Per-unitization and Equivalent Circuits

1.0 Normalization of voltage equations

We desire to normalize the voltage equations, that is, we desire to express

them in per-unit. The advantages of doing so are:
 The per-unit system offers computational simplicity by eliminating
units and expressing system quantities as dimensionless ratios.
 It eliminates the need for using arbitrary constants and simplifies some
of the mathematical expressions so that they may be expressed in terms
of equivalent circuits.
 The numerical values of currents and voltages are related to their rated
values irrespective of machine size.
 Impedances, when given on the machine base, lie on a relatively
narrow range so that errors can be easily detected.

There are several different possible normalization schemes. What Anderson

and Fouad (A&F) have done (see Appendix C in text) is to carefully compare
the merits of all of these schemes. In doing so, they developed some criteria,
guidelines, the most important of which is that the equations must be
independent of whether they are in pu or MKS for both the voltage equations
as well as the power expression (power invariance).

Note that machine manufacturers, when expressing their machine data in per
unit, may use a different system that does not satisfy the power invariance
property - they use Park’s original transformation (called “Q” in A&F, eq.
(4.22), instead of our “P”).

The choice made by A&F satisfies the above criteria; in addition, the A&F
choice ensures that the numerical values of the per-unit impedances are the
same as those provided by manufacturers using their system of
normalization.

1
In most undergraduate power system analyses courses, we learn that per-
unitization requires selection of two base quantities out of the following four:
V, I, Z, and S, and then the base quantities for the other two are computed.
The situation is the same here, except that we also must deal with speed (or
frequency). This necessitates that we must also select a base for either
frequency ( or f) or time, t.

In addition, we will also have need to compute base quantities associated

with flux linkage () and inductance (L or M).

Our approach will be to obtain the bases for the stator side and then the bases
for the rotor side.

One may note two excellent references on the subject of per-unitizing

synchronous machine models:
1. A. Rankin, “Per-unit impedance of synchronous machines,” AIEE
Transactions, 64, Aug., 1945.
2. M. Harris, P. Lawrenson, and J. Stephenson, “Per-unit systems with
special reference to electrical machines,” Cambridge University press,
Cambridge, England, 1970.

Other references that address this subject, besides A&F, include those by
- Sauer and Pai
- Concordia
- Padiyar
- Kundur
and also course notes from de Mello.

1.1 Stator side per-unitization:

We select our stator-side bases as:

 VB: the stator rated line-neutral voltage, rms.
 SB: the stator rated per-phase power, volt-amps
 B: the generator rated speed, in electrical rad/sec (=Re=377)

2
Then we may compute bases for the following 5 quantities:
SB
 current: I B 
VB
VB VB2
 impedance: X B  RB  Z B  
I B SB
1 2
 time: t B  (Note we could have used tB  but this
B B
would simply provide a different scaling and is therefore arbitrary). Note
that our choice of tB is the time required for the rotor to move one electrical
radian.
 Flux linkage: B  VB t B (This comes from the fact that
d 
v     Vt )
dt t
𝜆𝐵 𝑉𝐵 𝑡𝐵 𝑉𝐵 𝑋𝐵
 Inductance: 𝐿𝐵 = = = =
𝐼𝐵 𝐼𝐵 𝜔𝐵 𝐼 𝐵 𝜔𝐵

 B XB VB Vt
LB     BB
IB B B I B IB

Question: How does our choice of stator base quantities affect the per-unit
values of the d- and q-axis quantities? Note: although d- and q-axis quantities
associate with fictitious rotor windings, we view them to be stator quantities.

To answer this question, let V and I be the rms magnitudes of the a-phase
line-neutral voltage V and a-phase line current I, respectively. Then
the per-unit phasors are
V I
V u    Vu  I u    I u 
VB IB
Now let’s investigate the 0dq quantities.

3
To begin, note that the expressions for instantaneous voltages and currents
for each phase are:
va  2V sin(   ) ia  2I sin(   )
vb  2V sin(    120) ib  2I sin(    120)
vc  2V sin(    120) ic  2I sin(    120)
Use the Park’s transformation on the above to obtain:
 0   0 
v0 dq   3V sin   i0 dq   3I sin  
   
 3V cos    3I cos  
(This confirms our conclusion at the end of the last set of notes, “macheqts,”
that, for balanced conditions, the 0dq quantities are constants, i.e., DC.)

Now, per-unitize by dividing by VB and IB:

   
   
 0   0   0   0 
 sin     3Vu sin   sin     3I u sin  
3V

3I
v0 dq i0 dq
u  VB     IB   
  3Vu cos 
u
 3V  3I   3I u cos  
 cos   cos  
 VB   IB 
Observe about the above that
1. The per-unit d and q voltages are equal to the per-unit a-phase voltage
scaled by 3 sin  and 3 cos  , respectively.
2. The per-unit d and q currents are equal to the per-unit line current scaled
by 3 sin  and 3 cos  , respectively.

4
1.2 Rotor-side per-unitization:

Recall that in system per-unitization, we must select a single power base for
the entire system, independent of the fact that some sections of the system
are magnetically coupled through transformers, i.e., we do NOT choose
different power bases for different sides of a transformer.

The same restriction applies here, where the rotor circuit is magnetically
coupled to the stator circuit, i.e., the power base selected for the stator side
must also be the power base used on the rotor side. This is SB.

In addition, we are required to select the same time (or frequency) base for
both the stator side and the rotor side. This is tB (or B).

