Flux-linkage equations for 7-winding representation (similar to eq. 4.

11 in text)

a Laa Lab L ba Lbb b c Lca Lbc F = LFa LFb D L LDb Da Q LQa LQb G LGa LGb
Stator-stator terms:
Laa=Ls+Lmcos2 Lab=-[Ms+Lmcos2(+30)] Lac=-[Ms+Lmcos2(+150)] Lba=-[Ms+Lmcos2(+30)] Lbb=Ls+Lmcos2(-120) Lbc=-[Ms+Lmcos2(-90)] Lca=-[Ms+Lmcos2(+150)] Lcb=-[Ms+Lmcos2(-90)] Lcc=Ls+Lmcos2(-240)

Lac Lbc Lcc LFc LDc LQc LGc

LaF LbF LcF LFF LDF LQF LGF

i a LbD LbQ LbG ib LcD LcQ LcG ic LFD LFQ LFG iF LDD LDQ LDG iD iQ LQD LQQ LQG iG LGD LGQ LGG LaD LaQ LaG
Rotor-Stator terms:
LFa=MFcos LFb=MFcos(-120) LFc=MFcos(-240) LDa=MDcos LDb=MDcos(-120) LDc=MDcos(-240) LQa=MQsin LQb=MQsin(-120) LQc=MQsin(-240) LGa=MGsin LGb=MGsin(-120) LGc=MGsin(-240)

The above terms are defined as follows:

Stator-rotor terms:
LaF=MFcos LaD=MDcos LaQ=MQsin LaG=MGsin LbF=MFcos(-120) LbD=MDcos(-120) LbQ=MQsin(-120) LbG=MGsin(-120) LcF=MFcos(-240) LcD=MDcos(-240) LcQ=MQsin(-240) LcG=MGsin(-240)

Rotor-rotor terms:
LFF=LF LFD=MR LFQ= LFG =0 LDF=MR LDD=LD LDQ=LDG=0 LQF=LQD=0 LQQ=LQ LQG=MY

LGF=LGD=0 LGQ=MY LGG=LG

So the compact form of the flux linkage equations are

i abc L aa abc = [L ]i = L FDQG FDQG Ra

L aR i abc L RR i FDQG

(eq. L)

which, when expanded with the expressions for self and mutual inductances, become:

a L S + L m cos 2 M F cos M D cos M Q sin M G sin [ M S + L m cos 2( + 30)] [ M S + L m cos 2( + 150)] i a L S + L m cos 2( 120) [ M S + L m cos 2( 90)] M F cos( 120) M D cos( 120) M Q sin( 120) M G sin( 120) i b b [ M S + L m cos 2( + 30)] c [ M S + L m cos 2( + 150) [ M S + L m cos 2( 90)] L S + L m cos 2( 240) M F cos( 240) M D cos( 240) M Q sin( 240) M G sin( 240) i c M F cos M F cos( 120) M F cos( 240) LF MR 0 0 F = i F D i D M D cos M D cos( 120) M D cos( 240) MR LD 0 0 M Q sin M Q sin( 120) M Q sin( 240) LQ MY 0 0 Q i Q i M G sin M G sin( 120) M G sin( 240) MY LG 0 0 G G

(eq. L-ex)

Voltage equations

Consider the stator circuit appears as in Fig. 1:

Fig. 1 The current direction in the phases, which is out of the terminals for generator operation, produces a flux that is along the negative axis of the respective phase axis. We assume that the neutral conductor is not coupled with any other circuit. We can write a voltage equation for each of the phase windings as follows:

& va = ia ra a + vn & vb = ib rb b + vn & vc = ic rc c + vn

We may also write a voltage equation for the neutral circuit as follows: Now lets look at the rotor circuits. There are four of them. Fig. 2: D-Axis Field

& & & & vn = in rn Ln in = (ia + ib + ic )rn Ln (ia + ib + ic )

