1100 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 31, NO.

5, SEPTEMBEWOCTOBER 1995

Space Vector PWM Control of Dual Three-phase

Induction Machine Using Vector
Space Decomposition
Yifan Zhao, Student Member, IEEE, and Thomas A. Lipo, Fellow, IEEE

Abstract-The technique of vector space decomposition con- the current per phase in the machine is thereby reduced. In the
trol of voltage source inverter fed dual three-phase induction most common such structure, two sets of three-phase windings
machines is presented in this paper. By vector space decompo- are spatially phase shifted by 30 electrical degrees. In such
sition, the analytical modeling and control of the machine are
accomplishedin three two-dimensional orthogonal subspaces and systems, each set of the three-phase stator windings is excited
the dynamics of the electromechanical energy conversion related by a three-phase inverter, therefore the total power rating of the
and the nonelectromechanical energy conversion related machine system is doubled. In addition to enhancing power rating, it is
variables are thereby totally decoupled. A space vector PWM also believed that drive systems with such multi-phase redun-
technique is also developed based on the vector space decom- dant structure will improve the reliability at the system level
position to limit the 5th, 7th, 17th, 19th, . . . harmonic currents
which in such a system would be otherwise difRcult to control. [2], [3]. In particular, unlike in a normal three phase system,
The techniques developed in thii paper can be generalized for the loss of one phase in a multi-phase machine drive system
the control of an induction machine with an arbitrary number does not prevent the machine from starting and running.
of phases. However, when a multi-phase system is implemented with
a voltage source inverter with conventional six-step operation
I. INTRODUCTION or space vector PWM control, surprisingly large harmonic
currents have been observed [4]-[6]. Although this peculiarity
H IGH-POWER electric machine drive systems have found
many applications such as pumps, fans, compressors,
rolling mills, cement mills, mine hoists, just to name a few.
has drawn attention from people working on this kind of
system, little work has been done in developing a dedicated
At present, the most successful type of the high power drive control strategy to effectively suppress the harmonics due to
systems are cycloconverter-electric machine drives and syn- the complexity of the converter-machine model.
chronous machines fed by current source thyristor inverters. It is well known that the d-q-0 reference frame transfor-
Voltage source inverters, despite their advantage in line power mation has long been used successfully in the analysis and
factor over a cycloconverter as well as their advantage of being control of three-phase electric machines. In this technology,
the original three-dimensional vector space is decomposed
able to use low cost induction machines, are still limited to the
lower end of the high power range due to the limitations on into the direct sum of a d-q subspace and a zero sequence
the gate-turn-off type semiconductor power device ratings. subspace, which is orthogonal to the d-q. By virtue of this
In the past decade, multi-level inverter fed electric machine decomposition, the components which produce rotating MMF
drive systems have emerged as a promising approach in and the components of zero sequence are totally decoupled,
and thus the analysis and the control of the machine is
achieving high power ratings with voltage limited devices.
The typical structure of such systems is the three-level inverter simplified.
three-phase electric machine system [ 11. A three-level voltage For the analysis of multi-phase machines, a variety of
source inverter is a series switch type structure which operates transformations has been proposed in the past [S-[ 101. Sym-
metrical component theory and matrix theory have served as
with split-voltage dc bus. The voltage stress on each device is
only half of the total dc bus voltage and, thus, a doubled dc bus the theoretical foundations for these transformations. In this
paper, matrix theory and the concept of decomposition are
voltage can be achieved. The parallel circuit dual to the multi-
level system is, essentially, the concept of the multi-phase adopted in developing a transformation for dual three-phase
induction machines. A companion vector space decomposition
inverter fed electric machine drive system. In a multi-phase
machine drive system, more than three phase windings are control technique is also proposed. By vector space decompo-
implemented in the same stator of the electric machine, and sition, the machine current and voltage vectors in the original
six-dimensional vector space are mapped into three two-
Paper IPCSD 95-25, approved by the Industrial Drives Committee of the dimensional orthogonal subspaces. As a result, the machine
IEEE Industry Applications Society for presentation at the 1994 IEEE Industry
Applications Society Annual Meeting, Denver, CO, October 2-7. This work model can be easily described by three sets of decoupled
received support from the Electric Power Research Institute. Manuscript equations. The techniques presented in this paper can also be
released for publication March 10, 1995. used to analyze the current source inverter fed dual three-phase
The authors are with the Department of Electrical and Computer Engineer-
ing, University of Wisconsin, Madison, WI 53706 USA. machines and generalized to control multi-phase induction
IEEE Log Number 9412784. machine drive systems.
OO93-9994/95$04 .OO 0 1995 IEEE

Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 04,2020 at 14:52:11 UTC from IEEE Xplore. Restrictions apply.
ZHAO AND LIFO SPACE VECTOR PWM CONTROL OF DUAL THREE-PHASE INDUCTION MACHINE USING VECTOR SPACE DECOMPOSITION 1101

Xe
, here the standard basis vector set
Var mlk
- . . - - -
Fig. 1. Voltage source inverter fed dual three-phase induction machine drive
system.
1 0
0 0 k = 1 in (l), and w t = 0 and 7r/2 to make the resultant
0 vectors orthogonal. Two vectors, named d and q, are obtained
el = e6 = is implied.
0 '"" 0 as shown below
0 0
-0 - -1- d: [I,cOs(e), COS(^^), COS(^^), COS(^^), ~ ~ ~ ( g e ) ] ~ ,
(wt= 0)
As the variables are functions of time, the vector rotates
about the origin and spans a surface in the six-dimensional q: [0,sirl(8),sin(48), sin(58), sin(88), sin(90>]',
space. Geometrically, the proper control of the dual three- (wt= T / 2 ) .
phase machine is equivalent to positioning the vector on a
As pointed out before, the rest of the four vectors should
certain surface in the six-dimensional vector space and rotating
be selected such that they are orthogonal each other and also
this vector at a desired speed. This requirement makes the
orthogonal to vectors d and q . To find these four vectors the
modeling and control of the dual three-phase machine very
relations between the surfaces s k , which are spanned by a
difficult. However, the modeling and control of the dual three-
different order of harmonics when kwt changes from 0 to 27r,
phase induction machine can be greatly simplified with a
are examined and it is found that the surfaces corresponding
proper transformation which maps the description of a vector
to k = 1 , 3 , 5 are orthogonal each other, i.e.
with respect to the original six-dimensional frame spanned by
the six standard base vectors to a new reference frame. For ST os, -= ST o s , = S3'0S5 = 0 (0 5 wt 5 27r).
doing so, a set of six vectors should be chosen to form the
new base of the six-dimensional space, and, for purpose of Surfaces spanned by harmonics of higher order (k = 7,9,
decomposition, it is desirable that the six vectors be orthogonal etc.) can be shown to generate the same surfaces as the k =
1 , 3 , 5 surfaces. Therefore, with two orthogonal vectors being
each other.
For the dual three-phase induction machine in Fig. 1, suit- selected from each of the two surfaces as indicated below, the
able voltage and current vectors can be evolved by defining orthogonality of the new base vector set is guaranteed.
the following vector: With k = 5 we have
zl:[I,cOs(5e), COS(^^), cos(e), COS(^^), COS(^^)]*,
(wt = 0 )
z 2 : [0,sin(58), sin(8B), sin(@),sin(48), sin(9e)lT,
(wt = 7r/2).
where 8 = ~ / 6 and, k = 0 , 1 , 3 , 5 , .. ., which denotes the
order of the time harmonics. As the elements of the vectors With k = 3 we have
are functions of time, each vector spans a two-dimensional 01: [1,0,1,0,1,O]T, ( w t = O )
surface, which can be proven to be a subspace in the original
02: [0, 1,0, l , O , lIT, (ut = ~ / 6 ) .
six-dimensional space.
The surface spanned by the fundamental component vector With these six vectors used as the new basis of the six-
is of key importance to the electromechanical energy con- dimensional space, the following normalized coordinate trans-
version function of the machine. Therefore, two out of the formation matrix results (see (2), shown at the bottom of the
six new base vectors must be chosen from this surface. Let following page).

Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 04,2020 at 14:52:11 UTC from IEEE Xplore. Restrictions apply.
I102 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 31, NO. 5, SEF'TEMBEWWTOBER 1995

The transformation possesses the following properties:

1) The fundamental component of the machine variables
-
and the kth order harmonics with k = 12m f 1, (m =
1,2,3,. . .) are transformed into the d-q subspace or 'd-
"U, -
"bs
q plane. "c,
It should be pointed out that the d-q axes have been ["SI = 1 [is] =
"ds ids
chosen in such a manner that they coincide with the %,9
plane of rotation of the air gap flux. Therefore, these
-"f s -
variables will produce a rotating MMF in the machine -0 -
airgap and thus be electromechanical energy conversion 0
related. 0
2 ) Harmonics with k = 6m f 1,(m = 1,3,5,. . .), i.e., the [%I = 0 '
5th, 7th, 17th, 19th, . . . harmonics, are mapped into the 0
z1-z2subspace, or "21-z2plane." As the z1-z2 subspace -0 -
is orthogonal to the d-q subspace, it is expected that the
variables on this plane will not generate any rotating The resistance and inductance matrices in (3) and (4)
MMF in the airgap and thus be nonelectromechanical are defined as follows according to the machine structure:
energy conversion related. In this regard, these harmon- [R,],[Rr]are the stator and rotor resistance matrices
ics can be classified as a new type of zero sequence
component.
3) Zero sequence (m - 3) harmonics, which are also
nonelectromechanical energy conversion related, are [ I I , ] = / O0 0 T0s T , 0 0a / l
mapped into the O1 -02 subspace to form the conventional
0 0 0 0 T,
zero sequence components.
O O O O O r ,
0 0 0 0 0
111. MACHINEMODEL
The following assumptions have been made in deriving the
[
T,
O T r O 0 0 0 1
dual three-phase induction machine model:
1) Machine windings are sinusoidally distributed.
2 ) Flux path is linear.
3) Mutual leakage inductances are neglected.
4) Assume unity stator and rotor turn ratio. [L,,] is the stator self inductance matrix
Under these assumptions the voltage equations of the ma-
chine in the original six-dimensional space and the equations ' 1 0 0 0 0 0 -
transformed to the d-q-z1-z2-O1-02 reference frame can be 0 1 0 0 0 0
derived. 0 0 1 0 0 0
0 0 0 1 0 0
A. Machine Model in the Original Sir-Dimensional Space 0 0 0 0 1 0
.0 0 0 0 0 1,
The stator voltage equation is
- 1 a2 _ _21 -a 0
[Vsl = [&I * + P . ([A,])
[is]
1 0
2
__ _ _21
= [%I . + P . ([A,] + [Asr]>
[is]
0 1 _2a
[ i s ] + P ([Lss] [ i s ]+ [L,,] . [ i r ] ) . + Lm, 2
= [&31 '

The rotor voltage equation is

' ' (3) __
-&
2
_ -2
1
2
&
2
_ _1
2
1
0
_ -1
0

a2
1 A
_-
2

[Vr]= [K][GI + P ([A,])

1 -Y3
- 0 2 2 1
= [&I . [GI + P . +'[A,,])
([AT,] where Ll, and L,, are the stator leakage and magnetizing
= [RT]. \GI + P . ( [Lrr.]. [GI+ [L,,] . [ i s ] ) (4) inductance.

cos(e) cos(48) cos(58)

-1 cos(90)
0 sin(8) sin(48) sin(58) sin(88) sin(98)
1 1 cos(56) cos(88) cos(@) cos(il8) cos(98)
[TI = 0 sin(58) sin(88) sin(8) sin(48) sin(98)
1 0 1 0 1 0
-0 1 0 1 0 1

Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 04,2020 at 14:52:11 UTC from IEEE Xplore. Restrictions apply.
ZHAO AND LIPO: SPACE VECTOR PWM CONTROL OF DUAL THREE-PHASEINDUCTION MACHINE USING VECTOR SPACE DECOMPOSITION 1103

Shown at the bottom of the page are

[L,,] rotor self inductance matrix where L1, is the rotor
leakage inductance,
[L,,] stator-rotor mutual inductance matrix where 0, is the
rotor angular position,
[L,,] rotor-stator mutual inductance matrix.
The subscript ds is now, in this case, intended to denote
B. Transformation of the Voltage Equations the stator d-axis and not phase ds as in Figs. 1 and 2. The
to the New Reference Frame dynamic models of the machine in the three subspaces can be
derived from (6) and are illustrated as follows.
Applying the transformation (2) to the voltage equations
Machine model in d-q subspace:
( 3 ) and (4) yields
The stator voltage equation is