On the rotor side, we have 1 base left to choose. For transformers, we

typically choose the 1 remaining base as the voltage base (or current base)
according to the turns ratio. Here, however, we do not know a “turns ratio,”
and therefore we are left with problem of what, and how, to choose. (One
text treats the problem under the assumption that a “turns ratio” is known
between stator and rotor circuits - see the text by Padiyar, “Power System
Dynamics,” pp. 73-77.)

In making this choice, a problem results from the fact that stator power levels
are typically several times the rotor power levels. A&F give an interesting
comparison (see pg 95) of a typical stator-side per-phase power rating of 100
MVA and field winding ratings of 250v, 1000A (250kw). What are our
choices of the one remaining rotor-side base quantity in this case?
 Choose voltage base=rated voltage=250v, but then the current base is
IB=100E6/250=400000 amps, and per-unit values of field currents will be
very small.
 Choose current base=rated current=1000A, but then voltage base is
VB=100E6/1000=100000 volts, and per-unit values of field voltages will
be very small.

5
Analogy to transformers:
With transformers, we choose the base voltage (or current) on side 1, and
then we choose the base voltage (or current) on side 2 as that voltage
produced by the transformer on side 2 when the base voltage on side 1 is
applied. We did this because we wanted a per-unit circuit of the transformer
where the ideal transformer was eliminated.

Another way to think about what we do with transformers is that we select

current bases that would produce the same mutual flux between the two
windings, i.e., we choose IB1 and IB2 such that:
 IB1 produces 21 (the flux linkage from current IB1 in coil 1 that links coil 2),
 IB2 produces 12 (the flux linkage from current IB2 in coil 2 that links coil 1)
and 21=12. This is based on the fact that 21=M21IB1 and 12=M12IB2, where
M21=M21=M is the mutual inductance between the two transformer coils.

If coil 1 creates 1, and of this, only the mutual flux 21 links with coil 2,
then the difference is the leakage flux given by
L1=1-21 (eq. 1)
and illustrated in Fig. 1.
Leakage flux, L1

Mutual flux, 21

1

Fig. 1: Mutual and Leakage flux

Here, each of these three fluxes are given by

L1=l1IB1, 1=L1IB1, and 21=MIB1 (eq. 2)
where l1, L1, and M, are the leakage, self, and mutual inductances,
respectively. Substitution of (2) into (1) results in:
l1IB1=L1IB1- MIB1

6
and canceling IB1 gives:
l1=L1- M L1=l1+M
which implies that the self inductance is comprised of the leakage inductance
plus the mutual inductance.

Similar analysis results in L2=l2+M.

Back to synchronous machines:

The reason this is
We can apply the same concept to the synchronous machine as we applied beneficial is that it
will enable us to
to the transformer above. That is, develop a relatively
simple circuit to
represent direct-
We select the base currents for the four rotor-side windings F, D (Q, G) to axis pu quantities
and another one to
produce the same mutual flux in the air gap as produced by the stator-side represent
quadrature axis pu
base current IB flowing in the corresponding fictitious d-axis (q-axis) coil. quantities. The
form of this cct will
be “tee”. (see Per
App C, top of p.
We will begin by applying this idea to obtain the base current for the main 549) and pg. 34 of
field winding. these notes.

Base-current for main field winding, approach 1:

One can visualize the above concept for the case of the relationship between
the F-winding and the d-winding, in Fig. 2.
d-winding
IB
F-winding

md

IFB

Fig. 2: Base currents in d and F windings

We see from Fig. 2 that we select IFB, the field winding base current, as

7
 that current when flowing in the F-winding will produce a
mutual flux md equal to the same mutual flux that is produced
by a current IB flowing in the d-winding.

But how do we compute IFB?

From our previous set of notes (p. 28, “macheqts”), and also eq. 4.20 in text,
we derived
 
 L0 0 0 0 0 0 0 
 
 3 3
 0   0 Ld 0 MF MD 0 0   i0 
2 2  
   3 3  id
 d  0 0 Lq 0 0 MQ M G  
 q   2 2 i 
 q 
    3
  
F 0 MF 0 LF MR 0 0  iF 
2  
 D   3  iD 
   0 MD 0 MR LD 0 0  i

   2  Q
 
Q

G   3
0 0 MQ 0 0 LQ M Y  iG 
 2 
 3 
 0 0 M 0 0 M L 
 
G Y G
2
eq. (4.20’)
From this, we can see that
d  Ld id  kM F iF  kM DiD (eq. 3)
where k=(3/2).

Ldid is all of the flux produced by the d-winding, but only a part of this flux,
the mutual flux, links with the F-winding. Call this flux from the d-winding
that links with the F-winding md, given by md=Lmdid, where Lmd is the
mutual inductance associated with this flux (A&F call it a magnetizing
inductance).

8
As with the case of the transformer, the difference between the total flux
from the d-winding and the mutual flux is attributed to the leakage flux L,
so that,
ldid=Ldid- Lmdid (eq. 4)
Canceling the current id, we see that
ld=Ld-Lmd Ld=ld+Lmd (eq. 5)
When IB flows in the d-winding, so that id=IB, the mutual flux is given by
md=LmdIB (eq. 6)
Looking back at eq (3), we see that the flux from the F-winding that links the
d-axis winding is just kMFiF.