& v F = rF iF + F & v F = rF iF F
Fig. 3: D-Axis Damper

& 0 = rD iD D

Fig. 4: Q-Axis Damper

& 0 = rQ iQ Q
Fig. 5: Q-Axis Field

& 0 = rG iG G

Putting all of these equations together in matrix form, we have that:

ra va v b 0 0 vc v F = 0 0 0 0 0 0 0

& 0 0 0 0 0 0 i a a v n & r b 0 0 0 0 0 ib b v n & 0 rc 0 0 0 0 ic c v n & 0 0 rF 0 0 0 i F F + 0 i & 0 0 0 0 rD 0 0 D D 0 0 0 0 rQ 0 iQ 0 & Q 0 0 0 0 0 rG iG 0 & G

(eq 4.23)

We can write this more compactly, similar to eq. 4.26 in text:

& v abc R abc 0 i abc abc v n + & v = 0 R i FDQG FDQG FDQG 0 FDQG
Motivation for Parks Transformation

(eq. 4.26)

We would like to get the above equation into state-space form (xdot=Ax) so that we can combine it with our inertial equations (and then be able to apply numerical integration and solve them together). We notice, however, that we have two different types of state variables in the above equations: flux linkages () and currents (i). So we need to eliminate one of them, and this is not hard since we have that flux linkages can be easily expressed as functions of the currents that produce them. For example, for a single conductor, we write that =Li (see also first equation in these notes). But eq. 4.26 has derivatives on . Again, no problem, since d/dt=d(Li)/dt.

It is here that we run into trouble, since the inductances that we are dealing with are, in general, functions of , which is itself a function of time. Therefore the inductances are functions of time, and differentiation of flux linkages results in expressions like:

d dL di = i+ L dt dt dt

The differentiation with respect to L, dL/dt, will result in a time-varying coefficient on the state variable. When we replace, in eq. 4.26, the derivatives on with the derivatives on i, and then solve for the derivatives on i (in order to obtain x = A x ), we will obtain current variables on the & right-hand-side that have time varying coefficients, i.e., the coefficient matrix A will not be constant. This means that we will have to deal with differential equations with time varying coefficients, which are generally more difficult to solve than differential equations with constant coefficients. This presents some significant difficulties, in terms of solution, that we would like to avoid. We look for a different approach. The different approach is based on the observation that our trouble comes from the inductances related to the stator (phase windings): Stator self inductances Stator-stator mutual inductances Stator-rotor mutual inductances i.e., all of these have time-varying inductances. In order to alleviate the trouble, we will project the a-b-c currents onto the D and Q axes. In making these projections, we want to obtain expressions for the components of the stator currents that are in phase with the D and Q axes.

One can visualize the projection by thinking of the a-b-c currents as having sinusoidal variation IN TIME along their respective axes. The picture below illustrates for the a-phase.
d-axis ia a id iq a' q-axis

Fig. 6
Decomposing the b-phase currents and the c-phase currents in the same way, and then adding them up, provides us with:

id = k d (ia cos + ib cos( 120) + ic cos( + 120) ) iq = k q (ia sin + ib sin( 120) + ic sin( + 120) )

Here, the constants kd and kq are chosen so as to simplify the numerical coefficients in the generalized KVL equations that we will get. But note: we have transformed 3 variables ia, ib, and ic into two variables id and iq. This yields an undetermined system, meaning We can uniquely transform ia, ib, and ic to id and iq We cannot uniquely transform id and iq to ia, ib, and ic.

So we need a third current. We take this current proportional to the zerosequence current: (i-zero) We note that, under balanced conditions, i0 is zero, and therefore produces no flux at all. In fact, it is possible to show that i0 produces no flux which links the rotor windings at all (see Concordias book, pg. 14 and also Kimbark Vol III, pg. 60). The implication is that under all conditions, id and iq produce the exact same flux as ia, ib, and ic. We write our transformation more compactly as:

i0 = k0 (ia + ib + ic )

k0 k0 i0 k 0 ia id = k d cos k d cos( 120) k d cos( + 120) ib iq k q sin k q sin( 120) k q sin( + 120) ic 14444444 244444444 { 4 3{
i odq P i abc

i 0 dq = Pi abc

(eq. 4.3)

