TELKOM
NIKA
, Vol. 11, No. 12, Decem
ber 20
13, pp.
7222
~72
2
9
e-ISSN: 2087
-278X
7222
Re
cei
v
ed
Jun
e
26, 2013; Revi
sed
Jul
y
2
8
, 2013; Acce
pted Augu
st 20, 2013
Optimizing Multi-agent MicroGrid Resource
Scheduling by Co-Evolutionary with Preference
Sun Hongbi
n*
1
, Tian Chunguang
2
1
School of Elec
trical Eng
i
ne
eri
ng an
d Informa
tion,
Cha
ngc
hu
n Institute of
T
e
chn
o
lo
g
y
, Ch
angc
hu
n,
Jinli
n
, 130
01
2, Chin
a, T
e
lepho
ne: +
86-43
1-8
5
713
80
5
2
Electric Po
w
e
r Researc
h
Institute of Jilin El
e
c
tric
Po
w
e
r Co.
,
Ltd., Changch
un, Jinli
n
, 130
021, Ch
in
a,
T
e
lephon
e: +
86-43
1-85
71
39
9
8
*Corres
p
o
ndi
n
g
author, e-ma
i
l
:
w
i
n_s
hb@
16
3.com
A
b
st
r
a
ct
T
h
is p
aper
pr
esents
a
multi
-
age
nt fra
m
ew
ork for
th
e c
o
ntrol
of distri
b
u
ted
en
ergy r
e
sourc
e
s
orga
ni
z
e
d in Microgri
d
s
,
w
h
ich cons
ists of integrate
d
mi
crogrids
and l
u
mpe
d
loa
d
s. Multipl
e
ob
jecti
v
es
are co
nsi
dere
d
for lo
ad
ba
lan
c
ing
a
m
o
ng th
e fee
ders,
mi
ni
mi
z
a
t
i
o
n
of th
e
op
eratin
g cost
, mi
ni
mi
z
i
ng
th
e
emissio
n
, min
i
mi
z
i
ng
volta
ge profil
es,
mini
mi
z
i
n
g
active
po
w
e
r losses. T
h
e ag
ent re
pres
ents
messa
ge
of
micr
oGrid unit and
c
onstitutes
an auto
n
o
m
ic unit.
T
he netw
o
rk is ac
hi
eved
by the
evo
l
uti
on of th
e a
g
e
n
t
base
d
on th
e
semantic n
e
g
o
t
iation. Bas
ed
on the o
b
j
e
ct
iv
es is eva
l
uat
e
d
by me
mb
ers
h
ip fu
nctions.
We
prop
ose a
ne
w
Immu
ne C
o
-Evol
u
tion
ary
Algorit
hm w
i
t
h
Prefere
n
ce
to solve it. Si
mu
lati
on res
u
l
t
s
de
mo
nstrated t
hat the
prop
os
ed
meth
od
is e
ffective in i
m
pr
ovin
g perf
o
rma
nce a
nd
mana
ge
me
nt of micr
o-
sources.
Ke
y
w
ords
: mi
crogrid,
mu
lti-a
gent, opti
m
i
z
at
i
on mo
d
e
l, co-e
voluti
onary w
i
th prefere
n
ce.
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
As the ba
ckb
one of the p
o
w
er
network, the ele
c
tricity
grid i
s
no
w a
t
the focal poi
nt of
techn
o
logi
cal
innovation
s
[1]. The intelligent g
r
id
achi
eves o
p
e
ration
al efficien
cy thro
u
gh
distrib
u
ted
co
ntrol, monito
ring and
ene
rgy manag
em
ent. The ne
e
d
for mo
re fl
exible ele
c
tri
c
system
s,
cha
nging
re
gulat
ory a
nd
eco
nomic
scena
rios,
en
ergy saving
s and
environ
menta
l
impact
are p
r
oviding im
pet
us to
the
dev
elopme
n
t of
MicroGri
ds (MG),
whi
c
h
a
r
e
pre
d
icte
d t
o
play an in
cre
a
sin
g
rol
e
of the future p
o
w
er
sy
ste
m
s.
The MG u
n
its can me
et to the cu
stom
ers
load
dema
n
d
at comp
romi
se
co
st a
nd
emission
s
all
the time.
M
G
can
co
ntai
n vario
u
s
cle
an
and effici
ent
energy re
so
u
r
ce
s,
su
ch
as sola
r p
hotov
oltaic
(PV) m
odule
s
, small
wind
turbi
n
e
s
,
battery stora
ges, controll
able loa
d
s
and othe
r small ren
e
wa
ble, and it has a
n
ene
rgy
manag
eme
n
t
system
to reg
u
late p
o
wer fl
ow i
n
it
a
nd
p
r
ovide
co
nsi
d
erabl
e
cont
rol
over it. It can
not only be o
perate
d
effici
ently in its o
w
n di
stri
b
u
tio
n
network, bu
t also b
e
ca
pable to o
p
e
r
ate
in islan
d
ing m
ode when it is requi
red o
r
some
faults ha
ppen in u
p
stream network
[2].
