IEEE SYSTEMS JOURNAL, VOL. 6, NO.

1, MARCH 2012 85
A Feedback Linearization Control Scheme for the
Integration of Wind Energy Conversion Systems
into Distribution Grids
Federico Delno, Member, IEEE, Fabio Pampararo, Renato Procopio, Member, IEEE, and Mansueto Rossi
AbstractThis paper focuses on the development of a control
strategy for integration of wind energy conversion systems
(WECS) into the electrical distribution networks with particular
attention to the combined provision of energy and ancillary
services. Typically, a WECS is composed by a variable speed wind
turbine coupled with a direct driven permanent magnet (DDPM)
synchronous generator. This conguration offers a considerable
exibility in design and operation of the power unit, as its output
is delivered to the grid through a fully controlled frequency
converter. Here, a new control scheme to regulate electrical
and mechanical quantities of such generation unit is proposed,
aimed both at reaching optimal performances in terms of power
delivered to the grid and at providing the voltage support
ancillary service at the point of common coupling. The control
scheme is derived resorting to the feedback linearization (FBL)
technique, which allows both decoupling and linearization of
a non linear multiple input multiple output system. Several
numerical simulations are then performed in order to show how
the exibility of the DDPM wind generator can be fully exploited,
thanks to the use of the FBL approach, which assures indepen-
dent control of each variable and signicant simplications in
controller synthesis and system operation, thus making it easier
to integrate WECS into modern day smart grids.
Index TermsDirect drive, feedback linearization, permanent
magnet synchronous generator, wind power.
I. Introduction
T
HE INCREASED penetration of renewable energy re-
sources into the power delivery infrastructures is an effect
of the adoption at world level of governmental policies aimed
at facing climate changes and at designing a long term energy
scenario where power supply systems are technologically
improved to become more efcient and more friendly toward
environmental issues [1][6].
Among the different renewable production technologies,
wind power is the most promising for large scale applications,
due to its investment costs, which are lower than the costs for
power produced from other renewable systems, and thanks to
the growing technological progress, which is making available
Manuscript received September 15, 2010; revised February 25, 2011;
accepted May 25, 2011. Date of publication September 1, 2011; date of current
version February 23, 2012.
The authors are with the Department of Naval and Electrical
Engineering, University of Genoa, Genova I-16145, Italy (e-mail: federico.
delno@unige.it; fabio.pampararo@unige.it; renato.procopio@unige.it;
mansueto.rossi@unige.it).
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/JSYST.2011.2163002
wind turbine for large-sized power plants (on shore and off-
shore) and, at the same time, compact and low-speed triggered
systems, suitable for residential uses [7].
However, many utilities are reluctant to install signicant
wind capacity because of the intermittent nature of the re-
source, which prevents wind power plants to be controlled
in the same way as conventional fossil fuel bulk units.
Thus, research in the eld is presently very focused on the
development of suitable control strategies [8], [9], to improve
wind power integration into distribution networks and to
regulate it into a dispatchable resource. Among the different
kinds of wind generation systems, those based on direct driven
permanent magnet (DDPM) are gaining attention and growing
fast in terms of new installations, as it is expected that current
developments in gearless energy transmission with power-
electronic grid interfaces will lead to a new generation of quiet,
efcient, and economical wind turbines. In these systems,
optimum wind energy extraction is achieved by operating the
wind turbine generator in variable-speed/variable-frequency
mode, according to the wind speed, by maintaining the tip
speed ratio to the value that maximizes aerodynamic efciency.
The DDPM machine, jointly with new power-electronic
controls, has thus become an important and very employed
actor in the distributed generation world; it allows improv-
ing reliability of the wind mill and reducing maintenance
expenses, since its gearbox can be eliminated.
In the light of this state-of-the-art, the aim of this paper is
to present a new control scheme for integrating WECS into
the power grids. In particular, the following issues are faced:
1) optimal WECS performances, i.e., tracking of the max-
imum power extraction from the wind turbine;
2) optimal operating conditions of the permanent magnet
synchronous generator (PMSG), which works at mini-
mum stator current;
3) DC link voltage control;
4) control of the reactive power injection into the power
grid, in order to perform a voltage support at the point
of common coupling (PCC);
5) voltage support during grid fault conditions.
Such goals can be achieved by applying feedback lin-
earization (FBL) technique [10][14] at the control synthesis
stage, taking advantage of some peculiarities characterizing
such a technique, as the possibility to reduce a nonlinear
1932-8184/$26.00 c 2011 IEEE
86 IEEE SYSTEMS JOURNAL, VOL. 6, NO. 1, MARCH 2012
multiple-inputs, multiple-outputs system to a number of single
input, single output linear systems, without resorting to math-
ematical approximations (like the small signal analysis); this
makes it easier to design the regulators transfer functions and
allows a decoupled (independent) control of each regulated
variable.
This paper is organized in six parts. In Section II the whole
system model is derived, considering separately wind turbine,
synchronous generator, machine side pulse-width modulation
(PWM) inverter, DC link, network side PWM inverter and
grid, while in Section III the theory necessary to apply the
FBL and to synthesize the control is outlined. The next two
sections present simulations results aimed at highlighting the
system behavior both in normal operating conditions and in
the event of faults. In particular, Section IV is devoted to a
simulation campaign, performed in order to test the WECS
response to wind speed uctuations and reference variations,
showing also its ability to deliver reactive power to the grid
when the generated active power is low. Beside this latter
kind of ancillary services, utilities and transmission system
operators (TSOs) are becoming more demanding in terms of
requirements the system must full during low voltage and
fault conditions [15][17]; in order to assess the performances
of the proposed control scheme also when these contingencies
occur, in Section V the fault ride through capacity of the
system is tested using as a reference a voltage drop prole
of the kind prescribed by E.ON Netz GmbH [16]. Finally, in
Section VI some conclusions are drawn.
II. Wind Energy Conversion System
The considered WECS consists of a permanent magnet
synchronous generator driven by a wind turbine, a machine
side PWM inverter, an intermediate DC circuit, and a grid
side PWM inverter, as indicated in Fig. 1. In the following,
the mathematical models used to represent each one of these
components will be outlined.
A. Wind Turbine Model
The output mechanical power of the wind turbine is given
by [9]
P =
1
2
AV
3
w
c
p
(, ) (1)
where (kg/m
3
) is the air density, A (m
2
) is the area swept
by the blades, V
w
(m/s) is wind speed, c
p
is the performance
coefcient, () is the pitch angle of turbine blades, and is
the tip speed ratio expressed as follows:
=
2.237 V
w

