Journal of Electrical Engineering

www.jee.ro

MODELING AND SIMULATION OF SENSORLESS CONTROL OF PMSM WITH

LUENBERGER ROTOR POSITION OBSERVER AND SUI PID CONTROLLER

GHADA A. ABDEL AZIZ 1 , MOHAMED. I. ABU EL- SEBAH2

1,2
Electronic Research Institute, Dokki, Cairo, [email protected],
2
Collage of engineering and technology, Arab academy for science technology & maritime transport, [email protected]

A. SHALTOUT 3 and F. ISMAIL 4

3,4
Faculty of Engineering, Cairo University,[email protected]

Abstract: This paper presents an investigation and 1. Introduction

evaluation of the performance of the surface Permanent
Magnet Synchronous Motor drive under the Simplified
Universal Intelligent PID controller (SUI PID). The
P MSM has wide application in AC servo system,
compared to other machines because of its
advantages like quick response, excellent control
estimation of the rotor position and the angular speed in performances, small size , light weight, high power/
dynamic rate were derived by the use of the Luenberger
weight ratio, large torque/ inertia ratio, smooth torque
state observer for currents and MRAS (Model Reference
Adaptive System) observer. It also shows how to use a operation, controlled torque at zero speed, high speed
Luenberger state observer in a field oriented control operation, high torque production capability, high
(FOC) scheme to implement a sensorless vector control efficiency and compact structures [1].
strategy. The mathematical descriptions of the system The motor drives need the information of rotor speed to
and simulation results have been presented in this paper.
achieve speed control loop. Rotor speed can be
Key words: Surface Permanent Magnet Synchronous measured by sensors attached to rotor shaft which
Motor, Luenberger State Observer, SUI PID controller, transmit the motor speed to the drive controller.
MRAS observer, Sensorless vector control. However, using mechanical sensors placed on the
machine shaft have many drawbacks [2]-[3]. First, the
mechanical sensor presence increases the volume and
NOMECLATURE the global system cost. Therefore it requires an
Id , Iq d − q Stator currents available shaft end which can constitute a drawback
Vd , Vq d − q Stator voltages with small sized machines. Moreover, the installation
of this sensor requires a chock relating to the stator,
ωr Rotor angular speed ( rad s) operation which proves to be delicate and decreases the
reliability of the system. This requirement leads to the
θr Rotor position angle ( rad )
paper in the field of sensorless control schemes that
R Stator Resistance ( Ω ) estimate the speed of the rotor without any sensors
KT Motor torque constant attached to the rotating shaft [4].
F Fractional Coefficient In this paper, to overcome problems related to the
TL Load torque ( N.m ) mechanical speed sensors, the adaptive Luenberger
observer for rotor position and angular speed
J Rotor inertia ( Kg .m )
2
estimations is studied under the SUI PID controller for
P Number of Pole Pairs PMSM drive system.
K The matrix coefficient gain
∧ Refers to the estimated value 2. The control system block diagram
Fig. 1 shows the block diagram of sensorless control of
λ EMF constant PMSM. The system has been built using the matlab/
Kp Proportional gain constant simulink. The drive system consists of: - PMSM,
Current controller, Adaptive Luenberger observer for
Ki Integral gain constant estimating both rotor position & angular speed and the
ed The state error in the d-axis SUI PID controller. The estimated rotor angular speed
∧
eq The state error in the q-axis ω r is compared with the reference angular speed ω ref
M The SUI PI controller gain which it is set to 314 rad/sec. The resulting angular

1
 Journal of Electrical Engineering
www.jee.ro

Fig. 1. Block Diagram of Sensorless control of PMSM

speed error is processed in the SUI PID controller to The electro-magnetic torque:-
regulate the rotor speed. The output of this controller
is iq ref .The state frame transformations of Clark, Park,
Te =
3
2
[
P λ I q + ( Ld − Lq ) I q I d ] (3)
and inverse-Park are used to transform the currents and
voltages.
For constant flux operation I d maintained at zero,
2 PMSM model which produce electric torque
The stators of the PMSM and wound rotor synchronous