Con
c
u
r
rently, the power
system re
sea
r
che
r
s fo
cu
s on the poten
tial value of
multi-
agent sy
stem
(MAS) tech
n
o
logy to the power in
d
u
st
ry [3]. These rece
nt re
sea
r
ch wo
rks have
shown that M
AS is one of
the best technologies
for int
r
oducing di
stri
buted i
n
telligenc
e in power
system
s. Co
ordin
a
ting be
havio
r
of au
tonomou
s
ag
ents i
s
a
ke
y issue i
n
a
gent-o
rie
n
ted
techni
que,
which l
ead
s th
e MAS towa
rds the
syste
m
goal. MAS
is be
comi
ng
a sig
n
ifica
n
t and
gro
w
ing
interest in
po
wer engin
e
e
r
ing
pro
b
lem
s
[4
-5]. Energy reso
urce
sch
edulin
g is a
n
importa
nt opt
imization t
a
sk in the
dail
y
oper
ation
planni
ng of
any po
wer system, whi
c
h
is
typically han
dled by po
we
r syste
m
ma
nage
rs. Typi
cally, the probl
em is to mini
mize the
co
st
s
asso
ciated
with energy p
r
odu
ction, an
d start-
up a
nd sh
ut-do
w
n co
sts. It is a large
sca
l
e
nonlin
ear o
p
timization p
r
o
b
l
em for whi
c
h
,
there is no
exact sol
u
tio
n
techni
que [
6
]. Most of the
resea
r
ch wo
rk on u
n
it com
m
itment ha
s been d
one
i
n
centralized a
ppro
a
ch [7-9]
,
wherea
s very
little work h
a
s
be
en d
one
in dist
ributed
app
roa
c
h [1
0-11]. A ratio
nal meth
od o
f
building
MG
s
optimize
d
for
co
st and subj
ect to reliabilit
y cons
traints
have been p
r
ese
n
ted
in [12] the proble
m
of manag
em
ent of MG so
lved as
singl
e obje
c
tive and witho
u
t co
nsid
erin
g the
balan
cing
wi
th
the uppe
r gri
d
. In this pap
er, the age
nt approa
ch
is p
r
esented to h
andle the
s
e
challen
ges. Th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Optim
i
zing M
u
lti-age
nt MicroG
rid Resou
r
ce S
c
he
dulin
g by Co-E
vol
u
tionary (Su
n
Hong
bin)
7223
formulatio
n o
f
the MG control mo
del
and fuzzy prefe
r
en
ce
s
co-evolution
a
r
y algorithm
is
prop
osed to
re
solve
the
problem.
T
he
simulatio
n
results
sh
ow th
at the
app
ro
ach
can
signifi
cantly improve p
e
rfo
r
man
c
e a
nd a
dapt we
ll to the ch
ang
es o
f
dynamic en
vironme
n
ts.
2. MG Agen
t Neg
o
tia
t
ion
and Commu
nication Me
c
h
anism
As an atomi
c
unit of MG
control sy
stem, the MG
agent in
clu
des th
ree m
odule
s
.
Attributes de
scribe the ch
ara
c
teri
stic of
an agent
itself. Function
is desi
gne
d to evaluate the
matchin
g
a
b
il
ity of the me
ssage t
o
the
other
MG
m
o
bile ag
ents.
Behavior
co
n
t
ains inte
rfa
c
e
operation, inf
o
rmatio
n issue, and
en
ergy tran
smi
ssi
on. MG
age
nt
is a
n
ato
m
ic u
n
it of
MG
control pl
atform (M
G
C
P),
MGCP i
s
compo
s
ed
of
function
al m
odule
s
d
e
vel
oped
by java.
Inspired
by systemic
network, i
n
the e
n
v
ironme
n
t,
different a
gent
s may co
ntrib
u
te to differe
nt
servi
c
e
s
. MG optimizatio
n control result is achi
eved by
service co
m
positio
n of MG agent. It is a
novel comput
ing an
d probl
em-solving
e
n
vironm
ent
where an
a
ppli
c
ation se
rvice
is cre
a
ted out
of the intera
ction of multiple awa
r
e ag
en
ts
and the int
e
ra
ction bet
ween a
w
are ag
ents an
d their
environ
ment.
The id
eal m
o
del would
pla
c
e th
e platfo
rm on
every
MG unit
as a
network n
o
d
e
,
and fun
c
tiona
l merits refer to our p
r
eviou
s
wo
rk [1
3].
In orde
r to co
llaborate amo
ng age
nts, a
se
t of co
mmu
nicatio
n
me
chani
sm is n
e
eded.
The MG a
g
e
n
ts u
s
e RMI-IIOP as tra
n
sp
ort
p
r
otocol for comm
unication lan
guag
e (BNCL
)
messag
es.
RMI-IIOP provi
des th
e robu
stne
ss
of
CO
RBA and th
e
simpli
city of Java
RMI. We
pre
s
ent th
e
method
of M
G
ag
ent me
ssage
discov
ery ba
se
d o
n
the m
e
ssa
ge mat
c
hin
g
. A
matchin
g
me
ssage i
s
exchang
ed amo
ng agent
s to achi
eve MG agent me
ssa
ge matchi
ng
for
control model
. Based on th
e method of messag
e di
scovery in workflow, Semanti
c
discove
r
y of
atomic p
r
o
c
e
s
ses, delive
r
s a set of MG agent
th
at provide at
omic p
r
o
c
e
s
se
s whi
c
h a
r
e
sema
ntically
matchin
g
with tho
s
e of th
e age
nt
me
ssag
e, the o
p
timization
of
MG is a
c
hiev
ed
based on the
use of ontol
o
g
y to describ
e tasks an
d a
gent messa
g
e
.
3. Problem
Defini
tion fo
r MG Con
t
rol
Optimizatio
n
Model
The MG con
s
ist
s
of a gro
up of radi
al f
eede
rs, which coul
d be p
a
rt of a distri
butio
n
system. Th
e feede
rs
also h
a
ve the micro
sou
r
ces
co
n
s
istin
g
of a p
hotovoltaic, a
wind tu
rbin
e, a
fuel cell, a m
i
cro
turbine,
a die
s
el
gen
erato
r
, an
d b
a
ttery sto
r
ag
e. To
se
rve t
he lo
ad
dem
an
d
and ch
arge th
e battery, electri
c
al po
wer
can be p
r
o
d
u
c
ed eithe
r
directly by PV,
WT, DG, MT,
or
FC. Each co
mpone
nt of the MG
syste
m
is mod
e
led
sep
a
rately b
a
se
d on its
chara
c
te
risti
c
s and
constrai
nts. T
he MG agent
s interact as to utilize the
maximu
m quantity of ava
ilable generati
on
possibl
e. This is considered a ma
ximum
power utilizat
ion strategy.