mec
(2)
being
mec
(rad/s) the mechanical rotor speed.
The performance coefcient c
p
has been approximated
using the following nonlinear function:
c
p
=
1
2
_
0.022
2
5.6
_
e
0.17
. (3)
When the turbine operates with = 0, for each instanta-
neous wind speed,
mec
must be set to the value corresponding
to the maximum active power extraction from the wind. The
value of that maximizes the coefcient c
p
for each pitch
angle is obtained by solving
c
p

=
1
2
_
1.952 0.17

e
0.17
= 0. (4)
Once (4) is solved in the variable ,
opt
= 11.48, one can in-
sert the solution into [2] and then compute the optimal machine
angular frequency
opt
by using the following expression:

opt
=
2.237 V
w
p
11.48
(5)
where p is the pole pair of the generator (
opt
= p
mec
).
If the active power or the mechanical speed becomes greater
than their rated value, the pitch control mode (#0) is activated
acting as a limiter for those quantities.
B. PMSG Model
The PMSG dynamic equations are expressed in the d-q
reference frame. The model of electrical dynamics in terms
of the machine d-q axes voltages and currents is
v

md
= R
s
i
md
+ L
sd
i

md
+
e
L
sq
i
mq
(6)
v

mq
= R
s
i
mq
+ L
sq
i

mq

e
L
sd
i
md

e

p
(7)
while the model for the mechanical dynamic is
C
e
C
m
=
J
p

e
(8)
where the electromagnetic torque C
e
can be expressed as
C
e
= p
__
L
sq
L
sd
_
i
md
i
mq
i
mq

(9)
and the wind turbine torque C
m
is dened as
C
m
= P
_

mec
. (10)
The quantities appearing in the previous relationships are
here dened as J: moment of inertia,
p
: main ux, Ls
d
,
L
sq
: d-q synchronous inductances, and R
s
: stator resistance.
C. PWM Inverter Machine Side
For the machine side PWM converter, which acts as rectier
to connect the machine to the DC link, we used a rst
harmonic model, expressed in the d-q reference frame of
the permanent magnet generator
v
md
=
_
3/2m
a
/2 V
mDC
cos (11)
v
mq
=
_
3/2m
a
/2V
mDC
sin . (12)
The converter d-q axes voltages v
md
, v
mq
differ from that
of the machine (v

md
, v

mq
) because of the presence of the
parameters L
fm
and R
fm
(see Fig. 1), modeling a series lter.
In (11) and (12), m
a
and are the converter modulation
index and phase. Transforming these parameters in rectangular
coordinates, we can dene U
md
and U
mq
as follows:
U
md
= m
a
cos (13)
U
mq
= m
a
sin . (14)
DELFINO et al.: A FEEDBACK LINEARIZATION CONTROL SCHEME FOR THE INTEGRATION OF WIND ENERGY CONVERSION SYSTEMS 87
Fig. 1. Wind energy conversion system.
Equations (11) and (12) do not take into account the inverter
losses; a simple way to consider these losses in the rst
harmonic model is to add an equivalent resistance, whose
value is heuristically set to four thirds of the static switch
resistance, R
pwm
[18]. This way, and using the denitions (13)
and (14), (11) and (12) have to be modied as follows:
v
md
=
_
3/2V
mDC
/2 U
md
4/3R
pwm
i
md
(15)
v
mq
=
_
3/2V
mDC
/2 U
mq
4/3R
pwm
i
mq
. (16)
D. DC Link
The DC link electrical behavior is described simply by the
Kirchhoffs laws (Fig. 1)
V
DC
= R
mDC
i
mDC
+ L
mDC
i

