Sie sind auf Seite 1von 9

1444 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO.

6, JUNE 2009
Design and Implementation of a Robust Current
Controller for VSI Connected to the Grid
Through an LCL Filter
Ivan Jorge Gabe, Vincius Foletto Montagner, and Humberto Pinheiro, Member, IEEE
AbstractThis paper describes the design and implementation
of a discrete controller for grid-connected voltage-source inverters
with an LCL lter usually found in wind power generation systems.
First, a theorem that relates the controllability of the discrete dy-
namic equation of the inverter with LCL lter and the sampling
frequency is derived. Then, a condition to obtain a partial state
feedback controller robust to grid impedance uncertainties and
based on linear matrix inequalities is proposed. This controller
guarantees the stability and damping of the LCL lter resonance
for a large set of grid conditions without requiring self-tuning pro-
cedures. Finally, an internal model controller is added to ensure
asymptotic reference tracking and disturbance rejection, signif-
icantly reducing the impact of grid background voltage distor-
tion on the output currents. Experimental results are presented
to support the theoretical analysis and to demonstrate the system
performance.
Index TermsGrid-connected inverters, LCL lters, linear ma-
trix inequalities, partial state feedback, pulsewidth-modulation
(PWM) converters.
I. INTRODUCTION
W
ITH the recent increase of energy demands and con-
cerns about global warming and greenhouse gas emis-
sions, there is growing interest in the use of renewable energy
sources. Wind power is one of these resources that has under-
gone very fast expansion around the world. This growth has
been mainly supported by advances in technologies of variable
speed generation, enabled by the power electronics converters.
The main advantages of generation with variable speed turbines
include the lower stress on mechanical components, the reduced
acoustical noise, and a higher power capture by the turbine. The
two main concepts of variable-speed wind turbines are full and
partial power control topologies [1]. In both concepts, power
electronics converters contribute to raise the power levels as
Manuscript received September 9, 2008; revised December 17, 2008.
Current version published May 15, 2009. This work was suppored in part
by Coordenac ao de Aperfeicoamento de Pessoal de Nvel Superior (CAPES)
and in part by National Counsel of Technological and Scientic Development
(CNPq). Recommended for publication by Associate Editor M. Ponce-Silva.
I. J. Gabe is with the Power Electronics and Control Research Group
(GEPOC), Federal University of Santa Maria (UFSM), 97105-900 Santa Maria,
Brazil (e-mail: ivangabe@gmail.com).
V. F. Montagner is with the Power Electronics and Control Research Group
(GEPOC), Federal University of Pampa (UNIPAMPA), 97546-550, Alegrete,
Brazil (e-mail: vfmontagner@gmail.com).
H. Pinheiro is with Department of Electrical Energy Conversion (DPEE),
Federal University of Santa Maria (UFSM), 97105-900 Santa Maria, Brazil
(e-mail: humberto@ctlab.ufsm.br).
Color versions of one or more of the gures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identier 10.1109/TPEL.2009.2016097
well as to control the active and reactive power injected into the
grid [2], [3].
In some cases, the utilization of wind resources requires
generation to be located in remote regions, where grids with
high power transference capacity are generally not available [4].
Under these conditions, several issues may be of concern, in-
cluding limited power transference capability [5], thermal re-
strictions [6], and instability of the current controller due to
uncertainty of the grid impedance at the point of common con-
nection (PCC).
The converter topology used in most grid-connected high-
power wind turbines is the pulsewidth-modulation (PWM)
voltage-source inverter (VSI). The use of PWM requires an
output lter to limit the harmonic content of the grid-injected
currents. LCL lters allow for a reduction of the current har-
monic, due to PWM, to acceptable levels with lower inductance
values than can be achieved with Llters. The main limitation of
L lters is the need for high switching frequencies to avoid high
amounts of reactive power circulation. For high-power wind
turbines (WT), in the range of hundreds of kilowatts to a few
megawatt, low switching and sampling frequency are usually
deployed to limit switching losses, and as a result, LCL lters
offer better performance with reduced reactive power consump-
tion when compared with L lters [7], [8].
There are two main issues to be considered in the design
of the current controller for grid-connected converters with an
LCL lter. The rst is the grid background voltage distortion
and the second is the resonance of the LCL lter. PI resonant
controllers in stationary frames [9] or PI controllers in rotating
frames [10] have been considered to address the problemof grid
background distortion in the grid-injected currents. Both imple-
mentations are equivalent as shown in [9]. In this paper, the
stationary frame was selected due to its simplicity. The second
issue is the damping of resonance of the LCL lter. There are
two well-established techniques to damp the resonance of LCL
lters, which are passive and active damping. Passive damp-
ing [7], [11] consists of introducing additional passive elements
on the lter circuit. Generally, resistors in series with the ca-
pacitors are added. Usually, passive damping results in losses
typically around 1% of the converter nominal power [7]. These
losses may be unacceptable in some applications. For instance,
wind turbines operate typically at 30% of their nameplate ca-
pacity, so the relative damping losses become higher. Alter-