The m
a
jor concern i
n
the
desi
gn
of an
electr
i
c
al
sy
stem that
ut
ilizes M
G
sources i
s
the a
c
curate
sel
e
ctio
n
of output
po
wer th
at can
eco
nomi
c
ally
sati
sfy the
load
dem
and
,
Minimization
of the Co
st (t
he Op
eratin
g
Co
st, ac
tive powe
r
lo
sse
s
) mi
nimizi
ng
the emissio
n
.
Minimizi
ng, h
ence the sy
stem co
mpon
e
n
ts are foun
d
subje
c
t to: T
he network reco
nfiguratio
n
probl
em in
a
distri
bution
system i
s
to
find a
config
uration
with
minimum lo
ss an
d minim
u
m
deviation of the nod
es vol
t
age whil
e sa
tisfying t
he o
peratin
g con
s
traints u
nde
r
a ce
rtain loa
d
pattern. The
operating co
nstrai
nts a
r
e curre
n
t c
apa
city and radi
al operat
ing
stru
cture of the
system. The
mathemati
c
al
formul
at
ion
reco
nfiguratio
n p
r
oble
m
i
s
pre
s
ente
d
in
the literature
in
different way
s
. In this
paper, the pr
o
b
le
m formulation
is pre
s
ente
d
as:
i
L
i
i
i
i
i
loss
V
Q
P
r
F
1
2
2
2
(1)
loss
F
is the memb
ership fun
c
tio
n
for ac
tive power loss
es
,
i
r
represents t
he re
sista
n
ce
of the bran
ch
i.
i
P
,
i
Q
repre
s
e
n
t active power an
d rea
c
ti
ve powe
r
tha
t
flowing the termin
al of
the bran
ch i.
i
V
rep
r
e
s
ent
s t
he no
de volt
age of th
e termin
al of b
r
anch i.
i
L
repre
s
ent
s the
numbe
r of branche
s. Voltage variatio
n
may be
cau
s
ed by the Distribute
d
Ge
neratio
n outp
u
t
cha
ngin
g
.
The obje
c
tive
function is
d
e
velope
d accordin
g to the above mentio
ned a
s
sumpti
ons to
minimize the operating cost in $/h of the
MG in the followin
g
form:
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 12, Dece
mb
er 201
3: 722
2 – 7229
7224
N
i
i
i
i
ib
i
i
non
i
non
i
i
OM
F
C
C
F
P
F
P
F
P
F
P
F
F
1
_
_
cos
)
(
)
(
)
(
/
))
(
)
(
(
(2)
Whe
r
e F
(
Pi)
The op
eratin
g co
st of the
gene
ra
ting
unit i in $/h,
Ci Fuel
co
sts of the
gene
rating
u
n
it i in $/l fo
r
the DG, an
d i
n
$/kWh
for
FC a
nd M
T
,
Fi Fuel
co
nsu
m
ption
rate o
f
a
gene
rato
r
u
n
i
t
i,
OMi Ope
r
ation
a
n
d
mai
n
tenan
ce cost of a generating unit i in $/h.
)
(
_
non
i
P
F
is
operating cost without DG,
ib
i
C
F
is
distri
bution netwo
rk co
st.
The Co
st is
calcul
ated ba
sed on the Op
erating
Co
st and a
c
tive po
wer lo
sse
s
.
co
AF
is
define a
s
:
loss
co
F
F
AF
)
1
(
cos
(3)
Whe
r
e
]
1
,
0
[
is a weight to
co
AF
.
The atmo
sp
h
e
ric
polluta
nts such a
s
sulphu
r oxide
s
SO2,
carb
on oxide
s
CO2, and
nitroge
n oxid
es
NOx
cau
s
ed by fo
ssilfu
e
led the
r
mal
units
can
be
modele
d
sep
a
rately. The
total
emission of these polluta
n
t
s can b
e
exp
r
esse
d as:
)
exp(
)
(
10
)
(
)
(
/
)
(
2
1
2
_
i
i
i
i
i
N
i
i
i
i
i
non
i
i
po
P
P
P
P
E
P
E
P
E
E
(4)
W
h
er
e
α
,
β
,
γ
,
ζ
, a
nd
λ
are
no
nneg
ative co
efficien
ts of th
e ith
gene
rato
r e
m
issi
on
cha
r
a
c
teri
stics. In the emission mo
del in
trodu
ced, we prop
ose to evaluate the pa
ramete
rs
α
,
β
,
γ
,
ζ
, an
d
λ
u
s
ing th
e data
available. T
hus, the
emi
ssi
on p
e
r
da
y for the
DG,
FC, an
d MT
is
estimated, an
d the cha
r
a
c
teristi
cs of ea
ch gene
rato
r will be detache
d accordingly
.
Based o
n
th
e fuzzy eval
uation fun
c
tions, the mul
t
i-obje
c
tive optimization
model i
s
con
s
t
r
u
c
t
ed t
o
max
i
mize t
he sat
i
sf
a
c
t
i
o
n
s of
differen
t
objectives b
y
adjusting transfo
rme
r
tap-
cha
nge
rs a
n
d
shunt capa
ci
tors. The m
u
lti-obje
c
tive op
timization mo
del is re
prese
n
ted as:
K
k
Q
Q
P
P
P
P
P
K
k
Q
Q
K
i
P
P
K
i
P
P
MepF
F
Mep
Subject
E
Mep
AF
Mep
Min
j
j
N
i
batt
WT
PV
L
i
j
i
i
i
i
po
po
co
max
1
max
min
0
0
)
(
)
(
))
(
),
(
(
(5)
Whe
r
e Po
we
r balan
ce
co
nstrai
nts a
r
e
that
it meets the active
power bal
an
ce, an
equality co
nst
r
aint is imp
o
sed.