[ ]
motor are similar. In addition, there is no difference
3
between the back EMF produced by a permanent Te = P λI q = KT I q (4)
magnet and that produced by an excited coil. Hence, 2
the mathematical model of a PMSM is similar to that of
the wound rotor SM [5]. The electrical dynamic This torque equation of the PMSM similar to the
equation of the motor in the dq rotor reference frame regular dc machine and hence provides ease of control
can be written as:- where, the torque depends only on the quadrature
di d Vd R L component of the stator current.
= − id + q P ω r iq (1) The torque balanced equation is:-
dt Ld Ld Ld
di q Vq R L λPω r dωr
= − iq − d P ω r id − (2) Te = J + Fωr +TL (5)
dt Lq Lq Lq Lq dt

dωr [Te − Fωr − TL ]

Both stator currents in the rotor reference frame I d &
I q can be obtained by integrating equations (1) & (2). = (6)
dt J

The rotor position θ r in state space derivative form

is:-
dθ r
= ωr (7)
dt

2
 Journal of Electrical Engineering
www.jee.ro

The inverse Clark transformation can also be found as:

 
 0 
 ia   1
i  =  − 1 3   iα 
× 
 b  2 2  iβ  (9)
 ic   − 1 − 3
 
 2 2 
Fig. 2. Analytical model of PMSM 3.2. Park transformation
Park-transformation produces the system model; the
current and voltage vectors onto a rotating coordinate
3. Coordinate transformations time invariant system as shown on Fig.4.
In order to study the performance of PMSM model
under sensorless control, two different transformations
were made: The first one is Clarke-transformation
while the second is Park transformation.

3.1 Clarke transformation

Clarke-transformation transforms the 3-phase system
into a 2- phase time or vice versa.
.

Fig. 4. Park transformation

id   cosθ r sinθ r  iα 

i  =  × 
 q  − sinθ r cosθ r  iβ 
(10)

And the inverse Park transformation is:

iα  cosθr − sinθr  id 

i  =   × i 
Fig. 3. Clark transformation

  
β sinθr cosθ r   q
(11)

 −1 −1  i  4. Current controller
iα  21 2  a
2 ×i The current controller block used to regulate the stator
i  =    b ∧
current and it has five inputs: id ref , iq ref , ∆id , ∆iq & ω r
  0
β 3 3 − 3  i  (8)

 2 2   c 
and two outputs; Vd & Vq . The current controller has
been implemented according to the current controller
equations;

3
 Journal of Electrical Engineering
www.jee.ro

Vd   R − Pωr  idref  1 Ld 0 

The proposed SUI PI controller is a classical PI with a

V  =   × i  + ∆t ×  0 L 
universal PID constant, so no specific design is needed

  
q Pω r R   qref   q
and therefore no need for a system model. The SUI
PID controller constants K P and K I are the absolute
∆id   0  values of errors and integration of errors as shown in
×  +  
Fig.5.

∆iq  Pωr λ
(12)

Where; id ref is maintained at zero since there is no

flux weakening operation in this paper.

∆i d = id ref − I d (13)

∆i q = iq ref − I q (14)

5. The SUI PID speed controller

The speed controller was implemented to regulate the
rotor speed by comparing the reference angular speed
with the estimated angular speed. The output of this Fig. 5. SUI PID controller
controller is i q ref .
The PID controller consists of three terms [6]:- 6. Luenberger Adaptive Observer
The state space of PMSM model of equations 1&2 is
The first controller term: P = K P ∗ error (15)
constructed according to the following relations based
∫
The second controller term: I = K I ∗ error .dt (16) on the knowledge of inputs and outputs of the system:
derror
The third controller term: D = K d ∗ (17)
 dX
dt  = AX + BU
The following values can then be substituted into the  dt
Y = CX
PID controller equation:-

K P = ABS (error ) (18) (24)

