You are on page 1of 12

Published in IET Renewable Power Generation

Received on 26th January 2012


Revised on 10th November 2012
Accepted on 22nd November 2012
doi: 10.1049/iet-rpg.2012.0026
ISSN 1752-1416
Second-order sliding-mode controller design and
tuning for grid synchronisation and power control of a
wind turbine-driven doubly fed induction generator
Ana Susperregui
1
, Miren Itsaso Martinez
1
, Gerardo Tapia
1
, Ionel Vechiu
2
1
Department of Systems Engineering and Control, Polytechnical University College, University of the Basque Country
UPV/EHU, Plaza de Europa, 1, 20018 Donostia-San Sebastin, Spain
2
Ecole Suprievre des Technologies Industrielles Avances (ESTIA) Research, Technopole Izarbel, 64210 Bidart, France
E-mail: ana.susperregui@ehu.es
Abstract: This study presents a second-order sliding-mode control (2-SMC) scheme for a wind turbine-driven doubly fed
induction generator (DFIG). The tasks of grid synchronisation and power control are undertaken by two different algorithms,
designed to command the rotor-side converter at a xed switching frequency. Effective tuning equations for the parameters of
both controllers are derived. A procedure is also provided that guarantees bumpless transfer between the two controllers at the
instant of connecting the DFIG to the grid. The resulting 2-SMC scheme is experimentally validated on a laboratory-scale
7 kW DFIG test bench. Experimental results evidence both the high dynamic performance and the superior robustness
achieved with the proposed control scheme.
Nomenclature
cosj power factor
i
r
, i
s
rotor and stator current space-vectors
i
r
, i
s
components of rotor and stator current
space-vectors
L
lr
, L
ls
rotor and stator leakage inductances
L
m
, L
r
, L
s
magnetizing, rotor and stator inductances
L
r
rotor transient inductance
n general turns ratio
P number of pole pairs
P
r
, P
s
, P
t
rotor-side, stator-side and total active
powers
Q
s
stator-side reactive power
R
r
, R
s
rotor and stator resistances
s slip
v
g
, v
s
grid and stator voltage space-vectors
v
g
, v
r
components of grid and rotor voltage
space-vectors

s
angle between the stator-ux-oriented and the
stationary reference frames

s
angle between the grid-voltage-oriented and
the stationary reference frames

r
,
r
rotor electrical position and speed

rm
rotor mechanical speed

s
,
sl
synchronous and slip angular frequencies

s
stator ux space-vector

s
component of stator ux space-vector
Subscripts
D,
Q
direct- and quadrature-axis components, expressed
in the stationary reference frame
x, y direct- and quadrature-axis components, expressed
in the stator-ux-oriented reference frame
x, y direct- and quadrature-axis components, expressed
in the grid-voltage-oriented reference frame
, direct- and quadrature-axis components, expressed
in the rotor natural reference frame
Superscripts
R rated value
* reference value
1 Introduction
Standard vector control (VC) schemes conceived to command
wind turbine-driven doubly fed induction generators (DFIGs)
typically comprise proportional-integral (PI)-based cascaded
current and power control loops [13]. However, the system
transient performance degrades as the actual values of the
DFIG resistances and inductances deviate from those based
on which the control system tuning was carried out during
commissioning [4]. In addition, the tracking of the
maximum power point achieved by adopting classical VC
congurations exhibits the characteristic lags associated
with PI-based control schemes [5, 6].
www.ietdl.org
540 IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551
& The Institution of Engineering and Technology 2013 doi: 10.1049/iet-rpg.2012.0026
Overcoming those two drawbacks entails improving
robustness to DFIG parameter variations, as well as
achieving high dynamic performance power control. In this
framework, alternative control schemes for DFIGs are being
proposed, among which a strong research line focuses on
direct power control (DPC) [4, 7]. Several others explore
the option of applying sliding-mode control (SMC) [5, 810].
In particular, a rst-order SMC (1-SMC) conguration for
decoupled control of DFIG active and reactive powers is
described in [5]. In spite of conferring outstanding
robustness to parameter variations and superior tracking of
a rotor speed-dependent optimum power curve, the 1-SMC
structure put forward leads to a variable switching
frequency of the rotor-side converter (RSC) transistors. This
may cause broadband harmonics to be injected into the
grid, hence complicating the design of both the
back-to-back converter and the grid-side AC lter [5, 7].
This drawback is overcome in [10] through application of
the so-called boundary layer method. By adopting this
approach, the discontinuous rotor voltage control signals to
which 1-SMC leads are rst smoothed out. The resulting
voltage waveforms are then applied to the DFIG rotor by
means of a modulation technique, such as pulse-width
modulation (PWM) or space-vector modulation, which
drives the RSC transistors at a constant switching
frequency. Even if the described approach solves the
problem of variable switching frequency, it is only at the
expense of severely compromising the characteristics of
tracking accuracy, robustness and disturbance rejection
conferred by SMC. Indeed, within the boundary layer, the
feedback system no longer behaves as dictated by sliding
mode, and it is simply reduced to a system with no sliding
mode [11].
In this context, with the objective of keeping a constant
switching frequency without renouncing to the robustness
and tracking ability features gained with 1-SMC, a
PWM-based second-order SMC (2-SMC) scheme for DFIG
control is presented in this paper. This control conguration
is not only aimed at accurately tracking time-varying
reference values for the active and reactive powers to be
exchanged between the DFIG stator and the grid, but also
at ensuring the synchronisation required for smooth
connection of the DFIG stator to the grid [1215].
In addition, the problem of bumpless transfer between the
controllers in charge of synchronisation and power control,
at the instant of connecting the DFIG stator to the grid, is
solved straightforwardly.
Not only the design and implementation of the resulting
global control scheme are addressed in detail, but also its
analytical tuning. As an alternative to the conventional
trial-and-error method frequently adopted to adjust 2-SMC
algorithms [16], tuning equations for the global 2-SMC
system put forward are derived based on DFIG parameters
and intuitive specications for the closed-loop time
response.
Results of experimental evaluation of the proposed 2-SMC
scheme on a 7 kW DFIG prototype are reported.
2 Overview of DFIG models for power
control and grid synchronisation
When the DFIG stator is connected to the grid, its rotor-side
model, expressed in the stator-ux-oriented xy reference
frame depicted in Fig. 1, is given by [17]