N
i
batt
WT
PV
L
i
P
P
P
P
P
1
0
)
(
(6)
PL The total power dem
anded in kW, PPV The
output power of the photovoltaic
cell in
kW, P
W
T
Th
e outp
u
t po
wer of th
e
win
d
turbi
ne i
n
kw. Pbatt Th
e
output p
o
we
r of the
batte
ry
stora
ge kw.
Gene
ration
capa
city con
s
traints i
s
re
st
ricted by lo
wer and u
ppe
r limits for stable
operation,real
powe
r
output
of each ge
ne
rator, as follo
ws:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Optim
i
zing M
u
lti-age
nt MicroG
rid Resou
r
ce S
c
he
dulin
g by Co-E
vol
u
tionary (Su
n
Hong
bin)
7225
N
i
P
P
P
i
i
i
,.......
1
.
max
min
(7)
min
i
P
Minimum o
p
e
r
ating p
o
wer
of unit i,
max
i
P
Ma
ximum ope
ra
ting po
wer
of unit i,
i
P
,
max
,
i
P
represent the real runni
n
g
power an
d the maximum
permitted po
wer of the tra
n
sformer.
i
a
is penalty function p
a
ra
m
e
ter.
po
F
is
the membership func
tion for power.
b
i
i
i
n
i
P
P
K
i
i
i
po
a
P
P
F
1
)
(
max
,
)]
1
(
[
max
,
(8)
4. Fuzz
y
Pre
f
eren
ces
Co-Ev
o
lutionar
y
Algorithm
There a
r
e
ma
ny MO
soluti
on al
gorith
m
s allo
wing th
e
attainment of
these results, like
PESA-II [14],
NSGA-II [15]. An important is
s
u
e in
multiple objec
tive optimiz
ations
is
the
handli
ng of h
u
man p
r
efe
r
e
n
ce
s. Fin
d
ing
all Pareto
-o
p
t
imal solutio
n
s
is
not the fi
nal go
al. Such
prefe
r
en
ce
s can
u
s
ually be rep
r
e
s
ent
ed with
th
e
help
of fuzzy
logi
c. Ba
sed
on
prefere
n
c
e
relation
s [16-17] and indu
ced o
r
de
rs, these lingui
stic cate
gori
e
s
were tran
sformed into re
a
l
weig
hts an
d a weig
hted Pareto do
mina
nce relation
wa
s introd
uced.
In this pa
pe
r, the
novel
fuzzy p
r
efe
r
en
ce
s
evolu
t
ionary
algo
rithm (FP
-
EA) i
s
prop
osed. Su
ppo
se that th
e size of evol
utionary
p
opu
lation P is n, and Pt is t-th
gene
ration of
the popul
atio
n. Qt is a ne
w evol
ution
a
ry population
from Pt that
is upd
ated by
the sele
ction
,
cro
s
sove
r an
d mutation
o
perato
r
s, and
t
he
si
ze of Q,
is
al
so n. Let
Rt
=Pt
∪
Qt, and the
si
ze of
Rt is 2n. The non-domin
ated set P1 is gene
rat
e
d
from Rt, with the quick sort p
r
o
c
edu
re.
If|P1|>n, the clu
s
terin
g
p
r
o
c
ed
ure
is
use
d
to re
du
ce t
he si
ze
of P1
, and to
kee
p
the diversity of
P1 at the same time. T
h
e si
ze of P1
will
be n
after the cl
ust
e
ring process.
is equally
important,
is less important
,
is much less important,
is
not important,
!
is
important.
Definition
1:(Weig
h
ted d
o
m
inan
ce
relat
i
on) F
o
r a
given weight
s–v
e
ctor
)
....
1
k
w
w
w
summi
ng
to 1
and
a real
numbe
r
1
0
, a re
al vecto
r
)
,
)(
....
1
w
x
x
x
k
–
dominate
s
a
real
vec
t
or
)
....
1
k
y
y
y
written as
y
x
w
if and only if:
)
y
,
(x
I
y
x
i
i
1
k
i
i
w
w
(9)
Whe
r
e
y
x
y
x
y
x
I
0
1
)
,
(
The sta
nda
rd definition
of domina
n
ce co
uld be
obtaine
d by setting
1
and
k
w
w
n
/
1
...
1
. Note that in the stand
ard
definition of
domina
n
ce it is re
quired th
at at least
one of the
i
i
y
x
inequalitie
s i
s
stri
ct. Ho
wev
e
r thi
s
is
not a
problem si
nc
e th
ese two orders
are defin
able
in terms of ea
ch othe
r.
Definition
2:
(Weig
h
ted
sco
r
e).
The
nu
m
ber nin
e
i
s
u
s
ed h
e
re fo
r th
e g
r
ad
es of
relative
importa
nce betwe
en
o
b
j
e
ctives be
ca
use we
take the
well-known techni
que of
anal
ytic
hiera
r
chy proce
s
s (A
HP) for refe
ren
c
e. For e
a
ch
X
x
i
comp
ute weight a
s
no
rmalize
d
leaving sco
r
e
.
X
x
j
i
i
j
R
x
SL
R
x
SL
x
w
)
,
(
)
,
(
)
(
(10)
Definition 3:
(Fitness ev
aluation
)
. Suppo
se
there
are N indiv
i
dual
s in the
current
popul
ation po
p. The po
sitive stren
g
th
)
(
k
x
S
of each
individu
al
pop
x
k
.
