BULLETIN OF THE POLISH ACADEMY OF SCIENCES

TECHNICAL SCIENCES, Vol. 65, No. 1, 2017

DOI: 10.1515/bpasts-2017-0001

Nonlinear robust control for single-machine infinite-bus

power systems with input saturation
Y. WAN*
College of Automation Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China

Abstract. In this paper, a new control scheme is proposed to achieve stability for a single-machine infinite-bus power system. A power system
model simultaneously considering input saturation and time-varying uncertainties is presented. A sufficient condition for the system convergence
is given and based on this result, a switching excitation control law with auxiliary system is designed. The stability analysis and simulation
results all show that the developed controller is effective.

Key words: nonlinear robust control, input saturation, generator excitation, single-machine infinite-bus power systems (SMIB).

1. Introduction tion to the control problem of nonlinear systems with input

saturation. In [9], an auxiliary system is employed to address
Energy issue is one of the most pressing challenges. The main input constraints, but only boundedness of the “last modified
contents of this problem include generation, transmission and virtual error” is obtained. If the desired value of control input
distribution of energy. Power system is regarded as one of the greatly exceeds the constraints, then the auxiliary system will
most complex systems in the world [1] and in order to obtain be unstable and thus the convergence of the “last virtual error”
satisfactory and optimal operating state, many researchers to the origin cannot be ensured. In [10], the tanh function and
have developed various solutions from their different profes- the mean-value theorem are used to deal with the saturation
sional perspectives [2‒5]. This paper studies power system function, and then the redial basis function i.e. neural network is
transient stability by using model-based control systems tech- employed to approximate the unknown control gain. However,
nology [6]. since the tanh function is also a kind of bounded function, when
For many years, linear control has been used as a traditional there are large disturbances and the desired value of control
soft computing method to design excitation controllers, e.g. input goes beyond the range of tanh, the efficacy of the designed
automatic voltage regulator (AVR) and power system stabilizer control law may be limited. In [11], dynamic surface control
(PSS). Although these controllers have been simulatively vali- is used to equivalently transform a class of nonlinear system
dated, the robustness can not be ensured because the controllers into a linear system, and then the region of attraction which is
have been designed on the basis of linear power system model enlarged to allow some degree of input saturation is estimated
which is often not applicable in entire operation region. Thus via a linear matrix inequality method [12]. But this approach
it is necessary to preserve nonlinear characteristics of power does not give a solution for a case when the initial condition is
systems and study nonlinear control method. However, there out of this region of attraction.
are strongly nonlinear couplings between model parameters This article proposes a new nonlinear robust control frame-
and state variables, and the known values of the parameters work to tackle the stabilization problem for a class of nonlinear
are often not coincident with the actual situation. Thus, model system with time-varying uncertainties and input saturation.
errors are always subsistent and are necessary to be considered Compared with the prior works, the proposed method ensures
in control system synthesis from the viewpoints of nonlinear not only the boundedness of state variables but the global sta-
robust control [7, 8]. Furthermore, input constraints, such as bility of the closed-loop system. Furthermore, the approach pro-
excitation voltage saturation, are unavoidable in practical en- posed processes the saturation phenomenon directly and does
gineering. If this phenomenon is not considered in controller not require approximating the saturation function by a smooth
design, the prospective stability and performances of closed- function, which is often not invertible outside its range.
loop systems may not be ensured because the system dynamic The paper is organized as follows: the preliminaries and
behavior is uncontrolled when the desired control input goes a SMIB power system model are described in Section 2. In
beyond the limit. Section 3, a sufficient condition for the convergence of states
Due to the complex nature of nonlinear systems, in the ex- to the origin under saturation is shown in a Theorem, and then
isting literatures, most of results on input saturation/constraints on the basis of this result, a switching excitation control law
are based on linear systems. Only few researchers pay atten- with auxiliary system (SAEC) is designed, as well as stability
analysis for the closed-loop system with the designed controller.
The simulation results are presented in Section 4. Conclusions
*e-mail: [email protected] are drawn in Section 5.

Bull. Pol. Ac.: Tech. 65(1) 2017 3

Brought to you by | Gdansk University of Technology
Authenticated
Download Date | 3/2/17 11:05 AM
 x(0)
γ=
1, |E f d | < UB ,
Proof:
where δ is the power angle in rad, ω is the relative speed in can direc
rad/s, Pe is the active power in p.u., E f d is the field excitation
Y. Wan
voltage in p.u., UB > 0 is a bound of the field excitation voltage, LE
di (t) are bounded time-varying uncertainties, that is, there ex- inequality
istcoefficient
unknown constants ∗ ∗
in p.u., Hdis the0 such
i > inertia that inds,i ,ωi =
|di (t)| ≤
constant 1, 2; Pm
0 is the syn-
ischronous
the mechanical power in p.u., D is
0 the damping coefficient
speed in rad/s, Td0 and Td0 are the time constants of
in p.u., H is the inertia constant in s, ω0 is the synchronous
the excitation winding in s, xd is the d-axis synchronous reac- 3.2. SAE
speed in rad/s, T 0 and Td0 are the time constants of the ex-
tance in p.u., xd d0
is the d-axis transient reactance of the generator sired ope
citation winding in s, xd is the d-axis synchronous reactance
in p.u., xT and xL are the transformer and the line reactances in spectivel
in p.u., xd is the d-axis transient reactance of the 0 generator
0 in
p.u., Vs is the infinite bus voltage in p.u., xds = x d + xT + xL, tions of d
p.u., xT and xL 0 are 2the transformer
02 0 and the line0 reactances in
cxd = (xd ¡ x d)Vs /(2Td0 xds), c xd = Vs/(T d0 xds),  P1(δ, ω) =
p.u., Vs0 is the infinite bus voltage in p.u., xds =0 xd + xT + xL , erating at
–1/Td0 + ωcotδ, P 2 (δ) = c xd sin2δ, P3 (δ) = c xd sinδ.
   
cxd = (xd − xd )Vs 2 /(2Td0 xds2 ), cxd = Vs /(Td0 xds ), P1 (δ , ω) = hold, E ∗f d
Excitation 
−1/Td0 + ω cot δ , P2 (δ ) = cxd sin 2δ , P3 (δ ) = cxd sin δ .