natives of passive damping solutions have also been consid-
ered [11]; however, their performances are highly dependent
on the grid-side impedance. Furthermore, additional passive
0885-8993/$25.00 2009 IEEE
GABE et al.: DESIGN AND IMPLEMENTATION OF A ROBUST CURRENT CONTROLLER FOR VSI CONNECTED 1445
Fig. 1. Three-phase inverter with LCL lter connected to the grid.
elements are not desirable due to the increase in hardware size
and cost.
On the other hand, active damping can be achieved when
one includes a digital compensator in the current control loop.
Many papers in the literature address the design of active damp-
ing controllers. In [12], a lter was added on the reference
voltage for the modulator. Genetic algorithms have been used to
tune lter coefcients in order to optimize resonance damping
for a given system conguration. Dissipative-based controllers
for UPS have been developed in terms of phase variables and
transformed into abc coordinates [13], [14]. The resulting con-
trollers are similar to a proportional plus resonant controllers
with voltage reference feedforward and current feedback. Al-
though the dissipative-based approach has been demonstrated
to be a powerful design tool, to extend the results to high-power
grid-connected converters, two issues should be addressed. First,
the output lter should be replaced by an LCL lter and the
grid background voltage should be considered as a disturbance.
Second, to ensure the performance and stability under discretiza-
tion, an addition effort should be made since the controller is
designed in the continuous time domain [15]. In [3], the impact
of grid impedance uncertainties on the LCL lter resonance fre-
quency as well as the stability of the current loop on low- and
high-frequency ranges are addressed in detail. It is suggested
to use a leadlag controller tuned by a startup procedure that
identies the grid impedance at the PCC. A robust multiloop
control algorithm for grid-connected PWM inverters with LCL
lter is proposed in [16]. In that paper, an inner current capac-
itor loop improved the stability margins while a synchronous
frame PI or a P+ resonant controller was used to provide the
desired steady-state performance. However, additional sensors
were needed for its implementation.
So far, a solution that does not need tuning procedures or
adaptive approaches to ensure robustness for grid uncertainties
at the PCC has not yet been described in the literature.
This paper proposes a robust partial state feedback as active
damping for the LCL lter resonance. First, a design restriction
on the sampling frequency that guarantees the controllability
of the equivalent discrete-time system that represents the three-
phase grid-connected inverter with LCL lter is derived. Then,
partial state feedback gains are obtained by the solution of a
linear matrix inequality (LMI) [17] based on a theorem that
assures sufcient conditions for a pole location of the system
even under a large grid impedance uncertainty at the PCC. The
proposed controller does not need additional sensors for its im-
plementation since a partial state feedback is used. Finally, an
internal model controller [18] is included in the control loop to
meet the steady-state specications, which are grid background
voltage disturbance rejection and current reference tracking.
The presented methodology is new to this application. In ad-
dition, its good performance is demonstrated theoretically and
experimentally.
II. CONTROLLABILITY OF THE DISCRETE-TIME SYSTEM
Fig. 1 shows the equivalent circuit of a three-phase three-wire
inverter with an output LCL lter connected to the grid. The grid
impedance is considered a purely inductive reactance. This is
supported by the fact that WTare often connected to distribution
grids, where the systempresents less interaction with residential
urban area loads [3].
The three-phase circuit of Fig. 1 can be transformed into
two single-phase decoupled circuits by the well known abc to
transformation. Each of the single phase circuits can be
represented by a linear time-invariant state-space model of the
form
x(t) = Ax(t) + Bu(t) + F(t)
y(t) = Cx(t) (1)
where x is the normalized state vector selected as
[i
L
1
/I
base
v
C
f
/V
base
i
L
2
/I
base
]
T
, u is the normalized in-
verter output voltage, w is the normalized grid voltage, and
y is the normalized boost inductor current in the reference
frame. A, B, C, and F are matrices with appropriate dimen-
sions given in the Appendix. Note that the impact of the dc bus
voltage (V
cc
) on the loop gain can be eliminated by dividing the
control action u by V
cc
. Therefore, the dc-bus voltage does not
appear explicitly in the dynamic model. Also, for the stability of
the current loop and the LCL resonance damping analysis, the
grid can be represented by background voltage sources behind
inductances [3]. As a result, it is reasonable to assume that the
voltage sources are exogenous input that are not affected by
the converter currents; therefore, initially it is considered that
(t) = 0.
1446 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO. 6, JUNE 2009
TABLE I
500-kW WIND TURBINE PARAMETERS
Now, by considering that the PWM voltages are synthesized
in a DSPcontroller or microcontroller, it is convenient to analyze
and design the system in the discrete-time domain. The repre-
sentation in the discrete-time domain with a sampling period T
s
becomes
x((k + 1)T
s
) = Gx(kT
s
) + Hu(kT
s
)
y(kT
s
) = Cx(kT
s
) (2)
where G and H are given by
G = e
AT
s
H =
_
T
s
0
e
A(T
s
)
B d.
In addition, (2) can be modied to include the delay present
in the discrete controller implementation, which results in
x((k + 1)) = Gx(k) + Hu(k)
y(k) = Cx(k) (3)
where x = [x u
d
]
T
and G, H, and C are given by
G =
_
G H
0 0
_
H = [ 0 0 0 1 ]

