TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 13, No. 1, Janua
ry 201
5, pp. 33 ~ 4
1
DOI: 10.115
9
1
/telkomni
ka.
v
13i1.683
5
33
Re
cei
v
ed O
c
t
ober 1
2
, 201
4; Revi
se
d Novem
b
e
r
10, 2014; Accept
ed No
vem
b
e
r
29, 2014
Two Re
gional Power System PSO PID Control Research
Jingfan
g Wa
ng
Schoo
l of Information Sci
enc
e & Engin
eer
in
g, Huna
n Inter
natio
nal Ec
ono
mics Univers
i
t
y
,
Chan
gsh
a
, Ch
ina, postco
de:4
102
05
email: matl
ab_
b
y
s
j
@1
26.com
A
b
st
r
a
ct
In this p
a
p
e
r, w
e
prop
ose
a PID p
a
ra
me
ter
tuni
ng
of
particl
e sw
arm opti
m
i
z
at
io
n f
o
r
multi-
objective optimi
z
a
t
i
on char
acterist
ics of two regional powe
r system
PID c
ontroller design. By defining
a
compre
hens
ive
consid
erati
on
of system
o
u
tp
ut oversh
oot, rise time an
d th
e fitness functi
on ter
m
steady
-
state error i
ndi
cators, such a
s
the IT
AE, and i
n
a
ccor
d
a
n
ce w
i
th the p
e
rformanc
e re
quir
e
ments of
th
e
actual control
system
, appr
opriate weighting of
each index it
em
. Us
e
with base and im
prov
ed
part
i
cle
sw
arm alg
o
rith
m for multi-o
b
j
e
ctive opti
m
i
z
a
t
ion PID.
PSO
opti
m
i
z
at
ion a
l
gorith
m
h
a
s go
od gl
oba
l sear
ch
abil
i
ty a
n
d
hi
g
h
co
nver
genc
e
rate. Y
ou c
a
n
resp
on
d to
th
e PID c
ontro
lle
r tuni
ng
par
a
m
eters d
i
rectly
fro
m
the outp
u
t of the PID co
ntrol
syst
em e
a
sily
bal
ance
betw
een s
pee
d a
n
d
stabi
lity cont
rol syste
m
s. After
using PSO-PID control in
two regional pow
er system
，
the a
m
ou
nt of oversho
o
t in the
step total resp
on
s
e
dow
ns from 6
5
%
to 5%. Adjus
t
time
dow
ns from 1
2
seco
nds
to 3 seconds.
Ke
y
w
ords
: two regional pow
er system
, PID,
particle swarm
optimi
z
ation
(
PSO ), power
system
control
Copy
right
©
2015 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
The a
c
tive
ch
ange
s m
a
inly
impa
ct the f
r
eque
nc
y of
system
s (volt
age pha
se
),
reactive
cha
nge
s
hav
e little effect
on the
sy
ste
m
freq
uen
cy
,
whi
c
h m
a
inly
affect the volt
age
amplitud
e of
the system.
The active
and re
active
powe
r
co
ntrol ca
n be
con
s
id
ere
d
separately. Load
freque
ncy
co
ntrol
(LF
C
-l
o
ad frequ
en
cy co
ntrol
)
co
n
t
rols frequ
e
n
cy
an
d a
c
tive, automat
ic
voltage regul
ation (AV
R
-a
utomatic volt
age
reg
u
lator) a
d
ju
sts th
e re
active
po
wer an
d volt
age
amplitude. L
F
C ha
s be
en
applied to la
rge inte
rc
onn
ected p
o
wer
system. To
d
a
y, it is still the
basi
s
for ma
n
y
new co
ntrol
princi
ple.
Automatic G
eneration
Co
ntrol (A
GC-a
utomatic
gen
eration
control) pl
ay a rol
e
in the
operation of t
he po
we
r
system, whi
c
h
control
s
t
he
co
ntact line
acti
ve of intercon
necte
d sy
ste
m
.
[1]. The pu
rp
ose
of contro
l is when
the
system
volt
age a
nd freq
uen
cy are wi
thin the no
rmal
range, to ensure the
system as
much as possi
ble to achi
eve be
tter economy and reliability [2-4].