i
rx
=
v
rx
L

R
r
L

r
i
rx

L
m
L
s
L

r
|

c
s
| + v
sl
i
ry
(1)

i
ry
=
v
ry
L

R
r
L

r
i
ry
v
sl
i
rx
+
L
m
L
s
L

r
|c
s
|
_ _
(2)
where L

r
= L
r
L
2
m
_ _
/ L
s
_ _
and
sl
=
s

r
. Regarding the
active and reactive powers owing between the DFIG stator
and the grid, they may be expressed as [17]
P
s
=
3
2
L
m
L
s
|v
g
|i
ry
; Q
s
=
3
2
|v
g
|
L
s
|c
s
| L
m
i
rx
_ _
(3)
Expressions (1)(3) consider that currents owing into the
DFIG stator and rotor are positive, which implies that
generated active and reactive powers are represented as
negative.
On the other hand, given that power generation is not
protable for rotor speeds below a given value, the DFIG
stator is not connected to the grid until that threshold is
exceeded. The connection is not trivial, since the grid
voltage and that induced at the open stator of the DFIG
must coincide both in magnitude and phase to avoid
short-circuits. Therefore, prior to connection, the DFIG
must be commanded to achieve grid synchronisation [18].
Let a new xy reference frame be adopted when the
stator is disconnected from the grid, whose y quadrature
axis is collinear with the grid voltage space-vector, as
displayed in Fig. 2. As proved in [14], the open-stator
Fig. 2 xy reference frame for synchronisation
Fig. 1 Stator-ux-oriented xy reference frame, together with the
stationary DQ reference frame, and the rotor natural frame,
www.ietdl.org
IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551 541
doi: 10.1049/iet-rpg.2012.0026 & The Institution of Engineering and Technology 2013
model for the DFIG rotor side, expressed in that reference
frame, is given by

i
rx
=
v
rx

L
r

R
r
L
r
i
rx
+ v
sl
i
ry
(4)

i
ry
=
v
ry

L
r

R
r
L
r
i
ry
v
sl
i
rx
(5)
It turns out that, in steady state, if disconnected from the
grid, the DFIG stator ux and voltage space-vectors are,
respectively, collinear with x and y axes [14]. As a result,
v
s

s
, as depicted in Fig. 2. xy and xy reference frames
must hence be aligned to achieve synchronisation. Yet, for
a complete match up, v
s
must be not only collinear with v
g
,
but also identical in magnitude. As shown in [14], those
two requirements are satised if rotor current regulation is
carried out around set-points
i

rx
=
|v
g
|
L
m
v
s
; i

ry
= 0 (6)
3 Second-order SMC scheme
A global 2-SMC scheme for the DFIG is described in this
section. Owing to the different dynamic behaviours and
control targets to be dealt with when its stator is
disconnected from or connected to the grid, the relevant
DFIG model is considered to design the control law for
each of those two cases.
3.1 Selection of switching variables
Examination of the set-points in (6) evidences that rotor
current regulation must be carried out for grid
synchronisation. Therefore when the DFIG stator is
disconnected from the grid, the switching functions given
next are assumed
s
x
= i

rx
i
rx
+ c
x

_
i

rx
i
rx

_ _
dt (7)
s
y
= i

ry
i
ry
+ c
y

_
i

ry
i
ry

_ _
dt (8)
where the integral terms, weighted by c
x
and c
y
positive
constants, are added for steady-state response improvement,
as suggested in [19].
On the other hand, seeking to achieve close tracking
of the maximum power point, P

s
, as well as high
dynamic performance control of the reactive power, the
following switching variables are adopted when connected
to the grid
s
P
= P

s
P
s
+ c
P
_
P

s
P
s
_ _
dt (9)
s
Q
= Q

s
Q
s
+ c
Q
_
Q

s
Q
s
_ _
dt (10)
where c
P
and c
Q
are also positive constants to be tuned.
3.2 Synchronisation DFIG disconnected from
the grid
3.2.1 Design: Bearing in mind that the i

rx
and i

ry

set-points in (6) are constant, the following time derivatives


of switching functions s
x
and s
y
result from combination of
(7), (8), (4) and (5)
s
x
= c
x
i

rx
+
R
r
L
r
c
x

_ _
i
rx
v
sl
i
ry

1
L
r
v
rx
(11)
s
y
= c
y
i

ry
+
R
r
L
r
c
y

_ _
i
ry
+ v
sl
i
rx

1
L
r
v
ry
(12)
which reveal that the DFIG is of rst-order relative degree
during synchronisation. Accordingly, it may be commanded
by applying 1-SMC or 2-SMC [16]. As already mentioned,
a structure based on 1-SMC would however lead to a
variable switching frequency of the RSC transistors,
therefore complicating the design of the power converter
itself, as well as that of the grid-side AC lter. Aiming to
overcome that drawback, a PWM 2-SMC realisation,
known as the super-twisting algorithm (STA), is adopted [20].
Consequently, the voltage to be applied to the rotor is
computed according to control law
v
rx
= v
rx

ST
+ v
rx

eq
; v
ry
= v
ry

ST
+ v
ry

eq
(13)
where the terms with subscript ST are calculated, through
application of the STA, as
v
rx