Step 1
Fig. 1. Single-machine infinite-bus power systems 3. Preliminaries and controller design V 1 = (c1
3. Preliminaries and Controller Design tories is V
Beforepresenting presentingthethedesigned designed controller, some preliminaries tive cons
Before controller, some preliminaries z2 = ω −
and the preprocessing
and the preprocessing of control are given below. of control are given below.
2. Problem statement and system model Step 2
model [13, 14], the SMIB model does not directly reflect the is equivalent to Q(x) − S(x)
S(x)R(x) R(x) −1S(x) T > 0, where x ∈ Rn ,
model [13, 14], the SMIB model does not directly reflect the3.1. 3.1.isPreliminaries.
equivalentTto Generally,
Q(x) − S(x) power
S(x)R(x) R(x) systems −1 S(x) are under
T > 0, normal where x∈R
then
n,
V̇2 =
influences of power network components and loads, instead Preliminaries
Q(x) Q(x) Generally,
∈ R m×m , power
0 R(x) systems R(x) areTunder ∈ R nor-
s×s , and S(x) ∈
influences Considerofthepower SMIBnetwork system with components its components and loads, as shown instead in operating Q(x) == conditions
Q(x) T ∈rather m×m
Rrather than ,than0leading
<
<leading R(x)power =
=power R(x) factor T ∈operations s×s , and S(x) ∈
Roper-
mal operating m×s conditions factor c z d
of aggregated
of aggregated
Fig. 1. Compared node containing
node containing
with structure-preserving the information
the information of these
multi-machine
of these ele- ele- which Rm×s aredepend depend
the rare affinely affinely
cases [3,on 19],xx ..thus there exists a small con- 2 2 1 (t)
ationsRwhich are the rare caseson [3, 19], thus there exists a small (H/ω )[k
ments. modelFurthermore, [13, 14], the SMIB the model does notmodel
reduced-order directlyalso reflect the stant δDefinition
neglects m such that 1: π/2The > δset > C(P δm P>, ρ) 0. = {xx | x TT Px ≤ ρ} is said to 0be
be
ments. Furthermore, the reduced-order model also neglectsconstantDefinition δm such that π/2
1: The set C(P , ρ) = {x|
> δ > δ Pm > 0 and x P 3 (δx ) P > x 0.
≤ρ} is said to
positive c
the influences
flux linkages of power dynamics network by components
using the and loads,perturbation
singular instead of invariant ellipsoid when the following condition Q(x)condition S(x) holds: For aa
the aggregated
flux linkages dynamics
node containing by using the singular perturbation Schur
Schur invariant complements
complements ellipsoid [12].
[12]: when The the inequality
inequalityfollowing >> 0 is
holds:
0 D/H, For z
eventheso, so,information the of these elements. T 3
technique. Fortunately,
technique. Fortunately, even both the
both theoretical studies trajectory x (t), if x (0) ∈ C(P P,, ρ) ρ) thenS(x) (t)R(x)
x (t) ∈ C(P C(P P, ρ), ∀t ≥ ≥ 0,0,
Furthermore, the reduced-order model also theoretical neglects thestudies flux equivalent trajectory to xQ (x) if
(t), ¡TxS(0) (x)R ∈(x)
n×n
C(P –1P
S (x)T then > 0, xwhere ∈ x 2P,ℝρ), n ∀t
,
and the the experimental
experimental results results have have validated validated the the SMIB SMIB model model where 0 < TP = Pm×m ∈R R and ρ ρ > 0. T
and linkages dynamics by using the singular perturbation technique. 2 Q(x) where = Q0(x) < P2=ℝPT ∈ , 0 n×n < R and (x) =>R0. (x) 2 ℝs×s , and
is
is [13,well-suited
well-suited
Fortunately,
for
for
even
stability
stability
so, both
analysis
analysis
the theoretical
and
and feedback
feedback
studies and
control
control
the
de-
de- S(x) to Definition
m×s 2: The set C x.> is said to bexxaa∈∈region region of attraction
theexper- ℝto depend affinely on
2Definition −1 −1 T T nn
model
modelmodel
model [13,
[13, 14],
[13, 14],
14],14],thethethe SMIB
the SMIB
SMIBSMIB model model
modelmodel does does
does does not not
not directly
directly
directly
not directly reflect reflect
reflect
reflect thethe
the isonisis equivalent
equivalent
equivalent
is equivalent totoQ(x) Q(x)
Q(x) Q(x) −− S(x)R(x)
− − 2:
S(x)R(x)
S(x)R(x)
S(x)R(x) The set−1
S(x) S(x)
−1 S(x) C T
S(x) isT> T> said
0, > 0,
0, towhere
0,where
where where bex ∈ xR∈ nRR , R,,nn , of attraction
sign
model
signimental for
[13,
for industrial
14], results
industrial the SMIB
have power
power model
validated generators
generators does
the SMIB not [15,
directly
[15, model 16].
16].instead are Based
reflect the
well-suited
Based on is the
theequivalent when TT to Q(x)
the following
− S(x)R(x) condition
−1
S(x) T T > 0,
holds: where For x a
∈ R ,
trajectory xx(t),
(t), if if
influences
influences
influences
influences ofof of ofpower
power power
power network
network
network network components
components
components components and and
and loads,
and loads,
loads,
loads, instead
instead
instead Q(x)Q(x)
Q(x)Q(x) = =Q(x)
= =Q(x)
Q(x)Q(x)
T ∈
when ∈ TTR ∈∈ Rm×m
the
R
m×m m×m
Rm×m ,following
m×m 0,,< 00, <0<R(x)
<R(x)
R(x)
R(x) =condition
= = R(x)=R(x)
R(x) T ∈
R(x) ∈ TTR ∈holds:
R Rs×s
s×s
∈
s×s s×s
R,s×sand,,and
and For
, andS(x)
and S(x)
S(x) a∈trajectory
S(x) ∈∈∈
above
influences discussions,
of power from
network the perspective
components and of loads, robust instead control, Q(x) in Q(x) R , 0 R(x)
→P00, ρ) R(x) T R ,
∞.Px ∙ ρg is said to be S(x)
for stability analysis andthe feedback controlofof design for industrial xx (0)
Definition C1.on then
The xset as= tt →
= (0) ∈C ∈ (t)C(
< = → ∈ ∈
ofof of above
aggregated
aggregated
aggregated discussions,
node node
node containing
containing from
containing the the
the perspective
information
information
information ofof ofrobust
these these
these ele-ele-
ele-control,
ele-Rm×s RRm×sm×s in depend
depend affinely
affinely thenon
on x.x. fxjx ∞.
onxx. as
of
of aggregated
aggregated node
node containing
containing the
the information
information of these
these ele- Rm×s m×s
depend depend affinely ∈
affinely x.on x.
(t) →
order power to enhance
generators the applicability
[15, 16]. Based of on excitation
the above control
discussions, system, R depend affinely when the following condition holds:
ments.
ments.
ments. order
ments.
ments.
Furthermore,
Furthermore,
toFurthermore,
Furthermore, enhancethe
Furthermore, the
the
thethe the reduced-order
reduced-order
applicability
reduced-order reduced-order
reduced-order
model
model
of excitation
model model
model
also
also
alsoalso
alsoto
neglects
neglects
control
neglects neglectssystem,
neglects Definition
Definition
invariant
Definition
Definition 1:1: 1:
The 1:The
Theellipsoid
setset
The set
C(P,C(P,
set C(P, ρ)ρ)
C(P, ρ)=ρ) = =
{x= {x{x | {xx||xxPx
T T T
| xxPx Px
T
T≤ Px ≤ ρ}
≤ ≤ρ}ρ} is
ρ} issaid
isis said
said toto
said to
be
nto bebe Fornu ×n
be
uto
from the perspective of robust control, in order enhance the Definition
a trajectory 1: L
The x set ifC(P,x 1. ρ) Consider P , Px
u
then x ρ}
K x is said R P , Kbe ∈ Rnu ×n , and
time-varying uncertainties indicating the model errors should C( ρ) ∈ C( ρ), a awhere
the
the flux
flux linkages
linkages dynamics
dynamics byby using
using the
the singular
singular perturbation
perturbation EMMA
(t), (0) 2 = {x | ≤
= (t) 2 n ,K
the fluxthe
the
linkages
time-varying
flux
flux linkages
linkages
dynamics
uncertainties
dynamics
dynamics
by using by indicating
by using
using
the singularthe singular
the theperturbation
singular model perturbation
perturbation errors invariant invariant
invariant
should invariant ellipsoid
ellipsoid
ellipsoid ellipsoid LTwhen
when when
EMMA when the the
the 1.
following
the Consider
following
following
following condition
condition
condition u =holds:
condition Kholds:xholds:
holds: ∈For RFor
For u
a,K
For aK ∈R , and
technique.
technique.
technique. applicability
be considered.
considered. Fortunately,
Fortunately,
Fortunately, of excitation
Moreover,
even even
even so,so, in
so, control
both practice,
both
both thethe system,
the actuator
theoretical
theoretical
theoretical time-varying saturation
studies
studies
studies uncer- invariant
u- 0 x(t),
istrajectory <C(P PP
ellipsoid
x(t),
x(t), P= ρ)
, if Px(0) 2ℝ
is
x(0)
when
an n×n the
and ρ)
invariant following
ρ >then
ρ) 0.ellipsoid.
x(t)
x(t)
condition Then holds:
ρ),
ρ), u u≥ For
≤ uacis is equivalent
equivalent
technique.
be
technique. Fortunately,
Fortunately, Moreover, even
even so,
in both
practice,
so,errors both should the theoretical
actuator
the theoretical studies
saturation
studiesMore- trajectory
trajectory
trajectory
is u- C(P x(t),
x(t),
if , ρ)if
x(0)if is
ifTx(0) x(0)∈an ∈∈
C(P, ∈C(P,
C(P,
invariant
C(P,ρ) 
ρρ)then
ρ) then ellipsoid.
thenx(t) 
x(t)
∈ ∈ ∈
C(P,
 ∈ C(P, ρ),
C(P,
∈C(P,Then
C(P,ρ), ρ),
∀t ∀t
∀t
u u
≥ ∀t
 ∀t ≥TT 0,c
≥0, ≤≥0,
0, u
0,
and
and thethe
the
biquitous tainties
experimental
experimental indicating
and there resultsthe
results
exist modelhave
have
input validated
validated
constraint. thethethe be
SMIB
SMIB considered.
Simultaneously model
model trajectory
en- ∈ C(P, ρ 0. x (0)TTx(t)
then
and and
biquitous
and
experimental
the
the experimental
and
experimental thereresults results
exist
results
have have
input
have
validated validated
constraint.
validated
SMIB
the
the SMIB model
Simultaneously
SMIB modelwhere
model where
where
where
en-
where
0< 00<
0
Pto
0< <
to
P
=PP
Pthe
=P=
=
the
TP P∈ TT
inequalities
P∈ TR∈∈
inequalities
P
R Rn×n
n×n
R
n×n
Rn×n n×n
and and
and ρ ρ>
and
and
>
ρ
ρ
ρ>0.>0. x (0)0.−1 ≥
0. ≥ 00 and and KPP (u2KK/ρ)I ≥ 0 hold.
I ≥ 0 hold.
is isis issuring over,
well-suited
well-suited
well-suited
well-suited the in
for practice,
for
for
expected
for stability
stability
stability stability actuator
analysisanalysis
performances analysis
analysis saturation
and and
and and feedback
feedback
feedback
and is ubiquitous
the
feedback control
controlcontrol
control
control de-
laws and
de-
de-de- thereDefinition
remain- Definition
Definition
<
2:2:
=
The
The 2. The ∈
set
set CCsetisisCsaid isxsaid
said
x (0)
to
>
to
P
be
P
to
be −1be a region of attraction
a a region
region ofof K (u2cc /ρ)II when
attraction
attraction
issuring
well-suited theinput forconstraint.
expected stability performances analysis and and feedback
the control control laws de-remain- Definition
Definition 2: The
2: The set
The Cset C is said
C is is holds:to
(0)
said to be toFor a region
be aaatrajectory region of of attraction
of xattraction
attraction
sign
sign sign forfor exist
for industrial
industrial
industrial powerpower
power Simultaneously
generators
generators [15,
[15, 16].
16]. ensuring
Based
Based onon the
on onthe expected
the thefollowing
Definition following 2: condition
set said be region (t), if x(0) 2 C 
sign
ing
sign for
within
for industrial
industrial bounds power
power isgenerators generators
difficult,
generators
[15,
and 16].
[15,
[15, in 16].
this
16].
Based Based
case,
Based itthe
on is the
thea when
critical
when
whenwhen thethe
the following
following
the following condition
condition
condition condition holds:
holds:holds:
holds: For For
For a atrajectory
For a trajectory
atrajectory
trajectory x(t),x(t),
x(t),x(t),if ififif  ρ x (0)TT 
above
above ing
above
above
within
performances
discussions,
discussions,
discussions, bounds
discussions, problem from and
from
from is
the
from the the
the thedifficult,
control
perspective
perspective
perspective
the perspective
perspective
and
laws of ofin
remaining
of this
robust
robustrobust
of robust
case, within
control,
control,
robust control control, it
control, ofinis
bounds
in
in a critical
inx(0) x(0)is
when
x(0) then
CC
the
then
then x
following
Proof:
(t) x(t)
x(t) ! 0 asBased
0as
0 as t
condition! � on.
∞.
holds: the For
Schur
∞.∞.the Schur complements, x (0) P−1
a trajectory
complements, x(t), if ρ x (0) ≥
anddifficult,
above
and challenging
discussions,
challenging and in from
this
problem case,
foris astate-feedback
it
for state-feedback
critical of and challenging control,
control in
problem
of power
powerx(0)
∈∈
x(0) ∈∈
C
∈
thenC
C
x(t)
then
then Proof: x(t)
→
x(t) 
→ → 0→
→ 00as
Based t as→
as
tt → t∞.
→
t →
→ on ∞.
−1 ≥
order
orderorder
orderto toto enhance
enhance enhance
to enhance
enhance the the
the applicability
applicability
applicability
the applicability
applicability of of
of excitation
excitation
excitation
ofAs excitation control
controlcontrol
control system,
system,
system, system,  P KT  T x (0) P
systems
order to with input
the saturation. ofof such,
excitation
aa errors
nonlinear
control system, robust ex- LLemma 0EMMA and Kx nK P, ρ) is
00Kx can nube equivalently nnu×n transformed into
systems
time-varying
time-varying
time-varying
time-varyingfor state-feedback
with
uncertainties input
uncertainties
uncertainties
uncertainties control
saturation. indicating
indicating
indicating indicating power
As
the the
the such,
model
the model
modelsystems
model errors nonlinear
errorserrors with
should input
should
should
should robust satu-ex- L LEMMA
EMMA L 0
EMMA and 1. 1.
1. 1.Consider
1.ConsiderPConsider
Consider
K K Iuu
(u22c /ρ)Iu
Consider u=
= ≥
u=
≥ =Kx = Kx 2∈∈
can
Kx
∈ Rℝ∈ nR
R u ,n
be,RuK u,n,K K,2K
∈ K∈∈ ℝ∈
nuuequivalently
R RR nnuu,×n
Rnn,uand
u ×n ×n×n
u×n ,, and
and C(
,and transformed
and into
time-varying
citation ration. control
As uncertainties
such, scheme a in nonlinear indicating
considering robust the model
excitation
excitation errors input
control should constraint L EMMA 1. K Consider I u = Kx ∈ R , ∈ R , and
bebe be considered.
considered.
citation control Moreover,
Moreover, scheme in
in practice,
practice,
considering actuator
actuator excitation saturation
saturation is isis
input u-u-scheme C(P,
C(P, ρ)ρ) ρ) anisis invariant
an
an invariant
invariant ellipsoid.
(u /ρ)I
ellipsoid.
ellipsoid.
c Then
PThenThen kuk T∙ uc≤ isis
is
ucu−1/2 equivalent
equivalent
equivalent /ρ)IIIto the
isconstraint
considered.
be considered. Moreover, Moreover, practice,
in practice, actuator actuator saturation saturation u-is u-C(P, C(P, isρ) anis invariant
an Tinvariant ellipsoid. ρTTand
ellipsoid.   Then Then
−1/2
u
u K≤ ≤uP is
ccu cequivalent
is (u22c /ρ)I
equivalent
beis considered.
proposed
considering in Moreover,
this
excitation paper. in
input practice,constraint actuator is saturation
proposed in this u- paper. C(P, ρ) xxis (0)
an
(0) P
 xx(0)
Tinvariant
P (0)ρρx(0) ≤≤ ellipsoid.
Tρ
and
T P
Then
−1/2 u K
K T≤K
u
Pu T≤
P TTuc
−1/2 is ≤equivalent
≤ (u . Thus,
. Thus, xx (0)
(0) ∈ ∈
biquitous
biquitous
biquitousis proposed
biquitous and and
and there
and there
there
in
thereexist
thisexist
exist input
paper.
exist input
inputinput constraint.
constraint.
constraint. constraint. Simultaneously
Simultaneously
Simultaneously
Simultaneously en- en-
en-en- to
to the
the inequalities
inequalities
inequalities ρ  ρ x(0)
x(0)x(0) 00 and
and P 
P K TK K
000 hold.
hold. c T Px ≤
s notsuringbiquitous
directly and
reflect
Thus, there
on the the exist is of
basis input
equivalent
of the constraint.
classicalto Q(x) Simultaneously
SMIB − S(x)R(x) model
−1
[17, en-
S(x)
18]
toT the
and>to inequalities
to0,the
thewhere C(P
inequalities
C(P P
xP,∈ρ), ρ) Rx(0)n which
,x(0)
which
ρ x(0) is
≥ T≥
is an≥
¸ an
≥
0 and invariant
00 Tand
≥ invariant and KK K2cK P
P 22ellipsoid,
KT ≥¸
K ≥≥0 hold.
≥ 00 hold. one can
hold.can get x T Px ≤ get x
K T(uellipsoid, one
suring
suring Thus,
the the
the on
expected
expected
expected the basis
performances
performances
performances the and classical
and
and the the
the SMIB
control
control laws
laws model remain-
remain- [17, 18] and inequalities x(0) P−1
x(0)
−1
−1
PP P−1 (u (u
(u
/ρ)I c/ρ)I
c(u /ρ)I
2 /ρ)I ≥
suring the expected performances and T control
the control laws lawsremain- remain- 2c/ρ)I
Thus, on the basis of the classical SMIB model [17, 18] T ands×s ρ. Therefore, x(0) P−1 2
u u u x P c 1/2 P −1/2 T −1/2 )P P1/21/2x ≤
suring the expected performances and the control laws remain- u  K KP P −1/2
nentsing and loads, instead Q(x) Q(x) R m×m , 0 R(x) R(x) R , and S(x) = = (P
inging thewithin
within
within
ing the
above
within above
boundsbounds
bounds
bounds statements,
statements,is isis isdifficult,
difficult,
difficult,difficult, the the=
and system
system
and
and in
and in
inthis
in
∈
this
thisdynamic
dynamic
case,
this case,
case,
case,it itmodel
is <
itmodel
isis
it a considered
acritical
isa critical
a
=
considered
critical
critical in in
∈ ρ.2Therefore, ∈ 2
uu2 = u u = x P ρ (P T T 1/2  ρPρ
−1/2
T TTKT 
TK )PP x≤
nformation thewithin
ing above
thisof paper these statements,
bounds ele-
is is difficult,
proposed R m×stheasdepend systemand in
follows: dynamic
affinelythis case, on x. model
it is a critical considered Proof: in
Proof:
Proof:
Proof: (u
Based
Proof. Based
Based /ρ)x
2cBasedBasedon xon
onT
Tthe
onPon xthe
the
the ≤
Schur
the uSchur
Schur
Schur . complements,
2cSchur Thiscomplements,
complements, concludes the
complements,
complements, the x(0) x(0)
x(0)
ρproof.
ρ x(0)
≥ T ≥ 
≥ 0 and
andand
and challenging
challenging
thischallenging
challenging
and paper
challenging is givenproblem
problemproblem
given problem for for
as follows: for
follows: state-feedback
state-feedback
state-feedback
for state-feedback
state-feedback control control
control
control control of of ofof
powerpower
power
of power
powerT Proof:
 is (u
  Based /ρ)x
c TT on x P x ≤
the Schur u c . This concludes
complements,x(0) x(0) x(0)
x(0)x(0)
−1
P P−1P Px(0)
proof.
−1−1−1
 ¸
≥≥
der systems
model this
and
systems paper
alsowith
with is
neglects
input
input
problem as
saturation.
saturation.
for
Definition 1:a The set C(P, ρ) =ex- {x | x Px ≤ ρ} P PPsaid Kto
Definition be 3 [11]: For ς ∈ (0, 1], the P
constraint uuu ≤ ≤ uuc /ς /ς
systems systems with withinput inputsaturation.saturation. AsAs Assuch,
As such,
such,
such, aa nonlinear
anonlinear
nonlinear nonlinear robust robust
robust
robust ex-
ex-ex- 0 00 and and
and
0 and andKholds:
K TK
PDefinition KTT¸≥0 ≥≥can
0 0 0 3
be
can
≥ 00 can [11]:
can
can be
can be be
equivalently be For ς
equivalently
equivalently
equivalently
be equivalently transformed
∈ (0,
equivalently transformed 1], the
transformed
transformed
into
transformed x
constraint
transformed ∙ T
P x(0) u ρ
the singular systems δ̇ with
perturbation ω, input saturation. invariant As such, a when
ellipsoid nonlinear the robust ex-condition
following 0 KK2cK P
(u
(u 22 K
For
c/ρ)I 2 /ρ)Ia ≥
/ρ)I (0) c
citation
citation
citationcitation controlcontrol
control
=
δ̇ control
control ω, scheme
scheme scheme considering
considering
considering
scheme considering excitation
excitation
excitation
considering excitation input inputinput
input constraint
constraint
constraint
input constraint is
(u
is K an
/ρ)I c(u
extended
2c/ρ)I version of u
TTu
u  ≤ u , in
c in 2this this case, some degree
= Tan c extended version of u≤ c ,≤ case, some degree
(u
h the citation
theoretical studies scheme trajectory x(t),excitation if x(0) ∈ C(P,constraint ρ) then x(t) intointo
into ∈ x(0)
into x(0)
x(0)
C(P, and
x(0)ρ),
TTPx(0)
T Px(0)
PPx(0)–1/2
Px(0) K ≤0, T≤≤KPρ≤ρρ–1/2 ρisand
and and and PP2−1/2
P(u−1/2
−1/2
P−1/2 K
−1/2
I.K TK u
KPKKP
Thus, T
KP −1/2 ≤
−1/2
KPx−1/2
−1/2−1/2 ≤(u ≤C( (u
(u(uP
22/ρ)I.
2 /ρ)I.
cc ,/ρ)I.
2cρ) which is
is isproposed
isisproposed
proposed inin in
this this
this paper.
paper.paper. of T∀t ≥
saturation ρ ∙
allowed, c/ρ) namely T KP (0)
“ς 2 c /ρ)I.
-saturation".
lidatedisThus,
proposed
proposed
the ω̇
ω̇
SMIB = in
−(D/H)ω
in
model
this
this paper. paper. + (ω 0 /H)(P m − P e ) + d 1 (t), into
Thus,
Thus, (1)
(1)x(0)
x(0)
x(0)
x(0) of Px(0)
∈saturation
≤
ρ)
ρ)ρ) iswhich and
allowed, P K
namely “ς ≤
-saturation". (u c /ρ)I.
Thus, Thus, on onon =
the the−(D/H)ω
the basisbasis
basis ofof of
the where
the
the+ (ω
classical
classical
classical 00 /H)(P
< SMIB PSMIB=SMIB
SMIB mP−
T ∈PR)n×n
model model
model
emodel + [17, d[17,
[17,and (t),
1[17,
18] ρ18]
18]
18] >and
and 0.
and Thus,
and (1) Thus,
Thus, an x(0)invariant
x(0) ∈∈C(P, ∈C(P,
C(P,C(P,ρ)ellipsoid, which which
ρ)1/2 which which is
one isisan iscan ananan invariant
invariant
invariant
get xT P
invariant x ellipsoid,
ellipsoid,
ellipsoid,
∙ ρ. Therefore,
ellipsoid,
Thus,
Thus, on
on the
the basis
basis of
of the
the classical
classical SMIB model [17, 18] and ∈ TTC(P, is an invariant 2 22=2= Tellipsoid,
TTuu
nd feedback Ṗ control P de- ω) P P P sat(E d one
one can
can 2get
get T xx PxPx x P ρ.
ρ. P Therefore,
Therefore,
K KP P x uu2u/ρ)
uuT x u P
thethe
the
aboveabove
above Ṗ =
e statements,
statements,
statements, P (δ
1 (δ , ω) , the the
the
P +
esystem
system Definition
system P (δ
2 dynamic )
dynamic
dynamic + 2:P (δ
The
3model model )
model set
sat(E C is
considered
considered
considered f d ) +
said dto
in (t),
2in be
in one
a one
regioncan can
of
kuk get = get
attractionx u T Pxu x
L =T
T Px
≤ T≤≤ ρ.≤ ρ.
2.
( –1/2
Therefore,
“ς
T
Therefore,
-saturation"
–1/2u
) u 1/2
u u
is = u
equivalent
∙
2 =
(u T
=u= T
==x ∙ u2.
to thethec inequali-
inequali-
the above e = statements,
1 thee + system 2 (δ ) dynamic
+ 3 (δ model
) considered
f d ) + 2 (t),in one can get xL Px
EMMA ≤ ρ.
2. “ς Therefore,
-saturation" u
is =
equivalent c =to
[15,this
this
the above
16].
this paper
paperpaper Based
sat(Eis isis
statements,
givengivenon
given the
) = γE
as as
as
the
when
follows:
follows:
follows:
system the dynamic
following model
condition considered holds: in x
For TxxP
T
a
T
xPP
1/2T
T
1/2
1/2
P 1/2
(P
1/2
−1/2
(P
This
trajectory
(P
x P (Pties K ρKP
−1/2
(P
−1/2
K KTK
concludes
−1/2 
x(t),
KP
T
K
T KP
KP
T
T
ρ
−1/2
KP
ifEMMA
−1/2
x
−1/2
the
−1/2
)P T )P
1/2
)P

