You are on page 1of 12

Electrical Power and Energy Systems 29 (2007) 348–359

www.elsevier.com/locate/ijepes

Robust controller for synchronous generator with local load via VSC
J. Cabrera-Vázquez a, Alexander G. Loukianov b,*
, José M. Cañedo b, Vadim I. Utkin c

a
Universidad de Guadalajara, Centro Universitario de Ciencias Exactas e Ingenierı́as, Departamento de Electrónica, Av. Revolución No. 1500, Módulo
‘‘O’’, Apdo. Postal 44840, Guadalajara Jalisco, Mexico
b
Centro de Investigación y de Estudios Avanzados del IPN, Apdo. Postal 31-438, Plaza La Luna, C. P. 44550, Guadalajara, Jalisco, Mexico
c
Department of Electrical Engineering, The Ohio-State University, Columbus, OH 43210-1272, United States

Abstract

The objective of this paper is to design a nonlinear observer-based excitation controller for power system comprising a single synchro-
nous generator connected to an infinite bus with local load. The controller proposed is based on the using first singular perturbation
systems concepts and then Sliding Mode Control technique combining with Block Control Principle. To reduce ‘‘chattering’’ a nonlinear
observer with estimation of the mechanical torque and rotor fluxes is designed. This combined approach enables to compensate the
inherent nonlinearities of the generator and to reject external disturbances.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Synchronous generator; Singular perturbations; Sliding mode; Nonlinear observer; Lyapunov stability

1. Introduction trol algorithms based on a reduced-order power system


model, are sensitivities to the plant parameter variations
Design of robust stabilizing feedback controllers for and external disturbances, do not taken in consideration
power systems with modeled and unmodeled uncertainties practical limitation on the control and the effects of the
remains one of the important problems in control theory. exciter system and load dynamics. Additionally, these con-
A fruitful and relatively simple approach is based on the trol schemes do not consider the generator voltage regula-
use the concept of Variable Structure Systems with sliding tion and the plant state estimation problems.
mode [1]. This scheme provides, first and foremost, perfor- The purpose of this investigation is to design the excita-
mance robustness, and secondly decomposition and sim- tion control for the single machine infinite bus system
plicity of the control design procedure. (SMIB) with a local load (see Fig. 1). In this paper, we shall
It is known that the model of power system is highly resort to sliding mode [1] and block control [30] techniques
nonlinear and therefore, during last decade several modern combined with singular perturbation methods, to obtain a
control approaches including adaptive linear control [2–4], simpler controller, computationally law demanding, that
methods based on the passivity principle [5–9], fuzzy logic takes into account structural constraints. The main feature
and neural networks [10–15], control based on the direct of the proposed control is robustness to disturbances and
Lyapunov method [16,17], feedback linearization technique plant parameter variations. The advantage of singular per-
[18–23] have been used to design continuous nonlinear con- turbation method in the control application is the possibil-
trol algorithms which overcome the known limitations of ity of dealing with reduced order model instead of the full-
traditional linear controllers [24–29]. Basically, these con- order system. A switching function is formed using block
control linearization technique [30] and neglecting the fast
*
dynamics of the exciter system. But discontinuities in the
Corresponding author.
E-mail addresses: javier.cabrera@cucei.udg.mx (J. Cabrera-Vázquez),
sliding mode control can excite the unmodeled exciter
louk@gdl.cinvestav.mx (A.G. Loukianov), canedo@gdl.cinvestav.mx dynamics, leading to oscillations in the state vector. This
(J.M. Cañedo), utkin@ee.eng.ohio-state.edu (V.I. Utkin). phenomenon is known as ‘‘chattering’’ [1,31]. To prevent

0142-0615/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijepes.2006.09.001
J. Cabrera-Vázquez et al. / Electrical Power and Energy Systems 29 (2007) 348–359 349

where is = (id, iq)T, ir = (if, ig, ikd, ikq)T, Vs = (Vd, Vq)T,


Vr = (Vf, 0, 0, 0)T, id and iq are the direct-axis and quadra-
ture-axis stator currents; if is the field current; ikd, ikq and ig
are the direct-axis and quadrature-axis damper windings
currents; x is the angular velocity; Vd and Vq are the di-
rect-axis and quadrature-axis terminal voltages; Vf is the
excitation voltage; Rs and Rf are the stator and field resis-
tances; Rg, Rkd and Rkq are the damper windings resis-
tances; Ld and Lq are the direct-axis and quadrature-axis
Fig. 1. Single machine-infinite bus system, SMIB with local load. self-inductances; Lf is the rotor self-inductance; Lkd and
Lkq are the direct-axis and quadrature-axis damper wind-
ings self-inductances; and Lmd and Lmq are the direct-axis
these oscillations a nonlinear observer is designed and the and quadrature-axis magnetizing inductances.
ideal sliding mode in the auxiliary observer loop is The torque Te can be expressed in terms of the currents
produced. as follows:

2. Plant model T e ¼ ðLq  Ld Þid iq þ Lmd iq ðif þ ikd Þ  Lmq id ðig þ ikq Þ ð4Þ

The exciter system model is given by


The complete mathematical description of a single
machine infinite-bus system that will be considered includes sf V_ f ¼ V f þ bf u ð5Þ
the mechanical and electrical dynamics of the three phase
synchronous machine (including both the field and damper where u is the control input, and sf is the time constant.
windings) with exciter system and external networks with The mechanical input torque Tm is assumed to be con-
load constraints. stant. Thus,
The mechanical dynamics are described by the swing
equations given by [32] T_ m ¼ 0 ð6Þ
dd The equilibrium equations for the external networks of
¼ x  xs ð1Þ
dt the synchronous machine connected to an infinite bus can
2H dx be written for the generator dynamics as
¼ Tm  Te ð2Þ
xs dt   "1 #" # "R # 
d ied Le
0 VdV1 d
e
Le
x ied
where d is the power angle of generator; x is the angular ¼  ð7Þ
dt ieq 0 L1e VqV1 q x RLee ieq
velocity; xs is the rated synchronous speed, H is the inertia
constant; Tm is the mechanical torque applied to shaft; and and for the load dynamics as
Te is the electrical torque.        
After Park’s transformation, the electrical dynamics d iLd Vd iLd iq þ ieq
LL ¼  RL  xLL : ð8Þ
which include both the rotor and stator windings, using dt iLq Vq iLq id þ ied
the currents as the state variables, can be expressed as
The relations in terminals of the generator are given by
follows:       
" #     Vd 1 id  ied iq þ iLq
dis
dt is Vs ¼ ðI  Adq Þ RL þ xLL
L di ¼ GðxÞ þ ð3Þ Vq iq  iLq id þ ied
r ir Vr     " 1#  )
dt
2 3 f7 h2 1
Vd ied
Ld 0 Lmd 0 Lmd 0 þLL þ LL V f  Adq 1  Ax
f8 0 V iLq
60 Lq 0 Lmq 0 Lmq 7
q
6 7
6 7 ð9Þ
6 Lmd 0 Lf 0 Lmd 0 7
L ¼6 60
7;
6 Lmq 0 Lg 0 Lmq 7
7 where ie = (ied, ieq)T is the current in the transmission line,
6 7 iL = (iLd, iLq)T is the current in the loads, kr = (kf, kg, kkd,
4 Lmd 0 Lmd 0 Lkd 0 5
kkq)T is the rotor flux, V1 is the value of the infinite-bus
0 Lmq 0 Lmq 0 Lkq voltage; Le and Re are the transmission line resistance
2 3
Rs xLq 0 xLmq 0 xLmq and inductance, respectively, LL and RL are the load resis-
6 xLd Rs xLmd 0 xLmq 0 7 tance and inductance, respectively,
6 7
6 7
60 0 Rf 0 0 0 7 f7 ¼ a71 sinðdÞ þ a73 kf þ a75 kkd  a77 id þ xða74 kg
G ¼6
60
7
7
6 0 0 Rg 0 0 7  a76 kkq þ a78 iq Þ;
6 7
40 0 0 0 Rkd 0 5 f8 ¼ a81 cosðdÞ þ a84 kg þ a83 kkq  a88 iq
0 0 0 0 0 Rkq þ xða83 kf  a85 kkd þ a87 id Þ;
350 J. Cabrera-Vázquez et al. / Electrical Power and Energy Systems 29 (2007) 348–359
" # "L #
LL a71  LLLe 0 L
Le
0 where x = (x1, . . ., x10)T, x1 = d, x2 = x, x3 = kf,
Adq ¼ ; A1
dq ¼ ;
0 LL a81  LLLe 0 LL
Le
T
x4 ¼ ðkg ; kkd ; kkq Þ ¼ ðx4 ; x5 ; x6 Þ ;
T
" Re L #  
Le
L
LL x 1 0 T
x5 ¼ ðid ; iq Þ ¼ ðx7 ; x8 Þ ;
T
Ax ¼ Re LL
; I¼ ;
LL x Le
0 1 T
x6 ¼ ðied ; ieq Þ ¼ ðx9 ; x10 Þ ;
T

where ai,j, i = 7,8,j = 1, . . ., 7 are positive constant parame- z¼Vf


ters depended on the power system parameters. f3 ðxÞ ¼ a33 x3 þ a35 x5  a37 x7
Analyzing relations between the fluxes and currents
f2 ðxÞ ¼ a24 x7 x4  a25 x8 x5 þ a26 x7 x6 þ a28 x7 x8 ;
shows that the sensitivity of the fluxes with respect to
parameter variations is lower than that of the currents. b2 ðx5 Þ ¼ a23 x8 ; b5 ¼ ð0; b7 ÞT ; d 2 ¼ xs =2H ;
Therefore, it is more suitable the representation of the f 5 ðxÞ ¼ ðf7 ðxÞ; f8 ðxÞÞ ; T

electrical dynamics in terms of the stator current is and


rotor flux kr. Such model can be obtained from (3) using f 6 ¼ ðf9 ðx2 ; x6 ; V d ; V q Þ; f10 ðx2 ; x6 ; V d ; V q ÞÞT ;
the following transformation between the fluxes and f7 ¼ a71 sin x1 þ a73 x3 þ a75 x5  a77 x7
currents: þ x2 ða74 x4  a76 x6 þ a78 x8 Þ;
   
is is f8 ¼ a81 cos x1 þ a84 x4 þ a86 x6  a88 x8
¼T ð10Þ
kr ir þ x2 ða83 x3 þ a85 x5  a87 x7 Þ;
where 1 Re
f9 ¼ ðV d  V 1 d Þ x9 þ x2 x10 ;
Le Le
1 Re
f10 ¼ ðV q  V 1 q Þ x10  x2 x9 ;
Le Le
2 3 2 3
a44 0 a46 0 a48
6 7 6 7
A44 ¼ 4 0 a55 0 5; A45 ¼ 4 a57 0 5;
a64 0 a66 0 a68

