You are on page 1of 16

Sensorless Nonlinear Control of Doubly Fed

Induction Generators Using MRAS Observer

Naser Hashemnia *, Farzad Tahami †

Department of Electrical Engineering, Sharif University of Technology, Tehran, Iran,

e-mail: naserhashemnia@ee.sharif.edu *, † tahami@sharif.edu

Abstract— A novel model reference adaptive system (MRAS) observer for the estimation of the position and speed of a

doubly fed induction generator in a nonlinear control drive is proposed. MRAS makes the speed estimation a reliable scheme

especially when the motor parameters are poorly known or have large variations. In order to improve the estimation system,

a sliding mode control law has been used in the adaptation mechanism. The proposed MRAS method has been compared

with a conventional sensorless method in terms of robustness against parameter variations and precision. Simulation results

show the superiority of the proposed scheme to the conventional one.

Keywords— Doubly Fed Induction Generator, Feedback Linearization, Model Reference Adaptive System, Nonlinear

Control, Sliding mode control.

I. INTRODUCTION

Recently, the overall aim of most of the wind energy conversion systems (WECS) has been to provide a constant

frequency output voltage from a variable speed system. This has given rise to the term Variable Speed Constant

Frequency. A doubly fed induction generator (DFIG) can supply power at constant voltage and constant frequency while

its rotor speed varies. This provides more flexibility in power conversion and also better stability in frequency and

voltage control in the power systems to which such generators are connected. A DFIG consists of a wound rotor

induction generator (WRIG) with the stator windings directly connected to the three-phase grid and the rotor windings

connected to a back-to-back partial scale (20-30% rating) power converter as shown in Fig. 1 [1].
Fig.1. A grid connected doubly fed induction generator and its converters.

Such an arrangement provides flexibility of operation in sub-synchronous and super-synchronous speeds both in

generating and motoring modes. The power converter needs only be rated for a fraction of the total output power, the

fraction depending on the allowable sub- and super-synchronous speed range [1]. This results in lower converter cost and

reduced power loss.

In the drive system for a DFIG, rotor position needs to be either measured or estimated. This is firstly because a d-q

transformation needs the transformation angle which generally depends on rotor’s electrical angle. Secondly, in order that

the wind power plant to act in the Maximum Power Point Tracking (MPPT) mode, the counter torque of the generator

must be controlled in proportion to the squared rotor speed.

Various schemes have been proposed for the control of DFIGs which can be generally put into two categories:

Schemes that either neglect the nonlinear nature of the equations of the machine or try to take it into account by simple

feed forward compensation [2]-[11].

Schemes incorporating nonlinear control methods such as input-output linearization, back-stepping, sliding mode

control, etc [12]-[19].

It is rational to expect the second schemes to have better performance in terms of steady state error, decoupling

between quantities such as stator active-reactive power and the speed of tracking power commands. Therefore, a

nonlinear control method has been chosen in this paper to take advantage of the properties mentioned.

A drive system employing mechanical sensors has some drawbacks in terms of maintenance, cost, robustness, and

cabling between the sensor and the controller. The last problem is particularly crucial in wind power generation systems

because usually there is a large distance between the generator and the power electronic converter (which is at ground

level) [20].
Sensorless systems for DFIGs have been the subject of many previous studies. The method proposed in [3] uses the

rotor voltages and currents to design a torque angle controller. The major difficulty of this scheme is the method

employed for computation of rotor flux. The integration of rotor voltage at or near synchronous speed is equivalent to the

integration of the stator voltage at or near zero speed. Hence, similar problems of integrator saturation, resulting in

incorrect estimation of the rotor flux are inevitable. Use of this algorithm has to be restricted to a certain minimum slip

and operation in the synchronous region is not possible. On the other hand, although d-q transformation is not used, the

knowledge of rotor speed is still required if operation at MPPT mode is desired.

Systems proposed in [20]- [22] are based on open loop methods, where the estimated and measured rotor currents are

compared in order to derive the rotor position and the speed is then obtained via differentiation. However, as in any other

open loop control system, problems of making the estimator robust against parameter variations and model uncertainties

exist. Moreover, differentiation introduces some additional noise in the speed signal.

Model reference adaptive system (MRAS) observer based techniques have been proven to be one of the best available

methods, due to their high performance ability and straightforward stability analysis [23].

