 Open Access
 Total Downloads : 483
 Authors : Mamatha Gowda, Warsame H. Ali, Penrose Cofie, John Fuller
 Paper ID : IJERTV3IS050689
 Volume & Issue : Volume 03, Issue 05 (May 2014)
 Published (First Online): 20052014
 ISSN (Online) : 22780181
 Publisher Name : IJERT
 License: This work is licensed under a Creative Commons Attribution 4.0 International License
Design Implementation of Speed Controller Using Extended Kalman Filter for PMSM
Mamatha Gowda,
Dept. of ECE,
Prairie View A&M University, Prairie View, TX 77446
USA
Warsame H. Ali,
Dept. of ECE
Prairie View A&M University, Prairie View, TX 77446
USA
Penrose Cofie,
Dept. of ECE,
Prairie View A&M University, Prairie View, TX 77446
USA
John Fuller,
Dept. of ECE,
Prairie View A&M University, Prairie View, TX 77446
USA
Abstract A novel design impl1ementation of proportional integraldifferential equivalent controller using state observer based Extended Kalman Filter (EKF) for a Permanent Magnet Synchronous Motor (PMSM) is proposed. The EKF is constructed to achieve a precise estimation of the speed and current from the noisy measurement. Then, a proportional integral derivative (PID) controller is developed based on Linear Quadratic Regulator (LQR) to achieve speed command tracking performance. In the present method, the speed and qaxis current are estimated accurately using the EKF algorithm. The steady state and transient state response of the overall system is greatly enhanced and also the speed control is achieved effectively under load disturbance. The experimental results for the speed response and qaxis current as well as the control signal variations when the PMSM is subjected to the load disturbance are presented. The results verify the effectiveness of the proposed method.
Keywords – Permanent Magnet Synchronous Motor, Extended Kalman Filter, PID, Vector Control.

INTRODUCTION
The most utilized motor in the field of variable speed electric drive applications is the Permanent Magnet synchronous Motor (PMSM). The reason for this use is due to its higher full load efficiency and power factor, high torque/inertia ratio, and its wide range of speed control [14]. In PMSM drives, the vector control theory is applied in the dq reference frame where the flux and torque are controlled independently [57].
To measure the shaft position of the PMSM the optimal encoder is used. Speed signal is obtained through the discrete differentiation of the encoder position; this signal is very noisy and has an inherent delay. Thus, it adversely impacts the
1
performance of the system. To solve this problem, the speed can be estimated through the EKF algorithm. The estimated speed is very accurate and much better than that obtained using the encoder. The Extended Kalman Filter produces the optimal estimation of the states
based on the least square method. Here, a feedback system control is utilized to estimate the process. The EKF provides good tolerance for the mathematical model error and noises in the measurement.
Many research works have concentrated on speed and position estimation of PMSM using sliding mode method, high frequency signal injection method, adaptive control theory, fuzzy control, state observer, and the EKF approach
[816]. In all, the EKF method is more attractive as well as popular and is continuously being used in research and applications because it delivers rapid, precise, and accurate estimation. Also, in many applications, the EKF method is implemented because of its lowpass filter characteristics. The feedback gain used in EKF achieves quick convergence and provides stability for the observer.In this paper a vector control method is developed and implemented by means of the Extended Kalman Filter algorithm to provide the speed control. Here, state space representation of the model and observer is obtained. Utilizing the state space model, a PID controller is designed using Linear Quadratic Regulator (LQR) approach [17]. The accurate estimation of states is very essential to achieve better control and performance of the PMSM drives. Here the EKF is utilized for the precise
estimation of the rotor speed and stator qaxis current. Since the For the speed controller design, the system output is
drive speed and drive current measured directly from the machine terminals contain noise, they are not precise for speed
y 0
1isq
. (9)
control. In the proposed implementation, the speed and qaxis current are estimated accurately by introducing EKF algorithm theory. The proposed method yields a smooth and quick speed tracking. It also reduces the actual disturbance applied to the
mech
If the motor output is considered as the drive current, then the motor output is given by:
system, and provides better control of the control signal variation. The overall system performance under load condition
y' 1
0isq
. (10)
is greatly enhanced.