C = [ C 0 ] .
Note that u
d
is an additional state variable that has been included
in the state vector to represent the time delay of the digital
implementation.
In order to make the active damping possible, the oscillatory
modes of (1) with a discrete controller and the controllability of
(3), will be investigated. For this purpose, a theorem that relates
the controllability of the discrete system (3) with the parameters
of (1) and the sampling frequency is presented.
Theorem 1: Assume that the dynamic equation given in (1)
is controllable. The necessary and sufcient condition for the
discrete-time dynamic equation given in (3) to be controllable is
that Im[
i
(A)
j
(A)] = 2/T
s
for = 1, 2, . . ., when-
ever Re[
i
(A)
j
(A)] = 0.
Proof: Let (1) represent a single-inputsingle-output (SISO)
linear time invariant (LTI) system, where the matrix A is in
the Jordan canonical form with distinct eigenvalues. If
i
is an
eigenvalue of A, then
1
= e

1
T
S
is an eigenvalue of G. As-
sume that
1
= + j and
2
= j is the pair of complex
eigenvalues of A. Whenever = /T
s
for = 1, 2 . . .,
then
1
=
2
. As a result, G has two Jordan blocks associated
with the same eigenvalue; hence, the pair {G, H} is not con-
trollable [18].
To demonstrate a practical example of Theorem 1, consider
the wind turbine described in [3]. Table I gives the parameters
Fig. 2. State feedback diagram.
of a 500-kW wind turbine, where the grid-side inductance is
assumed as an uncertain parameter belonging to the interval
7.9 H L
g
79 H. (4)
The complex eigenvalues of the matrix A are given by

1,2
= j

L
1
+ L
o
L
1
L
o
C
f
(5)
where L
o
= L
g
+ L
2
.
If sampling period T
s
= 1/5000, then whenever the total grid-
side inductance L
o
reaches 64.6 H the condition Im[
i
(A)

j
(A)] = 2/T
s
is fullled, and accordingly with Theorem 1,
the controllability of the discrete equation (3) is lost. Indeed,
by computing the controllability matrix of the discrete system
(3) that is given by [
H H G H G
2
H G
3
], it is found that
for this value of total grid-side inductance, the controllability
matrix is not full rank, indicating that the discrete system is not
controllable. Therefore, in order to make it possible to damp
oscillatory modes associated to the LCL lter using a discrete
controller, the following inequality must be satised:
f
s
>
1