In a MRAS system, some state variables, xd and xq (e.g. rotor or stator flux linkage components, stator or rotor

current components or back e.m.f components) of the induction machine, which are obtained from measured quantities,

are used as the outputs of a reference model and are then compared with state variables x̂d and x̂q , which are

estimated, using an adaptive model. The difference between these state variables is then used in an adaptation

mechanism, which outputs the estimated value of the rotor speed ̂ r and adjusts the adaptive model until satisfactory

performance is obtained [24]. Such a scheme is shown in Fig. 2.

Fig.2. Model reference adaptive observer for the estimation of machine’s speed
As was reported in [25], rotor current based MRAS observers (RCMO) are the best candidate for grid connected

DFIGs in regards to such performance criteria as stability and robustness. Hence, the proposed method is this paper is

also based on a rotor current observer.

II. MODEL OF DOUBLY FED INDUCTION GENERATOR

Throughout this paper, a d-q reference frame, which has the d-axis coinciding with the grid voltage space vector will

be used, i.e. u sd  U , u sq  0 . This frame orientation can be referred to as "grid flux" oriented [15], [26]. The

differences between grid flux oriented and machine flux oriented controls are subtle, but significant in practice. At the

first glance, the difference is not large. As shown in Fig.3, the air-gap flux space vector (  gap ) of a large machine will

lag the stator voltage space vector by nearly 90 degrees, and not vary much from this position over a wide operating

range. Therefore grid flux orientation is basically the same as machine flux oriented plus a 90 degree offset [27].

Fig.3. Grid flux oriented dq axis [27]

However, determination of the stator voltage space vector angle is much faster, more accurate, and simpler to

implement than machine oriented flux, which requires an estimator. The flux cannot be measured directly, so it must be

calculated from current or voltage measurements and some estimated parameters. The flux estimator therefore has its

own set of errors due to slight variations in parameterisation or due to filter/observer dynamics. It also takes calculation

time [27]. Another advantage of grid flux oriented control concerns the unforced dynamics of a DFIG. In transient

conditions, the eigen values of the system under grid flux oriented control will be more damped, thus improving the

stability and transient behaviour of the system [26]. Moreover, using the line voltage vector reference frame, a simple and

smooth connection of the stator winding to the line grid can be realized during the system start-up [13].

On this frame, the electrical model of DFIG with the stator currents and voltages are considered as state variables [28]:
d 1
 .[  .( d iq   q id )  TL ]
dt J
did   u
  .id  2 .iq  . d  . q  sd   .urd
dt   

diq  
  .iq  2 .id  . q  . d   .u rq
dt  
d d
  Rs .id  0 . q  u sd
dt

d q
  Rs .iq  0 . d (1)
dt

where i d , i q ,  d and  q are stator two-axis currents and fluxes respectively. u rd and u rq are rotor two-axis voltages

and u sd is stator d-axis voltages. 0 ,  and  2  0   are stator voltage angular frequency, rotor speed and slip

frequency all in electrical units. J is the machine’s inertia and TL is load torque. The parameters  ,  ,  ,  and

 are defined as below:

3P  R L2 L
 ,   ( Ls  s ),   Ls (1  m ),   m ,
2   Lr L s Lr
(2)
R
 r
Lr

where Rr is rotor resistance, Rs is stator resistance, Ls is stator total inductance, Lr is rotor total inductance, Lm is

magnetizing inductance and P is number of poles.

Also, the active and reactive components of stator power (positive for motoring mode) will be:

3 3
P u sd id , Q u sd iq (3)
2 2
 
Regarding equation (3) and considering id and iq as reference values for active and reactive currents of the stator

side, the aim of nonlinear controller design will be the calculation of two-axis voltages u rd and u rq such that:

lim t   (id  id  )  0 and


(4)
lim t   (iq  iq  )  0

where it is assumed that the reference currents are constant and bounded or at least their first derivatives are constant

and bounded.
III. DESIGN OF OUTPUT FEEDBACK CONTROL ALGORITHM FOR DFIG

Defining stator current and flux errors as:

~ ~
id  id  id  , iq  iq  iq  ,
(5)
~d   d   d  , ~q   q   q 

and using equation (1) yields:

d~d ~
  Rs .( id  id  )  0 .(~q   q  )  u sd   d 
dt
(6)
d~q ~
  Rs .( iq  iq  )  0 .(~d   d  )   q 
dt

where:

 ( Rs .iq    q  )
d 
0
(7)
 ( Rs .id   u sd   d  )
q 
0


 d  and  q in (7) are the two-axis stator fluxes. Also, according to equation (1), the dynamic error equations of

stator currents will be:

did 
  .(id  id )  2 .(iq  iq )  .( d   d )
  

dt 
 u sd
.( q   q )    .urd  id
 

 

(8)
diq 
  .(iq  iq )  2 .(id  id )  .( q   q )
  

dt 

.( d   d )   .u rq  iq
 

Hence, according to the feedback linearization principle, the two-axis reference values of rotor voltage can be taken

as:

1  
u rd  .(  .id    2 .iq   . d   . q 
  
u sd
  vrd  id  )
 (9)
1  
u rq  .( .iq    2 .id   . q   . d 
  

 vrq  iq )

where vrd and vrq are virtual control inputs that will linearize the system equations.

Inserting equation (9) into equation (8) will make the error dynamics of the system as:
~ ~ ~  
id   .id   2 .iq  .~d  .~q  vrd

 
(10)
~
 ~ ~  ~  ~
iq   .iq   2 .id  . q  . d  vrq
 
~
~ d s d 0
~
   R .i   .
q
~
~ q s q 0
~
   R .i   .
d

Defining v rd and v rq as:

~
vrd  ki id
~ (11)
vrq  ki iq

where k i is the controller’s coefficient, and using Lyapunov theorem it can be proved that the stator two-axis currents

and fluxes converge to zero [27].

IV. SPEED IDENTIFICATION METHODS OF DFIG

In literature, there are numerous methods involving sensorless control of doubly-fed induction generators [21], [29],

[30], [31]. The authors in [22] were the first to propose rotor position sensing in the DFIG. The rotor speed is then

obtained by differentiating the rotor position. The disadvantage of this method is that as in many other existing sensorless

schemes, differentiation can lead to significant inaccuracy due to the noise inherent in measured signals. It should be

noted that the steady state error in the estimated speed may give rise to the following issues [32]:

• Reduced power capture

• Incorrect pitch control operation

• Incorrect operation for other control systems

In this section, a conventional speed sensorless method for DFIG as in [20] will be reviewed briefly and then the

RCMO scheme will be explained in detail. The sensorless algorithm in [20] is best understood using Fig.4.

Fig. 4. Location of different vectors in stationary coordinates [20]


Here, the rotor current vector is shown along with the stator and rotor axes. The rotor current vector ir has an angle

1 with respect to the stator coordinate system. The same vector makes an angle  2 with the rotor axis. If 1 and

 2 are computed, then the rotor angle can be determined as:   1   2 . Knowing the stator flux and current, the

S
rotor current in the stator reference frame, i.e., ir can be computed. The rotor current sensor gives the rotor current

directly in the rotor reference frame. From this information, the angle between the two reference frames can be computed

by using simple trigonometric relations. Speed is then estimated using the relation:

d d
est  Cos . Sin  Sin . Cos (12)
dt dt

It should be noted that using the simple differentiation:

est  d dt , has the drawback that  varies periodically between 0 and 2 , so the problems of discontinuity are

arise.

A second sensorless method described in [33] is based on determining the rotor speed from the rotor circuit equations.

Accordingly, the rotor voltage equation in the stationary reference frame is given by:

d r
u r  Rr ir   j r (13)
dt

Neglecting the derivative terms, the rotor speed can be estimated as:

u r  Rr ir
ˆ | | (14)
r

Hence, the rotor position can be found by integrating the rotor frequency:


 m   ˆ dt (15)

The disadvantage of this method is being an open loop method resulting in significant steady state errors due to the

variations in machine parameters.

In a MRAS system, the goal is to calculate one variable by two different models, and then construct the closed loop

system through the comparison of the model outputs. The adaptation mechanism adjusts the speed or slip frequency in

such a way that the results of the two models approach each other. There can be several adaptation mechanisms.

PI controllers are widely used in industrial control systems applications. They have a simple structure and can offer a

satisfactory performance over a wide range of operation. Therefore, the majority of adaptation schemes described in the

literature for MRAS speed observers employ a simple fixed gain linear PI controller to generate the estimated rotor
speed. However, due to the continuous variation in the machine parameters and the operating conditions in addition to the

nonlinearities present in the inverter, fixed gain PI controllers may become unable to provide the required performance

[34].