The rotor voltage equation is

where in (6), the voltage and current vectors transformed to

the new d-q-zl-22-01-02reference frame are defined as

dt +
L1r O3Lms I

3cos(l,) 3sin(B,)
-3sin(O,) cos(8,)

cos(+
cos(e,)
+e,)
cos(g+e,)
cos(e,) COS(% 1
+e,
Cos(e,)
COS(^+^,)
COS(% +e,)
+ e,)
COS
COS i
+ e,)
COS(%
%+e,)
76, + e,)
COS
COS
COS(%
i% + e,)i
%+e,
+e,
[Lsr] = Lm, COS(+ +e,) COS(B,) + e,)
COS(% +e,) COS(%
Cos & + e , c o s ( ~ + e , ) ~ ~ ~ ( 8 , )cos(;+e,)
cos[+ + e,i +e,) Cos(+ + e,) cos(e,)

Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 04,2020 at 14:52:11 UTC from IEEE Xplore. Restrictions apply.
1104 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 31, NO. 5, SEPTEMBEWOCTOBER 1995

Machine model in 21-22 subspace: transforms the rotor variables to the stationary reference frame,
The stator equation is is appropriate

d
dt
[i, ,]
izzs
= [-rs/LI" 0
-r./L1.] ' [e:]
+ [YS l L ? ]. *I:[ Since both btator and rotor variables are expressed in the
same reference frame now, it is necessary to distinguish them
with a newly defined variable expression
The rotor equation is
stator or rotor frame
variahled-or q-axis stator or rotor variable.
With the rotation transformation applied to (7) and (8),
Machine model in 01-02 subspace: the following stator and rotor combined d-q plane stationary
The stator equation is reference frame equation results:

The rotor equation is

It is observed that the stator equations in the z 1 - z ~and the

01-02 subspaces have the same form and the same parameters. where
This is due to the assumption that the stator mutual leakage
d
inductances can be ignored. With mutual leakage inductance p=-
considered, the leakage inductances in (9) and (10) will have dt
different values. The effect of the mutual leakage inductances +
Ls = L 1 S 3Lrns,
will be addressed in the future research work on this machine. Lr = Llr + 3Lm3,
It is also observed that there are no excitation terms in (9a) M = 3Lms.
and (loa). Therefore, this portion of the machine dynamics
can never be excited and, hence, will not be discussed further The electromagnetic torque of electric machine is expressed
in this paper. as
It is observed immediately from the above equations that all T - -1[ Z.I T . ( - [aL ] ) . [i].
the electromechanicalenergy conversion related variable com- "-2 80,
ponents are mapped into the d-q subspace, or "d-q-plane," and
the nonelectromechanical energy conversion related variable As the d-q plane is electromechanical energy conversion
components are transformed to the 21-22 and 01-02 planes. related, the current vector in (13) should include d-q plane
Hence, the dynamic equations of the machine are totally currents only, i.e.
decoupled. As a result, the analysis and control of the machine
can be greatly simplified.
It is also important to note that the nonelectromechanical
energy conversion related variables on the 21-22 and the 01-02
planes should be controlled to be as small as possible to reduce
the extra losses in the machine. However, as can be seen from For the same reason, the d-q plane inductance matrix as
(9) and (lo), only stator resistance and leakage inductance are illustrated in (15) should be used for the matrix [ L ]in (13).
associated with these variables. As a result, high switching

1
Lds,ds Lds,qs Lds,dr Lds,qr
frequency inverters are generally required in a voltage source
inverter fed multi-phase induction machine drive. This, clearly,
is a drawback of such a system.
In the analysis just completed, d-q reference frames were, in [Lqs,ds
[ L 1 = Ldr,ds
Lqr,ds
Lqs,qs
Ldr,qs
Lqr,qs
Lqs,dr
Ldr,dr
Lqr,dr
Lqs,qr
Ldr,qr
LqT9qT

effect, separately attached to the stator and the rotor, rotating

at zero and rotor angular speed, respectively. To express the
stator and rotor equations in the same stationary reference
frame and thus eliminate the sin and the cos terms in the
above equations, the following rotation transformation, which
L S