Our criteria for selecting IFB says that when IFB flows in the F-winding, the
mutual flux linking the d-winding should equal the mutual flux from the d-
winding linking the F-winding when it carries IB. Thus, we write that
md=LmdIB=kMFIFB (eq. 7)
And we see that
Lmd
I FB  IB (eq. 8)
kM F
MF and Lmd are generally provided in (or can be obtained from)
manufacturer’s data for a given machine1.
 MF can be computed as illustrated in Example 4.1 (which we review
below), using the magnetization curve,
 Lmd=Ld-ld, where manufacturer’s data sheets contain Ld and ld.
Therefore, once IB is selected, IFB may be computed.

Base current for main field winding, approach 2:

One may also develop a relation for IFB from the perspective of the flux
linking the field winding, i.e., instead of using eq. (3) from (4.20’), use:
F  kM F id  LF iF  M RiD (eq. 9)
Similar to eq. (5), the self inductance LF is comprised of the leakage and the
mutual, i.e.,

1
Note: in MKS units (i.e., henries), Lmd is not the same as kM, i.e., the reciprocal mutuals are not equal.

9
 LF=lf+LmF (eq. 10)
Inspecting eq. (9), we see that the flux from the d-winding linking with the
F-winding is kMFid, so that when id=IB and iF=IFB, we have that
LmFIFB=kMFIB (eq. 11)
and we see that
kM F
I FB  IB (eq. 12)
LmF
where, as before, MF is obtained per Example 4.1 below, LmF=LF-lF, and LF,
lF are obtained from manufacturer’s data sheet.

Base current for D-winding:

We select the D-winding base current, IDB according to the following criteria:
We select IDB, the D-winding base current, as that current when
flowing in the D-winding will produce a mutual flux md equal to the
same mutual flux that is produced by a current IB flowing in the d-
winding.
Similar analysis as for the F-winding results in
Lmd kM D
I DB  IB, I DB  IB (eq. 13)
kM D LmF
We may also utilize a similar procedure between D and F windings to obtain
LmF
I DB  I FB (eq. 14)
MR
Base current for Q-winding:
We select Q-winding base current, IQB according to the following criteria:
We select IQB, the Q-winding base current, as that current when
flowing in the Q-winding will produce a mutual flux mq equal to the
same mutual flux that is produced by a current IB flowing in the q-
winding.

10
Similar analysis as for the F-winding results in
Lmq kM Q
I QB  IB, I QB  IB (eq. 15)
kM Q LmQ
Base current for G-winding:
We select the G-winding base current, IGB according to the following criteria:
We select IGB, G-winding base current, as that current when flowing
in the G-winding will produce a mutual flux mq equal to the same
mutual flux that is produced by a current IB flowing in the q-winding.
Similar analysis as for the F-winding results in
Lmq kM G
I GB  IB, I GB  IB (eq. 16)
kM G LmG
We may also utilize a similar procedure between Q and G windings to obtain
LmQ
I GB  I QB (eq. 17)
MY
Summary:
Eqt. (8) together with eqts. (12-17) provide the ability to develop any of the
equations given as (4.54) in the text. These equations are referred to as the
“fundamental constraints among base currents” and are given by:
Lmd I B2  LmF I FB
2
 LmD I DB
2
 kM F I B I FB  kM D I B I DB  M R I FB I DB
Lmq I B2  kM Q I B I QB  LmQ I QB
2
 M Y I QB I GB
For example, recalling (8) is I  Lmd I and (12) is kM F , we
FB B
I FB  IB
kM F LmF
can multiply the left-hand-sides together and the right-hand-sides together to
Lmd kM F 2
obtain: I FB  kM L I B  LmF I FB  Lmd I B .
2 2 2

F mF

Now define the following factors:

11
 IB IB IB IB
kF  , kD  , k Q  , kG 
I FB I DB I QB I GB
Because we have the same power base on all stator and rotor circuits, we
obtain:
S B  VB I B  VFB I FB  VDB I DB  VQB I QB  VGB I GB
Then
VFB VDB VQB VGB
kF  , kD  , kQ  , kG 
VB VB VB VB
Note that these k-factors may be considered to be effective turns ratios.

We may also derive expressions for the resistance and inductance bases.
Note our desire is to be able to compute rotor-side bases as a function of
stator-side bases. The k-factors given above will be very handy here.
Rotor-side resistance bases:

VFB VFB I B VB
RFB    k F2 RB
I FB VB I FB I B
Likewise,
RDB  k D2 RB , RQB  kQ2 RB , RGB  kG2 RB
Rotor-side inductance bases:

VFB t B VFB I B VB t B
LFB    k F2 LB
I FB VB I FB I B
Likewise,
LDB  k D2 LB , LQB  kQ2 LB , LGB  kG2 LB

12
Rotor-stator mutuals:

Your text, pg. 95 refers to HW problem 4.18 which states that base mutuals
must be the geometric mean of the base self-inductances, i.e.,
M 12  L1B L2 B
Thus, we have that the base for the field winding to stator winding mutual
terms is given by (see eq. 4.57 in text):
M FB  LB LFB  LB k F2 LB  k F LB
Note that it is not the same as the base self inductance LFB given above.
Likewise, we get (see eq. 4.57 in text, except here we included MGB):
M DB  k D LB , M QB  kQ LB , M GB  kG LB
Rotor-rotor mutuals:
There are just 2 of them (see eq. 4.57 in text, except here we included MYB):
M RB  LFB LDB  k F2 LB k D2 LB  k F k D LB
Likewise,
M YB  kG kQ LB
2.3 Example 4.1, pg 97 of text
This is a good example that you should review carefully. Here is the first part
of it (p. 98 continues with it).