1/2
1/2
x
proof.
)P x
1/2
1/2≤x ≤
x ≤(u □□≤ (u2
(u 2
/ρ)x 2 2
/ρ)x
/ρ)x
cc 2cP/ρ)x
(u T T
Px
T
ςK
Px
K
Px
T
T TPx
≤ ≤≤ 
u ≤2uu
.
22. .2
ccu2c . This This
This This
≥ 00xand K TPx ≤≥u0
−1/2 −1/2 c c
this paper paper isf dgiven
given as follows:
follows: c . hold. This
f d ,, (0) T )P ≤ (ucP/ρ)x
ive ofthis robust sat(E is
control, f d ) =inγE as fx(0)
d ∈ C then x(t) → 0 as t → ∞. concludes
concludes
concludesconcludes the the
the
ties proof.
the proof.
proof.
x (0)
proof. x
P
(0)
 ≥
−1 and ςK
ςK
K (u22c /ρ)II ≥ 0 hold.
 concludes the proof.  ςK
δ̇ δ̇=
δ̇control ω,ω,ω,system, U /|E |, |E | ≥ U , x P −1 K I
xcitation = ω,
= (0)
ς ς∈ ςFor (u /ρ)I
δ̇δ̇ = (2)
Definition
Definition
Definition 3 3[11]:3 [11]:
3[11]: 3For For
For
[11]. 21],
1], the
the 1], constraint
constraint
the constraint u/ς
uccu/ς /ςuc/ϛ is
c
= ω, Definition
Definition [11]: For ∈(0,
∈∈
ςς ∈ ϛ1],
(0,
(0, (0, the1],
(0, constraint
the constraint u u
u ≤≤
u ≤ukuk
c≤ uc∙
/ς
γ U B
B /|E
fd
f dL|,EMMA |E f d | ≥ UB ,
u Kx nu (2) Definition
K nu ×n3 [11]: For (0, 1], the constraint u ≤ c/ς
the model ω̇
ω̇
ω̇ =ω̇−(D/H)ω
= errors
= =−(D/H)ω
−(D/H)ω
−(D/H)ωshould
γ =
= +
+ (ω1,+ + (ω
(ω /H)(P
(ω /H)(P
/H)(P
00 0 /H)(P −
mm m − −
P|E −P P
)e+ f))d1.
ePed ++) d
+ dConsider
(t),
(t),
11d1 (t),
(t),
B =
(2)
(1) (1)
(1)(1)∈ isR is
is
an an
isan
, extended
extended
∈an
extended
an R extended
extended Proof:,
version
Proof:
and
version
version
version Based
version
of
Based of
ofuofu
u on
of
on≤
u ≤≤u
kukL
L≤u,u EMMA
ccin
u,
∙, inin
, u
this
inthis
this
, 1
in
case,
this
1 and
case,
case,
this
and case, the
some
thesome
case,some Schur
some degree
some
degree
Schur degree
degree complements,
degree
complements, of one
one
actuator saturation ω̇ = −(D/H)ω is u- + 1,
0 (ω0 /H)(P
C(P, ρ)
m
is man
e Pf de )|| +
−invariant 1<dU U1 (t),
B ,
ellipsoid. Then (1)u is≤an u extended
is equivalent
c
version of u ≤ uc , in this case, some degree EMMA c c
ṖeṖṖ=
eeṖ= P1=
= PP (δ, ω)
(δ ,,ω)ω)Pω) eP P +P
++ PP (δ
(δ ))+ P
+ PP (δ |E))sat(E
) sat(E
(δ <
f d  f)fdd+
sat(E B d+
, d(t),
d2T2d(t),  (t), of
of saturation
saturation
of saturation c  can can
saturation
ofTsaturation
saturation is
is
is allowed, allowed,
allowed,
directly
directlyis allowed,
is allowed,
allowed, namely
get
namelynamely
getnamely the
namely
thenamely “ς
“ς
above
“ςabove -saturation".
-saturation".
-saturation".
“ς -saturation". result.
“ϛ-saturation”.
-saturation".
result.  
Ṗee =
11P
(δ
P11 (δ (δ ,, ω)
+ eeP
Pee 2+
(δ22P )2 +(δ )3++ (δ33P (δ sat(E
)) sat(E f d ρ)f)d+ )2++ (t),
P of is “ς
aint. Simultaneously
where δ is the
en-
power toP2the (δ
angle
P33 (δ
)inequalities
in rad, ω f d )x(0)
is the
d22(t), 0
relative
≥ and speed
K
in ≥ 0 hold.
wheresat(E
sat(E f d δ f dis γE γE
γE
the power angle in rad, ω isis the therelative relativespeed K
speed
2
 in
sat(E sat(E )) = = dγE ,, x(0) P −1 (u /ρ)I LL
LEMMA 2.
2. “ς
“ς -saturation"
-saturation" isis equivalent
equivalent anytoεεto to the
the inequali-
inequali-
the control laws remain-  2. “ςL-saturation" isequivalent the inequali-
f)fdd= f,fdd c EMMA
EMMA
LEMMA 2. “ς-saturation"
-saturation" is equivalent to the inequali-
  3. T[20] For 0the and
δ))= eeinequalities
R, the the following
following
where =
is the f d,, Lemma  2.
f dpower anglein in p.u.,rad, ω in “ϛ-saturation” TT Tis equivalent to the
sat(E f γE EMMA EMMA is >to ∈ R,
rad/s, P e f dis the
  active power E f d is the field excitation ρρ L x(0)
x(0)T TT T   2.
L EMMA“ς PP 3. ςς [20]
KK For
equivalent any > 0 and inequali- ∈
this case, rad/s, it P
is
rad/s,inPp.u.,
e a is e is
the
critical
UBU
the active
U/|E active power power in
|f|din p.u.,
Up.u.,
U,BU E
E field
f d is
is the the field (2) excitation
(2)
fieldcomplements,
(2)excitation ties
ties
ties ties
ρ x(0)
ρ x(0)TT≥
inequality ≥ 0≥ ≥ 00 and
and
and always
P ς K
P ς KT ≥
holdsς2K ≥ 0 ≥ ≥ 0 0 hold.
hold.
hold.hold.
¸≥000 and and ≥ 000 hold.
BUB/|E
/|E d f|, d|, |, |, |E |E
|E f|fdd≥ ≥ ≥
U B,,B (2)voltage, ρρP x(0)
x(0) P2(u
,, fdthe
fd0 22/ρ)I
voltage γ in γγ = = UUBBB/|E f/|E
> f dis
fProof:|, aaf bound
bound
|E
d|E
fBased
|≥
d | ≥of
B of UBthe the
on Schurexcitation (2) ties x(0)
x(0)x(0) P−1
x(0) inequality
P−1−1
P−1 −1 ≥ and
always
ς KςςK K
(uς cK (uholds
/ρ)Ic /ρ)I
c(u /ρ)I ¸
edbackvoltage control of γγp.u.,
power U 0
= dis field excitation voltage, (u2cc/ρ)I
voltage in
= 1,
= p.u.,
1, B >
1,ex-1, time-varying U B > 0
|E
|E
|E f d|E  is
f|fdd<| | a
< UK
< bound
U U U ,, of the field excitation x(0) PP−1
x(0) ςK
ddii (t) arerobust bounded uncertainties, that is, there
thereProof: ex-
h, a nonlinear (t) are 1,aretime-varying Pff dd|| < <BT,B U
B B,
voltage, bounded di(t) 0bounded and ∗
|E time-varying
K (u2c /ρ)I
B, ≥ 0
uncertainties, canthat
uncertainties, that is,Proof:
be∗ is,equivalently Proof:
ex- Proof:
Proof:
BasedBased
Based
transformed
Based
Based
onon onLon LLEMMA
EMMA
on L LEMMA
EMMA
EMMA
0 and
10 1≤ 1andand
≤11|e| |e|thethe
and
and
the
−
− ee tanh(e/ε)
Schur
the
the
tanh(e/ε)
Schur
Schur
Schur
Schur
complements,
complements,
complements, complements,
complements,
0.2785ε.
≤ 0.2785ε.
≤ one one
one one
one
(3)
(3)
xcitation
where
whereist unknown
input
istδunknown δ δ
there isis constraint
the
the
isexist constants
power
power unknown angle
angle d in
in
dconstants > rad,
rad,0 ω such
ω ω isis that
the
the |d
relative
0|d relative(t)| ≤ speedd
speed , iin=
in 1, 2;
can
can P directly
directly
Proof. getget the
the above
above result.
result.  
ρsuch that ,Pmdirectly ≤ (uBased theon Lemma 1 and the Schur complements, one can
where where isδthe power
the constants
power angle angle in ∗
i rad,
in 0
> Tsuch
rad, isωdthe ¤
that
is>≤ relative
the i
relative
i (t)| speed
≤ speed ∗
dPijd
speed
i in
, ii(t)j
=K can
in1,
∙KP 2; directly
dcan
¤
can m directly get the
2get above above result. result.
where δmechanical
is the power angle
into i x(0)in rad, Px(0) ωDiis the relative and −1/2 in T i −1/2 get
c /ρ)I. the above result. 
rad/s,
rad/s, is Pthe
rad/s,
rad/s,
is
rad/s,
the
e PiPisePis
e=
P[17,
is1,
the the
isthe
mechanical
e is 18]
e 2;active
the
the and
active
Pactive
m is
active the
power
active power
power
power
power
power
power inin
mechanical
Thus, inx(0)
in in
p.u.,
in
inp.u.,
p.u., p.u.,
E fEpower
p.u.,
p.u., Ed
p.u., ∈E fC(P, fEfdis
dDf
isis
theis
is
is
d is the
d
the
the
in the
field
the
the field
field
p.u.,
ρ) field
damping D excitation
excitation
excitation
field
damping
which is the
excitation
excitation
coefficient
damping
coefficient
is an invariant directly
LLEMMA 3.2.
ellipsoid, get
SAEC the above
Control result.
εε > Scheme □□ Let (δ ∗ , 0, P ) and z be the de-
SMIB model
voltagein
voltage
voltage p.u.,
in inin
p.u., H
p.u.,
p.u., U is
UU> Bthe>0
> 00inertia
is isisbound
a aabound bound constant
of of
of
the the
thefield in
field
field s, ω
excitation
excitation
excitation is the
voltage,synchronous
voltage,
voltage, L EMMA LEMMA
3.2.
EMMA 3.T3. 3.[20]
SAEC 3.[20]
[20] For
[20] For
For any
Control
For any
any ε>
any 0>
>
εScheme 0and
0 and 0and e∈
and ee∈ e∈Let
R, R,
∈ R,the
R, the
the
(δ ∗following
following
following
the 0, Pm
following
, following m ) and zi be the de-
voltage
in p.u., in H p.u., is
B BU
the > 0
inertia is a bound constant of the in
x field
T Px s, excitation
ω 0 ρ. is the voltage,
synchronous 2 L EMMA u u 3. [20] For any ε > 0 and e ∈ R, the i
mic voltage
dmodel
did(t) are
are
in
considered
bounded
bounded
p.u., U B
B
in > 0
time-varying
time-varying
is
one a boundcan  ofget
uncertainties,
uncertainties,
the field excitation
≤
that
that 0 is,
is, there
there
voltage,
Therefore,
ex-
ex- inequality
inequality u
inequality
inequality =sired
always
always always
sired always operating
holds =
holds
holds
operating point
holds point of (δ , ω, Pee ) and the error of (δ , ω, P ) and the error variablesvariables re- re-
speed
ii(t)
d(t)are
(t)
dspeed
bounded
are in rad/s,
bounded T
time-varying
time-varying
d0 x and
T P 1/2 T are
uncertainties,
uncertainties,
K the T KP time
that constants
is,
that there
1/2 is, x
ex-
there of
ex-
2 the ex-
inequality
T Px 2 always holds ˆ
i (t) arein rad/s,
bounded T
time-varyingand T are the time constants of the ex-
i d0
(P −1/2
uncertainties, −1/2 that
)P is, there
≤ (u ex- /ρ)x ≤ u . This T
istist
ist unknown
unknown
unknown
ist 4 constants
unknown constants
constants
constants
d0
∗dd>>
diin
∗ ∗
iids, ∗0 >
∗i> 0x0such
such 0suchd0 that
such that
that that |d
i|d i(t)|
(t)| ≤≤ ≤ ∗dd ∗∗
≤,iiidd,=
dsynchronous ,∗∗ii,= 1,=
= 1,
1,1,
2; 2;2;
Pm2; c
PPmmreactance spectively, z̄i =
c
spectively, z̄ = [z [z1 ,, .. .. .. ,, zzi ]]T ,, Bull. i=1∼4.
i=1∼4. Denote
Denote dd(3)
ˆjj asas2017the estima-
the estima-
citation winding is the i , ii = Pm
|dd-axis (t)|
i|d i (t)| 00≤ eetanh(e/ε)
tanh(e/ε) 0.2785ε.
0.2785ε. Pol. Ac.: Tech. (3)(3) 65(1)
(3)
ist unknown
citation winding constants ds,
ininconcludes
iin> d0
xp.u., such
is the
the that proof.
d-axis |d i (t)|  i≤
synchronous 1, 2; P
reactance
m 0≤
00≤
|e|≤ |e|
|e|
− −
|e|
∗|e|e−tanh(e/ε)
i−
ee tanh(e/ε)1 ≤
tanh(e/ε)
˜ ∗≤ ≤
0.2785ε.
i≤
ˆ 0.2785ε.
0.2785ε. (3)
is isis
theisthe
the mechanical
mechanical
mechanical
the
inthe p.u., mechanical 
xd is power
is the power
power
thepower power
d-axis in p.u., p.u.,
d
intransient D
p.u., D DDis is
is
D isthe the
the damping
damping
is reactance damping
the damping damping
i coefficient
coefficient
coefficient
of coefficient
the1],generator
generator in u tions
tions of ≤ d , and
d ∗jj ,toand
− d˜jj = = d −
d ∗jj − dUniversity
≤ d , j = 1,
ˆjj , j = 1,of2.Technology 2. When system (1-2) is op-
op-
is
in p.u., mechanical
x d-axis in
Definitionp.u.,
transient 3s, [11]:
ω
the
reactance
ω For of∈ coefficient
ςsynchronous the
(0, the constraint in ≤Broughtuof c /ς
you d by | Gdansk When system (1-2) is
inin in
p.u.,
in p.u.,
p.u., H
p.u.,x H HH is is
Handis
the the
the inertia
inertiainertia constant
constant
constant
dis the inertia  constant in s, in in
ins, s, ω 0is
ωand
0 isis
the
0 is of
the
the synchronous
synchronous
the synchronous erating atSchemethis point, the ∗ ,relation P ∗ ∗ ∗
E ∗f d = Pm /T

Pe ) +speeddin p.u.,p.u.,
1 (t), Trad/s, is the x(1) are
L inertia the
is  transformer
constant
an extended in s,0version ω 0 is the theu line
synchronous
≤ reactances
ucex-, in this3.2.3.2.
3.2.case,inSAEC
SAEC
3.2. SAEC some
SAEC Control
Control
erating Control
degree
Control at Scheme
Scheme
this Scheme point, Let Let
Letthe
(δ(δ
Let (δ ∗∗,,∗0,
0,∗P
relation 0, PP)mmPand
0, ))m|Pand
and
2and
Authenticated
(δ
z(δ zbe )+P
zi∗izbebe
thebethe
the (δ
3 de-
de-
the de- ∗)
de-
)E =P /Td0 
speed p.u.,
speed in
speed in inin x rad/s,
rad/s,
in rad/s, and
rad/s, T T T x and
Td0 andare
and
and T
and the
TT are transformer
are
are the the
the time time
time constantsand
constants
constants the of line
of
ofthe the
the reactances
ex- ex-ex- 3.2. SAEC in Control ∗ isScheme Let (δ , m
, 0, Pmthe ) 2
) and i )+P
zi be the de-
i 3 (δ m
T d0
d0L
TTsaturation are the the time time constants
in constants of the the fd d0
   Download (δfixes Date 3/2/17 11:05 AM
speed
p.u., V is the thed0 T
infinite ofd0
d0 d0 d0 are
bus voltage p.u.,namely  of ex- hold, Epoint constant ,,ω, ω, )and equilibrium value.
xxreactance xxd + xxsired xxoperating
L ,operating
d0 is synchronous
allowed, = “ς  -saturation".
+ T sired
sired
+ sired operating
operating
hold, point point
point
E ∗f dofis of
of of (δ, ω,
(δ
constant Pω,eP)PeeP)and and
and andthe the
the error
the error
errorerror variables
variables
variables variables re-re- re- re-
Pand fixes the equilibrium value.
) sat(E citation
citation dV 2sswinding
winding
winding inin in (δ
ins,s,
xds,xxdis dxis is isthe
the d-axis
d-axis synchronous dsreactance
reactance
d0
citation p.u.,
fd) +
citation winding (t),is infinite
s, thebus voltage
d-axis
the d-axis synchronous in p.u.,
synchronous =reactance T +sired L , operating point
fd of (δ
(δ ,, ω, e)) and the error variables re-
 (1)
L EMMA
is an  2. version
extended “ς -saturation"
of u ≤T isucequivalent
, in this case,tosome
the inequali-
degree + [k4 + c2 ω0 /(c˙ˆ 3 H)]z2 − (k + p )z then
)
ρ x(0)T
ties ofx(0)
saturation is
≥ 0 and
allowed,
P ςK
namely “ς 0 hold.
-saturation".
≥ − tanh(z2 /ε1 )d1 − f1 (z2 , dˆ1 )2 f2 (z̄33 )] 3
P−1 ς K (u2 /ρ)I c − dˆ2 tanh(z3 /ε2 ) − (H/ω ˆ
0 )[k2 d1 tanh(z2 /ε1 )
L EMMA 2. “ς -saturation" is equivalent to the inequali-
− (H/ω0 )[k2 + f1 (z2 , dˆ1 )]∗ Th
Proof: Based  on L EMMA
 