We may also operate on the voltages and fluxes in exactly the same way: (eq. 4.7) v 0 dq = P v abc , 0 dq = P abc This transformation resulted from the work done by Blondel (1923), Doherty and Nickle (1926), and Park (1929, 1933), and as a result, is usually called Parks transformation, and the transformation matrix P is usually called Parks transformation matrix or just Parks matrix. In Parks original paper, he used k0=1.0 and kd=1.0, and kq=-1.0 (he assumed the q-axis as leading the d-axis; if he would have assumed the qaxis as lagging the d-axis, as we have done, then he would have had kq=1.0). However, there are two main disadvantages with this choice:
8

1. The transformation is not orthogonal. This means that P-1PT. If the transformation were orthogonal (P-1=PT), then the power calculation, which is
1

p=v

T abc abc , is also given by

p=v

T 0 dq 0 dq . This can

be proven (see eq. 4.10 in text) since, from eqs. 4.3 and 4.7,

P v 0 dq = v abc and P i 0 dq = i abc , we may write:

p=v p=v =v

T abc abc

= P v 0 dq =v
T odq

Recalling that (ab)T=bTaT, the above is:

T abc abc

) (P (P ) (P
T 1 T

i 0 dq i 0 dq

) )= v

T odq

(P )(P 1 i 0 dq )

T odq 0 dq

2. The transformed mutual inductances, when per-unitized, do not provide that Mjk=Mkj, implying that the per-unit inductance matrix is not symmetric. This prevents us from finding a real physical circuit to use in modeling the transformed system. In order to overcome these problems, we (Anderson and Fouad) make a different choice of constants, according to:

k0 =

The choice of k0, when applied to eq. (i-zero) above, results in:

1 , 3

kd = kq =

2 3

1 2 1 1 1 i0 = (ia + ib + ic ) = ia + ib + ic 3 3 2 2 2 2 So we see that the factor is the multiplier on all three equations, 3
resulting in a Parks transformation (and the one that we will use) as:
9

 1 2 2 cos P= 3 sin

1 1 2 2 cos( 120) cos( + 120) sin( 120) sin( + 120)

Parks Transformation Applied toVoltage equations for 7-winding representation Now perform the Parks transformation on both sides of the voltage equation (eq. 4.23 or 4.26). Note that we apply P to only the a-b-c quantities, i.e., we leave the F-D-Q-G quantities alone (the rotor-rotor quantities are constants and therefore need no transformation) since these quantities are already on the rotor (and the rotor-rotor inductances are already constants). This means we need to multiply eq. (4.23 or 4.26) through by a matrix

Recall (4.26) is:

P 0 0 U where U4 is a 4x4 identity matrix. 4

& v abc R abc 0 i abc abc v n + & v = 0 R i FDQG FDQG FDQG 0 FDQG
(eq. 4.26)

Multiplying through by our matrix, we obtain: & P 0 v abc P 0 R abc 0 i abc P 0 abc P 0 v n + = 0 U 0 R 0 U i 0 U v & FDQG 0 U 4 0 4 FDQG 4 4 1 4 43 1442443 14444 2FDQG 44 1442443 4 2 4 4 444 FDQG 3
term 1

term 2

term 3

term 4

(eq. tve1)

10

We need to express eq. (tve1) in terms of 0-d-q quantities. In what follows below, we do this one term at a time. Our general procedure will be to replace the a-b-c quantities with 0-d-q quantities and then simplify. The easiest one is term 1, so we will begin with it. Term 1:

P 0 v abc Pv abc v 0 dq = v = v 0 U v 4 FDQG FDQG FDQG

Term 2: This term is:

Note that

P 0 R abc 0 i abc 0 U 0 R FDQG i FDQG 4

Substitution yields:

i 0 dq P 0 i abc i abc P 1 0 i 0 dq i = 0 U i i = i FDQG 4 FDQG FDQG 0 U 4 FDQG

1 0 i 0 dq P 0 R abc 0 P 0 U 0 R U 4 i FDQG FDQG 0 4

P R abc 0 P 1 0 i 0 dq P R abc P 1 = = 0 0 R FDQG 0 U 4 i FDQG

i 0 dq i FDQG R FDQG 0

Note that the upper left-hand element has a diagonal matrix in the middle of two orthogonal matrices.