13
 Why are some values “estimated for
academic study”?

Note this statement; What is “air gap line”?

Note their use of “peak phase voltage”.

The only thing that is perhaps not too clear is the computation of M F. I will
just review that part of it here.

Computation of MF: A&F make the statement,

“From the no-load magnetization curve, the value of field current
corresponding to the rated voltage on the air-gap line is 365 A.”
The “open-circuit characteristic” or “magnetization curve” plots
 Something proportional to exciting (field) current on horizontal axis
 Something proportional to the flux on the vertical axis.
under open-circuit conditions (the a-phase winding is open). Figure 3 below
illustrates.

14
 Air-gap line
Va
φ
λ
B Due to saturation of
the iron (a decrease in
permeability or
increase in reluctance)
for high MMF

iF
F=NiF
H=F/l

Fig. 3
The air-gap line is the Va vs. iF relation that results if the iron has constant
permeability. The solid line that bends to the right is the actual characteristic
that occurs, which shows that terminal voltage falls away from the air-gap
line as the field current is raised beyond a certain point. This falling away is
caused by saturation of the ferromagnetic material, resulting from the
decrease in permeability under high flux conditions. Figure 4 illustrates a
magnetization curve for a real 13.8 kV synchronous machine. The vertical
axis is line-to-line voltage.

15
 This synch gen is separately excited, and so
its field current If (here designated as “rotor
amps”) is supplied from the armature of a
separate DC gen. The separate DC gen has
field current IDCgen (on DC gen stator)
which creates field flux ϕ. The current If,
which is the armature current of the DC gen
and the field current of the synch gen,
increases with DC gen armature voltage Ea,
and Ea=kωϕ, where ϕ increases with IDCgen.
And so If and IDCgen are both indicators of
synch gen field strength.

If
Field
IDCgen Ea winding

+
-
The two magnetization curves to the left
plot line-to-line open cct voltage of the
synch gen against (a) IDCgen and (b) If.

Fig. 4

What is done in Ex. 4.1 (and what is actually done in industry to obtain M F),
is that the field current is determined corresponding to steady-state rated
open circuit terminal voltage. This voltage is VB=VLL-rated/sqrt(3). For Ex.
4.1, this is VB=15kV/sqrt(3)=8660 volts. This is the rms voltage, but A&F
indicate that we need the corresponding peak voltage:
Vpeak=√2(8660)=12,247.1 volts. But why do we need the peak voltage?

Let’s consider this question.

16
From first page of previous notes titled “Machine Equations,” or from eq.
(4.11’) in A&F, we have
a  Laaia  Labib  Lacic  LaF iF  LaDiD  LaQiQ  LaQiG
But ia=ib=ic=0 under open circuit conditions.
And damper currents iD=iQ=0 under steady-state conditions. Therefore
a  LaF iF  LaQiG
Recall that the g-winding models the Q-axis flux produced by the eddy-
current effects in the rotor during the transient period. But we are now
considering only the steady-state condition, iG=0. Therefore
a  LaF iF (*)
Now recall from first page of previous notes titled “macheqts,” or from eq.
(4.16’) in A&F, that LaF=MFcosθ, and substitution into (*) yields
a  M F iF cos (**)
Differentiating (**) results in
da d
  M F iF sin    Re M F iF sin  (***)
dt dt
Now recall the voltage equation for the a-phase:
va  ia ra  a  vn (#)
Substituting (***) into (#), we obtain
va  ia ra  Re M F iF sin   vn
But under open circuit conditions, ia=0, in=0 (implying vn=0) and we have
va  Re M F iF sin  (#*)
From (#*), we see that
V peak
V peak   Re M F iF  M F 
iF  Re
So we choose a point off the magnetization curve, for example, A&F choose
iF=365A, Vpeak=12,247.1volts (365A is the value of field current

17
corresponding to the rated voltage on the air-gap line, and
12,247.1/sqrt(2)=8660volts is the rated RMS line-to-neutral voltage
(corresponding to 8660sqrt(3)=15kV). Then
V peak 12,247.1
MF    89.006  103 henries
iF Re (365)(377)
And from this we can compute
kM F kM F
kF  
Lmd Ld  ld
where the denominator is comprised of data provided by the manufacturer.
The rest of Ex. 4.1 is just an application of our per-unitization formula.

There is an interesting paragraph in Appendix C, pg. 552 of your text, to

which I want to draw your attention. It says,
“Note that a key element in determining the factor kF, and hence all the
rotor base quantities, is the value of MF (in H). This is obtained from the
air gap line of the magnetization curve provided by the manufacturer.
Unfortunately, no such data is given for any of the amortisseur circuits.
Thus, while the pu values of the various amortisseur elements can be
determined, their corresponding MKS data are not known.
I provide some comments on certain sentences in this paragraph:
 “Note that a key element in determining the factor kF, and hence all the
rotor base quantities,” refers to the fact that we obtain LFB and RFB from:
LFB  k F2 LB
RFB  k F2 RB
 “This is obtained from the air gap line of the magnetization curve
provided by the manufacturer,” as we have seen above by using
V peak
MF 
iF Re

