Int
ern
at
i
onal
Journ
al of
R
obot
ic
s
and
Autom
ati
on (I
JRA)
Vo
l.
8
,
N
o.
4
,
D
ece
m
ber
201
9
,
pp.
256
~
268
IS
S
N:
20
89
-
4
856
,
DOI: 10
.11
591/
i
jra
.
v
8
i
4
.
pp25
6
-
268
256
Journ
al h
om
e
page
:
http:
//
ia
escore.c
om/j
ourn
als/i
ndex.
ph
p/IJRA
Optimi
zation
of
PID cont
ro
ll
er param
eters
using P
SO for t
wo
area l
oad frequ
en
cy
c
ont
ro
l
Br
ijesh K
umar
D
u
bey
1
,
N.
K.
Sin
gh
2
,
S
am
eer B
hambr
i
1
1
El
e
ct
ri
ca
l
and
E
le
c
troni
cs
Engi
n
ee
ring
,
PS
IT
Ka
npur
,
Indi
a
2
El
e
ct
ri
ca
l
Eng
in
ee
ring
,
RVIT
Bij
nor
,
Indi
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Oct
12
,
201
8
Re
vised
Ma
y
15
,
201
9
Accepte
d
Oct
6
,
201
9
In
th
is
p
ape
r
,
an
evol
ut
ionar
y
computing
appr
oac
h
for
det
ermining
the
opti
m
al
va
l
ues
for
the
pro
porti
onal
-
integra
l
-
der
iv
at
iv
e
(PID
)
cont
ro
ll
er
par
amete
rs
o
f
l
oad
fre
qu
ency
c
ontrol
(
LFC)
of
two
ar
ea
powe
r
s
y
st
em
is
pre
sente
d
.
Th
e
proposed
appr
o
ac
h
emplo
y
s
a
par
ticle
sw
arm
opti
m
iz
ation
te
chn
ique
to
fin
d
opti
m
um
par
a
m
et
ers.
Th
e
sta
t
e
spac
e
m
odel
of
two
area
power
s
y
s
te
m
a
nd
an
Ei
gen
v
alue
b
ase
d
obj
ec
t
i
ve
fun
cti
on
is
c
onsidere
d.
The
eff
e
ct
iv
enes
s
of
the
prop
osed
appr
o
ac
h
is
compare
d
with
int
egr
al
cont
rol
.
Sim
ula
t
i
on
result
s justi
f
y
the
proposed
ap
proa
ch
in
te
rm
s of
damping
the
osc
il
l
at
ions
,
i
m
prove
d
sett
l
ing
ti
m
e
,
l
ess ove
r/
under
shoots
.
Ke
yw
or
d
s
:
Eigen
v
al
ues
Loa
d
f
re
qu
e
nc
y con
t
ro
l
(LFC
)
P
arti
cl
e sw
a
rm
optim
iz
at
ion
(P
S
O)
PI
D
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Brijes
h Ku
m
ar Dub
ey
,
Ele
ct
rical
and
Ele
ct
ro
nics
E
nginee
rin
g
,
Pr
a
nv
ee
r
Si
ngh I
ns
ti
tute o
f
Te
chnolo
gy
(P
I
S
T) Kan
pur
,
I
nd
ia
.
Em
a
il
:
br
ijesh8
0d@r
e
dif
fm
ai
l
.co
m
1.
INTROD
U
CTION
In
powe
r
syst
e
m
s
,
bo
t
h
act
ive
an
d
reacti
ve
powe
r
dem
and
s
are
ne
ver
ste
ady
an
d
c
on
ti
nu
ously
change
with
ri
sing
or
fall
ing
tren
d.
Stea
m
inp
ut
to
t
urbo
ge
ner
at
or
s
(or
wa
te
r
input
to
hy
dro
ge
ne
rato
rs)
m
us
t
therefo
re
,
be
c
on
ti
nu
ously
re
gu
la
te
d
to
m
atch
t
he
act
ive
powe
r
dem
and
,
fail
ing
w
hich
the
m
achi
ne
s
pe
ed
will
var
y
with
c
ons
equ
e
nt
c
ha
ng
e
in
fr
e
quency
,
wh
ic
h
m
ay
be
highly
undesi
rab
le
.
I
n
br
ie
f
,
t
he
c
ha
ng
es
i
n
real
powe
r
a
ff
ect
t
he
syst
em
fr
e
qu
e
n
cy
,
w
hile
reacti
ve
po
we
r
is
le
ss
se
ns
it
ive
to
cha
nges
in
fr
e
quency
and
is
m
ai
nly
dep
e
nd
ent o
n
c
ha
nges
in
v
oltage
m
a
gn
it
ude.
T
he
qual
it
y
of p
ower
sup
ply
m
us
t
m
eet
certai
n
m
ini
m
u
m
sta
nd
a
rds
with
reg
a
rd
to c
onst
an
cy
of
volt
ag
e an
d
fr
e
qu
e
nc
y
[1
]
.
The
operati
on
al
obj
ect
ive
of
LFC
is
to
m
ai
ntain
reas
onably
un
i
form
fr
e
quency
,
to
div
i
de
l
oad
betwee
n
ge
ne
r
at
or
s
a
nd
t
o
co
ntr
ol
the
ti
e
-
li
ne
intercha
nge
s
chedules.