In the autom
atic ge
neration sy
stem,
whe
n
the lo
ad sudd
enly
increa
se
s,
before
the
spe
ed
contro
l system
ch
ange
s into th
e steam
of the ste
a
m en
gine, the turbine
spee
d h
a
s
drop
ped.
Flywhe
el d
e
tect
ed freque
ncy
error sig
nal i
s
sm
all, turbi
n
e
spe
ed
ca
n
be mai
n
taine
d
in
su
ch a
stea
d
y
state, it is c
onsta
nt, the speed i
s
lo
we
r than the
rate
d sp
eed, the
r
e is a
freq
uen
cy
deviation (ie,
freque
ncy a
d
justme
nt is
govern
o
r d
r
o
op). The
r
e i
s
a way to accumul
a
te such a
freque
ncy de
viation, or it is the freq
uen
cy deviati
on i
n
tegral u
n
it with integral frequ
en
cy offset
monitori
ng p
e
r
iod, the freq
uen
cy is b
a
ck to the nom
i
n
al valueb
ase
d
on the i
n
te
gral valu
e. When
the sy
stem l
o
ad
cha
nge
s
continuo
usly, t
he frequ
ency
of the
ge
nerator i
s
adju
s
t
ed to
the
rati
ng,
whi
c
h is calle
d the Automa
tic Gene
ratio
n
Control
(A
GC). In the interco
nne
cted
system, whi
c
h is
comp
osed
of several regio
n
s, AG
C di
stributes
load
b
e
twee
n regio
n
s, bet
wee
n
t
he po
we
r pl
a
n
t
and betwee
n
the gene
rat
o
rs i
n
ord
e
r t
o
achi
eve
ma
ximum eco
n
o
m
ic be
nefits. It also co
ntrol
s
the tie-line p
o
w
er
at the pla
nned valu
e, and to ens
ure the system'
s
freque
ncy rating. Of cou
r
se
,
the system m
u
st be sta
b
le
prem
i
s
e. Un
der the
ca
se
of large
di
stu
r
ban
ce
s an
d
accide
nts, AGC
will exit, and the accid
ent control p
r
og
ra
m is used.
In today's
society, acco
mpanie
d
wit
h
the contin
uou
s imp
r
ov
ement of
sci
ence an
d
indu
stry, it become
s
mo
re i
m
porta
nt for the re
q
u
ire
m
e
n
ts of techni
cal. Consequ
e
n
tly people pa
y
more a
nd mo
re co
ncern o
n
the increa
se of advanc
e
d
techn
o
logy for the past years. In the field
of indu
strial,
difficulty in controlling, t
a
rget
a
nd a
risin
g
comp
lexity provide
a pu
sh to t
he
indu
strial con
t
rol techn
o
log
y
just as the new ex
pe
ctat
ions. Previou
s
experi
e
n
c
e su
ch as Z
N
law
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 1, Janua
ry 2015 : 33 – 4
1
34
[5] can no lo
nger m
eet th
e latest control req
u
irem
e
n
ts, and in th
e geneti
c
alg
o
rithm ha
s b
een
put forward for
several de
cad
e
s, the
accuracy
and
e
fficiency h
a
ve
been
not a
b
le to satisfy the
need. A varie
t
y of improve
d
algorithm
s
based on ge
netic algo
rith
ms[6-9] were
broug
ht forward
freque
ntly, bu
t none
of the
m
co
uld
be g
enerally ap
p
licabl
e. Particl
e
swa
r
m o
p
timization
(PS
O
) i
s
an evol
ution
a
ry
comp
utation te
chni
qu
e, put fo
rward by
Dr. Eb
erha
rt a
nd
Dr. Kenn
edy[1
0].
Imitating from
the b
ehavio
r
of bird
s, PSO
algo
ri
thm su
ppo
se a
g
r
ou
p
of
bi
rd
s ran
dom sea
r
ch
f
o
r
food, whi
c
h i
s
the si
ngle
or several
be
st. In the part
i
cle si
mulatio
n
of bird
s se
arching fo
r fo
od,
according to t
he individu
al'
s
lo
cation, sp
eed, fi
tness a
nd othe
r indiv
i
dual vari
able
s
we ca
n obt
ain
the two
pola
r
value-parti
cl
e optimum
a
nd glo
bal
o
p
timum. The P
S
O algo
rithm
has bee
n p
r
oven
to be evoluti
onary
optimization alg
o
rith
m with a
hug
e
potential. I
n
this a
r
ticl
e, PID co
ntroll
er i
s
desi
gne
d ba
sed on
PSO
algorith
m
, it is u
s
ed
to
two re
gion
al po
wer sy
stem,
the expe
rime
nt
sho
w
s its adv
antage
s thro
u
gh the simul
a
tion results.