ST
= l
x

.....
|s
x
|
_
sgn s
x

_ _
+ w
x

_
sgn s
x

_ _
dt (14)
v
ry

ST
= l
y

.....
|s
y
|
_
sgn(s
y
) + w
y

_
sgn(s
y
) dt (15)
with
x
, w
x
,
y
and w
y
being positive parameters to be tuned.
The addends with subscript eq in (13), which correspond to
equivalent control terms, are derived by letting s
x
= s
y
= 0,
as dened in [19]. As a result
v
rx

eq
= L
r
c
x
i

rx
+
R
r
L
r
c
x

_ _
i
rx
v
sl
i
ry

_ _
(16)
v
ry

eq
= L
r
c
y
i

ry
+
R
r
L
r
c
y

_ _
i
ry
+ v
sl
i
rx

_ _
(17)
The sliding regime in manifold s
x
= s
y
= s
x
= s
y
= 0 can
also be attained by applying only the STA control terms in
(13). Accordingly, the equivalent control terms in (16) and
(17) are not strictly necessary. However, if incorporated, the
more accurately they are estimated, the lower is the control
effort let to be done by the STA. In conclusion, although
uncertainty in L
r
and R
r
leads to inaccurate v
rxeq
and v
ryeq
,
robustness of the overall control algorithm in (13) is not put
in jeopardy, as robustness does actually lie on the STA
control terms given by (14) and (15).
Equivalent control terms are hence included not only to
improve the system transient response [21], but also to ease
selection of constants c
x
and c
y
, as well as tuning of the
STA
x
,
y
, w
x
and w
y
gains. Owing to the analogy
existing between the procedures for tuning the algorithms
commanding i
rx
and i
ry
, only that related to i
rx
will be
presented in detail.
www.ietdl.org
542 IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551
& The Institution of Engineering and Technology 2013 doi: 10.1049/iet-rpg.2012.0026
3.2.2 Tuning: Substitution of control law (13) into (11)
produces
s
x
=
1
L
r
l
x

.....
|s
x
|
_
sgn s
x

_ _
+ w
x

_
sgn s
x

_ _
dt
_ _
(18)
Now, given that sgn(s
x
) = s
x
/|s
x
|, taking the time derivative
of (18) leads to
s
x
=
1
L
r
l
x

2
.....
|s
x
|
_ s
x
+ w
x

s
x

|s
x
|
_ _
(19)
Let us assume that, because of the rst term in the STA, the
reaching phase is satisfactorily completed and the sliding
regime is entered. From that moment on, |s
x
|
x
, with
x
close to zero. Considering the most unfavourable case, in
which |s
x
| =
x
, and using the denition of s
x
in (7), the
following expression can be worked out from (19)
e
x
+
l
x

2L
r
...
d
x

_ + c
x

_ _
..........,,..........
a
2
c
x
,l
x
( )
e
x
+
1
L
r
...
d
x

_
l
x
c
x

2
+
w
x

...
d
x

_
_ _
..............,,..............
a
1
c
x
,l
x
,w
x
( )
e
x

+
w
x
c
x

L
r
d
x

..,,..
a
0
c
x
,w
x
( )
_
e
x
dt = 0 (20)
where e
x
= i

rx
i
rx
. Taking the time derivative of (20),
the following differential equation reecting the e
x
error
dynamics while in sliding regime is obtained
e
x
+ a
2
e
x
+ a
1
e
x
+ a
0
e
x
= 0 (21)
Hence, once
x
is xed, adequate selection of c
x
,
x
and w
x
allows attaining certain target error dynamics,
established through the third-order characteristic equation
given next
p
2
+ 2j
x
v
nx
p + v
2
nx

_ _
p + aj
x
v
nx

_ _
= p
3
+ (2 + a)j
x
v
nx

........,,........
a
2t
p
2
+ 1 + 2aj
2
x

_ _
v
2
nx

.........,,.........
a
1t
p + aj
x
v
3
nx

....,,....
a
0t
= 0
(22)
which, provided that is selected high enough 10
gives rise to a pair of dominant poles with respect to a third
one placed at p =
x

nx
. As a result, it can be
considered that target error dynamics are entirely dened
via
x
damping coefcient and
nx
natural frequency.
Those designer-dened error dynamics would theoretically
be achieved just by tuning c
x
,
x
and w
x
so that a
2
= a
2t
,
a
1
= a
1t
and a
0
= a
0t
are simultaneously fullled.
Considering the expressions for a
2
, a
1
and a
0
provided in
(20), as well as those for a
2t
, a
1t
and a
0t
reected in (22),
the latter requirements lead to the following tuning equations
c
3
x
(2 + a)j
x
v
nx

........,,........
a
2t
c
2
x
+ 1 + 2aj
2
x

_ _
v
2
nx

.........,,.........
a
1t
c
x
aj
x
v
3
nx

....,,....
a
0t
= 0
(23)
l
x
= 2L
r
...
d
x

_
(2 + a)j
x
v
nx
c
x

_ _
(24)
w
x
=
L
r
d
x
aj
x
v
3
nx

c
x

(25)
It is important to note that the coefcients in (23) coincide
with those of target characteristic equation (22), except for
the signs of the squared and independent terms, which are
negative. It therefore turns out that the three possible values
for c
x
are equal to the roots poles of target
characteristic equation (22), although their real parts have
opposite signs. The real parts of the desired poles must
necessarily be negative to ensure stability, which implies
that the real parts of the three possible values for c
x
will
always be positive. As a result, since (23) is third-order, it
is guaranteed that at least one of the three solutions for c
x
will be both real and positive, as required. If (23) has more
than one feasible solution for c
x
, then the {c
x
,
x
, w
x
} set
leading to the best performance may be identied through
numerical simulation.
3.2.3 Implementation: The functional diagram provided
in Fig. 3 illustrates the procedure for implementing the
proposed synchronisation algorithm. Any variable present
in a given layer of the diagram is also available to the
layers inside, which implies that the algorithm must be
implemented from the outer to the inner layer labelled,
respectively, as 1
st
Step and 3
rd
Step at their bottom
left-hand corners.
In addition to the i

rx
and i

ry
set-points provided in (6), the
actual values of i
rx
and i
ry
are required to compute switching
functions s
x
and s
y
, as well as equivalent control terms
(16) and (17). Knowledge of
r
is also needed to calculate
the
sl
slip angular frequency present in (16) and (17). In
accordance with Figs. 1 and 2, the Parks e
j r

s
u
r
( )
transform is used to bring rotor current i
r
and i
r
components to the xy reference frame, where i
r
and i
r
are derived by applying the Clarkes transform to directly
measured three-phase rotor current. Based on Fig. 2, angle