0
M cos(0,)
-M sin(8,)
0
L,
M sin(0,)
M cos(8,)
M cos(0,)
M sin(0,)
Lr
0
- M sin(&.)
M cos(8,)
0
Lr
(15)
I
Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 04,2020 at 14:52:11 UTC from IEEE Xplore. Restrictions apply.
ZHAO AND LIPO: SPACE VECTOR PWM CONTROL OF DUAL THREE-PHASE INDUCTION MACHINE USING VECTOR SPACE DECOMPOSITION 1105

The electromagnetic torque of the dual three phase induction

machine is obtained by substituting (14) and (15) into (13) as

T, = 3Lms[2~s(r2,
cos(8,) - r;, sin(0,))
- i&(& sin(&) + vir C O S ( ~ , ) ) ]
= 3Lms(Z&2$,- i &.IZ q .ps) .

When the number of poles is greater than two

(16)
vgs
1)+ -1 fw-)+ lk
-
vb
- - -
(a)
- -

P
T, = 3 - L m , , a ( Z ~ s i-~i&i&.).
7. (17)
2
P is the machine pole number.
I
IV. SPACEVECTORPWM CONTROLSTRATEGY
In the dual three-phase machine drive system shown in
Fig. 1, the stator windings can be connected either to a single
neutral or to double neutrals. Neutral wires eliminated in both
cases will clearly reduce the dimension of the system.
In general, the six-phase voltage source inverter has a total
of 64 switching modes. By using the transformation matrix (2),
the 64 voltage vectors corresponding to the switching modes ‘uabs
- - 1 1 - 1 0 0 0 0
are projected on the three planes. A space vector PWM control vbcs - 1 0 1 - 1 0 0
strategy which covers the control of the volt-seconds applied vcds 1 0 0 1 - 1 0 0
to the machine in all the three planes will now be developed. = -1 0 0 0 1
vdes
vefs 1 0 0 0 0
A. Transformation of the Inverter Output Voltage Vectors n
to the d-q-21-.22-01-02 Reference Frame
With the machine stator windings connected to the inverter
without a neutral wire, only the line voltages of the machine
corresponding to the inverter switching modes are directly
known. On the other hand, the transformation matrix ( 2 ) 710
transforms the machine phase variables to the new d-q-iz1-22- vas
01-02 reference frame. Therefore, it is necessary to develop vbs
transformations from line voltages to phase voltages.
Fig. 2 shows two cases for the stator winding neutral point v ds
connection. In Fig. 2(a), the stator windings are connected to ves
a single neutral, the relation between the line and the phase -Vf 5
voltages are defined by a full rank matrix as illustrated in (18).
ext step is to combine the transformation ( 2 ) with the
‘1 -1 0 0 0 0 transformation (19) or (21) to transform the inverter voltage
0 1 - 1 0 0 0 vectors corre\ponding to the 64 switching modes to the d-q-
0 0 1 - 1 0 0 21-22-01-02 reference frame.
0 0 0 1 - 1 0 When the stator windings are connected to single neutral,
0 0 0 0 1 - 1 the inverter voltage vectors are transformed to all three planes
. 1 1 1 1 1 1 as shown in Fig. 3(a)-(c). Note that all the voltage vectors on
The transformation from line voltage to phase voltage can 01-02 plane now lie on a straight line. Therefore, the system

be derived from (18) by inverting the matrix. The resultant is, in actuality, five-dimensional.
transformation is shown in (19). The decimal numbers in the figures denote the switching
modes. When each decimal number is converted to a six-digit
-5 4 3 2 1 1 -
binary number, the 1’s in this number indicate that the upper
- 1 4 3 2 1 1
switches in the corresponding switching arms are “on,” while
-1- 2 3 2 1 1
the 0’s indicate the “on” state of the lower switches. The MSB
-1 -2 -3 2 1 1
(most significant bit) of the number represents the switching
-1 -2 -3 -4 1 1
state of phase a, the second MSB for phase b, and so on.
.-1 -2 -3 -4 -5 1- In the case when the stator windings are connected to double
Similarly, the transformation for the case of double neutral neutrals as in Fig. 2(b), the projections of the inverter voltage
connection (see Fig. 2(b)) can be derived and illustrated in vectors on the d-q, 21-22 planes remain the same as in the
(20) and (21). single neutral case. However, on the 01-02 plane all the vectors

Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 04,2020 at 14:52:11 UTC from IEEE Xplore. Restrictions apply.
1106 lEEE TRANSAC'MONSON INDUSTRY APPLICATIONS, VOL. 31, NO. 5, SEFTEMHER/OCTOBER 1995

q-axis B. Space Vector PWM Control

The machine model which has been developed has demon-
strated thqt only the voltages and currents in the d-q plane
are related to electromechanical energy conversion. Therefore,
the goal of the space vector PWM control is to synthesize
the d-q voltage vectors to satisfy the machine's torque control
requirements, and, at the same time, to maintain the average
volt-seconds on the z1-z2 and 01-02planes to be zerdduring
every sampling interval.
The space vector PWM strategy discussed in this paper
is based on the double neutral topology as illustrated in
Fig. 2(b). With stator windings connected to double neutrals,
the projection of the current vector on the 01-02 plane is
inherently zero and the space vector PWM is performed only
on the d-q and z1-z2planes.
The space vector PWM strategy is accomplished by the
3 35
following equation
(a)

where U$ is the projection of the lcth voltage vector on

the x-axis, and T k is the dwell time of that vector during
time interval T,. The quantities vl; and are the d-q plane
reference voltages.
During each sampling period T,,a set of five voltage vectors
must be chosen to guarantee that each Tk has a positive and
unique solution. There are numerous means for choosing such
a set. The set used in this paper is illustrated in Fig. 4. In
the method chosen, four adjacent voltage vectors are always
selected from the vectors which span the outermost polygon on
the d-q plane according to the position of the reference voltage
vector Viq. The fifth vector is chosen from the zero vectors
located at the d-q plane origin. For example, as indicated
by the dots in Fig. 4(a), voltage vectors 49, 48, 56, and 60
are selected when the reference voltage vector lies inside the
triangle area between voltage vectors 48 and 56. In Fig. 4(b),
the four voltage vectors spread out to cover the z1-z~plane,
thus making it possible that the average volt-second on this
plane be zeroed during every sampling interval.
It is also noted from Fig. 4(b) that the vectors on the z1-
z2 plane have the smallest amplitude. Therefore, the proposed
space vector PWM strategy will offer the maximum voltage
output capability on the d-q plane yet keep the harmonics on
the z1-z2 plane at a minimum.

(C) C. Simulation Results

Fig. 3. (a) Inverter voltage vectors projected on d-q plane (switching modes The proposed vector space decomposition control technique
at origin: 0, 21, 42, 63). (b) Inverter voltage vectors projected on z 1 - z ~
plane (switching modes at origin: 0, 21, 42, 63). (c) Inverter voltage vectors has been simulated and the results are shown in Fig. 5. For
projected on 01-02plane. purpose of comparison, the control using the conventional
space vector PWM technique and sine-triangle PWM have
are mapped to the origin, The resultant system is a four- also been simulated and the results are shown in Figs. 6 and
dimensional system, and the control of the dual three-phase 7. In all the simulations, the inverter switching frequencies
machine is further simplified. were kept the same.

Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 04,2020 at 14:52:11 UTC from IEEE Xplore. Restrictions apply.
ZHAO AND LIPO: SPACE VECTOR PWM CONTROL OF DUAL THREE-PHASEINDUCTION MACHINE USING VECTOR SPACE DECOMPOSITION 1107

30 I I
1

-30 1 I
14 I I I
I

15

3 35
(a) -30 I
25 50
z2-axis t (msec.)

14 a 48
Fig. 6. Simulation result using conventional space vector P W M technique.
From top to bottom: machine phase current, current on d-q plane, and current
on z 1 - z ~plane.

15 49
(b)
Fig. 4. Switching modes selection.

-30 ' I

30 I I I
I

0 25 m
t (*I.
Fig. 7. Simulation result using sine-triangle P W M technique. From top to
bottom: machine phase current. current on d-q plane, and current on z1-z~
plane.

