You are on page 1of 15

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO.

3, MAY/JUNE 2013

1343

Determination of Steady-State and Dynamic Control


Laws of Doubly Fed Induction Generator Using
Natural and Power Variables
Adeola Balogun, Member, IEEE, Olorunfemi Ojo, Fellow, IEEE,
Frank Okafor, Member, IEEE, and Sosthenes Karugaba, Member, IEEE

AbstractA doubly fed induction generator model is presented


whereby the natural and power variables are the state variables.
The natural variables are the electromagnetic torque (Te ), the
reactive torque (Tr ), the magnitude of the rotor flux linkage
(r ), the magnitude of the stator flux linkage (s ), and the
rotor speed (r ). The power variables are the real power (Pf )
and reactive power (Qf ) generated/absorbed by the grid-side
converter into/from the grid. Simulation of the dynamic natural
variable model of the induction machine is compared with a vector
variable simulation. Steady-state operating regions are established
for various power factor operations. The optimal stator power
factor operation is estimated. A direct control of torque and power
variables is developed. The robustness of the developed control
against rotor parameter variation is investigated using small signal
analysis and is compared with vector control. Results are shown
for a 5-hp machine.
Index TermsDoubly fed induction generator (DFIG), natural
variables, power variables, robustness, stator power factor, wind
energy conversion.

I. I NTRODUCTION

HE CONVENTIONAL ways of modeling induction machines entail obtaining the machine model in terms of
its flux linkages or rotor and stator currents as state variables,
as given in [1] and [2]. These variables in qd, however, do
not remain the same in all reference frames of transformation,
which implies that they differ from one reference frame to
the other. The reasons for this can be derived from the fact
that the reference frequency , which determines the angle
of transformation , is inherent in such models. Hence, the
qd state variables are dependent on choice of reference frame
Manuscript received September 12, 2011; revised August 12, 2012; accepted
August 30, 2012. Date of publication March 20, 2013; date of current version
May 15, 2013. Paper 2011-EMC-494.R1, presented at the 2010 IEEE Energy
Conversion Congress and Exposition, Atlanta, GA, USA, September 1216,
and approved for publication in the IEEE T RANSACTIONS ON I NDUSTRY
A PPLICATIONS by the Electric Machines Committee of the IEEE Industry
Applications Society.
A. Balogun and F. Okafor are with the Department of Electrical and
Electronics Engineering, University of Lagos, Lagos 101017, Nigeria (e-mail:
balog975@yahoo.com; cfrankok@yahoo.com).
O. Ojo is with the Center for Energy Systems Research, Department of
Electrical and Computer Engineering, Tennessee Technological University,
Cookeville, TN 38505 USA (e-mail: jojo@tntech.edu).
S. Karugaba is with the Department of Electrical Engineering, Dar es
Salaam Institute of Technology (DIT), Dar es Salaam, Tanzania (e-mail:
karugaba@gmail.com).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TIA.2013.2253532

of transformation. However, these state variables are invariant


after inverse transformation back to the abc reference frame.
Such models usually form the basis of various vector control
schemes for induction machine applications, which are evident
in [3][8]. In vector control schemes, the torque and power
quantities of the doubly fed induction generator (DFIG) are
regulated indirectly by some inner loop current control used
to generate the corresponding rotor voltage vectors. Because
vector control schemes are generally known to be dependent
on system parameters [8], alternative control schemes are considered in literature to avoid such dependences. Direct scalar
control schemes such as direct torque control (DTC) and direct
power control (DPC) have been alternative choices of interest
which are independent on system parameters. Conventional
DTC and DPC use nonlinear controllers such as hysteresis
regulators to control the torque and power directly [9][15].
Although some efforts have been made in literature to make the
switching frequency of the DTC and DPC techniques constant
[9], [12], tables of optimal switching are still usually required.
Models of squirrel-cage induction machines developed in
[16][18] had the torque quantities as state variables. In [17]
and [18], the model termed the natural variable was used to
achieve controller design for efficiency optimization. In [19],
a wound rotor induction machine of a DFIG was presented
in terms of torque variables, but details were not given about
modeling the grid-side converter (GSC) having power as state
variables. The multivariable dynamic model of the voltage
source converter presented in [20] had the power quantity as
the state variables. However, the converter was not connected
back to back to another converter.
In this paper, therefore, the methods of [16][21] are adopted
such that inputoutput linearization is used to linearly relate
the torque and power outputs of the DFIG to the rotor voltage
vectors. The resultant model is referred to as the natural/power
variable model because the state variables are the torque variables of the induction machine, the power variables of the GSC,
and the rotor speed r . Since the natural and power variables
are scalar quantities, they do not change with respect to change
in angle of reference frame transformation . Hence, they are
insensitive to the reference frame angular velocity . Therefore,
such a model can be effectively utilized in developing direct
control of torque and power schemes, which do not require any
form of inner loop current regulation. Another advantage is that,
since a linear relationship exists between the output variable

0093-9994/$31.00 2013 IEEE

1344

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 3, MAY/JUNE 2013

and the respective rotor voltage vector, then the switching of


the rotor voltage quantities becomes symmetrical.
II. I NDUCTION M ACHINE M ODEL
Fig. 1 shows a schematic representation of a DFIG driven by
a mechanical turbine via a gearbox. The stator and rotor voltage
equations of the induction machine in qd arbitrary reference
frame [1] are given in (1)(8). The model accounts for the core
loss with a core loss resistance rc , as shown in Appendix A,
such that kc = 1 + (rs /rc )
vqs = rs iqs + kc pqs + kc ds

(1)

vds = rs ids + kc pds kc qs

(2)

vqr = mqr vdc /2 = rr iqr + pqr + ( r )dr

(3)

vdr = mdr vdc /2 = rr idr + pdr ( r )qr

(4)
Fig. 1.

where
qs = Ls iqs + Lm iqr

(5)

ds = Ls ids + Lm idr

(6)

Lm iqs

(7)

qr = Lr iqr +

dr = Lr idr + Lm ids .

(8)

The qd components of the stator voltage (vqs and vds ) and the
rotor voltage (vqr and vdr ) in (1)(4) are the transformed values
of the abc reference frame stator voltage and rotor voltage, respectively. The stator voltage in (1) and (2) is expressed in terms
of the stator current (iqs and ids ), stator flux linkage (qs and
ds ), and the stator reference speed (). Similarly, in (3) and
(4), the rotor voltage is likewise expressed in terms of the rotor
current (iqr and idr ), rotor flux linkage (qr and dr ), stator
reference speed (), and r , which is the rotor speed. Ls and
Lr represent the stator and rotor self-inductances, respectively,
and rs and rr represent the stator and rotor resistances, respectively, while Lm is a value 3/2 the magnetizing inductance [1].
Observe in (1)(4) that the operator p = d/dt. A change of
variable from the abc reference frame to the arbitrary qd reference frame is achieved using the transformation matrix given
in (13) and (15) for the stator and rotor circuit variables, respectively. In (9)(12), fqd and fabc represent the voltage, current, and flux variables in the qd and abc reference frames,
respectively. The transformation angle given in (14) is in its
definite integral form and is defined in terms of an arbitrary
electrical speed . The angular displacement of the rotor r is
defined in terms of the rotor electrical speed r by (17)
fqds = Ts ()fabcs

(9)

fqdr = Tr ()fabcr

(10)

DFIG for wind power applications.

