Vous êtes sur la page 1sur 8

Nonlinear Model of Permanent-Magnet Synchronous Motors

Jun-qiang Lian, Shun-yi Xie , Wang Jian, Ping Hu

Nonlinear Model of Permanent-Magnet Synchronous Motors


1

Jun-qiang Lian, 2Shun-yi Xie , 3Wang Jian,4Ping Hu


Department of Weaponry Engineering., Naval University of Engineering,
Wuhan 430033, China
267122822@qq.com
2
Department of Weaponry Engineering., Naval University of Engineering,
Wuhan 430033, China
3
Higher Education Research Units., Naval University of Engineering,
Wuhan 430033, China
4
Department of Weaponry Engineering., Naval University of Engineering,
Wuhan 430033, China
1

Abstract
This paper develops a comprehensive nonlinear model of permanent-magnet synchronous
motors(PMSM) to be used for simulation and control models. The model is described in a new
expression to estimate the torque with low volume offline data. The offline data is constructed by
magnetic flue and cogging torque and inductance. When the magnetic saturation,the magnetic flue and
inductance is changed with the phase current and the position of rotor. In order to establish the
nonlinear relation among magnetic flue, inductance and phase current, and the position of rotor, least
squares support vector machine(LS-SVM) is used. The sample of LS-SVM is obtained from static finite
element analyses( FEA ) simulations. The experimental results show that the proposed model is more
accurate than the conventional model.

Keywords: Nonlinear Model, Least Squares Support Vector Machine, Finite Element Analyses
1. Introduction
The permanent-magnet synchronous motors(PMSM) are widely used in various electrical devices
due to their high torque-to-current, power-to-weight ratios and their high efficiency. The conventional
model of PMSM is d q axis frame, and this model assumes a constant value of induced
electromotive(EMF) due to permanent-magnets. The conventional model present method of calculating
the reluctance torque is performed under the assumption that all load current and the magnetic flue
flow only along the two-axis[1-3]. But due to the saturation in the rotor and stator cores the d q axis
quantities are no longer independent of each other. So the conventional model is difficult to accurate
calculate the magnet torque generated from direct- and quadrature-axis flux linkage due to permanentmagnets, it is difficult to calculate the inductance value, because the flux linkage due to permanentmagnets and inductance exist a nonlinear relation with current value and rotor position.
This paper presents a nonlinear model of PMSM, which uses the LS-SVM to predict flux linkage
due to permanent-magnets and inductance. The sample of LS-SVM is obtained from static finite
element analyses( FEA )simulations. The cogging torque is also taken account into the model,
the proposed model is build based on the Simulink model. By comparing with the simulation result
and experimental test, the proposed model is able to provide better accuracy.

2. Conventional model of PMSM


In the conventional two-axis machine model, The electrical subsystem of conventional two-axis
PMSM model is described by voltage equations[4,5]:

International Journal of Digital Content Technology and its Applications(JDCTA)


Volume6,Number7,April 2012
doi:10.4156/jdcta.vol6.issue7.15

119

Nonlinear Model of Permanent-Magnet Synchronous Motors


Jun-qiang Lian, Shun-yi Xie , Wang Jian, Ping Hu

d d
q Rs i d
dt
d q
d Rs iq
uq
dt
ud

(1)

The flux linkage equations of conventional two-axis PMSM model is described as:

d Ld id Lmf i f

(2)

q Lq i q
The torque equation of conventional two-axis PMSM model is described as:

Tem p( d iq q id ) p[ Lmd i f iq ( Ld Lq )id iq ]

(3)

In the equations(1), (2)and(3). ud and u q are direct and quadrature-axis voltage, id and iq are
direct and quadrature-axis current,

Rs

is phase resistance of armature coils, d and q are

direct- and quadrature-axis flux linkage.


Lmf

Ld

and

Lq are

inductance of direct- and quadrature-axis,

is permanent-magnet equivalent inductance, i f is permanent-magnet equivalent current,

respectively.