Nonlinear 1 androbust the Schur
 controlcomplements,
for single-machine oneinfinite-bus power ˙
withˆ1input (zˆ2 , dˆ1 ) f2 (z̄3 )] L EM
ties x(0)get
ρ x(0)T
≥ 0 and ς K (u2 /ρ)I
P ς KT
≥ 0 hold. ˆ1 (t)systems
∗ dtanh(z
− tanh( 2 /ε1f)d(z − f1saturation
3 2 , z3 , d1 )/ε1 )}/P3 (δ ) ,
n (2) can directly P−1the above result.  St
− (H/ω0 )[k2 + f1 (z2 , dˆ1 )]∗ (5) i =Th1
c
n
Lemma Proof: Based
3. [20] onFor L EMMAany any ε 1>and ε0>and the Schur
e 2 eℝ,∈ complements,
the
R, following one in- where sat(E ) ¡ α3z(z–3),= L EM
, in can
eed
L EMMA
directly
3. [20]
get the
For
above result. 
0 and the following where c3 ,c3∗, dppˆ313(t) andand ε2 arefεpositive
tanh( 32(z2are , z3 constants,
, dˆpositive
1 )/ε1 )}/P
z4 =constants,
3 (δ ) ,fd 4 C(z̄3
equality alwaysholds holds ), k4 = (H/ω0)(k1 ¡ k2 p2), f1(z2, St
inequality
-itation always sat(Ekf̂3d= ) (H/ω −̂ α03)(k (z̄13p),12 + k2kc31/c2= – (H/ω0 )(k1 p1 + k2 c1 /c(5) 2 ), not t
d1) = (d1/ε1)sech (z2/ε1), f2(z3) = –(c1/c2)z1 ¡ p2 z2 ¡2 (ω0/H ) i=1
L EMMA 3. [20] For any ε > 0 and e ∈ R, the followingk4 = (H/ω k),2fp3(z ˆ1 ) = (dˆ1 /ε1̂) sech (z2 /ε1 ), citat
moltage, 0 ≤ |e| − e tanh(e/ε) ≤ 0.2785ε. . (3) (3)
where z3 ¡cd3̂ 1, tanh(z 0 )(k p 12− /εand1 2ε),2 , zf3are d̂ 12),=
1, (z dpositive
[k2 + f1(zconstants,2 , d1)]z3. z = C(z̄
inpu3
3 2 4
eret ex- inequality always holds f2 (z̄3 ) = −(cα1 /c 2 )z 1 − kp2 z2=− (ω 0 /H)z 3 − dˆ1+ tanh(z /ε1),
2/c ), not
sat(E fd ) − (z̄
3 3 ), 3 (H/ω 0 )(k p
1 1 k c
2 1 2 t
s, 2; Pm 0 e tanh(e/ε) 0.2785ε. (3) f3 (z2Step, z3 , d4. ˆ1 )The = [kreal 2 +excitationf1 (z2 , dˆ1 )]z 3. ˆ
controller is designed as 2 pres
3.2. SAEC Control Scheme Let (δ ≤ |e| − ≤∗ , 0, P ) and z be the de- k4 = (H/ω0 )(k1 − k2 p2 ), f1 (z2 , d1 ) = (d1 /ε1 ) sech (z2 /ε1 ), citat ˆ
-ficient 3.2. SAEC control scheme. Let (δ¤, 0, Pm)mand zi be ithe desired Step 4: The real excitation controller is designed as
ofω,(δP, )ω, f2 (z̄3 ) = −(c1 /c2 )z1 − p2 z2 − (ω0 /H)z3 − dˆ1 tanh(z2 /ε1 ), inpu
e sired3.2.
ronous operating
operating SAEC point point of (δ, e and Pethe ) and error the error variables
variables respectively, re-
z̄ =Control . ,Scheme Let (δDenote ∗ , 0, Pm ) dand
spectively, z1»4. T
i ] , i=1∼4.
ˆj aszi the be the de-f¤ (z , z , dˆ ) = [k +Ef (z=, E
estima- f1d 2 d1f)]z
ˆ d1 3=. α3 (z̄3 )..(6) (6) prese
nhe ex- z–i = [z
sired 1, i…, zi]1point
operating
[z T,, .i.=
of (δ ,Denote
ω, P ) d̂ j asthe
and theerror estimationsvariables ofre- dj 3, 2 3 1 2
ctance of dd∗j˜j, =
tionsspectively,
and and dz̄j¤d ˜j = ̂ j, jd= ∗ − dˆ , j = 1, 2. When system (1-2) is op-Step 4: The real excitation controller is designed as
1, 2.j TWhen system (1–2)
e
is operating
nator in ¡d
i = [z1 ,j. . . , zi ] , ¤i=1∼4. Denote dˆj as the
0 estima-at L EMMA 4. Consider two sets: C1 (z̄3 , λ ) = {z̄3 |
, erating this at point,
this the
point, relation
˜ the relation P 2 (δ
ˆ ) P + P 3 (δ
∗
¤
)E ¤
= P
∗ m /T
E ∗ d0 hold,
P EU ¤
fdB < |α3 (z̄3 )| ≤ UB E +f dλ = E f d1 3=α3α(z̄
, sign(z
(δ )+P fd (δ ) = /T 3 ).= −1} and C2 (z̄3 )(6)
tions of d ∗ , and d d ∗ d , j 1, 2. When system (1-2) is op-
nces in j = − j = 2 3 m 3 ))
3 (z̄ =
is constant j
and fixes the j
equilibrium∗ value. ∗ ∗ f d d0
Lemma 4. Consider two sets: C1(z–3, λ) = fz–3jUB < jα3(z–3)j
hold,
= + xL , erating E ∗ is constant and fixes
f d at this point, the relation P2 (δ )+P3 (δ ) E f d = Pm /Td0 3 ∙ L the equilibrium value. 
{z̄ | P(z̄ 3 ) < p 3 }, in which λ named “saturation coefficien-
UEMMA B + λ, sign(z 4. 3Consider α3(z–3)) = –1g two andsets: C2(z–3) C =1fz (z̄–33jP(z
, λ )–3)=<{z̄ p33g, |
, ω) = Step hold, 1: E ∗ is constant and fixes∗
Define z
Step 1. Define z 1 = δ¡δ . Choose Lyapunov function = δ − δ ¤ .theChoose equilibrium Lyapunovvalue. function t” is a
in|αwhich given λ positive
named constant,
“saturation P(z̄ 3 )
coefficient’’ = P 3 (δ
is ) (|α
a given )| −
3 (z̄3pos-
f d 1 U B < 3 3 (z̄ )| ≤ U B + λ , sign(z α
3 3 3 (z̄ )) = −1} and C 2 3) =
(z̄
V1 =V(c Step 1: 2 , Define
the2 time z δ
derivative δ ∗ . Choose Lyapunov functionU )/|z |.
of V along system trajec- The set C(z̄ –, λ ) = {z̄3 | |α3–(z̄3 )| ≤ UB + λ } is a re-
11 /2)z
= (c1/2)z 1 1, the time 1 = derivative of V1 1 along system trajec-
− {z̄3 | itive
B
P(z̄3 )constant,
3
< p3 }, in P(zwhich
3 3) = Pλ 3(δ)(jα
named )j ¡ UB)/jz3coefficien-
3(z 3“saturation j. The set
toriesVtories1is=V̇(c 1is 1= V̇1−p
/2)z 2
=1 ,1–the cp11zc21time1z+12 + c1derivative
cz11zz12z,2,ininwhich of
which V1 along cc11 and systempp11 are trajec-
are posi-
posi-gion C(z of –attraction
3 , λ) = fz – if C (z̄
3 jjα 3 1(z–3)j 3,∙ λ )U⊆ C2λg
+ (z̄3is).a region of attraction if
2 + c z z , in which c and p are posi-t” is a –given positive constant, P(z̄3 ) = P3 (δ ) (|α3 (z̄3 )| −
B
tories
tive constants,
tive constants, is V̇1 = α1−p (zα111)(z c1=1z)1−p = –11pz111z1is 2 isthe thevirtualvirtual1 control control 1 of of ωω and and C1(z3, λ) µ C2(z–3).
U B )/|z T
3 |. The
HEOREM set 1.
C(z̄ 3 For
, λ ) SMIB
= {z̄3 |power |α3 (z̄3system )| ≤ UB which + λ } isisa ex- re-
s z2 =tive
inaries
zω2 = −constants,
ω¡α
α1 (z11(z ).1).α1 (z1 ) = −p1 z1 is the virtual control of ω andgion pressed of by
attraction the nonlinear
if C (z̄ , λ
dynamic
) ⊆ C (z̄equation
). (1-2) containing
z2 = ω − α1 (z1 ). 2 Theorem 1. For SMIB 1 3 power system 2 3 which is expressed by
StepStep 2: Augment 2:
Lyapunov function as V2 = =VV1++(c(c/2)z 2 /2)z 2 2 , influences of input saturation and time-varying uncertain-
2the
Step 2. Augment 2
Augment Lyapunov function as 22 V11 + (c2 2/2)z2 2, theT HEOREM
Lyapunov function as V , nonlinear dynamic 1. For SMIB equationpower (1–2) system containing which the influ-
is ex- whe
then V̇ ̇ ∑ p 2
c z 2− c z c z d ˆ ̂ tanh(z ties indicating
ences of inputthe saturation
model errors, and the excitation
time-varying controller indi-
uncertainties (5-6)
er- nor- then then 2 V̇V = −
22= =−– ∑i pi i ci i zi ¡ 2
−2c2 (ω (ω /H)z
0 0 /H /H)z )z22 z333 − − ˆ
¡cc222zz222dd111tanh(z tanh(z2 /ε /ε
22/ε11))1+
)pressed by the nonlinear dynamic equation (1-2) containing tune
+ +
i=1 i=1 ensures convergence of z̄
cating the model errors,3 the excitation controller (5–6) ensures to the origin under “λ level satura-
r- oper-c2 z2 dc+ as αα22(z̄(z̄ the
+ influences of input saturation and time-varying uncertain- wher
12 z(t),2cd 2the
2 1z(t), d1(t), virtual
the the
virtual control
virtualcontrol ofofPP
control e eis of ischosen
Pe is chosen
chosen as as2 )2α)=2= P–m2P)+
(z m=tion” convergence
if z̄3 (0) ∈ C(z̄ of 3z–,3λto). the origin under “λ level saturation’’ if
l small +ˆ+ ˆ1ktanh(z ̂ 11tanh(z c2, εp1ε2 are
ties indicating
are z–3(0) 2 C(z–3, λ). the model errors, the excitation controller (5-6) tune
(H/ω =0 )[k
(H/ω Pm 10+)[k z1(H/ω 1 z1k+
+ 20z)[k
k22+ z12z1d 1dtanh(z 2 z2 + 2d/ε
2 /ε )], where
1)], 2/ε1)],cc2where
where 2,, pp22and and 1and τ an
ε1 are positive constants, 1k/c =+cp1p1/c pp212,, kk–p 2 p + p −ensures
, Proof:convergence Based on theofdesigned z̄3 to thecontroller origin under (5-6), “λ the level following
satura-
 positive positive constants, constants, k1 k=1 = c1c/c 21 2+ 1D/H 2+p
D/H −1 D/H
− 1 2 2==1 k p
121 =+ p2 p
1 2+ − vary
0 > D/H, 0 D/H, +z3p2=¡ z3PD/H, P −z α=(z̄P ).¡ α2(z–2). Robust
Robust control controlfor forpower
powerthree
tion” cases
if
systems
Proof.
systems z̄3 (0)with are with
Based given
∈inputC(z̄ input3 ,the
onsaturation λ ).
to prove
saturation
designed L EMMA controller 4 and (5–6),T HEOREM the following 1.
e −eα23(z̄22 ).2e
= stabi
Robust control for power Casethree
systems 1: There
cases
with are is
input no
given saturation
to
saturation prove happened,
Lemma 4 and i.e.
Theorem|α 3 3 1.
(z̄ )| ≤ UB . τ and
Step 3: 22+ 2 Proof: Based on theisdesigned controller (5-6), the following man
Step
Step 3.3:Consider Consider
Consider Lyapunov Lyapunov
Lyapunov function
function
function
Bull. Pol. V3V V =33 =
Ac.: V 2V
=Tech.V+ 22+ +(c3(c
(c
XX(Y) 3 /2)z
/2)z3 /2)z
3 3+
2016 Invoking
3 + cases
Case LCaseEMMA 2: There2: 3There directly saturation for the
is saturation happened,
equation
happened, i.e.(4), UB the i.e.
< |α following
U3B(z̄<3 )||α≤3 (z̄3 )| vary
≤
23: 2Consider Bull. Pol. Ac.: Tech. XX(Y) three
2016
2 + Case UBcan + 1. are
λbe , Noand given in the to prove
meanwhile L EMMAsign(z 4 and
3 α33(z
T HEOREM
– )) ≥ 0. Thus, 1.
Step ˜
˜ 2
2 Lyapunov function V = V + (c /2)z results U Case + λ 2:, saturation
obtained
and There in the
is occurs,
meanwhile
saturation i.e. sign(z
jα
happened, (z̄33)jα∙ i.e.U
(z̄ BU.
)) Invoking
≥< 0.
|α Thus,
(z̄ )| ≤stabi
Pr
j .. Differentiating
DifferentiatingVV33,,,one
d∑˜2d.djDifferentiating one
one gets 3 Case 1: There is no saturation happened, i.e. |α33(z̄B3 )| ≤ U
B 3 3 3 .3
(1/2)+(1/2) (1/2) ∑
2 j=1j 3 gets 3 2 3
Lemma 3 directly for the equation (4), the following results B
man
ities
Invoking UBL+EMMA λ , and33indirectly
3 the meanwhile3
for the equation sign(z3 α(4), 3 (z̄3the 0. Thus,
)) ≥following
(1/2) j=1 ∑ d˜2j . Differentiating V3 , one gets canV̇be 3 ≤obtained− ∑ pi cV̇ 2
i z3i ≤ − 2 ∑ pi ci z2i + σ ,
i=1 3∑ i i i
3 V̇ ≤ − p c z (7) In
j=1 3 results can be
3 obtained Pr
V̇3 = − ∑
V̇3 = − ∑ p c
i 2i i z 2
+
pi ci zi + c3 P3 (δ ) z3 z4
c P
3 3 (δ ) z 3 4z
V̇3 ≤+−c3∑ i=1 pi)c|z 2 i=1
i z3i | sign(z
In
ities
3 i=1 P3 (δ 3 3 α3 (z̄3 ))(UB − |α3 (z̄3 )|) + σ (8)
where σ = 0.2785ε d31 (δ∗
|z3c|3sign(z
H/ω α+ 0.2785ε
))(UB 2−c3|α
∗
d23.(z̄3 )|)(7) + σ P(z̄
3 ≤−∑
i=1c31V̇ P (c2) + pi ci z2i03)+ (8)In3
V̇3 = − ∑ i=1
+ cp2izc2idzi1 (t)
2
+ c−3 P c32 z(δ 2 d)1 ztanh(z
∗
3 z4 2 /ε1 ) 3
+ 3σ(z̄,,3(7)
+i=1 c+2 zc23dz31d(t) − c z d1 tanh(z
2 (t) −2c32z3 d
∗∗
2 tanh(z3 /ε2 )
2 /ε1 )
(4) ≤ −+ ∑cp3iP3c3i z(δ 2
i + σ 3. |i=1
) |z sign(z3 α3 (z̄3 ))(UB − |α3 (z̄3 )|) + σ (8) In
+ +cc− 23zz(c23dd1H/ω (t)(t)− − cc2 zz2 dd1ztanh(z∗∗
tanh(zˆ Robust /ε
23(t) /ε12))control for power where
(4) (4)
Bull.
systems
Pol. σwithAc.:
= 0.2785ε ≤
Tech.
input
i=1− 3 ∑
XX(Y) p ∗ c z
2016 2
+ σ.
1 d1 (c¤2 + c3 H/ω0 ) + 0.2785ε2 c3 d¤2 .
saturation
i i i ∗ P(z̄3
32 0 ) f 33(z32 , 2 3 , d1 )d
+ −c(c 3 z3(c d2H/ω
3H/ω (t)0− )0f)3cf(z3 z(z
∗
3 d2 tanh(z
Robust
ˆˆ )d13∗(t)
1
/ε2 ) control for power (4)
systems where Case
with σ =3:
input
≤ − ∑
0.2785ε
There i=1 p
saturation cis1
i i i
dz 2(c2 + c3 H/ω
saturation
1 + σ . happened,
0 ) + 0.2785ε
i.e. U B2 c
< 3 d |α
1 .3 (z̄3 )| ≤
3 2 ,2z, 3z3, ,dd11)d
− 1 tanh( f 3 (z2 , z3 , d1 )/ε12),
ˆ UB + Case λ , and3: also the is inequality sign(z 3 α3 (z̄3 ))i.e. = −1 holds,
Step 3: Consider Lyapunov function
3
V3 = V2 + (c3 /2)z23 +Bull. Pol. Case Ac.: 2:Tech. There
i=1
XX(Y)Thereisλ occurs,
saturation
2016 saturation happened, happened,
– i.e. UB < |α UB (z̄ < |α≤ 3 (z̄3 )| ≤
ˆ Case
i.e. 2.
z̄ Saturation
C In i.e.
this U
case, < one
jα (zcan )j get
∙ U the + λ, 3and3 )|
following in≤
Step the 3: −Consider(c
− (c3 H/ω H/ω Lyapunov
) f
0 ) f 3 (z
(z , z , d
function
, z3 , df1d)d ˆ )d ∗(t) V = V + (c ˆ
/2)z + Case U 2:3 ∈ There
λ , 1 (z̄
and 3 ,
is ).
saturation
also the inequalityB
happened, 3 3
i.e.
sign(z U α B< |α (z̄ )| holds,
1 tanh( f3 as(z2 , z3 , d1 )/ε λCase
B, and3: in There isαsaturation happened, i.e.
3 3UThus,
2 expecting 3 0
action 3 2
of2sat(E 3 1 1
) is chosen
3 2 3 3 1 ), + B (z̄ ))
3 =3 −1
3 < |α3 (z̄3 )| ≤
3 α3 (z̄3 )) ≥ 0. B
Uthe +meanwhile the meanwhile – sign(z
Binequality sign(z 3 3(z3)) ¸ 0. Thus,
(1/2) 2∑ d˜2˜− 2 . Differentiating V , one gets
j (c H/ω ) f (z , z 3 , d ˆ )d ∗
tanh( f (z , z , d ˆ )/ε ), U B + U λ
i.e. , + and λ z̄,3 in
and
∈ the
C 1alsomeanwhile
(z̄ 3 , λ
the ). In
inequalitysign(z
this 3 α
case, (z̄
sign(z
3 one
3 )) α ≥ can 0.
(z̄ Thus,
get
)) = the
−1 following
holds,
the∑ d˙ˆ13= action
d j . Differentiating V 33, 1one ) f1d )gets 3
B 3 3
(1/2) j=1 expecting c2 z02 tanh(z 3of 2sat(E
2 /ε
1is chosen 3 as 2 3 1 1 32
the
j=1 expecting action of sat(E ) is chosen as i.e.inequality z̄−33 ∈ pCpc1c(z̄ 232 , λ ). In this case, one can get the following
the expecting 3˙ˆ action of sat(E fd ) is
+ c3 (H/ω0 ) f3 (zf 2d, z3 , d1 ) tanh( f3 (z2 , z3 , d1 )/ε1 ), chosen
ˆ as ˆ V̇
V̇ 3 V̇ ≤3
3inequality
≤
≤
−
−
∑ ∑ ∑ p i i i zi2zi i
c z (9)
3d1 = c22z2 tanh(z2 /ε1 ) i=1i i2 i
V̇3 = − ∑ p˙ˆi ci z2i + c3 P3 (δ ) z3 z4 i=1
V̇3 = − ∑ PP323∑
˙
dˆ1 p= di 2cci=z +zc33tanh(z P3 (δ ) z13)z24), V̇+ 3−≤
i=1
(δ))p|z i c33i||(p
2
zsign(z +σ,
i=1 +2izcc233tanh(z
(H/ω 20/ε)3f/ε ˆ
3 (z2 , z3 , d1 ) tanh( f 3 (z2 , z3 , d1 )/ε1 ),
ˆ cc33− (δ |z i 3 |z3 |/P 3 α33 (z̄3 ))(UBB − |α33 (z̄
(δ ) +U − |α (z̄3 )|) 3 )|) + σ (8) (9)
c P 2sign(z3 α3 (z̄3 ))(UB − |α3 (z̄3 )|) + σ (8) (8)
i=1
+αc32ˆ˙(z̄ z23d)+ = {−P
1 (t) c −
(H/ω c1 (δz , ω)
d )
∗ Pe − P2 (δ
f tanh(z
(z , z /ε
, d ˆ ) − k3 z1
) tanh( f (z , z , d ˆ )/ε ), and
+
V̇3furthermore,
≤− 3
3
3
∑ (δ
i=1 ) |z 3 |
pi ciifzi the condition C1 (z̄3 , λ ) ⊆ C2 (z̄3 ) holds,
+ c2dz22= d1 (t)c3 z3− 3 tanh(z
2 2
c2 z2 d01∗31/ε 3 2 3/ε )
tanh(z 2
2 ), 2 12 − (k2 + p3 )z3
1 3 2 3 1
(4)
1
−∑ 3 −
picc3i P
i=1 z22i3a+ (δ σ) .|z3 |(pp3> |z30|/P 3 (δ ) +UB − |α3 (z̄3 )|) + σ (9) ,
+ c˙ˆ3 z3 d2 (t)+−[kc4 3+z3cd2∗2∗ωtanh(z 0 /(c3 H)]z then ≤there exists constant such that
α+ cd332z)3= d2c(t) 3 z3−tanh(z z3, dω) 3 /ε 2 ) (4) ≤ − ∑ − pc i cP i z i + σ |z. |(p |z |/P − |α σ
− d1ˆc2(δ tanh(z
3 (z̄ = {−P 3tanh(z 32/ε P2e),− P32/ε (δ2 ) − k3 z1 ˆ and i=1 i=1 3 3
furthermore, (δ ) 2 3if the 3 3condition 3 (δ ) +U (z̄ )|)
C1B(z̄3 , λ3) ⊆3 C2 (z̄3 ) holds, + ,
3 /εˆ1 )d 2 ) − (H/ω0 )[k2 d1 tanh(z2 /ε1 )
α− −
3 (z̄
(c
(c3 3)H/ω 3 H/ω
= {−P +0[k 0 ) f 3
(δ
) f41tanh(z
3+
(z
(z2c,,ω)
2 ,
2zω
z 3 , d
,0Pd/(c
eˆ1− )d ˙ˆ3P
1
1H)]z
(t)
2(t)(δ )2 − k(k z p3 )z3
ˆ31 )21f+
Case
Caseand then 3:
3: There
there
furthermore,
There
V̇ ≤
3exists
is
∑ifa constant
is saturation
−
saturation
p c
the z 2
i i i happened,
− pc
condition
happened,
P (δ
p3 >3 0 Csuch ) i.e.
|z 3| +
1 (z̄3U ,UλBσB)<
that .< |α3 (z̄(10)
) ≤
⊆|αC32(z̄(z̄33)|)| holds,
− 3/ε ) d − f (z , d (z̄ )] i=1 – i.e. 3also≤
(5) − (c3 H/ω ) f 3 c22 ω03/(c
(z , 2
z , d1ˆ )d 1 ∗
tanh(1 2
f (z ,2
z 3
, d ˆ )/ε ), U Case+ λ 3., Saturation
and also occurs,
the i.e.
inequality U B < sign(z
jα 3 (z 3 α
)j ∙ (z̄ U B)) + = λ, and
−1 holds,
− (c3 H/ω +−0[k 0ˆ4 +
d−)2f(H/ω3tanh(z
(z2 , z03)[k 3,/εdˆ2112+ )d3)H)]z ∗1 2 − (k
f1tanh(
1− (H/ω ˆf30(z
32+
)[k 2 pˆ33 )z
2 ,2zd31, d
ˆ113)/ε112),/ε1 )
tanh(z
B
UBthe +thenλ , and therealso exists
ofthe
a inequality
constant
– 2 three p > 0 such
–1sign(z
3
αi.e.3that3 = −1
3 (z̄z–33)) –proofholds,
2 , d1 )]∗ i.e. The z̄inequality resultssign(z
3 ∈ C1 (z̄3 , λV̇).3 ≤
the3αIn above
3(zthis case, 2holds,
cases
one 3conclude
can get)2|zthe the
C 3, λ). In
1(zfollowing of
∑
3)) =
(z
the expecting action ˆ
− d∗2dtanh(z of sat(E ) is chosen as ˆ i.e. z̄ ∈ C (z̄ , λ ). In − this p
case,
i c i z −
one pc can
3 P 3 (δ
get the 3 | + σ .
following (10)
2 d1 tanh(z L 4 and
this case, one can get the following T 2 1. i inequality
3 /εff2d˙ˆ))(z − ,(H/ω )[k)}/P /ε ) EMMA 3 1 3 HEOREM
the expecting action − tanh(z ˆof
1 (t) sat(E
2tanh(
/ε 1f)dd31 is − 2 chosen
zf31,(z ,0dˆas
dˆ12)/ε 11) f 2 (z̄33(δ )] ) , 2 1 inequality
inequality Step 5: Note V̇3 ≤that, − ∑forpi cgiven
i=1 2
pc3 P3 (δ ) |zλ3 |,+pσi , . ci , ε j , (10)
i zi − parameters
d˙ˆ˙ˆ1 = c2 z− 2−tanh(z
tanh(z
(H/ω 2 /ε
2 /ε 1 1) ) d ˙ˆ − f (z , dˆ ) f (z̄ )]
+ f1 (z2 , d1 )]∗
1 1 2ˆ 1 2 3 (5) i = 1,The 2, 23, results
j = 1, 2,of the
how i=1 above
to estimate three thecasesregion conclude
of attraction the proof of
d1 = c2 z2 tanh(z 0 )[k1) 2
where+ cc− , p
2 /ε
and
f (z +, z2f31,(zε are
dˆˆ12),,tanh( positive
ddˆˆ11)]∗ constants,
, dˆˆ1))/ε z = V̇C(z̄
3 ≤
V̇3 ≤L3The ,− λ
EMMA
∑)∑2
is p the
results i c
p c4i zandi z
2
2
key
i of Tthe point.
HEOREM However,
above three 1. cases conclude the (proof
our goal in this paper is
9)
(9) of
sat(E+f dc)3∗(H/ω
33(H/ω ˆ1 (t)3 0tanh(
d(H/ω α30(z̄
))[k
)0f33),3(z22f2,3z(z 3k,32d,1z= )3tanh( )/εff313(z (z2 , z3 (δ