11

Fact: If P is orgthogonal, then P R abc P = diagonal having equal elements on the diagonal. You can test this as follows. Let

R abc if R abc is

It is easy to prove as follows. If R is a diagonal matrix with all of its diagonal elements the same, call them r, then R=rU. Then ARAT= ArUAT= rAUAT=rAAT=rU=R. Here, we will assume ra=rb=rc which is very typical of synchronous machines and simply implies that all phase windings are equal length with the same type of conductor, which is always the case. Therefore term 2 is just:

0 1 0 A = 0 0 1 . It is easy to show this is orthogonal using A AT=U. 1 0 0 2 0 0 Then try multiplying A R AT where R = 0 2 0 . 0 0 2

P 0 R abc 0 i abc i 0 U 0 R FDQG FDQG 4 P R abc P 1 0 i 0 dq R abc 0 i 0 dq = = 0 R FDQG i FDQG 0 R FDQG i FDQG
12

Repeating our equation (tve1) here for convenience. & P 0 v abc P 0 R abc 0 i abc P 0 abc P 0 v n + = 0 U 0 R 0 U i 0 U v & FDQG 0 U 4 0 4 FDQG 4 4 1 4 43 1442443 14444 2FDQG 44 1442443 4 2 4 4 444 FDQG 3
term 1

term 2

term 3

term 4

and recalling what we have done so far:

TERM 2:

P 0 v abc Pv abc v 0 dq TERM 1: = v = v 0 U v 4 FDQG FDQG FDQG

P 0 R abc 0 i abc 0 U 0 R FDQG i FDQG 4 P R abc P 1 0 i 0 dq R abc 0 i 0 dq = = 0 R FDQG i FDQG 0 R FDQG i FDQG
Substituting, we obtain:

& v 0 dq R abc 0 i 0 dq P 0 abc P 0 v n & + = 0 R FDQG i FDQG 0 U 4 FDQG 0 U 4 0 v FDQG 4 3 44 2444 4 4 123 1 4 3 1442443 14 244 4 4
term 1 term 2 term 3 term 4
eq. (tve2)

Now we observe that terms 3 and 4 have variables not in terms of 0-d-q quantities. We work on term 4 next (before term 3) because it is easier. Term 4: Observe that vn=[vn vn vn]T. Therefore, when we multiply Pvn, we get elements in the second and third rows of P being scaled by the same constant (vn) and then summed. Consider these elements in the second and third rows of P, below.

13

 1 2 2 P= cos 3 sin

1 1 2 2 cos( 120) cos( + 120) sin( 120) sin( + 120)

So the product of the second row with vn, or of the third row and vn, will include a summation of symmetrical components, which will be zero. So the only non-zero element in Pvn will be the product of the first row of P and vn, which is

1 3

1 3

But recall the voltage equation indicates that: Also, recall that

v n 1 3vn v n = 3 3 v n

(*)

& & & & vn = in rn Ln in = (ia + ib + ic ) rn Ln (ia + ib + ic ) (**)

Substitution of (***) into (**) yields:

1 i0 = (ia + ib + ic ) ia + ib + ic = 3i0 3

(***)

& vn = ( 3i0 ) rn Ln ( 3i0 )

and replacing vn in (*) with this, we have:

14

3rn i0 P 0 v n Pv n 0 U 0 = 0 = 4

& 3Ln i0 0 0 n 0 dq 0 = 0 0 0 0

(*#)

where n0dq is the first 3 elements and 0 is the last 4 elements. Now recall eqt. (tve2), repeated here for convenience:

& v 0 dq R abc 0 i 0 dq P 0 abc P 0 v n & + = 0 R FDQG i FDQG 0 U 4 FDQG 0 U 4 0 v FDQG 4 3 44 2444 4 4 123 1 4 3 1442443 14 244 4 4
term 1 term 2 term 3 term 4

and substitute in eqt. (*#) to obtain

& v 0 dq R abc 0 i 0 dq P 0 abc n 0 dq & + = 0 R FDQG i FDQG 0 U 4 FDQG 0 v FDQG 44 2444 4 4 2 123 1 4 3 1442443 1 3
term 1 term 2 term 3 term 4
eq. (tve3)