mDC
+ V
mDC
(17)
V
DC
= R
nDC
i
nDC
+ L
nDC
i

nDC
+ V
nDC
(18)
i
mDC
+ i
nDC
+ C
DC
V

DC
= 0. (19)
E. Grid Side PWM Inverter
The equations for the grid side inverter are identical to those
of the machine side inverter
v
nd
=
_
3/2V
nDC
/2 U
nd
4/3R
pwm
i
nd
(20)
v
nq
=
_
3/2V
nDC
/2 U
nq
4/3R
pwm
i
nq
(21)
where U
nd
, U
nq
, and R
pwm
have the same meaning of the
analogous quantities already dened for the previous converter.
This time, the d-q reference frame adopted for the AC side
converter phase voltages and currents rotates at the network
angular frequency.
F. Grid Model
The system is supposed to be connected to an innite bus
through a resistance R = R
tr

LV/MV
+R
fn
+R
l
+R
tr

MV/HV
and an
inductance L = L
tr

LV/MV
+L
fn
+L
l
+L
tr

MV/HV
modeling two
(LV/MV and MV/HV) transformers longitudinal parameters,
a series lter and a MV line parameters (as seen from the
transformer low voltage side), respectively. With the equivalent
resistance R
pwm
also included in R, the link between grid axes
voltages v
gridd
, v
gridq
, and the quantities dened in the previous
section is the following:
v
nd
= Ri
nd
+ Li

nd
+ Li
nq
+ v
gridd
=
_
3/2V
nDC
/2 U
nd
(22)
v
nq
= Ri
nq
+ Li

nq
Li
nd
+ v
gridq
=
_
3/2V
nDC
/2 U
nq
. (23)
The d-q reference frame has been chosen so that v
gridq
= 0,
while v
gridd
has been set to the rate value; this way, the active
and reactive power delivered to the grid becomes
P
n
= v
gridd
i
nd
(24)
Q
n
= v
gridd
i
nq
. (25)
III. Control System
In order to apply the FBL, the equations representing the
system have to be rewritten in the form [10]
_
x

= f (x) + g (x) u
y = h(x)
(26)
where u is the input variables vector, x is the vector of the
system state variables, and y the output variables vector.
Based on (6), (7), (8), (19), (22), and (23), and neglecting
the DC link series parameters (thus V
mDC
= V
DC
= V
nDC
and
the DC link model reduces to a power balance between the
converters and the DC link capacitor), the system state vector
can be identied as follows:
x =
_
i
md
i
mq

m
V
DC
i
nd
i
nq

T
(27)
while the system input variables are represented, in our case,
by the converter parameters U
md
, U
mq
, U
nd
and U
nq
u =
_
U
md
U
mq
U
nd
U
nq

T
. (28)
As a consequence, considering the equations derived in the
previous sections, and dening
_
L
md
= L
sd
+ L
fm
L
mq
= L
sq
+ L
fm
(29)
R
m
= R
s
+ R
fm
+
_
4
_
3
_
R
pwm
. (30)
88 IEEE SYSTEMS JOURNAL, VOL. 6, NO. 1, MARCH 2012
f(x) and g(x) can be written in the following forms:
f (x) =

_
R
m
i
md

m
L
mq
i
mq
_
/L
md
_
R
m
i
mq
+
m
L
md
i
md
+
m

p
_
/L
mq
p
2
__
L
sq
L
sd
_
i
md
i
mq
i
mq

/J pC
m
/J
0
_
Ri
nd
Li
nq
v
pccd
_
/L
_
Ri
nq
+ Li
nd
_
/L

(31)
g (x) =

3
2

V
DC
/L
md
0 0 0
0 V
DC
/L
mq
0 0
0 0 0 0
i
md
/C
DC
i
mq
/C
DC
i
nd
/C
DC
i
nq
/C
DC
0 0 V
DC
/L 0
0 0 0 V
DC
/L