2. Particle Sw
arm Optimi
z
a
tion PID Design
2.1. PID Summar
y
Descri
p
tion
In actual in
d
u
strial
pro
c
e
ss, the
r
e
will
be
an e
r
ror of variable,
it is the re
gu
lator for
controlling a
c
cording to the prop
ortion
a
l
, integral
an
d differential of the deviation, it is known
PID reg
u
lato
r (al
s
o
kno
w
n
as PI
D cont
rolle
r), the
sy
stem di
agra
m
is
sho
w
n
i
n
Figu
re
1.
PID
control
algo
rithm i
s
simple
and
e
a
sy to
achi
eve, the
r
e a
r
e
goo
d
cont
rol
pe
rformance
and
h
i
gh
stability, it is
widely
used i
n
indust
rial
process
control. The current industri
al
control l
oop
are
using PID control thought at 90
percent. The
actual
process will inevitabl
y exist
nonlinear,
uncertaintie
s
and
other co
mplicatin
g fa
ctors, th
ere
wil
l
be th
e follo
wing
difficulti
e
s: it i
s
difficu
lt to
establi
s
h the
pre
c
ise co
ntrol of
the syst
em model; tu
ning metho
d
is too dep
end
ent on traditio
nal
para
m
eters,
PID co
ntrolle
r paramete
r
tu
ning oft
en
be
poo
r, poo
r p
e
rform
a
n
c
e,
poor ada
ptab
ility
to the o
p
e
r
ati
ng e
n
viron
m
e
n
t. In re
sp
on
se to the
s
e
p
r
oblem
s, for a
long time,
pe
ople
have
be
en
improve
d
PID
controlle
r p
a
ram
e
ters of
self-tuni
ng te
chn
o
logy to
meet the
co
n
t
rol requi
rem
ents
of complex condition
s an
d
high indi
cato
rs.
Figure 1. Con
t
rol System Block Dia
g
ra
m
Whe
n
the system is conti
nuou
sly cont
rolled,
the rel
a
tionship bet
wee
n
the pro
portion
al,
integral, diffe
rential will exi
s
t in Fo
rmula
(1) f
r
om
the
input e (t
) to
the output
u (t) of the PI
D
controlle
r.
0
1(
)
()
[
(
)
(
)
]
pd
i
de
t
ut
K
e
t
e
t
d
t
T
Td
t
(
1
)
e (t) = r (t
) -y
(t), Kp is the prop
ortio
nal g
a
in, Ti
is
the integral time,
Td is
the derivative time.
In different i
n
dustri
a
l p
r
o
c
e
s
ses,
co
ntrol
pur
p
o
ses will
be different, therefo
r
e,
wh
en PID
controlle
r pa
rameters a
r
e t
uned, pe
rformance indi
ca
tors a
r
e often
cho
s
en
acco
rding to
spe
c
i
f
ic
requi
rem
ents.
There are
two commo
n
perfo
rman
ce index, whi
c
h i
s
ba
sed
on individu
al
perfo
rman
ce
indicators of the clo
s
ed
-lo
op re
sp
o
n
se
characte
ri
stics, such as
the attenuati
o
n
ratio, the maximum dynami
c
deviation, the adju
s
ti
ng t
i
me or the o
s
cillation p
e
rio
d
, and the other
one is the e
r
ror pe
rform
a
n
c
e from the
start time
poin
t
to a stable time until the entire respon
se
curve
sha
pe. Individual pe
rforman
c
e in
di
cators a
r
e int
u
itive, simple
and cle
a
r m
eanin
g
, but it is
difficult to accurately mea
s
ure, error p
e
rform
a
n
c
e i
s
more accu
rate comp
are
d
to use yet more
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Two Regio
nal
Power S
yste
m
PSO
PID
Control Re
se
arch (Jingfa
n
g Wan
g
)
35
trouble. In this pape
r, the integral of time and
absol
u
t
e erro
r pro
d
u
c
t (ITAE)of the step re
spo
n
s
e
c
u
r
v
e o
n
th
e c
l
os
ed
-
l
oo
p
c
o
n
t
ro
l
s
y
s
t
em is
s
e
le
c
t
ed as
a fitn
e
s
s
fu
nc
tio
n
o
f
PSO
. T
h
e
e
r
r
o
r
perfo
rman
ce i
ndicator is
sh
own in Equ
a
tion (2
).