And so now the only a-b-c variables remaining are in term 3. So lets work on term 3.
Term 3: Term 3 is:

& & P 0 abc P abc =& 0 U & FDQG FDQG 4

15

(4.30)

& 1. Obtain P abc in terms of the 0-d-q quantities. 2. Express all of term 3 in terms of currents instead of flux linkages.
To begin this task, recall that 0 dq

So we need to do two things:

= P abc , and take derivatives of

both sides. Note in differentiating the right-hand-side, we need to account for the fact that P is time-dependent. Thus:

& Solving for P abc , we obtain:

& & & 0 dq = P abc + P abc

& & & P abc = 0 dq P abc

(#)

But the right-hand side still has abc . We can eliminate this using

abc = P 1 odq

Substitution into eq. (#) yields:

Substitution of eq. (4.31) into eq. (4.30) above yields:

& & & 1 P abc = 0 dq P P odq & Now we have expressed P abc in terms of the 0-d-q quantities.

(4.31)

& & & & 1 P 0 abc P abc 0 dq P P 0 dq = = 0 U & FDQG FDQG FDQG & & 0 4
term 3a term 3b

& So we have accomplished our objective 1, which was to obtain P abc in terms of the 0-d-q quantities. Lets substitute the above equation into eq. (tve3)

16

& v 0 dq R abc 0 i 0 dq P 0 abc n 0 dq & + = 0 R FDQG i FDQG 0 U 4 FDQG 0 v FDQG 44 2444 4 4 2 123 1 4 3 1442443 1 3
term 1 term 2 term 3 term 4
eq. (tve3)

to obtain

& & v 0 dq R abc 0 i 0 dq 0 dq P P 1 0 dq n 0 dq + + & = 0 R FDQG i FDQG FDQG 0 v FDQG 103 44 2444 4 4 4243 2 123 1 4 3 1 24 1 4 3
term 1 term 2 term 3a term 3b term 4
eq. (tve4)

Now we need to accomplish our objective 2, which is to express all of term 3 in terms of currents instead of flux linkages. To do this, lets investigate terms 3a and 3b one at a time. Lets start with term 3a. Term 3a: So term 3a is:

& 0 dq & FDQG

Our goal is to see if we can express this in terms of currents, which means we will need to use inductances. Lets start by looking at the same expression but without the derivatives, since we know how to write this using Parks transformation and a-b-c flux linkages. This is:

0 dq P 0 abc = 0 U 4 FDQG FDQG

abc L aa = L FDQG Ra

(eq. 3a-1)

Now to write eq. (3a-1) in terms of the 0dq/FDQG currents (instead of 0dq/FDQG flux linkages), recall from eq. (L), pg. 2, repeated here for convenience

L aR i abc L RR i FDQG
17

(eq. 3a-2)

that the vector of abc/FDQG flux linkages on the right of (eq. 3a-1) is related through the inductance matrix to the abc/FDQG currents. Now recall that the abc/FDQG currents may be related to the 0dq/FDQG currents using the inverse Park Transformation according to:
1 i abc P 0 i 0 dq i = 0 U 4 i FDQG FDQG

(eq. 3a-3)

Substitution of (3a-3) into (3a-2) and then what results into (3a-1), we have

0 dq P 0 L aa = 0 U L 4 Ra FDQG

Performing the above matrix multiplication, we obtain.

L aR P 1 0 i 0 dq L RR 0 U 4 i FDQG

Now we need to go through each of these four matrix multiplications. I will here omit the details and just give the results (note also in what follows the definition of additional nomenclature for each of the four submatrices): Submatrix (1,1):

0 dq P L aa P 1 = 1 L Ra P FDQG

P L aR i 0 dq L RR i FDQG

L0 1 P L aa P = 0 0

0 Ld 0

0 0 L 0 dq Lq

where L0=LS-2MS, Ld=LS+MS+(3/2)Lm, and Lq=LS+MS-(3/2)Lm.