All of the variables defined in (1)(4) change from one reference frame to the other. However, the electromagnetic torque
Te , reactive torque Tr , stator flux linkage magnitude s , rotor
flux linkage magnitude r , and rotor speed r remain the same
in every reference frame, and that is the reason why they are
referred to as natural variables





2 cos cos 2
cos  + 2
3
3



(13)
Ts () =
sin + 2
3 sin sin 2
3
3
t
=

()d + (0)
0


2 cos
Tr () =
3 sin



cos  2
3 
sin 2
3

(14)


cos  + 2
3 
(15)
sin + 2
3

where
= r
r =

dr
.
dt

(16)
(17)

When (1)(4) are rearranged in terms of the flux linkages as


state variable equations, a conventional simulation model in
statespace form is obtained as given in
pqs = vqs rs iqs ds

(18)

pds = vds rs ids + qs

(19)

pqr = vqr rr iqr ( r )dr

(20)

pdr = vdr rr idr + ( r )qr .

(21)

III. NATURAL /P OWER VARIABLE M ODEL OF DFIG


where
(fqds )T = [fqs fds ]
T

(fqdr ) = [fqr fdr ].

(11)
(12)

The outputs of the DFIG that are generally of interest to


control are the following: the electromagnetic torque (Te ), the
reactive torque (Tr ), the stator real power (Ps ), the stator
reactive power (Qs ), the GSC real power (Pf ), and the GSC

BALOGUN et al.: DETERMINATION OF STEADY-STATE AND DYNAMIC CONTROL LAWS OF DFIG

reactive power (Qf ). These outputs vary nonlinearly with the


magnitude of the rotor voltage of the DFIG. However, since
DFIG can be regarded as a multi-inputmulti-output (MIMO)
system, then a feedback (inputoutput) linearization theory
described in [22] for a MIMO system can be adopted. The
theory summarily entails differentiating the outputs until the
control inputs appear.
A. Natural Variable Model of Induction Machine and MSC
The stator flux linkage referred equations for the electromagnetic torque, reactive torque, and square of the magnitude of
stator flux linkage magnitude (ss ) are given in
Te = k(ds iqs qs ids )

(22)

Tr = k(ds ids + qs iqs )

(23)

ss = 2qs

2ds .

Te = ko (dr iqs qr ids )

(25)

Tr = ko (dr ids + qr iqs )

(26)

rr = 2qr + 2dr

L1 pids = vdr
r

rr
Lr
qs + rl iqs
vqs
Lm
Lm k c

Lr
ds ( r )L1 ids
Lm

(28)

rr
Lr
ds + rl iqs
vds
Lm
Lm k c

Lr
qs + ( r )L1 iqs
Lm

pTe = k

(29)

ds L ids
L

vqs

kc

qs L iqs
L

Lm kds
Lm kqs
vqr +
vdr
L Lr
L Lr
 

qs + L iqs vqs
pTr = k
L
kc




ds + L ids vds
rr

ss
L
kc
L Lr

rs  2
rT
T + Tr2
Tr + r Te +

L
kss e

Lm kds
Lm kqs
vqr +
vdr
L Lr
L Lr


vqs
vds
2rs
= 2 qs
+ ds
Tr
+
kc
kc
kkc
+

pss

pr =

vds
kc

kr
rT
Te r Tr
ss
L
L

(32)
(33)

where
rl =

vqr =

rr Ls
Lr rs
+
Lm
Lm k c

L1 = L m

Lr Ls
Lm

mqr vdc
2

(30)

(31)

P
(Te Tm )
2J

(27)

where k = 3P/4 and ko = 3P Lm /4Lr .


However, the stator flux linkage referred model is considered
throughout this paper. Both stator flux linkage referred and rotor
flux linkage referred models were given in [21].
The method of obtaining the natural variable model in a
statespace form in the stator reference frame stems from
differentiating (22)(24) until the rotor voltage vectors appear.
If the rotor flux linkage and the rotor current are eliminated in
(20) and (21) using (5)(8) and simplification is made with (18)
and (19), then (28) and (29) are obtained such that the stator
currents are the state variable. Hence, combining (18), (19),
(28) and (29) into the derivative of (22)(24) enables the qd
rotor voltage vectors to appear. Hence, (30)(32) are obtained.
The q-axis and d-axis components of the rotor voltage in (30)
and (31) represent half the product of the respective modulation
index (mqr and mdr , respectively) of the machine-side converter (MSC) and the dc-link voltage (vdc ). This incorporates
the switching actions of the MSC into the machine dynamics.
The rotor speed dynamics is given in (33) to form a complete
model in natural reference frame. In (33), Tm represents the
mechanical load torque
L1 piqs = vqr



(24)

Similarly, the rotor flux linkage referred equations for Te , Tr ,


and rr (square of the magnitude of rotor flux linkage) are
given in

1345

rT =

rs
rr Ls
+
kc
Lr

L = Ls

vdr =

L2m
Lr

mdr vdc
.
2

B. Power Variable Model of GSC


The voltage equation at the output of the GSC across the filter
network toward the point of shunt injection of current into the
grid is given as
vqf = mqf vdc /2
= rf iqf + Lf piqf + Lf idf + nvqL

(34)

vdf = mdf vdc /2


= rf idf + Lf pidf Lf iqf + nvdL
3
Cd pvdc = (io id )
2
3
= (mqr iqr + mdr idr mqf iqf mdf idf ).
4

(35)

(36)

In (34) and (35), the inverter voltages vqf and vdf are expressed in terms of the filter network (rf , Lf ), the current (iqf
and idf ) injected through the shunt transformer into the grid,
and the voltage at the grids point of common coupling (vqL and
vdL ). The q-axis and d-axis components of the GSC ac voltage
modulation index are mqf and mdf , respectively. Equation (36)
establishes Kirchhoffs current law at the dc-link between the
two converters in terms of the dc capacitor current, the MSC
dc output current (io ), and the GSC dc input current (id ). Cd is

1346

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 3, MAY/JUNE 2013

the capacitance of the dc-link capacitor. In (34)(36), the real


power and reactive power injected in shunt into the grid by the
GSC are as follows:
3
Pf = n(vqL iqf + vdL idf )
2
3
Qf = n(vdL iqf vqL idf ).
2

(37)
qs = s
(38)

The power model of the GSC is obtained by differentiating (37)


and (38) and making appropriate substitutions and simplifications, where n in (34)(38) represents the turns ratio of the
shunt injection transformer. If the stator terminal connection
to the grid is , for example, then vqs = vqL and vds =
vdL in (37) and (38). Rearranging (37) and (38) results in
(39) and (40)
2
(vqs Pf + vds Qf )
3nvss
2
=
vds Pf Vqs Qf )
3nvss

iqf =

(39)

idf

(40)