3. Proposed nonlinear model of PMSM


3.1. Torque calculated by magnetic co-energy
In electric motors, the output torque is expressed as follow[5-7]:

Te

W (i, )

Where W is magnetic co-energy, i is armature current, respectively. The


follow:

W Wi Wmi Wm

(4)
W

can be expressed as

(5)

Where Wi is the magnetic co-energy due to only the armature current, Wmi is the magnetic coenergy due to interaction of armature and the magnet, Wm is the magnetic co-energy due only
the magnet. Through analyzing the three energy terms by energy theory, they can expressed as
follow[8]:

dWi 1 dL 2
p
i
2 d
d
dWmi
d
pi
d
d

dWm
Tcogging
d

(6)
(7)
(8)

120

Nonlinear Model of Permanent-Magnet Synchronous Motors


Jun-qiang Lian, Shun-yi Xie , Wang Jian, Ping Hu

In the formula (6), (7) and (8), p is the number of pole pairs, L is inductance, is
magnetic flux, Tcogging is cogging torque. Then the torque equation can expressed as follow[1315]:
Te

1 dL 2
d
Tcogging
p
i pi
2 d
d

(9)

The voltage, flux linkage equation are given in equation (10) and (11).

d
dt

(10)

Li m

(11)

V Ri

3.2. Inductance and magnetic flux linkage calculate use LS_SVM


3.2.1. Theory of LS-SVM

The main objective of regression estimation is to approximate a function f(x) from a given set of
samples ( xi , yi ) . SVM approximates the function as follows[9]:

f ( x) ( x) b

(12)

{ ( x )}

D
i 1

denotes a set of non-linear transformations from the low dimensional space to the high
dimensional feature space. SVM regression is formulated as minimization of the following function:
min ( )

n
1
2
C (i i* )
2
n 1

(13)

*
Subject to: yi xi b i and xi b yi i . where is called tube size, i and i* ,
i are the slack variables, C is the regularization parameter. By introducing Lagrange multipliers,
decision function (12) takes the following form:
n

f ( x) ( ai ai* )K ( x, xi ) b

(14)

i 1

*
Where K ( x, xi ) is kernel function. ai and ai are Lagrange multipliers which are obtained by
maximizing the dual form of the function (13). The dual form is as follows:

i 1

i 1

max W (a ) yi (ai ai* ) (ai ai* )


n

Subject to: xi (ai ai* ) 0 ,


i 1

0 ai , a*j C

1 n
(ai ai* )(a j a*j )( xi j )
2 i 1

(15)

i 1, 2, , l .

LS-SVM is trained by the following equation:


1
1 n
min J T ei2
2
2 i 1

(16)

Subject to the equality constraints: yi T ( xi ) b ei .


Then the optimization problem is expressed as the following linear equation:

121

Nonlinear Model of Permanent-Magnet Synchronous Motors


Jun-qiang Lian, Shun-yi Xie , Wang Jian, Ping Hu

1v

Where y [ y1 , , yn ] ,

b 0
1Tv

1
K I a y

1v [1,1]T , a [a1 , , an ] , K kij ,


nn

then we gain the nonlinear model of LS-SVM:

f ( x) ( ai ai* )K ( x, xi ) b

(17)

i 1

3.2.2. LS-SVM model for calculate inductance and magnetic flux linkage

Theoretically the inductance and magnetic flux linkage of PMSM is not only the function of rotor
position but also the winding currents, then we use LS-SVM express the nonlinear relation between
them. the LS-SVM model for calculating the inductance and magnetic flux linkage of PMSM can be
expressed as follow:
n

