TELKOM
NIKA
, Vol. 11, No. 12, Decem
ber 20
13, pp.
7769
~77
7
2
e-ISSN: 2087
-278X
7769
Re
cei
v
ed
Jul
y
14, 201
3; Revi
sed Aug
u
st
25, 2013; Accepted Sept
em
ber 6, 201
3
Biologically Inspired Optimization of Building District
Heating Networks
Leiming Shang
1
, Xiaomin
Zhao*
2
1
School of Co
mputer Scie
nc
e and T
e
chno
l
o
g
y
, Hua
i
b
e
i N
o
rmal Un
iversit
y
, Hu
aib
e
i, 23
5
000, Ch
in
a
2
School of Ph
ysics and El
ectronic Informati
o
n
, Huai
bei N
o
r
m
al Univ
ersit
y
,
Huai
bei, 2
350
00, Chi
n
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: xmzha
o
1
2
0
1
@
12
6.com
A
b
st
r
a
ct
In this
pap
er w
e
sh
ow
that a
biol
og
ical
ly i
n
s
p
ire
d
mo
delc
a
n b
e
succ
essfu
lly a
p
p
lie
d to
p
r
obl
ems
of
bu
ild
in
g opti
m
a
l
districthe
ating
netw
o
rk.
T
he mod
e
l is ba
sed on
phys
i
o
l
ogic
a
l obs
ervat
i
ons
of
th
e
tru
e
sli
m
e
mold P
h
ysaru
mp
olyce
p
hal
u
m
, but ca
n
also
be
us
e
d
for path-fi
nd
ing
in the
co
mp
lica
t
ed netw
o
rks o
f
ma
z
e
s an
d ro
ad
maps. A s
t
rategy of opti
m
a
lly b
u
il
di
ng
heati
ng d
i
stri
butio
n netw
o
rk
w
a
s guide
d
by
themod
el an
d a
w
e
ll-tu
ned
a
n
t
col
ony alg
o
ri
thm an
d
g
e
n
e
ti
c alg
o
rith
m. T
h
e resu
lts in
dic
a
te that a
l
tho
u
g
h
there ar
e not
large-sc
al
e efficiency s
a
vi
ngs to
b
e
ma
de, the
bi
olo
g
ica
lly i
n
sp
ired a
m
oe
boi
d
mov
e
me
ntmod
e
l is ca
pa
ble
o
f
findin
g
resu
lts of equ
al or
b
e
tter opti
m
a
lity
than a c
o
mpa
r
abl
e ant co
lo
n
y
alg
o
rith
m an
d gen
etic al
gorith
m
.
Ke
y
w
ords
:
a
m
o
e
b
o
id
mov
e
ment mod
e
l,
comp
lex net
w
o
rk, ant colony al
gorith
m
,
heatin
g netw
o
rk
opti
m
i
z
at
ion
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Distri
ct heati
ng
n
e
two
r
ks (DHN) se
rve to
tran
sp
ort h
o
t wate
r
or st
eam from tre
a
tment
works to in
dividual custo
m
ers an
d u
s
ually rep
r
e
s
e
n
t a sig
n
ifica
n
t capital i
n
vestment i
n
th
e
developm
ent of the urban
environ
ment [1].
The p
r
obl
em
of desi
gnin
g
a DHN to opt
imally meet p
e
rform
a
n
c
e
criteria
on
one
hand,
su
ch a
s
d
e
li
vering
suffici
ent wate
r p
r
essure for
hi
gh ri
se
buildi
ngs
and fi
re
fighting; whi
l
st
minimizi
ng co
st crite
r
ia, su
ch a
s
the co
st of
material,
excavation, freque
ncy of maintena
nce is
kno
w
n to
be
NP ha
rd [2].