s
is calculated as
arctan
v
gQ
v
gD
90

(26)
while
r
, as well as
r
, can either be measured using an
incremental encoder or estimated by an observer, if a
sensorless scheme is to be implemented.
Similarly, rotor voltage v
r
and v
r
natural-frame
components are obtained through application of the inverse
e
j r

s
u
r
( )
Parks transform to v
rx
and v
ry
control signals. The
inverse Clarkes transform is then applied to v
r
and v
r
in
order to achieve the three-phase rotor voltage waveforms to
be supplied as reference signals to the PWM algorithm.
3.3 Power control DFIG connected to the grid
3.3.1 Design: Considering that |

c
s
| in (1) turns out to be
negligible when grid connected, combination of (9) and
(10) with (1)(3) leads to the time derivatives of switching
www.ietdl.org
IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551 543
doi: 10.1049/iet-rpg.2012.0026 & The Institution of Engineering and Technology 2013
variables s
P
and s
Q
given next
s
P
=

P

s

3
2
L
m
L
s
|v
g
|
R
r
L

r
c
P
_ _
i
ry
+ v
sl
i
rx
+
L
m
L
s
L

r
|c
s
|
_ _ _ _
+ c
P
P

s
+
3
2
L
m
L
s
L

r
|v
g
|v
ry
(27)
s
Q
=

Q

s

3
2
|v
g
|
L
s
c
Q
|c
s
| +
R
r
L

r
c
Q
_ _
L
m
i
rx
v
sl
L
m
i
ry
_ _
+ c
Q
Q

s
+
3
2
L
m
L
s
L

r
|v
g
|v
rx
(28)
which prove that the DFIG is also of rst-order relative degree
when connected to the grid. Therefore the design and tuning
of power controllers is analogous to that presented in
the preceding section. Thereby, the rotor is fed with the
following xy axes-referred voltage
v
rx
= v
rxST
+ v
rxeq
; v
ry
= v
ryST
+ v
ryeq
(29)
The STA terms are then obtained as
v
rxST
= l
Q
.... .
|s
Q
|
_
sgn(s
Q
) w
Q
_
sgn(s
Q
) dt (30)
v
ryST
= l
P
....
|s
P
|
_
sgn(s
P
) w
P
_
sgn(s
P
) dt (31)
where
Q
, w
Q
,
P
and w
P
are positive constants to be tuned.
On the other hand, the equivalent control terms in (29) are
computed by zeroing (27) and (28), which yields
v
rxeq
=
2L
s
L

r
3L
m
|v
g
|

s
+ c
Q
Q

s
Q
s
_ _
_ _
+ R
r
i
rx
L

r
v
sl
i
ry
(32)
v
ryeq
=
2L
s
L

r
3L
m
|v
g
|

s
+ c
P
P

s
P
s
_ _ _ _
+ R
r
i
ry
+ v
sl
L

r
i
rx
+
L
m
L
s
|c
s
|
_ _
(33)
3.3.2 Tuning: Given that the algorithms governing P
s
and Q
s
are adjusted by following the same method, only
tuning of that controlling the active power will be dealt
Fig. 3 Functional diagram of the proposed synchronisation algorithm
www.ietdl.org
544 IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551
& The Institution of Engineering and Technology 2013 doi: 10.1049/iet-rpg.2012.0026
with. Substitution of control law (29) in (27) leads to
s
P
=
3
2
L
m
L
s
L

r
|v
g
| l
P
....
|s
P
|
_
sgn s
P
_ _
+ w
P
_
sgn s
P
_ _
dt
_ _
(34)
expression which turns out to be identical to that in (18), once
factor (3L
m
|v
g
|)/(2L
s
L
r
) and subscript P are, respectively,
replaced by 1/(L
r
) and x. Consequently, tuning equations
(23)(25) remain valid for c
P
,
P
and w
P
just by substituting
L
r
and subscript x with (2L
s
L
r
)/(3L
m
|v
g
|) and P. Therefore
c
3
P
(2 + a)j
P
v
nP
c
2
P
+ 1 + 2aj
2
P
_ _
v
2
nP
c
P
aj
P
v
3
nP
= 0
(35)
l
P
=
4L
s
L

r
...
d
P
_
(2 + a)j
P
v
nP
c
P
_ _
3L
m
|v
g
|
(36)
w
P
=
2L
s
L

r
d
P
aj
P
v
3
nP
3L
m
|v
g
|c
P
(37)
3.3.3 Implementation: Fig. 4 displays the functional
diagram corresponding to the proposed power control
scheme. Given that it is analogous to that described in
Section 3.2, only the main differences will be highlighted
here.
On the one hand, it is essential to note that the actual values
of P
s
and Q
s
are computed as
P
s
=
3
2
v
gD
i
sD
+ v
gQ
i
sQ
_ _
; Q
s
=
3
2
v
gQ
i
sD
v
gD
i
sQ
_ _
(38)
hence avoiding the use of the DFIG parameter-dependent
expressions in (3), which are only to be used for controller
design. Regarding the reference values for active and
reactive powers, a maximum power point tracking (MPPT)
strategy provides P