18
 We are able to get MF in this way because we can directly control the
current iF, with no other circuits energized (as a result of the open-circuit,
steady-state conditions), and directly measure the induced voltage at the
a-phase terminals.
 “Unfortunately, no such data is given for any of the amortisseur
circuits.” It is not possible to directly control the currents iD, iQ, and iG,
since their corresponding circuits do not have sources. The only way to
energize these circuits is via a transient condition, but there is no way to
provide a transient condition that will also not energize other circuits,
which would result in the measured terminal voltage being induced from
the mutual inductance between itself and the other circuits as well.
 “Thus, while the pu values of the various amortisseur elements can be
determined, their corresponding MKS data are not known.” In example
4.1, the text puts an asterisk by some of the parameters (L D, LQ, kMD,
kMQ, rD, and rQ), indicating they were “estimated for academic study”).
This is because manufacturer’s datasheets do not always include the
parameters for the amortisseur (and g-winding) circuits, simply because
they are hard to measure (based on the comments of the previous bullet).
However, it is possible to obtain the per-unit value (not the MKS value)
of some of the amortisseur circuit parameters (specifically, the mutual
inductances), because, as we shall see in Section 3.0 below, in per-unit,
all direct-axis mutuals are equal and all quadrature-axis mutuals are
equal! In other words:
 D-axis mutuals:
F-d winding mutual, kMF
D-d winding mutual, kMD
F-D winding mutual, MR (is called MX in some texts)
That is, we will show that in per-unit, kM Fu  kM Du  M Ru
 Q-axis mutuals:
G-q winding mutual, kMG
Q-q winding mutual, kMQ
G-Q winding mutual, MY
That is, we will show that in per-unit, kM Qu  kM Gu  M Yu

19
2.4 Applying the bases to voltage equations:
Recall our voltage equation as written in MKS units:
ra  3rn 0 0 0 0 0 0 
 v0     i0 
 v  3 3
 0 rb Lq 0 0  M Q  M G  i 
 d   2 2  d 
 vq   3 3   iq 
   0  LD rc  MF  MD 0 0  
  vF     2 2  iF 
 0   0 0 0 rF 0 0 0  iD 
   0 0 0 0 rD 0 0  
 0    iQ 
   0 0 0 0 0 rQ 0  i 
 0   0  G 
 0 0 0 0 0 rG 
 
 L0  3Ln 0 0 0 0 0 0 
 
 3 3 
 0 Ld 0 MF MD 0 0   i0 
2 2
   i 
M G   
 0 3 3 d
0 Lq 0 0 MQ
 2 2  
iq
 3  
 0 MF 0 LF MR 0 0  iF 
 2  
 3  iD 
 0 MD 0 MR LD 0 0  i 
 2  Q 
 3  
 0 0 MQ 0 0 LQ M Y  iG 
2
 
 0 0
3
MG 0 0 MY LG 
 2 

Let’s normalize them using our chosen bases to obtain the equations in per-
unit. The per-unit equations should appear as above when done, except that
everything must be in per-unit.
Step 1: Replace all MKS voltages on the left with
 the product of their per-unit value and their base value (use VB
for the first 3 equations and VFB, VDB, VQB, VGB for the last four
equations),
and replace all currents on the right with
 the product of their per-unit value and their base value (use IB
for the first 3 equations and IFB, IDB, IQB, IGB for the last four
equations).
This results in a relation similar to eq. 4.60 in the text, as follows….

20
 ra  3rn 0 0 0 0 0 0 
 v0uVB     i0u I B 
 v V  3 3
 0 rb Lq 0 0  MQ  MG  i I 
 du B   2 2   du B 
 vquVB   3 3   iqu I B 
   0  LD rc  MF  MD 0 0  
 vFuVFB     2 2   i Fu I FB 
 0   0 0 0 rF 0 0 0 
 iDu I DB 
   0 0 0 0 rD 0 0  
 0    iQu I QB 
   0 0 0 0 0 rQ 0  
 0   0   iGu I GB 
 0 0 0 0 0 rG 
 
 L0  3Ln 0 0 0 0 0 0 
 
 3 3 
 0 Ld 0 MF MD 0 0   i I 
2 2 0u B
   i I 
M G  
 0 3 3 du B 
0 Lq 0 0 MQ
 2 2  iqu I B 

 3  
 0 MF 0 LF MR 0 0   iFu I FB 
 2  
 3  i Du I DB 
 0 MD 0 MR LD 0 0  i I 
 2   Qu QB 
 3   
 0 0 MQ 0 0 LQ M Y  iGu I GB 
2
 
 0 0
3
MG 0 0 MY LG 
 2 
(eq. 4.60’)
Step2: For each of the equations in the above, we need to divide through by
the voltage base. For those equations containing , we replace it with
=uB (B=Re). Then we do some algebra on each equation to express
the coefficients of each current and current derivative as per-unitized self or
mutual inductances. As an example, the second equation is done for you in
the text; here, I will do the last equation, corresponding to the G-winding….
3
0(VGB )  rG iGu I GB  M G iqu I B  M Y iQu I QB  LG iGu I GB
2
Step 2a: Divide through by VGB to obtain:

21
  rG 3 MG  M L
0 iGu  iqu  Y iQu  G iGu
VGB 2 VGB VGB VGB
I GB IB I QB I GB
The first term has a denominator of RGB. The last three terms are not so
obvious. We desire them to have denominators of MGB, MYB, and LGB,
respectively, where, from above, we recall M GB  kG LB , M YB  kG kQ LB , LGB  kG2 LB ,
VGB I I
where k G   B , and k Q  B .
VB I GB I QB
Step 2b: Let’s multiply the denominator of the last three terms by VB/VB.
This results in:

 rG 3 MG  MY  LG 
0 iGu  iqu  iQu  i
RGB 2 VGB VB VGB VB VGB VB Gu
VB I B VB I QB VB I GB
Step 2c: Let’s multiply the denominator of the last two terms by IB/IB. This
results in:

 rG 3 MG  MY LG
0 iGu  iqu  iQu  iGu
RGB 2 VGB VB VGB VB I B VGB VB I B
VB I B VB I B I QB VB I B I GB

VGB I I
Step 2d: Recall the k-factors (pg 96 of text): k G   B , and k Q  B .
VB I GB I QB
Substitution yields:
r 3 MG  MY  LG 
0  G iGu  iqu  iQu  iGu
RGB 2 VB VB 2 VB
kG kG kQ kG
IB IB IB

Step 2e: We are close now, as we need MGB=kGLB, MYB=kGkQLB, and

LGB=(kG)2LB, respectively, on the denominator of the last three terms. Recall

22
that LB=VB/(BIB), so we need to divide top and bottom on the denominators
of the last three terms by B. Doing so yields:

 rG 3 MG MY LG
0 iGu  iqu  iQu  iGu
RGB 2 V V 2 VB
kG B  B kG kQ B  B kG B
BIB B IB B IB

Step 2f: And substituting in LB results in:

 rG 3 MG  MY L
0 iGu  iqu  iQu  2 G iGu
RGB 2 k G LB  B k G k Q LB B k G LB B

Step 2g: Recalling that MGB=kGLB, MYB=kGkQLB, and LGB=(kG)2LB , we

may write:

 rG 3 MG  MY  LG 
0 iGu  iqu  iQu  iGu
RGB 2 M GB B M YB B LGB B
which results in

3 M Gu  M L
0  rGu iGu  iqu  Yu iQu  Gu iGu
2 B B B
Step 2h: However, we still have one problem. Recall that we want the
equations to be identical in pu to their form in MKS units. But in the last
equation, we still have B, which does not appear in our MKS equation. We
can take care of it, however, by recalling that B=1/tB, so that:

23
 1  1 diqu diqu diqu
iqu    ;
B 1 dt d  t  d
tB
 tB 
1  1 diQu diQu diQu
iQu    ;
B 1 dt d  t  d
tB
 tB 
1  1 diGu diGu di
iGu    Gu
B 1 dt
d  t  d
tB
 tB 
where =t/tB is the normalized time.

With this last change, we can write, finally, that

3
0  rG iGu  M Gu iqu  M Yu iQu  LGu iGu
2
which is the per-unitized form of the last equation in eq. (4.60’).

Note that it is exactly the same form as the original equation in MKS units.

Similar work can be done for the other equations (and you should try to do
one of the others yourself), resulting in equations similar to eq. 4.74 in your
text:

24
 vd   r 0 0 Lq kM Q kM G   id 
 v   0 rF 0 0 0 0  i 
 F
   F 
 0   0 0 rD 0 0 0  iD 
      
 v q   Ld  kM F  kM D r 0 0   iq 
 0   0 0 0 0 rQ 0  iQ 
    
 0   0 0 0 0 0 rG  iG 
 Ld kM F kM D 0 0 0   id 
kM LF MR 0 0 0  iF 
 F  
kM MR LD 0 0 0  iD 
 D  
 0 0 0 Lq kM Q kM G   iq 
 0 0 0 kM Q LQ M Y  iQ 
  
 0 0 0 kM G MY LG  iG 
(eq. 4.74’)

Note that in the above equation,

 The “u” subscript was dropped; however, all parameters are in
per-unit.
 We have dropped the zero-sequence voltage equation since we
will be interested in balanced conditions for stability studies. (A
system having a three-phase fault, considered to be, usually, the
most severe, is still a balanced system. This does not mean that
we cannot analyze unbalanced faults using stability programs. It
is possible to analyze the effects of unbalanced faults on the
positive sequence network represented in stability programs – see
Kimbark Vol I, pp. 220-221). Otherwise, eq. (4.74’) is precisely
the same as eqt. 4.39 in your text (see the same equation as 4.39,
except for the “G” winding included, in the notes called
“macheqts”– we called it there, eq. 4.39’).

25
 The equations are rearranged to better display the coupling and
decoupling between the various circuits. This coupling can be
well illustrated by a figure similar to Fig. 4.3 in your text, given
below as Fig. 4.3’. Notice that the coupling between the F and D
windings is captured by MX. We have called this mutual
inductance MR in our work above to remain consistent with A&F.
These are physically-realizable circuits for which KVL in each of
the 6 circuits results in the 6 equations of 4.74’ above.
r0+3rn i0
+
L0+3Ln v0

-
rF

iF 
+
vF LF ra id
- kMF +
rD MX Ld vd
iD 
+  kMD
-
vD=0 LD - +
-  q

“Dot” convention:
a. If the reference current direction enters the dotted terminal of a coil, the reference polarity
of the voltage that it induces in the other coil is positive at its dotted terminal.
b. If the reference current direction leaves the dotted terminal of a coil, the reference polarity
of the voltage that it induces in the other coil is negative at its dotted terminal.
rG