0
dt
|
e(t)
|
t
ITAE
(
2
)
2.2. PSO
PID Design
PSO get i
n
spi
r
ation
from
thi
s
m
odel
an
d i
s
u
s
e
d
to
solve optimi
z
atio
n p
r
obl
ems.
In PSO,
each individu
al sol
u
tion of
the optimiza
t
ion pro
b
lem
is a bi
rd in t
he search
sp
ace.
We
call
it
"particl
e". All the p
a
rti
c
le
s have
an
ad
aptative
valu
es
whi
c
h
is
determi
ned
b
y
function
to
be
optimize
d
, fitness value i
s
determin
ed
by the per
formance index
(2), ea
ch p
a
rt
icle ha
s a sp
eed
whi
c
h d
e
term
ines th
e dire
ction a
nd di
st
ance of t
heir
flight. Then the pa
rticle
s
are to follo
w
the
curre
n
t optim
al pa
rticle
se
arch in th
e
solution
sp
a
c
e
.
PSO is initi
a
lize
d
to a
g
r
oup
of rand
om
particl
es (ran
dom
solution
), and then
fin
d
the optim
al
solutio
n
by iteration. In
ea
ch ite
r
ation, t
h
e
particl
es
by tracking t
w
o "e
xtremes" to u
pdate thei
r o
w
n: the first "extreme valu
e" is the o
p
timal
solutio
n
whi
c
h is foun
d b
y
the pa
rticl
e
itself, thi
s
solutio
n
i
s
called th
e ind
i
vidual extre
m
e
point_Be
s
t, other
extreme i
s
. the enti
r
e p
opulatio
n to
find the o
p
tima
l solutio
n
, it is namely glo
b
a
l
extreme g
r
ou
p_Best. Alternatively, you can n
o
t hav
e
global extrem
e, but accordi
ng to the actu
al
situation, the use of local o
p
timum, so it is not
sele
cte
d
for the entire populatio
n but only as p
a
rt
of a colle
ction
of particl
es, t
hen the extre
m
es p
a
rt
i
c
le
colle
ction i
s
n
o
t global mini
mum, while i
s
a
local extre
m
e
.
PSO algorithm is iterativ
e in (3).
1
12
11
()
(
)
||
||
kk
k
k
k
k
id
id
g
d
g
d
id
id
id
id
kk
k
id
i
d
id
k
id
m
m
k
id
m
m
v
w
v
c
rand
p
x
c
r
and
p
x
xx
v
vV
xX
(3)
In the formul
a,
k
id
v
rep
r
e
s
e
n
ts the
k-th
gen
eration, th
e i
-
th parti
cle, th
e d-dime
nsio
nal
veloc
i
ty
v
.
k
id
x
repre
s
ent
s the
k-th g
ene
rati
on, the i-th p
a
rticle, the d
-
dimen
s
ion
a
l locatio
n
x
.
w
is
inertia fac
t
or.
is con
s
traint factor
o
f
speed
ratio.
k
id
p
is the optimal value for
the position o
f
the individual particl
es.
k
g
d
p
is optimal value for the gro
up locatio
n
.
1
c
,
2
c
is
accele
rating f
a
ctor.
g
d
r
and
,
id
r
and
is ran
dom num
ber
betwe
en [0,1].
Vmm
is spe
ed ra
nge,
a
nd
it
i
s
the bou
nd
arie
s
of
th
e spe
ed size
o
f
the
individu
al
movement, it appea
rs a
s
a cha
nge in
the individua
l trend.
Xm
m
is ra
nge for the p
opulatio
n
,
and it i
s
the
scop
e which p
opulatio
n indi
viduals
ca
n
a
c
tivity, That is rang
e of p
a
ramete
rs,
an
d it
is
p
K
,
i
T
and
d
T
thre
shold amo
unt in PID [11, 12].
