TRANSFORMATION OF CONVERTER DYNAMIC MODELS
Charles Sao
University of Toronto
email: cksao@ieee.org
Peter W. Lehn
University of Toronto
email: lehn@ecf.utoronto.ca
Abstract
This paper proposes a reference frame transformation method which
applies appropriate rotational operators to the block diagram repre
sentation of a converter space vector model to transform it from one
reference frame to another. Unlike reference frame transformation
using matrices, it can easily be applied to converter models of aver
age complexity. A table of reference frame transform pairs and three
examples are provided to demonstrate the use of this method. Time
domain simulation results are used to validate the reference frame
transformation in each example.
Keywords—Voltage source converter; space vector model;
reference frames; Park transformation; Clarke transforma
tion; dqframe; αβframe.
1 Introduction
Reference frame transformation, such as Park transfor
mation [1], [2], [3], of converter dynamic models using ma
trices can be tedious in all but the simplest cases. This
paper provides a reference frame transformation method
that can easily be applied to converter models of average
complexity. In this method, reference frame transformation
is achieved by adding appropriate rotational operators to
the inputs and outputs of the block diagram representation
of the converter space vector model [4] and then eliminating
those operators through block diagram manipulation. This
transformation method is applicable to a simple system,
like a set of threephase current signal ﬁlters, as well as to
complete closed loop converter models including the plant,
controller and ﬁlter elements. Three application examples
of the proposed transformation method are presented in
this paper to highlight its relevance.
Section 2 of this paper describes the proposed transfor
mation method and tabulates the reference frame trans
form pairs for some common transfer functions. Section
3 presents a typical application of the proposed method,
where a system containing both synchronous dqframe com
ponents and stationary abcframe components is converted
into a common reference frame. Section 4 demonstrates
how a given dqframe current controller for a converter may
be transformed into an equivalent phase current controller
that tracks sinusoidal references.
2 Reference Frame Transformation
Any threephase voltage or current may be broken into a
space vector, and an associated zero sequence component,
as per (1) and (2).
i
(S)
=
2
3
(1i
a
+ ai
b
+ a
2
i
c
) (1)
i
0
=
1
3
(i
a
+ i
b
+ i
c
) (2)
Where a = e
j2π/3
. The superscript (S) is employed to in
dicated that the space vector is in the stationary reference
frame. The above transformation of phase quantities into
a stationary frame and an associated zero sequence compo
nent may be shown to be equivalent to the familiar Clarke
Transform [5]. In speciﬁc, the stationary frame space vector
is merely a linear combination of the αβframe quantities:
i = i
α
+ ji
β
. (3)
As quantities are transformed from one rotating refer
ence frame to another, only the space vector is trans
formed, while the zero sequence component remains un
altered. Since the zero sequence component is invariant of
the reference frame, it will be neglected in the subsequent
analysis.
Fig. 1(a) shows the space vector model of a dynamic
system with input u
∗(X)
and output u
(X)
. This model,
which is formulated in rotating reference frame X, may be
transformed into its equivalent in rotating reference frame
Y using the following two step procedure.
Step 1: add rotational operators to its input and output
as shown in Fig. 1(b). The two operators are exact inverses
of each other. The rotation angle is given by Δθ = θ
Y
θ
X
where θ
X
and θ
Y
are the angles of Xframe and Yframe
with respect to the common stationary reference frame S.
Δθ need not be constant as reference frames X and Y may
be rotating at diﬀerent angular frequencies. The output
operator transforms the original system output u
(X)
into
u
(Y )
, its equivalent in Yframe. Conversely, the input op
erator transforms the incoming signal in the Yframe u
∗(Y )
into the Xframe quantity u
∗(X)
. The input transformation
is necessary because the input of the original model must
be a vector in Xframe.
Step 2: move operator e
jΔθ
through the model block dia
gram until it reaches the output as shown in Fig. 1(c). This
step transforms the transfer function of the original system
model in Xframe into its equivalent in Yframe, while leav
ing constant gain and summation blocks unchanged. The
rotational blocks themselves cancel each other.