where aij (i = 2, . . ., 8; j = 1, . . ., 9), b3 and b7 are positive


constant parameters depending on Rs, Rf, Rg, Rkd, Rkq, Re,
RL, Ld, Lq, Lkd, Lkq, Lmd, Lmq, Le, LL and V1. It is assumed
The transformation (10) reduces system (3) to the following that the power angle x1, the terminal voltage Vg, the speed x2
form: and the stator currents x7 and x8 are available for measure-
" #     ment, and that the control input u(t) should remain in:
dis
dt is Vs
dkr
¼ A e ðxÞ þ Be ð11Þ juðtÞj 6 U m ð14Þ
dt
kr Vr
with Um > 0.
with Ae(x) = TL1G(x)T1 and Be = TL1. The torque
Te can be expressed now in terms of the currents and fluxes 3. Ideal sliding mode
as
T e ¼ a23 kf iq þ a24 kg id  a25 kkd iq þ a26 kkq id þ a28 id iq ð12Þ The control objectives are first to stabilize the rotor
angle x1 and make the speed x2 be equal to the rated syn-
where a2,i, i = 1, . . ., 8 are positive constant parameters de- chronous speed xs. Secondly, the voltage regulation is
pended on the generator parameters. important as well. To simplify the control algorithm, we
The state variables as well as the parameters of the model set sf = 0 in (13d) that results in
(1)–(12) are expressed in per unit. Combining Eqs. (1)–(12)
and using relationship (10), the complete model of the SMIB z ¼ bf u ð15Þ
is represented in the nonlinear state-space form: Substituting (15) in (13a)–(13f) and (13g), yields
x_ 1 ¼ x2  xs ð13aÞ x_ 1 ¼ x2  xs ð16aÞ
x_ 2 ¼ f2 ðx4 ; x5 Þ þ b2 ðx5 Þx3 þ d 2 T m ð13bÞ x_ 2 ¼ f2 ðx4 ; x5 Þ þ b2 ðx5 Þx3 þ d 2 T m ð16bÞ
x_ 3 ¼ f3 ðx4 ; x5 Þ þ b3 z ð13cÞ x_ 3 ¼ f3 ðx4 ; x5 Þ þ b3 u ð16cÞ
sf z_ ¼ z þ bf u ð13dÞ x_ 4 ¼ A44 x4 þ A45 x5 ð16dÞ
x_ 4 ¼ A44 x4 þ A45 x5 ð13eÞ x_ 5 ¼ f 5 ðxÞ þ b5 u ð16eÞ
x_ 5 ¼ f 5 ðxÞ þ b5 z ð13fÞ x_ 6 ¼ f 6 ðx2 ; x6 ; V d ; V q Þ ð16fÞ
x_ 6 ¼ f 6 ðx2 ; x6 ; V d ; V q Þ ð13gÞ where b3 ¼ b3 bf and b5 ¼ b5 bf .
J. Cabrera-Vázquez et al. / Electrical Power and Energy Systems 29 (2007) 348–359 351

3.1. Speed control design where x11 is the steady-state value of angle x1 corre-
sponded to the value of mechanical torque Tm. Addition-
The subsystem (16a), (16b) and (16c) has the Block Con- ally, from rx(t) = 0 for t P ts it follows:
trollable form [30]. Therefore, this subsystem can be linear-
x3 ðtÞ ¼ x3d ðx4 ðtÞ; x5 ðtÞ; T m Þ;
ized using block control technique by the following
transformation: 1 ð27Þ
x3d ¼ ½f2 ðx4 ; x5 Þ þ d 2 T m 
b2 ðx5 Þ
z1 ¼ x1 :¼ u1 ðxÞ
The basic property of the control for sliding mode is that
z2 ¼ x2  xs :¼ u2 ðxÞ
  the control provides the subspace {n = (x1,x2,x3)T = (x11
1 f2 ðx4 ; x5 Þþ ,xs,x3d)T,g 2 R7} invariant where g = (x4,x5,x6)T. The
rx ¼ x3  :¼ u3 ðx; T m Þ
b2 ðx5 Þ þd 2 T m  k z ðx2  xs Þ dynamics of vector g on this subspace are the zero dynam-
ð17Þ ics. To obtain these dynamics, it is necessary to substitute
the equivalent value of the control on the invariant sub-
where kz > 0. The transformed subsystem (16a), (16b) and space, uxeq ðx11 ; xs ; x31 ; g; T m Þ (24) and the values
(16c) then is represented in new variables z1, z2 and rx as x1 = x11, x2 = x21 = xs (26) and x3 = x3d (27) in (16d),
z_ 1 ¼ z2 (16e) and (16f), that yields
z_ 2 ¼ k z z2 þ b2 ðx5 Þrx ð18Þ g_ ¼ f g ðx11 ; xs ; x3d ; g; V d ; V q ; Þ
r_ x ¼ fx ðx; T m Þ þ 
bx ðxÞu ð19Þ þ bg uxeq ðx11 ; xs ; x3d ; g; T m Þ ð28Þ
where where fg = (f4(g), f5(x11,xs,x3d,g), f6(xs,g,Vd,Vq))T,
fx ¼ ou 3
ox2
ðf2 þ b2 x3 þ d 2 T m Þ þ ou 3
ox3 3
f þ ou
ox4
3
ðA44 x4  A45 x5 Þ þ f4(g) = A44x4 + A45x5, bg ¼ ð0; 0; 0; b7 ; 0; 0; 0Þ
T
and

ou3
f , 
b ¼ 
b  b 7
ða x þ a x  a 
28 8 Þ; and bx ðtÞ is a
x b7 ¼ b7 bz .
ox5 5 x 3 a23 x8 24 4 26 6
positive function of time. Dividing the right part of this system into linear and
A discontinuous control law is proposed as nonlinear parts, then the zero dynamics can be represented
u ¼ U m signðrx Þ; U m > 0: ð20Þ as