PSO with
a similar g
eneti
c
algo
rithm i
s
based
on a
n
iterative opti
m
ization
tool
s. But it
did not use th
e cro
s
sove
r(crossove
r) a
n
d
variation (m
u
t
ation) of gen
etic algo
rithm
.
The particle
s
only se
arch i
n
the solutio
n
sp
ace a
s
b
i
rds foll
ow th
e optimal p
a
rticle sim
u
lati
on. Com
p
a
r
ison
with Ge
netic
algorith
m
, the advanta
g
e
s
of PSO i
s
simple
an
d e
a
sy to a
c
hi
e
v
e without m
any
para
m
eters n
eed to
be
adj
usted,
with
b
e
tter robu
stn
e
ss a
nd fe
asi
b
ility, and m
o
re im
po
rtant i
s
that PSO algorithm is e
a
si
er to impleme
n
t and und
erstand.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 1, Janua
ry 2015 : 33 – 4
1
36
3. T
w
o
Regi
onal Po
w
e
r
Sy
stem Model
3.1. Basic G
e
nera
tor Co
ntrol Loop
s and L FC-Lo
a
d Frequ
e
nc
y
Control
Figure 2. Sch
e
matic dia
g
ra
m of a synch
r
onou
s ge
nera
t
or's L
F
C a
n
d
AVR
In intercon
n
e
cted
sy
ste
m
s, loa
d
fre
quen
cy
cont
rol
(LF
C
) a
nd a
u
tomati
c voltage
regul
ation (A
VR) a
r
e in
stalled on e
a
ch gene
rato
r, load freq
ue
ncy co
ntrol
(LFC) rin
g
s a
n
d
automatic vol
t
age
reg
u
lation
(AVR) ri
n
g
a
r
e
sh
own
in sch
e
mati
c
diagram
of th
e Fig
u
re
2. T
he
controlle
r is
set to run a
particular
state, it
detects sm
all ch
an
ges in lo
ad, maintainin
g the
freque
ncy
an
d amplitud
e
of the voltag
e of the g
e
n
e
rato
r withi
n
an allo
wa
bl
e ra
nge. Acti
ve
cha
nge
s a
r
e
mainly dete
c
ted in th
e a
ngle of t
he g
enerator
roto
r, ie freq
uen
cy ch
ang
es
are
achi
eved,
chang
es in
re
active p
o
wer are
ma
i
n
ly
achi
eveed
by gen
erato
r
v
o
ltage
amplit
ude
detectio
n
. Excitation syste
m
time const
ant is mu
ch
smalle
r than
the origin
al motivation for the
time cons
tant, s
o
it's
muc
h
fa
s
t
er
in tr
a
n
s
i
en
t de
c
a
y, an
d it w
ill n
o
t
a
ffe
c
t
th
e d
y
namic
cha
r
a
c
teri
stics of L
F
C, a
n
d
t
herefore,
LFC
co
ntrol l
oop a
nd AV
R control loo
p
ca
n b
e
vie
w
ed
indep
ende
ntly of each oth
e
r.
LFC'
s g
oal i
s
to
keep
th
at the sy
ste
m
frequ
en
cy is the
nomi
nal fre
quen
cy, load is
distrib
u
ted
be
tween th
e g
e
nerato
r
s, co
n
t
rol tie line
po
wer is fo
r the
plann
ed valu
e. It detects the
freque
ncy a
n
d
tie-line
power chan
ge
s, such
as by d
e
tecting
a freq
u
ency e
r
ror
sig
nal
f and tie-
line p
o
wer error
sig
nal
P
tie
, th
e
er
r
o
r
s
i
gna
l is a
m
p
lifie
d, mixe
d
an
d tra
n
s
for
m
ed
in
to
th
e ac
tive
control sig
nal
P
V
, and it i sent to the pri
m
e mover to
get torque in
creme
n
t.
Prime move
r
bring
s
gene
rator o
u
tput p
o
we
r
cha
nge
P
g
, it will cha
nge the
f and
P
tie
, so that it
rem
a
in
s with
in the all
o
wa
ble rang
e.