In [29], a PI controller has been used as the adaptation mechanism which tunes the estimated rotor speed so that the

estimated current approaches the real current measured by rotor current sensors. As the study is based on small signal

model of the MRAS system, there is no guarantee for global convergence of the sensorless scheme. Moreover, all MRAS

observers are implemented in the stationary reference frame, where the electrical states are usually sinusoidal functions of

time in steady state. Hence, it is difficult to design controller parameters and the observer might become inaccurate or

even unstable in digital implementation [35].

In [36], an adjustable model is obtained based on neural networks and backward differentiation method. The speed

identification is performed by training a two-layer neural network using the back propagation technique. However, there

are some differences between the actual speed and the MRAS identification speed during the startup process. Also,

stability of the sensorless scheme has not been proved. In [37], a rotor current based MRAS observer is used as in [29],

the difference being that a hysteresis controller has been used in the adaptation mechanism instead of a PI controller. The

major advantage is that it is not necessary to adjust any PI parameters. However, no proof has been given to demonstrate

the stability of the system.

V. THE PROPOSED MRAS METHOD

Considering the drawbacks of the MRAS method using a PI controller as the adaptation mechanism, a novel

adaptation scheme, based on sliding mode (SM) theory is proposed is this paper. A new speed estimation adaptation law

is derived using Lyapunov theory to ensure estimation stability as well as fast error dynamics. The work has been

inspired by [34] where a sliding mode MRAS observer was used for a cage induction motor.

The electrical equations of a DFIG in the synchronous rotating reference frame, choosing stator flux and rotor current

components as state variables are [38]:

 ps   A11 A12  s   B11 B12  us 


 pi    A     (16)
 r   21 A22  ir   B21 B22  ur 

where:
1 L
A11   I  s J , A12  m I ,
s s
Lm  L
A21  I  r m J,
 s Lr Ls  Lr Ls
 L2m R 
A22     r  I  (s  r ) J ,
L L L (17)
 r s r 
B11  I , B12  0,

Lm 1
B21   I , B22  I,
 Lr Ls  Lr
1 0 0  1
I  0 1 , J  1 1 
   

And:

T T
s  [ds qs ] , ir  [ids iqs ] ,
2 d
L L
s  s ,  1 m ,p
Rs ( Ls Lr ) dt

The above equations can be used to give an estimation of rotor current using stator voltage and current and rotor

voltage, the estimation being dependant on rotor speed. The value of rotor current in the synchronous frame can also be

estimated using the readings of stator voltage and current sensors only:

s ( stat )   (vs ( stat )  Rs is ( stat ) ) dt ,

s ( stat )  Ls is ( stat )
ir ( stat ) 
Lm
 jst
ir  ir ( stat ) e

where the subscript “stat” denotes the stationary reference frame. The above estimation of rotor current is independent of

rotor speed.

Define the speed tuning signal as:

  irq iˆrd  ird iˆrq (18)

Choose a sliding surface as:

s     k  dt , k  0 (19)

The error dynamics at the sliding surface s  0 will be forced to exponentially decay to zero.
Using the following Lyapunov function candidate:

1 2
v s (20)
2

and forcing the time derivative of v to be negative definite, the state trajectory will be attracted toward the sliding surface

s.

The time derivative of the Lyapunov function in (20) can be calculated as:

v  ss  s (  k  ) (21)

Replacing the values of iˆrd and iˆrq using (16),  can be written as:

  f1  ˆ r f 2 (22)

Therefore,

v  s ( f1  k  ˆ r f 2 ) (23)

The derivative is negative definite if:

 0 for s  0

( f  ke - wˆ f )  0 for s  0 (24)
1 r 2
 0 for s  0

This can be guaranteed if:

f  k M
ˆ r  1  sign( s ) (25)
f2 f2

where sign s  is the signum function.

Therefore, an estimation of rotor speed can be obtained using (25).

VI. SIMULATION RESULTS

In this section, the performance of the proposed sensorless nonlinear method is simulated using Matlab/ Simulink®

and its effectiveness is investigated. Comparison between the MRAS-based method and the conventional sensorless

method confirmed the superiority of the proposed scheme.