T
he
change
in
t
he
f
reque
ncy
an
d
ti
e
-
li
ne
powe
r
are
sen
s
ed
,
w
hich
is
a
m
easur
e
of
t
he
cha
ng
e
in
ro
t
or
a
ngle
,
i.e
.
,
t
he
er
r
or
t
o
be
correct
ed.
T
he
error
sign
al
,
i.e.
,
∆f
a
nd
∆Pti
e
,
ar
e
am
plifie
d
,
m
ixed
,
a
nd
t
ransform
ed
into
a
real
power
c
omm
and
sig
na
l
∆P
v
,
wh
ic
h
is
se
nt
to
the
pr
im
e
m
ov
e
r
to
cal
l
f
or
an
inc
rem
ent
in
the
to
rque.
The
pr
im
e
m
o
ver
,
the
refor
e
,
br
i
ng
s
change
in
t
he
ge
ne
rato
r
outp
ut
by
an
am
ount
∆Pg
w
hich
will
change
th
e
values
of
∆
f
and
∆
Pti
e
with
in
the
sp
eci
fied
tolera
nce.
In
this
pap
e
r
m
at
he
m
at
ic
a
l
m
od
el
for
the
two
area
syst
e
m
is
dev
el
op
e
d
in
or
der
to
analy
ses
an
d
desig
n
t
he
co
nt
ro
l
syst
em
.
A
sta
te
sp
ace
m
od
el
is
de
velo
ped
by
li
nea
ri
zi
ng
t
he
m
at
hem
at
ic
al
equ
at
ion
s
of
diff
e
re
nt
c
om
po
ne
nts
an
d
th
us
the
sta
te
m
atr
ix
is
form
ed.
An
Ei
gen
valu
e
based
obj
ect
i
ve
f
un
ct
io
n
to
place
the m
od
es in
bet
te
r
reg
i
on
on
the co
m
plex pl
ane is c
onside
r
ed
a
nd opti
m
um
co
ntro
ll
er
pa
ram
et
ers
are fo
und.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
R
ob
&
A
uto
m
IS
S
N:
20
89
-
4856
Op
ti
miz
atio
n o
f PID co
ntr
oller p
arameters
usi
ng PSO f
or
t
wo... (B
rij
esh K
. Dubey)
257
2.
SY
STE
M
I
N
VESTIG
ATED
Durin
g
norm
al
op
e
rati
on
,
t
he
real
powe
r
tra
nsfer
red
over
t
he
ti
e
-
li
ne
is
give
n
by
(as
s
how
n
in
Fig
ur
e
1
a
nd 2)
,
12
1
2
1
2
s
in
T
EE
P
X
=
(1)
Wh
e
re
,
1
2
1
2
=−
.
Fo
r
a sm
al
l dev
ia
ti
on
i
n
t
he
ti
e
-
li
ne
fl
o
w
f
rom
the nom
inal value:
12
12
12
12
12
s
dP
P
d
P
=
=
(2)
Wh
e
re
Ps is t
he
slo
pe of p
ower a
ng
le
c
urve
at
init
ia
l op
erat
ing
a
ngle
call
ed
syn
ch
ronizi
ng
powe
r
c
oeffici
ent.
12
1
2
1
2
0
c
o
s
T
EE
P
X
=
(3)
(
)
1
2
1
2
s
PP
=
−
(4)
A
r
e
a
1
A
r
e
a
2
X
t
i
e
P
1
2
Figure
1. Tw
o area sy
ste
m
E
1
∠
δ
1
E
2
∠
δ
2
X
1
X
2
X
t
i
e
P
1
2
X
T
=
X
1
+
X
t
i
e
+
X
2
Figure
2. Ele
ct
rical
equivale
nt
f
or a t
wo area
syst
e
m
A
blo
c
k
diagr
a
m
rep
resen
ta
ti
on
of
tw
o
a
rea syst
e
m
with
LFC
co
ntainin
g
on
ly
p
rim
ary
l
oop
is
s
ho
w
n
in
F
ig
ure
3
wi
th
eac
h
a
rea
r
epr
ese
nt
ed
by
an
e
quivale
nt
inerti
a
M
,
loa
d
dam
pin
g
co
ns
ta
nt
D
,
t
urbine
a
n
d
gove
rn
i
ng
syst
e
m
with
an
eff
ect
ive
sp
ee
d
droop
R
.
T
ie
-
li
ne
is
re
presented
by
sy
nchr
on
i
z
in
g
t
orq
ue
coeffic
ie
nt.
A
p
os
it
ive
re
pr
es
ents
an
i
ncr
eas
e
in
po
wer
tra
ns
fe
r
f
r
om
area
1
to
a
rea
2
a
nd
it
is
eq
ui
va
le
nt
to
increasin
g
l
oa
d i
n
a
rea
1
a
nd dec
reasin
g
lo
ad
in ar
ea
2.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
I
nt
J
R
ob
&
A
uto
m
,
Vo
l.
8
,
No.