Th
e first
step
in
the an
alysi
s
and
de
sign
o
f
control syste
m
s is to esta
blish a math
e
m
atical
mo
de
l of the system, and the most pop
ular m
odel
of the two method
s are
the transfe
r functi
on me
thod and the
state variabl
e method. State
variable m
e
th
od ca
n be a
p
p
lied to linea
r and nonli
nea
r
system
s, an
d in ord
e
r to
use the tran
sfer
function
met
hod
and
line
a
r e
quatio
n
of state, th
e
syste
m
mu
st first be
lin
eari
z
ed, i
e
,
with
rea
s
on
able
a
s
sumption
s
and
app
roxi
mations,
the
mathem
atical eq
uation
s
are
line
a
ri
zed to
obtain t
r
an
sf
er fun
c
tion
of
a ge
ne
rator
model, lo
ad
model, p
r
ime
mover
mod
e
l
and th
e
spe
e
d
control syste
m
model. The
s
e mod
e
ls a
r
e in Referen
c
es [1].
3.2 AG
C-Au
tomatic Gen
e
ration Contr
o
l in the T
w
o
Area Sy
stem
Whe
n
the l
o
ad sudd
enly
increa
se
s,
befor
e
the
speed
co
ntrol
system
ch
ange
s the
amount of st
eam into the
steam engi
n
e
, the s
pee
d
of the turbine ha
s been
dropp
ed.Eerror
sign
al of flywheel d
e
tecte
d
frequ
en
cy is
small, tu
rbine
ca
n be
maint
a
ined i
n
su
ch
a ste
ady
state,
its spe
ed is consta
nt, the speed is lo
we
r than the
rate
d spe
ed, there is a frequ
en
cy deviation (i
e,
adju
s
tment freque
ncy of governo
r is droo
p). Th
e
r
e is a way to accumulate
su
ch a freq
u
ency
deviation, o
r
it is the frequ
ency d
e
viatio
n integ
r
al
, th
e freq
uen
cy
offset is
moni
tored fo
r p
e
ri
od
with inte
gral
unit, the fre
q
uen
cy is ba
ck to the
nomin
al value
ba
se
d on
the i
n
te
gral val
ue.
When
the sy
stem l
o
ad
cha
nge
s
continuo
usly, t
he frequ
ency
of the
ge
nerator i
s
adju
s
t
ed to
the
rati
ng,
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Two Regio
nal
Power S
yste
m
PSO
PID
Control Re
se
arch (Jingfa
n
g Wan
g
)
37
whi
c
h i
s
cal
l
ed the
Aut
o
matic Ge
ne
ration
Co
ntrol (AG
C
). In
the inte
rco
nne
cted
syst
em
comp
osed of
several regi
ons, AGC b
e
t
ween regio
n
s
distrib
u
te load
s betwe
e
n
the powe
r
plant
and b
e
twe
e
n
the gen
erat
ors i
n
o
r
de
r to achieve
m
a
ximum e
c
on
omic b
enefits. It also co
ntrols
the tie-line
p
o
we
r at the
plann
ed valu
e, and the
system's f
r
equ
ency i
s
en
su
red in
rating.
Of
cou
r
se, the system mu
st be sta
b
le. In
the ca
se
of large
distu
r
b
a
n
ce
s a
nd a
ccidents, AG
C
will
exit, and it is
applie
d to accide
nt control
prog
ram [1].
A gro
up
of closely
gene
rators,
sp
eed
unity, gen
erator
rotor ha
s the
same
resp
on
se
cha
r
a
c
teri
stics, su
ch
a
s
tu
rbine
are
ca
lled a
s
relate
dgen
erato
r
s
whi
c
h i
s
a
re
pre
s
entative
by
LFC ri
ng, it is call
ed the
control area. AGC in
two
region
s is u
s
e
d
to underst
and multi-zo
n
e
system A
G
C,
co
nsid
er two
equivale
nt g
enerator
s a
r
e
on b
ehalf
of two
regi
onal
system
s. A
G
C
model of two
-
zo
ne
system
is in Figu
re
3 [1]. Ar
ea control e
r
ror (ACEs) i
s
u
s
e
to establi
s
h
two
area
s of the system simul
a
tion diag
ram.
And the
frequ
ency re
sp
on
se of the powe
r
is dete
r
mine
d
in each regi
o
n
. Figure 4
sh
ows the
SIMULINK
simula
tion diagram [1].
