-Saturated

SG model

-Unsaturated SG model

50

VIII. APPENDIX: SYNCHRONOUS

GENERATOR
DATA
Per-unit parameters and magnetising curve approximation:
R, = 0.0062 R f = 0.000152 Rd,. = 0.0254 R,, = 0.055
L, = 0.14
= 0.91

L,
-20.54
12

-lds,lqs,lf,ldr.lqr

-0.8,

12.5

:
13

model

: : :
13.5
14
Time (s)
!

:
14.5

-lds,lqs.lf,Fdm,Fqm

I
15

y,

131

d-axis

[41
[51

12

12.5

13

13.5 14
Time (s)

14.5

15

[71

Fig. 2. Comparison of results obtained with two different state-space

models of a saturated salient pole synchronous machine (F ly).

elements of the system matrix are equal to zero and only four
elements are saturation dependent. State-space variables in
both of these models are those that are usually of interest, i.e.,
stator current components and excitation current (which is
not the case in pure winding flux linkage model, where
currents have to be recalculated using state-space variables)
with considerably simpler structure of the system matrix
compared with pure d-q winding current model.
All the models (except for the pure winding flux linkage
model) contain explicit terms that describe dynamic crosssaturation. However, the number of these terms differs for
various models. All the models are easily adaptable to a
smooth air-gap synchronous machine.
VII. CONCLUSION

Application of generalised flux concept for saturation

modelling in salient pole synchronous machines is discussed
and a number of novel models are derived. The models differ
with respect to the selected state-space variable set. Relatively
simple overall model structure is arrived at in a number of
cases. The accuracy of saturation representation in all the
novel models is the same as in the two already existing
models, as verified by simulation. It is believed that the
models will find application in synchronous machine
simulation.

= 0.3426

Lqmn= 0.53

m2 = 0.58242

L,,

= 0.242

OB =

120n

IX. REFERENCES

model

-1.6

L,

= 1.1212 *0.8029ivei,

= 2.7093

-0.2

I-

Ld = 0.2244

[81
[91

P.Vas, Electrical machines and drives, A space-vector theory

approach, Oxford: Clarendon Press, 1992.
LBoldea, S.A.Nasar, A unified analysis of magnetic saturation in
orthogonal axis models of electric machines, Electric Machines and
Power Systems, vol. 12, no. 3, 1987, pp. 195-204.
R.G.Harley, D.J.N.Limebeer, E.Chirricozzi, Comparative study of
saturation methods in synchronous machine models, IEE Proceedings,
Pf. B, vol. 127, no. 1, 1980, pp. 1-7.
G.Xie, R.S.Ramshaw, Nonlinear model of synchronous machines with
saliency,IEEE Trans. Energy Conv., vol. 1 , no. 3, 1986, pp. 198-204.
I.Iglesias, L.Garcia-Tabares, J.Tamarit, A D-Q model for the selfcommutated synchronous machine considering the effects of magnetic
saturation, IEEE Trans. Energy Conv., vol. I,no. 4, 1992, pp. 768-776.
S.A.Tahan, LKamwa, A two-factor saturation model for synchronous
machines with multiple rotor circuits, IEEE Trans. Energy Conv., vol.
10, no. 4, 1995, pp. 609-616.
S.A.Tahan, I.Kamwa, P.Viarouge, M.Ferfa, A generalised model of
saturated synchronous machines, IMACS Journal on Mathematics and
Computers in Simulation. vol. 38, 1995, pp. 425-435.
P.M.Anderson, AAFouad, Power system control and stability,
Piscataway, NJ: IEEE Press, 1994.
J.O.Ojo, T.A.Lipo, An improved model for saturated salient pole
synchronous motors,lEEE Trans. Energy Conv.. vol. 4, no. 1, 1989, pp.
135- 142.
R.Wamkeue, LKamwa, X.Dai-Do, A detailed model of grounded
synchronous machines for saturated unsymmetrical transients, Proc. of
Int. Conf ELECTRIMACS96. St. Nazaire, France, 1996, pp. 333-340.
V.Brandwajn, Representation of magnetic saturation in the synchronous
machine model in an electro-magnetic transients program, IEEE Trans.
PowerApp. andSysf., vol. 99, no. 5, 1980, pp. 1996-2001.
F.P.de Mello, L.N.IIannett, Representation of saturation in synchronous
machines,IEEE Trans. on Power Systems, vol. 1, no. 4, 1986, pp. 8-14.
A.M.El-Serafi, J.Wu, Saturation representation in synchronous machine
models, Elec. Machines &Power Sysfems, vol. 20, 1992, pp. 355-369.
L.Pierrat, E.Dejaeger, M.S.Garrido, Models unification for the saturated
synchronous machines, Proc. Int. Con$ on Evolution and Modern
Aspects ofSynch. Machines, Zurich, Switzerland, 1991, pp. 44-48.
E.Levi, A unified approach to main flux saturation modelling in D-Q
axis models of induction machines, IEEE Trans. Energy Conv., vol. 10,
no. 3, 1995, pp. 455-461.
E.Levi, Main flux saturation modelling in double-cage and deep-bar
induction machines, IEEE Trans. Energy Conv., vol. 11, no. 2, 1996,
pp. 305-31 1.
E.Levi, General method of magnetising flux saturation modelling in d-q
axis models of double-cage induction machines, IEE Proc. - Elecfr.
PowerAppl., vol. 144, no. 2, 1997, pp. 101-109.
E.Levi, Modelling of magnetic saturation in smooth air-gap synchronous
machines, IEEE PES Summer Meeting, Denver, CO, 1996, Paper No.
96SM 364-0 EC.

Emil Levi (M1992) received his PhD from the University of Belgrade,
Yugoslavia in 1990. From 1982 till 1992 he was with the University of Novi
Sad, Yugoslavia. Since 1992 he is with Liverpool John Moores Ilniversity,
Liverpool, UK, where he is currently Reader in Electrical Power Engineering.