Вы находитесь на странице: 1из 6

3500

An Improved Distributed Control Strategy for


Parallel Inverters
Tianzhi Fang, Xinbo Ruan, Senior Member, IEEE, Lan Xiao and Aizhong Liu
Aero-Power Sci-tech Center, College of Automation Engineering
Nanjing University of Aeronautics and Astronautics, Nanjing, 210016, China
Tel: +86-25-84892053, Fax: +86-25-84892053
E-mail: fangtianzhi@126.com
AbstractParallel operation of inverters improves the
reliability, maintainability, expandability, and standardization.
Many control methods have been proposed. The master-
slave method can not realize the redundancy because the
slave modules are dependent on the master module, the
wireless control method using frequency and voltage droop
technique presents too poor output characteristics.
Distributed control strategy achieves hot swap and
redundancy, the output characteristics is relatively poor due
to the inductor current feedback loop. This paper introduces
the load current feed-forward control to the distributed
control strategy to improve the output characteristics.
Meanwhile, the functions of output current limiting and
current sharing are remained. The output characteristics
and circulating current with and without the load current
feed-forward control are compared in this paper. The
prototype of parallel inverters is built to verify the
effectiveness of the improved method.
I. INTRODUCTION
Parallel redundant operation of inverters improves the
reliability, maintainability, expandability and standardization.
The key issue is ensuring the output current be equally
shared by the constitutive inverter modules. Many control
methods have been proposed, including master-slave
control [1-3], wireless control [4-7] and distributed
control [8-16]. The master-slave method has simple
current sharing control circuit, but it can not realize the
redundancy because the slave inverters can not work
alone and the whole system will collapse in case the
master module fails. The wireless control, namely the
frequency and voltage droop technique, requires no
interconnections among the modules and it is easy to
achieve redundancy, however the system output
characteristics is very poor. The distributed control can be
classified into two categories, i.e., the power deviation
control method [8-11] and the average current control
method [12-16]. The former one is based on the theory
that the active and reactive power is separately
determined by the phase and amplitude of each modules
output voltage, respectively. It needs complicated control
circuit for current sharing. The latter one is based on
average current control. Particularly, the strategy
proposed in [14-15] decouples the parallel control into
the synchronization of voltage reference and the
averaging of current reference. The instantaneous current
sharing is acquired with simple circuit, and hot swap and
redundancy are easy to achieve. The inductor-current
feedback control is employed in this method and it
induces poor output characteristics.
In this paper, the load current feed-forward control is
introduced into the average current control method to
improve the output characteristics. Meanwhile, the
functions of output current limiting and current sharing
are kept. The output characteristics and circulating
current between the original and improved methods are
compared, and experimental results are presented to
verify the effectiveness of the improved method.
II. CONTROL METHOD OF SINGLE INVERTER
The circuit block diagram and the control block
diagram of single inverter are illustrated in Fig.1 and
Fig.2, respectively. The voltage and current double
close-loop control is employed. The voltage outer loop
adopts a proportional-integral (PI) regulator with the
transfer function of G
v
(s)=K+(1/Ts). The current inner
Fig. 1. Circuit block diagram.
(a)
(b)
Fig. 2. Control block diagram of single inverter.
978-1-4244-1668-4/08/$25.00 2008 IEEE
3501
loop is normally the inductor current feedback. In order
to improve the inverter output characteristics, the load
current feed-forward, as shown with broken line arrow in
Fig.2(a), is introduced, where K
i
and K
o
are inductor
current sampling ratio and load current sampling ratio,
respectively. When K
o
is equal to K
i
, the output voltage of
inverter is independent of the load and the optimal output
characteristics is achieved. At this time, the current loop