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 12, Dece
mb
er 201
3: 722
2 – 7229
7226
)
,
2
,
1
(
N
k
is cal
c
ulated. Suppo
se
))
(
(
min
,
,
2
,
1
min
k
N
k
x
S
S
,
)
(
max
,
,
2
,
1
max
k
N
k
d
d
. The fitnes
s
of each in
dividual
)
,
,
2
,
1
(
N
k
pop
x
k
is calcul
ated acco
rdin
g to the following formulat
ion:
2
max
min
)
/
(
)
1
)
(
(
)
(
d
d
S
x
S
x
fit
k
k
k
(11)
Algorithm : FP-EA Algorith
m
Pt , t = 0 ;//
Set t = 0. Generat
e an i
n
itial population P[t], for each
X
x
i
comp
ute
weig
ht as no
rmalize
d
leavi
ng score:
X
x
j
i
i
j
R
x
SL
R
x
SL
x
w
)
,
(
)
,
(
)
(
While ( t
≤
T
)
do //T is maximum numb
e
r of generatio
n
s
{
2
max
min
)
/
(
)
1
)
(
(
)
(
d
d
S
x
S
x
fit
k
k
k
//
Cal
c
ul
ate the fitness value of each individual in
Pt,
)
,
,
2
,
1
(
N
k
P
x
t
k
Qt = make
-n
ew-pop
(Pt ) // Use sele
ction, cro
s
sover and mutat
i
on to cre
a
te
a new
popul
ation Qt
Rt = Pt
∪
Qt // Combin
e pa
rent and children pop
ulation
I
f
(|
P
t
+
1 |
<
=
N)
Th
en
{
P
t
+ 1
=
P
t
+
1
∪
sele
ct
- b
y
- r
and
om
(
Rt -
Pt +
1,
N
- | Pt
+
1 | ) } // randomly selecte
d
N - | Pt + 1 |
element
s and
joined into Pt + 1
Els
e
if (| Pt +
1 | >
N)
Then {
c
rowdi
ng - dista
n
ce - assign
ment
(Pt + 1)
// Calculate
crowdi
ng di
stan
ce.
Sort (Pt +
1 ,
≥
n) // Sort in desce
nding
o
r
de
r usi
ng
≥
n
Pt + 1 = Pt + 1 [1: N]} // Choose the first
N eleme
n
ts
t =
t +
1}
It can be
pro
v
ed that the time co
mplexi
ty of
Algorithm (FP-EA) i
s
less than
O (nlogn
). It
is better than
O (n2) in the NSGA II.
5. Results a
nd Analy
s
is
This sample
system
is used to
sim
u
lat
e
t
he transfo
rmer lo
ading
s, line flow profiles,
and
syste
m
l
o
sse
s
of th
e
microgri
d
. Be
side
s, the
pa
ramete
rs of t
he di
strib
u
tio
n
tra
n
sfo
r
me
r,
con
d
u
c
tor, g
eneration, a
nd load a
r
e
describ
ed
i
n
the followi
ng su
bsectio
n
s. The rela
ted
para
m
eters for sim
u
lation
of the MV/L
V distribut
io
n
transfo
rme
r
are liste
d in Table 1. Thi
s
transfo
rme
r
i
s
4
0
0
k
VA, 20
kV/0.4kV, a
n
d
its l
e
a
k
age
impeda
nce i
s
0.01
+j0.04
pu
. The lo
catio
n
s
and
ca
pa
cities of th
e
DG
s
interconn
ecte
d to the
net
work a
r
e
as foll
ows: A 10
kW p
hotovolta
i
c
gene
ration
systems a
nd
a 10
kW
win
d
turbin
e ge
nerato
r
a
r
e
con
n
e
c
ted. A 10kW fuel
cel
l
gene
ration
system is con
necte
d to system with
three-p
h
a
s
e inv
e
rter. A 30 kW microtu
r
bi
ne
gene
rato
r is conne
cted to system with th
ree
-
ph
ase inverter.
Figure 1. Dail
y Load Cu
rve
s
for the
Th
re
e Load Type
s of the Microg
rid
De
m
a
n
d
(%)
Ti
m
e
(
h
r
)
Res
i
d
e
n
t
i
a
l
I
n
dus
t
r
i
a
l
C
o
m
m
e
r
ci
al
De
m
a
n
d
(%)
Ti
m
e
(
h
r
)
Res
i
d
e
n
t
i
a
l
I
n
dus
t
r
i
a
l
C
o
m
m
e
r
ci
al
Res
i
d
e
n
t
i
a
l
I
n
dus
t
r
i
a
l
C
o
m
m
e
r
ci
al
Res
i
d
e
n
t
i
a
l
Res
i
d
e
n
t
i
a
l
I
n
dus
t
r
i
a
l
I
n
dus
t
r
i
a
l
C
o
m
m
e
r
ci
al
C
o
m
m
e
r
ci
al
10
0
90
80
70
60
50
40
30
20
10
0
0
3
6
9
1
2
1
5
1
8
2
1
De
m
a
n
d
(%)
Ti
m
e
(
h
r
)
Res
i
d
e
n
t
i
a
l
I
n
dus
t
r
i
a
l
C
o
m
m
e
r
ci
al
De
m
a
n
d
(%)
Ti
m
e
(
h
r
)
Res
i
d
e
n
t
i
a
l
I
n
dus
t
r
i
a
l
C
o
m
m
e
r
ci
al
Res
i
d
e
n
t
i
a
l
I
n
dus
t
r
i
a
l
C
o
m
m
e
r
ci
al
Res
i
d
e
n
t
i
a
l
Res
i
d
e
n
t
i
a
l
I
n
dus
t
r
i
a
l
I
n
dus
t
r
i
a
l
C
o
m
m
e
r
ci
al
C
o
m
m
e
r
ci
al
10
0
90
80
70
60
50
40
30
20
10
0
0
3
6
9
1
2
1
5
1
8
2
1
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Optim
i
zing M
u
lti-age
nt MicroG
rid Resou
r
ce S
c
he
dulin
g by Co-E
vol
u
tionary (Su
n
Hong
bin)
7227
Table 1 .The
MV/LV Distrib
u
tion Tra
n
sfo
r
mer
Capacit
y
(kVA)
Primary
Side (kV)
Secondar
y
Side (kV)
R(pu)
X(pu)
400 20
0.4
0.01
0.04
Table 2. The
Real Po
we
r Output Cu
rve
s
for Fou
r
Types
The po
wer g
eneration
s
of the PV and
WT ar
e cal
c
ulated by the propo
se
d formula
s
with in
sulatio
n
, temperatu
r
e, and
win
d
spe
ed-
rel
a
ted pa
ram
e
te
rs. Additio
nal
ly, the powe
r
gene
rated by
the fuel-cell gene
ration
system and
mi
crotu
r
bi
ne ge
nerato
r
is
cal
c
ulate
d
und
e
r
the minimizat
i
on of total fu
el co
st in the micro
g
rid by
direct search method. Th
ese
curve
s
a
r
e
use
d
as the p
o
we
r gen
eration data for a
full day's anal
ysis.