4
,
Decem
ber
2019
:
256
–
268
258
1
(
1
+
T
g
1
s
)
1
(
1
+
T
t
1
s
)
1
(
2
H
1
s
+
D
1
)
1
(
1
+
T
g
2
s
)
1
(
1
+
T
t
2
s
)
1
(
2
H
2
s
+
D
2
)
1
/
R
1
1
/
R
2
P
s
/
s
-
+
+
-
+
∆
P
L
2
∆
P
L
1
-
-
+
-
∆
P
1
2
∆
P
r
e
f
1
∆
P
r
e
f
2
∆
P
v
1
∆
P
m
1
∆
P
v
2
∆
P
m
2
-
∆
ѡ
1
∆
ѡ
2
Figure
3. Tw
o area sy
ste
m
w
it
h
pri
m
ary LFC loop
Both a
reas
will
h
a
ve
a
sam
e st
eady sta
te
freq
uen
cy
de
viati
on
for
a l
oad ch
ang
e
of
i
n
are
a
1
.
12
=
−
an
d
1
1
2
1
1
mL
P
P
P
D
−
−
=
(5)
2
1
2
2
m
P
P
D
+
=
(6)
Fr
om
the
gove
rnor spe
ed
ch
a
racteri
sti
cs th
e
change i
n
m
echan
ic
al
powe
r i
s:
11
22
/
/
m
m
PR
PR
=
=
(7)
1
12
12
11
L
P
DD
RR
=
−
+
+
+
(8)
Wh
e
re
,
11
1
22
2
1
1
BD
R
BD
R
=+
=+
(9)
21
2
12
12
12
1
11
L
DP
R
P
DD
RR
+
=
−
+
+
+
(10)
In
norm
al
op
erati
ng
c
onditi
on
po
wer
syst
e
m
is
op
erated
so
that
dem
a
nd
of
a
reas
is
sat
isfie
d
at
no
m
inal fr
e
que
ncy.
A
sim
ple co
nt
ro
l st
rateg
y shoul
d
incl
ude the
foll
owin
g functi
ons:
‐
Fr
e
qu
e
ncy a
pproxim
at
e
l
y at
n
om
inal value.
-
Ma
intai
nin
g t
he
ti
e
-
li
ne
fl
ow
at
ab
out sc
he
dule
.
-
Each a
rea s
houl
d
ab
sorb
it
s own loa
d
c
hang
es.
2.1.
Tie
-
li
ne b
ias c
on
tr
ol
Conve
ntion
al
LFCs
a
re
base
d
on
ti
e
-
li
ne
bias
c
ontrol
w
her
e
eac
h
a
rea
te
nds
to
co
nt
ro
l
it
s
area
con
t
ro
l
er
ror
(
ACE)
to
ze
ro.
T
he
co
ntr
ol
error
c
onsist
s
of
li
nea
r
c
ombinati
on
of
f
r
equ
e
ncy
an
d
ti
e
-
li
ne
error
[2
].
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
R
ob
&
A
uto
m
IS
S
N:
20
89
-
4856
Op
ti
miz
atio
n o
f PID co
ntr
oller p
arameters
usi
ng PSO f
or
t
wo... (B
rij
esh K
. Dubey)
259
1
n
i
j
i
j
i
A
C
E
P
K
=
=
+
(11)
The
area
bias
factor
K
i
determ
ines
the
am
ou
nt
of
i
nteracti
on
duri
ng
the
distu
rb
a
nc
es
in
th
e
neig
hbori
ng
ar
ea. In
orde
r
to
get sati
sfacto
ry
p
e
rfor
m
ance a
rea
bias f
act
or
is sel
ect
ed
as
:
1
()
ii
i
KD
R
=+
(12)
1
1
2
1
1
2
2
1
2
2
A
C
E
P
B
A
C
E
P
B
=
+
=
+
(13
)
12
P
an
d
21
P
are
the
de
viati
on
s
f
r
om
the
sc
heduled
intercha
nges.
ACEs
are
the
a
ct
uating
sig
na
ls
that
a
r
e
us
e
d
to
c
hang
e
the
ref
e
re
nce
set
po
i
nts
,
an
d
w
he
n
the
ste
ady
sta
te
is
reached
a
nd
will
be
zer
o.
T
he
blo
c
k
diag
ram
o
f
tw
o area sy
ste
m
LFC w
it
h i
nteg
r
al
co
nt
ro
l i
s
shown i
n
Fi
gure
4.
1
(
1
+
T
g
1
s
)
1
(
1
+
T
t
1
s
)
1
(
2
H
1
s
+
D
1
)
1
(
1
+
T
g
2
s
)
1
(
1
+
T
t
2
s
)
1
(
2
H
2
s
+
D
2
)
1
/
R
1
1
/
R
2
P
s
/
s
-
+
∆
P
L
2
∆
P
L
1
-
-
+
-
∆
P
1
2
∆
P
v
1
∆
P
m
1
∆
P
v
2
∆
P
m
2
∆
ѡ
1
∆
ѡ
2
K
I
1
/
s
K
I
2
/
s
B
1
B
2
-
+
-
-
+
+
+
A
C
E
1
+
A
C
E
2
Figure
4. Bl
oc
k d
ia
gram
o
f
t
wo area
syst
em
LFC w
it
h
in
te
gr
al
c
on
t
ro
l a
ct
ion
In
t
his
pa
per
the
inte
gr
al
c
ontrol
bl
ock
is
re
placed
by
PID
con
t
ro
ll
er
a
nd
sta
t
e
sp
ace
m
od
el
of
t
he
syst
e
m
Ẋ
=
AX
is
obta
ine
d.