L( x ) ( ai ai* )K ( x, xi ) b

(18)

i 1

( x) (ai ai* )K ( x, xi ) b

(19)

i 1

Where x represents the winding current i and the rotor position . The sample of LS-SVM is
obtained from static finite element analyses( FEA )simulations.
Laa (ia , ib , ic , )
Lab (ia , ib , ic , )
Lac (ia , ib , ic , )
Lbb (ia , ib , ic , )
Lba (ia , ib , ic , )
Lbc (ia , ib , ic , )
Lcc (ia , ib , ic , )
Lca (ia , ib , ic , )
Lcb (ia , ib , ic , )

ia
ib
ic

Figure 1. LS-SVM model for inductance

ia
ib
ic

ma (ia , ib , ic , )
LS_SVM
Model for PM flux
linkage

mb (ia , ib , ic , )

mc (ia , ib , ic , )

Figure 2. LS-SVM model for PM flux linkage

4. Simulation use nonlinear model of PMSM


4.1. Cogging torque and sample for LS-SVM model use FEA

In the nonlinear model of PMSM, the cogging torque data and the sample for LS-SVM model of
PMSM are obtained use the method of FEA. We firstly build the FEA model of PMSM. The
mathematics formula for magnetic vector potential and boundary condition express as equation (20),
FEA model of PMSM show as Figure 3. Its parameters list in Table 1. We use ansoft Maxwell to solve
the FEA model of PMSM by parameterization method. We can obtain the data which is needed for
nonlinear model of PMSM[10],[11],[12].
2 AZ

2 AZ 2 AZ
1 AZ
1 AZ

J Z , AZ 0, AZ AZ ,

J m
x 2
y 2
1 n 12 2 n 12

(20)

122

Nonlinear Model of Permanent-Magnet Synchronous Motors


Jun-qiang Lian, Shun-yi Xie , Wang Jian, Ping Hu

Figure 3. FEA model of PMSM


Table 1. Parameters of PMSM
The out
The out The inner
diameter of diameter of diameter of
stator(mm) stator(mm) stator(mm)
371

174.2

174

The inner
diameter of
stator(mm)

Pole
pairs

54

Slot
number
24

rated
rated
voltage(V) power(KW)
220

15

4.2. Simulink block of nonlinear model PMSM

The Simulink block of nonlinear model PMSM is build based on formula (4)-(19), which contain
mainly 3 blocks as Figure4-6.

Figure 4. The simulink block for voltage calculation of nonlinear model PMSM

123

Nonlinear Model of Permanent-Magnet Synchronous Motors


Jun-qiang Lian, Shun-yi Xie , Wang Jian, Ping Hu

Figure 5. The main simulink block for nonlinear model PMSM

Figure 6. The simulink block for control frame of nonlinear model PMSM
4.3. Analyze the result of simulation and experiment

Firstly, we use the method of FEA, and cogging torque data and 50 samples for LS-SVM
model of PMSM are obtained. The parameters of LS-SVM model as: 3000 , 0.1 . Then the
data which is input simulink blocks are obtained. The simulation time is 46.5s, the result of
simulation is shown as Figure7-8, In order to validate the accuracy of the proposed model, we
put up the experiment platform, and conventional simulation model of PMSM is build too. The
result of experiment and simulation are shown as Figure 9-10, the average amplitude of torque
ripples of proposed model is 0.23N*m, the average amplitude of torque ripples of conventional

124

Nonlinear Model of Permanent-Magnet Synchronous Motors


Jun-qiang Lian, Shun-yi Xie , Wang Jian, Ping Hu

model is 0.12N*m, the average amplitude of torque ripples of experiment is 0.26N*mm, So the
accuracy of the proposed model is much higher than that of conventional model.

Figure 7. Result of simulation for torque


nonlinear model

Figure 8. Result of simulation for cogging torque of


of nonlinear model
15

Torque(N*m)

10

-5

-10
0

Figure 9. Result of simulation for torque


of conventional model

0.05

0.1
t(s)

0.15

0.2

Figure 10. Measured torque of experiment

5. Conclusions
We have created a nonlinear model of PMSM that accounts for several nonlinear factors,
which affects the operation of PMSM. In order to establish the nonlinear relation among
magnetic flux linkage, inductance and phase current, and the position of rotor, the least squares
support vector machine(LS-SVM) is used. The sample of LS-SVM is obtained from static finite
element analyses( FEA )simulations. The simulation results show that the proposed model is
more accurate than the conventional model. It can be used in simulation of PMSM control
frames.