+ iG  ra
vG=0 LG
kMG +
-
rQ MY Lq vq

  -
+ kMQ +
vQ=0 LQ -
-  d

Fig 4.3’

26
Now let’s make some definitions:

r 0 0 0 0 0  0 0 0 Lq kM Q kM G 
0 r 0 0 0 0   0 0 0 0 0 0 
 F 
0 0 rD 0 0 0  0 0 0 0 0 0 
R ; N  
0 0 0 r 0 0   Ld  kM F  kM D 0 0 0 
0 0 0 0 rQ 0  0 0 0 0 0 0 
   
0 0 0 0 0 rG   0 0 0 0 0 0 
 Ld kM F kM D 0 0 0 
kM LF MR 0 0 0 
 F
kM MR LD 0 0 0 
L D 
 0 0 0 Lq kM Q kM G 
 0 0 0 kM Q LQ MY 
 
 0 0 0 kM G MY LG 
 vd   id 
 v  i 
 F  F
 0  i D 
v ; i 
v
 q   iq 
 0  iQ 
   
 0  iG 

With these definitions, we rewrite eqt. (4.74’) in compact notation:

v  ( R   N )i  Li (eq. 4.75)

We may solve eq. (4.75) for di/dt so that it is in state-space form:

i   L 1 ( R   N )i  L1 v (eq. 4.76)

27
3.0 Per-unit mutuals (See Section 4.11)
A useful observation regarding per-unit values of MF, MD, and MR:

Recall our definitions of the D-axis k-factors:

I B VFB
kF  
I FB VB
I B VDB
kD  
I DB VB
and that we developed (see eqs. 4.54, pg 11 of these notes):
Lmd I B2  LmF I FB
2
 LmD I DB
2
 kM F I B I FB  kM D I B I DB  M R I FB I DB (18)
1 2 3 4 5 6
From the first and fourth expression in eq (18), we have:
Lmd I B2  kM F I B I FB
Thus,
I B kM F
 (19)
I FB Lmd
Likewise, from the first and fifth, and from the fourth and sixth
expressions in eq (18), we have:
Lmd I B2  kM D I B I DB kM F I B I FB  M R I FB I DB
Thus,
I B kM D IB M
 and  R (20)
I DB Lmd I DB kM F
From the definitions of the k-factors, and eqs (19) and (20), we have:
kM F and k  kM D  M R
kF  D
(21)
Lmd Lmd kM F
And from eq. 4.57 in text (also see p13 of these notes), we find

28
 M FB  k F LB , M DB  k D LB , M RB  k F k D LB (22)
which we obtained by using the fact that base mutuals must be the
geometric mean of the base self-inductances (see prob 4.18).

Now, recall the elements in the per-unitized voltage equations as

given by eq. 4.74’ (see page 24 of these notes).
 vd   r 0 0 Lq kM Q kM G   id 
 v   0 r 0 0 0 0  i 
 F   F  F 
 0   0 0 rD 0 0 0  i D 
 
   
 v q    L d  kM F  kM D r 0 0   iq 
 0   0 0 0 0 rQ 0  iQ 
    
 0   0 0 0 0 0 rG  iG 
 Ld kM F kM D 0 0 0   id 
kM L M 0 0 0  
 F F R  i F 
kM D M R LD 0 0 0  iD 
   
 0 0 0 L q kM Q kM G   iq 
 0 0 0 kM Q LQ M Y  iQ 
  
 0 0 0 kM G M Y LG  iG 
4.74’
In particular, consider the mutual terms in the last matrix for the
direct axis. This is the upper left-hand 3x3 block, in the blue box.
These terms, in pu, are by definition the ratio of the term in MKS to
the appropriate base. Therefore:
kM F
 Stator-field mutual: kM Fu  .
M FB
Substituting for MFB from eq. (22) and then kF from eq. (21)
results in:

29
 kM F kM F kM F Lmd Lmd
kM Fu      Lmdu
M FB k F LB kM F LB LB
kM D
 Stator-D-winding damper mutual: kM Du  .
M DB
Substituting for MDB from eq. (22) and then kD from eq. (21)
results in:
kM D kM D kM D Lmd Lmd
kM Du      Lmdu
M DB k D LB kM D LB LB
MR
 Field-D-winding damper mutual: M Ru 
M RB
Substituting for MRB from eq. (22) and then kF and kD from eq.
(21) (using the 2nd expression for kD in eq. (4)) results in:
MR MR M L kM F Lmd
M Ru    R md   Lmdu
M RB k F k D LB kM F M R LB LB
Important fact: In per-unit, all d-axis mutuals are numerically equal!
We will define a new term for them, LAD, as the per-unit value of
any d-axis mutual inductance, so that:
LAD  Lmdu  kM Fu  kM Du  M Ru
Also note that, since the mutual is the difference between the self
and the leakage, this implies
Ldu-ldu=LDu-lDu=LFu-lFu=LAD
The above relations are given in eqs. 4.107 and 4.108 in your text.

We can go through a similar process for the q-axis mutuals (from

4.74’, we see that these are the terms in the lower right-hand block
of the matrix, kMQ, kMG, and MY). I will leave this for you to do.
The result is:

30
 LAQ  Lmqu  kM Qu  kM Gu  M Yu
Lqu-lqu=LQu-lQu=LGu-lGu=LAQ
The above relations are given by eq. 4.109 in your text, except for
the addition of the G-term and the Y-term.

LAD and LAQ are very important for drawing the equivalent circuits.