Sliding mode stability can be analyzed using the following _


g¼A g gþgg ðx11 ;xs ;x3d ;g;T m ;V d ;V q Þ
2 3
candidate of Lyapunov function: a44 0 a46 0 a48 0 0
60 a55 0 a57 0 0 0 7
1 6 7
V x ¼ r2x : ð21Þ 6 7
2 6 a64 0 a 0 a 0 0 7
6 66 68 7
Find the time derivative of (21) along the trajectories of 6 7
whereAg ¼6 xs a74 a75 xs a76 a77 xs a78 0 0 7;
system (19) with control (20): 6 7
6 a84 xs a85 a86 xs a87 a88 0 0 7
6 7
V_ x ¼ fx ðx; T m Þrx  U m 
bx ðxÞsignðrx Þrx 6 7
40 0 0 0 0 a99 xs 5
6 jrx j½U m bx ðxÞ  fx ðx; T m Þ: ð22Þ 0 0 0 0 0 xs a10;10
2 3
0
Under the following condition:
6 a x ðg;T Þ 7
U m > juxeq ðx; T m Þj ð23Þ 6 53 3d m 7
6 7
60 7
6 7
with equivalent control uxeq [1] calculated from r_ x ¼ 0 (19) 6 a cosx þ b ux ðx ;x ;x ;g;T Þ 7
gg ¼6 71 11 7 eq 11 s 3d m 7:
as 6 7
6 a81 sinx11 7
6 7
bx ðxÞÞ1 fx ðx; T m Þ
uxeq ðx; T m Þ ¼ ð ð24Þ 6 d ðV V 1 Þ 7
4 9 d d 5
the value of V_ x (22) is negative. Therefore, the state will d 9 ðV q V 1
q Þ
reach the manifold rx = 0 after finite time interval [0,ts]
[1]. Once this is achieved, the sliding motion on rx = 0 is It can be shown that the matrix Ag is Hurwitz, and the
governed by the following second-order system: nonlinear term gg(Æ) is bounded. Therefore, a solution of
Eq. (28) is ultimately bounded [33].
z_ 1 ¼ z2
ð25Þ
z_ 2 ¼ k z z2 3.2. Voltage regulator design
with desired eigenvalue kz, that corresponds to the linear-
The second objective of the excitation control is to reg-
ized mechanical dynamics. If kz > 0 then limt!1z2(t) = 0
ulate the voltage Vg at the terminals of the synchronous
and
generator,
x1 ðtÞ ! x11 ¼ const qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
as t ! 1 ð26Þ V g ¼ jVs j ¼ V 2d þ V 2q : ð29Þ
x2 ðtÞ ! x21 ¼ xs
352 J. Cabrera-Vázquez et al. / Electrical Power and Energy Systems 29 (2007) 348–359
     
To derive the voltage dynamics, we use the balanced r_ x fx ðx; T m Þ bx ðxÞ
¼ þ u
equation for the external net (9). This equation can be rep- r_ v f v ðx; T m ; V_ ref Þ bv ðxÞ
resented as
      There is only one input for the control, that is, the excite-
Vd hd ðxÞ bd ment of the synchronous machine. Therefore, to achieve
¼ þ z; z ¼ bf u ð30Þ
Vq hq ðxÞ 0 two objectives, a combined discontinuous control is chosen
where of the form
 
LL u ¼ U m signðrÞ ð35Þ
hd ¼ ða73 x3 þ a74 x2 x4 þ a75 x5 þ a76 x2 x6 Þ  
L11 rx for jz2 j > e e1 for jrv j > e3
    r¼ and e¼
RL Le þ LL a77 Le xs LL Le þ LL x2 a78 Le rv for jz2 j 6 e e2 for jrv j 6 e3
þ x7 þ x8
Le L11 Le L11 ð36Þ
 
LL Re  RL Le LL V 1
d
þ x9 þ ; with e1 > 0, e2 > 0, e3 > 0 and e2 < e1.
Le L11 Le L11
  It is assumed that the mechanical dynamics are slower
LL than the electric ones and they take advantage the
hq ¼ ða87 x2 x3 þ a84 x4 þ a83 x2 x5 þ a86 x6 Þ
L22 resources of the control, first to stabilize the angle x1 and
   
xLL Le þ LL a85 Le x2 RL Le þ LL a88 Le speed x2 and then to regulate the generator voltage Vg. For
þ x7 þ x8
Le L22 Le L22 U m > max juaeq ðx; T m Þj; a ¼ x; v ð37Þ
  a
RL Le  LL Re LL V 1
q
þ x10 þ ; the state vector first converges to the manifold rx = 0 in a
Le L22 Le L22
  finite time and then the sliding mode motion, described by
1
L11 ¼ 1  LL a71  ; the linear system (25), ultimately confines to the vicinity
Le jz2j < e1 (Fig. 2). Thereafter, the control algorithm regu-
 
1 L L b7 bf lates the voltage Vg. After the voltage error convergences
L22 ¼ 1  LL a81  and bd ¼ :
Le L11 so that jrvj 6 e3 is satisfied, the action of the controller de-
creases to the narrow part of the layer for the speed error
It is considered that bd  0 since bd is very small. Then, z2, from e1 to e2 (e1 > e2). When there is jz2j 6 e2, the con-
the pattern of the voltage Vg can be obtained. From (29), trol (32) ensures the finite time convergence of rv to zero.
(30), (16a)–(16e) and (16f) we have The vicinity jz2j 6 e2 can be made arbitrarily small by
dV g choosing e2 small enough. In the limit, as e2 ! 0 we have
¼ fv ðx; T m Þ þ bv ðxÞu ð31Þ
dt jz2j ! 0.
where Note, due to the synchronism property of the machine,
  in steady state x(1) = xss we have
LL V d a73 V q a87 x2 
bv ¼ þ b3 uxeq ðxss ; T m Þ ¼ uveq ðxss ; T m Þ
V g L11 L22
 
LL V d ðRL Le þ LL a77 Le Þ V q x2 ðLL Le þ LL a85 Le Þ that means, one excitation control input achieves two con-
þ þ b7 bf ;
Vg Le L11 Le L22 trol objectives.

ohq
fv ¼ V1g V d ohox
d
þ V q ox
x_ and bv (t) is the positive function
for t P 0.
With the purpose of regulating the voltage Vg, a discon-
tinuous control is defined as
u ¼ U m signðrv Þ; rv ¼ V g  V ref ð32Þ
where Vref is a reference signal. Then
r_ v ¼ f v ðx; T m Þ þ bv ðxÞu ð33Þ
 _
where f v ¼ fv ðx; T m Þ  V ref . Similar to the speed controller
case, it can be shown that for
U m > juveq ðx; T m Þj; uveq ¼ b1 
v ðxÞf v ðx; T m Þ ð34Þ
the control error rv(t) vanishes at one finite time.