_
L
1
+ L
o
L
1
L
o
C
(6)
where f
s
= 1/T
s
. Note that the inequality (6) has been obtained
from Theorem 1 and (5) by conning the eigenvalues of (1)
within the primary strip in the s plane, that is, = 1.
One way to guarantee that the controllability will not be lost is
to choose a minimumvalue of L
2
to a given sampling frequency
f
s
. In this way, even with grid impedance uncertainty, Theorem
1 is satised within the whole range of L
g
. For the previous
case, the minimum value of L
2
must be greater than 64.6 H.
Once this condition is fullled, it is possible to actively damp
the oscillatory modes of the LCL lter with a discrete controller.
In the next section, a robust partial state feedback design for the
discrete system (3) is developed.
III. ROBUST PARTIAL STATE FEEDBACK DESIGN
The robust partial state feedback aims to assign the discrete-
time system poles to mitigate the resonance of the LCL lter
in the current control loop given in Fig. 2. Another desired
characteristic of the closed-loop system is that all the poles
remain inside the unit circle for any inductance value inside the
interval dened for the operation of the system from weak grid
conditions to stiff grid conditions. One limitation imposed on
the control design is that (i
L
2
) will not be available. Therefore,
a partial state feedback control law is used. This results in a
GABE et al.: DESIGN AND IMPLEMENTATION OF A ROBUST CURRENT CONTROLLER FOR VSI CONNECTED 1447
reduction in the number of sensors, making the controller design
more challenging.
The LMI conditions presented in [19] and applied to grid-
connected inverters with LCL lters in [20] can assure a robust
pole location for the uncertainty systems under investigation.
Rewriting (1) as a function of L
o
, one has that
x(t) = A(L
o
)x(t) + Bu(t) + F(L
o
)w(t). (7)
The uncertainty in L
g
results in variations in almost all entries
of the matrices G and H of the discrete system, which can be
expressed as
x(k + 1) = G()x(k) + Hu(k) (8)
where each entry of G can be included in a convex polyhedral
set. As a result, G() belongs to the polytope P dened as
P = G() R
44
: G() =
N

i=1

i
G
i
,
N

i=1

i
= 1,
i
0, i = 1, . . . , N (9)
[21], [22].
Note that the representation of (8) is a version of (3) including
the parametric uncertainty (vector ).
Let the partial state feedback control law be represented by
u(k) = Kx(k), K = [ k
11
k
12
0 k
14
] . (10)
The design aim is to nd k
11
, k
12
, and k
14
inside the set
S = [ k
11
k
12
k
14
] R
3
with
k
11
k
11
k
11
k
12
k
12
k
12
e k
14
k
14
k
14
that is dened by the control designer.
The vector of gains K has to assure that the closed-loop
system
x(k + 1) = G
cl
()x(k), G
cl
() = G() + HK, G() P
(11)
is stable and that the eigenvalues of G
cl
() (close-loop poles)
remain inside the circle C, with center d and radius r chosen a
priori by the designer, (placed inside the unit circle), shown in
Fig. 3.
The next theoremprovides a sufcient LMI condition to solve
the problem stated earlier.
Theorem 2: Given that the values of r and d dene a region
in C to the pole location and given that the gains k
11
, k
12
, and
k
14
of the controller belong to S. If there exists a symmetric
positive denite matrix P R
44
such that
_
rP (G
i
+ HK)

P dP
P(G
i
+ HK) dP rP
_
> 0,
i = 1, . . . , N (12)
then G
cl
() is stable and the system eigenvalues belong to C.
Fig. 3. Circle with radius r and center d for the allocation of the closed-loop
poles.
Proof: If Theorem 2 holds, based on Schur complement and
on convexity [21], one has
(G() + HK dI)

r
P
(G() + HK dI)
r
P < 0 (13)
which ensures the stability with the prescribed pole location for
G
cl
() [19].
It is worth mentioning that the condition from Theorem 2 is
a condition for analysis. Given a controller candidate extracted
from set S, the feasibility of Theorem 2 certies that this con-
troller ensures the pole location for the closed-loop systemwhen
the grid-equivalent inductance on the PCC belongs to . Some
convex conditions for synthesis of partial state feedback are
available in the literature, based on structural constraints on ma-
trix variables of the problem (see, for instance, [19]). However,
such conditions based on quadratic stability (with and without
slack variables) were tested for the problem under investigation
here not providing a controller in the entire R
3
. On the other
hand, Theorem 2 allows more than one solution for the prob-
lem, as it is tool for validation of robust stabilizing controllers
with guaranteed performance in terms of pole location for the
application under consideration.
Let us again consider the 500-kW WT system described in
Table I. Fig. 4(a) shows the eigenvalue location of the system
without partial state feedback. The bold line indicates how the
undamped eigenvalues of (3) change from weak to stiff grid.
The points 1 and 3 indicated by show the undamped eigen-
values for stiff grid and 2 and 4 indicated by show the eigen-
values for weak grid. The eigenvalue associated with the time
delay is indicated by 5 and the third eigenvalue associated with
the LCL lter is indicated by 6. Fig. 4(b) shows that all the
eigenvalues are well damped when the grid inductance varies
in a range between 0.01 per unit (p.u.) and 0.1 p.u. The feed-
back gains obtained in the LMI framework for this example are
K = [ 1.78 1.10 0 0.50 ] for one circle centered in d = 0
and a radius r = 0.95.
In the next section, an internal model controller is added to
the system and a design procedure to the overall system.
IV. STEADY-STATE CONTROLLER DESIGN
In the proposed design procedure, the LMI conditions in
Section III ensure robust (with respect to the grid impedance)
1448 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO. 6, JUNE 2009
Fig. 4. Eigenvalue location of the system given in Table I. (a) Eigenvalue
location without state feedback under stiff and weak grid condition.
(b) Eigenvalue location with state feedback under stiff and weak grid
conditions.
Fig. 5. Complete block diagramof the closed-loop system with the stabilizing
partial state feedback controller and the internal model steady-state controller.
asymptotic stability for the closed-loop system but are not suit-
able to cope with tracking of sinusoidal reference signals. Then,
the resonant controller is used to ensure tracking, based on the
internal model principle [9], [23]. This is equivalent to the pro-
portional + resonant structure considered in [3], [9]. The con-
troller structure used to eliminate the grid background voltage
is given by (14), where i is the compensated harmonic order
G
c
(z) =
N(z)
(z)
= Z
_