)}/P , z 3p , d )/ε , k
1 ), 4
),c not
−
to i=1 i on
focus
i
this estimation, but instead to redesign the (9)
ex-
− (H/ω 0 2
)(k 3
1 1 1 + 1 2 1 2 /c ), L EMMA Step
i=1 4 5: and Note
T HEOREM that, for
1.  given parameters λ , p i , ci , ε j ,
d˙ˆk˙ˆ42 = c(H/ω 3 z∗ dˆ10 )(k
3 tanh(z (t)1tanh( 3−/εk22),fp32(z ),2 ,fz13(z, d2 ,1d)/ε
ˆ ˆ ) 1=)}/P (dˆ13/ε(δ ), 2
1 ) sech (z2 /ε1 ),
(5) citation i − 1, ccontrol
3P 2,3 3,(δ )law j |z=3 |(p (5-6)
1, 32,|zto |/P
3how stabilize
3 (δto theBnonlinear
)estimate
+U − |αthe )|) + σof
3system
3 (z̄region with
, attraction
d2 = c3 z3 tanh(z3 /ε2 ), 1 Step
− c3 P35:(δ )Note
= |z3 |(pthat, 3 |z3 |/P for3 (δgiven ) +UBparameters− |α3 (z̄3 )|)λ+, σp, i , ci , ε j ,
αwhere f (z̄ ){−P
3 (z̄32) =3 c , ˆ1−(c
= p(δ3 ,1ω) /cand2P )ze 1−− εP22p(δ z2) −
2are − kpositive
(ω
3 z01
/H)z3 − dˆ1 tanh(z2 /εz1(5)
constants, ), = input C(z̄ saturation. Then
λ is the key the proposed
point. However,SAEC control
our goal strategy is
α3 (z̄3f)3 (z =2{−P 3 (δ , ω) P − P (δ ˆ ) − k z 4 and i =
presented
1,
furthermore, 3 2,
, 3,
)
as
j
follows
= if 1, the2, how
condition to estimate
C 1 (z̄ 3 , λthe) ⊆region C2 (z̄3in ) this
of holds, paper is
attraction
sat(E f d )c,+z33,− , d11 ) = [k2 + f1 (z22 , d1 )]z3 .3 1
α
e
k p k c and furthermore,
not to focus ifonthe this condition
estimation, C1 (z̄but 3, λ ) ⊆ C2to
instead ) holds, the ex-
(z̄3redesign
where [k
4:44+
p
+ 3 (z̄
3 c2real ω0 /(c
and
3 ), ε are
32H)]z2 controller
3 = positive
(H/ω
− (k + p )zdesigned 0 )(k constants,
1 1 + 2 z1 /c
as 2 4 then there
= 2 then
), C(z̄there 3 , λ exists
) is the a  key
constant point. p > However,
0 such that our goal in this paper
5 is
Bull. Step Pol. + [k Ac.: The c2 ω65(1)
Tech. 0 /(c H)]z2 − (k2ˆ2+ p33is
excitation
32017 )z3ˆ3 citation exists control a constant law p>
(5-6) 0 χ),
to such
stabilize that
z̄ the λ
nonlinear system with
sat(E
k4 = f(H/ω d )−− α3 (z̄
)(k
dˆ20 tanh(z 1− 3 ),k 2 p2 ), k3(H/ω f1= (z2 ,(H/ω d1 ) d= (d11/ε
)(k
ˆ10tanh(z p11/ε )+ sech k c(z/c/ε
) 22 1 2 (6) 1 ), not to focus
2 ), on this sign(P(z̄ estimation,
3 ) − but 3instead
∈ C 1 3to redesign the ex-
(z̄ , ),
3 /εE2 f)d− 0 )[k α32d(z̄ BroughtL to = you22by | Gdansk University of Technology
kf42 (z̄ =3 )(H/ω =− dˆ−(c 20 tanh(z
)(k11/c−23)z k/ε21p2 )2−), −= p(H/ω E f d1−=
2f 1z(z2 2 ,0d ˆ(ω
)[k ) 2 0=ˆ/H)z
1 (
).ˆ −2/ε
3tanh(z
d 3/ε 2 ) d ˆ11)tanh(z
sech (z /ε
2 /ε ),
1 ), input
citation saturation.
control
V̇ ≤ − ∑pi ci i zi 2i i − pc33P33(δ )Authenticated
law
−1, Then
(5-6)
p c z 2 the
to
− pc proposed
stabilize
P (δ )the
otherwise,
|z SAEC + σ . control
3 |nonlinear system strategy
(10) with is
(z32), z3=,−dLˆ−(c
f23(z̄ tanh(z 4.
222 )z11 ˙ˆ1−
/ε ) fd ˙ − f ˆ (z ,two
ˆ
Consider d d .ˆ
1
) fsets:
(z̄ )]
1 1
C ˆ λ
2 1
presented
input V̇33≤as−
saturation. ∑
follows Then the proposed |z
SAEC + σcontrol
3 |11:05 .AM (10) is
strategy
/ε1 )d11−2pf21z1(z 1 d31 tanh(z2 /ε31 ),
1 )
EMMA = [k
1 /c + (z , 12 −)]z2 3ˆ(ω 1 0 /H)z2 3 3 − (z̄ , ) = {z̄ | i=1 Download
1, z̄3  ≤ τ or L = −1,
Date | 3/2/17
− tanh(z 2≤ 2 , d1α) f(z̄ 2 (z̄3 )] i=1
UB <4:|α
Step ˆ The
3 (z̄3 )| real B + λ , sign(z
Uexcitation ˆ controller 3 3 3 )) is −1} and Cas
=designed 2 (z̄3 ) = presented as s =follows
 UB +z̄3λ ,∈and
i.e. 3 , λ ).
C1 (z̄also theIninequality this case,sign(z one can get the following 3 ∆ f d1
3 α3 (z̄3 )) = −1 holds,
001
inequality the
E f d2sinceswitching
areρrealized is small, strategy in
theCases is proposed
variable 1,2 χ can andreach to
Case make a3 small logical
respectively.
neighborhood choice Thus, be- Step
i.e. z̄3 ∈ C1 (z̄3 , λ ). In this case, one can get the following tween
the of them
the
switching manifold basedstrategy χ0on = P(z̄ the
is 3proposedpower
) + l∆ , then systems toP(z̄ make <χ
3 )operating and the
logical conditions.
condi- be-where
choice
can activate the dynamic system ρ χ̇ = P(z̄3 )− χ +ltion 1 0 T , d = [ d d ]T . wh0
inequality2 ∆ . As z̄ such, C holds.
As such,
tween them 3 thebased
∈ transient
(z̄
2 3 0pon )| 1 the
03 =χ performances z
power systems z1 z2
operating 
are improved. conditions.timewh Ade+
V̇3 ≤ − ∑ pi ci zi since ρ is small, the variable χ can reach a small
2
Y. neighborhood
Wan Consequently, Step
on the V:basis Choose of T Lyapunov 1 function
and L as Ve4,= (1/2)ēT3 Q
i=1
2
of the manifold χ = P(z̄3 ) + l∆ , then P(z̄3 ) <(9) As
χ andthe such,
theconvergence the
condi- where transient performances HEOREMare improved. 
EMMA
ten+
A
V̇3 ≤ − ∑ pi ci z2i tion z̄ ∈ C (z̄ )| of z̄3 0to<theQorigin = QTcan ∈ Rbe 3×3 are to be chosen later. V̇
ensured. Thus
e=
− ci=1 3 P3 (δ ) |z3 |(p33|z3 |/P 2 3 (δ 3 )p3+U=χ holds. B − |α3 (z̄3 )|) + σ , (9)
3.3. Some Discussions time derivative In the above
of Veofalong subsection,
the system Etrajectories the SAEC is ten
Therefore, the main advantages sub-controllers f d1 and =
and−furthermore, if Consequently,
the condition Con (z –the basis of –T HEOREM 1 and
, λ) µ C (z ) holds, then control
3.3. EL
3.3. Some
EMMA scheme
Discussion.
are Discussions
4,
realized is proposed
In
in the
Cases In
above 1,2 to
the andensure
above
subsection,
Case stability
subsection,
3 the SAEC
respectively. of power
the control
Thus, SAEC sys-
and furthermore, c3 P3 (δ ) |z ifthe the
3 |(p condition
3 |z 3 |/P3 (δof
1
)C+U 3 −λ |α
z̄13(z̄toB3 ,the ⊆ 2
)origin 3
3 (z̄C32)|) (z̄3+)beσholds,, control tems
f d2
V̇eexist T
3A
T
Qē T
ē QDd sys- giv
convergence can ensured. theeven scheme isifproposed
there is proposed control
k to 3 +input
ensure saturation
stability ofchoice andeven
power time-
there exists a constant p > 0 such that scheme =ē
to is ensure stability of zlogical
power systems
switching strategy proposed to3 make be-
then there exists a constant Therefore, p > 0
the
and furthermore, if the condition C1 (z̄3 , λ ) ⊆ C2 (z̄3 ) holds, if such
main that
advantages of sub-controllers varying
tems E
even
therethem and
uncertainties.
if
is control there
basedinput exist But T control dueT to input the auxiliary
2saturation Tdynamic and sys-
time- sur
giv
tween f d1 on=ē the 3saturation
power
[A k Q+ and time-varying
systems
(1/(2ε operating
))QDD conditions. Q]ēuncertain-
3 where A
E f2d2 are realized in Cases 1,2 and Case 3 respectively. temties.
varying Asand such, Thus,
But since
switching
uncertainties.
the transientthestrategy auxiliaryBut
performances are
due dynamicneeded
to the system
are auxiliaryin SAEC,
improved. T and  2switching
dynamic the costsys- 2ofA +2sur if t
then there exists a constant p2 > 0 such that − (1/2)εd z − (1/ε)D Qē3  + (1/2)ε d z  ,(ς
V̇3 ≤ − ∑ pi ci zi − pc3 P3 (δ ) |z3 | + σ ..(10)
the switching strategy is proposed to make logical
(10)temcontrol choice
strategy
andsystem be-
are needed
switching is often strategy inmore SAEC, thanthe
are that
needed cost of in of control
general
SAEC, system
linear the control cost is oftenuatio if t
tween 2
i=1 them based on the power systems operatingoften
system.
conditions.
3.3. more
Some
Thus, than where
Discussions
an that
alternative
control system is often more than that of general linear control Aof general
=
In Athe when
control linear
above there control
scheme is
subsection, no system.
is saturation,
shown
the Thus,
SAEC in and this A = Ak
V̇3 ≤ −As
The results of thei=1
∑ pi ci zthe
such,
2
) |z3 | + σ . are improved.
pc3 P3 (δperformances
i −transient
above three cases conclude the proof ofsystem.
(10) an
subsection
alternative control
control Thus, scheme
to facilitate A is+
an alternative (ς
proposed −
k
scheme
1)br
engineers to whenis shown
ensure
control there
to trade in
is
stability
scheme this
ς
off between of subsection
-saturation, power
is showncost εsys- is to
in this
k
disturbanc
and T
The results of the above three cases conclude the proof of facilitate tems even engineers
if tenuation
there toexist trade off between
constant.
control input cost and performance:
saturation and time- given ς
L EMMA 4 and of T HEOREM 
1.three performance:
subsection a sufficient
to facilitate engineers condition to trade for allowingoff between γ-saturation
cost and
The results
Lemma 4 and the 3.3.above
Theorem Some 1. □□ cases conclude
Discussions In the above the proof subsection, of avarying
sufficient
the SAEC condition for
uncertainties. But allowingdue to theγ-saturation auxiliary dynamic of a kind sys- of sures co
Step 5: Note that, for given parameters λ , p , c , ε ,
of a
jperformance: kind of linear
a sufficient excitation T HEOREM condition control input
fortheallowing
3.presented,
For is
error presented,
and dynamic γ-saturation and a(12-13)
1.  is proposed to ensurei stability costsystem
i of linear excitation control input
) L EMMA 4 and T HEOREM control scheme power
tem and sys- switching strategy areisneeded in SAEC, a the
method for
of if there
i =Step1,Step
2, 3,5:5. jNote
= 1,that,
Note 2,that,
tems howforeven to given
given
for estimate
ifparameters theλ,region
thereparameters
exist pi, ci,λεof
control , ip=
,jinput attraction
, 1, 2, 3,
saturation
c , ε method
of, athe kind
and for
estimation of
time- thelinear estimation
given
of excitation
the ςregion
∈ of 1],
(0, the
control
of region
ρattraction
> input
0, of0< of ε attraction
is <is1,presented,
given. the control is given.and
inputa u = rē3
i i j control system is often more than that general linear control and
=) C(z̄
i =31, ,jλ= is2,the
2,) 1,
3, =key
j how point.
1,to2,varying
estimate
how to However,
theestimate
region of
uncertainties. ourthe goal
attraction
But dueintoC(z
region this
of paper
– auxiliary
the is theismethod
3,attraction
λ) Next,
dynamic
system. Next, for
the sys- thesystem
the
Thus, estimation
sures
system (1-2)
an alternative (1–2)of
convergence the
isisequivalently of ē3scheme
region
equivalently
control toof attraction
atransformed
ball is of
transformed shownradius isin εgiven.
into into
thisan anς -satura
under
ifethere Q fol
and
=, not
C(z̄to focus
key point. onHowever,
thistem estimation,
and goalbut
ourswitching instead
in this strategypapertoare isredesign
notneeded to focus the
in SAEC, ex-error
on error
Next,thesystem.
subsection system.
costtheofsystem toLet Let
facilitate be
iei(1-2)beexist thetheisstate
engineers state satisfying
variables
variables
equivalently
to trade off of this
thiserror
ofbetween
transformed errorcost system system
into
and an
3 , λ ) is the key point. However, our goal in this paper is fol
citation thiscontrol
estimation, but instead
control tosystem to is redesign
often
the more the excitation
than that ofcontrol
general and
linear e–control , .1.,Let
.…, T
e]Tii],be ,i=1∼3.
i= the1»3.
,, on law (5-6) stabilize nonlinear system with and
error ēisystem. ,aeisufficient stateAvariables of 2this error
γ-saturation T system
performance:
= i [e=1[e condition T for allowing
not to focus thissystem.estimation, Thus, but instead
an alternative to redesign
control schemethe ex- Q + (1/(2ε ))QDD Q < 0,
,, input law (5–6) to stabilize
saturation. Then thethe nonlinear
proposed systemcontrol
SAEC with input strategysatura-is isand shown
Stepofēia=kind in[eDefine
I: this
1 ,of . , eei ]1T=
. . linear , i=1∼3. δ − δ¤ ∗ .control
excitation The input is presented, ofT eand a
citation control law (5-6) to stabilize the nonlinear system with T time derivative 2 1 along
presentedtion. Then the
as follows subsection
proposed SAECto facilitate
control engineers
strategy isto trade
presented off between
as Step
method cost I. and
Define
for the e
estimation
1 = δ ¡ δof . A
The
the kς Q
time
region + derivative
(1/(2ε
of attraction ))QDD of is e Q
given.
1 along < 0,
, inputfollows
saturation. Then the proposed SAEC control strategy system
is Step
γ-saturation
I: Define e1is=ėδ1 =
trajectories − δ−q.1The ∗ e1 +time e2 , inderivative
which  q1of is epositive
1 alongand theP
 performance: a sufficient condition for allowing system Next, trajectories is ė 1 = –q1e1 + e2, in which Q q T is pos-
β1the system is )(1-2) isis equivalently transformed ς r into
constant,
system trajectories (ea 1 )Let = −q ė11e= −q the e virtual
+ e2 , in control
which of q ω1/ρ) isand ≥an
positive e0, = decP
presented as follows of a kind 3of
sign(P(z̄ )− linear
χ), excitation
z̄3 ∈ C1 (z̄ control
3 , λ ),
input is presented, itive
error constant, and
system. β1(e ei1 be=the
1–q e11is1variables
1state the virtual of control
thisςerror r (u12cof ω and2 followi
system
L =  method for the estimation of the region of attraction ω −e βis=
constant, 1 (e 1β ).1 (eβ 1(e) ). = −q e1 is the virtual control of ω and e2 = dec
) −1, otherwise, 2 given.
and ēω ¡
i = [e11, . .1.and , ei ]Tthen , 1i=1∼3. the region of attraction can be estimated by using or
sign(P(z̄
Next, 3
the) − χ),
system z̄
(1-2) 3 ∈ C
is 1 (z̄ 3 , λ
equivalently ), transformedω Step
− β (e
1 into II: ).
1 I:an Then ė = −(q ∗+ r )e − (ω /H)e + d (t), the
L = Step Define 2
e
following 1 = δ − δ2
inequality . The 2 2
time 0
derivative 3of e 1
1
along con
or
|) error
−1, system. Let e otherwise,
be the state variables of virtual
this Step
Step
error control
II:
II.
system Then
Then of ė Pė is
–(q
= chosen
−(q r + as r β )e (ē − ) =(ω P /H)e+ d (H/ω +the d virtual
)(r(t), e the
+
1, z̄3  ≤ τ or L = −1, i system trajectories 2 = e2 is ė2 1
+ = 2
−q2 )e 2
–1 1
e ¡2 +2(ω e
2 02
2
/H
, in )e which
30 m+ 1 (t),
q
1 3 is 0
positive
1 1 1 conR
Proo
=| s =  and ēi = [e1 , . . . , ei ]T , i=1∼3. q2 econtrol
virtual 2 ), where
constant, control ofβP1eq(e of2is1 )is
chosen
P =e positive
is−qchosen as
e βis2constant,
(e as2)β
the =2 (ē
virtual Pm2 )+ rcontrol
=Pēm3q(0)
ρ (H/ω
1= +10D/H
T
)(r
of ω
(H/ω
1≥ e1and−
0. +0q)(r qe2122,e1=2re = linR
),21 +
2,
1, Step otherwise, ≤ τ eor L− 1 1
Q2 r = D/H 2– q , decay o
=-
z̄3I: Define 1=δ =δ−1, ∗ . The time derivative of e along
where q11q(e ,2e1is). qpositive βconstant, r1 = q1D/H ē3 (0)–q
r1 = q1,1 D/H
−1
s = (11)qD/H 2 eω 2 ),− −where
1β 3= 2– Pis e− positive 2 (ē2 ).constant, 2 − q1 , r12 =or not. is
lin
−- system
2, trajectories is ė1 = −q1 e1 + e2 , in which
otherwise, D/H 1e3is−
qStep qPe1III:
=positive ,¡ e3β2Then(e 2P ). − ė2 β=2 (ē
P(z̄ 3 ) − χ + l∆ , L = −1, Step II: = Differentiating
e Proof: −(q 2The er23)e
).2 +objective , 2 one − (ω gets
/H)e3 ė+3design
of0control d= 1 (t), −(q isthe to3 guarantee
+conclud sat
is
ρ χ̇ =  constant, β1 (e1 ) = −q1 e1 is the virtual(11) control (11)of ω  and e2 =
−- 0, (e1χ).+ l∆ , L = −1, otherwise, 1/T Stepvirtual − q control
III:
2 )e 3 + of
u Pe rē
decay
Differentiating
− is of chosen
3 +Lyapunov
d z2
as
, e β32, (ē
where ) =q Pmgets
function
2one 3 isV+epositive
whether
(H/ω ė3 0= 0 ς
)(r e1 +3 + Rema
-saturation
constan-
1−(q happ
the
sat
P(z̄
ω −3β)1− Step
q2 e2 ),III. Differentiating e3, constant,
one gets theeṙ 3equation
= –(q3 + 1/T q=2)
d0 2, ¡
ρ χ̇ = where qor 2 isnot. positive
 Thus, from 1 = q1 D/H (14) −and q2d0 rL2EMMA 2, onee
- 1/T
t, r eD/H = − r q
u1¡theH(q )e –3 +−
re31+ dPz2 u 1/T
− rē
, where 3 )/ω+ d , , r
q0 is positive where
4  constant,
= (H/ω q is )(r positive q
1 r3 = 2r1 H(q21 –2
− 1 constan-
− q r + linear ς -s
the
(z̄Step II: Then ė2 = −(q 2 z2 3
E f d2 = α30, 3 , χ) = α3 (z̄3 )|otherwise, , 2 + r2 )e2 − (ω0 /H)e3 + 3d+
3d0 1 (t),
 0− q1 , e3 =conclude e −d0 β 2 (ēthis 2 ).3 proof.
0
- virtual control of Pe pis3 =χ 2 2− –q23r∗],2) +is ofplo in
0
chosen as β2 (ē2 ) = Pm + t, q(H/ω r
/T
2 d0 –1/T
3 =
0 )(r
),
Step r
d0 1
1 r H(q
e
)/ω =
1 + 0[, r
−r
1 4−
III: Differentiating
3 1/T
=(H/ω
r 4Remark−q d030 )/ω
)(r
], d1 ¡q
1: 0
z1 The,= 2 r¡qd
4 = r
(t),
2
e31¤, main 2(H/ω
+q
one z2d 2 /T
= 0 )(r
d0
differences P ),1(δ
gets 2ė3 =betweenr −= ) q[
− –r
2−(q3Pr4 (δ q
2 3 + SAEC and ς -s
αf3qds saturati
g- EEf fd2 =E 2 ,e32,),χ) = α3q(z̄ d− drr+ d3z2 P P02)d + ωP cotδ (δd))2−
(z̄ is)|positive 2
d = where p3 =χ , constant, r1 = q1 D/H z1 q= ),1δ, −

