Machine Transformations

J. McCalley
 Space vectors

We have already considered this concept, but

without calling it a “space vector,” when we
discussed the rotating magnetic field in the slide
deck called “Wind GeneratorsSteady-StateBasics”
(see slides 8-12). We repeat slide 12 from that
slide deck in the next slide.

2
 Rotating magnetic field
• Now let’s consider the magnetic field from all three
windings simultaneously.
 cos  0 t  I a  cos 
Ba ( , t )  Bmax (1)
 2   2 
 cos  0 t  I a 
Bb ( , t )  Bmax  
cos    (2)
 3   3 
 2   2  (3)
 cos  0 t  I a 
Bc ( , t )  Bmax  cos   
 3   3 
• Add them up, then perform trig manipulation to obtain:

3 Bmax
Babc ( , t )  cos 0 t  I a    (4)
2
Notice that each of (1), (2), and (3) has a spatial maximum which is fixed in space, as
dictated by the second term of each the 3 expressions. But the spatial maximum of
(4) rotates. This is a characteristic of a rotating magnetic field. One can observe this
using the following: http://educypedia.karadimov.info/library/rotating_field.swf
www.youtube.com/watch?v=eQk0OznWTjM
The shape of the individual winding fields throughout the airgap are spatially fixed, but
their amplitudes pulsate up and down. In contrast, the amplitude of the composite is
fixed in time, but it rotates in space. What you will see in the video are just the
variation of the maximum field point.
 Space vectors
Consider the current ia=Iacosωt. Let’s represent it as a vector having
•a magnitude equal to the peak amplitude of the variable and
•an angle equal to the angular position of the peak MMF produced by this current
around the circumference of a machine.
The vector is then given by:

i a  ia (t )0
We may think of it as oscillating up and down in a direction as shown below.

4
 Space vectors
We may do the same for phase currents b and c, in which case we represent them as

ib ic
j ( 2 / 3 ) 1 3
Identify a positive rotation of 2π/3 as: a  e   j
2 2
j ( 4 / 3 ) 1 3
Then a positive rotation of 4π/3 is: a 2
 e   j
2 2
Because the b and c phase windings are spatially displaced from the a phase winding
by 120 and 240 degrees, respectively, then:

i b  aib
i c  a 2 ic
The total effect of a set of instantaneous currents in all three phase windings is then
found by vector addition according to:

i  i a  ib  ic
This is not the same as the sum of the currents: ia+ib+ic, which is zero! 5
 Space vectors
Time-domain waveforms Current vectors and their summation for
the “instant” indicated on the other plot
+b axis

+a axis

+c axis

Observe: at the “instant,” ia is positive and The current vector for each phase
just after its peak, ib and ic are negative. oscillates up and down its axis, but the
composite (space vector) i rotates.

i  i a  i b  i c  ia  aib  a 2ic (Eqt 1)

6
 Space vectors
Let’s add them mathematically:

i  i a  i b  i c  ia  aib  a 2ic (Eqt 1)

Now use:
1 3 1 3
a  e j ( 2 / 3 )    j a 2  e j ( 4 / 3 )    j
2 2 2 2
And we obtain:
 1 3  1 3
i  ia     j 
ib     j 
ic
 2 2   2 2 
1 3
 ia   ib  ic   j  ib  ic 
2 2

Let’s assume that the three quantities ia, ib and ic are balanced 3-phase quantities.
Then ia+ib+ic=0ia=-(ib+ic). Making this substitution in the above results in:

1 3 3 3
i  ia  ia  j  ib  ic   ia  j  ib  ic 
2 2 2 2
7
 Space vectors
3 3
i ia  j  ib  ic 
2 2
It will be convenient for us later on to scale this vector by 2/3 resulting in :

2 2 3 3 
is  i   ia  j  ib  ic    ia  j 3  ib  ic 
3 3 2 2  3
1
 ia  j  ib  ic 
3
The above relation concentrates the effects of the three phase currents into a
single complex variable. Very nice!
2
Note, because i s  i and i  ia  aib  a ic that
2
3

is 
2
3

ia  aib  a 2ic  (Eqt 2)
So the above is how to obtain a space vector given values of the balanced currents.

Now, let’s consider the reverse process….