k
s
s
2
+ (i)
2
_
. (14)
The proposed current control loop is shown in Fig. 5.
The gain k determines the performance of the internal model
controller, which means how fast the internal model controller
tracks the reference and/or rejects the disturbance. A single
gain k is considered for all resonant controllers. This simplies
Fig. 6. Root locus of the closed-loop system assuming G
cl
(z) = 1.
signicantly the design; however, if no satisfactory result is
found, then resonant controllers with different gains give an
additional degree of freedom. The reference current i
L
1
ref
can
be obtained, for example, from the capacitor voltage positive
sequence. For more details see [24] and [25].
V. INTERACTION BETWEEN THE CONTROLLERS
The partial feedback gains belonging to set S were interac-
tively tested in the LMI condition of Theorem 2. Thus, it is
likely that more than one vector K allocates the closed-loop
poles inside the circle centered in d and with radius r speci-
ed a priori. Assume Q to be the set of all vectors of gains
K = [k
11
k
12
0 k
14
], with [k
11
k
12
k
14
] S that sat-
isfy the LMI condition (12). The problem now is to select one
of these vectors of Qthat even with the inclusion of the internal
model controllers does not compromise pole allocation.
In order to gain insight into this problem, the impact of the
closed-loop poles associated with G
c
(z) in G
cl
(z) and vice
versa will be considered. It is also important to consider in this
analysis the closed-loop zeros, that is, the zeros of G
cl
(z).
Let us assume initially that G
cl
(z) does not have dynamics,
that is, G
cl
(z) = 1. In this case, it is possible to nd gains k of
the resonant controller such that the closed-loop system stays
stable. Note that the departure angle of the root locus of the
controller complex poles are pointing inside the unit circle, as
shown in Fig. 6. However, G
cl
(z) = 1. This means that for the
design of the internal model controller gain k, as well as to
select the vector K from the set Q, the angular contribution of
G
cl
(z) in G
c
(z) should be considered. It is reasonable to assume
that in the frequency around the internal model controller poles,
the phase of G
cl
(z) is decreasing with the frequency. As a
result, ensuring that the departure angles of the root locus of the
complex poles point toward the interior of the unit circle for all
grid conditions, it is possible to guarantee the stability of the
closed-loop poles associated with the internal model controllers
for some gain k.
The departure angle of the root locus of the complex
conjugate poles of the internal model controllers is given by

p
= + (15)
GABE et al.: DESIGN AND IMPLEMENTATION OF A ROBUST CURRENT CONTROLLER FOR VSI CONNECTED 1449
TABLE II
SETUP PARAMETERS
where can be expressed as
=

G
c
(z)(z e
j(i)T
s
)G
cl
(z). (16)
To guarantee that the departure angle points to inside the unit
circle,
p
must belong to the interval given by
_

(e
j(i)T
s
) +

2
_
<
p
<
_

(e
j(i)T
s
) +
3
2
_
. (17)
The next step is to derive a criterion for the selection of a
vector of gain K from Q. The selected K will be the one that
optimize the angular difference
d
, which is given by

d
= min{
1
,
2
} (18)

1
=
_

p

_

(e
j(i)T
s
) +

2
_

_
(19)

2
=
_

p

_

(e
j(i)T
s
) +
3
2
_

_
. (20)
Table II shows the setup parameters used in the experimental
results. Considering this system, the resulting gain vector K is
given by K = [ 1.8 0.9 0 0.4 ]. For a stiff grid condition,
the angular margin is
d
= 36.05