A 5 kW DFIG, whose nominal parameters are reported in table 1, is used.

TABLE I. DFIG PARAMETERS

Stator Resistance (R1) 0.95 Ω Magnetizing Inductance (Lm) 0.041 H


Stator Inductance (L1) 0.094 H Pole Pairs 3
Rotor Resistance (R2) 0.45 Ω Rotor Inertia (J) 0.05 Kg.m2
Rotor Inductance (L2) 0.022 H Nominal Torque (Nm) 50 Nm

The simulated scenario is described in the next paragraph.

The generator is first operating at a speed of 100 elect-rad/s with the reference active and reactive commands equal to

-600 Watts and -300 VARs respectively (The minus sign denotes production and the plus sign denotes consumption). At

the third second the reactive command changes from -300 VARs to +300 Vars. This process continues till the speed

changes to 150 elect-rad/s in the fifth second. Finally, the active power command changes from -600 Watts to -800 Watts

in the seventh second. Of course, it is worthy to mention that in a wind power plant, neither the active nor the reactive

power commands are actually constant. The active power reference is a cubic function of rotor speed while the reactive

power reference is usually the output of the loss minimization subsystem.

The waveforms of active and reactive powers are depicted in Fig.7. It is seen that there is a good reference power

tracking. The rotor’s active and reactive current components in the rotor reference frame are shown in Fig. 8 and its

voltage components are demonstrated in Fig.9. In order to compare the two sensorless methods, the rotor electrical speed

has been estimated using two observers, namely the conventional and the MRAS observers. It is evident that the MRAS

has had a better performance in terms of precision (Fig.10). Specifically, the conventional method has some steady state

speed error.

In order to compare the robustness of the two speed estimation schemes, the value of stator and rotor resistances has

been decreased by a value of 5%. The outputs of the estimators are given in Fig.11. It is obvious that the MRAS has

again a better performance. Fig.12 shows the estimator outputs when there is a 5% increase in stator and rotor resistances.

2000
Stator active power (W)

-2000
0 1 2 3 4 5 6 7 8 9 10
time (sec)
Stator reactive power (Var)

500

-500

-1000
0 1 2 3 4 5 6 7 8 9 10
time (sec)

Fig.7. Stator active and reactive powers


Rotor q-axis current (A) Rotor d-axis current (A)
50

-50
0 1 2 3 4 5 6 7 8 9 10
Time (sec)
50

-50
0 1 2 3 4 5 6 7 8 9 10
Time (sec)

Fig.8. Rotor d and q-axis currents


Rotor d-axis voltage (v)

200

-200
0 1 2 3 4 5 6 7 8 9 10
Time (sec)
Rotor q-axis voltage (v)

200

-200
0 1 2 3 4 5 6 7 8 9 10
Time (sec)

Fig.9. Rotor d and q-axis voltages

500
MRAS Estimation
Estimated rotor electrical speed (rad/s)

Conventional Estimation
450

400
Steady State Error

350

300
Steady State Error

250
0 1 2 3 4 5 6 7 8 9 10
Time (sec)

Fig.10. Estimated rotor electrical speed (Comparison between the two schemes)

500
MRAS Estimation
Estimated rotor electrical speed (rad/s)

Conventioal Estimation
450

400
Steady State Error

350

300

Steady State Error


250
0 1 2 3 4 5 6 7 8 9 10
Time (sec)

Fig.11. Estimated rotor electrical speed with a 5% decrease in resistances


500
MRAS Estimation

Estimated rotor electrical speed (rad/s)


Conventional Estimation
450

400
Steady State Error

350

300

Steady State Error


250
0 1 2 3 4 5 6 7 8 9 10
Time (sec)

Fig.12. Estimated rotor electrical speed with a 5% increase in resistances

VII. CONCLUSION

A sensorless nonlinear control based on MRAS observers was suggested. Sliding mode control was used in the

adaptation mechanism to improve the speed estimation scheme. This method offers the advantage of combining the

precise and decoupled dynamics of nonlinear control with good speed tracking and simple implementation of MRAS

observers. Simulation results validate the effectiveness of the proposed approach.

REFERENCES

[1] R. Datta and V. Ranganathan, Variable-Speed Wind Power Generation Using a Doubly Fed Wound Rotor Induction Machine: A Comparison

with Alternative Schemes, IEEE Power Engineering Review, Vol. 22, No. 7, 2002, pp. 52-55.