18

Submatrix (1,2):

P L aR

0 3 = MF 2 0
0 0 1 L Ra P = 0 0

0 3 MD 2 0

0 0 3 MQ 2

0 0 Lm 3 MG 2

Submatrix (2,1):

3 MF 2 3 MD 2 0 0

Submatrix (2,2) (note that this submatrix is unchanged from the original inductance matrix):

0 LT m 3 MQ 2 3 MG 2 0

L RR

Using the defined nomenclature above for the 4 elements, we finally have:

LF M = R 0 0

MR LD 0 0

0 0 LQ MY

0 0 L RR MY LG

0 dq L 0 dq = T FDQG L m

L m i 0 dq i FDQG L RR
19

Expanding L0 0 0 d 0 q = F 0 D 0 Q G 0 0

0 Ld 0 3 MF 2 3 MD 2 0 0

0 0 Lq 0 0 3 MQ 2 3 MG 2

0 3 MF 2 0 LF MR 0 0

0 3 MD 2 0 MR LD 0 0

0 0 3 MQ 2 0 0 LQ MY

0 i0 id 3 M G i 2 q 0 i F iD 0 i Q M Y iG LG 0

(4.20) Compare this to eq. (L-ex) on page 2 to see big improvement in simplicity.
Aside: It is convenient here to note from the above matrix relation that d and q are given by: 3 3 0 dq = L 0 dq i 0 dq + L m i FDQG d = Ld i d + M F iF + M D iD 2 2 3 3 q = Lq i q + M Q iQ + M G iG 2 2 We will use this in developing term 3b below.

One nice surprise from the above is that THE MATRIX IS CONSTANT!!! As a result of this nice surprise, we may differentiate both sides to get:

20

& & 0 dq L 0 dq L m i 0 dq = T & &FDQG FDQG L m L RR i

or, again, when expanded, is:
& 0 & d & q & F = & D & Q & G L0 0 0 0 0 0 0 0 Ld 0 3 MF 2 3 MD 2 0 0 0 0 Lq 0 0 3 MQ 2 3 MG 2 0 3 MF 2 0 LF MR 0 0 0 3 MD 2 0 MR LD 0 0 0 0 3 MQ 2 0 0 LQ MY 0

($)
& 0 i0 i & 3 d M G & iq 2 & 0 i F & i D 0 i & Q & M Y i G LG

Substitution of ($) for term 3a into eq. (tve4), repeated here for convenience,

& & v 0 dq R abc 0 i 0 dq 0 dq P P 1 0 dq n 0 dq + + & = 0 R FDQG i FDQG FDQG v FDQG 0 0 2 44 2444 4 4 123 1 4 3 1 24 4243 1 3 4 3 1
term 1 term 2 term 3a term 3b term 4
eq. (tve4)

results in

& & v 0 dq R abc 0 i 0 dq L 0 dq L m i 0 dq P P 1 0 dq n 0 dq + + & T = 0 v FDQG 103 0 4R FDQG 4i44 L m L RR i FDQG 14243 2 FDQG 4 4 123 1 4 2 4 3 1444 444 2 3
term 1 term 2 term 3a term 3b term 4
eq. (tve5)

We are almost done! The only remaining term which contains flux linkages is term 3b.
21

Term 3b:

& P P 1 0 dq Recalling term 3b is: 0

& 1 we see that we need to expand the product P P . First, recall that:

Also, recall that

1 2 2 P= cos 3 sin

1 1 2 2 cos( 120) cos( + 120) sin( 120) sin( + 120)

& = Re + & (t )

And note carefully that P is a function of time because the angle is a function of t. Therefore we need to differentiate P. This is not hard and results in:

= Re t + (t ) + / 2

& Now taking the product P P 1 , we obtain:

0 2 2 & 1 sin PP = 3 3 cos

0 & = d P = 2 sin P dt 3 cos

sin( 120) sin( + 120) cos( 120) cos( + 120) 0 0

1 cos sin 2 1 cos( 120) sin( 120) 2 1 cos( + 120) sin( + 120) 2

0 0 sin( 120) sin( + 120) cos( 120) cos( + 120)

22

& Note in the above that row 1 is all zeros because row 1 in P is all zeros. On the other hand, column 1 is all zeros because the multiplication of rows
2 and 3 in
1 & P by column 1 of P yield a sum of symmetrical terms.