.
(32)
As system output variables, we chose the following quantities:
h(x) =
_

m
i
md
i
nq
V
DC

T
. (33)
This way, we can directly control the machine angular fre-
quency (in order to set its value to the optimal one), the
DC link voltage and the reactive power delivered to the grid
[according to (25), by regulating i
nq
], while, by acting on i
md
,
we can minimize the machine current.
The goal of the feedback linearization is to obtain a trans-
formed system whose states are the output variables and, even-
tually, some of their derivatives. This structure is achieved by
considering each one of the outputs and, repeatedly, applying
to it the Lie derivatives [10] along the vector f and along the
vector gu, until at least one input variable appears in the latter;
as an example, for the output h
1
=
m
we have to compute
_
L
f,h
1
= h
1
, f = [0, 0, 1, 0, 0, 0] , f
L
g,h
1
= h
1
, gu = [0, 0, 1, 0, 0, 0] , gu = 0
(34)
where the symbol , represents the scalar product and (.)
is the gradient with respect to state variables.
Since in the last expression none of the inputs appear
explicitly (actually, the expression is identically zero), we have
to apply the Lie derivatives once again, this time to L
f,h1

L
2
f,h
1
=

L
f,h
1
, f
_
=
=
p
2
J
__
L
sq
L
sd
_
i
mq
,
_
L
sq

L
sd
_
i
md

p
,
C

(
m
)
p
, 0, 0, 0
_
, f
_
L
g
L
f
h
1
=

L
f,h
1
, gu
_
=
=

3p
2
V
DC
J

8
_
_
L
sq
L
sd
_
i
mq
U
md
L
md
+
_
L
sq
L
sd
_
i
md
U
mq
L
mq

p
U
mq
L
mq
_
.
(35)
As can be seen, now the inputs u
1
and u
2
(i.e., U
md
and U
mq
)
appear in the last expression. For the output h
2
= i
md

L
f,h
2
(x) = h
2
, f = [1, 0, 0, 0, 0, 0] , f
L
g,h
2
(x) = h
2
, gu =
_
3
8
_
V
DC
U
md
L
md
_
.
(36)
So, for this output we can stop at the rst order derivative.
Fig. 2. Block diagram of the equivalent linearized system and of the
regulators (the symbol indicates the reference values).
In a similar way, for the control output h
3
= i
nq
_
L
f,h
3
(x) = h
3
, f = [0, 0, 0, 0, 0, 1] , f
L
g,h
3
(x) = h
3
, gu =

3/8V
DC
U
mq
/L
(37)
and for the control output h
4
= V
DC

L
f,h
4
(x) = h
4
, f = 0
L
g,h4
(x) = h
4
, gu =
=

3/8
_
U
md
i
md
+ U
mq
i
mq
+ U
nd
i
nd
+ U
nq
i
nq
_
/C
DC
.
(38)
Taking into account the Lie derivatives denition and (26),
from the previous results it follows that [10]

m
= L
2
f,h
1
(x) + L
g
L
f
h
1
(x, u)
i

md
= L
f,h2
(x) + L
g
h
2
(x, u)
i

nq
= L
f,h3
(x) + L
g
h
3
(x, u)
V

DC
= L
f,h
4
(x) + L
g
h
4
(x, u) .
(39)
Now, if we dene a set of ctitious input variables V
i
(i =
1, ..., 4) so that
_
V
1
= L
2
f,h
1
(x) + L
g
L
f
h
1
(x, u)
V
i
= L
f,h
i
(x) + L
g
h
i
(x, u) , i = 2, ..., 4
(40)
by combining (39) and (40) (i.e., by equating the two l.h.s.),
it is apparent that, as anticipated, the system has been re-
formulated in terms of the new state variable
m
and

m
,
i
md
, i
nq
, and V
DC
, with
m
(and

m
) depending (linearly)
only on the ctitious input V
1
, i
md
on V
2
and so on (the
system has also a sixth new state variable, corresponding to an
internal dynamics; see [10] for details). The stability of the
zero dynamics (sufcient to ensure the stability of the internal
one) has been veried but here all the steps are not reported
for simplicity. Thanks to this result, the very simple control
scheme of Fig. 2 can be applied, where R
hi
are the regulators.
In order to employ this scheme, a relation between the actual
and the ctitious inputs is needed. So, if we rewrite (40) as
_
L
g
L
f
h
1
(x, u) = V
1
L
2
f,h
1
(x) = T
N1
(x)
L
g
h
i
(x, u) = V
i
L
f,h
i
(x) = T
Ni
(x) , i = 2, ..., 4
(41)
or
E(x) u = T
N
(x) (42)
DELFINO et al.: A FEEDBACK LINEARIZATION CONTROL SCHEME FOR THE INTEGRATION OF WIND ENERGY CONVERSION SYSTEMS 89
where the matrix E depends only on the system parameters
(e.g., J, C
DC
, and so on) and x, then the required relationship is
_
U
md
U
mq
U
nd
U
nq