[2] I. Boldea, Variable Speed Generators. CRC Press, 2005, pp. 11-22.

[3] W. Leonhard, Control of Electrical Drives, Springer, 2001, pp. 400-418.

[4] M. Yamamoto and O. Motoyoshi, Active and reactive power control of doubly fed wound rotor induction generator, Power Electronics

Specialists Conference, Vol. 11, Jun 1990, pp. 455 – 460.

[5] E. Bogalecka , Power control of a double fed induction generator without speed or position sensor, Fifth European Conference on Power

Electronics and Applications, Volume. 8, Sep 1993, pp. 224 – 228.

[6] N. P. Quang, A. Dittrich and A. Thieme, Doubly-fed induction machine as generator: control algorithms with decoupling of torque and power

factor, Vol. 80, No. 5, 1997, pp. 325-335.

[7] Xu, Longya and Wei. Cheng, Torque and reactive power control of a doubly fed induction machine by position sensorless scheme, IEEE Trans

on Industry Applications, Vol. 31, No.3, May/Jun 1995 , pp. 636 - 642.

[8] R. Pena, J. C. Clare and GM. Asher, Douvelbly fed induction generator using back-to-back PWM converters and its application to variable-

speed wind-energy generation, IEE Proc on

Electric Power Applications, Vol.43, No. 3, May 1996, pp. 231- 241.

[9] B. Hopfensperger , D. J. Atkinson and R. A. Lakin, Stator-flux-oriented control of a doubly-fed induction machine with and without position

encoder, IEE Proc on Power Applications, Vol. 147, No. 4, Jul 2000 pp. 241 – 250.
[10] R. Datta, V. T. Ranganathan, Decoupled control of active and reactive power for a grid-connected doubly-fed wound rotor induction machine

without position sensors, IEEE

Industry Applications Conference, 1999, Vol. 4, 1999, pp.. 2623- 2630.

[11] H. Azaza and A. Masmoudi, Implementation of a dual vector control strategy in a doubly-fed machine drive, European Trans on electrical

power, 2000, pp. 541-555.

[12] S. Peresada, A. Tilli and A. Tonielli, Robust active-reactive power control of a doubly-fed induction generator, IECON,

Vol. 3, No. 31, Aug-4 Sep 1998, pp.1621 – 1625.

[13] S. Peresada, A. Tilli and A. Tonielli, Robust output feedback control of a doubly-fed induction machine, IECON, Vol. 3, 1999, pp. 1348 –

1354.

[14] S. Peresada, A. Tilli and A. Tonielli, Dynamic output feedback linearizing control of a doubly-fed induction motor, Proceedings of the IEEE

International Symposium on Industrial Electronics, ISIE, Vol. 3, 1999, pp. 1256 – 1260.

[15] S. Peresada, A. Tilli and A. Tonielli, Indirect stator flux-oriented output feedback control of a doubly fed induction machine, IEEE Trans on

Control Systems Technology, Vol. 11, No. 6, Nov. 2003, pp. 875-888.

[16] N. Patin, A. Nassani, E. Monmasson, J.-p. Louis, Sliding mode control of a doubly-fed induction generator, European Conference on Power

Electronics and Applications, Sept. 2007, pp. 1-9.

[17] P. Vidal, P. David, F. Bonnet, Mixed control strategy of a doubly fed induction machine, Vol. 90, No. 5, May 2008, pp. 337-346.

[18] P. N. Lan, N. P. Quang and P. Buechner, A non-linear control algorithm for improving performance of wind generator using doubly-fed

induction generator, Proceeding of European Wind Energy Conference, Feb/March 2006.

[19] D. D. Li, Decoupled Power Control of Wind Turbine Based on Doubly-Fed Induction Generator, IEEE Conf on Industrial Electronics and

Applications, May 2006, pp. 1-5.

[20] R. Datta and V. T. Ranganatham, A simple position-sensorless algorithm for rotor-side field-oriented control of wound-rotor induction

machine, IEEE Trans on Industrial Electronics, Vol. 48, No. 4, Aug 2001, pp. 786-793.

[21] M. Abolhassani, P. Niazi, H. Toliyat and P. Enjeti, A sensorless integrated doubly-fed electric alternator/active filter (IDEA) for variable

speed wind energy system, in Proc. 38th IAS Annu. Meeting, Oct 2003, Vol 1, pp. 507-514.