On the
othe
r ha
nd,
the e
n
vironm
ent is becoming
a
n
increa
sin
g
l
y
importa
nt in the philo
so
ph
y of the modern
so
ciety, and inte
rnati
onally peo
ple
have be
com
e
more
a
w
are
of the e
n
viro
nment, an
d t
he
con
s
e
que
nce
s
of glo
b
a
l
wa
rmin
g [3]. To
con
s
id
er
the
gro
w
ing
dem
and
s for
a re
ductio
n
of th
e CO
2 emi
ssions, a
nd the
govern
m
ent’
s
requi
reme
n
t
s
on the distri
ct
heating to operate
with a
green
er p
r
ofi
l
e, propo
sal f
o
r optimi
z
atio
n of the supp
ly
of distri
ct hea
ting netwo
rk
wa
s pro
p
o
s
e
d
in this pap
e
r
.
In the past de
cad
e
s, a larg
e variety of
computation
a
l algorith
m
s ha
ve been devi
s
ed for
this task whi
c
h inclu
de well
kno
w
n techn
i
que
s in ope
rational re
se
arch such as li
near, dyna
mi
c
and intege
r
prog
ram
m
ing
.
In recent years ho
we
ve
r, a variety of nature-i
n
spired an
d meta-
heuri
s
tic
algo
rithms
su
ch a
s
gen
etic al
g
o
rithm
s
, simu
lated ann
eali
ng and ta
bu
sea
r
ch [4] ha
ve
been
widely
investigate
d
as u
s
eful re
sea
r
ch
tools for DHN de
sign.Amo
n
g
s
t meta-he
u
ri
sti
c
algorith
m
s
aforem
ention
e
d
,
ant colo
ny a
l
gorithm
(ACOs) [5] ha
s b
een g
ene
rally
accepte
d
to
be
one of the m
o
st succe
s
sfu
l
solution
sfo
r
DHN
optimi
z
ation[6]. Whil
st ACO
s
hav
e provid
ed go
od
solutio
n
s to h
eating dist
rib
u
tion optimization pro
b
lems
for s
o
me time, the s
t
eady inc
r
eas
e
in
the
compl
e
xity of the network i
n
formatio
n b
e
ing k
ept by the heating
co
mpanie
s
mea
n
s that ACO
s
are
no lo
nge
r al
ways
suit
able. Thi
s
i
s
in pa
rt due t
o
the lon
g
ru
nning time
s i
n
cu
rred by th
e
algorith
m
du
e
and in
pa
rticular, the
high
numb
e
r
of
o
b
jective fun
c
ti
on evalu
a
tion
s requi
red
by
evolutiona
ry tech
niqu
es. An increa
sing
numbe
r
of el
ements in th
e netwo
rk
an
d more d
e
tail
ed
24-h
o
u
r
sim
u
lation stu
d
ie
s have seen t
he complexit
y
of a singl
e
netwo
rk sim
u
lation in
cre
a
se
massively. Therefo
r
e, sci
en
tists are
con
s
tantly
looking
for techni
que
s whi
c
h might
deliver ACO-
cla
ss results,
but with fewe
r obje
c
tive function
cal
c
ula
t
ions.
The o
p
timizat
i
on of
DHN h
a
s two
crite
r
i
a
: 1) the
ne
w method
sh
o
u
ld be
gu
ara
n
teed to
find the sho
r
t
e
st ro
ute. Th
at
mean
s sto
c
ha
stic b
a
se
d pro
c
e
s
se
s are o
u
t of co
nsid
eratio
n; and
2) the
ne
w
al
gorithm
s
sh
o
u
ld b
e
capa
bl
e of flexib
le
a
daptability an
d re
-routing. I
n
re
ce
nt yea
r
s,
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: 776
9 – 7772
7770
sci
entist find
a completel
y
new method that fu
lfills these crite
r
ia is dem
on
strated u
s
in
g
a
mathemati
c
al
path findin
g
model d
e
rive
d from
o
b
servation of an
amoeb
oid o
r
gani
sm, the t
r
ue
s
lime mold [7].