s
as a function of
r
, while Q

s
is
established directly or based on P

s
and on the demanded
cos j* power factor.
On the other hand, angle
s
is required instead of
s
to carry
out the Parks transform and inverse transform reected in the
diagram of Fig. 4. In this case, as evidenced by Fig. 1,
s
is
calculated as
r
s
= arctan
c
sQ
c
sD
(39)
where
sD
and
sQ
are in turn estimated as
c
sD
=
_
v
gD
R
s
i
sD
_ _
dt; c
sQ
=
_
v
gQ
R
s
i
sQ
_ _
dt
(40)
Fig. 4 Functional diagram of the proposed power control scheme
www.ietdl.org
IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551 545
doi: 10.1049/iet-rpg.2012.0026 & The Institution of Engineering and Technology 2013
Aiming at avoiding drift, bandpass lters are used rather than
pure integrators to implement (40) [22]. Accordingly, |
s
| in
equivalent control term (33) is computed from
sD
and
sQ
as follows
|c
s
| =
...........
c
2
sD
+ c
2
sQ
_
(41)
3.4 Bumpless transfer
So far, the separate performance of each of the two DFIG
operating states disconnected from and connected to the
grid has only been considered, but an undesirable
phenomenon may appear if the switch between the two
controllers is not properly carried out. If a direct transfer is
accomplished, discontinuities arise in the control signals at
the instant of connection, owing to the magnitude mismatch
between the rotor voltage components generated by the two
controllers. This effect produces high stator current values,
leading the machine to exchange excessive power with the
grid at the instant of connection. Aiming to attenuate or
even avoid this bump, it is required to apply the same
values of the control signals previous to and just after the
transfer, that is
v
rx
= v
rx
; v
ry
= v
ry
(42)
Setting the focus on the DFIG rotor voltage components when
connected to the grid, appropriate combination of (29)(33)
produces
v
rx
= l
Q
.....
|s
Q
|
_
sgn(s
Q
) w
Q
_
sgn(s
Q
) dt

2L
s
L

r
3L
m
|v
g
|

s
+ c
Q
Q

s
Q
s
_ _
_ _
+ R
r
i
rx
L

r
v
sl
i
ry
(43)
v
ry
= l
P
....
|s
P
|
_
sgn s
P
_ _
w
P
_
sgn(s
P
) dt

2L
s
L

r
3L
m
|v
g
|

s
+ c
P
P

s
P
s
_ _ _ _
+ R
r
i
ry
+ v
sl
L

r
i
rx
+
L
m
L
s
|c
s
|
_ _
(44)
Condition (42) can be satised by carefully selecting the
initial values for integrals
_
sgn(s
Q
) dt and
_
sgn(s
P
) dt at
the instant of connection. Thus, substitution of (43) and
(44) into (42) allows deriving those initial values as
_
sgn(s
Q
) dt
t=0
=
R
r
i
rx
L

r
v
sl
i
ry
l
Q
.... .
|s
Q
|
_
sgn(s
Q
) v
rx

w
Q

2L
s
L

r

Q

s
+ c
Q
Q

s
Q
s
_ _
_ _
3L
m
|v
g
|w
Q
(45)
_
sgn(s
P
) dt
t=0
=
R
r
i
ry
+ L

r
v
sl
i
rx
l
P
....
|s
P
|
_
sgn s
P
_ _
v
ry

w
P

2L
s
L

r

P

s
+ c
P
P

s
P
s
_ _ _ _
3L
m
|v
g
|w
P
+
L
m
v
sl
|c
s
|
L
s
w
P
(46)
4 Experimental validation
4.1 Description of the experimental rig
Experimental validation of the global control scheme
presented throughout Section 3 is carried out on the 7 kW
DFIG test bench shown in Fig. 5. The electric parameters
of the DFIG under consideration are collected in Table 1.
The values of resistances and inductances were identied in
a recent study reported in [23]. The DFIG is driven by a
15 kW armature-controlled dc motor, whose rotational
speed can be commanded at will via a commercially
available controller. The eMEGAsim OP4500 F11-13
platform by Opal-RT is used to perform rapid control
prototyping of the synchronisation and power control
algorithms reected, respectively, in the functional diagrams
of Figs. 3 and 4. The state of the synchronisation stage is
displayed by means of a conventional lamp synchronoscope.
The sample rate for the global control algorithm, as well
as the frequency for the triangular carrier of the PWM
Table 1 7 kW DFIG electric parameters
Parameter Value
|v
g
| 400
.....
2/3

V
|i
s
|
R
16
..
2

A
|i
r
|
R
24
..
2

A
R
s
370 m
R
r
145.8541 m
L
ls
4.86 mH
L
lr
1.2138 mH
L
m
37.6812 mH
N 2.001
L
s
= L
ls
+ nL
m
80.2601 mH
L
r
= L
lr
+ L
m
/n 20.045 mH
P 2
Fig. 5 Experimental setup
www.ietdl.org
546 IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551
& The Institution of Engineering and Technology 2013 doi: 10.1049/iet-rpg.2012.0026
modulator, are xed to 5 kHz. In addition, the technique
proposed in [24] is implemented in order to compensate the
2.3 s dead time inherent to the RSC. The integral terms
included in both the switching functions and the STA itself
are digitally implemented based on Tustins trapezoidal
method. Yet, with the aim of avoiding derivative ringing
[25], Eulers rectangular method is adopted to discretise the
time derivatives of Q

s
and P

s
included in equivalent
control terms (32) and (33).
Tuning equations (23)(25) and (35)(37) have been
directly applied to derive the values of the global 2-SMC
algorithm parameters reected in Table 2.
x, y
and
nx, y
have been selected so as to achieve synchronisation of
stator voltage to that of the grid in 105 ms, showing no
overshoots. Given that
x, y
= 1, two values are possible for
c
x, y
; namely,
nx, y
and
nx, y
. However, numerical
simulation revealed that the {c
x, y
,
x, y
, w
x, y
} parameter
set leading to the best performance was that resulting from
selecting c
x, y
=
nx, y
. Similarly, as a result of the values
xed for
P, Q
and
nP, Q
, possible errors arising in active
and reactive powers would theoretically vanish in 50 ms,
with no overshoots. As for the previous case, c
P, Q
are
chosen to be equal to
nP, Q
.
For simplicity, the power signal feedback method is
applied in this case to dene the maximum power point to
be attained, P

s
. Nevertheless, it is to be noted that the
proposed 2-SMC power control algorithm is independent of
the MPPT strategy adopted to establish the time-varying
reference value for P
s
. As a consequence, tip-speed ratio or
hill-climb searching strategies could equally be used to
dene P

s
[26].
In particular, the optimum power curve of a high-power
DFIG installed in an actual wind turbine is taken as starting
point. It was obtained by the manufacturer via off-line
experimentation and consists in an empirical function
expressing the total output power to be generated, P