2
2
+ vds
.
where vss = vqs
Substitute (39) and (40) in (34) and (35), and then, rearrange
yield (41) and (42). If (39)(42) are used to make appropriate
substitutions in the time derivatives of (37) and (38)

Lf piqf =

2rf
mqf vdc

(vqs Pf + vds Qf )
2
3nvss

2e Lf
(vds Pf vqs Qf ) nvqs
3nvss
2rf
mdf vdc

=
(vds Pf vqs Qf )
2
3nvss

Lf pidf

2e Lf
(vqs Pf + vds Qf ) nvds
3nvss

(41)

(42)

then (43) and (44) evolve as the dynamic power model of


the GSC, with the real power and reactive power as the state
variables
pPf =
+
pQf =

rf
3n2 vss
3mqf vdc nvqs
Pf + e Q f
+
Lf
2Lf
4Lf
3mdf vdc nvds
4Lf

(43)

rf
3mqf vdc nvds
Q f e Pf +
Lf
4Lf
3mdf vdc nvqs
.
4Lf

flux linkage magnitude so that the d-axis component becomes


zero, given in (45). Such orientation is similar to the method
of alignment done in vector control. Hence, with appropriate
substitution and simplification of (36), (46) evolves

pvdc

ds = 0
(45)




Ls T e
3mqr qs
Ls T r
3mdr
=
+
+
4Cd Lm
Lm qs k
4Cd Lm qs k
+

1
[Pf (mqs vqs + mds vds )
2Cd nvss
+Qf (mqs vds mds vqs )] .

IV. O PEN -L OOP S IMULATION OF I NDUCTION M ACHINES


For the purpose of simulation, the rotor of the wound rotor
induction machine is electrically shorted so that it acts like a
squirrel-cage induction machine. Hence, the rotor voltage is
zero. The machine parameters are given in Appendix A.
A. Vector Variable (Conventional) Simulation
The vector variable model of (18)(21) along with (32) is
simulated such that the stator flux linkage and rotor flux linkage
are the state variables. Observe that the reference velocity is
evident in this model, which implies that the qd components
of the state variables are dependent on the choice of the reference frame of transformation. The implication of this is that
they are not bound to the same orientation or angular direction
in the different reference frames.
B. Natural Variable Simulation
In order to ease the simulation of the natural variable model
of (30)(33), it is essential, therefore, to align the stator flux
linkage along the q-axis component flux linkage such that the
d-axis component is equal to zero (45). An angular velocity
given in (47), which is obtained by substituting (45) into (19),
gives such an orientation. Integrating (47) gives the precise angle of transformation. Recall that the transformation angular
displacement in definite integral is given in (14). Consequently,
the qd stator currents are given in
1
(vds rs ids )
s

(47)

iqs =

Tr
ks

(48)

ids =

Te
.
ks

(49)

=
(44)

(46)

C. DC-Link Dynamics
The MSC and the GSC are linked by (36). The MSC dc
output current (io ) and the GSC dc input current (id ) are
expressed in terms of rotor current with MSC modulation index
and filter current with GSC modulation index, respectively. For
ease of analysis, the stator flux linkage is aligned such that
the q-axis of the stator flux linkage corresponds to the stator

C. Open-Loop Simulation Results


MATLAB/Simulink is employed for simulation. Observe
that the electrical speed of 376.9911 rad/s is thrice the mechanical speed at 60 Hz because the induction machine is a six-pole
machine.

BALOGUN et al.: DETERMINATION OF STEADY-STATE AND DYNAMIC CONTROL LAWS OF DFIG

Fig. 2.

No-load transient of the electromagnetic torque.

Fig. 3.

No-load transient of the reactive torque.

1347

Fig. 4. No-load transient of the square of the magnitude of the stator flux
linkage.

Fig. 5. No-load transient of the stator angular velocity in radians per second.

The results obtained from simulating the vector variable


model and the natural variable model are presented in Figs. 29.
Observe that the results from the two models overlap one
another, which validates that the natural variable model can
also be effectively used to predict the electrical dynamics of the
machine. No-load transients of the electromagnetic torque for
both simulations are shown in Fig. 2 to decay to zero. Figs. 35
show the transients of the reactive torque, the square of the
magnitude of the stator flux linkage, and the stator angular
velocity.
At steady state, some step changes are introduced by the
mechanical torque. At 2 and 4 s, 16 and 16N m are applied,
respectively, to compare the responses of both models. In
Figs. 69, the corresponding changes in the rotor speed, electromagnetic torque, reactive torque, and stator flux linkage are
observed to be the same for both models. In Fig. 7, the
electromagnetic torque was positive for 16-N m change
in load torque and negative for 16N m change in load
torquerespectively, motoring and generating modes as
expected.
The reactive torque, like the reactive power as regards to
power, is the inactive torque component. In fact, it is related to
the reactive power by a factor of 2/P , where P is the number
of magnetic poles of the machine. The sign convention adopted

Fig. 6. Stepped response of the rotor electrical speed.

for this paper is such that a negative reactive torque implies that
reactive torque is supplied to the machine. Hence, in Figs. 3 and
8, the reactive torque remains negative irrespective of change
in load (mechanical) torque because the machine operates at
lagging power factor. Fig. 5 shows the stator frequency which
is computed to align the stator flux. Such alignment ensures that
the d-axis stator flux remains zero at all time. Fig. 9 shows the

1348

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 3, MAY/JUNE 2013

Fig. 7. Stepped response of the electromagnetic torque.

Fig. 8. Stepped response of the reactive torque.

tor frequency, (30)(33) and (43)(46) translate to (50)(56),


respectively,



qs
Vds
Tr
Te Vqs

+
0 =k
kqs kc
L
kqs kc


Lm kqs Mdr Vdc
Te
kr
r Tr
ss +
rT
(50)
L
L
L Lr
2

 
Vqs
qs
Tr
Te Vds
rr

ss
0 =k
L
kqs kc
kqs kc
L Lr

Tr
rs  2
rT
Te + Tr2
+ r Te +
L
k

ss
Lm kqs Mqr Vdc
+
(51)
L Lr
2


Vqs
rs
+
Tr
(52)
0 = 2 qs
kc
kkc
P
(Te Tm )
(53)
0=
2J
2
3n Vss
rf
Pf + e Q f
0=
Lf
2Lf
3Mqf Vdc nVqs
3Mdf Vdc nVds
+
+
(54)
4Lf
4Lf
3Mqf Vdc nVds
rf
Q f e Pf +
0=
Lf
4Lf
3Mdf Vdc nVqs