, the middle of the impedance


range results in
d
= 29, 22

, and for a weak grid condition,


the angular margin is
d
= 22

. Note that in this case, the weak


grid condition represents the worst case in terms of stability.
Now, the impact of the internal model controllers on the
lter modes will be investigated. Fig. 7(a) shows the eigenvalue
location of G
cl
(z), while Fig. 7(b) shows the eigenvalue location
of the system including the internal model controller poles.
For this system, it is possible to see that the inclusion of the
internal model controller does not signicantly affect the lter
pole location within the considered inductance interval. Fig. 8(a)
shows the systemroot locus under stiff grid conditions for k > 0.
On the other hand, in Fig. 8(b), the same root locus is drawn for
a weak grid condition. It is possible to see that under the weak
grid condition, the angular margin limits the maximum internal
model controller gain to k = 1.54. This gain is normalized with
k = 250, in this way. To avoid instability in all grid conditions,
the internal model controller gain k must be less than 250
1.54 = 385, or k < 385.
VI. EXPERIMENTAL RESULTS
The experimental results were obtained using a three-
phase grid-connected inverter of 11 kVA controlled by a DSP
Fig. 7. Eigenvalue location for the system given in Table II. (a) Eigenvalue
location of G
cl
(z) to stiff and weak grid conditions. (b) Eigenvalue location
including the internal model controllers in the fundamental, fth, and seventh
harmonics.
Fig. 8. Root locus diagram for a gain k
1, 5, 7
= 250 and K = [1.8 0.9
0 0.4]. (a) Root locus for stiff grid condition. (b) Root locus in weak grid
conditions.
1450 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO. 6, JUNE 2009
Fig. 9. Inductor L
1
current with different L
o
values. (a) Transient response
on the current of L
1
with L
o
= 750 H, (within design interval). (b) Current
in L
1
to L
g
= 1000 H (value above the design interval).
TMS320F2812 with the parameters given in Table II. The grid
impedance interval is assumed in a range from L
o
= 0 under
stiff grid conditions, that is, S
k
= to L
o
= 880 H under a
weak grid condition that results in a short-circuit power at the
PCC of S
k
= 15. The lter parameters were chosen to fulll the
controllability conditions stated by (6).
The feedback gains obtained under the LMI framework are
K = [ 1.8 0.9 0 0.4 ] . (21)
Gain k of the internal model controllers is k = 250.
To demonstrate the systembehavior on stability limits, the l-
ter output is short-circuited, and the total grid-side inductance L
o
is changed. Fig. 9(a) shows the L
1
current when L
o
= 750 H,
which is within the assured stability interval. A step in cur-
rent reference from 5 to 25 A demonstrates the system dy-
namic response. On the other hand, a weak grid condition, with
L
o
> 925 H, which corresponds to S
k
= 13 is considered in
Fig. 9(b), which shows a low-frequency oscillation associated
with the seventh harmonic resonant controller, as predicted by
the root locus analysis. Values of L
o
lower than the minimum
value of the stability interval can lead to system instability as
Fig. 10. Filter currents in full-power operation. (a) Three-phase currents in the
L
1
inductors. (b) Spectrum of the current in phase a. (c) Three-phase currents
in the L
2
inductor. (d) Spectrum of the current in phase a.
well. However, this case will never be reached, since the mini-
mum value of L
o
is ensured by the appropriate selection of L
2
.
These results show that the system behaves as predicted by the
theoretical analysis.
Fig. 10 shows the main waveforms of the PWM inverter with
the LCL lter operating at full power connected to the grid.
GABE et al.: DESIGN AND IMPLEMENTATION OF A ROBUST CURRENT CONTROLLER FOR VSI CONNECTED 1451
Fig. 11. Filter states and grid voltage.
In this case, a lter with lower p.u. values of inductances and
capacitance is considered to demonstrate the generality of the
proposed design procedure, as well as the performance at full
power. In this case, the lter has been designed to meet the
current harmonic requirements of the IEEE 1547 standard. The
current controller parameters have been obtained as described
in Sections II, IV, and V. Fig. 10(a) shows the boost inductors
currents. Note that a higher harmonic content appears as a result
of a lower value of inductance. Fig. 10(b) shows the i
L
1
a
spec-
trum, where the rst harmonic group appears at the switching
frequency f
c
. Fig. 10(c) shows the i
L
2
currents and Fig. 10(d)
shows the current spectrum. Note that the lter reduces the out-
put harmonics to meet the standard requirements. These results
demonstrate that the proposed control is feasible using different
lter sizes.
In Fig. 11, it is possible to see that the output current and the
grid voltage are in phase resulting in a power factor close to
unity.
VII. CONCLUSION
This paper has proposed the design of a robust current con-
troller for grid-connected converters with an LCL lter. In ad-
dition, it demonstrates how to obtain the feedback gains for a
given interval of grid inductance that assure the active damping
of the LCL lter oscillatory modes. Consequently, the controller
does not need self-tuning or adaptive approaches in the case
of grid impedance uncertainty. Moreover, low-frequency con-
trollers are included to assure steady state performance. Further-
more, a procedure for controller design, which avoids instability
under grid impedance uncertainty, is described. Two sets of ex-
perimental results are reported. The rst one demonstrates that
the experimental stability limits are in very good agreement
with the theoretical predictions. Second, results in full-power
operation are presented to demonstrate the proposed control
performance.
APPENDIX
MATRIXES
The matrices of the continuous- and discrete-state space mod-
els of the three-phase grid-connected inverters with LCL lter
in coordinates are given by
A =
_