(
K I = ABS ∫ error.dt ) (19)
 derror  Where;
K d = ABS 
 dt 
 (20)
[
X = id iq] ; U = [V
T
d Vq λ ]T ;
Y = [i i ] ;
Selecting the controller constants from the above T
explanation leads to a simple design algorithm and d q
simplified adaptive weighting for the three terms.
 −R  1 
Pω r 
The proposed intelligent PID controller is constructed
by using the multi degree of freedom controller  L  0 0 
(MDOF) concept [7], [8]. Applying the intelligent PID A= d  B = L
controller eliminates the need to know the system
 − Pω r − R; 1 − Pωr  ;
steady state gain. 0 
1  Ld   L L 
C m = C * M * error + C * (1 − error ) (21)
M
1 0
1 C C= 
Cm = C * error * ( M − ) +
M M
(22)
0 1 

For M ff 1 C m = C * error * M (23) For SPMSM, Ld = Lq = L . To estimate the rotor

position and the angular speed; the stator current in d &

4
 Journal of Electrical Engineering
www.jee.ro

q directions ( id & i q ) in the rotating frame are used as adjustable one. The rotor angular speed is built around
the following adaptive
daptive mechanism
mechanism:
the estimated variables and the rotor speed ωr as the
regulated variable. ∧ ∧ ∧ λ t ∧ λ ∧
The second order Luenberger observer is given by
by: ωr = Kp (id iq −iq id − eq ) + Ki ∫(id iq −iq id − eq )dt
 ∧ ∧ ∧ ∧
L 0
L
 d X = A1 ( ω r ) X + B o U + K ( Y − Y ) (26)
 dt Where;
 ∧ ∧
 Y = C X ∧ ∧
ed = id − id & eq = i q − i q .

(25) Where; ed & eq are the state errors.

∧ The tracking performance of the speed estimation and
Where; ωr is the estimated angular speed (rad/sec) and the sensitivity to noise are depending on
o proportional
K is a constant coefficient matrix which is and integral coefficient gains. The integral gain K i is
characterized by an equation containing a term which
corrects the current state estimates by an amount chosen to be high for fast tracking of speed. While, a
proportional to the prediction error. The prediction low proportional K p gain is needed to attenuate high
error is the difference between the estimated and actual frequency signals denoted as noises [[11].
current.
rrent. This correction ensures stability and By estimating the rotor angular speed, the estimated
convergence of the observer even when the system ∧
being observed is unstable [9]. rotor position θ r can be obtained by:
∧ T T ∧ ∧
∧ ∧
 ∧ ∧ ∧ θ r = ∫ ω r . dt
X = id iq  ; Y = id iq  ; (27)
    MRAS structure for estimating the rotor angular speed
 ∧ 
and the rotor position is illustrated in Fig. 6.
 − R The scheme consists of 2 models; reference model and
∧ P ωr  adjustable model with adaptation mechanism. The
A1 (ω r ) =  L  reference model block represents the actual system
 ; having unknown parameters values. The adjustable
− P ω
∧
− R  model block has the same structure of the reference
 r
L  one, but with adjustable parameters instead of the
unknown ones. The adaptation mechanism block
1  updates the adjustable model with the estimated
L 0 0  parameter until satisfactory performance is achieved.
Bo =  ∧ 
The PMSM model was considered to be the reference
1 − Pωr  .
model and the Luenberger observer was considered as
 adjustable
justable model to obtain the estimated stator currents
 0 L L  [12].

7. Rotor angular speed observation

The rotor angular speed is reconstructed using the
model reference adaptive system (MRAS). The MRAS
principle is based on the comparison of the outputs of
the two estimators. The first one is independent of the
observed variable named as model reference,
eference, while the
second is the adjustable one. The error between the two
models feeds an adaptive mechanism to turn out the
observed variable [10].
In this paper,, the actual system is considered as the
model reference and the observer is used as the Fig. 6. MRAS Structure