is equivalent to the capacitor-current feedback. In the
case of K
o
=K
i
, when the load current feed-forward is at
point A, the current-limiting component actually limits
the capacitor current by limiting the capacitor current
reference. However, the load current can not be limited
for the capacitor current does not include the information
of the load current. When the load current feed-forward is
at point B, the current-limiting component limits the
inductor current by limiting the inductor current reference.
The load current can be limited for the inductor current
being the sum of the load current and the capacitor
current.
With well-designed current tracking performance, the
inner loop (referring to the dashed frame in Fig.2(a)) can
be simplified as the current follower whose amplification
K
I
is equal to 1/K
i
[11, 15]. So Fig.2(a) is equivalent to
Fig.2(b) when the load current feed-forward is at point B.
III. CONTROL STRATEGY OF PARALLEL SYSTEM
The control block diagram of multi-module parallel
system is shown in Fig.3. The voltage reference signals of
all modules, U
r1
(s), U
r2
(s), , U
rN
(s), are generated by
DSP and high precision digital-to-analog converter and
they are synchronized by DSP, so they have almost the
same amplitude, frequency and phase, i.e.,
1 2
( ) ( ) ( ) ( )
r r rN r
U s U s U s U s = = = "" (1)
Fig. 3. Control block diagram of parallel system.
Fig. 4. Control block diagram of two-module parallel system.
In the conventional average current control method
[14-15] (see Fig.3 excluding the dashed arrow), the
output characteristics of single or parallel inverters is
poor due to the inductor-current feedback control.
If the added load current feed-forward is at point A and
K
oj
K
ij
, it is equivalent to the capacitor-current feedback
control for single inverter and the optimum output
characteristics is achieved accordingly. In this strategy
the capacitor current reference signals of all inverter
modules are averaged as the common current reference
signal, which is followed by capacitor current of all
inverter modules that reach coherence ultimately. But the
coherent capacitor current can not promise the coherent
load current because the former does not include the
information of the latter. So current sharing can not be
obtained, and moreover, the current-limiting component
can only limit the capacitor current but can not limit the
load current.
If the added load current feed-forward is at point B and
K
oj
K
ij
, the output characteristics independent of the
load can also be achieved, which will be analyzed in the
next section. The inductor current reference signals of all
inverter modules are averaged as the common one, which
is followed by the inductor current of all inverter modules
that reach coherence finally. The load current will be
shared as long as the output filter capacitors are equal.
Meanwhile the current-limiting component can limit the
load current by limiting the inductor current. In the
following text, the control strategy of the inductor-current
feedback combined with and without the load current
feed-forward (added load current feed-forward is at point
B and K
oj
K
ij
) are called strategy #1 and strategy #2,
respectively.
IV. COMPARISON BETWEEN TWO CONTROL STRATEGIES
OF PARALLEL SYSTEM
A. Comparison of Output Characteristics
For no loss of generality, a two-module parallel system
is taken as the example to compare the output
characteristics and circulating current between the two
strategies. Fig.4 shows the control block diagram, where
the current loops are simplified as the current followers.
Voltage references, namely U
r1
(s) and U
r2
(s) can be
considered unique one, that is, U
r1
(s)=U
r2
(s)=U
r
(s) for
their synchronous and identical in amplitude. The transfer
function of the parallel system can be deduced as
1 2 1 2
1 2
2
1 2 3
1 1
( ) ( )
( )
( )
( )
I I
o
r
K K K K s
T T
U s
s
U s s s
|
( | |
+ + + +
( |
\ .
= =
A + A + A
(2)
where
1 2 1 1 2 2 1 1 2
( )( ) 2( )
o o I f I f f f
K K K C K C C C A = + + ,
1 1 2 2
2 1 2 1 1 2 2
2 ( )
( )( )
( )
o I o I
I I v v
K K K K
K K K K K K
Z s
+
A = + + + ,
1 2
3 1 2
1 2
( )
v v
I I
K K
K K
T T
| |
A = + +
|
\ .
.
According to (2), the amplitude frequency
characteristics under resistive load can be expressed as
3502
2
2 2 2
1 2 1 2
1 2
1 2
1 1
( ) ( )
( )
I I
K K K K
T T
j
e
| e
(
| |
( + + + +
|
( \ .