T
= E
1
(x) T
N
(x) . (43)
We veried that the matrix E(x) is invertible; this check is
not reported here for brevity.
As anticipated, among the reference values in Fig. 2, that
referring to i
md
is chosen in order to minimize the stator
current module (and then the generator Joule losses), for each
value of the electrical torque C
e
.
The peculiarities of the proposed control strategy if com-
pared with conventional ones can be highlighted as follows.
Traditional controls usually adopt some sort of multi-level
architecture like the one proposed in [19]; an external level
(corresponding to a number of outer control loops with their
P or PI regulators) is responsible for maintaining the wind
turbine at the optimum speed, making at the same time the grid
side converter deliver the desired reactive power and the active
power corresponding to the mechanical one produced by the
wind turbine (or, as an alternative to the former, performing
voltage regulation). The outputs of this outer level are the
reference signals for the converter axis currents, controlled by
inner loops (equipped with another set of PI regulators), and
constituting an intermediate control level whose outputs are the
converter parameters U
md
, U
mq
, U
nd
, and U
nq
. In [19], a third
level, named internal, is identied as the one responsible for
producing, on the basis of these latter parameters, the actual
control signals to the converters switches; furthermore, another
control loop is used to regulate the DC link voltage. So, from
an architectural point of view, a rst difference between
traditional schemes and the one proposed in this paper can
be identied in the fact that this latter does not need such
a hierarchical structure, treating the system as a whole and
requiring in principle only a regulator per each control channel.
Furthermore, in traditional control schemes, the practical
decoupling, for instance, of active and reactive powers deliv-
ered to the grid is usually obtained by a proper choice of
the regulators (and of the related proportional and integral
gains); as an example, in [19], the steady-state decoupling
of these powers is achieved by means of the two inner loop
PI conventional controllers and proper feedback of the grid
side converter output currents. Even if a decoupled control is
implemented for the grid side converter, a cross coupling still
exists between the two grid side axis currents and the DC link
voltage [apparent in (22) and (23)], so this control cannot be
considered as fully decoupled. This problem is faced in [19] by
trying to maintain the DC link voltage as constant as possible
by means of the associated PI regulator. Similar considerations
hold for the scheme presented in [20], where the DC link
voltage is maintained close to the specied reference level by
acting on the active power delivered to the grid, in order to
make it match that of the power input to the generator side
converter. On the contrary, in the proposed approach, the actual
decoupling of all control channels is a priori guaranteed by
the FBL theory both in transient and in steady state.
A third important difference can be identied which has
practical implications in the control system design, i.e., in
TABLE I
Modeled System Parameters
Parameter Description Value
P
m
Rated generator power 2 MW
V
m
Rated generator voltage (L to L) 4 kV
p Number of pole pairs 11
J Moment of inertia 2.52510
6
kgm
2