(55)
4Lf




Ls T e
Ls T r
3Mqr qs
3Mdr
+
0=
+
4
Lm
Lm qs k
4
Lm qs k
1
[Pf (Mqs Vqs + Mds Vds )
+
2nVss
(56)
+Qf (Mqs Vds Mds Vqs )] .
A. Power Balance
Equation (56) establishes power balance between the two
converters in terms of the modulation indexes and the state
variables. In general, it is desired to operate the GSC at unity
power factor to improve efficiency [4]. For such a situation,
Qf = 0, which means that the GSC does not consume or
generate reactive power from or into the grid. Hence, (54)(56)
translate to (57)(59)
rf
3n2 Vss
3Mqf Vdc nVqs
Pf
+
Lf
2Lf
4Lf
3Mdf Vdc nVds
+
4Lf
3Mqf Vdc nVds
3Mdf Vdc nVqs

0 = e Pf +
4Lf
4Lf


3Mqr qs
Ls T r
3Mdr Ls Te
0=
+
+
4
Lm
Lm qs k
4Lm qs k
Pf
(Mqs Vqs + Mds Vds )
+
2nVss




2
2
Te
Tr
3
s Te
Ps =
rs +
rs +
2
qs k
qs k
k


3 Tr
2s Tr
s =
Qs =
2 k
P
2kQG
Tr =
3s

0=

Fig. 9. Stepped response of the stator flux linkage.

stepped response of the stator flux linkage which is the square


root of the magnitude of ss .
V. S TEADY-S TATE O PERATING L IMITS
At steady state, all derivatives become zero. Hence, when
the reference frame is the synchronous speed, i.e., at sta-

(57)
(58)

(59)
(60)
(61)
(62)

BALOGUN et al.: DETERMINATION OF STEADY-STATE AND DYNAMIC CONTROL LAWS OF DFIG

2rs QG
3qs s




2rs QG 2
2 =
Vds = Vss Vqs
Vss
3qs s


T

2r
s e s
s2 2ss + Vss
ss
k

2
2 
rs Te
2rs QG

=0
k
3qs s

b b2 4a c
ss =
2a

Vqs =

where
a = s2


2rs Te s
b = Vss
k

2
2 
rs Te
2rs QG

.
c =
k
3s

1349

(63)
(64)

Mqr


2L Lr
=
Lm kqs Vdc
  

qs
Vqs
Tr
Te Vds
rr
k

ss
L
kqs kc
kqs kc
L Lr


Tr
rs  2
rT
Te + Tr2 .
+ r Te +
(68)
L
kss

(65)
(66)

B. Unity Power Factor Operation


The DFIG can operate at a general unity power factor with
the stator and GSC set at unity power factor. Such unity power
factor operation will require that the MSC supplies the machine
with the reactive power needed for excitation. The machine
parameters for this study are given in Appendix A. Fig. 10
shows the turbine power, the stator power, the rotor power,
and the sum of the stator power and rotor power at general
unity power factor operation of stator and GSC. The difference
between the turbine power and the sum of the stator power
and rotor power is accounted by the copper and core losses
of the machine, assuming that all other losses were neglected
for the purpose of analysis. In Figs. 1117, the unity power
factor is compared to leading and lagging stator power factors.
It is clearly shown, therefore, that a general unity power factor
operation requires a higher power rating of MSC than lagging
stator power factor but a lower MSC power rating for leading
stator power factor

bmds b2mds 4amds cmds
(69)
Mds =
2amds

When the stator real power (Ps ) and reactive power (Qs ) are
expressed in terms of the natural variables, then (60) and (61)
evolve. If the stator reactive power is equated to the grid reactive
power (QG ) because the GSC is set not to exchange reactive
power with the grid, then (62) is obtained. Substituting (62)
into (52) yields the q-axis of the stator voltage, which is given
in (63). The d-axis of the stator voltage is given in (64), from
which (65) is obtained by using (49) to substitute for the d-axis
stator current. Therefore, the stator flux linkage is determined
from the square root of (66) while specifying the grid reactive
power. Hence, appropriate substitution and simplification in
(50) and (51) and (57)(59) yield (67)(71). The qd vectors where
of the MSC modulation indexes are given in (67) and (68),
2
respectively. Similarly, the qd vectors of the GSC modulation amds = 3rf Vdc Vss
2
2
4e Lf
indexes are given in (69) and (70), respectively. Equation (71)


gives the real power of the GSC. Observe that if (59) is
2
rf Vds
3Vss
multiplied by Vdc , then the first two expressions after the bmds =
+ Vqs
and
2e Lf e Lf
equality sign in terms of Mqr and Mdr represent the rotor

2
2
real power, which is given in (72). The rotor reactive power
rf Vds
Vqs
3Vss
+
+ Vqs
is given in (73). Equation (74) gives the mechanical torque as cmds = 2 L
e Lf
e f
a quadratic function of the shafts mechanical speed m (in




3Vss Mqr qs
Ls T r
3Vss Mdr Ls Te
radians per second) [13], where kopt is a constant which is

+
+
2
Lm
Lm qs k
2Lm qs k
obtained from the mechanical torque versus shaft speed curve


of associated wind turbine. Hence, the turbines mechanical
rf Vqs
2Vss
shaft power Pm of (75) in terms of Tm and m is further
e Lf + Vds Mds + Vdc


Mqs =
(70)
expressed in terms of Te , damping coefficient Bm , and rotor
rf Vds
+
V
qs

L
e f
electrical speed r (in radians per second), where m = 2r /P
3Vss Vdc
3Vss Vds
and P = number of poles of the machine. The copper and core
4e Lf Mds + 2e Lf
power losses of the generator are given in (76)


(71)
Pf =
rf Vds
+
V
qs


e Lf
2L Lr


Mdr =
3Mqr Vdc qs
Ls T r
3Mdr Vdc Ls Te
Lm kqs Vdc
(72)
+
Pr =
+
 


4
Lm
Lm qs k
4Lm qs k
Te Vqs
qs
Vds
Tr


k

+
3Mdr Vdc qs
Ls T r
3Mqr Vdc Ls Te
kqs kc
L
kqs kc
(73)
+
Qr =


4
Lm
Lm qs k
4Lm qs k
Te
kr
rT
r Tr
ss
(67)
2
Tm = kopt m
(74)
L
L

1350

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 3, MAY/JUNE 2013

Fig. 10. Turbine power, stator power, rotor power, and sum of stator power
and rotor power at unity power factor operation of stator and GSC.

Fig. 11. Rotor voltage versus rotor electrical speed for unity, leading, and
lagging stator power factors.


2r 2r
Pm = T m m = T e + B m
P
P

2

2
Tr
Te
+ AP l
PL = AP l
kqs
kqs
+ BP l

(75)

Tr
Te
+ CP l
+ DP l
kqs
kqs

where

AP l = 1.5 rs + rr
BP l =

3rr Ls qs
L2m

CP l =

3e rs qs
rc


DP l = 1.52qs

Ls
Lm

rr
+ kc
L2m

(76)

e
rc

2 
.