5
 Journal of Electrical Engineering
www.jee.ro

8. Simulation results After loading the motor torque oscillates around Te =

Based on the motor parameters indicated in the
1.8 N.m .
appendix, the simulation of the sensorless control of
Fig. 10 shows the d-axis current, the actual and
PMSM has been carried out using the Simulink in
Matlab. The stator current is decoupled using the field estimated stator current at d axis I d . The estimated
∧
oriented control into both id & i q . In this paper the
current I d tracks the actual d axis current I d of the
motor was tested at no load and sudden loading
motor at t=0.05 sec and both become 0 A at t=0.06 sec
conditions with constant stator resistance at 2Ω. and both oscillate around the zero axis.
The PWM switching frequency of the inverter is set to
10 KHz. The DC link voltage of the inverter is After loading at t=1 sec, both estimated and actual id
maintained at 200 V. The reference angular speed of are equal to zero.
the motor is set to 314 rad/sec. Fig. 11 shows the q-axis current, the actual and
Using only the three phase stator currents the estimated stator current at the q axis, I q : the estimated
quadrature and direct axis currents, rotor angular speed ∧
and position are estimated by the Luenberger state q axis current I q tracks the actual q axis current I q at
observer. t=0.06 sec and both of them have the same
The estimated angular speed was regulated via
characteristics of the motor electromagnetic torque Te
conventional PI controller with low K p =2 and above
in both cases during the stating up with no load and
Ki =5000 and the constant coefficient matrix K =1000. after loading at t=1 sec and both oscillates around 4A.
Fig. 7 shows the actual, estimated and the reference Fig. 12 shows the rotational parameters (Sin and
angular speeds. Cos of the estimated rotor position).
At starting, the motor has rapid changing, both the 350
actual and the estimated speeds track the target speed at
300
t=0.062 sec. The system response converges to the Target speed
model reference one. 250
∧
Speed(rad/sec)

The estimated rotor angular speed ω r tracks the actual 200

rotor angular speed ω r in a very short time nearly 150 Actual speed
0.003 sec by the virtue of the SUI PI controller. 100
After loading, both the estimated and actual rotor
angular speeds dropped but in a very short time and by 50
the virtue of the SUI PI controller they returned back 0 Estimated speed
again to the reference value (target) in less than 0.002
sec. -50
0 0.5 1 1.5 2
Time(sec)
At t=0.04 sec an error occurs with angular speed
error=65 rad/sec and at t=0.04 sec the error converges Fig. 7. The actual, target and estimated angular
to zero in 0.0572 sec thanks to the adaptive control speeds at starting and then a sudden loading
algorithm as shown on Fig. 8. 70
At t=1 sec, with a sudden loading, the speed error
60
decreases and then return back again to zero as both
Speederror(rad/sec)

Actual and estimated speeds are equal. 50

The speed error between the estimated and the 40

reference is the input to the SUI PI controller to
30
regulate it and the output of the controller is iq ref .
20
Fig. 9 shows the motor electromagnetic torque, Te 10
which starts smoothly and at t=0.05 sec it becomes 1.8 0
N.m then from t=0.06 sec to t=0.128 sec, the
-10
electromagnetic torque was oscillating around 0 N.m 0 0.5 1 1.5 2
Time(sec)
and then it returned back again to 1.8 N.m when load
and then it returned back again to 1.8 N.m .
Fig. 8. The angular speed error

6
 Journal of Electrical Engineering
www.jee.ro

10 1.5
Cos(theta estimated)
Sin (theta estimated)

Sin & Cos ( theta estimated)

1
5
0.5
Te(N.m)

0 0

-0.5
-5
-1

-10 -1.5
0 0.5 1 1.5 2 0 0.1 0.2 0.3 0.4 0.5
Time(sec) Time(sec)

Fig. 9. The Electromagnetic torque Te Fig. 12. The Sin & Cos (estimated theta)
50
Conclusions
This paper presents a sensorless control of PMSM
using the SUI PID controller and Luenberger adaptive
observer to estimate both rotor angular speed and the
Estimated Id rotor position. The efficiency of using luenberger
adaptive observer has been studied and indicated on the
Id(A)