p
PM ux 166.8 Wb
L
sd
Stator d-axis inductance 0.367 H
L
sq
Stator q-axis inductance 0.250 H
R
s
Stator resistance 0.08
L
fn
Filter inductance 33 mH
R
fn
Filter resistance 0.078
L Inductance between grid side conv. and grid 0.0295 H
R Resistance between grid side conv. and grid 0.086
C
DC
DC link capacitance 100 F
the choice of regulators types and in their proportional and
integral gains settings. This task is usually accomplished by
an heuristic procedure, perhaps making use of trial and error
through several simulations, in a way similar to that outlined
in [20] for the grid side converter controller; the proportional
gain of this controller was chosen by increasing its value
until borderline stability was reached and then taking the
25% of this limit gain to provide a conservative margin
of 4. The integral gain term was then chosen as the 10% of
the proportional one to ensure a small inuence on the loop
gain margin. The application of the FBL makes this same
task straightforward; as the whole system is reduced to the
elementary scheme of Fig. 2, the closed loop behavior of
each control channel with a given regulator and related gain
values can be predicted a priori. An application example of
this feature is illustrated in Section V.
IV. Simulations in Normal Operation
The simulations have been performed on a real system,
whose electrical parameters are listed in Table I.
In order to test the efciency of the new FBL-based control
scheme, different disturbances and reference variations have
been considered, with the aim to highlight the system ability
to:
1) generate or absorb reactive power to or from the grid,
even when the active power is low (Fig. 3);
2) respond to wind speed V
w
variations (Fig. 4);
3) work at optimal current, machine side; I
m
decreased to
the minimum value at t = 30 s (Fig. 9).
As previously stated, one of the main advantages of this
method is the complete decoupling of each control channel;
as a matter of fact, the step variations imposed to h
2
= i
md
at
t = 30 s or h
3
= i
nq
at t = 5 s (Figs. 8, 3, respectively)
to minimize the machine current in order to decrease the
generator losses and to regulate the reactive power at the grid
side, do not affect the machine speed (Fig. 6). This means that
after the beginning of the simulation, the machine speed is set
to the optimum speed value which guarantees maximum power
extraction for the actual wind speed magnitude (performance
coefcient c
p
close to its maximum value), until the wind rises
90 IEEE SYSTEMS JOURNAL, VOL. 6, NO. 1, MARCH 2012
Fig. 3. Reactive power and its reference at the grid-side.
at t = 10 s, without being affected by modication of h
3
. Since
the new V
W
would make both generated power and rotor speed
exceed their rated values, the pitch angle is increased (Fig. 5)
in order to limit the rotor speed.
As anticipated, the variation of h
2
at t = 30 s does not
affect the machine speed, which remains constant till the wind
decreases to 5 m/s (t = 80 s), the pitch angle is set back to
zero and the speed control loop regulates
m
in order to reach
the new optimal value. Fig. 7 shows that both the active power
P
m
generated by the machine and that delivered to the grid
P
n
, follow the behavior of
m
, as expected.
As far as the behavior of the reactive power delivered to
the grid is concerned (Fig. 3), since the regulator is purely
proportional and the transfer function between V
3
and h
3
= i
nq
is an integrator, the response of the control loop (the third one
in Fig. 2) is that of a simple pole, whose time constant can be
tuned by varying the regulator proportional gain; this means
that the closed-loop response of this control channel can be
shaped in a very straightforward way. Similar considerations
could be done for the other control channels, but the feature
is of particular interest for this one, as the reactive power,
being an important ancillary service, is subjected to stringent
regulations by system operators [17], [18], especially in terms
of the system response time when given reference variations
are considered; the developed control scheme can help the de-
signer in fullling these requirements at the control synthesis
stage.
Most importantly, by comparing Figs. 3 and 7, it is evident
that every variation of the active power delivered to the grid
has no effect (not even transient) on the reactive power (and
vice versa); when, after about 80 s, the active power falls below
0.2 MW (i.e., below the 10% of the rated power), the system
continues to inject into the grid the same value of the reactive
power till t = 100 s when, in response to a reference variation,
the reactive power exchanged with the grid becomes negative.
This latter result highlights the system ability to generate or
absorb the required value for the reactive power, even if the
active power is low.
Finally, Fig. 9 shows the effective reduction of machine
current (6%) comparing a control which would impose
i
md
= 0 and the proposed approach. The picture is related
to the generator working point at t = 50 s. The red line is
Fig. 4. Wind speed.
Fig. 5. Pitch angle of turbine blades.
the hyperbole dened by (9), that can be seen as a constraint
(i
md
and i
mq
values must ensure the required electromagnetic
torque), while the blue line is the circle that represents
the current absolute value. As a matter of fact, the curves
corresponding to constant absolute values of the machine
current are circles in the i
md
, i
mq
plane; among them, the one
identifying the allowable minimum for the current module is
the one tangent to (9); this is because the function gradient
and the constraint gradient have to be parallel in the minimum
point. As a consequence, the couple (i
md
, i
mq
) that minimizes
the machine current, providing at the same time the required
electromagnetic torque, is the one identied by the label new
working point in Fig. 9. Figs. 8 and 10 show how, in response
to the step imposed to h
2
= i
md
at t = 30 s, i
mq
, after a transient,
reaches the optimum value corresponding to this point.
V. Behavior During Fault Ride Through
Fault ride through capabilities is nowadays demanded by a
number of TSO grid codes [15], [16] both for conventional and
renewable generation units. Prescriptions differ considerably
in terms of the shape of the limit curves for the voltage [17]
(which identify the severest voltage drop the generation unit
must withstand) and also in terms of the behavior the generator
must comply with during the fault (essentially, the active and
reactive currents to be injected into the grid).
Here, the voltage limit curve (Fig. 11, V
PCC
) and the pre-
scription about the reactive current to be delivered as indicated
DELFINO et al.: A FEEDBACK LINEARIZATION CONTROL SCHEME FOR THE INTEGRATION OF WIND ENERGY CONVERSION SYSTEMS 91
Fig. 6. Machine angular frequency and its reference (optimal value).
Fig. 7. Active power at machine and grid side.
Fig. 8. Active current at machine side.
by E.ON Netz [16] are taken as a reference. In particular, as
far as the latter aspect is concerned, for a voltage drop/rise
(with respect to the rated value) exceeding a dead band of
10%, the reactive current fed to the grid must be equal to
I
nq