C. Leading and Lagging Stator Power Factor Operation


In the case of the motoring convention in which the model
is based upon, which is evident in (1)(4), a negative stator
current should normally imply that power flows to the grid, and
when positive, it should mean that power flows back into the
machines stator. This is true for the stator reactive power, but
for the stator, it is vice versa. The reasons for this are derived
from the fact that the stator flux orientation which aligns the
q-axis of the stator flux linkage with the stator flux magnitude
imposes that the Ids controls electromagnetic torque Te and that
Iqs controls the reactive torque Tr , as evident in (22) and (23).
Notice in (60) and (61) that the real power is mainly regulated
by Te while Tr controls the reactive power. It is intuitive that a
negative Te implies that the machine is generating and a positive
Te means motoring.
Consequently, a negative Te will imply that Ids is positive
for generating. Therefore, a negative stator real power means

Fig. 12. Rotor current versus rotor electrical speed for unity, leading, and
lagging stator power factors.

generating operation and vice versa. However, from (62), based


on the orientation of the stator current, the stator reactive power
is negative for lagging power factor and positive for leading
power factor. Take notice that the power sign convention of the
GSC of Fig. 16 differs from the rotor power of Fig. 15 because
the GSC was modeled based on (34) and (35) such that positive
power flow means power flow toward the grid and vice versa.
For a situation whereby the DFIG is required to supply the
grid with reactive power, i.e., a leading stator power factor, the
rating of the MSC must be raised so as to account for the extra
reactive power consumption by the grid. This definitely implies
more losses because of the increase in the rotor voltage and
current, which is revealed in Figs. 11, 12, and 14. However,
Figs. 1116 show that the lagging stator power factor gives the
best operating regimes since it accounts for the least losses and
certainly the least size of MSC power rating. Furthermore, it is
evident in Fig. 18 that there exists an optimal lagging reactive
power for the various power profiles. This is dealt with in the
next section.

BALOGUN et al.: DETERMINATION OF STEADY-STATE AND DYNAMIC CONTROL LAWS OF DFIG

Fig. 13. Stator current versus rotor electrical speed for unity, leading, and
lagging stator power factors.

Fig. 14. Core and copper losses versus rotor electrical speed for unity, leading,
and lagging stator power factors.

Fig. 15. Rotor real power versus rotor electrical speed for unity, leading, and
lagging stator power factors.

D. Estimation of Optimal Stator Power Factor


The losses in (76) account for the copper and core losses
of the generator. The windage and frictional losses can be
considered to be constant over the entire operating region [4],

1351

Fig. 16. Rotor reactive power versus rotor electrical speed for unity, leading,
and lagging stator power factors.

Fig. 17. GSC real power versus rotor electrical speed for unity, leading, and
lagging stator power factors.

Fig. 18. Copper and core losses versus stator reactive power for 380, 392, and
410 rad/s rotor electrical speeds.

[23], while the stray load loss is also assumed to be constant


[4]. Therefore, it is intuitive from the results obtained so far
that the minimal copper and core losses correspond to a lagging
stator power factor, as evident in Fig. 18. A lagging stator power
factor will guarantee the least power rating of the MSC [24].

1352

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 3, MAY/JUNE 2013

Fig. 19. Rotor apparent power versus rotor speed for unity, leading, and
lagging stator power factors.

Fig. 21. Optimal steady-state results of stator real power and rotor real power
versus rotor electrical speed.

stator power factor to supply reactive power when required for


system stability [6], [7]
 Pm Pm 
 T
 2AP l
Tr 
 e
(77)
 PL PL  = kqs Tr + BP l = 0
Te
Tr
kqs BP l
Tr =
(78)
2AP l
3e qs BP l
Qs =
.
(79)
4AP l
VI. DYNAMIC C ONTROL
A. MSC Torque Control

Fig. 20. Optimal copper and core losses versus rotor electrical speed.

The stator reactive power that corresponds to the minimal losses


can be obtained analytically from the optimal Tr . Therefore, a
Jacobi matrix obtained from (76) and (75) with respect to Te
and Tr is given in (77), which yields the optimal Tr given in
(78) and consequently the optimal stator reactive power given
in (79). Hence, the optimal value is about Qs = 1250 V AR
with a qs = 0.499. Take notice that this optimal stator reactive
power agrees with the common minimum value of the reactive
power for the three rotor electrical speed conditions of Fig. 18.
The negative sign implies a lagging power factor, which is
intuitive from the results of Fig. 18. Therefore, this value of
stator reactive power gives an optimal lagging stator power
factor of 0.9389. The apparent power (V A) profile of the
MSC for unity stator power factor and stator reactive power
absorption/generation is given in Fig. 19. Observe in Fig. 20
that the copper and core losses are lower than those obtained in
Fig. 14. Hence, Fig. 20 gives the optimal copper and core losses
for the generator. Fig. 21 gives the rotor real power and stator
real power at optimal stator reactive power.
Although the optimal stator reactive power corresponds to a
lagging stator power factor, for most applications, the DFIG is
operated at either unity stator power factor or even at leading

An independent control of the torque variables is achieved


by field orientation of qs = s and ds = 0 so that vdr regulates the electromagnetic torque while vqr controls the reactive
torque. Hence, (30) and (31) yield (80) and (81)
L pTe + rT Te



qs L iqs vds
 vqs
= L k ids

kc
L
kc


Lm kqs mdr vdc
L r Tr kr ss +
(80)
Lr
2
L pTr + rT Tr
 

qs + L iqs vqs
rr
 vds
= L k
ids

ss
L
kc
k
L Lr
 c


Lm kqs mqr vdc
L rs  2
Te + Tr2 +
+ L r T e +
.
kss
Lr
2
(81)
Therefore, the dynamic controllers are given in
L pTe + rT Te = KT e (Te Te ) = T e
L pTr +

rT Tr = KT r (Tr

Tr ) = T r .

(82)
(83)

In Laplace domain, (84) and (85) are obtained


(pL + rT )Te = KT e (p)eT e

(84)

(pL + rT )Tr = KT r (p)eT r

(85)

BALOGUN et al.: DETERMINATION OF STEADY-STATE AND DYNAMIC CONTROL LAWS OF DFIG

where KT e and KT r are PI controllers and eT e and eT r are


the errors between the reference current and the actual value.
If it is assumed that KT e = KT r = KT , then, in general, the
open-loop transfer function for (84) and (85) is given
HT (p) =

IT e
IT r
KT (p)
.
=
=
eT e
eT r
(pL + rT )

C. Rotor Speed Control


Speed control of the wind generator is essential to enable
an optimal power extraction from the wind turbine system.
Therefore, the electromagnetic torque reference command Te
can be generated from an outer loop speed control

(86)
2J
pr = (Te Tm ) .
P

The PI controller is defined as


KT (p) = Kpt + Kit /p.