In this p
ape
r
we inve
stigat
e the a
ppli
c
at
ion of a
n
am
oeboi
d move
ment mod
e
l t
o
solve
the proble
m
of DHN optim
ization.
We
d
e
scrib
e
th
e
a
pplication of t
he tru
e
slime
mold
syste
m
s
to a district h
eating network and comp
a
r
e it with
a well-tune
d ant colo
ny algorit
hm, with mixed
results.
T
he remaind
e
r of
this se
ction di
scus
se
s heat
ing
di
strib
u
tio
n
net
work op
timization an
d
previou
s
rese
arch into usi
n
g biologi
cally inspi
r
ed a
pproache
s to this purpo
se.
2. Method
DHNs a
r
e p
a
r
t of the e
nergy sup
p
ly sy
st
em, comp
ri
sing
of num
b
e
r of inte
rcon
necte
d
element
s su
ch as pipe
s, n
ode
s, pump
s
, valves,
and reservoi
rs of
varying sh
ap
es an
d si
zes.
The no
de
s re
pre
s
ent
com
b
ined
points
of headi
ng d
e
mand
(e.g.
hou
sing o
r
in
dustri
a
l e
s
tate
s)
on the
syste
m
. The pu
rp
ose
of the n
e
twork i
s
to
deliver h
o
t water o
r
st
ea
m to the de
man
d
node
s from t
he heatin
g treatment wo
rks or othe
r so
urce thro
ugh
out the day and und
er varying
deman
d co
nd
itions.
There are m
any options t
o
be con
s
id
e
r
ed
when opt
imizing a DHN, but in most ca
se,
an exi
s
ting
n
e
twork is alre
ady in
pla
c
e
makin
g
it
difficult to
attemp
t major
stru
ctural
chang
e i
n
the existin
g
desi
gn.
Chan
ging th
e po
si
tion of
the n
e
twork elem
ents
i
s
con
s
i
dere
d
a
maj
o
r
stru
ctural
cha
nge
and
wou
l
d be
very
co
stly. As
thi
s
i
s
a
la
rge
cap
i
tal investm
e
nt and
mayb
e
there
is som
e
en
ergy waste after
use, t
he h
eat
ing
compani
es ine
v
itably want
modificatio
n
s to
last for lo
ng ti
me pe
riod
s, typically 50
-1
00 year
s. Th
erefo
r
e, opti
m
ization
sh
o
u
ld be
con
s
id
ered
and imple
m
e
n
ted before constructio
n
.
The pla
s
mo
d
i
um of true
slime mold Ph
ysaru
m
poly
c
ephal
um can
tackl
e a ma
ze a
nd
some
othe
r t
y
pes of
geo
metrical pu
zzle, and
ca
n
succe
ssfully o
p
timize
su
rvival tasks [8]. The
chall
enge
is t
o
extract
a m
a
thematical a
l
gorithm
fo
r t
h
is n
a
tural
co
mputation. T
he bo
dy of the
plasm
odiu
m
contai
ns a
n
e
twork of tu
bes, whic
h e
nable
s
nut
rie
n
ts and
ch
e
m
ical
sign
als to
circulate th
ro
ugh the o
r
ga
nism. When f
ood sou
r
ce
s
are p
r
e
s
ente
d
to a starve
d plasmodi
u
m
that has spre
ad over the e
n
tire su
rfa
c
e
of an agar
pl
a
t
e, parts of the orga
nism concentrate ov
er
the food
source
s a
nd
are
con
n
e
c
ted
by only a
fe
w tube
s. The
pa
th co
nne
cting
these p
a
rts
of
the pla
s
m
odi
um a
r
e
the
shorte
st p
o
ssi
b
le, even
in
a
ma
ze [9].Th
e
phy
siolo
g
ical me
ch
anism of
tube form
atio
n ha
s be
en e
s
tabli
s
he
d: tubes thi
c
k
en i
n
a given
dire
ction
whe
n
shuttle streami
n
g
of the protopl
asm pe
rsist
s
in that directio
n for
a certai
n
time [10]. This implie
s po
sitive feedback
betwe
en flux and tube thi
c
kne
s
s, as the
cond
ucta
nce
of the sol is greate
r
in a thicker
cha
n
n
e
l.