3.3. Speed and voltage combined controller

Using (19) and (33) the equation of the motion projection


of the system on the subspace rx and rv can be written as Fig. 2. Voltage error rv and speed error z2.
J. Cabrera-Vázquez et al. / Electrical Power and Energy Systems 29 (2007) 348–359 353

4. Real sliding mode with solutions close to those of the ideal system. An asymp-
totic observer serves as a bypass for the high-frequency
The key feature of the designed sliding controller is its component; therefore the unmodeled dynamics are not ex-
robustness to uncertainties. During the reaching time per- cited. Preservation of sliding stage modes in systems with
iod, the task of forcing the trajectories toward the sliding asymptotic observers enabled successful applications of
manifold rx = 0 and rv = 0 and maintaining them, is discontinuous control.
achieved by the control (35), provided the plant parameters
variations satisfy the inequality (37). However, the pro- 4.2. Observer- based solution
posed controller provides the desired performance of the
closed-loop system only in the ideal case, i.e. in the absence To prevent chattering and estimate unmeasured excita-
of unmodeled dynamics (13d), sf = 0. tion flux x3, rotor flux x4 = (x4,x5,x6)T, and mechanical tor-
que Tm, a nonlinear observer is proposed as
4.1. Chattering problem
x4 ; x5 ðtÞÞ þ d 2 Tb m þ l1 ðx2  ^x2 Þ
^x_ 2 ¼ f^ 2 ð^ ð41aÞ
In the case sf50, the switching control leads to finite fre- _
Tb m ¼ l2 ðx2  ^x2 Þ ð41bÞ
quency oscillations in the state vector, ‘‘chattering’’ [1,31],
since discontinuities in the control excite the unmodeled ^x_ 3 ¼ f^ 3 ð^
x4 ; x5 ðtÞÞ þ b3 u ð41cÞ
dynamics (13d). In this case we have ^_ 4 ¼ A44 x
x ^4 þ A45 x5 ðtÞ ð41dÞ
r_ x ¼ fx ðx; T m Þ þ bx ðxÞz
where ^x3 ; x^4 ¼ ð^x4 ; ^x5 ; ^x6 ÞT and Tb m are the estimated vari-
1 bf ables, l1 and l2 are observer gains, f^ 2 ð^ x4 ; x5 ðtÞÞ ¼
z_ ¼  z þ u ð38Þ
sf sf a23 x8 ðtÞ^x3 þ a24 x7 ðtÞ^x4  a25 x8 ðtÞ^x5 þ a26 x7 ðtÞ^x6 þ a28 x7 ðtÞ
where bx ¼ b3  a23b7x8 ða24 x4 þ a26 x6  a28 x8 Þ and the deriva- x8 ðtÞ and f^ 3 ð^
x4 ; x5 ðtÞÞ ¼ a33^x3 þ a35^x5  a37 x7 ðtÞ.
tive V_ x (22) depends on the variable z only, but not on Considering the stator current x5(t) = is(t) in (41a)–
the control input u: (41c) and (41d) as a known (measured) function of time,
the dynamics of the estimation errors can be obtained by
V_ x ¼ fx ðx; T m Þrx þ bx ðxÞzrx : ð39Þ subtracting the Eqs. (16b)–(16d), (6), (41a)–(41c) and
In order to examine the system behavior, assume steady- (41d) as the following linear system with time varying
state conditions with parameters:
    
zðtÞ ¼ bf uðtÞ ¼ bf U max : e_ 1 A11 A12 ðtÞ e1
¼ ð42Þ
The step response of the exciter system (38) for the first e_ 2 0 A22 e2
switch at ts from u(t) = Umax to u(t) = Umax at rx = 0 is
where e1 = (e2, em)T, e2 = (e3, e4, e5, e6)T, ei ¼ xi  ^xi ,
given by
i = 2, . . ., 6, em ¼ T m  Tb m ,

tts  
zðtÞ ¼ bf U max 1  2e sf : ð40Þ l1 a29
A11 ¼ ;
l2 0
For some initial time interval Dt = t  ts variable z(t) < bf  
u(t) = bfUmax and V_ x > 0 in (39) results for the case a23 ðtÞ a24 ðtÞ a25 ðtÞ a26 ðtÞ
A12 ðtÞ ¼ ;
fx(x,Tm)rx > 0. It is only after the decay of the exponential 0 0 0 0
term in (40), i.e. after some delay period D t(sf), that 2 3
a33 0 a35 0
j zðtÞ j>j b1
x ðxðtÞÞfx ðxðtÞ; T m Þ j is established once more 60 a44 0 a46 7
and V_ x < 0 indicates convergence to sliding manifold 6 7
A22 ¼ 6 7;
4 a53 0 a55 0 5
rx = 0. During this delay period the trajectory has a devi-
ation from ideal motion. Similar derivations hold for the 0 a64 0 a66
next switch from u(t) = Umax to u(t) = Umax. Repetition a23 ðtÞ ¼ a23 x8 ðtÞ; a24 ðtÞ ¼ a24 x7 ðtÞ
of this process creates the ‘‘zig-zag’’ motion or chattering. and a25 ðtÞ ¼ a25 x8 ðtÞ;
Chattering results in low control accuracy and high heat
losses in electrical power circuits. This phenomenon has a26 ðtÞ ¼ a26 x7 ðtÞ:
been considered as serious obstacles for applications of
It is easy to see that the spectrum of the system (42)
sliding mode control in many papers and discussions. A re-
consists of
cent study and practical experience showed that chattering
caused by unmodeled dynamics may be eliminated in sys-
(1) the eigenvalues of the block A11, which can be
tems with asymptotic observers. This idea was proposed
assigned by appropriatechoice
 of the observer

gains
by Bondarev et al. in [31]. In spite of the presence of 0 dm
unmodeled dynamics, ideal sliding mode is possible. It is l1 and l2, since the pair ; ½ 1 0  is obser-
0 0
described by a singularly perturbed differential equation vable, and
354 J. Cabrera-Vázquez et al. / Electrical Power and Energy Systems 29 (2007) 348–359