(87)

Dead-time, transport, and sampling delays introduced by the


converter and the analog-to-digital conversion process are
modeled by introducing into (86) a delay factor Hrd (s) =
esTrd [6], [25], which can be simplified as given in
HT d (p) =

1
.
(pTT d + 1)

(88)

Hence, (89) is obtained as the new closed-loop transfer function


HT o (p) =

1
KT (p)
.
(pL + rT ) (pTT d + 1)

(89)

For a condition of the controllers zero cancelling undesired


pole of the plant, the gains of the PI controller are selected such
that Kpt /Kit = L /rT [26]. Therefore, (90) is obtained
HcT (p) =

(p2 L

Kpt
.
Trd + pL + Kpt )

(90)

When the denominator of (90) is compared with the secondorder Butterworth polynomial of p2
+ 2n p + n2 at optimal
damping design, such that = 1/ 2, then n 2 = 1/Trd ,
and n2 = Kpt /(L TT d ). Hence, Kpt = L /(2TT d ), and consequently, Kit = rT /(2TT d ).

An independent power control is achieved with voltage orientation of vds = vs and vqs = 0. Therefore, vdf controls the
real power generation/absorption by the GSC into/from the
grid, while the GSC reactive power generation/absorption is
controlled by vqf
3mdf vdc nvds
3n2 vss
+
(91)
2
4
3mqf vdc nvds
Lf pQf + rf Qf = Lf e Pf +
.
(92)
4
Lf pPf + rf Pf = Lf e Qf

Similarly, the closed-loop transfer function of (91) and (92) is


given in
KpP Q
(p2 Lf TP Qd + pLf + KpP Q )

(94)

The speed controller and the reference Te are given in (95) and
(96), respectively
r = Kr (r r ) = (Te Tm )

(95)

Te

(96)

= r + Tm .

D. DC-Link Voltage Control


Control of the dc-link voltage is achieved by the d-axis outer
loop control of the GSC. It is assumed that the GSC does not
generate/absorb reactive power into/from the grid. Therefore,
(97) is obtained from (46)


3mqr qs
Ls T r
+
pvdc =
4Cd Lm
Lm qs k


Ls T e
3mdr
Pf mds vds
+
. (97)
+
4Cd Lm qs k
2Cd nvss
A multiplication of (63) by vdc results in a situation such that
the first two terms of the RHS of (97) represent the real power
of the MSC while the last term represents the real power of the
GSC. Hence, further simplification of the last term eventually
results to Pf
Cd pvdc = (Pr + Pf )/Vdc .

(98)

Therefore, the controller is developed as follows:

B. GSC Power Control

HcP Q (p) =

1353

(93)

where KpP Q = Lf /(2TP Qd ) and KiP Q = rf /(2TP Qd ) for


pole-zero cancellation.




Vdc = KVdc (vdc


vdc ) = Pr + Pf /Vdc
Pf = Vdc Vdc Pr

(99)
(100)

where Pf is the reference command of the GSC real power.


The overall power variable control schemes for the GSC is
illustrated in Fig. 22.

E. Robustness Against Rotor Parameter Variation


The impacts of parameter variation on the PI controllers for
the direct control of torque and power variables and on the
PI controllers for vector control are investigated using small
signal model of the DFIG. If a small perturbation is performed
on (80) and (81) by setting the state variables x = xo + x
about an equilibrium state xo and by neglecting higher order
terms, then a small signal linear dynamic model of (101) and
(102) is obtained. Hence, the transfer functions that relate small

1354

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 3, MAY/JUNE 2013

KT r

iqs vqs
ids vds
+
kc
kc


qs vqs
rr
+
ss
k
r
kc
L

o
= kL

o Te + rs Lo (2Te Te + 2Tr Tr )
+ r L
kss
+

mk
L
(qs vqr + qs vqr ) .
r
L

(104)

If the rotor voltage vectors of (103) and (104) are substituted by


using (101) and (102), then the closed-loop transfer functions
of (105) and (106) are obtained. The coefficients of equations
(105) and (106) are given in Appendix B
Te
A3 p3 + A2 p2 + A1 p + A0
=
Te
B 5 p5 + B 4 p4 + B 3 p3 + B 2 p2 + B 1 p + B 0

(105)

Tr
C 3 p3 + C 2 p2 + C 1 p + C 0
=
. (106)
Tr
D 5 p5 + D 4 p4 + D 3 p3 + D 2 p2 + D 1 p + D 0
Similarly, for vector control, the closed-loop transfer function
of (107) is obtained
Fig. 22. Decoupled natural and power variable controls of DFIG. (a) MSC
control. (b) GSC control.

change in the output to small change in the input variables are


derived
L pTe + rT Te
 

iqs vds
ids vqs
kqs vds
= kLo
+

kc
kc
kc
r Lo Tr kr ss
Lm k
+
(qs vdr + qs vdr )
(101)
Lr
L pTr + rT Tr
 

iqs vqs
ids vds
= kL
+
kc
kc


qs vqs
rr
+
ss + r L Te
k
kc
Lr
rs L
+
(2Te Te + 2Tr Tr )
kss
Lm k
+
(qs vqr + qs vqr ).
(102)
Lr
Let us assume that the estimated parameters used in selecting
the gains of the PI controllers do not match the actual parameters of the plant. In that case, (103) and (104) can be assumed
to be the basis for the controller design, where indicates an
estimated value
 

iqs vds
kqs vds
o ids vqs +

KT e = k L
kc
kc
kc
o Tr kr ss
r L
+

mk
L
(qs vdr + qs vdr )
r
L

(103)

iqr
idr
a 3 p3 + a 2 p2 + a 1 p + a 0
=
=
iqr
idr
b4 p 4 + b 3 p 3 + b 2 p 2 + b 1 p + b 0

(107)

where
a3 = kpr L3
2
a2 = kpr
L2 + 2kpr rr L2
2
a1 = 2kpr
rr L + kpr rr L2
2 2
a0 = kpr
rr

b5 = L4 Tdr
b4 = L4 Tdr + rr L3 Tdr + L4 + L3 rr
b3 = 2kpr L3 + kpr L2 rr + kpr Tdr L2 rr + 2rr L3 + L4

2

+ L2 rr3 + (e r )(L L)
L2 Tdr
b2 = 4kpr L2 rr + rr2 Lkpr + rr2 L2

2
2

+ (e r )(L L)
L2 + rr2 kpr
2
b1 = 2kpr Lrr2 + 2kpr
rr L
2 2
b0 = kpr
rr .

The root loci of (105)(107) are obtained when variations


are introduced to the actual rotor resistance and rotor leakage
inductance. The rotor parameters are of interest because the
converters are connected to the rotor circuit. The root loci obtained from (105) and (106) for the direct control of the torque
variables in Figs. 23 and 24 reveal that the poles are restrained
to the left hand plane even if the actual rotor resistance and rotor
leakage inductance are each 0.5 times the estimated values. This
infers that, unlike the vector control, the controllers developed
for the decoupled direct control of the torque variables are

BALOGUN et al.: DETERMINATION OF STEADY-STATE AND DYNAMIC CONTROL LAWS OF DFIG

1355

lr and rr = 0.5
Fig. 23. Loci of poles of (105) when Llr = 0.5L
rr .

lr , rr = 0.5
Fig. 25. Loci of poles of (107) when Llr = 0.5L
rr .

lr and rr = 0.5
Fig. 24. Loci of poles of (106) when Llr = 0.5L
rr .