We
n
o
w dem
onstrate
the appli
c
ation of
our
model
to DHN
de
sig
n
. Grey b
a
ckgrou
n
d
and
white line
s
in Fig
u
re
1
sho
w
the
stre
et stru
ctur
e (netwo
rk) of west di
st
rict of
a Chin
ese cit
y
.
Blue, gree
n a
nd yellow
re
ctangle
s
represent apa
rtmen
t
s (si
n
k n
ode
s)
with differe
nt deman
ds o
f
energy. We
use thi
s
m
ap a
s
ma
ze
in the
mo
del of
pla
s
modium
of t
r
ue
slim
e m
o
ld
Physarumpol
yceph
alum. T
he optimi
z
ed
sho
r
test a
nd
the most efficient heatin
g n
e
twork
will b
e
fund by the organi
sm.
In this case, we u
s
e
pattern
re
cog
n
ition techniq
u
e
s
to obtai
n street and
ap
artmen
t
blocks. To do
so, it is ne
ce
ssary to po
ssess
the no
de
data, in whi
c
h each nod
e
corre
s
p
ond
s to
a jun
c
tion i
n
the st
reet
net
work, a
nd the
node
c
onn
ection data th
at
corre
s
p
ond to
the di
stan
ce
s
(or
co
st) between conn
ecte
d node
s.
On
ce the data are set by sele
cting the sou
r
ce node a
nd al
l
the sin
k
nod
e
s
, it is easy to obtain the op
timized net
wo
rk u
s
ing Phy
s
arum
solver.
We d
e
velop
e
d
a math
em
atical mo
del
for ad
aptive
netwo
rk construction
to e
m
ulate
Physarum b
e
havior, ba
se
d
on feed
ba
ck
loop
s bet
we
e
n
the thickn
e
ss
of ea
ch tu
be an
d inte
rn
al
protopl
asmic
flow in whi
c
h
high rates o
f
str
eaming
stimulate an increa
se in tube diamete
r
,
whe
r
ea
s tub
e
s tend to
decli
ne at lo
w flow
rate
s. The initial sha
pe of a
plasm
odiu
m
is
rep
r
e
s
ente
d
by the map with object
s
be
ing extra
c
ted.
The edg
es
repre
s
e
n
t plasmodial tube
s
in
whi
c
h protopl
asm flows, and node
s
are
junction
s bet
wee
n
tubes.
Suppo
se that
the pressu
re
s
at node
s i an
d j are pi an
d
pj, resp
ectiv
e
ly, and
that the two no
de
s are conn
ecte
d by a cylind
e
r
of length
Lij
and
ra
diu
s
rij. Assuming
t
hat fl
ow i
s
la
minar an
d
follows
the
Ha
gen-P
o
iseuill
e
equatio
n, the flux through the tube is:
Q
π
δ
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
e-ISSN:
2087
-278X
The De
sig
n
o
f
Wild Anim
als Monito
ring
System
base
d
on 3G an
d Internet…
(Ch
e
nyi
ngZe
n
g
)
7771
W
h
er
e
δ
i
s
th
e visco
s
ity of
the flui
d, an
d Dij
=
Π
r4/
8
δ
is a me
as
ur
e
o
f
th
e co
nd
u
c
tivity
of the tu
be.
As the
len
g
th
Lij i
s
a
con
s
tant, t
h
e
be
ha
vio
r
o
f
th
e
n
e
t
w
o
r
k
is
de
sc
r
i
be
d b
y
th
e
con
d
u
c
tivities, Dij, ofthe edges.