t
, as a
cubic function of the rotor speed,
rm
. That function is then
scaled in power level to derive a plausible optimum power
curve for the 7 kW DFIG under consideration.
Bearing in mind that the following relation holds for a
DFIG
P
t
= P
s
+ P
r
(47)
where [18]
P
r
sP
s
=
v
s
v
r
v
s
P
s
=
v
s
Pv
rm
v
s
P
s
(48)
it turns out that
P
s

v
s
Pv
rm
P
t
(49)
Accordingly, based on the optimum power curve providing
P

t
, the rotor speed-dependent reference value for the
stator-side active power is computed as
P

s
=
v
s
P
P

t
v
rm
(50)
thus evidencing that the resulting P

s
is a quadratic function of
the rotor speed, as P

t
is, in turn, described as a cubic function
of
rm
.
In order to avoid setting unachievable reference values for
P
s
and Q
s
, it is required to observe the feasibility region of the
DFIG, represented by the shaded area of Fig. 6. The DFIG
feasibility region is obtained by intersection of two
semicircles. The dashed line delimiting one of those
semicircles displays the restriction because of the rated rotor
current, and its analytical expression is found to be [2]
P
2
s
+ Q
s

3|v
g
|
2
2L
s
v
s
_ _
2
=
3
2
L
m
L
s
|v
g
||i
r
|
R
_ _
2
(51)
On the other hand, the analytical expression for the
semicircumference in solid line is given by
P
2
s
+ Q
2
s
=
3
2
|v
g
||i
s
|
R
_ _
2
(52)
which corresponds to the constraint imposed by the rated
value of the stator current. Examination of the feasibility
region reveals the limited ability of this particular DFIG to
generate reactive power. In effect, Q
s
can be injected into
the grid only when generated P
s
is below 4.547 kW.
4.2 Experimental results
Aiming at showing some of the most illustrative results of the
global 2-SMC scheme put forward, a test is conducted where
the DFIG is subjected to the sequence of operation modes
reected in Table 3. Additionally, the 15 kW dc motor is
commanded so that, during the test, the DFIG rotor speed
evolves with time as shown in Fig. 7.
The time progress of the DFIG rotor speed has been
conceived with the aim of evaluating the proposed
synchronisation and power control algorithms under
Table 2 Specifications set to tune the 2-SMC arrangement,
and resulting parameter values
Specification Value Parameter Value

x
,
y
0.05 A c
x
, c
y
55.2381

x
,
y
1
x
,
y
5.4469

nx
,
ny
55.2381 rad/s w
x
, w
y
30.5811
10

P
35 W c
P
, c
Q
116

Q
15 VAr
P
1.5453 10
1

P
,
Q
1
Q
1.0116 10
1

nP
,
nQ
116 rad/s w
P
48.2032
10 w
Q
20.6585 Fig. 6 Feasibility region for the 7 kW DFIG
www.ietdl.org
IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551 547
doi: 10.1049/iet-rpg.2012.0026 & The Institution of Engineering and Technology 2013
realistic conditions. In an actual wind turbine, rotor speed
variations are expected to be inappreciable during
synchronisation, as it only takes a few cycles 105 ms in
this particular case. Considering this, rotor speed is
regulated around 1220 rpm when testing the synchronisation
scheme. However, it should be noted that the
synchronisation algorithm described in Section 3.2.1 is
designed to operate with no restriction on rotor speed.
Regarding the second part of the test, devoted to analyse
the presented power control algorithm, rotor speed is varied
in order that a signicant part of the optimum power curve
is to be tracked both up and downwards. For that purpose,
starting at second 9, rotational speed is accelerated from
1220 to 1620 rpm, and subsequently decelerated back to
1220 rpm from second 16 onwards. The acceleration and
deceleration displayed in Fig. 7 emulate, respectively, those
experienced by a DFIG installed in a real wind turbine of
H = 0.9346 s inertia constant while undergoing a sudden
increase followed by a sudden decrease of wind speed.
The initial rotor positioning stage is activated manually at
instant p, and it lasts 0.5 s, until instant s. The purpose of
this operation mode is to obtain an accurate estimate of the
rotor position at instant s,
r0
. From instant s on, the
encoder measures the increment of the rotor electrical angle
with respect to
r0
,
r
. Thus, the absolute rotor position
required for synchronisation and power control is derived as

r
=
r0
+
r
. The procedure described in [14] is followed
for initial rotor positioning.
Grid synchronisation is automatically launched at instant s.
Fig. 8 zooms at the beginning of the synchronisation process.
In particular, Fig. 8a displays the transient response of the
voltage induced at the terminals of the DFIG open stator,
showing that synchronisation with grid voltage is
accomplished at approximately the specied settling time of
105 ms. For clarity, only voltages corresponding to phase
A are represented. The v
ra
, v
rb
and v
rc
voltage components
shown in Fig. 8b are those supplied as inputs to the
PWM modulator driving the RSC transistors during
synchronisation. The resulting three-phase rotor current is
displayed in Fig. 8c.
After the lamps of the synchronoscope turn off, indicating
that grid synchronisation has been achieved, the DFIG stator
is manually connected to the grid at instant c. As a result of
the bumpless transfer between the 2-SMC algorithms in
charge of synchronisation and power control, grid
connection at zero power turns out to be smooth. This is
substantiated by Figs. 9a and b, which zoom, respectively,
at the active and reactive powers interchanged between the
DFIG stator and the grid around instant c. As expected,
they are both equal to zero prior to connection, and they are
kept within the ranges of 80100 W and 200250 VA
just after connecting the DFIG stator to the grid. The
three-phase rotor voltage and current leading to such a
Table 3 Operation modes of the DFIG during the test
Operation mode Time range, s
global control algorithm is idle 0p
initial rotor positioning ps; s = p + 0.5
grid synchronisation sc
connected to the grid with no power exchange c5
linear increase of generated power up to
3.752 kW
56
generation according to optimum power curve 923
Fig. 8 Grid synchronisation
a Grid and DFIG open-stator voltages
b Reference values of the three-phase rotor voltage fed into the PWM
modulator
c Three-phase rotor current
Fig. 7 DFIG rotational speed
Fig. 9 Smooth connection at zero power
a Active power
b Reactive power
c Reference values of the three-phase rotor voltage fed into the PWM
modulator
d Three-phase rotor current
www.ietdl.org
548 IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551
& The Institution of Engineering and Technology 2013 doi: 10.1049/iet-rpg.2012.0026
smooth connection are those shown in Figs. 9c and d,
respectively. If the manoeuvre of connecting the DFIG
stator to the grid at zero power was ideal, both the rotor
voltage and current previous to instant c would remain
unchanged after connection.
As evidenced by Fig. 10, power generation is started at
second 5. Between seconds 5 and 6, P