ref
= 2
(V

grid

pre fault V

grid)
V

grid

rated
I
n

rated
+I
q

prefault
(44)
while for a drop greater than 50%, at least the rated value must
be guaranteed. Fig. 12 graphically represents the constraint
imposed by (44). Furthermore, the voltage control must take
place within 20 ms after fault recognition, by providing the
required current.
Fig. 9. Machine current minimization.
Fig. 10. Reactive current at machine side.
Fig. 11. Voltage prole requested by E.ON for fault ride-through.
As illustrated in Fig. 11, the fault ride through transient
takes place at t = 0 s; the grid voltage drops to zero and in
response, the control system raises the reference for I
nq
from
the pre-fault value to the rated one (Fig. 15). The actual value
of I
nq
follows its reference with a time behavior given by the
third control loop sketched in Fig. 2. The resulting voltage
support action can be appreciated in Fig. 11; as can be seen,
the PCC experiences a voltage drop less severe than the one
affecting the grid.
Reactive and active powers delivered to the grid during the
fault are shown in Figs. 13 and 14, respectively; it can be
observed that the reactive power fed to the grid does not
92 IEEE SYSTEMS JOURNAL, VOL. 6, NO. 1, MARCH 2012
Fig. 12. Principle of voltage support by reactive current injection during grid
faults specied by E.ON (reproduction from [16]).
Fig. 13. Reactive power at the PCC.
vanish, thanks to the just highlighted voltage support at the
PCC.
After 150 ms, the grid voltage starts rising linearly and,
when the 0.5 pu value is reached, the reactive current is
decreased as prescribed by (44); in the same time interval, the
active power increases and, after a transient, nally reaches
the pre-fault value. On the contrary, the nal values of both
reactive power and I
nq
differ from the pre-fault ones as,
according to the prescribed limit curve, the grid voltage after
the fault is lower than 0.9 pu and, as a consequence, the system
must continue to deliver reactive power.
The cited constraints about the control system response time
of 20 ms can represent an example of how easily the system
transient behavior can be shaped, thanks to the inherent
simplicity of the block diagram of the equivalent linearized
system in Fig. 2. Using a purely proportional regulator, the
I
nq
closed loop transfer function is simply
F (s) =
R
Inq
/s
1 + R
Inq
/s
=
R
Inq
s + R
Inq
(45)
i.e., a simple pole whit time constant = 1/R
inq
; so, if we
consider the steady state reached after, say, 4, we can
calculate the value of R
inq
that allows us to fulll the
constraint
0.02 s = 4 R
Inq
=
4
0.02
= 200. (46)
Fig. 14. Active power at the PCC.
Fig. 15. Reactive current at the grid side.
Fig. 16. Reactive current at the grid side (rst 50 ms of the transient) for
different R
inq
values.
Fig. 16 shows (in an extended view of the rst 50 ms of the
transient) the response obtained for I
nq
with different values
of R
inq
; the diagram conrms the obtained value.
VI. Conclusion and Perspectives of Future Work
This paper discussed issues related to the management of
wind renewable energy in distribution networks. The presented
control scheme, together with the mathematical model needed
for its implementation, offers the possibility of a fully decou-
pled control of the main variables which characterize WECS
DELFINO et al.: A FEEDBACK LINEARIZATION CONTROL SCHEME FOR THE INTEGRATION OF WIND ENERGY CONVERSION SYSTEMS 93
operation, allowing to regulate the reactive power injected into
the grid independently from the maximum power point track-
ing and maintaining, at the same time, the DC link voltage at
its rated value and the generator at its maximum power factor.
Moreover, the very simple system structure resulting from
the application of the FBL greatly simplies the regulators
synthesis, thus making it easier to meet regulatory constraints
on control loop dynamic behavior. It should be underlined that
such features can help integrating wind distributed generation
resources in modern smart grids, in view of their ever growing
application both for active power production and ancillary
services provision [21].
Future research activities will be devoted to the extension of
the presented approach to systems equipped with short-term
storage for power smoothing applications.
References
[1] D. Kline, Benets to the United States of increasing global uptake
of clean energy technologies, U.S. DOE Natl. Renewable Energy Lab.
(NREL), Tech. Rep. NREL/TP-6A2-47807, Jul. 2010.
[2] National Energy Policy Recommendations, position statement by IEEE-
USA Board of Directors, Feb. 2010.
[3] European Parliament. (2008, Dec.). EUs Climate and Energy Pol-
icy [Online]. Available: http://ec.europa.eu/climateaction/eu

action/
index

en.htm
[4] P. Capros, PRIMES model analysis for EUSUSTEL project,
ICCS/NTUA DG Research FP6, Dec. 2006.
[5] A. Negri, A systemic approach for the electric power system, in Proc.
CIGR