Fig. 26. Experimental results showing the actual torque variables. (a) Electromagnetic torque. (b) Reactive torque.

robust against parameter variations. However, Fig. 25 shows


that, for a vector control scheme, two poles crossed over to the
instability plane for such a rotor parameter variation.
VII. E XPERIMENTAL R ESULTS
The results given in Fig. 26 were obtained from an experimental set-up that was achieved by a 5-hp wound rotor
induction generator, which serves as the DFIG. A 5-hp dc
machine served as the prime mover to emulate the wind turbine.
Two Semikron converters were connected back to back, one
to the rotor electrical circuit and the other to the grid with
a common dc-link. A ds1104 dSPACE board was used for
signal conditioning and controller implementation for the machine. The rotor speed was set at 350 rad/s, and the controlled
variables were observed to track their respective commands.
Fig. 26(a) shows the electromagnetic torque which tends toward
about 22 N m, while Fig. 26(b) shows the reactive torque that
was set to operate at 9.94, which corresponds to 1250 VAR
stator reactive power.

VIII. C ONCLUSION
A DFIG model with the torque and power variables as state
variables has been presented. The results obtained from the
natural variable simulation were shown to conform to vector
variable simulation. Steady-state operating regions were established for various stator power factor operations. The highest
apparent power rating of the MSC was noticed for leading stator
power factor. Optimal stator reactive power was estimated.
Direct control of torque and power variables was obtained using
feedback linearization. It was revealed by the small signal analysis that, unlike the vector control, the developed controllers
were robust to variations in actual and estimated parameters of
the machine.
A PPENDIX A
PARAMETERS
5-hp, 60-Hz, 220-V lineline (rms), and six-pole woundrotor induction motor
0.65 ;
Stator resistance (rs )
Rotor referred resistance
0.41 ;

1356

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 49, NO. 3, MAY/JUNE 2013

2
o rs Lo )
B1 = 2kpt
rT Lo + 2kpt rT2 Lo + (
rs L

2Tr Lo kpt rT
kss

2 2
B0 = kpt
rT

C3 = kpt L3o ,


o rs Lo )
C2 = kpt L2o kpt + 2rT + (
rs L


C1 = kpt Lo 2kpt rT + rT2 ,
2 2
rT ,
C0 = kpt

Fig. 27. qd equivalent circuit model of the induction machine including


core-loss resistance. (a) q-axis. (b) d-axis.

D5 = L4o Td
D4 = L3o (Td rt + Lo ) + L3o Td (kpt + rT )

Stator leakage inductance (Lls )


Rotor referred leakage inductance (Llr )
Magnetizing Inductance (Lm )
Core loss resistance (rc )

2.6 mH;
2.6 mH;
0.0441 H;
800 (Fig. 27).

A PPENDIX B
D EFINITION OF C OEFFICIENTS

3
o rs Lo ) 2Tr Lo Td
+ (
rs L
kss

D3 = L2o rT (Lo + kpt Td )




o rs Lo ) 2Tr Lo
+ Td kpt Lo + Td Lo rT + L2o (
rs L
kss
o Lo ) 2
+ kpt L3o + L2o Td r2 (L
2

A3 = kpt L3o ,


2Tr

rs Lo rs Lo )
+ rT
kpt + rT + (
kss


o rs Lo ) 2Tr + kpt r2
A1 = kpt Lo 2kpt rT + rT2 + (
rs L
T
kss

A2 = kpt L2o

+ kpt rT (Td rT + Lo )
o rs Lo ) 2Tr Lo
+ (Td kpt rT + Lo kpt + Lo rT )(
rs L
kss
2

o rs Lo ) 2Tr rT Lo L2 2 (L
o Lo ) 2
+ (
rs L
o r
kss

2 2
rT ,
A0 = kpt

B5 = L4o Td

o rs Lo )
o Lo )(
+ (L
rs L

o rs Lo ) 2Tr Lo Td + L4o (Lo + Td rT )


B4 = L3o Td kpt + (
rs L
kss
B3 = kpt L3o + kpt L2o Td rT + L2o kpt (Lo + Td rT )
o rs Lo )
+ (Lo + Td rT )(
rs L

2Tr L2o
kss

o Lo ) 2
+ rt L3o L2o Td r2 (L

2Te L2o r
kss

o Lo ) 2
+ Lo r2 (L
2
o rs Lo ) 2Tr Lo kpt rT ,
D1 = 2kpt
rT Lo + 2kpt rT2 Lo + (
rs L
kss
2 2
D0 = kpt
rT .

R EFERENCES
2

o rs Lo ) 2Te Lo Td r
o Lo )(
(L
rs L
kss
2 2
o rs Lo )
B2 = kpt
Lo + (
rs L

o rs Lo ) 2Te Lo Td r
o Lo )(
+ (L
rs L
kss


D2 = rT L2o (kpt + rT ) + Lo Td rT Lo + L2o (kpt + rT )

2Tr L2o kpt


kss

+ 2kpt L2o rT + kpt Lo rt (Lo + Td rT )



2
2
o rs Lo ) 2Tr rT Lo Lo r (L
o Lo )
+ (
rs L
kss
2
o rs Lo ) 2Te Lo r
o Lo )(
(L
rs L
kss

[1] P. Krause, O. Wasynczuk, and S. D. Sudhoff, Analysis of Electric Machinery and Drive Systems. Piscataway, NJ, USA: IEEE Press, 2002.
[2] B. Ozpineci and L. M. Tolbert, Simulink implementation of induction
machine modelA modular approach, in Proc. Int. Elect. Mach. Drives
Conf., 2003, pp. 728734.
[3] M. Shin, D. Hyun, S. Cho, and S. Choe, An improved stator flux estimation for speed sensorless stator flux orientation control of induction
motors, IEEE Trans. Power Electron., vol. 15, no. 2, pp. 312318,
Mar. 2000.
[4] W. Qiao, W. Zhou, J. M. Aller, and R. G. Harley, Wind speed estimation
based sensorless output maximization control for a wind turbine driving
a DFIG, IEEE Trans. Power Electron., vol. 23, no. 23, pp. 11561169,
May 2008.
[5] R. Pena, J. C. Clare, and G. M. Asher, Doubly fed induction generator
using back-to-back PWM converters and its application to variable-speed
wind-energy generation, Proc. Inst. Elect. Eng.Elect. Power Appl.,
vol. 143, no. 3, pp. 231241, May 1996.