Table 1
Calculated mathematical model parameters and their variations
Nominal Increment Nominal Increment Nominal Increment
a23 0.48514 – a48 0.322 – a74 0.26017 –
a24 0.13789 – a53 30.321 – a76 1.5008 –
a25 0.46667 – a55 33.333 – a78 1.0377 –
a26 0.800 – a57 5.000 – a81 685.44 –
a28 0.020 – a64 9.849 – a84 13.630 –
am 0.00003763 – a66 14.286 – a86 20.133 –
a33 0.71222 +0.11745 a68 7.143 – a88 46.799 –
a35 0.64556 +0.12911 a71 711.30 – a83 0.84848 –
a37 0.11067 +0.02213 a73 26.046 0.521 a85 0.96364 –
a44 2.776 – a75 28.759 0.118 a87 0.88207 –
a46 2.579 – a77 42.203 0.020 b7 345.08 –

(2) the eigenvalues of the matrix A22: ^4 ; x5 ; Tb m Þj;


U m > j^uxeq ðx1 ; x2 ; ^x3 ; x
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1 ^uxeq ¼ ð^bx ð^ 1
^4 ; x5 ; Tb m Þ
x4 ; x5 ÞÞ f^ x ðx1 ; x2 ; ^x3 ; x
k3;4 ¼  ða33 þ a55 Þ  ða33  a55 Þ2 þ 4a35 a53 ;
2 2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1 2
sliding mode is established after finite time, and r ^x ¼ 0
k5;6 ¼  ða44 þ a66 Þ  ða44  a66 Þ þ 4a46 a64 ð43Þ holds exactly thereafter. Due to the observer (41a)–(41c)
2 2
^x ¼ 0 will
and (41d) stability, the sliding mode motion on r
which are real and negative. coincide after some transient time period in the observer,
with the motion on rx = 0, which is described by the linear
The parameters of A12(t) and its derivatives are reduced-order system (25).
bounded. Therefore, the linear system (42) with time vary-
ing parameters is asymptotically stable. The resulting esti-
mates ^xi , i = 2, . . . ,6 and Tb m are employed in the control 2
X1 [rad.]
law (20), in particularly forming the estimate of switching 1.8

function rx (17) as 1.6

1 1.4
^x ¼ ^x3 
r ½f2 ð^x4 ; x5 Þ þ d 2 Tb m  k z ðx2  xs Þ :
b2 ðx5 Þ 1.2

¼u ^4 ; x5 ; Tb m Þ
^ 3 ðx2 ; ^x3 ; x ð44Þ 1

0.8
that allows definition of an ideal sliding mode controller
0.6
for the observer loop as
0.4
rx Þ
u ¼ U m signð^ ð45Þ
0.2
Time [sec.]
Then 0
0 2 4 6 8 10 12 14 16 18 20

^_ x ¼ f^ x ðx1 ; x2 ; ^x3 ; x
r ^4 ; x5 ; T^ m Þ þ ^
bx ð^
x4 ; x5 Þu ð46Þ Fig. 3a. Angle x1.

where f^ x ¼ o^ u3
ox2
ðf2 þ b 2x3 þ d 2 T m Þ þ o^ u3 ^
o^x3
f 3 þ o^
u3
o^
x4
ðA44 x
^4 

o^
u3
A45 x5 Þ þ ox5 f 5 þ o^u3 ^  b7
l2 ðx2  ^x2 Þ,bx ¼ b3  a23 x8 ða24^x4 þ a26^x6  2
ob
Vg [p.u.]
Tm
^
a28 x8 Þ and bx ðtÞ it is a positive function of the time. 1.8

1.6
Stability of the auxiliary observer loop is examined via a
similar Lyapunov function as in (21) 1.4

1.2
1 2
Vb x ¼ r
^ ð47Þ
2 x 1

0.8
Substitution of (46) under control (45) into time derivative
of (47) results in 0.6

_ 0.4
Vb x ¼ f^ x ðx1 ; x2 ; ^x3 ; x
^4 ; x5 ; Tb m Þ^
rx  U m ^
bx ð^
x4 ; x5 Þsignð^
rx Þ^
rx
0.2
rx j½U m ^
P j^ bx ð^ ^4 ; x5 ; Tb m Þ
x4 ; x5 Þ  f^ x ðx1 ; x2 ; ^x3 ; x 0
Time [sec.]

0 2 4 6 8 10 12 14 16 18 20

Hence, under similar assumption as in (23) Fig. 3b. Terminal voltage Vg.
J. Cabrera-Vázquez et al. / Electrical Power and Energy Systems 29 (2007) 348–359 355

385 -0.45
X2, X2est, (p.u.)
X4, X4est [p.u.]

-0.5

-0.55
380

-0.6

-0.65
375

-0.7

-0.75
Time, (sec.)
370 Time, [sec.]
0 2 4 6 8 10 12 14 16 18 20
-0.8
0 2 4 6 8 10 12 14 16 18 20
Fig. 3c. Speed x2 and its estimate ^x2 .
Fig. 3f. q-axis(1) damper winding flux x4 and its estimate ^x4 .

1.5 1
Tm, Tmest (p.u.) X5, X5est [p.u.]

0.95

Tm 0.9
1
0.85

0.8

Tmest
0.5 0.75

0.7

Time, (sec.) 0.65


0
0 2 4 6 8 10 12 14 16 18 20 Time, [sec.]
0.6
0 2 4 6 8 10 12 14 16 18 20
Fig. 3d. Mechanical torque Tm and its estimate Tb m .
Fig. 3g. d-axis damper winding flux x5 and its estimate ^x5 .

1.4
X3, X3est [p.u.]
1.3 -0.45
X6, X6est [p.u.]

1.2 -0.5

1.1
-0.55

1
-0.6
0.9
-0.65
0.8

0.7 -0.7

0.6 -0.75
Time, [sec.] Time, [sec.]
0.5
0 2 4 6 8 10 12 14 16 18 20 -0.8
0 2 4 6 8 10 12 14 16 18 20
Fig. 3e. Excitation flux x3 and its estimate ^x3 .
Fig. 3h. q-axis(2) damper winding flux x6 and its estimate ^x6 .