E Conf. Int. Standards Promote Energy Efciency Reduce Carbon
Emissions, Mar. 2009, pp. 19.
[6] C. W. Gellings, Energy efciency, a renewed imperative, ELECTRA,
vol. 240, Oct. 2008.
[7] U.S. DOE: Ofce of Energy Efciency and Renewable Energy. 20%
Wind Energy by 2030 [Online]. Available: http://www1.eere.energy.gov/
windandhydro/pdfs/41869.pdf
[8] K. Tan and S. Islam, Optimum control strategies in energy conversion
of PMSG wind turbine system without mechanical sensors, IEEE Trans.
Energy Convers., vol. 19, no. 2, pp. 392399, Jun. 2004.
[9] P. M. Anderson and A. Bose, Stability simulation of wind turbine
systems, IEEE Trans. Power Appl. Syst., vol. PAS-102, no. 12, pp.
37913795, Dec. 1983.
[10] J. J. Slotine and W. Lee, Applied Nonlinear Control. Englewood Cliffs,
NJ: Prentice-Hall, 1991.
[11] J. K. Hedrick and A. Girard. (2005). Control of Nonlinear Dy-
namic Systems: Theory and Applications, ch. 8 [Online]. Available:
http://www.me.berkeley.edu/ME237/notes.html
[12] F. Zhang and B. Fernandez-Rodriguez, Feedback linearization control
of systems with singularities, in Proc. ICCS, Jun. 2006, pp. 412424.
[13] T.-S. Lee, Input-output linearization and zero-dynamics control of
three-phase ac/dc voltage source converters, IEEE Trans. Power Elec-
tron., vol. 18, no. 1, pp. 1122, Jan. 2003.
[14] A. Megretski, Lecture 13: Feedback linearization, Dept. Electr. Eng.
Comput. Sci., Massachusetts Instit. Technol., Cambridge, Tech. Rep.
6.243j, 2003.
[15] RTE. Gestionnaire du R eseau de Transport delectricit e [Online].
Available: http://clients.rte-france.com/htm/fr/mediatheque/telecharge/
reftech/18-12-09

complet.pdf
[16] E.ON Netz GmbH, Grid Code: Grid Connection Regulations for High
and Extra High Voltage, Apr. 2006.
[17] G. Jo os, Wind turbine generator low voltage ride through requirements
and solutions, in Proc. IEEE Power Energy Soc. General Meeting
Conversion Delivery Electr. Energy 21st Century, Jul. 2008, pp. 17.
[18] F. Delno, R. Procopio, M. Rossi, and G. Ronda, Integration of
large-size photovoltaic systems into the distribution grids: A P-Q chart
approach to assess reactive support capability, IET Renew. Power
Gener., vol. 4, no. 4, pp. 329340, 2010.
[19] M. G. Molina, A. G. Sanchez, and M. Rizzato Lede, Dynamic modeling
of wind farms with variable-speed direct-driven PMSG wind turbines,
in Proc. IEEE/PES T&D Transmission Distribution Conf. Expos. Latin
America, Nov. 2010, pp. 18.
[20] G. Ramtharan, A. Arulampalam, J. B. Ekanayake, F. M. Hughes, and N.
Jenkins, Fault ride through of fully rated converter wind turbines with
AC and DC transmission systems, IET Renewable Power Generation,
vol. 3, no. 4, pp. 426438, 2009.
[21] H. Weber and M. Kleimaier, Grid integration of wind generation, in
Proc. Int. CIGRE Symp. Integr. Wide-Scale Renewable Resources Into
Power Delivery Syst., Jul. 2009, pp. 16.
Federico Delno (M03) was born in Savona, Italy,
on February 28, 1972. In 1997, he received the M.S.
degree (with honors) in electrical engineering and
the Ph.D. degree, both from the University of Genoa,
Genoa, Italy, in 1997 and 2001, respectively.
He is currently an Associate Professor with the
Department of Naval and Electrical Engineering,
University of Genoa. He is the author or co-author of
more than 90 scientic papers published in reviewed
journals or presented at international conferences.
His current research interests include mainly power
system transients, integration of renewables into the power delivery systems,
and smart power grids.
Dr. Delno is a member of the European Commission Task Force on Smart
GridsExpert Group 3: Roles and Responsibilities of Actors involved in the
Smart Grids Deployment and a member of the Italian Expert Committee on
Energy Efciency, at the Economic Development Ministry.
Fabio Pampararo was born in Genova, Italy, on
February 14, 1984. In 2008, he received the M.S.
degree with honors in electrical engineering from the
University of Genoa, Genova, where he is currently
pursuing the Ph.D. degree with the Department of
Naval and Electrical Engineering.
He works on power quality improvement in distri-
bution networks, electric machine modeling, power
systems control, and integration of renewables into
the power delivery system. He is the author or co-
author of more than ten scientic papers published
in reviewed journals or presented at international conferences.
Renato Procopio (M03) was born in Savona, Italy,
on March 6, 1974. He received the Bachelors cum
laude degree in electrical engineering in 1999 and
the Ph.D. degree in 2004, both from the University
of Genoa, Genova, Italy.
He is currently a Researcher with the Department
of Naval and Electrical Engineering, University of
Genoa. He works on power systems control and opti-
mization, power quality improvement in distribution
networks, and power system transients. He is the
author or co-author of more than 80 scientic papers
published in reviewed journals or presented at international conferences.
Mansueto Rossi was born in Savona, Italy, on
April 10, 1974. In 2000, he graduated cum laude in
electrical engineering from the University of Genoa,
Genova, Italy, where he received the Ph.D. degree
in 2004.
He is currently a Researcher with the Department
of Naval and Electrical Engineering, University of
Genoa. He is the author or co-author of more than
80 scientic papers published in reviewed journals
or presented at international conferences. His current
research interests include power system transients,
smart power grids, and power system control and optimization.