The following sections use a controlled sinusoidal pulse
width modulated converter as an example to demonstrate
1424400384 2006
IEEE CCECE/CCGEI, Ottawa, May 2006
2270
Space Vector
Model in Reference
Frame X
) X (
u
) X *(
u
θ Δ j
e
θ Δ − j
e
) Y (
u
) Y *(
u
Space Vector
Model in Reference
Frame Y
θ Δ j
e
θ Δ − j
e
(b)
(c)
) Y *(
u
) Y (
u ) Y (
u
) X (
u
Space Vector
Model in Reference
Frame X
) X (
u
) X *(
u
(a)
Figure 1: The Reference Frame Transformation Procedure.
Gating
Signal
sabc
v
L R
abc
i
DSP with
dqFrame
Control
Converter
tabc
v
fabc
i
τ +s 1
1
Voltage
Sesnsor
sabc
v
Figure 2: The Converter Used in the Examples.
the application of the block diagram reference transforma
tion method. Fig. 2 shows the considered converter and its
digital controller, as taken from [6]. The converter is con
nected to the ac supply through inductance L and resistance
R. The converter switching frequency is well above the fun
damental frequency of the ac supply. The dc side of the
converter is connected to a strong dc bus with suﬃciently
high voltage. As may be seen from the ﬁgure, most of the
control system is described in the synchronous dqframe
within the DSP, however, analog ﬁlters on the feedback
current signals are described in the stationary abcframe.
3 dqFrame Model of ThreePhase
Current Signal Filters
This ﬁrst example transforms the model of the feedback
current signal ﬁlters in Fig. 2 into an equivalent in the
synchronous dqframe, which will henceforth be referred to
as Xframe for convenience. The transformation allows the
entire closed loop converter system to be represented in a
single rotating reference frame. This facilitates both com
pensator tuning and stability analysis.
Before reference frame transformation is possible, the
space vector model of the ﬁlters in the stationary reference
is found. Considering an identical ﬁlter transfer function
G(s) on each of the three phase quantities, we have:
i
(S)
filt
= 1G(s)i
a
+ aG(s)i
b
+ a
2
G(s)i
c
(4)
i
(S)
filt
= G(s)i
(S)
. (5)
(a)
(b)
τ ω + + ) j s ( 1
1
X
τ +s 1
1
X
j
e
θ
) X (
i
) X (
f
i
) S (
i
) S (
f
i
X
j
e
θ
X
j
e
θ −
X
j
e
θ −
Cancellation
) X (
i
) S (
f
i
) X (
f
i
) e i (
X
j ) S (
f
θ −
Figure 3. The Space Vector Model of the abcFrame Current
Signal Filter
Thus for balanced systems, the transfer function of phase
a is identical to the space vector transfer function in the
stationary frame.
To transform the Sframe ﬁlter model into its equivalent
in Xframe, rotational operators are added to the model
input and output, as shown in Fig. 3(a). The rotation
angle θ
X
is the angle of Xframe with respect to Sframe.
The operator e
jθX
is then pulled through the transfer func
tion, transforming it into its equivalent in Xframe as shown
in Fig. 3(b). The relationship between i
(X)
and i
(S)
f
in
Fig. 3(a) is used as a starting point to derive the equiva
lent transfer function in Xframe. That relationship can be
written in time domain as:
τ
di
(S)
f
dt
+ i
(S)
f
= i
(X)
e
jθX
(6)
Multiplying both sides of (6) by e
−jθX
and subtracting
and adding τje
−jθX
dθX
dt
i
(S)
f
to the left hand side results in:
τe
−jθX
di
(S)
f
dt
−τje
−jθX
dθ
X
dt
i
(S)
f
+ τje
−jθX
dθ
X
dt
i
(S)
f
+
i
(S)
f
e
−jθX
= i
(X)
(7)
The ﬁrst two terms of (7) make up the derivative of
(i
(S)
f
e
−jθX
). Substituting these two terms with
d(i
(S)
f
e
−jθ
X
)
dt
and rearranging the terms yield:
d(i
(S)
f
e
−jθX
)
dt
=
−(i
(S)
f
e
−jθX
) −jω
X
τ(i
(S)
f
e
−jθX
) + i
(X)
τ
(8)
where ω
X
=
dθX
dt
is the angular frequency of Xframe.