2 3 ωP
q2 /T e1/Tcot
d0d0
 12=(t),
q=2d [2−r
)e 3(t) + ru=−q
−
linear
4− (qrē (δ)
3H/ω
23excitation
2], d¡
+ z1dz2= ,d
(δcontrol
)(t),
11(t),
where ¤ q¤in
du3z2eis=this +
PP2subsection
3(δ
positive
(t)P¡(δ
sat(E
constan-2 are:
(q∗2) (i)
fd) −
+the SA
plo
theview
for
D/H − q1 , e3 = Pe − β2 (ē2 ). H/ω u P P represents the control
g-
)d (t), = (δ)sat(E ) ¡ (δ )E tio
where the Epositive fd = E constants
f ds , ρ, τ and l ∆ are parameters to be ωP
P 3 (δe t,
∗
cot r ) 0 Eδ=
∗ 1
+ r represents
d
H(q
f d 1 error 2 (t) is
− −3of 1/T (q the
2 H/ω
initiative

)/ω control
fd
,0 )d
r 1
3
control
= variable
(t), (H/ω u
fd
action =)(r of P
from
3
− error
(δq 2)the
−sat(Esystem
q perspective
r f+ d ) to
− SA
ς -satur“a
of
)- Step III: Differentiating e3 , one gets be =∗ 3−(q
variable
ė3 designed  ∗ 3of + 1 system d0 to be 0 designed 4 0
later. 1 2 2 2
tuned χ(0) are P E later.
represents saturation”, the but
control the latter
variable is ofof passive
error system
system analysis
to is
tiofk
whereappropriately by the engineer, .
= pparameters (δ ) d r = [ −r3 r4 −q3 ], dz1 = d1 (t), dz2 = P2 (δ ) − P2 (δ ) +
∗ ploying
-) the positive 1/T constants
d0 − q2 )e ρ,3 + τ uand
− rēl∆3 + dz23, where q3 is be3 q2 /T
to positive d0f),
constan- the viewpoint of “estimation of region of attraction” allow
Als
 l are parameters to bebe Step designed
ωP
2 e cot
Step IV:
IV. δThe +later.
The saturation
dsaturation
2 (t) − (q2on H/ω Eon 0 )d E1f(t),dbecan u be = Pequivalently sat(E f d ) trans-
3 (δ ) transformed − SAEC ck is
tunedwhere the positive
appropriately
T HEOREM 2.t,by constants
3 the
rThere H(q1ρ,−
= rengineer,
1 exist
and
τ 1/T
small χ(0) =0p, 3r.4parameters
positive
∆ = (H/ω0 )(rρ, 1 − qP2 − q r ς -saturation. can
fd(ii) Lyapunov equivalently function V is obtained
u∙c ≤tion.by
)/ω ∗ 2 2 ∗ + ing
Als
H
- tuned appropriately by  the engineer, χ(0) = p3.
d0 formedStep 3 (δ into
into IV:
the dthe
) E f constant
The constant
represents saturation constraint theconstraint
control
on Eucf dvariable on c on
ucan u, be u,ofequivalently
namely,namely,
error juj system|u|3
∙ u≤ctrans- to
τ and l∆Tsuch that, inq2the /Td0presence
), r = [ −r3ofr4input
HEOREM 2. There exist small positive parameters ρ,
−q3 ], d saturation
z1 = d1 (t),and dz2 = time-P2 (δ ) −
U P be P (δ
designed
2
∗) +
P later. ploying
∗ E ∗ .
constructive
Generally
design,
speaking,
and the
for
control
γ
gain
1], if 1 , know
pis p2ing
,p
g formed ∙3 UBm (δ into
P3(δmthe
) − ) ¡constant
(δ P3(δ )E
) ¤
f dfdconstraint
¤
. Generally on u, namely,
uspeaking, fortoγ practical
2 |u| ∈(0, 1],
(0, ≤ uoperating
ifc ≤ −1
varying ωPethe cot δcontroller
+ d2 (t) −(5), (q2 H/ω 1 (t), u global = P3 (δB) sat(E − 3TheSAEC ∗tocan beonspecified c according co
l∆ uncertainties,
τ andTheorem such that,2. There in
P
the exist
∗ E ∗small positive
presence represents
of input the
(11)
parameters achieves
0 )d
saturation
control
ρ,and
variable
τ and time-
of
l∆U γ-saturation
errorB P 3
Stepf)d )−
γ-saturation
(δ
system m
IV:
toP 3happens
happens (δ ∗saturation
tion.) E to
However,
fd
E
. E , there
,
Generally
fd f d there
only
f d can
Emust exist
must be exist
speaking,
the
equivalently
a
structure constantfora
of γ
the ς
constant∈ 2 trans-
(0,
(0,
Lyapunov 1]
1],ς∈ ifAlso, arei
−1
functio
g stabilitysuchof SMIB power (δ )
systems (1-2) and(11) the and transient perfor- formed into the constant Uconstraint ing coe
P3u(δ on)3u, namely, E u∗f dc ,≤cor-
3 |u|∗ )≤
varying that, in the
uncertainties, presencethe f dof input
controller saturation
(5), time-varying
achieves global such
γ-saturation
1] such that thathappens uknown,/ςu∙c /ς P,3(δ U QmPcan
cmust ¤
− fdP
¤
3,a(δ
correspond- ς the is g
are
∗ ) Eto ∗ .EGenerally there
. (0, juj ∙|u|is c≤ f d≤
B
the mB)/γ
matrix ¡ (δexist
/γ be)E got constant
by solving ∈ LMI
mances are improved
uncertainties, the
be by designed
controllerusing later.
the
(5), switching
(11) achieves control
global strategy.
stability Uingly,
B P 3 (δ ) −
ς-saturation
m P 3 (δ happens to u. Based speaking,on the forabove γ ∈ 1],
analysis,
(0, if −1/Td0(
g. stability of SMIB power Stepsystems IV: The(1-2) saturationand the ontransient
E f d can be perfor- respondingly,
(0, 1]
equivalently such
γ-saturation trans- thatςhappens -saturation
|u|Also, ≤ uin f
to
d
c /ς order
E re
happens
≤ UtoB Pdecrease (δ
3 must tomu. ) /γ Based
− aP3constant
control (δon ∗
gain, )the ∗
E fthe above
, cor-are prop
ςd ∈inherent
isdagF
of SMIB power systems (1–2) and the transient performances let u be specified as u = f d–,3,thereand the real exist excitation control
mances
Proof:are In improved
general,formed withby usingappropriate
into the theconstant
switching
choice control
of
constraint τ, the uc strategy.
oninequal-
u, namely, analysis,
respondingly,|u|1]≤such let
uc ≤uthat ςbe specified
-saturation
ing coefficients as
happens uof= Ppower rēto
¤3 ) , u.and Based
¤systems the(δare real
∗on E∗∗fexcitation
the
) reserved above such as −D areF
g are improved by using the switching ∗ ) E control strategy. (0,
input is designed |u|as≤ Efd  uc= /ς[re ≤– U
3 +B P33(δ m
(δ )Efd /γ ]/P− P
3(δ), 3 ∗then, d , cor-
one can is give
ities |E f d1In UB andBwith
| ≤general, U P (δ ≤ τ are
) − P
z̄3 3 mappropriate 3 equivalent.
(δ ∗ . Generally speaking, control
for γ
analysis, ∈ (0,
respondingly, inputlet
1], u
if isbe designed
specified
−1/T
ς -saturation . as
(iii) as E
For
happens u
fd the
== rē
[rē , and
time-varying
to3u.3 Based+ P the(δ
3 on the real
) E
uncertainties,excitation
]/P
f d above 3 (δ adaptive
), tro
are
From
Proof: γ-saturation happens
fd
choice
to E ,
of τ, the
there must
inequal-
exist a
obtain
constant
the following
ς are proposed
error system:
d0
in SAEC, and a disturbance attenuation condi
) In Case
Proof. 1 and Case 2: one can get E = f dE f d1 . then,
control one
analysis, input can ∈
let obtain
isu designed
be the
specified following
as asE u = = error
rē [rē, and system:
+ P
the (δreal
∗ ) E ∗
excitation]/P (δ ), are: are
tro
(i)
f d1 | In≤ general, with appropriate choice of τ, the inequalities ∗ ∗ f d 3 3 3
ities |E UB and (0,–∈ z̄
1] C 3such τ are|u|equivalent.
≤ that f
≤ uz̄c /ς∈
d
≤C UB(z̄P3 (δ m ) /γ −i.e. P3then, )E
(δ control , cancor- isdesigned
give infollowing
the latter scheme.
3 fd
) In
In jE Case
fd1j ∙
Case 1U3, if
B and
and z̄
Case 3kz3k 2:∙one (z̄
1 τ are ,
3can λ ) and
equivalent.
get E f d3=happens / E f d12 . 3 p3 =χ)| one f d input obtainis the
˙3 =
as E f d = error
[rē 3 + system:
P 3 (δ ∗ ) E ∗ ]/P (δ ),
fd 3 trol has
are
inp
respondingly, ς -saturation to u. Based onthen, no-saturation:
the above
one can obtain From ēthe the Aē
followingabove3 + Dd z ,,(12)
discussions,
error system: the advantages (12) ofare thebettSA
≥In
P(z̄In3 ) Case χ:Casethe 1 and
logic Case
judgment 2: one ofcantheget E
auxiliary = E .
system in (11) ma
has
3, if z̄analysis,
3 ∈– C1 (z̄ let3 ,u–λbe) and
specified z̄3– ∈
fd /as Cu2=
fd1 (z̄–rē
3 )|3 ,p3and i.e.
the real no-saturation:
excitation are: (i)ē˙ global Aē Dd
convergence can be ensured when (12) there are
In Case 3, if z 3 2 C1(z 3, λ) and z 3 2 / C (z 3)j p3 ==χ χ Pi.e. ∗ ς -saturation: ē˙ē˙33 = = Aē +
3 + (ς −
Aē33 + Ddand
,
z 1)brē + Dd ,
3 z (13)has ma ada
er
P(z̄3 )P(z ≥– χ: the logic control judgment input is of designed
the auxiliary as E f d2system
= [rē3 + in (11) ∗
3 (δ ) E f d ]/P no-saturation:
3 (δ ), trol input 3= saturation z , time-varying uncertainties. (12) (ii)cht
3) ¸ χ: the logic judgment of the auxiliary system in (11)3 ς -saturation:  are better ˙
ē3 =performances Aē3 + (ς − 1)brē  that 3 + of Ddlinear z ,,(13) (13) er main to
then, one can obtain the following error system: ς -saturation: ˙
1 ē3 = Aē3 +0(ς − 1)brē3 + Ddz , than controller
(13) er costs sin
can activate the dynamic system ρχ̇ = P(z–3) ¡ χ + l∆. As such, −q1
 0 has −(qadaptive , b T tim
to
where A =  21+r2 ) mechanism
−ω0 /H
0 and auxiliary
= [ 0 0 decision
1 ] , D= system.
to obtai
since ρ is small, the variable no-saturation: χ can reach ē˙ 3 =a small
Aē3 +neighborhood
Ddz , 3 −q
(12) −q1 01 −(q +1/T1
0  −q ) und
tim
of the manifold χ = P(z–3) + l∆, then P(z ˙
–
3)Aē< χ and the condi-where where AA = = 0 main −(q2characteristics
0 −(q
+r2 )
2 +r2 )
3 −ω0 /H
−ω0 /H d0 of the 2 linear
, ,bonlyb==[controller[ 0 0T1 ]T ,are:
0 0 1] , D =
D (i) it has
=time-va
ς -saturation: ē 3 = 3 + (ς − 1)brē 3 + Dd z , (13) 00 er costs 00 than that of 
SAEC, off-line analysis und
undernee
is s
tion z–3 2 C2(z–3)j p3=χ holds. −(q  −q)2 )
−(q33 +1/T +1/Td0 d0−q 2
  4 to obtain the region of attraction considering ς -saturation
Consequently, on the basis−q of1 Theorem 1 1 and 0Lemma 4, the  0 T
]T , D = 00 10time-varying d uncertainties. (ii) it is suitable for power syst
T
convergence can z–activate
of where 3 to the Athe= dynamic
origin 0 −(q2system
can be+rensured. 2) ρ−ωχ̇ = 0 /H P(z̄3 )−, χb+l =4∆[.0As 40 1such, 1 , dz = [ z1 z2 ] .
d