_
0
1
L
1
V
base
I
base
0
1
C
f
I
base
V
base
0
1
C
f
I
base
V
base
0
1
L
o
V
base
I
base
0
_

_
B =
_

_
1
L
1
V
base
I
base
0
0
_

_
C = [ 1 0 0 ]
F =
_

_
0
0

1
L
o
V
base
I
base
_

_
.
REFERENCES
[1] F. Blaabjerg, Z. Chen, and S. Kjaer, Power electronics as efcient inter-
face in dispersed power generation systems, IEEE Trans. Power Elec-
tron., vol. 19, no. 5, pp. 11841194, Sep. 2004.
[2] F. Blaabjerg, Wind power-a power source nowenabled by power electron-
ics, in Proc. 9th Brazilian Power Elecron. Conf., 2007, vol. 1, pp. 116.
[3] M. Liserre, R. Teodorescu, and F. Blaabjerg, Stability of photovoltaic and
wind turbine grid-connected inverters for a large set of grid impedance
values, IEEE Trans. Power Electron., vol. 21, no. 1, pp. 263272, Jan.
2006.
[4] H. Bindner and P. Lundsager, Integration of wind power in the power
system, in Proc. 28th Annu. IEEE IECON, Nov. 2002, vol. 4, pp. 3309
3316.
[5] G. Ledwich and H. Sharma, Connection of inverters to a weak grid, in
Proc. 31st Annu. IEEE PESC, 2000, vol. 2, pp. 10181022.
[6] J. Tande and K. Uhlen, Wind turbines in weak grids-constraints and
solutions, in Proc. 16th Int. Inst. Electr. Eng. Conf. Exhib. Elect. Distrib.
2001 Part 1: Contrib. CIRED, vol. 4, pp. 1821.
[7] M. Liserre, F. Blaabjerg, and S. Hansen, Design and control of an LCL-
lter-based three-phase active rectier, IEEE Trans. Ind. Appl., vol. 41,
no. 5, pp. 12811291, Sep./Oct. 2005.
[8] M. Lindgren and J. Svensson, Control of a voltage-source converter
connected to the grid through an LCL-lter-application to active ltering,
in Proc. IEEE PESC, May 1998, vol. 1, pp. 229235.
[9] D. Zmood and D. Holmes, Stationary frame current regulation of pwm
inverters with zero steady-state error, IEEE Trans. Power Electron.,
vol. 18, no. 3, pp. 814822, May 2003.
[10] M. Liserre, R. Teodorescu, and F. Blaabjerg, Multiple harmonics control
for three-phase grid converter systems with the use of pi-res current con-
troller in a rotating frame, IEEE Trans. Power Electron., vol. 21, no. 3,
pp. 836841, May 2006.
[11] T. Wang, Z. Ye, G. Sinha, and X. Yuan, Output lter design for a grid-
interconnected three-phase inverter, in Proc. IEEE PESC, 2003, vol. 2,
pp. 779784.
[12] M. Liserre, A. DellAquila, and F. Blaabjerg, Genetic algorithm-based
design of the active damping for an LCL-lter three-phase active rectier,
IEEE Trans. Power Electron., vol. 19, no. 1, pp. 7686, Jan. 2004.
[13] G. Valderrama, A. Stankovic, and P. Mattavelli, Dissipativity-based adap-
tive and robust control of ups in unbalanced operation, IEEE Trans.
Power Electron., vol. 18, no. 4, pp. 10561062, Jul. 2003.
[14] G. Escobar, P. Mattavelli, A. Stankovic, A. Valdez, and J. Leyva-Ramos,
An adaptive control for ups to compensate unbalance and harmonic
distortion using a combined capacitor/load current sensing, IEEE Trans.
Ind. Electron., vol. 54, no. 2, pp. 839847, Apr. 2007.
[15] R. Costa-Castello and E. Fossas, On preserving passivity in sampled-
data linear systems, in Proc. Amer. Control Conf., Jun. 2006, vol. 1,
pp. 43744378.
1452 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO. 6, JUNE 2009
[16] E. Twining and D. Holmes, Grid current regulation of a three-phase volt-
age source inverter with an lcl input lter, IEEE Trans. Power Electron.,