0 0 0 0 0 0 2 = 0 0 3 / 2 = 0 0 3 0 0 0 0 3 / 2

This provides that:

& 1 P P 0 dq

0 0 0 0 0 = 0 0 d = q 0 0 q d

Some comments on speed voltages -q and d. These speed voltages together account for the voltages induced in the (fixed) phase windings as a result of the spatially-moving magnetic field from the rotor. They represent the fact that a flux wave rotating in synchronism with the rotor will create voltages in the stationary armature coils. Speed voltages are so named to contrast them from what may be called transformer voltages, which are induced as a result of a time varying magnetic field. You may have run across the concept of speed voltages in Physics, where you computed a voltage induced in a coil of wire as it moved through a static magnetic field, in which case, you may have used the equation Blv where B is flux density, l is conductor length, and v is the component of the velocity of the moving conductor (or moving field) that is normal with respect to the field flux direction (or conductor).

23

 The first speed voltage term, -q, appears in the vd equation. The second speed voltage term, d, appears in the vq equation. Thus, we see that the q-axis flux causes a speed voltage in the d-axis winding, and the d-axis flux causes a speed voltage in the q-axis winding. Now we are in a position to obtain term 3b. Using the expressions for d and q obtained in the Aside of page 20 above, we get: 0 0 3 3 Lq iq M Q iQ M G iG 2 2 q 3 3 d & PP 1 0 dq Ld id + 2 M F i F + 2 M D i D speed = 0 = = 0 0 0 (&) 0 0 0 0 0 0 where 0 0 0 L i 3 M i 3 M i speed = q q Q Q G G ; 0 = 0 2 2 3 3 L d i d + M F i F + M D i D 0 2 2 Now recalling eq. (tve5),

& & v 0 dq R abc 0 i 0 dq L 0 dq L m i 0 dq P P 1 0 dq n 0 dq + + & = 0 R FDQG i FDQG L T m L RR i FDQG 0 0 v FDQG 2 44 2444 4 4 123 1 4 3 1444 444 14243 1 3 2 3
term 1 term 2 term 3a term 3b term 4
eq. (tve5)

we substitute (&) to obtain:

24

& v 0 dq R abc 0 i 0 dq L 0 dq L m i 0 dq speed n 0 dq + + & = 0 R FDQG i FDQG L T m L RR i FDQG 0 0 v FDQG 44 2444 2 4 4 123 1 4 3 144 2444 1 2 3 1 3 4 3 4 4
term 1 term 2 term 3a term 3b term 4
eq. (tve6)

Putting it all together: Lets re-write the voltage equation eq. (tve6) by substituting in complete expressions for all vectors and submatrices in terms 1, 2, 3a, 3b, and 4, as obtained above: Term 1 Term 2
0 i0 0 i d 0 iq 0 i F 0 i D 0 iQ rG iG L0 0 0 0 0 0 0 & 3 Ln i 0 0 0 0 0 0 0 0 Ld 0 3 MF 2 3 MD 2 0 0 0 0 Lq 0 0 3 MQ 2 3 MG 2

Term 3a
0 3 MF 2 0 LF MR 0 0 0 3 MD 2 0 MR LD 0 0 0 0 3 MQ 2 0 0 LQ MY & 0 i0 i & 3 d M G & iq 2 & 0 i F & i D 0 i & Q & M Y i G LG 0