DESCRIPTION OF MOTOR MODEL
The stator voltage and stator flux linkages equations in the rotor references frame are [5]:
vsq (t)
mech
Ls s Rs
1
TL
1
Js B
p
isq (t)
2 fd
vsd
Rs isd

L d i
s dt sd

m
Ls isq
, (1)
fd
2
p
v R i

L d i
(L i
) , (2)
sq s sq
s dt sq
m s sd fd
sd Ls isd fd , (3)
Fig1. Qaxis subsystem of a simplified PMSM
sq
Ls isq
(4)



PID PARAMETERS TUNING WITH STATE FEEDBACK AND STATEFEEDFORWARD LQR
where
vsd , vsq , isd , isq , sd , sq are daxis and qaxis voltages,
The controller is designed as a singleinputsingleoutput (SISO)
currents and flux linkages respectively. Rs and Ls are the stator winding resistance and inductance respectively.
The electromagnetic torque generated and the acceleration is
system. The following discussion is based on the qaxis controller/observer design [18].
The state space model of the simplified PMSM, G1(s), with reference to the qaxis is:
given by:
x1 (t) A1 x1 (t) B1u1 (t) ,
y1 (t) C1 x1 (t)
x1 (0) x10
(11a)
(11b)
T p
i , (5)
where, x (t) R21 , u (t) R11 , y (t) R11 , and A1,B1 andC1 are
em 2
d
fd sq
T T B
1 1 2
constant matrices.
dt mech
em L
J
J mech
, (6)
The entire system output is the motor output plus the load disturbance and this is given by:
p
m 2
mech
(7)
y(t) y1 (t) d(t)
where y(t) R11 , d(t) R11 .
(12)
where mech
poles.
is the speed of the motor and p is the number of
The state space model of the speed controller, G2(s), can be written as:
Since the torque developed by the motor is directly proportional to the motor drive current, it is difficult to control
x 2 (t) A2x2 (t) B2 u2 (t),
x2 (0) x20
, (13a)
the drive current directly. Therefore the drive current is
y2 (t) C2x2 (t) u1(t) , (13b)
indirectly controlled through the input voltage. The simplified
u2 (t) y(t) Ec r(t)
(13c)
model of the PMSM qaxis subsystem is shown in Fig.1. In
2 2 2 2 2
vector control, assuming ids=0, the statespace model of the
where x (t) R 21 , u (t) R11 , y (t) R11 , r(t) R11 , and A , B
, C , E are constant matrices.
PMSM qaxis subsystem is derived as [18]: 2 c
R p
In order to convert the PID tuning problem to an optimal
sq
s
Ls
( fd )
Ls 2 isq
1
L v
(8)
design, we modify the closed loop cascade system into an
augmented system with d(t)=0. The result is the following
p
B
s sq
equation:
mech
( fd )
e
2 J
J
mech
0
x (t) A x (t) B u (t) E r(t) , (14a)
e
e
e
e
1
ye (t) y1 (t) Ce xe (t)
where
(14b)
A A1 0 , B B1 , E 0
, x x1 (t) , C C
0.