=
B + B
(3)
where
}
1 2
1 1 2
1 2
2
2
2 1 1 2 2 1 1 2
( )
[( )( ) 2( )]
v v
I I
o o I f I f f f
K K
K K
T T
K K K C K C C C e
| |

B = + +
|
\ .
+ +
,
2
2 1 1 2 2
2 1 2 1 1 2 2
2 ( )
( )( )
o I o I
I I v v
K K K K
K K K K K K
R
e
+ (
B = + + +
(

.
In the case of single inverter, we can obtain
2 2 2
2
2 2
2 2
1
( )
( )
1
I
I v I o
f I v
K K
T
j
K K K K
C K K K
T R
e
| e
e e
+
=
| | | |
+ +
| |
\ . \ .
(4)
The rms value of the output voltage U
o
develops with the
same trend as ||(je)| for the relationship of U
o
=||(je)|U
r
.
Expression [2(K
o1
K
I1
+K
o2
K
I2
)]/R in (3) will be 2/R under
strategy #2 where K
o1
K
o2
0, and then ||(je)| and U
o
decrease with the increasing load. This conclusion can
also be drawn for single inverter for expression
(1K
Ij
K
oj
)/R in (4) will be 1/R under strategy #2 where K
oj
0 (j=1,2)). Obviously strategy #2 results in poor output
characteristics. Under strategy #1 where K
oj
=K
ij
=1/K
Ij
(j=1,2), both expressions (1K
Ij
K
oj
)/R and
[2(K
o1
K
I1
+K
o2
K
I2
)]/R are equal to zero, which promises
that the output voltage of parallel system and that of
single inverter are both independent of the load, so the
output characteristics is greatly improved.
B. Comparison of Circulating Current
From Fig.4, we can derive the following equations.
1 1 1
( ) ( ) ( )
o I g f o
I s K I s sC U s = (5)
2 2 2
( ) ( ) ( )
o I g f o
I s K I s sC U s = (6)
| |
1 2
( ) ( ) ( ) ( ) ( ) ( )
o o o o
U s I s Z s I s I s Z s = = + (7)
Substitution of (5) and (6) into (7), yields
1 2
1 2
1 ( ) ( )
( ) ( )
( )( )
f f
g o
I I
Z s s C C
I s U s
Z s K K
+ +
=
+
(8)
The circulating current can be derived from (5), (6) and
(8) as
| |
1 2
1 2 2 1
1 2 2 1
1 2
1 2 1 2
( ) ( ) ( ) 2
( ) ( ) ( ) ( ) 2
( )
( ) ( )
2( )
H o o
I I g f f o
I f I f
I I
o o
I I I I
I s I s I s
K K I s s C C U s
s K C K C
K K
U s I s
K K K K
=
(
= +

= +
+ +
(9)
Eq. (9) can be rewritten in phasor form as
1 2 2 1
1 2
1 2 1 2
1 2
1 2
1 2 2 1
1 2
1 2 1 2
2( )
2( )
2( )
I f I f
I I
H o o
I I I I
I I
o
I I
I f I f
I I
o o
I I I I
K C K C
K K
I I j U
K K K K
K K R
I
K K Z
K C K C
K K X
j U I
K K K K Z
e
e
- - -