5. Simulation results
The parameters of the synchronous machine, transmis-
The performance of the proposed sliding mode obser- sion and exciter systems and load, all in p.u., except where
ver-based controller was tested on the complete power sys- indicated, are Rs = 0.003, Rf = 0.021, Rg = 0.725,
tem (see Fig. 1). Rkd = 10.714, Rkq = 8.929, Re = 0.05, Ld = 1.81,
356 J. Cabrera-Vázquez et al. / Electrical Power and Energy Systems 29 (2007) 348–359

2 1
X1, [rad.] Tm, Tmest (p.u.)
1.8 0.9

1.6
0.8
1.4 -20% Lmd
0.7
1.2
Tm
0.6
1
0.5
0.8
Tmest
0.6 0.4

0.4 0.3
- 20 % Lmd
0.2
0.2
Time, [sec]
0
0 5 10 15 20 25 30 35 40 0.1
Time, (sec.)
Fig. 4a. Angle x1. 0
0 5 10 15 20 25 30 35 40

Fig. 4d. Mechanical torque Tm and its estimate Tb m .

1.2
Vg, [p.u.]
1.4
X3, X3est, [p.u.]
1
1.3

0.8 1.2
-20% Lmd

0.6 1.1

1
0.4
0.9
0.2 -20% Lmd
0.8

0
0.7
Time, [sec]
-0.2 0.6
0 5 10 15 20 25 30 35 40
Time, [sec]
0.5
Fig. 4b. Terminal voltage Vg. 0 5 10 15 20 25 30 35 40

Fig. 4e. Excitation flux x3 and its estimate ^x3 .

378 -0.45
X2, X2est [p.u.]
X4, X4est, [p.u.]

377.5 -0.5

377 -0.55

376.5 -0.6
-20% Lmd

376 -0.65
-20% Lmd

375.5 -0.7

375 -0.75

Time, [sec] Time, [sec]


374.5 -0.8
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40

Fig. 4c. Speed x2 and its estimate ^x2 . Fig. 4f. q-axis(1) damper winding flux x4 and its estimate ^x4 .

Lq = 1.76, Lkd = 1.831, Lkq = 1.735, Lmd = 1.66, For these parameters we obtain the parameters of math-
Lmq = 1.61, Le = 0.3, H = 3.525 s, bf = 1, sf = 0.015 s, ematical model (13a)–(13f) and (13g) presented in Table 1,
dref = 1.3314, xs = 377 rad s1, Tm = 0.9463 and V1 = 1. considered as nominal parameters. After Lmd experienced
J. Cabrera-Vázquez et al. / Electrical Power and Energy Systems 29 (2007) 348–359 357

1 -0.45
X5, X5est, [p.u.] X6, X6est, [p.u.]

0.95
-0.5

0.9
-0.55

0.85
-0.6
0.8
-20% Lmd
-0.65
0.75 -20% Lmd

-0.7
0.7

0.65 -0.75

Time, [sec.] Time, [sec.]


0.6 -0.8
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40

Fig. 4g. d-axis damper winding flux x5 and its estimate ^x5 . Fig. 4h. q-axis(2) damper winding flux x6 and its estimate ^x6 .

an increment of 20%, some parameters of model (13a)–(13f) same Table 1. The controller gain was adjusted to kz = 7
and (13g), namely, a33, a35, a37, a73, a75 and a77 are modified and the observer gains were chosen as l1 = 200 and
and their corresponding increments are presented in the l2 = 187, resulting in the eigenvalues k1 = k2 = 100. The

Fig. 5. Comparison of regulators.