STATE OBSERVER BASED EXTENDED KALMAN
e B C A
e 0 e B E e x (t) e 1
FILTER
2 1 2
2 c
2
The resulting statefeedback LQR for the augmented system is:
The state space model of the simplified PMSM given by (11a)
u (t) K x (t) K x (t) K x (t)
(15)
and (11b) is the nonlinear system. To apply the EKF algorithm,
1 e e
1
2
where K R12 , K
1 1 2 2
R12 .
the system needs be discretized and linearized [19]. The discrete approximated equation is given by:
The quadratic cost function J for the system is given as:
xk (I AT)xk1 BTuk , (21a)
T
J 0 [xe
(t)Q x (t) u T (t)Ru (t)]dt
(16)
yk cxk . (21b)
e 1 1
where Q 0 is the state variation and R 0 is the control energy consumption.
The optimal statefeedback control gain which minimizes the performance index is given by:
The nonlinear stochastic equation is:
xk f (xk1, uk ,0)
yk h(x k ,0)
L
L
2
(1 TRs ) i (Tfd p)
T v
(22a)
(22b)
1 T
(17)
sq mech
f (x k , u k ) s s
sq
s
(23a)
Ke R Be P
(T p fd ) i
L
(1 TB
where matrix P>0 is the solution of the Riccati equation,
2 J sq
J ) mech
e e e e
PA A TP PB R1B TP Q 0
(18)
h(xk ,0) H xk . (23b)
e
Let (Ae, Be) be the pair of the given open loop system and h>0 represents the prescribed degree of relative stability. Then the closedloop system (Ae – BeR1B TP) has all its eigenvalues lying
The Jacobian matrices of partial derivative of f and h with respect to x are given by:
in the left of the h vertical line in the complex splane.
Ak
f (xk ,uk )  x
x k
x k
, (24a)
The solution of the revised Riccati equation is given by:
H h(xk )  x
x
, (24b)
P(A

hI) (A hI)T P PB R1B TP Q 0
(19)
k x k k
e e e e
where P 0, h 0.
From (23)
1
Ak
TRs Ls

( Tfd
Ls

p )
2
(24c)
Controller(G2(s))
Motor(G1(s))
(T p fd )
1 TB
2 2 J J
Hk 1
0. (24d)
Fig.2. PID controller system
The total control law is equivalent to a PID controller with
The PMSM states x can be estimated by using EKF algorithm during each sampling time interval as follows:

Time update step
[K1x1(t)] acting as proportional and derivative controller andx pk f (x k1 , u,0)
(25a)
[K2] acting as integral controller.P A P
A T W Q
W T
By choosing the desired values of h and the weighting
pk k
k1 k
k k1
k
matrices Q and R, the control gain can be determined. The block diagram of the designed augmented system including the

Measurement update step
K P H T (H P H T V RV T )1
controller is shown in Fig.2.
k pk k
k pk k
k k
(25b)
x ck x pk K k (yk Hk (x pk ,0)
The calculated control gains are given by:
P (I K H )P
ck k k pk
K1 0.2241 0.0030 , (20a)
K2 – 0.1333 . (20b)
where Wk and Vk are zero mean white Gaussian process and measurement noise with covariance Q and R.
The important and difficult part in the design of the EKF is choosing the proper values for the covariance matrices Q and R
Variable
Physical Meaning
Value
Unit
Rs
Armature Resistance
0.1127
Ls
Armature Inductance
3.63e4
H
J
Moment of
Inertia
1.267e4
Kg m2
B
Damping Coefficient
2.485e4
N m / rad / s
fd
Stator Flux Linkage
0.0131
V / rad / s
p
poles
10
TL0
Static Friction
0.0237
N m
The chosen values of Q, R and P are:
Fig.3. Block diagram of PMSM drive
Table.1 PMSM parameter
Q = 0.008 0
0 1.5
; R= [0.02] (26a)
The reference speed is changed from 0 to 100rad/sec and the speed, current and control signal responses are recorded under no
P = 1 0 (26b)
0 1