8
 Space vectors
The reverse process….

Consider that we know the space vector i s and that we want to find the
instantaneous values of the individual phase currents. How to do this?
We project the vector i s onto the respective a, b, and c axes, as shown below.
+b-axis
The analytic equivalent of this projection is:
ic
ia  Rei s 
Take the real part of the is vector.
ai s
is i
Rotate the is vector forward by 240°.
The projection of this rotated vector
on the a-axis (via taking real part) is 
ib  Re a 2 i s  ib ia
+a-axis
the same as the projection of the
original vector on the b-axis.
Rotate the is vector forward by
120°. The projection of this rotated
vector on the a-axis (via taking real
ic  Reai s  +c-axis
a2 i s
part) is the same as the projection is may be decomposed to other
of the original vector on the c-axis. sets of 3 vectors (e.g., 3 vectors
Recall (Eqt 1): i  i a  i b  i c  ia  aib  a 2ic collinear with is having 1/3 the
Recall (Eqt 2): is 
2
3
ia  aib  a 2ic  magnitude), but if we impose that
the 3 vectors be on the a, b, and c
 Space vectors
We have represented the phase currents as space vectors. In doing so, however,
the only thing required was they were balanced three-phase quantities (see slide
7). Any other variables can be similarly represented as long as they are balanced
three-phase quantities, e.g., currents, voltages, and fluxes.

Let’s generically refer to any such variables as xa, xb, and xc and the corresponding
space vector as x s

10
 α-β transformation
The space vector x s can also be represented by two phase magnitudes, called xα
and xβ in the real-imaginary complex plane, as illustrated below.

Whereas vectors corresponding to

xa, xb, and xc oscillate up and down
xc s x s
the a, b, and c axes, respectively, the
vectors corresponding to xα and xβ
xb oscillate up and down the α and β
xa x
axes, respectively.

We can express this relationship mathematically according to: x s  x  jx 

We also know (from Eqt 2, slide 8) that : x s 
2
3
xa  axb  a 2 xc 
The α-β components of the space vector can be calculated from the abc
magnitudes according to:
 
x  Re x s   Re xa  axb  a 2 xc
2
3
 
x   Im x s   Im xa  axb  a 2 xc
2
3
2  1 3 1 3 
2  1 3 1 3   Im xa  (   j ) xb  (   j ) xc 
 Re xa  (   j ) xb  (   j ) xc  3  2 2 2 2 
3  2 2 2 2 
2 1 1  2 3 3 
  x a  xb  xc    xb  xc 
3 2 2  3 2 2 
 α-β transformation
  2 1 1 
x  Re x s   xa  xb  xc 
3 2 2 
  2 3
x  Im x s   xb 
3 
xc 
3 2 2 
These two relations can be represented in matrix form as follows:
 1 1 x 
 x  2  1    a
2 2 x 
x    3

3  b 
   3 0  x 
 2 2   c 
We define the matrix as the T-matrix:

 1 1 
1  
2 2 2 
T  
3 3 3
0 
 2 2 
This transformation is also called the
“Clarke transformation” for the person
who developed it, Edith Clarke.
α-β transformation E. Clarke, Circuit Analysis of AC
Power Systems. New York:

 
Wiley,
2 1 1 
x  Re x s   xa  xb  xc  1943, vols. I, II.
3 2 2 

2 3 3 
 
x  Im x s  
3 2
xb  xc 
2 

February 10, 1883 - October 29, 1959

AT&T, MIT, GE, U-Texas

https://babel.hathitrust.org/cgi/pt?
id=uc1.b4164139;view=1up;seq=332;si
 α-β transformation
  2 1 1 
x  Re x s   xa  xb  xc 
3 2 2 
  2 3
x  Im x s   xb 
3 
xc 
3 2 2 

We will often represent the Clarke transformation in one of the two equivalent ways:

Note that
2 3 3 
x  Im x s    xb  xc 
3 2 2 
3 3 1 1
 xb  xc  xb  xc
3 3 3 3
 α-β transformation
  2 1 1 
x  Re x s   xa  xb  xc 
3 2 2 
  2 3
x  Im x s   xb 
3 
xc 
3 2 2 

It is affirming to observe, in a balanced system (which implies that xa+xb+xc=0),

  2
3
1
x  Re x s   xa   xb  xc  
2


2 1  23 
  x a   x a     x a   xa
3 2  32 
Using xα=xa, we can solve for xb and xc in terms of xα and xβ, in which case we obtain
the inverse Clarke transformation as
 