0 simulation results in the simulink/ matlab software

package.
Actual Id
This technique of estimation for both rotor angular
speed and rotor position instead of using mechanical
sensors; increases the system reliability and reduces the
system cost.
The adaptive with control algorithm for the rotor
-50
0 0.5 1 1.5 2 angular speed associated with the SUI PI speed and
Time(sec) current controllers show good performances at starting
and at sudden loading condition.
Fig. 10. The actual and estimated d-axis current id

50
APPENDIX

Motor parameter used in the simulation:-

Estimated Iq
Power 1.7 KW
Frequency 150 Hz
Iq(A)

0 Line Voltage 380 V - Star

Actual Iq Line Current 3.4 A - Star
No. of Pole Pairs 3
Rated angular Speed 314 rad/sec
Stator Resistance 2 Ω / phase
d- axis Inductance (Ld ) 7.75mH
-50
0 0.5 1 1.5 2 q- axis Inductance Lq( ) 7.75mH
Time(sec)
EMF Constant 0.098V/rad/sec
Fig. 11. The actual and estimated q-axis current i q

7
 Journal of Electrical Engineering
www.jee.ro

Friction Coefficient 0
Moment of inertia 0.00037 Kg.m 2
coefficient
Load Torque 1.5 N.m

REFERENCES:
[1] Wang Song, Shi Shuang - Shuang, Chen Chao, Yang
Gang and QuZhi-jian, ”Identification of PMSM based on
EKF and Elman Neural network”, Proceedings of the IEEE
International conference on Automation and Logistics
Shenyang, China August 2009.
[2] S. Nakashima, Y. Inagaki, I. Miki, "Sensorless Initial
Rotor Position Estimation of Surface Permanent Magnet
Synchronous Motor ", IEEE Trans. Ind. Appl. ,Vol. 36, No.
6, November 2000, pp 1598-1603.
[3] J. Jang, J. Ha, M. Ohto, K. Ide, S. Sul, "Analysis of
Permanent-Magnet Machine for Sensorless Control Based on
High-Frequency Signal Injection", IEEE Trans. Ind. Appl.,
vol.40, N°.6, November/December 2004, pp 1554-1604.
[4] K. Nabti, K. Abed, and H. Benalla, “Sensorless direct
torque control of Brushless AC machine using luenberger
observer”, Journal of Theoretical and Applied Information
Technology, 2005 - 2008 JATIT.
[5] P. PILLAY and A. KRISHNAN, “Modeling of
Permanent Magnet Motor Drives” , IEEE Transactions On
Industrial Electronics, VOL. 35, NO.4, November 1988 pp:
537 – 541.
[6] Khater , F.M.H., Ahmed , F.I., and Abu-Elsebah , M.I.,
"Sensorless PM Motor with Multi Degree of Freedom Fuzzy
Control”, World Academy of Science Engineering and
Technology 6, 2005.
[7] Dhaouadi , R., Mohan , N., and Norum , I., “Design and
implementation of an extended Kalman filter for state
estimation of a Permanent Magnet Synchronous Motor”,
IEEE Trans. Power Elect., July 1991, pp. 491- 497.
[8] Schroedl , M., “Sensorless Control of Permanent Magnet
Synchronous Motors”, Electric Machines and Power
Systems, Vol. 22, 1994, pp. 173-185.
[9] www.st.com
[10] Grouz Faten, Sbita Lassaâd, "Speed Sensorless IFOC of
PMSM Based On Adaptive Luenberger Observer", World
Academy of Science, Engineering and Technology 44, 2010.
[11] S. Suwankawin, and S. Sangwongwanich., “Design
strategy of an adaptive full-order observer for speed
sensorless induction motor drives- tracking performance and
stabilization”, IEEE trans. on power electron. 53 (1), pp: 96
-119, 2006.
[12] H. Kraiem, M. Ben Hamed, L. Sbita, and M. N.
Abdelkrim, “DTC Sensorless Induction Motor Drives based
on MRAS Simultaneous of Rotor Speed and Stator
Resistance,” International Journal of Electrical
Power Engineering: IJEPE, 2 (5): 306-313, 2008.