358 J. Cabrera-Vázquez et al. / Electrical Power and Energy Systems 29 (2007) 348–359

remaining observer eigenvalues were calculated using (43) as [2] Ghandakly A, Idowu P. Design of a model reference adaptive
k3 = 0.123, k4 = 33.922, k5 = 0.883 and k6 = 16.179. stabilizer for the exciter and governor loops of power generators.
IEEE Trans Power Syst 1990;5:887–93.
The mechanical torque disturbance is presented in the [3] Mao C, Malik O, Hope G, Fan J. An adaptive generator excitation
second 10. Also a short circuit in the 30th s with a release controller based on linear control. IEEE Trans Energy Conver
time of 150 ms. Figs. 3a–3g and 3h show the behavior of 1990;5:673–8.
the variables of states when a short circuit is caused in [4] Ghandakly A, Dai J. An adaptive synchronous generator stabilizer
the 6th s with a release time of 150 ms. The system SMIB design by generalized multivariable pole shifting (GMPS) technique.
IEEE Trans Power Syst 1992;7:1239–44.
undergoes a step change in the torque at the 10th s. The [5] Pogromsky AY, Fradkov AL, Hill DJ. Passivity based damping of
nominal value 0.902 p.u. goes down 0.5 p.u. The dynamics power system oscillations. In: Proceedings of the IEEE CDC, Kobe,
related to the weight of the coils have not been modeled for Japon; 1996. p. 512–7.
this system. Figs. 4a–4g and 4h show the behavior of the [6] Ortega R, Stankovic A, Stefanov P. A passivation approach to power
state variables, and also, besides the two previous perturba- systems stabilization. In: Proceedings of the IFAC NOLCOS,
Enshede, Netherlands; 1998. p. 320–5.
tions, an internal perturbation in introduced with Lmd [7] Catherine FM. Robust output feedback controllers for power system
decreasing 50% from the nominal value of 1.66 p.u. This stabilization. Thesis Doctor of Philosophy in Electrical and Computer
variation takes place in the interval from 20 to 30 s. The Engineering, Georgia Institute of Technology; 2000.
system has no dynamics modeled for the one reeled of [8] Xi Z, Cheng D. Passivity based stabilization and H1 control of
the excitatory one. hamiltonian systems with dissipation and its applications power
systems. Int J Control 2000;73(18):1686–91.
Fig. 5 shows the comparison that became between a sys- [9] Galaz M, Ortega R, Bazanella A, Stankovic A. An energy-shaping
tem of conventional control (line dashed) with respect to approach to excitation control of synchronous generator. In:
the proposed control (line solid). Proceedings of the American Control Conference, Arlington, VA;
June 2001. p. 817–2.
6. Conclusions [10] Shi J, Herron L, Kalam A. A fuzzy logic controller applied to power
system stabilizer for a synchronous machine power system. In:
Proceedings of the IEEE region 10th conference Tencon 92; 1992. p.
A nonlinear controller based on the combination of the 346–50.
sliding mode control, block control linearization and singu- [11] Lim C, Hiyama T. Comparison study between a fuzzy logic stabilizer
lar perturbations techniques has been proposed. A model, and self tuning stabilizer. Comp Ind 1993;21:199–215.
which takes into account all of the interactions between [12] Dash P, Mishra S, Liew A. Design of a fuzzy PI controller for power
system applications. J Intell Fuzzy Syst 1995;3:155–63.
the electrical and mechanical dynamics and load con- [13] Hoang P, Tomsovic K, Design and analysis of an adaptive fuzzy
straints, has been described. A nonlinear observer for the power system stabilizer. In: Proceedings of the IEEE PES Winter
estimation of the excitation and rotor fluxes and the Meeting; 1996. p. 7.
mechanical torque, has been designed. This new controller [14] Hernández Romero N, Romero D.R. Control de excitación tipo
has been tested through simulation under three very impor- ANFIS para un Generador Sı´ncrono’’. RVP-AI/02-GEN-II. Reunión
de Verano, Acapulco; 2002.
tant perturbations in the power systems: [15] Hsu Y, Chen H. Tuning of power system stabilizers using an artificial
neural network. IEEE Trans Energy Conv 1991;6:612–9.
• variation of the mechanical torque; [16] Bazanella AS, Silva AS, Kokotovic P. Lyapunov design of excitation
• a large fault (a 150 ms short circuit); control for synchronous machine. In: Proceedings of the 36th IEEE
• variation of parameters. CDC, San Diego, California, USA; 1997. p. 211–6.
[17] Machowski J, Robak S, Bialek JW, Bumby JR, Abi-Samra N.
Decentralized stability-enhancing control of synchronous generator.
The results of the simulation show that with the pro- IEEE Trans Power Syst 2000;15(4):1336–45.
posed observer-based sliding mode controller, we can elim- [18] Chapman JW, Ilic MD, King CA, Kaufman H. Stabilizing a
inate many of the problems encountered with other types multimachine power system via decentralized feedback linearizing
of controllers, such as eliminating oscillations and high excitation control. IEEE Trans. Power Syst 1993;8(3):830–9.
[19] Wang Y, Hill DJ, Middleton RH, Gao L. Transient stability,
variations after internal and external perturbations to the enhancement and voltage regulation of power systems. IEEE Trans.
system. This controller takes into account all of the electri- Power Syst 1993;8(2):620–7.
cal and mechanical dynamics. This controller is much more [20] Mielczarski W, Zajaczowski AM. Multivariable nonlinear controller
robust, simple to design, and uses much less CPU power for synchronous generator. Optim Contr Appl Meth 1994;15:49–65.
than other alternatives. [21] Lahdhiri T, Alouani AT. On the robust control of synchronous
generator. In: Proceedings of the ACC. Philadelphia, Pennsylvania;
June 1998. p. 3798–01.
Acknowledgements [22] Akhkrif O, Okou Fr, Dessaint L, Champagne R. Application of
multivariable feedback linearization scheme for rotor angle stability
The authors thank the support of CONACYT Mexico, and voltage regulation of power system. IEEE Trans Power Syst
Grant 46069Y. 1999;14(2):620–8.
[23] Ramos R, Bretas NG, Costa LF. Feedback linearization improving
an LMI-based design: application power system. In: Proceedings of
References the 15th triennial word congress, Barcelona, Spain; 2002 IFAC.
[24] Schleif FR, Hunkins HD, Martin GE, Hattan EE. IEEE excitation
[1] Utkin VI, Guldner J, Shi J. Sliding mode control in electromechanical control to improve power line stability. IEEE Trans Power Appl Syst
system. London, UK: Taylor and Francis; 1999. 1968.
J. Cabrera-Vázquez et al. / Electrical Power and Energy Systems 29 (2007) 348–359 359

[25] de Mello FP, Concordia C. Concepts of synchronous machine [29] Son KM, K Park J. On the robust LQR control of TCSC for damping
stability as affected by excitation control. IEEE Trans Power Appl power system oscillations. IEEE Trans Power Syst 2000;15(4):
Syst. 1969:316–29. 1306–12.
[26] Crenshaw ML, Miller WJ, Schulz PRP, Temoshok M. Althyrex [30] Loukianov AG. Nonlinear block control with sliding mode. Automat
excitation system with power stabilizer. In: IEEE Paper 70 CP 563- Rem Contr 1998;59(7):916–33.
PWR; 1970. [31] Bondarev AG, Bondarev SA, Kostylyeva NYe, Utkin VI. Sliding
[27] El-Sherbiny MK, Mehta DM. Dynamic system stability: part I – modes in systems with asymptotic state observers. Automat Rem
investigation of the effect of different loading and excitation systems. Contr 1985;46:679–84.
IEEE Trans Power Appl Syst 1973. [32] Anderson PM, Fouad A. Power system control and stability. New
[28] Leirbukt AB, Chow JH, Sanchez-Gazca JJ, Larsen EV. Damping York: IEEE Press; 1994.
control design based on time-domain identified models. IEEE [33] Khalil HK. Nonlinear systems. New Jersey: Prentice Hall; 1996.
Transaction on Power Systems, Vol. 5 pp. 172-178, 193.11.

You might also like