under small signal disturbances.
Therefore, the ρmain
since is small, the0 variable
advantages χ−(q
can reach
of0 sub-controllers
3 +1/T a2Esmall
d0 −q ) neighborhood
fd1 and Step V: Choose Lyapunov function as Ve = (1/2)ēT3 Qē3
of the
Efd2 are realized in Cases 1, 2χand
manifold = P(z̄
Case ) < χthe
l∆ , then P(z̄3Thus,
3 )3+respectively. and the condi- where 0 < Q = QT ∈ R3×3 are to be chosen later. Thus th
tion 4 z̄ C
switching strategy is proposedp3to
∈ (z̄ )| holds.
=χmake logical choice between Step V. Choose
timeLyapunov
derivativefunction as Vthe
of Ve along (1/2)e–T
Bull.–Pol. Ac.: Tech. XX(Y)
Qe3, whereis
3 2 3 e =system 3trajectories
Consequently, on the basis of T 1 and L 4, T 3×3
them based on the power systems operating conditions. As such, 0 < Q = Q 2 ℝ are
HEOREM EMMA to be chosen later. Thus the time deriv-
□□ can be ensured. ative of Ve along
T
the transient the convergence
performances z̄3 to the origin
areofimproved. V̇e =ē Ak T Qētrajectories
the3system
T
3 + ē3 QDdzis
Therefore, the main advantages of sub-controllers E f d1 and =ēT3 [Ak T Q + (1/(2ε 2 ))QDDT Q]ē3 (14
E f d2 are realized in Cases 1,2 and Case 3 respectively. Thus, T 2 2 2
− (1/2)εdz − (1/ε)D Qē3  + (1/2)ε dz  ,
6 the switching strategy is proposed to make logical choice be- Bull. Pol. Ac.: Tech. 65(1) 2017
tween them based on the power systems operating conditions.Brought where
to you A whenUniversity
Akby=| Gdansk saturation, and Ak = Akς =
there isofnoTechnology
As such, the transient performances are improved.  A + (ς − 1)br when there is ςAuthenticated
-saturation, ε is disturbance at
Download Date | 3/2/17 11:05 AM
tenuation constant.
3.3. Some Discussions In the above subsection, the SAEC
   0 1TT0 T z = [ dz1 dTz2 T
,,ch, 0000 1010001010 1,, ddz,z d = [[ddz1z1 ddz2z2]]T..] .
=
dood Step V: Choose Lyapunov function as V= T Qē ,
d Step
Step V:
V: Choose
Choose Lyapunov
Lyapunov function
function as
as VVee = =
e (1/2)ē(1/2)ē
(1/2)ē TTQē
Qē3 33,, 3
ndi- where 0 Q Q T ∈ Nonlinear 3×3 arerobust to becontrol
chosen for single-machine
later. 33
Thus infinite-bus power systems with input saturation
-- where where 00 < <Q <
Q= =Q =Q ∈ T
T ∈R R R
3×3
3×3 are to
are to be be chosen
chosen later. later. Thus Thus the thethe
timetime
time derivative
derivative
derivative of V
of of
Vee V e along
along
along thethe
the system
system
system trajectories
trajectories
trajectories is is
is
A 4,
,, T T T Lyapunov function V3 is obtained by employing constructive
TT 3 A
V̇V̇ee V̇ e =ē
=ē
=ē 33A AkkTTQē QēQē
k
33 +
TT ē3 QDdz
+3ēē+
33QDdQDdzz design, and the control gain p1, p2, p3 of SAEC can be specified
ddand
T T 2 T (14)(14)according to practical operating condition. However, only the
=ē
TT 3 [A TT k Q + (1/(2ε 22 ))QDD TT Q]ē3
=ē33[A
=ē [Akk Q Q+ +(1/(2ε
(1/(2ε ))QDD ))QDD Q]ē Q]ē33 (14)
(14)
,,us, − (1/2)εd z − (1/ε)D T
TT Qē223  + (1/2)ε
2 2
22 d22z  ,
2 structure of the Lyapunov function Ve is known, the matrix Q
--be- −− (1/2)εd
(1/2)εd z
z −− (1/ε)D
(1/ε)D Qē
Qē 3
3  ++ (1/2)ε
(1/2)ε d
d zz , , can be got by solving the LMI (15). Also, in order to decrease
ns. control gain, the inherent damping coefficients of power sys-
.. where where
where AAkkA=
where =kA
A
= A
A=when
when
A
when
when there
there
there
there is
isis
is no saturation,
nosaturation,
no
no saturation,
saturation, and and
andA
and
=AAkA A=
k=
kA
=
=
Akς A
Akς
kς+ == =tems are reserved such as ¡D/H, ¡1/Td0 0
. (iii) For the time-
AA+ A + (ς −
1)br
k
1)br when when there there is ς -saturation,
isis ςςς-saturation,
-saturation,ε εis ε is ε
is
k
is
disturbance
kς
disturbance at-
at- varying uncertainties, adaptive laws are proposed in SAEC, and
+(ς (ς+− (ς 1)br
− ¡ 1)br whenwhenthere thereis -saturation, disturbance
disturbance atten- at-
tenuation
tenuation
tenuation constant.
constant.
uationconstant.
constant. a disturbance attenuation condition is give in the latter scheme.
C
C EC From the above discussion, the advantages of the SAEC are:
T HEOREM 3.ForFor thethe error dynamic system (12-13) andand(i) global convergence can be ensured when there are control
--ys- TTHEOREM
Theorem HEOREM 3. For 3. the
3. For the
error error
error
dynamic dynamic
dynamic systemsystemsystem
(12–13) (12-13)
(12-13)
and given and
me-
-- given
given
given ςςς2∈ ∈ς(0, ∈1],
(0,
(0, (0,
1],
1], ρ 1],
ρ>
ρ> ρ 0,
0,
> >
0<
0, 00,
0< ε<0< ε<
ε< 1,
<ε the
< 1,
1, the
1, the
control
the control
control
control input uinput
input
input = ureu–= rē33rēen-
u rē
=
3 ensures
= 3 en-input saturation and time-varying uncertainties. (ii) better per-
en-
ys- sures convergence –of ē3 to a ball of radius ε under ς -saturation
-- sures suresconvergence
convergenceofof
convergence ofeē3ē33toto toaaaball ballofof
ball ofradius
radiusε εεunder
radius under
under ς-saturation
ςς-saturation
-saturation if formance is obtained than that of linear controller since it has
of if there
there exist
exist
exist Q QQ satisfying
satisfying
Q satisfying adaptive mechanism and auxiliary decision system. The main
ff ifif there
there exist satisfying
trol characteristics of the linear controller are: (i) lower costs than
ll T
TTA Q + (1/(2ε
2
22 ))QDD
T
TT Q < 0, that of SAEC, only off-line analysis is needed to obtain the
ss his AA QQ ++ (1/(2ε
(1/(2ε ))QDD
))QDD QQ << 0,
0,
ATTkςQ+ T
Q(1/(2ε 2 T
TT Q < 0,
region of attraction considering ς-saturation and time-varying
ddand AAkςkς Q + + (1/(2ε
(1/(2ε 22 ))QDD
))QDD
))QDD  QQ< < 0, (15)
0, (15)
(15) (15)uncertainties. (ii) it is suitable for power systems under small
nnion  Q TTς rT signal disturbances.
0, 0,
QQ ςςrr 2 ≥
aa d a ς r22(uc /ρ)≥ ≥ 0,
ςςrr (u(ucc/ρ)
/ρ)
n.
andand
and then
then then thethe
the region
region
region of of
of attraction attraction
attraction can
can can bebebe
be estimated
estimated
estimated
estimated by by byusing
by using
using thethe4. Simulation results
the
nnan following and then the region
inequality of attraction can using the
following
following following inequality
inequality
inequality
mem
m
  ρ ē3TT(0) T  In this section, the developed control strategy for a SMIB
ρρ ēē33(0)
ē
(0) −1
Q ≥≥ ≥
0.
0. . 0.
(16) (16)
(16) (16)power system is shown to be effective by a comparative simu-
gong
3 (0) −1
−1
g ēē33(0)
(0) Q Q
ive lation study of the SAEC and the excitation controller which is
ee Proof:
Proof: Proof: The
The The objective
objective
objective of
of of control
control
control design
design design is
is is
to
to to guarantee
guarantee
guarantee the
the thedesigned by using the approach proposed in [17]. The transient
=2 = decay
= decay
decay of of
Proof.
of Lyapunov
The
Lyapunov
Lyapunov objective function
function
function of control VVee Vwhether
e whether
design
whether ς -saturation
isςςto-saturation
guarantee
-saturation the happensresponses of power angle, relative speed, active power and con-
decay
happens
happens
of Lyapunov function e whether happens2,orone
the or or
or not.
not. not. Thus,
Thus, Thus, from
from from thethe
the
Vequation
equation
equation (14)
(14)
ς-saturation
(14) andand
and LLEMMAL EMMA
EMMA 2, 2, oneonenot. cancantrol input are shown in Figs 2–5 with the same initial condition
can
e
e conclude conclude Thus, from
this equation
proof. (14) and Lemma 2, one can conclude this δ(0) = 1.309 (75°), ω(0) = 5, Pe(0) = 0.7 and also with the same
+ conclude proof.
this
this □□ proof. 
proof.  limit on the control input Efd. The values of the parameters and
1 Remark
+
+
2 = Remark
Remark 1: 1:The
1: The The main
main main differences
differences
differences between
between
between SAEC
SAEC SAEC andand
and thetheoperating point used in the simulation are respectively listed as
the
= linear
= linear
linear excitation
excitation
excitation control
control
control in this
in in
thisthis subsection
subsection
subsection are:
are: are: (i)
(i) (i)
the the former
former
Remark 1. The main differences between SAEC andthe theformer
linear follows: 1) Values of system parameters: ω0 = 314.159, D = 5,
is
is of
is of
of excitation initiative
initiative
initiative control control
control
control in this action
action action
subsection
from
from from are:
the
the the (i) perspective
the formerof
perspective
perspective is of
of of “anti-H = 7.1, V = 1, x = 1.79, x0 = 0.17, x = 0.02, x = 0.8,
ini-
“anti-
“anti-
+ s d d0 T L
+ saturation”,
+ saturation”,
saturation”,tiative control but
but butthe
the the
actionlatter
latter latter
from is
is is
the
of
of of passive
perspective
passive
passive system
of
system
system analysis
“anti-saturation’’,
analysis
analysis from
from fromx0ds = x0d + xT + xL = 0.99, Td0 = 7.65, Td0 = 5, UB = 6;
--an- thethethe viewpoint
but
viewpoint
viewpoint the latter of of
of is of“estimation
passive system
“estimation
“estimation of of
of region
region region
analysisof of
of attraction”
from
attraction”
attraction” the viewpoint allowing2) Operating point: (δ, ω, Pe) = (0.3475, 0, 0.81); 3) Values
allowing
allowing
+ + ς -saturation.
+2 ςς-saturation. of “estimation
-saturation. (ii)(ii)
(ii) of Lyapunov
region
Lyapunov
Lyapunov function
of attraction’’
function
function VV33 Vis
allowing is
3 obtained
is obtained
ς-saturation.
obtained byby
by (ii)em-of designed parameters: c1 = c2 = c3 = 1, p1 = p2 = p3 = 0.5,
em-
em-
) + ploying
ploying constructive constructive
constructive design, design,
design, and and
and the the
the control control
control gain gain pp11,, pp22,, pp33 pof
gain p 1 , p 2 , of3 of
+ ploying
+
) − SAEC
SAEC can can
can be be specified
be specified
specified according according
according to to
to practicalpractical
practical operating operating
operating condi- condi-
condi-
− SAEC
−
tion. However, only thethe 2.2structure of
thethe Lyapunov function VVee Ve
oo to tion.tion. However,
However, only
only the structure
structure ofof the Lyapunov
Lyapunov function
function 6
is
is known,
is known,
known, the the
the matrix matrix
matrix Q Q can Q
can can be
be gotgot bygot by
by solving solving
solving the the
the LMI LMI
LMI (15). (15). (15).
2 be
ns- Also,
Also, in in
in order order
order to to
to decrease decrease
decrease control
control gain, gain,
gain, the the
the inherent inherent
inherent damp- damp-damp-
-- Also, 1.8 control
4
≤ ing coefficients
ing coefficients
coefficients of
of power power
power systems systems
systems are are
are reserved reserved
reserved such such
such as as
as −D/H, −D/H,
≤c ing
≤  of 1.6
−D/H,
, if −1/T  . (iii) For the time-varying uncertainties, adaptive laws 2
ff −1/T d0.. (iii)
(iii) ForFor the the time-varying
time-varying uncertainties, uncertainties, adaptive adaptive laws laws
−1/Td0 d0
ς ∈ are proposed in SAEC, and
1.4 a disturbance attenuation condition
ω(rad/s)