s
is raised linearly
from 0 to 3.752 kW value which, according to (50),
corresponds to the optimum stator-side power for a 1220
rpm rotational speed and established as dictated by (50)
for the remaining time. In view of the feasibility region
depicted in Fig. 6, Q

s
is set in order that the DFIG operates
with a 0.81 lagging inductive power factor all
through the test. Figs. 10a and b evidence the excellent
performances of P
s
and Q
s
when tracking their respective
time-varying reference values. In the worst case, while
operating at supersynchronous speeds, chatter in reactive
power approximately reaches 3% of the 7 kW rated
power. However, during the rest of the experiment, chatter
present in P
s
and Q
s
represents < 1.5%.
Aiming at further assessing the dynamic performance of
the proposed power control scheme, the experiment
discussed above is repeated, but dening Q

s
as the
succession of 1 kVAr steps displayed in Fig. 11b. The
different set-points for Q
s
are carefully selected so that
the resulting (P
s
, Q
s
) operating points fall inside the DFIG
feasibility region shaded in Fig. 6. Fig. 11b conrms the
outstanding dynamic response of Q
s
when reacting to
sudden set-point changes. Moreover, as shown by Fig. 11a,
tracking of P

s
is not affected by the steps in reactive power.
The reection of those reactive power steps as abrupt
variations in rotor current is clearly observable by
examining Figs. 11b and d.
In order to evaluate the robustness of the proposed global
control scheme, the two tests above are repeated once
x, y
,
w
x, y
,
P, Q
and w
P, Q
are re-tuned by deliberately assuming
incorrect values for L
m
, L
s
and L
r
. In particular, L
m
is
considered to be a 50% higher than its actual value in
Table 1, which in turn leads to adopt incorrect values for
Fig. 10 Tracking of optimum stator-side power and Q

s
a Active power
b Reactive power
c Reference values of the three-phase rotor voltage fed into the PWM
modulator
d Three-phase rotor current
Fig. 11 Tracking of optimum stator-side power and reactive power
steps
a Active power
b Reactive power
c Reference values of the three-phase rotor voltage fed into the PWM
modulator
d Three-phase rotor current
www.ietdl.org
IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551 549
doi: 10.1049/iet-rpg.2012.0026 & The Institution of Engineering and Technology 2013
both L
s
and L
r
see their L
m
-dependent expressions in
Table 1. Moreover, not only those inductances but also R
r
is deliberately assumed to be incorrect a 50% lower than
its actual value when computing equivalent control terms
(16), (17), (32) and (33).
The main results of the rst robustness experiment are
shown in Figs. 12 and 13. The grid synchronisation detailed
by Fig. 12a, as well as the active and reactive power
tracking displayed in Figs. 13a and b, remain highly
satisfactory in spite of the aforementioned 50% mismatches
in L
m
and R
r
, hence supporting the robustness of the
proposed global 2-SMC scheme. In addition, the low-power
ow between the DFIG stator and the grid reected by
Figs. 12b and c at instant c corroborates that smoothness of
connection is also preserved. Rotor voltage and current
signals have not been provided to avoid reiteration with no
signicant added information.
To conclude, the second robustness experiment consists in
repeating the test whose results are provided in Fig. 11 under
the afore-cited parameter mismatches. Its most representative
results are shown in Fig. 14, where a still excellent tracking of
both the optimum stator-side power and the reactive power
steps can be observed.
5 Conclusions
A PWM-based 2-SMC scheme both for grid synchronisation
and power control of a DFIG has been presented. It
allows keeping the tracking accuracy and robustness
features characteristic of standard SMC, while leading to a
xed switching frequency of the RSC transistors.
Experimentation conducted on a 7 kW DFIG test bench
proves that high dynamic performance control and superior
robustness against DFIG parameter variations are achieved
when applying the proposed global 2-SMC scheme. In
addition, bumpless transfer between the grid
synchronisation and power control operating regimes is
guaranteed, which results in smooth connection of the
DFIG stator to the grid.
A systematic methodology to tune all the parameters
involved in the presented 2-SMC realisation is also
provided. Its main benet lies in deriving the values of
those parameters via direct application of tuning equations,
hence eluding the generally time-consuming task of
trial-and-error adjustment suggested in [16], which is still
the most usual practice. The satisfactory experimental
results obtained support the effectiveness of the proposed
tuning method.
6 Acknowledgments
The authors thank Giovanna Santamara, from Jema Energy,
for her helpful advice. We are also grateful to Manu
Larramendi and J. Ignacio Susperregui for their help in
Fig. 12 Synchronisation and grid connection under 50%
mismatches in L
m
and R
r
a Synchronisation
b Active power
c Reactive power
Fig. 13 Tracking of optimum stator-side power and Q