Figure 3. AGC blo
ck di
agram
of a two-zone sy
stem [1]
Figure 4. Example simul
a
tion blo
ck di
ag
ram [1]
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 1, Janua
ry 2015 : 33 – 4
1
38
4. Simulation of PSO-PID Experimen
t
al Tes
t
s
Based o
n
two re
gion
s turbin
e mod
e
l
in Figure 4,
PSO-PID co
ntrol is de
sig
ned, and
matlab
simul
a
tion i
s
in
Fig
u
re
5. PSO-P
ID control
si
mulation
wo
rk flo
w
chart
di
agra
m
i
s
in
Fi
gure
6. Step tra
cki
ng an
d key resp
on
se in
dicators bef
o
r
e
usin
g PSO-PI
D
control are
in Figu
re 7. A
nd
step tra
c
king
and key re
sp
onse indi
cato
rs by u
s
ing P
S
O-PID control are in Fig
u
re 8.
Figure 5. PSO-PID control
simulation bl
ock diag
ram
Figure 6. PSO-PID control
simulation
work flo
w
cha
r
t diagram
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Two Regio
nal
Power S
yste
m
PSO
PID
Control Re
se
arch (Jingfa
n
g Wan
g
)
39
Figure 7. Step tracking a
n
d
key re
spo
n
s
e indi
cato
rs i
n
no PID cont
rol
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 13, No. 1, Janua
ry 2015 : 33 – 4
1
40
Figure 8. Step tracking an
d key re
spo
n
s
e
indi
cato
rs
by using PSO
-PID co
ntrol
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Two Regio
nal
Power S
yste
m
PSO
PID
Control Re
se
arch (Jingfa
n
g Wan
g
)
41
By compa
r
in
g the sim
u
lat
i
on re
sult
s b
e
fore a
nd aft
e
r u
s
ing PS
O-PID
cont
ro
l
，
1) the
amount
of overshoot i
n
th
e step
total resp
on
se
d
o
wns from 6
5
%
to 5%. 2) Adj
u
st time
do
wns
from 12
seco
nds to
3
se
cond
s. 3)
The
main p
a
ra
m
e
ters:
Fre
que
ncy deviatio
n
step
re
spo
n
s
e
and po
we
r de
viation step resp
on
se
are improve
d
sig
n
i
ficantly.
5. Conclusio
n
and Outlo
o
k
In this
pape
r, for multi-o
b
jective
optimizat
ion
cha
r
acte
ri
stics o
f
two regio
n
a
l po
we
r
system PID controll
er
desi
gn, we
prop
os
e a PID param
eter tuning of
particle
swarm
optimizatio
n. First, by defin
ing a com
p
re
hen
sive
co
nsi
deratio
n of system
output
overshoot, ri
se
time and the
fitness fu
nct
i
on term
ste
ady-state
error indi
cato
rs, su
ch a
s
th
e ITAE, and in
accordan
ce
with the
pe
rforman
c
e
re
quire
ment
s
o
f
the a
c
tual
co
ntrol
syst
em, app
ro
pri
a
te
weig
hting of each index item. After that, the t
ape-ba
sed a
nd imp
r
oved pa
rticle
swarm al
go
rithm
are u
s
e
d
for multi-obj
ecti
ve optimizati
on PI
D. PSO optimizatio
n algo
rithm
has
goo
d gl
obal
search
ability and high
conv
ergence
rate. The PID
contro
ll
er
tuni
ng parameters can
be
respon
ded di
rectly from the
output
of the PID control system, it is easily bala
n
ce betwe
en spe
e
d
and sta
b
ility
control syste
m
s. Impleme
n
tation of
the algorithm do
es not dep
en
d on the act
ual
controlled o
b
j
e
ct model, it is with a wi
de
rang
e of pra
c
ticality.
Particle
swa
r
m algo
rithm
is u
s
ed to
o
p
timize th
e p
a
ram
e
ters of
PID co
ntrol
system
desi
gn, throu
gh si
mulation
experim
ents can
be
se
en
, PSO algorit
hm ha
s g
ood
rob
u
stn
e
ss
and
dynamic
qua
lity. Optimize performan
ce and e
ffici
ency of the
algorithm t
han the ge
n
e
tic
algorith
m
ha
s improved to some exte
nt. When
th
e transfe
r fu
nction
with large la
gs, PSO
algorith
m
is a
b
le to meet the system
re
quire
ment
s for the adaptiv
e PID param
eters. Th
eref
ore,
the use of pa
rticle swa
r
m
optimizatio
n algorith
m
PID param
eter o
p
timization m
e
thod is a
kin
d
of
good
practi
ca
l value.In the
power
co
ntrol
syst
em, f
r
eq
uen
cy deviati
on
step
re
spo
n
se
an
d po
wer
deviation ste
p
respon
se a
r
e improve
d
si
gnifica
ntly
Referen
ces
[1]
Saad
at, Hadi.