Taking Laplace transform on (8) results in:
(i
(S)
f
e
−jθX
)(s)
i
(X)
(s)
=
1
1 + (s + jω
X
)τ
(9)
Since (i
(S)
f
e
−jθX
) = i
(X)
f
in the block diagram of Fig. 3
(b), the equivalent ﬁlter model in Xframe is:
2271
) X (
i
) X *(
i


+
 

) s ( R
Current Control Plant
3 Σ 2 Σ 1 Σ ) s ( C
L j
X
ω
) X (
f
i
τ ω + + ) j s ( 1
1
X
φ j
Ke
L j
X
ω
) X (
i
) X (
t
v P
) X (
e
) X (
u
) X (
s
v
+
Figure 4: The Closed Loop Converter Model in XFrame.
0.2 0.22 0.24 0.26 0.28 0.3
5
0
5
10
15
20
25
30
35
Time (s)
C
u
r
r
e
n
t
(
A
)
Step Responses with abcFrame Filter and XFrame Equiv. Filter
iqref
iq  abcFrame Filter
iq  XFrame Equivalent Filter
Figure 5. The Step Responses of Converter Models with abc
Frame Current Signal Filter and the Equivalent in XFrame.
i
(X)
f
(s)
i
(X)
(s)
=
1
1 + (s + jω
X
)τ
(10)
This ﬁlter model may be used to synthesize the closed
loop converter model in which the transfer functions of the
plant, the current control and the ﬁlter are all formulated
in Xframe. This is shown in Fig. 4. The transfer functions
are R(s) =
1
sL+R
and C(s) =
Kps+Ki
s
where L, R, K
p
and
K
i
are converter interface inductance, resistance, PI com
pensator proportional gain and integral gain respectively.
The constant gain Ke
jφ
compensates for the attenuation
and phase shift of the feedback current signal due to the
ﬁlter. This model is the space vector equivalent of the dq
frame converter model presented in [2], [6].
To validate the transformation, the model of a typical D
STATCOM [6] is simulated, ﬁrst using phase current ﬁlters,
and then using the equivalent synchronous frame model
presented in Fig. 4. (The system parameters are L =0.637
mH, R = 0.15 Ω, K
p
= 0.28 V/A, K
i
= 41.8 V/A/s, τ =
2 ms, K = 1.2524, φ = 37 degrees and ω
X
= 377 rad/s.)
Fig. 5 shows the responses of i
q
, the imaginary component
of i
(X)
, to a step change in the reference i
∗
q
, the imaginary
component of i
∗(X)
, from 0 to 10 A. Perfect agreement be
tween the two step responses validates the reference frame
transformation of the current signal ﬁlter from Sframe to
Xframe.
TABLE I
Common Reference Frame Transform Pairs
Description TF in X TF in Y
Gain K K
Rotation e
jρ
e
jρ
Integrator 1/s 1/(s + jΔω)
Derivative s (s + jΔω)
First Order Pole 1/(1 + sτ) 1/(1 + (s + jΔω)τ)
First Order Zero (1 + sτ) (1 + (s + jΔω)τ)
It may be noted in this example that the ﬁlter transfer
function undergoes a frequency shift of ω
X
as a result of
the transformation from Sframe to Xframe whose angular
frequency is ω
X
rad/s above that of Sframe. Similarly, Ta
ble I shows that a transfer function undergoes a frequency
shift of Δω=ω
Y
ω
X
when it is transformed from Xframe
to Yframe. ω
X
and ω
Y
are the angular frequencies of X
frame and Yframe. Some common reference frame trans
form pairs are given in Table I.
4 Converter Phase Current Controllers
This example transforms the closed loop converter model
of Fig. 4 into its equivalent in the stationary reference
frame S. Such a transformation may be performed when
one wishes to derive a converter controller which regulates
the abc phase currents to track sinusoidal references. The
resultant current controller is dynamically equivalent to the
space vector current controller of Fig. 4.
Step 1 in performing the transformation is to add rota
tional operator e
−jθX
to the input and e
jθX
to the output
of the model block diagram of Fig. 4. The rotation angle
θ
X
is the angle of Xframe with respect to Sframe. To
facilitate the next step, operator e
jθX
is moved to the left
of node P, resulting in an additional operator e
−jθX
on the
feedback path as shown in Fig. 6.
Step 2 of the transformation is to move the two e
−jθX
operators (one at the input and one at in the feedback path)
along the signal paths through all the blocks towards the
output where they cancel the e
jθX
term. The following
paragraphs focus on the derivation of the equivalent current
controller in abcframe because the plant transfer function
in abcframe, which models the inductance L and resistance
R, is wellknown as
1
sL+R
and the current signal ﬁlter model
in abcframe is given in Fig. 2.