BALOGUN et al.: DETERMINATION OF STEADY-STATE AND DYNAMIC CONTROL LAWS OF DFIG

[6] B. C. Rabelo, W. Hofmann, J. Lucas da Silva, R. Gaiba de Oliveira, and


S. Rocha Silva, Reactive power control design in doubly fed induction
generators for wind turbines, IEEE Trans. Ind. Electron., vol. 56, no. 10,
pp. 41544162, Oct. 2009.
[7] M. Kayk and J. V. Milanovic, Reactive power control strategies
for DFIG-based plants, IEEE Trans. Energy Convers., vol. 22, no. 2,
pp. 389396, Jun. 2007.
[8] M. P. Kazmierkowski, R. Krishnan, and F. Blaabjerg, Control in
Power Electronics: Selected Problems. London, U.K.: Academic, 2002,
pp. 190207.
[9] N. R. Nik Idris and A. H. Mohamed Yatim, Direct torque control
of induction machines with constant switching frequency and reduced
torque ripple, IEEE Trans. Power Electron., vol. 51, no. 4, pp. 758767,
Aug. 2004.
[10] J. Rodrguez, J. Pontt, S. Kouro, and P. Correa, Direct torque control
with imposed switching frequency in an 11-level cascaded inverter, IEEE
Trans. Ind. Electron., vol. 51, no. 4, pp. 827833, Aug. 2004.
[11] G. Abad, M. A. Rodrguez, and J. Poza, Two-level VSC based predictive
direct torque control of the doubly fed induction machine with reduced
torque and flux ripples at low constant switching frequency, IEEE Trans.
Power Electron., vol. 23, no. 3, pp. 10501061, May 2008.
[12] D. Zhi and L. Xu, Direct power control of DFIG with constant switching frequency and improved transient performance, IEEE Trans. Energy
Convers., vol. 22, no. 1, pp. 110118, Mar. 2007.
[13] G. S. Buja and M. P. Kazmierkowski, Direct torque control of PWM
inverter-fed ac motorsA survey, IEEE Trans. Ind. Electron., vol. 51,
no. 4, pp. 744757, Aug. 2004.
[14] G. Poddar and V. T. Ranganathan, Direct torque and frequency control
of double-inverter-fed slip-ring induction motor drive, IEEE Trans. Ind.
Electron., vol. 51, no. 6, pp. 13291337, Dec. 2004.
[15] J. Maes and J. A. Melkebeek, Speed-sensorless direct torque control of
induction motors using an adaptive flux observer, IEEE Trans. Ind. Appl.,
vol. 36, no. 3, pp. 778785, May/Jun. 2000.
[16] E. Bogalecka and Z. Krzeminski, Control system of doubly-fed induction
machine supplied by current controlled voltage source inverter, in Proc.
6th Int. Conf. Elect. Mach. Drives, 1993, pp. 168172.
[17] G. Dong and O. Ojo, Efficiency optimization control of induction motor using natural variables, IEEE Trans. Ind. Electron., vol. 53, no. 6,
pp. 17911798, Dec. 2006.
[18] G. Dong, Sensorless and efficiency optimized induction machine control
with associated converter PWM modulation schemes, Ph.D. dissertation,
Tennessee Technol. Univ., Cookeville, TN, USA, 2005.
[19] Z. Krzeminski, Sensorless multiscalar control of double fed machine for
wind power generators, in Proc. PCC, Osaka, Japan, 2002, pp. 334339.
[20] A. Tabesh and R. Iravani, Multivariable dynamic model and robust control of a voltage-source converter for power system applications, IEEE
Trans. Power Del., vol. 24, no. 1, pp. 462471, Jan. 2009.
[21] A. Balogun and O. Ojo, Natural variable simulation of induction machines, in Proc. IEEE AFRICON, Sep. 2325, 2009, pp. 16, [CDROM].
[22] J. Slotine and W. Li, Applied Nonlinear Control. Englewood Cliffs, NJ,
USA: Prentice-Hall, 1991, pp. 207275.
[23] IEEE Standard Test Procedure for Polyphase Induction Motors and Generators, IEEE Std. 112-2004, Nov. 2004.
[24] D. Aguglia, P. Viarouge, R. Wamkeue, and J. Cros, Analytical determination of steady-state converter control laws for wind turbines equipped
with doubly fed induction generators, IET Renew. Power Gener., vol. 2,
no. 1, pp. 1625, Mar. 2008.
[25] D. G. Holmes, T. A. Lipo, W. Y. Kong, and B. P. McGrath, Optimized
design of stationary frame three phase ac stationary ac current regulators,
IEEE Trans. Power Electron., vol. 24, no. 11, pp. 24172426, Nov. 2009.
[26] C. Lascu, L. Asiminoaei, I. Boldea, and F. Blaabjerg, High performance
current controller for selective harmonic compensation in active power
filters, IEEE Trans. Power Electron., vol. 22, no. 5, pp. 18261835,
Sep. 2007.

1357

Adeola Balogun (M10) was born in Kano, Nigeria.


He received the B.Sc., M.Sc., and Ph.D. degrees
in electrical and electronics engineering from the
University of Lagos, Lagos, Nigeria, in 1998, 2002,
and 2011, respectively.
He was a Visiting/Research Scholar with the Center for Energy Systems Research, Tennessee Technological University, Cookeville, TN, USA, in 2008
and 2009. Presently, he is a Lecturer in the Department of Electrical and Electronics Engineering,
University of Lagos, Akoka-Lagos, Nigeria. His current research interests include electric machines, power electronics, control of
electric drives, and renewable energy conversion systems.

Olorunfemi Ojo (M87SM95F10) was born in


Kabba, Nigeria. He received the Bachelors and
Masters degrees in electrical engineering from Ahmadu Bello University, Zaria, Nigeria, and the Ph.D.
degree from the University of Wisconsin, Madison,
WI, USA.
He is currently a TVA Chair Professor of electrical
and computer engineering with Tennessee Technological University, Cookeville, TN, USA. His current
research interests include the areas of electric machine analysis and drive control, switching converter
technology, and modern control applications in converter-enhanced power and
distributed energy generation systems.
Dr. Ojo is a Fellow of the Institution of Electrical Engineers, U.K. He is
currently the Chair of the Industrial Power Conversion Systems Department of
the IEEE Industry Applications Society. He is an Associate Editor of the IEEE
T RANSACTIONS ON P OWER E LECTRONICS and a member of the Editorial
Board of IET Power Electronics.

Frank Okafor (M01) received the B.Sc., M.Phil.,


and Ph.D. degrees in electrical engineering from the
University of Lagos, Lagos, Nigeria, in 1984, 1987,
and 1993, respectively.
He was a Research Scholar with the Technical University of Chemnitz, Chemnitz, Germany, in
2000. He is currently a Professor of electrical and
electronics engineering with the University of Lagos.
His research interests include the areas of renewable
energy systems, control engineering, power systems,
electric drives, and electromagnetic compatibility.
Dr. Okafor is a Fellow of the Nigerian Society of Engineers.

Sosthenes Karugaba (S07M12) was born in


Muleba, Kagera, Tanzania. He received the B.Sc.
degree in electrical engineering from the University
of Dar es Salaam, Dar es Salaam, Tanzania, in 1999
and the M.S. degree in electrical engineering and the
Ph.D. degree in engineering from Tennessee Technological University, Cookeville, TN, USA, in 2008
and 2012, respectively. Presently, he is a Lecturer
in the Department of Electrical Engineering, Dar
es Salaam Institute of Technology, Dar es Salaam,
Tanzania.
His research interests include power electronics, electrical machines, and
motor drives.

You might also like