Po
w
e
r S
y
stem
Anal
ys
is. Mcgra
w
-Hi
ll Co
lle
ge
; Har/Dsk Su. 1998: 52
7-5
85.
[2]
Cha
udh
uri B,
Majumd
er R,
Pal BC. Wi
de
-ar
ea m
eas
ure
m
ent-bas
ed st
abil
i
zi
ng c
ontr
o
l of
po
w
e
r
s
y
stem c
ons
id
erin
g si
gna
l tr
ansmissi
on
de
la
y
.
IEEE Transactions on
Power Systems
. 200
4; (0
4):
197
1-19
79. do
i
:
10.110
9/
T
P
WRS.200
4.83
56
69.
[3]
LUO Ke,
L Hong-li, etc. T
w
o la
yer
int
e
r-are
a d
a
mpi
ng c
ontro
l
of p
o
w
e
r
s
y
ste
m
s consi
der
ing
time-d
ela
y
base
d
on LMI. Po
w
e
r S
y
stem
Protecti
on a
nd
Contro
l. 201
3; (24): 33-3
6
.
[4]
Hsu Mi
ng-R
en,
Ho W
en-
Hsie
n, Cho
u
J
y
h-
H
o
rng. Sta
b
le
a
nd q
u
a
d
ratic o
p
timal c
ontrol f
o
r T
S
fuzzy
-
mode
l-bas
ed
ti
me-del
a
y
c
ontr
o
l s
y
stems.
IEEE Tra
n
s
a
c
tion
s on
System
s Ma
n an
d C
y
be
rn
e
t
i
cs-Pa
rt
A: Systems an
d Hu
ma
ns
. 200
8; (04): 933-9
4
4
.
[5]
JG Zigeler, Nichlos. Optimizat
i
on settin
g
for automatic co
ntroller.
T
r
an. AS
ME
. 1942; 64(
11): 756-
76
9.
[6] A
Visioli.
T
u
ni
ng
of PID c
ont
rollers
w
i
th fu
zz
y
lo
gic.
Proc.
Inst. Elect. E
ng. C
ontr. T
heor
y Ap
plic
at.,
200
1; 148(
1): 1–8.
[7]
T
L
Seng, MB Khal
id, R Yus
o
f.
T
uning
of a
neuro-f
u
zz
y
c
ontrol
l
er b
y
g
e
netic al
gor
ithm
.
IEEE Trans.
Syst., Man, Cy
bern. B,
199
9; 29: 226
–2
36.
[8]
Y Mitsukura, T
Yamamoto, M Kaneda.
A desig
n of self-tuni
ng PID
c
ontroll
ers using a g
enetic
alg
o
rith
m.
Proc. Amer. Contr.
Conf., San
Diego, CA. 1999;
1361–1365.
[9]
RA Kro
h
li
ng, J
P
Re
y.
Desi
gn
of o
p
tima
l d
i
sturba
nce r
e
j
e
cti
on PID
co
ntroll
er
s us
ing
g
e
n
e
t
ic al
gorithm
.
IEEE Trans. E
v
ol. Com
p
ut.,
2001; 5; 78
–82.
[10]
J Kenn
ed
y, R
Eberh
a
rt.
Particle sw
arm opti
m
i
z
at
io
n.
Proc. IEEE Int. Conf. Neural
Net
w
orks, vol. IV
,
Perth, Australi
a. 1995; 1
942
–
194
8.
[11]
Y Shi, R Eb
e
r
hart.
A modif
i
ed
particl
e s
w
arm opti
m
i
z
e
r
.
Proc. IEEE Int. Conf. Evol. Com
put.,
Anchor
ag
e, AK. 1998; 69
–73.
[12]
L W
ang. Appli
c
ation of ad
ap
tive
gen
etic al
gorithms in PI
D controll
er d
e
sig
n
. J. Xi’An
Univ. Sci. &
T
e
ch., 2005; 2
5
(1): 93-9
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.