S
ta
te
m
at
rix
of
order
11
⨯
11
i
s
f
or
m
ed
with
sta
te
var
ia
bles
1
g
P
,
1
v
P
,
1
m
P
,
1
,
1
A
C
E
,
12
P
,
2
g
P
,
2
v
P
,
2
m
P
,
2
,
2
A
C
E
res
pecti
ve
ly
.
The
pa
ram
e
te
rs
of
t
he
syst
e
m
wh
ic
h
ar
e
to
be
optim
iz
e
d
us
i
ng
PS
O
ar
e
1
p
K
,
1
l
K
,
1
D
K
,
1
R
,
1
B
,
2
p
K
,
12
K
,
2
D
K
,
2
R
,
2
B
r
e
sp
ect
ively
[
3].
Bl
oc
k
diag
ram
o
f
tw
o area sy
ste
m
w
it
h
PID c
ontrol
le
r
is s
how
n
in
Figure
5.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
I
nt
J
R
ob
&
A
uto
m
,
Vo
l.
8
,
No.
4
,
Decem
ber
2019
:
256
–
268
260
1
(
1
+
T
g
1
s
)
1
(
1
+
T
t
1
s
)
1
(
2
H
1
s
+
D
1
)
1
(
1
+
T
g
2
s
)
1
(
1
+
T
t
2
s
)
1
(
2
H
2
s
+
D
2
)
1
/
R
1
1
/
R
2
P
s
/
s
-
+
∆
P
L
2
∆
P
L
1
-
-
+
-
∆
P
1
2
∆
P
v
1
∆
P
m
1
∆
P
v
2
∆
P
m
2
∆
ѡ
1
∆
ѡ
2
P
I
D
1
(
s
)
P
I
D
2
(
s
)
B
1
B
2
-
+
-
-
+
+
+
A
C
E
1
+
A
C
E
2
Figure
5. Bl
oc
k diag
ram
o
f
t
wo area
syst
em
LFC w
it
h
PID c
on
t
ro
l act
io
n
PI
D
1(s)
=
1
11
l
pD
K
K
K
s
s
++
(14)
PI
D
2(s)
=
2
22
l
pD
K
K
K
s
s
++
(15)
3.
PAR
TI
CLE S
WA
RM OPTI
MIZ
ATION
The
P
SO
al
gor
it
h
m
was
fir
st
introd
uced
by
Dr
.
Russ
el
C.
Eberha
rt
a
nd
Dr
.
Jam
es
Kenned
y
(19
95)
,
insp
ire
d
by
soc
ia
l
beh
avi
or
of
bir
d
fl
oc
king
or
fish
sc
hool
ing
.
The
syst
e
m
is
init
ializ
ed
with
a
po
pu
la
ti
on
of
rand
om
so
luti
on
s
a
nd
sea
rc
hes
f
or
op
ti
m
a
by
up
dating
ge
ner
at
io
ns
[4
]
.
H
oweve
r
,
un
l
ike
GA
,
PS
O
has
no
evo
l
ution
oper
a
tors
s
uch
as
cro
ss
over
a
nd
m
utati
on
.
I
n
PSO
,
the
pote
ntial
so
luti
ons
,
cal
le
d
par
ti
cl
es
,
fly
thr
ough
the
pr
ob
le
m
sp
ace
by
fo
ll
owin
g
t
he
cu
rr
e
nt
opti
m
u
m
par
ti
cl
es
.
Eac
h
par
ti
cl
e
kee
ps
trac
k
of
it
s
coor
din
at
es
in
the
pro
blem
sp
ace
wh
ic
h
are
associat
ed
with
th
e
best
so
l
ut
ion
(
fitness
)
it
has
ac
hieve
d
so
fa
r.
(
The
fitness
va
lue
is
al
so
st
ored
)
t
his
val
ue
i
s
cal
le
d
pbest
.
Anothe
r
"
best"
value
t
hat
is
tr
acked
by
the
pa
rtic
le
swar
m
op
ti
m
i
zer
is
t
he
best
val
ue
,
obta
in
ed
s
o
far
by
a
ny
pa
rtic
le
in
the
neig
hbors
of
the
pa
rtic
le
.
T
his
locat
ion
is
cal
l
ed
l
best.
W
he
n
a
par
ti
cl
e
ta
ke
s
al
l
the
popula
ti
on
a
s
it
s
t
opologica
l
nei
gh
bors
,
the
best
va
lue
is
a g
lo
bal
best a
nd is cal
le
d g
be
st.
In
t
he
nu
m
erical
i
m
ple
m
entation
of
t
his
sim
plifie
d
s
ocial
m
od
el
,
each
pa
rtic
le
has
th
re
e
a
tt
ribu
te
s
;
the
posit
ion
ve
ct
or
in
the
sea
r
ch
s
pace
,
t
he
c
urren
t
directi
on
vector
,
the
be
st
posit
ion
i
n
it
s
track
a
nd
t
he
best
po
sit
io
n of t
he swarm
. Th
e
pr
ocess
ca
n be
outl
ined
a
s foll
ows
[5
]
:
Step
1.
