TP03: Backstepping Control of Permanent Magnet Synchronous Motors

PMSM MODEL
class of feedback linearisable nonlinear systems exhibiting
The setting in the state form of the PMSM model allows the constant uncertainty, and it guarantees global regulation
simulation of this latter. In the rotor rotating ( d  q ) and tracking for the class of nonlinear systems
reference frame, the PMSM stator current model is transformable into the parametric-strict feedback form. The
described as follows [18], [19]: backstepping design alleviates some limitations of other
approaches [18, 20, 21 ]. It offers a choice of design tools to
 x  f ( x)  Bu  DTL accommodate uncertainties and nonlinearities and can


 x  x1 x2 x3   id iq r
T T
 avoid wasteful cancellations.
The idea of backstepping design is to select recursively
 some appropriate functions of state variables as pseudo-
u d  Vd 
T
 b1 0 0 (1)
u  u  V  ; B   0 b 0 control inputs for lower dimension subsystems of the
  q q  2  overall system. Each backstepping stage results into a new

 D  0 0 d 
T pseudo-control design, expressed in terms of the pseudo-
control designs from the preceding design stages. When the
With the following expression of field vector f (x) : procedure terminates, a feedback design for the true control
input results and achieves the original design objective by
 f1 ( x)  a1 x1  a2 x2 x3 virtue of a Lyapunov function, which is formed by
 summing up the Lyapunov functions associated with each
 f 2 ( x)  a3 x2  a4 x3  a5 x1 x3 (2)
 f ( x)  a x  a x  a x x individual design stage [21, 17, 22].
 3 6 2 7 3 8 1 2 The control objective in this case is to force the PMSM
The components of this vector are expressed according to speed ( r  x3 ) to follow its reference x3* and maintain in
the PMSM parameters as follows: the same time the direct current ( id  x1 ) to zero under load
 Rs Lq R f L torque disturbance. The application of the backstepping
a1   ; a2  ; a3   s a4   ; a5  d control strategy to the PMSM in this case is divided into
 Ld Ld Lq Ld Lq two steps (see [18, 20]).

 np f
2
f np f
2

a 6   ; a7   ; a8  ( Ld  Lq ) 1. Speed regulator:
 J J J
 np 1 1 This first step consists to identify the error e which
d   ; b1  ; b2   represents the error between real speed r  x3 and
 J L d Lq
 reference r*  x3* . In this case we control x3 by x2 .
Where : id , iq : d, q axis stator current;
Let the Lyapunov function:
Vd , Vq : d, q axis stator voltage;
1 2 1
Ld , Lq : d, q axis stator inductance; V1  e  ( x3  x3* ) 2 (3)
2 2
Rs : Stator resistance; Whose derivative is:
f : Rotor permanent magnet flux.
V 1  e e  ( x3  x3* )( x3  x3* ) (4)
r : Mechanical rotor speed ( r  n p  )
The error derivative is given by:
f : Viscous friction coefficient
LT : Load torque e  a6 x2  a7 x3  dTL  x3* (5)
J : Moment of Inertia
If we selected stabilizing functions as follows:
As presented in the appendix we take in this paper in 1
PMSM with smooth poles Ld  Lq  L in this case (a8  0) x 2*  ( a7 x3  K1e  dTL  x 3* ) (6)
a6
The use of the classical controllers such as the Where k1  0
proportional and integral controller (PI) is insufficient to
provide good speed tracking performance. To overcome
Then the derivative of Lyapunov function V 1 is written as:
these problems, a robust controller based on backstepping
control approach is proposed.
V1   K1e2  0 (8)
III. BACKSTEPPING CONTROL TECHNIQUE This guarantees convergence of the speed  r to its
The Backstepping is a systematic and recursive design reference x3* with robustness respect to load torque
methodology for nonlinear feedback control. This approach disturbance.
is based upon a systematic procedure for the design of
feedback control strategies suitable for the design of a large

1
2. Direct and Quadrature currents regulator: 1
uq  (  f2 ( x )  x 2*  K 4 ( x 2  x 2* )  a6 ( x3  x3* )) (20)
The second step consists to control the currents id  x1 and b2
iq  x2 by the voltages ud  Vd and uq  Vq ; where From the second term of (14)
x1  x1*  0 and x2  x2* eq eq  ( x2  x2* )( f1 ( x)  b2uq  x2* ) (21)
Consider the following Lyapunov function:
By replacing (20) in (21) we get:
1 1 1
V  e2  eq2  x12 (9) eq eq   K 4 ( x2  x2* ) 2  a6 ( x3  x3* )( x2  x2* ) (22)
2 2 2
Where e  ( x3  x3* ) and eq  ( x2  x2* ) Where K 4  0 . Finally, by grouping terms (19) and (22) we
The derivative of V with respect to time is: obtain:

V  e e  eq eq  x1 x1 (10) V   K 4 ( x 2  x 2* )2  a6 ( x3  x3* )( x 2  x 2* ) 

(23)
a ( x  x * )( x  x * )  V
6 2 2 3 3 1
From (10) and in order to control x2 by u d . The term of the
derivative V can be written as: By simplification:

x1x1  x1 ( f1 ( x)  b1ud ) (11) V   K 4 ( x2  x2* )2  V1 (24)

If we take the first control low as: From (8) and (24) we get:
1 V  K 4 ( x2  x2* )2  K1 ( x3  x3* )2 (25)
ud  (  f1 ( x )  K 3 x1 ) (12)
b1
Finely, from the foregoing, it is clear that it suffices to
Then this term is written: properly select the different gains K i ( i  1, 2, 3, 4 ) for the
set-negativity of the derivative of the complete Lyapunov
x1 x1   K3 x12  0 (13)
function ( V  0 ) overall V defined by (25). This implies
Where K 3  0 then the convergence of x1 to 0 is ensuring. that all the error variables are globally uniformly bounded
and maintain the system closed loop performance in
The remaining terms of (10) Let e  e e  eq eq as: presence of load torque disturbances.

e  ( x3  x3* )( x3  x3* )  ( x2  x2* )( x2  x2* ) (14)

Rated Values Power 22 w
The first term can be written as: Frequency 50 Hz
np 2
e e  ( x3  x3* )( a6 x2  a7 x3  dTL  x3* ) (15)
Rated parameters Rs 3.4 
or by adding and subtracting the term a6 x 2* we get: Ld 0.0121 H
Lq 0.0121
e e  ( x3  x3* )( a6 ( x2  x2* )  a6 x2  a7 x3  dTL  x3* ) (16) H
f 0.4212 H
By simplification :
J 0.0001 Kg.m2
e e  a6 ( x2  x2* )( x3  x3* )  f 0.0005 IS
(17)
( x3  x3* )( a6 x2*  a7 x3  dTL  x3* )

By replacing the term a6 x 2* presented in (6) in this last

equation we get:

e e  a6 ( x 2  x 2* )( x3  x3* ) 
(18)
( x3  x3* )(  K1e  dTL )

From (6) and (18) we get:

e e  a6 ( x2  x2* )( x3  x3* )  V1 (19)

Otherwise if one chooses:

a1 a2 a3 a4 a5 a6 a7 a8 a9 b1 b2 d

-297.52 1 -297.52 -1.074 -1 520 -0.5 0 0.026 82.644 82.644 -20000