are proposed
proposed in in SAEC,
SAEC, and and aa disturbancedisturbance attenuation attenuation condition condition
δ(rad)

∈ are
∈
is give in
thethe latter scheme.
--or- is is give
give in the
in latter
latter scheme.
scheme. 1.2 0

ove From
ee From
From thethe
the above
aboveabove discussions,
discussions,
discussions, 1 thethe
the advantages
advantages
advantages of of
of thethe
the SAEC
SAEC SAEC
ion
nn are: are: (i) global convergence can be ensured when there are con- −2
are: (i) (i) global
global convergence
convergence 0.8cancan be be ensured
ensured when when there there are are con- con-
δ ),
,, trol trol input saturation and time-varying uncertainties. (ii) there
trol input
input saturation
saturation and and time-varying
time-varying uncertainties.
0.6 uncertainties. (ii) (ii) therethere
areare better performances than that of linear controller since itit it
−4
are better
better performances
performances than
than
0.4
that
that of
of linear
linear controller
controller since
since
hashas
has adaptiveadaptive
adaptive mechanism
mechanism
mechanism andand
and auxiliary
auxiliary
auxiliary decision
decision
decision system.The
system.
system. The The
)12)
) main characteristics of the
0.2 linear controller are: (i) it has low- −6
main characteristics
main characteristics of of the the linear linear
0 controller
controller
1 2 are:are: (i) 3(i) itit has
has 4 low-low-5 6 7 8 9 10 0 1
13)
)) er er costs than that of SAEC, only off-line analysis is needed t(s)
er costs
costs than than that that of of SAEC,
SAEC, only only off-line
off-line analysis analysis is is needed
needed
to to
to obtain
obtain obtain thethe
the region
region
region of of
of attraction
attraction
attraction considering
considering
considering ςς ς -saturation
-saturation
-saturation andand
and
time-varying uncertainties. (ii) it is Fig. 2.for
suitable Transient
power response
systems curves of the power angle
D time-varying uncertainties.
= = time-varying
= uncertainties. (ii) (ii) itit isis suitable
suitable for for power
power systems systems
under
under
under small
small small signal
signal
signal disturbances.
disturbances.
disturbances. 2 5
Bull. Pol. Ac.: Tech. 65(1) 2017 47
Bull. Pol. Ac.: Tech. XX(Y) 2016
Bull. Pol.
Bull. Pol. Ac.:
Ac.: Tech.
Tech. XX(Y)
XX(Y) 2016
2016 Brought to you by | Gdansk University of Technology
1.5 3
Authenticated
Download Date | 3/2/17 11:05 AM 2
)
 Y. Wan

2.2 6
4
2
4
1.8
2
ω(rad/s)
1.6
2
0
1.4

ω(rad/s)
δ(rad)

1.2 0
−2
1
−2
0.8
−4
0.6
−4
−6
0.4
9 10 0 1 2 3 4 5 6 7 8 9 10
0.2 t(s) −6
0 1 2 3 4 5 6 7 8 9 10 0 1

Fig. 3. Transient responset(s)

curves of the relative speed

4
2 5

3 4
6
2
1.5 3
4
1
Efd (p.u.)

0
1 1

Efd (p.u.)
Pe (p.u.)

2
−1 0
ω(rad/s)

−2
0.5 −1
0
−3 −2
−2
−4
0 −3

−5 −4
9 10 −40 1 2 3 4 5 6 7 8 9 10
−0.5 −5
0 1 2 3 4 t(s)
5 6 7 8 9 10 0 1
−6 t(s)
9 10 0 1 2 3 4 5 6 7 8 9 10
t(s)
Fig. 4. Transient response curves of the active power

1
Efd (p.u.)

−1

−2

−3

−4

−5
9 10 0 1 2 3 4 5 6 7 8 9 10

t(s)
Fig. 2. Transient response curves of the power angle

8 Bull. Pol. Ac.: Tech. 65(1) 2017

Brought to you by | Gdansk University of Technology
Authenticated
Download Date | 3/2/17 11:05 AM
 Nonlinear robust control for single-machine infinite-bus power systems with input saturation

ε1 = ε2 = 0.05, ρ = 0.5, τ = 2, l∆ = 0.1; 4) Time-varying uncer- [4] S. Mei, Y. Ni, G. Wang, and S. Wu, “A study of self-organized
tainties: d1(t) = 2sin(ωt), d2(t) = 0.03sin(0.01ωt). criticality of power system under cascading failures based on
In Figs 2–5, the solid lines and the dash lines represent ac-opf with voltage stability margin,” IEEE Transactions on
the results with the proposed SAEC and the controller in [17] Power Systems, 23 (4), 1719–1726 (2008).
respectively. From these figures, the SAEC globally stabilizes [5] P. Pruski and S. Paszek. “Assessment of Polish Power System
the system (1–2) and provides better performance than the angular stability based on analysis of different disturbance wave-
forms,” Bull. Pol. Ac.: Tech 63 (2), 435–441 (2015).
controller given in [17], which does not consider the influ-
[6] A. Isidori, Nonlinear Control Systems, Springer Science&Busi-
ence of input saturation. The simulation results verify the pro-
ness Media, 1995.
posed theoretical results i.e. Theorem 1 and Theorem 2, and [7] S. Mei, T. Shen, and K. Liu, Modern Robust Control Theory and
the superiority of the proposed control method over existing Application, Beijing: Tsinghua, 2003.
methodologies that ignore input saturation within the design [8] L. Y. Sun, S. Tong, and Y. Liu, “Adaptive backstepping sliding
is demonstrated. mode control of static var compensator,” IEEE Transactions on
Control Systems Technology 19 (5), 1178–1185 (2011).
[9] Y. Li, T. Li, and X. Jing, “Indirect adaptive fuzzy control for
5. Conclusions input and output constrained nonlinear systems using a barrier
lyapunov function,” International Journal of Adaptive Control
A new control synthesis framework was developed to design and Signal Processing 28 (2), 184–199 (2014).
a kind of nonlinear robust generator excitation controller [10] M. Chen, G. Tao, and B. Jiang, “Dynamic surface control using
for power systems. Compared with the conventional control neural networks for a alass of uncertain nonlinear systems with
schemes, by employing adaptive auxiliary decision mechanism, input saturation,” IEEE Transactions on Neural Networks and
global convergence and better transient performances have been Learning Systems 26 (9), 2086–2097 (2014).
[11] B. Song and J. K. Hedrick, “Simultaneous quadratic stabilization
achieved theoretically and simulatively even with control input
for a class of non-linear systems with input saturation using dy-
saturation and time-varying uncertainties.
namic surface control,” International Journal of Control 77 (1),
The practicability of the proposed control law is discussed 19–26 (2004).
as follows: firstly, as a model-based approach, it is important to [12] S. Boyd, L. El Ghaoui, E. Feron, and V. Balakrishnan, Linear
verify the accuracy of the model used in this paper. Fortunately, Matrix Inequalities in System and Control Theory, Philadelphia,
as stated in Introduction, the SMIB model has been validated to SIAM, 1994.
be well-suited for power system analysis and control. Moreover, [13] W. Dib, A. E. Barabanov, R. Ortega, and F. Lamnabhi- Lagar-
time-varying uncertainties have been considered, which has in- rigue, “An explicit solution of the power balance equations of
creased the exactness of the classical SMIB model. Secondly, structure preserving power system models,” IEEE Transactions
only local measurable information is needed for the designed on Power Systems 24 (2), 759–765 (2009).
control law in which the parameters can be tuned flexibly in [14] W. Dib, R. Ortega, A. Barabanov, and F. Lamnabhi-Lagarrigue,
terms of practical operating conditions and engineering require- “A globally convergent controller for multi-machine power sys-
ments. tems using structure-preserving models,” IEEE Transactions on
Automatic Control 54 (9), 2179–2185 (2009).
[15] M. A. Arjona, R. Escarela-Perez, G. Espinosa-Perez, and J. Al-
Acknowledgements. This work was supported by the Na-
varez-Ramirez, “Validity testing of third-order nonlinear models
tional Natural Science Foundation of China under Grants
for synchronous generators,” Electric Power Systems Research
61403194, the Natural Science Foundation of Jiangsu Prov- 79 (6), 953–958 (2009).
ince under Grant BK20140836, the Foundation of Jiangsu Key [16] P. V. Kokotovic and P.W. Sauer, “Integral manifold as a tool for
Laboratory “Internet of Things and Control Technologies” and reduced-order modeling of nonlinear systems: A synchronous
NUAA Research Funding for Introduction of Talent under Grant machine case study,” IEEE Transactions on Circuits and Sys-
1003‒56YAH14007. tems 36 (3), 403–410 (1989).
[17] T. Shen, S. Mei, Q. Lu, W. Hu, and K. Tamura, “Adaptive non-
linear excitation control with l2 disturbance attenuation for
References power systems,” Automatica 39 (1), 81–89 (2003).
[1] P. M. Anderson and A. A. Fouad, Power System Control and [18] Y.Wan and J. Zhao, “Extended backstepping method for sin-
Stability, Second Edition, Wiley-IEEE Press, 2002. glemachine infinite-bus power systems with smes,” IEEE Trans-
[2] G. M. Dimirovski, Y.W. Jing,W. L. Li, and X. P. Liu, “Adaptive actions on Control Systems Technology 21 (3), 915–923 (2013).
back-stepping design of tcsc robust nonlinear control for power [19] G. Gurrala and I. Sen, “Power system stabilizers design for
systems,” Intelligent Automation & Soft Computing 12 (1), 75– 87 interconnected power systems,” IEEE Transactions on Power
(2006). Systems 25 (2), 1042–1051 (2010).
[3] M. Galaz, R. Ortega, A. S. Bazanella, and A.M. Stankovic, “An [20] X. Jin, “Adaptive fault tolerant control for a class of input and
energy-shaping approach to the design of excitation control of state constrained mimo nonlinear systems,” International Journal
synchronous generators,” Automatica 39 (1), 111–119 (2003). of Robust and Nonlinear Control 26 (2), 286–302 (2016).

Bull. Pol. Ac.: Tech. 65(1) 2017 9

Brought to you by | Gdansk University of Technology
Authenticated
Download Date | 3/2/17 11:05 AM