Gen
e
rate t
he
in
it
ia
l swar
m
inv
olv
in
g N
pa
rtic
le
s at ra
nd
om
.
Step
2.
Ca
lc
ulate
the new
directi
on
ve
ct
or
for
eac
h partic
le
b
a
sed
on it
s att
rib
utes.
Step
3.
Ca
lc
ulate
the
new
sea
rc
h
po
sit
ion
of
eac
h
pa
rtic
le
f
r
om
the
c
urren
t
s
earch
posit
ion
an
d
it
s
ne
w
directi
on
vecto
r.
Step
4.
If
te
rm
inati
on
co
ndit
ion i
s sa
t
isfie
d
,
sto
p.
Ot
herwise
,
go to
ste
p 2.
As
t
he
pa
rtic
le
can
fly
in
D
-
dim
ension
searc
h
sp
ace
,
t
he
po
sit
ion
a
nd
velo
ci
ty
of
i
-
t
h
par
t
ic
le
can
be
represe
nted
as:
Xi=[
x
i1
,
x
i2
,
x
i3
,
x
i4
,
……
.x
iD
]
(16)
Vi=[
vi1
,
vi2
,
vi
3
,
vi4
,
……
viD]
(17)
W
it
h
incre
ased
it
erati
on
,
th
e
swar
m
will
m
ov
e
t
ow
a
rds
it
s
global
best
posit
ion
by
kee
ping
track
of
their
per
s
onal
best.
In D
-
dim
ensio
nal sea
rch sp
ace
the
pbes
t of i
-
th
pa
rtic
le
can be
re
pr
es
ented
a
s:
Pb
est
=[
p
i1
,
p
i2
,
p
i3
,
p
i4
,
……
..p
iD
]
(18)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
R
ob
&
A
uto
m
IS
S
N:
20
89
-
4856
Op
ti
miz
atio
n o
f PID co
ntr
oller p
arameters
usi
ng PSO f
or
t
wo... (B
rij
esh K
. Dubey)
261
and gbest
of th
e who
le
swa
rm
is prese
nted
as
:
gb
e
st=
[g
1
,
g
2
,
g
3
,
g
4
,
………
..g
D
]
(19)
T
he
ne
w
dire
ct
ion
vecto
r
o
f
the
i
-
th
par
t
ic
le
at
tim
e
t
,
is
cal
culat
ed
by
t
he
fo
ll
owin
g
sc
hem
e
i
ntrod
uced by
Sh
i a
nd E
berha
rt.
1
1
2
1
(
)
2
(
)
t
t
t
t
t
t
i
d
i
d
i
d
i
d
t
t
t
d
i
d
V
v
c
R
p
b
e
st
x
c
R
g
b
e
st
x
+
=
+
−
+−
(20)
1
t
R
an
d
2
t
R
are
ra
ndom
nu
m
ber
s
be
tween
0
a
nd
1.
t
id
v
an
d
t
id
x
is
t
he
vel
ocity
an
d
po
sit
io
n
of
t
he
i
-
th
par
ti
cl
e
in d
-
t
h
dim
ension
at
i
ts
tim
e
track
t.
t
id
p
b
e
st
is
the
be
st
pos
it
ion
of
the
i
-
th
par
ti
cl
e
(
per
sonal
best
)
in
d
-
t
h
dim
ension
in
it
s
track
at
tim
e
t
and
t
d
g
b
e
st
is
the
best
posit
ion
of
the
swa
rm
i
n
d
-
th
dim
ensi
on
at
ti
m
e
t.
Ther
e a
re three
p
aram
et
ers
suc
h
as the i
ner
ti
a o
f
t
he
pa
rtic
le
t
,
an
d
tw
o
pa
r
a
m
et
ers
c1
an
d c2.
c
1
an
d
c
2 are
the
le
ar
ning
fa
ct
or
s
w
hich
de
te
rm
ines
the
re
la
ti
ve
influ
e
nc
e
of
the
c
ogniti
ve
a
nd
so
ci
al
c
om
po
ne
nts
to
updat
e
the posit
io
n
a
nd
velocit
y com
pone
nt.
The
n
,
ne
w po
si
ti
on
of the i
-
th
par
ti
cl
e at t
i
m
e t
,
1
t
id
X
+
,
is
cal
culat
ed fr
om
:
11
t
t
t
id
id
id
X
X
V
++
=+
(21)
Wh
e
re
t
id
X
is
t
he
current
posit
io
n
of
the
i
-
th
pa
rtic
le
at
tim
e
t.