EXPERIMENTAL RESULTS
Experiments are conducted in order to verify and validate the result obtained through simulation. The PMSM speed controller is implemented using DS1104 dSPACE board. The model implementation is achieved in the real time (RTI) in dSPACE board where the simulink code is converted directly into DSP code in the expansion box. The controldesk which acts as user interface is used to regulate the output of the system. The block diagram used in the implementation is as shown in the Fig.3.
The PMSM parameters are given in Table 1. Based on optimal control theory, the desired control gain K1 and K2 are determined and the Kalman filter gain is obtained using EKF algorithm. The output drive current and speed are estimated through the EKF algorithm as previously discussed. The inputs are taken directly from the machine terminals. At the output, the motor response is checked and the speed control is observed.
load condition with nonobserver. Fig.4 (a) shows the Speed response, Fig.4 (b) shows the actual Isq current response, and Fig.4(c) shows the control signal Vsq response for the input reference change.
Similarly speed response, actual Isq current response, and control signal Vsq response are observed and recorded under load variations. The experiment results in Fig. 5(a)(b)(c) are shown for when load increase and Fig. 6(a)(b)(c) are shown for when load decrease. It is observed from the results that the speed deviation is around 22rad/sec and control signal variation is nearly 1.26Volts with the transient time of 1.45s. It is also observed that the measured speed signal is too noisy and there is an over shoot in the control signal.
120
100
Speed (rad/sec)
80
60
40
20
0
20
Actual
Reference
0 1 2 3 4 5
Time (sec)
Fig.4(a)
Current Isq (Amps)
10
5
0
5
10
Actual Isq
X: 1.439
Y: 2.397
0 1 2 3 4 5
Time(sec)
Fig.4 (b)
10
Current Isq (Amps)
5
0
5
10
X: 2.378
Y: 8.185
10
Actual Isq
X: 1.142
Y: 1.864
0 1 2 3 4 5
Time (sec)
Fig.5 (b)
10
Control signal (Volts)
8
6
4
2
Vsq
0
9
Control signal (Volts)
8
X: 0.9218
Y: 6.923
7
6
5
X: 1.069
Y: 6.597
Vsq
0 1 2 3 4 5
Time (sec)
0 1 2 3 4 5
Time (sec)
Fig.4 (c)
Fig.4. Experimental results for the input reference change under no load with Nonobserver

Speed response (b) Actual Isq current response
(c) Control signal Vsq response
140
Fig.5 (c)
Fig.5. Experimental results for the load increase with Non observer
(a) Speed response (b) Actual Isq current response
(c) Control signal Vsq response.
X: 1.727
Y: 121.7
140
Actual
Speed (rad/sec)
120
100
80
60
Actual
Reference
X: 1.087
Y: 77.23
0 1 2 3 4 5
Time (sec)
Fig.5 (a)
120
Speed (rad/sec)
100
80
60
Reference
0 1 2 3 4 5
Time (sec)
Fig.6 (a)
Current Isq (Amps)
10
5
0
5
10
Actual Isq
X: 1.758
Y: 1.447
120
100
Speed (rad/sec)
80
60
40
20
0 EKF speed
0 1 2 3 4 5
Time (sec)
20
Reference speed
0 1 2 3 4 5
Time (sec)
Fig.6 (b)
Fig.7 (b)
10
10
X: 1.704
Y: 8.586
Vsq
Control signal (Volts)
9
X: 1.565
Y: 8.13
8
X: 2.982
Y: 6.889
7
6
5
Current Isq ( Amps)
0
5
10
Actual EKF
0 1 2 3 4 5
Time (sec)
5
0 1 2 3 4 5
Time (sec) 8
Fig.7 (c)
Fig.6 (c)
Fig.6. Experimental results for the load decrease with Non observer
(a) Speed response (b) Actual Isq current response
(c) Control signal Vsq response.
With the reference speed of 100rad/s, the experimental results with the EKF are shown in Fig.7 under no load condition. It is observed that the motor speed tracks the reference speed quickly and smoothly. Fig.7 (a) shows the Speed response, Fig.7 (b) shows the estimated speed response Fig.7 (c) shows the actual Isq current response, and Fig.7 (d) shows the control signal Vsq response for the input reference change. The actual current and estimated current have same value. The control signal variation is also smooth without any overshoots. It is evident that the proposed method estimates the rotor speed and qaxis current exactly.
120
100
Speed (rad/sec)
80
60
40
20
EKF
0 Actual
6
Control signal Vsq (Volts)
4
2
0
Vsq
2
0 1 2 3 4 5
Time (sec)
Fig.7 (d)
Fig.7. Experimental results for the input reference change under no load with EKF
(a) Speed response (b) Estimated Speed response