 0 
 xa   1
 x    1 3   x 
 
 b  2 2   x 
 xc   1 3
  
 2 2 
 α-β transformation
 
 0 
 xa   1
 x    1 3   x 
 
 b  2 2   x 
 xc   1 3
  
 2 2 
We often represent the inverse Clark transformation in two equivalent ways:
 Some observations on xα and xβ
2 1 1 
  x  x 
Let’s expand  3 a 2 b 2 xc  using xa  X cost;
x xb  X cos(t  120); xc  X cos(t  240)
 
2 X X  2X  1 
x   X cost  cos(t  120)  cos(t  240)    cost   cos(t  120)  cos(t  240) 
3 2 2  3  2 

Let’s apply the trig identity: cos(u  v )  cos u cos v sin u sin v

2X 1 
x   cos t   cos(t  120)  cos(t  240 ) 

3 2 
2X  1 
  cos t   cos t cos 120  sin t sin 120  cos t cos 240  sin t sin 240  
3  2 
Recalling cos120=-1/2, sin120=√3/2, cos240=-1/2, sin240=-√3/2, the above becomes
2X 1 1 3 1 3 
x   cos t    cos t  sin t   cos t  sin t  
3  2 2 2 2 2 
2X  1  2X 3 
  cos t    cos t     cos t   X cos t
3  2  3 2 

From this we learn that:

•xα=xa (we knew that from slide 14)
•xα is a sinusoidally varying quantity. This is because of
our choice of 2/3 as
•The amplitude of xα is equal to the amplitudes of xa, xb, and xc. the coefficient in
Similar work on xβ leads to the following conclusions: the transformation.

•xβ is a sinusoidally varying quantity.

•The amplitude of xβ is equal to the amplitudes of xa, xb, and xc.
•xα and xβ are 90 degrees phase-shifted (we also knew this already)
 Other transformations
When we transform variables to the α-β transformation we have just established, we
are said to be working in the Stator Reference Frame.

This reference frame is aligned with the stator, and the rotational speed of this
reference frame, since it is aligned with the stator, is 0. It is a fixed reference frame.

The space vector referred to it rotates at the synchronous speed ωs. We will denote
the corresponding space vector with a superscript “s” (stator) according to:
s
x  x  jx
We can also define a space vector aligned with the rotor. The reference frame in this
case is called the D-Q reference frame and it rotates with angular speed of ωm.
Therefore the space vector referred to it rotates at the slip speed of ωr.
r
x  xD  jxQ
Finally, we can also define a space vector aligned with the synchronous reference
frame, at a speed of ωs. The space vector referred to it does not rotate, that is, it
presents constant real and imaginary parts. This is called the d-q reference frame.
a
x  xd  jxq
 Other transformations
The below figure shows the three different reference frames discussed on the
previous slide: stator (α-β); rotor (D-Q); and synchronous (d-q).

)
θ
0-
m
9
o s(
c
x β

s θm
co
x α

We illustrate just one of the transformations, to the D-Q coordinates:

xD=xαcosθm+xβcos(90-θm)=xαcosθm+xβsinθm
We can make a similar projection for xQ to find… 19
 Other transformations
We can make a similar projection for xQ to find…

 xD  cos m sin  m   x 
 x    sin  cos m   x  
 Q  m

 m    m dt   mt

20
 Other transformations
You can find a book on electric machine theory at http://s1.downloadmienphi.net/file/downloadfile5/192/1388763.pdf. This
book contains ch 3, titled “Reference frame theory.” Below are some cut-outs from that chapter.

P. Krause, O. Wasynczuk, and S. Sudhoff, “Analysis of electric 21

machinery and drive systems,” 2nd edition, IEEE Press, 2002.
 Reference frame transformations
Our objective now is to represent balanced, but time-varying (dynamic) behavior of a
DFIG. There will be three types of equations that we will need:
•Voltage equations
•Flux linkage equations
•Motion equations
We will do this by transforming to the d-q coordinate reference frame
We desire to achieve this objective because the control becomes much easier.

One can see, for example, on the more

detailed block diagram on the right, when
in the d-q frame, we may specify Te by idr
22
and Qs by iqr.
 Reference frame transformations
We will come back, later, to the Clarke (α-β) transformation, because we will find it
very useful in designing the switching sequence for the RSC and the GSC.

23