A
fter
t
he
i
-
th
par
ti
cl
e
cal
c
ulate
s
the
ne
xt
search
di
recti
on
vecto
r
1
t
i
V
+
in
co
ns
ide
rati
on
of
the
cu
rr
e
nt
sea
rch
directi
on
ve
ct
or
t
id
v
,
the
dire
ct
ion
vecto
r
go
i
ng
from
the
current
sea
rch
po
sit
io
n
t
id
X
to
th
e
best
searc
h
posit
ion
i
n
it
s
track
t
id
p
b
e
st
and
the
di
recti
on
vecto
r
goin
g
f
ro
m
t
he
c
urre
nt
sea
rch
posit
ion
t
id
X
to
t
he
be
st
searc
h
posit
ion
of
t
he
swa
rm
t
d
g
b
e
st
,
it
m
ov
es
from
t
he
c
urre
nt
pos
it
ion
t
id
X
to
th
e
ne
xt
sea
rch
pos
it
ion
1
t
id
X
+
cal
culat
ed
by
Eq
uati
on
(21)
.
In
gen
e
ral
t
he
pa
r
a
m
et
er
t
is
set
to
la
rg
e
values
in
t
he
ea
rly
sta
ge
f
or
global
s
earch
,
w
hile
it
is
set
to
a
sm
al
l
value
i
n
th
e las
t st
age fo
r
loca
l search
.
The
ine
rtia
we
igh
t
is
use
d
t
o
con
t
ro
l
the
im
pact
of
the
previ
o
us
vel
ociti
es
on
t
he
c
urr
ent
vel
ocity
,
influ
e
ncin
g
the
trade
-
off
bet
ween
t
he
glob
al
and
l
ocal
ex
per
ie
nce.
Alth
ough
Z
heng
cl
aim
ed
that
PS
O
with
increasin
g
ine
r
ti
a
weig
ht
pe
rfor
m
s
bette
r
,
li
near
decr
easi
ng
of
t
he
ine
rtia
wei
ght
is
rec
om
m
end
ed
by
S
hi
a
nd
Eber
ha
rt.
m
a
x
m
i
n
m
a
x
m
a
x
*
ww
w
w
i
t
e
r
i
t
e
r
−
=−
(
22
)
Wh
e
re
m
a
x
w
and
m
i
n
w
are
m
axi
m
u
m
and
m
ini
m
u
m
of
inerti
a
weig
ht
value
res
pecti
ve
ly
,
m
a
x
i
t
e
r
is
m
axi
m
u
m
it
erati
on
num
ber
a
nd
it
er
is
th
e
cu
rr
e
nt
it
erat
ion
.
A
s
o
-
cal
le
d
c
onstric
ti
on
f
act
or
K
,
is
fact
or
that
i
ncr
ea
s
es
the
al
gorithm
’s
ab
il
ity
to
co
nv
e
rg
e
t
o
a
good
so
luti
on
a
nd
can
ge
ner
at
e
higher
qual
it
y
so
luti
on
tha
n
the
conve
ntion
al
P
SO
a
ppr
oac
h.
I
n
this
case
,
t
he e
xpressi
on u
se
d
to
up
date the
p
a
rtic
le
’s
vel
oc
it
y b
ecom
es
1
1
2
*
[
1
(
)
2
(
)
]
t
t
t
t
t
t
i
d
i
d
i
d
i
d
t
t
t
d
i
d
V
K
v
c
R
pbe
st
x
c
R
gbe
st
x
+
=
+
−
+−
(
23
)
Wh
e
re
(
)
1
2
2
2
24
K
=
−
−
−
12
cc
=+
,
> 4
(24)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
I
nt
J
R
ob
&
A
uto
m
,
Vo
l.
8
,
No.
4
,
Decem
ber
2019
:
256
–
268
262
4.
PROBLE
M
F
ORMUL
ATI
ON
Pr
ovi
de
a
sta
t
e
m
ent
that
w
ha
t
is
e
xpect
ed
,
as
sta
te
d
in
t
he
"
I
ntrod
uction"
c
hap
te
r
ca
n
ulti
m
a
te
l
y
resu
lt
in
"R
es
ults
an
d
Disc
ussi
on
"
c
ha
pter
,
so
the
re
is
c
om
patibil
ity.
More
ov
e
r
,
it
c
an
al
so
be
a
dded
t
he
pros
pect
of
th
e
de
velo
pm
ent
of
resea
rch
r
esults
a
nd
a
ppli
cat
ion
pr
os
pe
ct
s
of
furthe
r
stud
ie
s
into
th
e
ne
xt
(b
ase
d o
n resu
l
t and disc
us
si
on)
.
4.
1.
Ob
jecti
ve
f
unc
tio
n
Accor
ding
to
a
pp
e
ndix B
the
equ
at
io
n o
f D
-
con
t
our
is
giv
e
n by
F(
z)
=
Re
(z
)
–
m
in[
-
ζ
⃓
Im
g(
z
)
⃓
,
a]
=
0
(25)
Wh
e
re z
ϵ
A
is
a point
on th
e
D
-
c
onto
ur
a
nd
A
r
ep
rese
nts th
e com
plex
pla
ne
.
Def
i
ning J
a
s
:
J=m
ax[
Re
(
i
)
–
m
in(
-
ζ
|Im
i
|
,
a)]
(26)
i = 1
,
2
,
3......
n
Wh
e
re
n
is
th
e
num
ber
of
E
igen
val
ues
.
i
is
the
i
-
th
E
ige
n
value
of
t
he
s
yst
e
m
at
an
operati
ng
point.