Actual and estimated Isq current response

Control signal Vsq response
The performance of the proposed technique under load disturbance is also verified. The load is varied from – 0.2Nm to 0.2Nm. The Speed response, estimated Speed response, actual Isq current response, and control signal Vsq response are observed and recorded under load variations. The experimental results in Fig. 8(a)(b)(c)(d) are shown for when load increase and Fig. 9(a)(b)(c)(d) are shown for when load decrease. From the results it is evident that speed deviation is reduced from 22 rad/sec to 10 rad/sec and it quickly follows the reference speed.
20
Reference 0 1 2 3 4 5
Time (sec)
Fig.7 (a)
The voltage is also reduced from 1.25v to 0.05v and with less transient time of around 1.2s. The measured speed signal is smooth and contains less noise compared to the nonobserver.
120
Speed (rad/sec)
110
100
EKF
120
Speed (rad/sec)
110
EKF
Reference
90
X: 1.541
Y: 89.18
80
0 1 2 3 4 5
Time (sec)
Fig. 8 (a)
10
Reference
X: 1.644
Y: 110.1
100
90
80
0 1 2 3 4 5
Time (sec)
Fig. 9 (a)
X: 1.663
Y: 2.325
Current Isq ( Amps)
5
10
X: 1.459
Y: 2.353
Current Isq ( Amps)
0 Actual
5
5
Actual 0
10
0 1 2 3 4 5
Time (sec)
Fig. 8 (b)
5
10
0 1 2 3 4 5
Time (sec)
10
Current Isq ( Amps)
5
0
5
10
X: 1.663
Y: 2.325
10
Current Isq ( Amps)
5
Actual 0
EKF
Fig. 9 (b)
X: 1.458
Y: 2.32
Actual EKF
0 1 2 3 4 5
Time (sec)
Fig. 8 (c)
8
X: 2.565
Y: 6.589
Control signal Vsq (Volts)
Vsq
X: 1.345
Y: 6.529
7
5
10
8
0 1 2 3 4 5
Time (sec)
Fig. 9 (c)
X: 1.605 Vsq
Y: 7.268
Control signal Vsq (Volts)
7
6
X: 1.404
Y: 6.629
X: 2.677
Y: 6.589
X: 1.527
Y: 5.917
6
5
4
0 1 2 3 4 5
Time (sec)
Fig. 8 (d)
Fig.8. Experimental results for the load increase with EKF

Speed response (b) Actual Isq current response (c)Estimated Isq current response
(d) Control signal Vsq response.
5
4
0 1 2 3 4 5
Time (sec)
Fig. 9 (d)
Fig.9. Experimental results for the load decrease with EKF
(a) Speed response (b) Actual Isq current response

Estimated Isq current response

Control signal Vsq response.
From the obtained result, it is observed that the speed tracking is smooth and quick. The disturbance is quickly rejected by the system and attains the actual speed. With the implementation of EKF algorithm, the transient time and control signal variation are reduced compare to the non observer.


CONCLUSION
A novel implementation of proportionalintegraldifferential equivalent controller using state observer based Extended Kalman Filter (EKF) for a Permanent Magnet Synchronous Motor (PMSM) is proposed. The proposed method accurately estimates the speed and current. Also the proposed approach mitigates speed deviation caused by the load disturbance and attenuates the control signal variation. The steady and transient state responses are also improved. The EKF algorithm method attains good speed tracking. The system performance is enhanced significantly using the proposed method.
REFERENCES

Chaoui, H. , Sicard, P., Adaptive vector control of permanent magnet synchronous machines with speed estimation, Proceedings of 38th Annual Conference on IEEE Industrial Electronics Society ,2012 , pp: 1660 1665.

Yang Jun , Liu Huixian , Li Shihua , Chen Xisong, Nonlinear disturbance observer based robust tracking control for a PMSM drive subject to mismatched uncertainties, Proceedings of 31st Chinese Control Conference (CCC), 2012,pp: 830 835.

Du Limin , Zhang Ke , Jin Guang , Linear auto disturbance rejection controller for vectorcontrolled PMSM drive system, Proceedings of International Conference on Transportation, Mechanical, and Electrical Engineering (TMEE),2011, pp: 879 882.