- a 7 ~ 1

0 25 50
t (mscc.) three-phase induction machine compared with the conven-
tional space vector PWM. However, as being compared with,
Fig. 5. Simulation result u&ng proposed technique. From top to bottom:
Machine phase current; current on d-q plane; and current on 21 -22 plane.
the control technique proposed in this paper, the sine-triangle
PWM technique produces a larger amplitude of harmonic cur-
For the conventional space vector PWM, two adjacent rents on the zI-z2 plane because the voltage vectors generated
voltage vectors plus one zero voltage vector at the origin are by the sine and the triangular carrier wave crossing cannot be
selected from the d-q plane. Since only the control over d-q guaranteed to have minimum projections on the z1-z2 plane.
From the comparison of the simulation results, it is apparent
plane variables is exercised, the harmonic currents on the z 1 - z ~
that much better control performance over previous approaches
plane are now left free to flow. As a result, the amplitudes of
can be achieved by using the proposed technique.
these currents will be expected to be quite large. Fig. 6 shows
the simulation results for this case, where considerable large
fifth and seventh harmonic currents can be observed. v. EXPERIMENTAL RESULTS
The simulation results in Fig. 7 suggests that the sine- The dual three-phase induction machine drive system ex-
triangle PWM technique is superior for the control of the dual amined in this paper has been constructed to test the pro-

Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 04,2020 at 14:52:11 UTC from IEEE Xplore. Restrictions apply.
IIO8 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 31, NO. 5, SEPTEMBEWOCTOBER 1995

I DSP56000 mMl? 0 a.. llop:ooooooo

I
L,_______--_____,,__I E r: w.87mvrr
Fig. 8. Test system block diagram. Fig. 9. Experimental results using the conventional space vector PWM.
Upper trace: current spectrum; lower trace: machine phase current, 5 A
(50 mv)/div.

posed analysis and control technique. The implementation uses

IGBT’s for the power switching devices and a MOTOROLA
DSP56000 microprocessor for the controller. A five horse-
power, four-pole induction machine with thirty-six stator slots
was rewound to form a six-pole, dual three-phase machine for Wru
the test. The system block diagram is shown in Fig. 8, and the mxv

operating principle illustrated as follows. 0

v r u
START: 0 Hx
nn -
0.64.cI W I *TOP: L 000 HI
I
The d-g plane is divided into 12 sections as indicated in 0 TIM€ 0
200
Fig. 4, and the position of the voltage reference vector Viq .Volt

was coded accordingly and used as the addresses for RAMI so

and RAM2. During each sampling period T,, five out of =volt
MXV
thirteen switching modes were picked up from RAM2 and
stored in the switching modes stack. Meanwhile, five voltage 400 1
START 0 0.C mop: 4 0 0 . I u c
vectors corresponding to the five switching modes in the d- E is Ht y: ae.m DYPY

q-zl-zgreference frame were picked up from RAMI and Fig. 10. Experimental results using the proposed space vector PWM. Upper
trace: current 3pectrum; lower trace: machine phase current, 5 A (50 mv)/div.
stored in the voltage vector stack for the purpose of dwell
time calculation. The dwell times for each of the five voltage
vectors are calculated from (22) and used to control the timer
can be observed compared with the conventional method. The
interrupt interval. Gating signals are sent out in the timer
experimental results have sufficiently verified the correctness
interrupt subroutine and their widths are determined by the
of the analysis and the feasibility of the proposed dual three
timer interrupt intervals.
phase induction machine control technique.
To compare the performance of the proposed space vector
PWM with the conventional method, both strategies were im-
plemented in software. Sampling frequencies of 2 kHz for the VI. CONCLUSION
proposed technique and 4 kHz for the conventional approach This paper has presented a unified approach to space vector
have been chosen to maintain the switching frequency for an control of dual three-phase induction machines. While the
individual IGBT in both cases to be equal at 2 kHz. Although machine examined specifically deals with the dual three-
this sampling frequency is far below the switching capability phase connection, it is clear that the approach may be readily
of an IGBT, the switching speed is limited by the instruction extended to any machine having sinusoidally wound stator
execution speed and the minimum achievable timer interrupt windings with any number of phases. It has often been stated
interval of the DSP. For the same reason, the machine for the that machines having a phase number greater than one are “all
tests was operated at 15 Hz rather than the rated 60 Hz. equivalent” from the modeling point of view. That is, it is said
The experimental results corresponding to the conventional that all multi-phase machines effectively reduce to the same
space vector PWM are shown in Fig. 9. As expected, the rt-g-0 equivalent circuit model. This paper has demonstrated
harmonic currents on the z1-z~ plane, i.e., the 5th, 7th, 17th, that this perception is false. In particular, even when the phase
and 19th, etc., are extraordinarily large due to the lack of currents sum to zero, m - 3 zero sequence and 6n f 1 (n
control over currents on this plane. The performance of the odd number) zero sequence components exist. It is further
proposed space vector PWM is illustrated in Fig. 10. Although shown that careful attention to the control of these added
the 5th and 7th harmonic currents are still appreciable due to zero sequence components are necessary to provide optimum
the switching frequency limitation, very substantial reductions current regulation of a multi-phase machine.

Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 04,2020 at 14:52:11 UTC from IEEE Xplore. Restrictions apply.
ZHAO AND LIPO: SPACE VECTOR PWM CONTROL OF DUAL THREE-PHASEINDUCTION MACHINE USING VECMR SPACE DECOMPOSInON 1109

ACKNOWLEDGMENT [9] R. H. Nelson and P. C. Krause, “Induction machine analysis for arbitrary
displacement between multiple winding sets,” IEEE Trans. Power App.
The authors are indebted to the member companies of and Svsr., vol. PAS-93, UP. 841-848, MadJune 1974.

[ I l l R. J. Schilling and H. Lee, Engineering Analysis-A‘kctor Space

REFERENCES Approach. New York: Wiley, 1988.

[I1 A. Nabae, 1. Takahashi, and H. Akagi, “A new neutral-point-clamped

PWM inverter,” IEEE Trans. Ind. Applicar., vol. IA-17, no. 5, pp.
5 18-523, Sept./Oct. 1981.
I21 T. M. Jahns, “Improved reliability in solid-state ac drives by means of Yifan Zhao (S’91) received the B.S. degree from
multiple independent phase-drive units,” IEEE Trans. Ind. AppLicaf.,vol. Jiaozuo College of Mining, P.R.C., in 1982, and the
IA-16, no. 3, pp. 321-331, May/June 1980. M.S. degree from the China University of Mining
131 J.-R. Fu and T. A. Lipo, “Disturbance free operation of a multiphase and Technology, in 1985, both in electrical engi-
current regulated motor drive with an opened phase,” IEEE Trans. Ind. neering.
Applicat., vol. 30, no. 5, pp. 1267-1274, Sept./Oct. 1994. From 1985 to 1990, he was with the Indus-
141 K. Gopakumar, V. T. Ranganathan, and S. R. Bhat, “Split-phase trial Automation Department, China University of
induction motor operation from PWM voltage source inverter,” IEEE
Mining and Technology. He is currently a Ph.D.
Trans. Ind. Applicat., vol. 29, no. 5, pp. 927-932, Sept./Oct. 1993.
candidate at the Department of Electrical and Com-
PI M. A. Abbas, R. Christen, and T. M. Jahns, “Six-phase voltage source puter Engineering, University of Wisconsin, Madi-
inverter driven induction motor,” IEEE Trans. Ind. Applicar., vol. IA-20,
no. 5, pp. 1251-1259, Sept./Oct. 1984. son. His research interests include the analysis,
design, and control of electric machine drives, power electronics circuits, and
I61 E. E. Ward and H. Harer, “Preliminary investigation of an inverter fed
5-phase induction motor.” IEE Proc., June 1969, vol. 116(B), No. 6, microprocessor-based control systems.
pp. 980-984.
[71 C. L. Fortescue, “Method of symmetrical co-ordinates applied to the
solution of polyphase networks,” AIEE Trans., vol. 37, pp. 1027-1 115,
1918.
181 D. White and H. Woodson, Electromechanical Energy Conversion. Thomas A. Lipo (M’64-SM’71-F‘87), for a photograph and biography, see
New York: Wiley, 1959. p. 1078 of this issue.

Authorized licensed use limited to: Universidad de Malaga. Downloaded on March 04,2020 at 14:52:11 UTC from IEEE Xplore. Restrictions apply.