s
under 50%
mismatches in L
m
and R
r
a Active power
b Reactive power
Fig. 14 Tracking of optimum stator-side power and reactive power
steps under 50% mismatches in L
m
and R
r
a Active power
b Reactive power
www.ietdl.org
550 IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551
& The Institution of Engineering and Technology 2013 doi: 10.1049/iet-rpg.2012.0026
modifying the 7 kW DFIG test bench. This work has been
developed within the Intelligent Systems and Energy (SI+E)
research group of the University of the Basque Country UPV/
EHU, and has been funded by the Spanish Ministry of
Economy and Competitiveness, under project code
DPI2012-37363-C02-01 and grant code BES-2008-002563,
as well as by the Basque Government (Spain), under research
grant IT677-13, and the UPV/EHU, under unit of formation
and research UFI11/28.
7 References
1 Hopfensperger, B., Atkinson, D.J., Lakin, R.A.: Stator-ux oriented
control of a doubly-fed induction machine without position encoder,
IEE Proc. Electr. Power Appl., 2000, 147, (4), pp. 241250
2 Tapia, A., Tapia, G., Ostolaza, J.X., Senz, J.R.: Modeling and control
of a wind turbine driven doubly fed induction generator, IEEE Trans.
Energy Convers., 2003, 18, (2), pp. 194204
3 Pea, R., Clare, J.C., Asher, G.M.: Doubly fed induction generator
using back-to-back PWM converters and its application to
variable-speed wind-energy generation, IEE Proc. Electr. Power
Appl., 1996, 143, (3), pp. 231241
4 Xu, L., Cartwright, P.: Direct active and reactive power control of DFIG
for wind energy generation, IEEE Trans. Energy Convers., 2006, 21,
(3), pp. 750758
5 Susperregui, A., Tapia, G., Zubia, I., Ostolaza, J.X.: Sliding-mode
control of doubly-fed generator for optimum power curve tracking,
Electron. Lett., 2010, 46, (2), pp. 126127
6 Tapia, G., Tapia, A.: Wind generation optimisation algorithm for a
doubly fed induction generator, IEE Proc. Gener. Transm. Distrib.,
2005, 152, (2), pp. 253263
7 Zhi, D., Xu, L.: Direct power control of DFIG with constant switching
frequency and improved transient performance, IEEE Trans. Energy
Convers., 2007, 22, (1), pp. 110118
8 Beltran, B., Ahmed-Ali, T., Benbouzid, M.E.H.: High-order
sliding-mode control of variable-speed wind turbines, IEEE Trans.
Ind. Electron., 2009, 56, (9), pp. 33143321
9 Ben Elghali, S.E., Benbouzid, M.E.H., Ahmed-Ali, T., Charpentier, J.F.,
Mekri, F.: High-order sliding mode control of DFIG-based marine
current turbine. Proc. 34th Annual Conf. on IEEE Industrial
Electronics (IECON 2008), Orlando, USA, November 2008, pp.
12281233
10 Hu, J., Nian, H., Hu, B., He, Y., Zhu, Z.Q.: Direct active and reactive
power regulation of DFIG using sliding-mode control approach, IEEE
Trans. Energy Convers., 2010, 25, (4), pp. 10281039
11 Young, K.D., Utkin, V.I., zgner, .: A control engineers guide to
sliding mode control, IEEE Trans. Control Syst. Technol., 1999, 7,
(3), pp. 328341
12 Blaabjerg, F., Teodorescu, R., Liserre, M., Timbus, A.V.: Overview of
control and grid synchronization for distributed power generation
systems, IEEE Trans. Ind. Electron., 2006, 53, (5), pp. 13981409
13 Arbi, J., Ghorbal, M.J.B., Slama-Belkhodja, I., Charaabi, L.: Direct
virtual torque control for doubly fed induction generator grid
connection, IEEE Trans. Ind. Electron., 2009, 56, (10), pp. 41634173
14 Tapia, G., Santamara, G., Telleria, M., Susperregui, A.: Methodology
for smooth connection of doubly fed induction generators to the grid,
IEEE Trans. Energy Convers., 2009, 24, (4), pp. 959971
15 Peresada, S., Tilli, A., Tonielli, A.: Power control of a doubly fed
induction machine via output feedback, Control Eng. Pract., 2004,
12, (1), pp. 4157
16 Bartolini, G., Ferrara, A., Levant, A., Usai, E.: On second order sliding
mode controllers, in Young, K.D., zgner, . (Eds.): Variable
structure systems, sliding mode and nonlinear control (Springer
Verlag, 1999, 1st edn.), pp. 329350
17 Vas, P.: Sensorless vector and direct torque control (Oxford University
Press, 1998, 1st edn.)
18 Tazil, M., Kumar, V., Bansal, R.C., et al.: Three-phase doubly fed
induction generators: an overview, IET Electr. Power Appl., 2010, 4,
(2), pp. 7589
19 Utkin, V., Guldner, J., Shi, J.: Sliding mode control in
electromechanical systems (Taylor & Francis, 1999, 2nd edn.)
20 Levant, A.: Sliding order and sliding accuracy in sliding mode control,
Int. J. Control, 1993, 58, (6), pp. 12471263
21 Rashed, M., Goh, K.B., Dunnigan, M.W., MacConnell, P.F.A.,
Stronach, A.F., Williams, B.W.: Sensorless second-order
sliding-mode speed control of a voltage-fed induction-motor drive
using nonlinear state feedback, IEE Proc. Electr. Power Appl., 2005,
152, (5), pp. 11271136
22 Pea, R., Crdenas, R., Proboste, J., Asher, G., Clare, J.: Sensorless
control of doubly-fed induction generators using a rotor-current-based
MRAS observer, IEEE Trans. Ind. Electron., 2008, 55, (1),
pp. 330339
23 Zubia, I., Zatarain, A., Alcalde, C., Ostolaza, X.: In situ electrical
parameter identication method for induction wind generators, IET
Electr. Power Appl., 2011, 5, (7), pp. 549557
24 Lin, J.L.: A new approach of dead-time compensation for PWM voltage
inverters, IEEE Trans. Circuits Syst. I, Fundam. Theory Appl., 2002,
49, (4), pp. 476483
25 strm, K.J., Hgglund, T.: PID controllers: Theory, design and
tuning (Instrument Society America, 1995, 2nd edn.)
26 Wang, Q., Chang, L.: An intelligent maximum power extraction
algorithm for inverter-based variable speed wind turbine systems,
IEEE Trans. Power Electron., 2004, 19, (5), pp. 12421249
www.ietdl.org
IET Renew. Power Gener., 2013, Vol. 7, Iss. 5, pp. 540551 551
doi: 10.1049/iet-rpg.2012.0026 & The Institution of Engineering and Technology 2013

You might also like