[22] L. Morel, H. Godfroid, A. Mirzaian and J. M. Kauffman, Double fed induction machine, Converter optimization and field oriented control

without position sensor, Proc. Inst. Elect. Eng, Vol. 145, No.4, Jul 1998. pp. 360-368.

[23] A. R. Haron and N. R. N. Idris, Simulation of MRAS-based Speed Sensorless Estimation of Induction Motor Drives using

MATLAB/SIMULINK, PECON 2006, Nov. 2006, pp. 411-415.

[24] P. Vas, Sensorless Vector and Direct Torque Control, Oxford University Press, 1998, pp.400-423.

[25] R. Cardenas, R. Pena, J. Clare, G. Asher and J. Proboste, MRAS Observers for Sensorless Control of Doubly-Fed Induction Generators,

IEEE Trans on Power Electronics, Vol. 23, No. 3, May 2008, pp. 1075-1084.

[26] A. Petersson, Analysis, Modeling and Control of Doubly-Fed Induction Generators for Wind Turbines, Ph.D. Thesis, Chalmers University of

Technology, Gothenburg, Sweden, 2005.

[27] T. K. A. Brekken, A novel control scheme for doubly fed wind induction generator under unbalanced grid voltage conditions, PhD Thesis,

University of Mineesota, July 2005.

[28] A. Farrokh payam, Control of slip power wound rotor induction machine drive based on adaptive nonlinear control, MSC Thesis, Isfahan

University of technology, Jan 2006.


[29] R Cardenas, R. Pena, J. Proboste, G. Asher and J. Clare, MRAS observer for sensorless control of standalone doubly fed induction

generators, IEEE Trans on Energy Conversion, Vol. 20, No.4, December, 2005, pp 710-718.

[30] H. Huang, Y Fan, RC. Qiu and XD. Jiang, Quasi-steady-state rotor EMF oriented vector control of doubly fed winding induction generators

for wind-energy generation, Electric Power Components and Systems, Vol. 34, No.11, Nov 1, 2006, pp 1201-1211.

[31] O.A. Mohammed, Z. Liu and S. Liu, A novel sensorless control strategy of doubly-fed induction machines, 2005 IEEE International

Conference on Electric Machines and Drives, 2005, pp. 315-319.

[32] K. Gogas, Design of robust speed and position sensorless decoupled P-Q controlled doubly-fed induction generator for variable speed wind

energy applications, MSC Thesis, McGill University, Montreal, Quebec, Canada, Feb 2007.

[33] E.H. Kim, S. B. Oh, YH. Kim and C.H. Kim, Power control of a doubly fed induction machine without rotational transducers, Proc. Of 3rd

International Conference on Power Electronics and Motion Control, Vol. 2, Beijing, China, Aug. 15-182000, pp. 951-955.

[34] S. M. Gadoue, D. Giaouris, J. W. Finch, MRAS Sensorless Vector Control of an Induction Motor Using New Sliding Mode and Fuzzy Logic

Adaptation Mechanisms, IEEE Trans on Energy Conversion, Vol. pp, No. 99, 2010, pp. 1-9.

[35] Y. Sheng, V. Ajjarapu, Sensorless Control of the Doubly-Fed Induction Generator for Wind Energy Generations Using a Speed-Adaptive Full-

Order Flux Observer, Twenty-Fourth Annual IEEE Applied Power Electronics Conference and Exposition APEC 2009, 2009, pp. 1951 – 1957.

[36] Li Lan, Liu Yanli, Rotor Speed Identification of Doubly-Fed Generator System Based on Neural Network, International Conference on E-

Learning, E-Business, Enterprise Information Systems, and E-Government, 2009, pp.146 – 149.

[37] Marques, G.D, Pires, V.F, Sousa, S, Sousa, D.M, Evaluation of a DFIG rotor position-sensorless detector based on a hysteresis controller,

International Conference on Power Engineering, Energy and Electrical Drives, 2009. pp. 113 – 116.

[38] M. Yin, G. Li, M. Zhou, G. Liu, C. Zhao, Study on the control of DFIG and its responses to grid disturbances,

IEEE Power Engineering Society General Meeting, October 2006, pp. 1-6.

You might also like