Mohamed, Y.A.R.I., Design and Implementation of a Robust Current Control Scheme for a PMSM Vector Drive With a Simple Adaptive Disturbance Observer, IEEE Transactions on Industrial Electronics, Vol.54, No.4, pp. 1981 1988,August 2007.
Minneapolis, 2000

P.Pillay and R. Krishnan, Application Characteristics of Permanent Magnet Synchronous and Brushless DC Motors for Servo Drives, IEEE Trans. Ind. Appl., vol. 27, no.5, pp. 986996, Sept./Oct. 1991.

Mohan, N. Advanced Electric Drives, Analysis, Control and Modeling using Simulink, MNPERE, Minneapolis, 2001.

Benchabane.F, Titaouine.A, Bennis,O.,Yahia, K.,Taibi, D., Systematic fuzzy sliding mode approach with Extended Kalman Filter for Permanent magnet Synchronous Motor, Proceedings of the IEEE international conference on Systems Man and Cybernetics (SMC), pp.21692174, 2010.

Bolognani S, Calligaro S, Petrella R, Sensorless control of IPM motors in the lowspeed range and at standstill by HF injection and DFT processing, IEEE Transactions on Industry Applications, vol. 47, pp. 96 104, January/February 2011.

Junggi lee, Jinseok Hong, Kwanghee Nam, Sensorless control of surface mount permanentmagnet synchronous motors based
on a nonlinear observer, IEEE Transactions on Power Electronics, vol. 25, No.2,pp. 290297, February 2010.

Zheng Zedong, Li Yongdong, Fadel M, Xiao Xi, A Rotor Speed and Load Torque Observer for PMSM Based on Extended Kalman Filter, Proceedings of the IEEE international conference on Industrial Technology,pp.233238,2006.

Yin Zhonggang, Zhang Ruifeng, Zhong Yanru, Cao Yu, Speed and Flux Estimation of Permanent Magnet Synchronous Motor for Sensorless Vector Control Based on Robust Extended Kalman Filter, IEEE International Symposium on Industrial Electronics (ISIE),2012, Pages: 748751.

P. Tomei C. M. and Verrelli, ObserverBased Speed Tracking Control for Sensorless Permanent Magnet Synchronous Motors with Unknown Load Torque, IEEE Trans. on Automatic Control, vol. 56, no. 6, pp. 14841488, June 2011.

TaegJoon Kweon, and DongSeok Hyun, Highperformance speed control of electric machine using lowprecision shaft encoder, IEEE Trans. Power Electronics, vol 14, pp. 838849A, September 1999.

Shedbalkar, K., Dhamangaonkar, A.P., Patil, A.B., Speed estimation using extended Kalman filter for PMSM, International Conference on Emerging Trends in Electrical Engineering and Energy Management (ICETEEEM), 2012, pp: 433 435.

A.Bado, S.Bolognani., Effective estimation of speed and rotor position of a PM synchronous motor drive by a Kalman filtering technique, in Proc. IEEE PESC'92, pp. 951 957, 1992.

Shieh, L. S., H.M. Dib, and S. Ganesan, Linear quadratic regulators with eigenvalue placement in a specified region, Automatica, Vol. 24, No. 6, 1988, pp. 819823.

Yongpeng Zhang, Akujuobi, C., Ali, W.H., Tolliver, C.L., Leangsan Shieh Load disturbance resistance speed controller design for PMSM, IEEE Transaction on Industrial Electronics, Vol. 53, No. 4, pp.11981208,August 2006.

E. R. Kalman, A new approach to linear filtering and prediction problems, Transactions of the ASMEJournal of Basic Engineering, vol. 82, no. Series D, 1960, pp. 3545.

G. Welch, and G. Bishop, An introduction to the Kalman filter, Technical Report 95041, Univ. of North Carolina at Chapel Hill, Dept. of Computer Science, 1995.

Dhaouadi R., Mohan N., Norum L., Design and implementation of an extended Kalman filter for the state estimation of a permanent magnet synchronous motor, IEEE Transactions on Power Electronics, vol. 6, pp. 491497, Jul 1991.

Silverio Bolognani, Luca Tubiana, and Mauro Zigliotto, Extended Kalman filter tuning in sensorless PMSM drives, IEEE Trans. Industry Applications, vol. 39, NO. 6, pp. 276281, November /December 2003.