The transfer function C(s) of Fig. 6 undergoes a fre
quency shift of −ω
X
in step 2. Thus, its equivalent in
Sframe is C
(s) =
Kp(s−jωX)+Ki
(s−jωX)
as ascertained from Ta
ble 1. Fig. 7(a) shows the block diagram of C
(s) as a PI
compensator with a frequency shift of −ω
X
in the integral
block. This block may be decomposed into real and imag
inary components as shown in the block diagram in Fig.
7(b).
Step 3 transforms the block diagram in Fig. 7(b) into
its equivalent in abcframe. The upper two branches of Fig.
2272


+



) s ( R
Current Control Plant
3 Σ 2 Σ 1 Σ ) s ( C
φ j
Ke
X
j
e
θ −
X
j
e
θ
X
j
e
θ −
) X (
t
v
τ ω + + ) j s ( 1
1
X
L j
X
ω L j
X
ω
) X *(
i
) X (
f
i
) X (
i
) X (
i ) S *(
i
) S (
i
P
) X (
u
) X (
e
) S (
s
v
+
X
j
e
θ −
) X (
s
v
Figure 6: Reference Frame Transformation from XFrame to SFrame.
(b)
+
+
) S (
u
) S (
e
p
K
) s (
s K
2
X
2
i
ω +
) s (
K j
2
X
2
i X
ω +
ω
+
(a)
+
+
) S (
u
) S (
e
p
K
) j s (
K
X
i
ω −
abc
u
abc
e
(c)
+
+
p
K
) s (
s K
2
X
2
i
ω +
) s (
K
2
X
2
i X
ω +
ω
+
)
2
( R
π
Figure 7: (a),(b)Space Vector Compensator C
(s) in SFrame; (c)Current Loop Compensator in abcFrame.
7(b) comprise real transfer functions, (i.e. ones having real
valued coeﬃcients), call it G(s). As per (4) and (5) any
real transfer function G(s) acting on a space vector in the
Sframe is equivalent to an identical G(s) acting on each
of the individual phase quantities in the abcframe. This
equivalence assumes no 0sequence component at the input
to the abcframe transfer functions.
The third branch in Fig. 7(b) consists of the j operator
and a real transfer function. As with other rotational oper
ators, the j operator needs yet to be transformed to obtain
its abcframe equivalent. An arbitrary space vector rota
tional operator e
jρ
may be transformed into the abcframe
using the relation:
R(ρ) =
2
3
⎡
⎣
cos ρ cos(ρ +
2π
3
) cos(ρ +
4π
3
)
cos(ρ +
4π
3
) cos ρ cos(ρ +
2π
3
)
cos(ρ +
2π
3
) cos(ρ +
4π
3
) cos ρ
⎤
⎦
(11)
A proof of (11) is given in the Appendix. The abcframe
equivalent of the operator j is R(
π
2
) and Fig. 7(c) shows
the resulting abcframe current controller, where, again, the
equivalence is valid provided e
abc
contains no zero sequence
term. Fig. 8 shows the entire system block diagram in the
abc frame. It should be noted that the constant gains Ke
jφ
and jω
X
L in the current controller of Fig. 6 also become
KR(φ) and ω
X
LR(
π
2
) respectively in Fig. 8
To validate the transformation, the converter models of
Fig. 4 and 8 are simulated in PSCAD/EMTDC. The sys
tem parameters in this example have the same value as in
Section 3. To ensure that the step inputs to the two mod
els are equivalent, i
∗
abc
in Fig. 8 is derived by transforming
i
∗(X)
in Fig. 4 into abcframe. Moreover, output i
abc
is
transformed into a space vector in Xframe so that the step
responses of the two models can be compared. Fig. 9 shows
the responses of i
q
, the imaginary component of i
(X)
, to a
step change in i
∗
q
, the imaginary component of i
∗(X)
, from 0
to 10 A. For both models, i
q
rises to track the new reference
input. There is complete agreement between the two step
responses, validating the reference frame transformation of
the converter model from Xframe to abcframe.
5 Conclusions
This paper proposes a reference frame transformation
method which applies appropriate rotational operators to
the block diagram representation of a converter space vector
model to transform it from one reference frame to another.
Unlike reference frame transformations using matrices, it
can easily be applied to converter models of average com
plexity.