A
neg
at
ive
value
of
J
im
plies
t
hat
al
l
the
E
ig
en
values
li
e
on
the
le
ft
of
the
D
-
c
on
t
our.
If
J
is
po
sit
iv
e
that
i
m
plies
E
igen value is
ly
ing
on the
r
i
gh
t
of c
on
t
our.
On
these
f
act
s
obj
ect
ive
fun
ct
ion
F ca
n be
de
fine
d
as
:
F=
0
0
J
if
J
if
J
(27)
Wh
e
re
⍺
is a la
rg
e
posit
ive
num
ber
.
The o
pti
m
iz
ati
on pr
ob
le
m
can now
be
sta
te
d
as:
Mi
ni
m
iz
e F S
ubj
ect
t
o
,
m
i
n
m
a
x
m
i
n
m
a
x
m
i
n
m
a
x
m
i
n
m
a
x
p
i
p
i
p
i
l
i
l
i
l
i
D
i
D
i
D
i
i
i
i
K
K
K
K
K
K
K
K
K
R
R
R
(28)
1
ii
i
BD
R
=+
(29)
Wh
e
re
K
pi
,
K
li
,
K
Di
are
PID
con
t
ro
ll
er
pa
ra
m
et
ers
R
i
an
d
B
i
are
s
pee
d
c
har
ac
te
risti
cs
a
nd
area
bias
f
act
or
resp
ect
ively
.
i=
1
,
2 f
or
c
ontr
ol ar
ea
f
irst
an
d
sec
ond res
pe
ct
ively
.
4.2.
Al
go
ri
thm
Step1.
In
it
ia
li
ze
the
se
t
of
p
a
rtic
le
s
w
it
h
posit
ion
v
al
ue
(
popula
ti
on
)
,
it
s u
pper
lim
it
and
l
ow
e
r
li
m
it
,
ran
dom
velocit
ie
s
,
ine
r
ti
a w
ei
gh
t
,
acc
el
era
ti
on
c
onst
ants
,
it
erm
ax
,
e
tc
.
Step2.
Linearize
syst
e
m
an
d
cal
culat
e the eig
en
v
al
ues for eac
h pa
rtic
le
f
r
om
the system
m
od
el
lin
g.
Step3.
Ca
lc
ulate
the obj
ect
iv
e f
unct
ion J.
Step4.
Set
al
l
po
sit
io
n
va
lues
as
loc
al
best
val
ues
and
fitness
val
ues
of
obj
ect
iv
e
functi
on
as
l
ocal
fitness
.
Find gl
ob
al
fitness a
nd it
s c
orrespo
nd
i
ng pos
it
ion
v
al
ue.
Step5.
Set i
te
r=
1
a
nd
com
pu
te
iner
ti
a w
ei
ght
by
m
a
x
m
i
n
m
a
x
m
a
x
*
ww
w
w
i
t
e
r
i
t
e
r
−
=−
(30)
Update
velocit
y usin
g velocit
y update e
qu
at
i
on for al
l pa
rtic
le
s b
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
R
ob
&
A
uto
m
IS
S
N:
20
89
-
4856
Op
ti
miz
atio
n o
f PID co
ntr
oller p
arameters
usi
ng PSO f
or
t
wo... (B
rij
esh K
. Dubey)
263
1
1
2
1
(
)
2
(
)
t
t
t
t
t
t
i
d
i
d
i
d
i
d
t
t
t
d
i
d
V
v
c
R
p
b
e
st
x
c
R
g
b
e
st
x
+
=
+
−
+−
(31)
A
lso
chec
k
it
s
uppe
r
a
nd lo
w
er lim
it
s.
Step6.
Update
posit
ion
value o
f parti
cl
es b
y
11
t
t
t
id
id
id
X
X
V
++
=+
(32)
Also
chec
k
it
s
uppe
r
a
nd lo
w
er lim
it
s.
Step7.
Linearize
syst
e
m
and
cal
culat
e
E
igen
val
ues
for
eac
h
par
ti
cl
e
and
c
orres
pondin
g
fitne
s
s
value
s
of
obj
ect
ive
fun
ct
ion
.
Step8.
Now
c
om
par
e
these
fitness
va
lues
to
the
previo
us
fitne
ss
values
.
Mi
nim
um
fitness
values
will
be
sel
ect
ed
as
l
oc
al
fitness
val
ues
a
nd
it
s
c
orres
pondin
g
posit
ion
val
ues
as
local
best
valu
es.
Fin
d
m
ini
m
u
m
o
f
al
l fit
nesses
,
e.
g
.
global
fitness
value
a
nd it
s c
orres
pondin
g p
os
it
ion val
ue.
Step9.
If
num
ber
of ite
rati
on r
eac
hes
it
erm
ax
go
t
o st
ep 10
,
oth
er
w
ise
go to st
ep 5
.
Step10.
Partic
le
w
it
h
m
ini
m
u
m
f
it
ness
v
al
ue
is the
op
tim
u
m
p
arti
cl
e.
4.3.
Re
s
ult
PI
D
p
a
ram
et
ers
an
d
s
pee
d
c
ha
racteri
sti
cs of
bo
t
h
a
reas a
re
al
lowed to
va
r
y wit
hin
f
ollo
wing
ranges
:
5
.