3. Results
We
ob
se
rve
d
Physaru
m
co
nne
cting
a tem
p
late
of 4
2
n
o
d
e
s th
at rep
r
ese
n
ted
geog
rap
h
ical
locatio
n
s
of apartm
ent in
the we
st
pa
rt
of the city. The Physaru
m pla
s
modi
u
m
wa
s allo
wed t
o
gro
w
from
sou
r
ce nod
e
and initially filled much of the availabl
eland spa
c
e, b
u
t
then con
c
e
n
trated on ap
artment
s
thin
ning out
th
e
network to
leave
a
su
bset
of larg
er,
interconn
ecti
ng tube
s. T
he re
sult i
s
sho
w
n i
n
Figure 1. Red line
s
sho
w
that po
ssi
b
le
con
n
e
c
tion
s betwe
en no
d
e
s.
In Figu
re 2
Amoeboi
d a
n
d
GA are b
o
th
run fo
r 3
0
0
0
0
0 fitness
evaluation
s
. Am
oeboi
d
contin
ue
s to evolve the quality of solutions u
n
til over 200 0
00 fitness evalu
a
tions a
n
d sh
o
w
s
eviden
ce of f
u
rthe
r imp
r
ov
ement. On
th
e final
ite
r
atio
n pe
rform
e
d
by Amoeboi
d
the differen
c
e
betwe
en the
fitness of th
e iteration
s
best
soluti
on
f(Sib) an
d the mea
n
av
erag
e fitness of
s
o
lutions f(
Ф
) was 92
1,0
83,790. T
he
differen
c
e
be
tween
the f(
Ф
) a
nd f(Si
b
)
indi
cate
s th
at
further impro
v
ement on
the qu
ality o
f
solu
tion
s
could b
e
a
c
h
i
eved given
more
fitness
evaluation
s
a
s
clo
s
e f(
Ф
)
and f(Sib
)
values in
dicate that stagn
a
tio
n
is o
c
curring
.
Wherea
s the
greate
r
the
differen
c
e b
e
t
ween f(
Ф
) a
nd f(Sib) th
e
more
explo
r
ation i
s
bei
ng condu
c
te
d.
Amoeboi
d ap
pears to
be t
he alg
o
rithm
of choi
ce
he
re
as it ha
s a
c
hieved a
mu
ch fitter sol
u
tio
n
than ACO. G
A
improve
s
t
he q
uality of
solutio
n
s slo
w
er than
the
Amoeboi
d an
d provide
s
a
l
e
ss
fit final sol
u
tion a
nd th
erefore
is po
ore
r
in b
o
th
re
sp
ects. Am
oeb
o
id h
a
s a
c
hie
v
ed a fitter fina
l
solutio
n
than
the GA whi
c
h
requi
re
s more fit
ness eval
uation
s
than
both the Amo
e
boid a
nd th
e
ACO.
Figure 1. Street Structure a
nd Optimized
Heatin
g Net
w
ork
Figure 2. Co
mpari
s
o
n
of Co
st among
Thre
e
Algorithm
s: genetic al
gorit
hm (GA), Ant colo
ny
optimizatio
n (ACO) a
nd the
Amoeboid
so
lver
4. Conclusio
n
Overall, we
concl
ude that the Physaru
m
net
wo
rks
sho
w
e
d
cha
r
acteri
stics si
milar to
those
of th
e
real
heatin
g
netwo
rks in t
e
rm
s of
cost,
tran
sp
ort
efficien
cy, an
d f
ault tole
ran
c
e.
Ho
wever, th
e
Physarum n
e
tworks
self-orga
n
ized
without centralized control o
r
explicit glo
b
a
l
informatio
n by a proce
ss of selective
reinforc
em
e
n
t of preferred route
s
an
d simultan
eo
us
removal of re
dund
ant co
nn
ection
s.