A table of reference frame transform pairs and two ap
plication examples are provided in the paper. The ﬁrst
example transforms a set of converter phase current signal
ﬁlters into its equivalent in dqframe, (which is called X
frame in section 3), to derive the complete dqframe model
of the converter and its controls for further use. The sec
ond example derives a controller to regulate converter phase
currents by transforming the spacevector converter model
from dqframe, (which is called Xframe in section 4), to
2273
abc
e
+
 

) s ( R
2 Σ 1 Σ
tabc
v
abc
i
fabc
i
) ( KR φ
abc
u
abc
i
*
abc
i
)
2
( LR
X
π
ω

3 Σ
sabc
v
+
Current Control Plant
τ +s 1
1
+
+
p
K
) s (
s K
2
X
2
i
ω +
) s (
K
2
X
2
i X
ω +
ω
+
)
2
( R
π
Figure 8: Closed Loop Converter Model in abcFrame.
0.2 0.22 0.24 0.26 0.28 0.3
5
0
5
10
15
20
25
30
35
Time (s)
C
u
r
r
e
n
t
(
A
)
Step Responses of XFrame and abcFrame Current Controllers
iqref
iq  XFrame Controller
iq  abcFrame Equivalent
Figure 9: Step Responses of VSC Models in Fig. 4 and Fig. 8
Sframe. The resultant model is transformed further to de
rive the abcframe current controller. Simulation results
for the examples demonstrate the validity of the proposed
transformation method.
References
[1] P.C. Krause et. al., ANALYSIS OF ELECTRIC MA
CHINERY. Piscataway, NJ: IEEE Press, 1995.
[2] C. Schauder and H. Mehta, “Vector analysis and con
trol of modern static VAR compensators,” Proc.Inst.
Elec. Eng., Pt. C, vol. 140, no. 4, pp. 299306, July
1993.
[3] R. Wu et. al., “Analysis of an ACtoDC voltage source
converter using PWM with phase and amplitude con
trol ,” IEEE Transactions on Industry Applications,
vol. 27, no. 2, pp. 355  364, MarchApril 1991.
[4] A. Sonnenmoser, P. Lehn, “Line current balancing
with a uniﬁed power ﬂow controller,” IEEE Transac
tions on Power Delivery, vol. 14, no. 3, pp. 1151  1157
, July 1999.
[5] E. Clarke, CIRCUIT ANALYSIS OF AC POWER
SYSTEMS. New York, NY: Wiley, 1950, vol. 1.
[6] C. K. Sao et. al., “A benchmark system for digital
timedomain simulation of a pulsewidthmodulated
DSTATCOM,” IEEE Transactions on Power Deliv
ery, vol. 17, no. 4, pp. 11131120, October 2002.
Appendix A: Rotation Matrix
Given the space vector deﬁnition of (1), the abcframe
components of y
(S)
can be expressed as:
y
a
= 0.5(y
(S)
+ conj(y
(S)
)) (12)
y
b
= 0.5(a
2
∗ y
(S)
+ a ∗ conj(y
(S)
)) (13)
y
c
= 0.5(a ∗ y
(S)
+ a
2
∗ conj(y
(S)
)) (14)
Consider the rotational transformation:
y
(S)
= e
jρ
i
(S)
(15)
Substituting (1) in (15) yields:
y
(S)
=
2
3
(e
jρ
i
a
+ e
j(ρ+2π/3)
i
b
+ e
j(ρ+4π/3)
i
c
) (16)
Substituting (16) in (12) results in the following equa
tions
i
a
=
1
2
∗
2
3
((e
jρ
+ e
−jρ
)i
a
+ (e
j(ρ+
2π
3
)
+ e
−j(ρ+
2π
3
)
)i
b
+
+(e
j(ρ+
4π
3
)
+ e
−j(ρ+
4π
3
)
)i
c
) (17)
=⇒i
a
=
2
3
cos ρ cos(ρ +
2π
3
) cos(ρ +
4π
3
)
i
abc
(18)
where i
abc
is a 3x1 matrix with elements i
a
, i
b
and i
c
.
Similarly,
i
b
=
2
3
cos(ρ + 4π/3) cos ρ cos(ρ + 2π/3)
i
abc
(19)
i
c
=
2
3
cos(ρ + 2π/3) cos(ρ + 4π/3) cos ρ
i
abc
(20)
Combining (18), (19) and (20) results in the rotation ma
trix of (11).
2274