= +
+ +

=
+
(

+
(
+ +
(

(10)
1 Lf
I
-
2 Lf
I
-
1 o
I
-
2 o
I
-
1 Cf
I
-
2 Cf
I
-
o
I
-
Fig.5 Circuit diagram of two-module parallel system
<
o U
2
<
H I
1
<
Lf I
2
<
Lf I
1
<
o I
2
<
o I
I I 1 2
< <
Cf Cf =
<
o I
<
o U
2
<
H I
1
<
o I
2
<
Cf I
1
<
Cf I
1 2
< <
Lf Lf I = I
<
o I
2
<
o I
(a) (b)
Fig. 6. Phasor diagram of parallel system: (a) different K
I
, same C
f
, and
(b) different C
f
, same K
I
.
where Z=R+jX. The real and the imaginary parts of (10)
represent the active and reactive circulating current
respectively. From (10), we can draw the following
conclusions: 1) the difference of K
I
results in both active
and reactive circulating currents, and 2) the difference of
C
f
leads to reactive circulating current. This can also be
explained by the phasor diagram. The circuit diagram and
the phasor diagram of two-module parallel system are
shown in Fig.5 and Fig.6. Expression / 2
H Lf
I I
- -
= A can
be got from Fig.6(a) under different K
I
and the same C
f
.
Since st/2 (where is the angle between the inductor
current and the output voltage), so the conclusion 1) can
be drawn for |st/2 or |e[t/2, t], where | is the angle
between the circulating current and the output voltage.
Whereas expression / 2
H Cf
I I
- -
= A can be obtained
from Fig.6(b) under different C
f
and the same K
I
. Then
the conclusion 2) can be drawn for |=u=t/2, where u is
the angle between the capacitor current and the output
voltage.
At resistive load, (10) can be simplified as
1 2 2 1
1 2
1 2 1 2
2( )
I f I f
I I
H o o
I I I I
K C K C
K K
I I j U
K K K K
e
-

= +
+ +
(11)
and the rms value of the circulating current is
2 2 2 2 2
1 2 1 2 2 1
1 2
( ) 4( )
2( )
I I o I f I f o
H
I I
K K I K C K C U
I
K K
e +
=
+
(12)
Besides the circulating current, current sharing effect
can also be described by current sharing precision which
is given as
2 2 2 2
1 2 2 1 1 2
1 2
4 ( ) ( )
/ 2
I f I f I I
H
o I I
K C K C R K K
I
I K K
e