Based o
n
the evaluatio
n function
s,
the
multi-objective opti
m
ization m
o
del is
con
s
t
r
u
c
t
ed t
o
max
i
mize t
he sat
i
sf
a
c
t
i
o
n
s of
differen
t
objectives b
y
adjusting transfo
rme
r
tap-
cha
nge
rs an
d shunt
cap
a
c
itors.
Th
e coordi
nation
control
strateg
i
es
we
re di
scussed
above,
it
can
be u
s
e
d
to rea
c
h the
target of max
i
mizing th
e e
fficiency of M
i
cro
g
ri
d. In this pa
per, the
novel fuzzy preferen
ce
s evolutiona
ry
alg
o
rithm (FP
-
E
A
) is propo
se
d.
Figure 3
an
d
4
sho
w
s th
e rel
a
tion
ship
of the
co
st
and
emissio
n
obje
c
tives o
f
non
-
dominate
d
so
lutions obtain
ed by m
u
lti-o
b
jectiv
e
opti
m
ization.
The
co
st of th
e
non-domi
nate
d
solutio
n
s thu
s
ap
pea
rs to
be inversely
prop
ortion
al
to their emi
ssi
on
s. It can se
e that the
Pareto optim
al set has a
numbe
r of non-d
o
min
a
ted
solution
s. It
can be
con
c
l
uded that the
prop
osed ap
proa
ch i
s
capabl
e of exploring m
o
re efficient a
nd non
-infe
r
i
o
r solution
s of
optimizatio
n probl
em
s.
As can
be
se
en in
Figu
re
3 and
4, the
dist
rib
u
tion
chara
c
te
risti
c
s of the a
p
p
r
o
x
imate
weig
hted Pa
reto optimal la
yer are different in di
ffere
nt prefe
r
en
ce
s ci
rcum
st
an
ce
s.
Figu
re 3
is
)
(
)
(
po
co
E
Mep
AF
Mep
Pareto
cu
rve
,
the curve
re
flects th
e feat
ure
s
of l
e
ft sp
arse a
nd
righ
t den
se
based on
)
(
po
E
Mep
p
r
eferen
ce.
Fig
u
re
5(c) i
s
)
(
)
(
po
co
E
Mep
AF
Mep
Pareto
cu
rve,
it showed
more o
b
viou
s feature of
left sparse a
nd right
de
n
s
e, wh
en it is more prefe
rre
d obje
c
tive
function
)
(
po
E
Mep
. We
ighted Pa
reto
method
can
obtain a
p
p
r
ox
imate Pareto
optimal
soluti
on in
different preferen
ce
s to m
eet the need
s of decisi
on
make
r.
Figure 2. The
Pareto Opti
mal Front in
Multi-
objec
tive Optimiz
a
tion(A)
Figure 3. The
Pareto Opti
mal Front in
Multi-
objec
tive Optimiz
a
tion(B)
These g
r
ap
h
s
sho
w
very
clea
r
sepa
ration of Pare
to fronts
obt
ained
usi
ng
different
prefe
r
en
ce
s. It perform
s well on the
co
nverge
nc
e a
nd the diversity. The traditional meth
od
s
0
.
70
0
.
7
5
0.
80
0.
85
0
.
90
0.
95
0
0
.
70
0
.
7
5
0.
80
0.
85
0
.
90
0.
95
0
0
.
028
0
.
030
0
.
032
0
.
034
0
.
036
0
.
038
0
.
040
(E
po
)
M
ep
Me
p
(
A
F
co
)
))
(
(
))
(
(
po
co
E
Me
p
f
AF
Me
p
f
0
.
70
0.
7
5
0
.
80
0.
8
5
0
.
90
0.
9
5
0
0.
0
2
8
0.
0
3
0
0.
0
3
2
0.
0
3
4
0.
0
3
6
0.
0
3
8
0.