vol. 18, no. 3, pp. 888895, May 2003.
[17] S. Boyd, L. El Ghaoui, E. Feron, and V. Balakrishnan, Linear Matrix
Inequalities in System and Control Theory. Philadelphia, PA: SIAM,
1994.
[18] C. T. Chen, Linear System Theory and Design. New York, NY: Holt,
Rinehart and Winston, 1984.
[19] V. Montagner, V. Leite, and P. Peres, Discrete-time switched systems:
pole location and structural constrained control, in Proc. 42nd IEEEConf.
Decis. Control 2003, Dec., vol. 6, pp. 62426247.
[20] I. Gabe, J. Massing, V. Montagner, and H. Pinheiro, Stability analysis
of grid-connected voltage source inverters with LCL-lters using partial
state feedback, in Proc. Eur. Conf. Power Electron. Appl., Sep. 2007,
pp. 110.
[21] S. Boyd, L. E. Ghaoui, E. Feron, and V. Balakrishnan, Linear Matrix
Inequalities in System and Control Theory. Philadelphia, PA: SIAM,
1994.
[22] P. Gahinet, A. Nemirowskii, A. J. Laub, and M. Chilali, LMI Control
Toolbox Users Guide. Natick, MA: The MathWorks Inc., 1995.
[23] F. Botteron and H. Pinheiro, Discrete-time internal model controller
for three-phase pwm inverters with insulator transformer, in Proc. Inst.
Electr. Eng.Electr. Power Appl., Jan. 2006, vol. 153, no. 1, pp. 5767.
[24] R. F. de Camargo and H. Pinheiro, Synchronisation method for three-
phase PWM converters under unbalanced and distorted grid, Proc. Inst.
Electr. Eng.Electr. Power Appl., vol. 153, pp. 763772, Sep. 2006.
[25] P. Rodriguez, R. Teodorescu, I. Candela, A. Timbus, M. Liserre, and
F. Blaabjerg, New positive-sequence voltage detector for grid synchro-
nization of power converters under faulty grid conditions, in Proc. 37th
IEEE PESC, Jun. 2006, pp. 17.
IvanJorge Gabe was born in Ibirub a, Brazil, in 1983.
He received the B.S. degree in electrical engineering
in 2006 and the Master degree in 2008 from the Fed-
eral University of Santa Maria (UFSM), Santa Maria,
Brazil, where he is currently working toward the Ph.D
degree.
His current research interests include wind power
generation and grid-interconnected converters.
Vincius Foletto Montagner received the Ph.D. and
the Postdoctoral degrees in electrical engineering
from the University of Campinas, Campinas, Brazil,
in 2005 and 2006, respectively.
Currently, he is a Professor at the Federal Univer-
sity of Pampa, Santa Maria, Brazil, and a Researcher
with the Power Electronics and Control Research
Group. His current research interests include con-
trol theory and applications.
Humberto Pinheiro (M90) was born in Santa
Maria, Brazil, in 1960. He received the B.S. degree
from the Federal University of Santa Maria, Santa
Maria, Brazil, in 1983, the M.Eng. degree from the
Federal University of Santa Catarina, Florian opolis,
Brazil, in 1987, and the Ph.D. degree fromConcordia
University, Montreal, QC, Canada, in 1999.
From 1987 to 1990, he was a Research Engineer
with a Brazilian UPS company and then joined the
Pontifcia Universidade Cat olica do Rio Grande do
Sul, Brazil, where he lectured on power electronics.
Since 1991, he has been with Federal University of Santa Maria. His current
research interests include uninterruptible power supplies, wind power systems,
and control applied to power electronics.
Dr. Pinheiro is currently a member in the IEEE Power Electronics Society.

Das könnte Ihnen auch gefallen