+
= =
+
(13)
From (13), we can see that current sharing precision
decreases with the increasing load.
3503
The circulating current, current sharing precision, and
current sharing effect under the two strategies can be
considered the same because the discussion above can be
applied to both strategy #1 and strategy #2.
V. EXPERIMENTAL RESULTS
The prototype of two-module parallel inverters is built
to verify the theoretical analysis. The specifications are as
follows: Output voltage: V
o
=115VAC/400Hz; Output
voltage sampling ratio: K
v
=0.031; PI voltage regulator
G
v
(s): K=3.92, T=34.68s; Output filter inductor:
L=0.6mH.
Fig.7 gives the experimental waveforms at full
resistive load under the two strategies, here we purposely
let the output filter capacitors are the same, i.e.,
C
f1
=C
f2
=15F, meanwhile K
I
is different, i.e., K
I1
=3.33,
and K
I2
=5. It can be seen that there is active and reactive
circulating current under both strategies for |e[t/2, t].
This result matches conclusion 1) in section IV.B.
Moreover,we can also see that the circulating current
under two strategies is almost the same (I
H1
=2.18A,
I
H2
=2.16A) while the rms value of output voltage under
strategy #1 is higher than that under strategy #2
(U
o1
=115.49V,U
o2
=112.69V).
Fig.8 gives the measured curves when K
I
is different
(C
f1
=C
f2
=15F, K
I2
=5). Fig.8(a) presents the curves of
output characteristics under the two strategies when K
I
is
different (K
I1
=3.33, 6.67). It can be seen that strategy #1
has a better output characteristics than strategy #2.
Fig.8(b) depicts the curves of the rms value of the
circulating current at different load under two strategies
when K
I
is different (K
I1
=3.33, 4.50, 5.50, 6.67). Fig.8(c)
shows the curves of current sharing precision under the
two strategies when K
I1
is equal to 6.67. It can be seen
(a)
(b)
Fig.7. Experimental waveforms at full resistive load (C
f1
=C
f2
=15F,
K
I1
=3.33, K
I2
=5,): (a) strategy #1, and (b) strategy #2
from Fig.8(b) and (c) that the curves with the same
parameters under the two strategies are highly close,
which means that the two strategies has the same current
sharing effect.
(a)
(b)
(c)
Fig.8 Measured curves under two strategies when K
I
is different
(C
f1
=C
f2
=15FK
I2
=5): (a) output characteristics, (b) circulating current,
and (c) current sharing precision.
Fig.9 gives the experimental waveforms at full
3504
resistive load under the two strategies with the same K
I
and different C
f
(K
I1
=K
I2
=5, C
f1
=30F, C
f2
=15F). It can
be seen that there is only reactive circulating current
under both strategies for |=t/2. This result matches
conclusion 2) in section IV.B. Furthermore, the
circulating current under the two strategies is almost the
same (I
H1
=2.45A, I
H2
=2.47A) whereas the rms value of
output voltage under strategy #1 is higher than that under
strategy #2 (U
o1
=116.42V, U
o2
=113.94V).
Fig.10 gives the measured curves under the two
strategies when C
f
is different (K
I1
=K
I2
=5, C
f2
=15F).
Fig.10(a) presents the curves of output characteristics,
from which, it can be seen that strategy #1 has a better
output characteristics than strategy #2. Fig.10(b) depicts
the curves of the rms value of the circulating current at
different load. Fig.10(c) shows the curves of current
sharing precision when C
f1
is equal to 20F. From
Fig.10(b) and (c), we can see that the curves with the
same parameters under the two strategies are highly close,
which means that the two strategies have the same current
sharing effect.
Fig.11 depicts the dynamic current-limiting waveforms
of parallel system under strategy #1 when the load steps
up and down between the rated load current and limited
current (the specifications are as before except for K
I
=5
and C
f
=15F). The output current of each inverter is
limited at 1.5 times of the rated load. As it can be seen,
the current-limiting component of each inverter limits the
load current by limiting the inductor current when the
added load current feed-forward is at point B, and parallel
system has excellent dynamic performance.
(a)
(b)
Fig.9. Experimental waveforms at full resistive load when K
I
is different
(K
I1
=K
I2
=5, C
f1
=30F C
f2
=15F): (a) strategy #1, and (b) strategy #2.
O
u
t
p
u
t