0
4
0
(E
po
)
Me
p
Me
p
(
A
F
co
)
))
(
(
))
(
(
po
co
E
Me
p
f
AF
Me
p
f
<
Output cu
rves
(Ti
m
e)
0
2
4
6
8
10 12 14 16 18 20 22
24
PWT
0 2 3 0 1 1 2 4 8 6 4 0
0
PPV
0 0 0 0 0 5 6 5 4 3 0 0
0
PFC
10 10 10 10 10 10 10 10 10 10 10 10
10
PMT
30 30 25 28 30 30 30 30 30 30 30 30
30
Evaluation Warning : The document was created with Spire.PDF for Python.
e-ISSN: 2
087-278X
TELKOM
NIKA
Vol. 11, No
. 12, Dece
mb
er 201
3: 722
2 – 7229
7228
solve the
mul
t
i-obje
c
tive a
probl
em i
s
to
transl
a
te the
vector
of obje
c
tives into
on
e obje
c
tive b
y
averagi
ng th
e obje
c
tives
with a wei
g
h
t
vector. The
most profo
u
nd dra
w
b
a
ck of traditional
algorith
m
s
i
s
their sen
s
itivity
to
ward
s
we
ights o
r
de
m
and level
s
. T
h
is di
scu
ssi
o
n
su
gge
sts th
at
the cl
assi
cal
method
s to
the p
r
obl
em
s
of MG
co
ntro
l optimization
model
a
r
e i
n
adeq
uate
an
d
inco
nvenient to
use.
Figure 4. The
Pareto Opti
mal Front in
Multi-obj
ectiv
e
Optimizatio
n
(C)
Table 3 Com
pare a
nd an
a
l
ysis of different
prefe
r
en
ces ap
proxim
a
t
e weighte
d
Pareto
optimal laye
r. The
pro
g
ra
ms
of lo
w
carbo
n
di
sp
atch
are
de
sig
ned
as L
C
DP1 (4
0%, 60
%),
LCDP2 (2
5%, 75%), LCDP
3 (0, 100%),
EWD (equ
al weig
hts di
spa
t
ch).
Table 3. The
Thre
e Prog
ra
ms Of Lo
w Carbo
n
Di
spat
ch
EWD
LCDP1
LCDP2
LCDP3
Mep (AFC
o )
5.7%
5.6%
5.8%
5.7%
Mep(Epo)
85%
81%
79%
73%
It shows that Low-ca
r
bo
n
powe
r
sche
duling st
rate
gy can also redu
ce the li
ne loss,
redu
cin
g
emi
ssi
on
s from four indi
cato
rs in Table 3.
Compa
r
ed
with equal
weight strate
gy,
carbon powe
r
sche
duling policy
redu
ce
s greater
extent to redu
ce
emission
s, b
u
t the co
st of
power g
ene
ration in
cre
a
ses
slightly. The re
sult
s ob
tained u
s
ing
our p
r
o
posed
techni
que t
o
minimize the
total co
st and
total emissio
n
s
were
com
pare
d
with
so
me co
nventio
nal strategie
s
of s
e
ttings
.
6. Conclusi
on
This p
ape
r
pre
s
ent
s a
gene
ral fram
ewo
r
k
fo
r th
e co
ntrol of
distrib
u
ted
energy
resou
r
ces o
r
g
anized in Microg
rid
s
. A agent is in com
m
unication wi
th other age
n
t
s by passing
a
messag
e. Me
ssage
re
ceiv
ed a
r
e h
andl
ed by the
m
e
ssage i
n
terp
reter of a
n
ag
ent, the age
n
t
s
have the
abil
i
ty to dynamically mo
del
communi
ty ba
sed
on
negot
iation in the
orga
nizationa
l
model of com
putation with
observe its e
n
vir
onm
ent a
nd exch
ang
e
messag
e am
ong the a
gent
s.
The formul
ation of the MG cont
rol m
odel and
fu
zzy prefe
r
en
ces evolution
a
r
y algorithm
is
prop
osed to
re
solve
the
problem.
T
he
simulatio
n
results
sh
ow th
at the
app
ro
ach
can
signifi
cantly improve p
e
rfo
r
man
c
e a
nd a
dapt we
ll to the ch
ang
es o
f
dynamic en
vironme
n
ts.
Ackn
o
w
l
e
dg
ement
This
wo
rk
wa
s supp
orted i
n
part by the
Ke
y Project
of the Nation
al Natu
re Sci
e
n
c
e
Found
ation o
f
China
(No.
6053
4020
). T
h
is
work
wa
s sup
p
o
r
ted b
y
a gra
n
t fro
m
the Nation
al
High T
e
chn
o
logy Re
se
a
r
ch
and
De
velopment
P
r
og
ram
of Chin
a
(8
63 Program) (No.
2012AA0
401
04). A Proje
c
t
Suppo
rted b
y
Scientific Re
se
arch F
und
of Jilin Provi
n
cial Ed
ucation
0
.
70
0.
75
0.
80
0.
85
0.
90
0.
95
0
0.
028
0.
030
0.
032
0.
034
0.
036
0.
038
0.
040
(E
po
)
Me
p
Me
p
(F
lo
s
s
)
))
(
(
))
(
(
po
lo
s
s
E
Me
p
f
F
Me
p
f
<<
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
Optim
i
zing M
u
lti-age
nt MicroG
rid Resou
r
ce S
c
he
dulin
g by Co-E
vol
u
tionary (Su
n
Hong
bin)
7229
(201
202
68).
A Proje
c
t Su
pporte
d by
Scientific
an
d Te
chn
o
logi
cal Pla
nnin
g
Proje
c
t of
Jilin
Province (20
1203
32). A P
r
oje
c
t Supp
orted by Scie
ntific and
Te
chn
o
logi
cal Plan
ning Proje
c
t
of
Jilin Provin
ce
(2010
056
5).
Referen
ces
[1]
Rahm
an S, Pipattanas
omp
o
r
n
M,
T
e
klu Y.
Intellig
ent Dis
tributed Aut
o
n
o
mous Pow
e
r
System
(IDAPS)
. IEEE
Po
w
e
r En
gin
e
e
r
ing Soc
i
et
y
G
ener
al Meeti
n
g
.
2007:1-8.