Ackn
o
w
l
e
dg
ement
This work was supp
orted
in part by a gr
ant from
College a
n
d
University Nature
Scien
c
e
Fou
ndation
-
Fu
nd
ed Proje
c
t o
f
Anhui P
r
ov
ince,
Chi
na
(No.
KJ2
012
B170 a
n
d
No.
KJ201
3B24
8).
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: 776
9 – 7772
7772
Referen
ces
[1]
Dale
nb
äck JO. Lar
ge-Sca
l
e
S
o
lar
He
ating:
S
t
ate of the
Art.
Prese
n
tatio
n
at Euro
pe
an S
u
stain
abl
e
Energy W
eek
.
Brussels, Bel
g
i
u
m. 2012: 1
8
–
22.
[2]
Micha
e
l R G
a
r
e
y
,
Davi
d S J
ohns
on. C
o
mp
uters an
d Intra
c
tabilit
y: A Gu
i
de to th
e T
heor
y of NP-
Compl
e
ten
e
ss. W
.
H.
F
r
eeman
. 1979. ISBN 0
-
716
7-10
45-
5.
[3]
Patrick Lesl
i
e,
Joshu
a
Pearc
e
, Rob Harra
p, S
y
lv
ie D
ani
el.
T
he applic
ation
of smartphon
e
technol
og
y
to eco
nom
ic a
nd
envir
onm
en
tal a
nal
ys
is
of bu
ild
in
g e
ner
g
y
co
nservati
o
n
strateg
i
es.
I
n
ternati
o
n
a
l
Journ
a
l of Sust
ain
abl
e Ener
gy
. 2012; 31(
5): 295-3
11.
[4]
F
r
ed Glover. T
abu Se
arch - P
a
rt 2.
ORSA Journa
l on Co
mp
uting
. 19
90; 2(
1): 4–32.
[5]
M Dori
go
et L
M
Gambard
e
l
l
a
. Ant C
o
lo
n
y
S
y
stem
: A
Coo
perativ
e L
earn
i
ng
Ap
pro
a
ch to t
h
e
T
r
aveling Sal
e
sman Prob
lem.
IEEE Transactions o
n
Evol
uti
onary C
o
mput
ation
. 19
97; 1(
1): 53-66.
[6]
Z
a
tirostami Ah
mad. Ce
ntral
Heati
ng S
y
ste
m
Optimizatio
n
.
Australia
n J
o
urna
l of B
a
sic
and
App
lie
d
Scienc
es
. 201
1; 5(10): 15
44-
154
8.
[7]
Atsushi T
e
ro, et al. Ru
les fo
r Biol
ogic
a
ll
y I
n
spir
ed Ad
apti
v
e Net
w
o
r
k De
sign.
Sci
ence
. 201
0:
32
7-
439.
[8]
Atsushi T
e
ro,
R
y
o Kob
a
y
as
hi
,
T
o
shi
y
uki Na
kagak
i. Ph
y
s
ar
um solver: A bi
olo
g
ica
l
l
y
ins
p
ir
ed metho
d
of road-n
e
t
w
or
k navig
atio
n.
Physica A.
200
6
;
(363): 115
–11
9.
[9]
Even, Shimo
n
, Graph Alg
o
rith
ms (2nd ed.),
Ca
mbri
dg
e Uni
v
ersity Press.
201
1: 46–
48.
[10]
A Baucer, B B
u
lln
he
imer, RF
Ha
rtl, C Strau
ss. Minimizi
ng
total tardi
ness
on a si
ng
le ma
chin
e usi
n
g
a
n
t
co
lo
ny
op
timi
za
ti
on
.
Centr
a
l Euro
pe
an J
ourn
a
l for Ope
r
ations R
e
se
ar
ch an
d Econ
o
m
ics
. 20
00
;
8(2): 125-
14
1.
Evaluation Warning : The document was created with Spire.PDF for Python.