0
0
5
.
0
0
5
.
0
0
5
.
0
0
5
.
0
0
5
.
0
0
0
.
0
2
0
.
0
8
p
l
D
K
K
K
R
−
−
−
(33
)
Ti
m
e con
sta
nts of diffe
ren
t m
od
el
c
om
po
ne
nt
s asso
ci
at
ed
wi
th both
a
reas
a
re
giv
e
n
in
app
end
i
x A.
The
de
sired
P
S
O
pa
ram
et
ers
are
giv
e
n
in
T
a
ble
1.
w
it
h
t
he
desire
d
set
ti
ng
of
dam
pin
g
rat
io
‘ζ’
(
0.5
)
and
re
fer
e
nce
li
ne
value
‘a
’
(
-
1.5)
al
gorithm
is
r
un
se
ver
al
t
i
m
es.
Am
on
g
al
l
the
r
uns
m
os
t
optim
um
ru
n
a
n
d
it
s
corres
ponding
pa
ram
et
ers
are
sel
ect
ed
.
O
pti
m
iz
ed
pa
ra
m
et
ers
an
d
c
orrespo
nd
i
ng
E
ig
en
values
as
show
n
i
n
Table
1
-
3.
Table
1
.
O
pti
m
iz
ed
pa
ram
et
ers
K
p1
-
1.8
787
K
l1
-
2.3
640
K
D1
-
0.9
364
R
1
0.065
2
B
1
15.93
01
K
p2
-
0.1
353
K
l2
-
0.5
873
K
d2
-
0.5
212
R
2
0.065
3
B
2
16.20
87
Table
2.
O
pti
m
iz
ed
Eige
n val
ues
(
o
s
ci
ll
at
or
y
m
od
e)
Eigen
values
Dam
pin
g
-
1.4
953 +
2.4192
i
,
-
1.4
953
-
2.419
2
i
0.525
8
-
1.4
934 +
1.3570
i
,
-
1.4
934
-
1.357
0i
0.740
1
-
1.4
762 +
0.8555
i
,
-
1.4
762
-
0.855
5i
0.865
2
-
1.4
887 +
0.5015
i
,
-
1.4
887
-
0.501
5i
0.947
7
-
0.2
653
,
-
0.0
600
,
-
0.1
125
1
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
I
nt
J
R
ob
&
A
uto
m
,
Vo
l.
8
,
No.
4
,
Decem
ber
2019
:
256
–
268
264
Table
3.
PS
O p
arm
et
ers
C1
2
C2
2
Wmax
0.9
0
Wmin
0.1
0
Vm
ax
100%
o
f
P
m
a
x
Vmin
100%
o
f
pmin
Po
pu
l
ation
500
Iteratio
ns
1000
C1
2
5.
SIMULATI
O
N
ST
UDY
The
syst
em
is
si
m
ulate
d
in
MATLAB
e
nviro
nm
ent
for
s
te
p
c
ha
ng
e
in
load
(
0.2
pu
)
i
n
a
rea
fir
st
,
giv
e
n
at
one se
cond
a
nd
is c
he
cked
out for
50 seco
nds.
C
om
par
ison
of r
e
sp
onses
of
1
w
,
12
P
,
AC
E1
,
1
m
P
,
2
w
,
ACE
2
,
2
m
P
between
c
onve
ntion
al
i
nteg
ral
co
ntr
ol
a
nd
PI
D
PS
O
is
c
arr
ie
d
ou
t
as
sho
wn
i
n
Figure
6
-
12.
Figure
6. Fr
e
quency
de
viati
on
respo
ns
e
of a
rea f
i
rst w
it
h i
nt
egr
al
c
on
t
ro
l
(
das
hed li
ne)
a
nd P
ID co
ntr
ol
(so
li
d l
ine)
Figure
7. P
ow
e
r
interc
ha
nge
r
esp
on
se
of a
re
a first
with i
ntegr
al
c
ontrol
(
da
sh
e
d
li
ne) an
d PI
D
c
on
t
ro
l
(so
li
d l
ine)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
R
ob
&
A
uto
m
IS
S
N:
20
89
-
4856
Op
ti
miz
atio
n o
f PID co
ntr
oller p
arameters
usi
ng PSO f
or
t
wo... (B
rij
esh K
. Dubey)
265
Figure
8.
A
rea
con
t
ro
l
e
rror (
ACE)
res
pons
e
of ar
ea
f
ir
st wi
th integ
ral c
on
t
ro
l
(d
a
sh
e
d
li
ne
)
a
nd PID c
on
trol
(so
li
d l
ine)
Figure
9. De
vi
at
ion
in
m
echan
ic
al
pow
e
r res
pons
e
of a
rea fi
rst w
it
h
i
nteg
r
al
co
nt
ro
l
(das
hed li
ne) an
d P
ID
con
t
ro
l
(s
olid l
ine)
Figure
10.
Fr
e
qu
e
n
cy
dev
ia
ti
on r
es
ponse
of
area sec
ond
wi
th integ
ral c
on
t
ro
l
(d
a
sh
e
d
li
ne
)
a
nd PID c
on
trol
(so
li
d l
ine)
Evaluation Warning : The document was created with Spire.PDF for Python.