v
o
l
t
a
g
e

(a)
(b)
C
u
r
r
e
n
t

s
h
a
r
i
n
g

p
r
e
c
i
s
i
o
n

(c)
Fig.10. Measured curves under two strategies when K
I
is different
(K
I1
=K
I2
=5C
f2
=15F): (a) output characteristic, (b) circulating current,
and (c) current sharing precision
3505
Time: [5ms/div]
io2 [10A/div]
io1 [10A/div]
uo [100V/div]
(a)
Time: [5ms/div]
io2 [10A/div]
io1 [10A/div]
uo [100V/div]
(b)
Fig.11 Dynamic current-limiting waveforms of parallel system: (a)
suddenly apply load, and (b) suddenly remove load
VI. CONCLUSIONS
This paper proposes an improved distributed control
strategy based on the average current control method.
Comparing with the conventional method, the control
strategy of the inductor-current feedback combined with
the load current feed-forward is adopted in the novel
technique. The improved strategy greatly improves the
output characteristics of single or parallel inverters and
meanwhile, the functions of the output current limiting
and current sharing effect are kept the same as the
original one.
Comparison of output characteristic and circulating
current between the conventional and improved strategy
is presented in the paper. Moreover, by theoretical and
phasor means, the paper points out two key factors,
namely the difference of K
I
and that of C
f
of inverters
operating in parallel, determine circulating current under
both the two strategies.
ACKNOWLEDGEMENT
This work is supported by Fok Ying Tung Education
Foundation under Award Number 91058 and the Delta
Power Electronics Science and Education Development
Foundation.
REFERENCES
[1] J.F. Chen, C.L. Chu, Combination voltage-controlled and
current-controlled PWM inverters for UPS parallel operation,
IEEE Trans. Power Electron., vol. 10, no. 5, pp. 547-558, Sept.
1995.
[2] H. van der Broeck, U. Boeke, A simple method for parallel
operation of inverters, in Proc. IEEE INTELEC98 Conf., 1998,
pp. 143-150.
[3] W.-C. Lee, T.-K. Lee, S.-H. Lee, K.-H. Kim, D.-S. Hyun, et al.,
A master and slave control strategy for parallel operation of
three-phase UPS systems with different ratings, in Proc. IEEE
APEC04 Conf., 2004, pp. 456-462.
[4] A. Tuladhar, Advanced control techniques for parallel inverter
operation without control interconnections, Dissertation for
Doctor Degree in the University of British Columbia, 2000.
[5] S. J. Chiang, J. M. Chang, Parallel control of the UPS inverters
with frequency-dependent droop scheme, in Proc. IEEE
PESC01, Conf., 2001, pp. 957-961.
[6] J. M. Guerrero, L. Garca de Vicua, J. Matas, J. Miret, and J.
Cruz, Steady-state invariant frequency and amplitude droop
control using adaptive output impedance for parallel-connected
UPS inverters, in Proc. IEEE APEC05, Conf., 2005, pp.
560-566.
[7] K. De Brabandere, B. Bolsens, J. Van den Keybus, A. Woyte, J.
Driesen, et al., A voltage and frequency droop control method
for parallel inverters, IEEE Trans. Power Electron., vol. 22, no.
4, pp. 1107-1115, July. 2007.
[8] Takao Kawabata, Shigenori Higashino, Parallel operation of
voltage source inverters, IEEE Trans. Ind. Applicat., vol. 24, no.
2, pp. 281-287, March/April 1988.
[9] H. Oshima, Y. Miyazawa, and A. Hirata, Parallel redundant UPS
with instantaneous PWM control, in Proc. IEEE INTELEC91,
Conf. 1991, pp. 436-442.
[10] J. Tan, H. Lin, J. Zhang, J. Ying, A novel load sharing control
technique for paralleled inverters, in Proc. IEEE PESC03, Conf.,
2003, pp. 1432-1437.
[11] L. Chen, L. Xiao, C. Gong, Y. Yan, Circulating currents
characteristics analysis and the control strategy of parallel system
based on double close-loop controlled VSI, in Proc. IEEE
PESC04, Conf., 2004, pp. 4791-4797.
[12] Y.-K. Chen, T.-F. Wu, Y.-E. Wu, and C.-P. Ku, A current-sharing
control strategy for paralleled multi-inverter systems using
microprocessor-based robust control, in Proc. IEEE
TENCON01,Conf., 2001, pp. 647-653.
[13] Y.-K Chen, Y.-E Wu, T.-F Wu, and C.-P. Ku, CWDC strategy for
paralleled multi-inverter systems achieving a weighted output
current distribution, in Proc. IEEE APEC02, Conf., 2002, pp.
1018-1022.
[14] Y. Xing, L. Huang, S. Sun, and Y. Yan, Novel control for
redundant parallel UPSs with instantaneous current sharing, in
Proc. IEEE PCC02, Conf., 2002, pp. 959-963.
[15] Y. Xing, L. Huang, and Y. Yan, Redundant parallel control for
current regulated inverters with instantaneous current sharing, in
Proc. IEEE PESC03, Conf., 2003, pp. 1438-1442.
[16] Z. He, Y. Xing, and Y. Hu, Low cost compound current sharing
control for inverters in parallel operation, in Proc. IEEE
PESC04, Conf., 2004, pp. 222-227.

Вам также может понравиться