v0 ra v d 0 vq 0 v F = 0 0 0 0 0 0 0

0 0 0 0 0 rb 0 0 0 0 0 0 0 0 0 0 0 rc

0 0 rF

0 0 0 rD

0 0 0 0 rQ 0 0 0 0 0

Lq iq L i + d d +

0 3 M Q iQ 2 3 M F iF + 2 0 0 0 0

3rn i0 3 M G iG 2 3 M D iD + 2

Term 3b

Term 4

25

Now, observe that each of the non-zero elements of term 3b and term 4 is multiplied by a current or current derivative, and that terms 2 and 3a both get multiplied by vectors of currents or current derivatives, respectively. Therefore, we may fold-in Term 3b and Term 4 into the Terms 2 and 3a by combining parts of the non-zero term 3b and 4 elements with the appropriate matrix element in terms 2 and 3a. For example, we may fold in the -Lqiq term in row 2 of term 3b by including Lq in row 2 (since we are dealing with the second equation), column 3 (since we need the term that multiplies iq) of term 2. Note that since term 2 has a minus sign out front, we do not include the minus sign of -Lqiq when we fold it in. The circle and arrow above illustrate this folding-in operation. The complete results of all fold-in operations are provided in what follows:
0 0 ra + 3rn v0 v 0 Lq rb d vq rc LD 0 vF = 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 L0 + 3 Ln 0 Ld 0 0 3 0 MF 2 3 0 MD 2 0 0 0 0 0 0 3 MF 2 rF 0 0 0 0 0 Lq 0 0 3 MQ 2 3 MG 2 0 3 MF 2 0 LF MR 0 0 0 0 3 MD 2 0 rD 0 0 0 3 MD 2 0 MR LD 0 0 0 3 MQ 2 0 i0 3 M G i 2 d iq 0 0 iF 0 0 iD 0 0 iQ 0 rQ i G 0 rG 0 0 0 0 & i0 i &d 3 3 MQ M G i&q 2 2 & iF 0 0 & iD 0 0 i& Q i& LQ M Y G MY LG

26

It is of interest to rearrange the ordering of the variables so that the voltage equations for all d-axis windings are together and the voltage equations for all q-axis windings are together because this will emphasize the presence or absence of the various couplings that we have. The result of this reordering of the variables is as follows:
0 r + 3rn v0 0 r vd v 0 0 F 0 v D = 0 = 0 v q L D 0 v Q = 0 0 0 v G = 0 0 0 0 L0 + 3L n Ld 0 3 MF 0 2 3 MD 0 2 0 0 0 0 0 0 0 0 rF 0 3 MF 2 0 0 0 3 MF 2 LF MR 0 0 0 0 0 0 rD 3 MD 2 0 0 0 3 MD 2 MR LD 0 0 0 0

L q
0 0 r 0 0 0 0 0 0 LQ

0 3 MQ 2 0 0

3 MQ 2 3 MG 2

0 i0 3 M G id 2 0 i F i 0 D iq 0 0 i Q rQ 0 i G rG 0 0 0 0 0 & i0 & id 0 0 i F & & i D 0 0 & iq 3 3 MQ M G & i 2 2 Q & LQ M Y i G MY LG

(eq. 4.39)

27

Some observations about the transformed voltage equations: 1. The first matrix gives a. Resistive voltage drops b. Speed voltage drops, svd (terms with ). These svds Occur in the d- and q- circuits, to represent the fact that a flux wave rotating in synchronism with the rotor will create voltages in the stationary armature coils Do not occur in circuits physically located on the rotor, since there is no motion between the rotating flux wave and the rotor windings. Are caused by currents in the field windings of the other axis: the d-circuit svd is caused by iq, iQ, and iG the q-circuit svd is caused by id, iD, and iF 2. The matrices are almost constant, except for the svd terms in the first matrix, but even these terms are practically constant since we only see small changes in . The constancy of the matrices is the main motivation behind the Parks transformation. 3. The variables have been reorganized so that all d-axis circuits are together and all q-axis circuits are together. This makes it easy to observe any coupling/decoupling between different sets of circuits. 4. The second matrix gives voltage induced by current (or flux) variation. Note that there is no coupling between the d-axis circuits (d, F, D) and the q-axis circuits (q, Q, G). This is because these two sets of circuits are orthogonal. Finally, some comments about the Parks transformation: 1. id and iq are currents in a fictitious pair of windings fixed on the rotor. 2. These currents produce the same flux as do the a,b,c currents. 3. For balanced steady-state operating conditions, we can use i0dq = Piabc to show that the currents in the d and q windings are dc! The implication of this is that:

28

 The a,b,c currents fixed in space, varying in time produce the same synchronously rotating magnetic field as The d,q currents, varying in space, fixed in time! From Kimbark, Vol. III:

29

Another interesting paragraph from Kimbark Vol. III

30