6. References
[1] Noriya Nakao, Kan Akatsu, Torque Ripple Compensation Based on Instantaneous Torque
Estimation in Permanent Magnet Synchronous Motors, 2011 IEEE international Electric &
Drives Conference, pp.989-994, 2011.
[2] O. A. Mohanmmed. S. Liu, Z. Liu, Phase-variable model of PM synchronous machines for
integrated motor drive, IEE. Proc-Sci. Meas. Technol, Vol.151, No.6, pp. 423429, 2004.
[3] Calum Cossar, Mircea Popescu, A General Magnetic-Energy-Based Torque Estimator: Validation
via a Permanent-Magnet Motor Drive, IEEE TRANSACTIONS ON INDUSTRY
APPLICATIONS, vol. 40, no. 4, pp.1210-1216, 2008.
[4] Ping Zhou, Dingsheng Lin, Determination of - Axis Parameters of Interior Permanent Magnet
Machine, IEEE TRANSACTIONS ON MAGNETICS, vol. 40, no. 4, pp.3125-3128, 2011.
[5] W. N. Fu, P. Zhou, D. Lin, S. Stanton, and Z. J. Cendes, Magnetic Force Computation in
Permanent Magnets Using a Local Energy Coordinate Derivative Method, IEEE
TRANSACTIONS ON MAGNETICS, vol. 40, no. 2, pp. 683-686.
[6] P. Zhou, W. N. Fu, D. Lin, S. Stanton, and Z. J. Cendes, Numerical modeling of magnetic
devices, IEEE TRANSACTIONS ON MAGNETICS, vol. 40, no. 4, pp. 18031809, 2004.

125

Nonlinear Model of Permanent-Magnet Synchronous Motors


Jun-qiang Lian, Shun-yi Xie , Wang Jian, Ping Hu

[7] L. Parsa and L. Hao, Interior permanent magnet motors with reduced torque pulsation, IEEE
Trans. Ind. Electron., vol. 55, no. 2, pp. 602609, 2008.
[8] Z. Xu and M. F. Rahman, An adaptive sliding stator flux observer for a direct-torque-controlled
IPM synchronous motor drive, IEEE Trans. Ind. Electron., vol. 54, no. 5, pp. 23982406, 2007.
[9] Suykens, J.A.K., Vandewalle, J.: Least squares support vector machine classifiers. Neural
Processing Letters vol. pp. 9293300, 1999.
[10] Z. Q. Zhu and D. Howe, Influence of design parameters on cogging torque in permanent magnet
machines, IEEE Trans. Energy Conv. vol. 15, no. 4, pp. 407412, 2000.
[11] L. Lu, Z. Chen, B. Yao, Q. Wang: Desired Compensation Adaptive Robust Control of a LinearMotor-Driven Precision Industrial Gantry With Improved Cogging Force Compensation, IEEE
Trans. Mechatronics, vol. 13, no. 6, pp. 617624, 2008.
[12] Y. Liu, Z. Q. Zhu, D. Howe: Direct Torque Control of Brushless DC Drives with Reduced
Torque Ripple, IEEE Trans. Industry Applications, vol. 41, no. 2, pp. 599608, 2005.
[13] Qibei Lu, , Chunhua Ju, "Research on Credit Card Fraud Detection Model Based on Class
Weighted Support Vector Machine", JCIT, vol. 6, no. 1, pp. 62 ~ 68, 2011
[14] Linkai Luo, Meixiang Luo, Lingjun Ye, Hong Peng, Fan Yang, "Maximum Robustness Criterion
on Kernel Selection of Support Vector Machine", JCIT, vol. 6, no. 1, pp. 294 ~ 306, 2011
[15] Wei LIU, Yuhua YAN, "Coal Gangue Interface Detection based on IMF Energy and SVM",
JDCTA, vol. 5, no. 4, pp. 160 ~ 166, 2011

126