They are also important in dealing with saturation because they

provide for the definition of the per-unit mutual flux (we will see
this in our development of the flux-linkage state-space model).

4.0 Equivalent Circuits (See Section 4.11)

Let’s return to the voltage equations that we had before we folded
in the speed voltage terms. They were:
 
 L0 0 0 0 0 0 0 
 
 3 3 
0 Ld 0 MF MD 0 0   i0 
 v0   ra 0 0 0 0 0 0   i0   2 2
 v        i 
M G   
 d  0 rb 0 0 0 0 0  id   3 3 d
0 0 Lq 0 0 MQ
 vq  0 0 rc 0 0 0 0   iq   2 2  iq
      3  
  vF     0 0 0 rF 0 0 0 iF    0 MF 0 LF MR 0 0  iF 
 0    
0 0 iD  
2
  
0 0 0 rD 0
  3  iD 
 0  0 0 0 0 0 rQ 0  iQ   0 MD 0 MR LD 0 0  i 
      2  Q 
 0  0 0 0 0 0 0 rG  iG   3  
 0 0 MQ 0 0 LQ M Y  iG 
2
 
 0 0
3
MG 0 0 MY LG 
 2 
 0  3rni0  3Lni0 
    0

 q  
 d   0 
   
 0  0 
 0   0 
   
 0   0 
 0   
   0 

Assume all of the above is in per-unit (but we have dropped the u-

subscript).

31
There is some advantage to re-writing these equations in terms of
LAD and LAQ. For example, consider the d-axis equation. It is:

vd  rid  Ld id  kM F iF  kM D iD  q

Recall that Ld=Lmd+ld Lmd=Ld-ld

Let’s modify the d-axis voltage equation by adding and subtracting

ld (did/dt) :
vd  ri d  Ld id  l d id  l d id  kM F iF  kM D iD  q ,
which can be written as:
vd  rid  l d id  [( Ld  l d )id  kM F iF  kM D iD ]  q
The advantage to this is that, in per-unit, we recall that
Ld-ld=kMF=kMDLAD. Therefore,
vd  ri d  l d id  L AD [id  iF  iD ]  q

Let’s repeat this for the G-axis equation, which is, from the matrix
equation at the beginning of this section:
vG  0  rG iG  kM G iq  M Y iQ  LG iG

Let’s modify the G-winding voltage equation by adding and

subtracting lG (diG/dt):

vG  0  rG iG  kM G iq  M Y iQ  LG iG  lG iG  lG iG

which can be written as:
vG  0  rG iG  kM G iq  M Y iQ  ( LG  lG )iG  lG iG
The advantage to this is that, in pu, we have LG-lG=kMG=MYLAQ.

Therefore
vG  0  rG iG  lG iG  L AQ (iq  iQ  iG )

32
Repeating this procedure for the F, D, and q equations, and then
summarizing, we obtain:

D-axis relations:
vd  ri d  l d id  L AD [id  iF  iD ]  q
 v F  rF i F  l F iF  L AD [id  iF  iD ]
v D  0  rD i D  l D iD  L AD [id  iF  iD ]

Q-axis relations:
vq  ri q  l q iq  L AQ (iq  iQ  iG )  d
vQ  0  riQ  lQ iQ  L AQ (iq  iQ  iG )
vG  0  rG iG  lG iG  L AQ (iq  iQ  iG )

We desire to draw circuits that are characterized by these equations.

Note:
 The D-axis relations are coupled through the LAD terms.
 This term, for each equation, may be represented by a single
“center” branch.
 The other terms, for each equation, may be represented as single
branches which feed the center branch.
This results in the circuit of Fig 4.5 in your text.

Similar reasoning results in the circuit of Fig. 4.6 in your text.

We redraw these circuits below.

33
 vd  ri d  l d id  L AD [id  iF  iD ]  q
 v F  rF i F  l F iF  L AD [id  iF  iD ] id

v D  0  rD i D  l D iD  L AD [id  iF  iD ]

ld r

lF lD

rF iF rD iD LAD iF+iD+id vd

vF -
+

-- +

q

Direct-axis equivalent circuit:

The above is the same as Fig. 4.5 in your text

vq  ri q  l q iq  L AQ (iq  iQ  iG )  d

vQ  0  riQ  lQ iQ  L AQ (iq  iQ  iG )
iq
vG  0  rG iG  lG iG  L AQ (iq  iQ  iG )
lq r

lG lQ

rG iG rQ iQ LAQ iQ+iG+iq vq

+ --

d

Quadrature-axis equivalent circuit:

The above is the same as Fig. 4.6 in your text, except we have
included the G-circuit

34
The ability to draw these circuits is a direct result of the LAD and LAQ
relations that occur only in per-unit. Therefore, it is important to be
in the per-unit system when utilizing these circuits. And the “equal
mutuals” effect came as a result of the fact that we chose our base
currents according to the following criteria (see p. 7 of these notes):
We select the base currents for the four rotor-side windings F,
D (Q, G) to produce the same mutual flux in the air gap as
produced by the stator-side base current IB flowing in the
corresponding fictitious d-axis (q-axis) coil.
See Appendix C of your text, at the top of p. 549, for articulation of
this fact.

These equivalent circuits are useful for:

 Remembering the voltage relations.
 Gaining physical understanding of relations between d-q-F-D-Q-
G quantities.
 Reading the literature, where you will see them often.

35