[2]
Khoday
ar ME, Barati
M, Sh
a
h
id
ehp
our
M. Integrati
o
n
of
Hig
h
R
e
li
abi
lit
y Distrib
utio
n S
y
stem
i
n
Microgri
d
Oper
ation.
IEEE Transactions on S
m
art Grid
. 20
1
2
; 3(4): 199
7-2
006.
[3]
Sortomme E, Venkata SS,
Mitra J. Microgrid
Protecti
on
Using C
o
mm
unic
a
tion-Ass
i
s
t
ed Dig
ital
Rela
ys.
IEEE Transactions on Power Deliv
ery
. 2010; 25(
4): 2789-
279
6.
[4]
McArthur SDJ, Davidson EM, Catterson VM
. Multi-a
g
ent s
y
stems f
o
r po
w
e
r
en
g
i
ne
erin
g
app
licati
ons Pa
rt I: concepts,
appr
oac
h
e
s, a
nd tech
nica
l ch
alle
ng
es.
IEEE Transactions
on Power
System
. 20
07; 22(4): 17
43-
17
52.
[5]
Mait
y
I, Ra
o
S. Simulatio
n
and Pric
in
g Mecha
n
ism A
nal
ysis
of a
Solar-P
o
w
er
ed
Electrical
Microgri
d
.
IEEE System
s Journal
. 20
10; 4 (3
): 275-28
4.
[6]
Mishra S,
Mal
l
e
sham
G, Jha
AN. Desi
gn
of
cont
rol
l
er
an
d
communic
a
tio
n
for freq
ue
nc
y r
egu
latio
n
of a smart microgri
d
.
Institution of Engin
eer
in
g and T
e
ch
no
l
ogy
. 201
2; 6(4)
: 248-25
8.
[7]
Srinivas
an D,
Chaz
elas J.
A
priority list-b
a
s
ed evo
l
uti
ona
ry algor
ith
m
to solve lar
ge s
c
ale u
n
it
commitment pr
obl
e
m
. IEEE In
ternational Conference
on Pow
e
rCon. 2004: 1746-1751.
[8]
Che
ng CP, Li
u
CW
, Liu CC. Unit commitme
n
t b
y
Lagr
an
gi
an rel
a
xati
on a
nd g
enetic
alg
o
rithms
.
IEEE Transactions on Power
System
. 20
00; 30(2): 70
7–
714
.
[9]
Hern
and
ez-Ara
mburo
CA , Gr
een
T
C
, Mugni
ot N.
F
uel
co
n
s
umptio
n mi
ni
mizatio
n
of
a
microgri
d
.
IEEE Transactions on Industr
y Applic
ations
.
200
5; 41(3): 67
3-68
1
[10]
Nag
a
ta T
,
Oh
ono M, Kub
o
k
a
w
a
J.
A mu
lti
-
age
nt appr
oa
ch to unit co
mmit
me
nt prob
le
ms
. IEEE
Po
w
e
r En
gin
e
e
r
ing Soc
i
et
y
W
i
nter Meetin
g. 2
002: 64
–6
9.
[11]
Jing
Y, Jia
n
zh
o
ng Z
,
Bi
n H,
R
ongta
o
L. Opti
mal sh
ort-term
gen
eratio
n sc
h
edu
lin
g
w
i
t
h
m
u
lti-a
gent
s
y
stem u
n
d
e
r
a der
eg
ulate
d
po
w
e
r mar
k
et.
IEEE International Jour
nal of
Com
p
utational
Cog
n
itio
n
. 200
5: 61–6
5.
[12]
Moham
ed F
,
Koivo H.
System Mo
del
li
ng
and Onl
i
ne O
p
timal Man
a
g
e
m
e
n
t of MicroGrid w
i
th
Battery Storage
. 6th Intern
ati
ona
l Co
nfere
n
c
e on
Re
ne
w
a
ble E
ner
gies
a
nd Po
w
e
r Qu
al
it
y
.
2
007:
26–
28
[13]
Sun HB, Di
ng
YS.
T
he Evolution
a
r
y
St
abi
li
t
y
Affected b
y
F
l
o
w
of Ener
g
y
i
n
the Bi
o-
Net
w
ork
Architecture.
N
euroc
o
m
puti
n
g
. 2008; 72:1
62-
169
[14]
Corne DW, Jerram NR, Kno
w
les JD.
PESA-II: Region based select
ion in evolutionar
y
mu
ltio
bjectiv
e
opti
m
i
z
a
t
i
on.
Procee
din
g
s
of the Ge
netic
a
nd E
v
oluti
onar
y
C
o
mputati
on
Confer
ence. 2
001: 28
3-2
90.
[15] Deb K, Pratap
A, Agra
w
a
l S.
A fast and elitist multi - objec
tive genetic algorithm: NSGA-II,
IEEE
T
r
ansactio
n
s o
n
Evoluti
o
n
a
ry Co
mp
utation
. 200
2; 6(2): 182
-197.
[16]
Eppe S, Lo
pez
-I
banez M, Stu
t
zle T
.
An experi
m
ent
al stud
y of pref
erenc
e mode
l inte
gr
ation i
n
t
o
mu
lti-ob
jectiv
e
optimi
z
a
t
i
on
heur
istics.
IEEE Conference on Evolutio
nar
y
Computation. 2011:
275
1 -27
58.
[17]
Lakshm
i
R, B
h
arathi
SG. PS
O base
d
Optim
a
l P
o
w
e
r
F
l
o
w
w
i
t
h
H
y
b
r
i
d
D
i
s
t
ributed
Gen
e
r
a
tors a
nd
UPFC PDF.
TELKOMNIKA Indo
nesi
an Jo
u
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2012; 1
0
(3): 4
09-4
18.
Evaluation Warning : The document was created with Spire.PDF for Python.