I
n
d
o
n
e
s
i
a
n
J
o
u
r
n
a
l
o
f
E
l
e
c
tr
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
a
n
d
C
o
m
p
u
te
r
S
c
i
e
n
c
e
V
ol
.
8,
N
o.
1,
O
c
t
ob
er
20
17
,
pp
.
137
~
145
D
O
I
:
10.
115
91/
i
j
eec
s
.
v
8
.i
1
.
pp
13
7
-
14
5
1
37
R
ec
ei
v
ed
J
une
6
,
2
01
7
;
R
e
v
i
s
ed
A
ugus
t
1
9
,
201
7
;
A
c
c
ept
e
d
S
ept
e
mb
er
2
,
201
7
De
v
elo
p
m
e
n
t
o
f
H
y
b
r
id
A
r
t
if
i
cia
l Neu
r
al N
et
w
o
r
k f
o
r
Q
ua
nti
f
y
i
n
g E
ne
r
g
y
S
a
v
i
n
g us
i
n
g M
e
a
s
ur
e
me
nt
a
nd
V
er
if
icat
i
o
n
W
a
n
N
a
z
i
r
a
h
W
a
n
M
d
A
d
n
a
n
*
1
,
N
o
fr
i
Y
e
n
i
ta
D
a
h
l
a
n
2
,
Is
m
a
i
l M
u
s
ir
in
3
1
,2
,3
F
a
c
ul
t
y
of
E
l
e
c
t
r
i
c
a
l
E
ng
i
ne
er
i
ng,
U
ni
v
er
s
i
t
i
T
ek
nol
o
gi
M
A
R
A
,
S
hah A
l
a
m
,
S
el
a
ngor
,
M
al
ay
s
i
a
1
F
ac
u
l
t
y
of
E
n
gi
n
eer
i
ng a
nd L
i
f
e S
c
i
enc
es
,
U
ni
v
er
s
i
t
i
S
e
l
an
go
r
,
B
es
t
ar
i
J
ay
a,
S
e
l
ang
or
,
M
al
ay
s
i
a
*C
or
r
es
po
ndi
ng a
ut
hor
,
e
-
m
a
i
l:
w
an_naz
i
r
ah@
un
i
s
e
l
.
ed
u.
m
y
1
,
nof
r
i
y
en
i
t
a0
12@
pp
i
nan
g.
ui
t
m
.
edu.
m
y
2
,
i
s
m
ai
l
bm
@
s
al
a
m
.
u
i
t
m
.
edu
.
m
y
3
A
b
st
r
act
T
hi
s
pa
per
pr
e
s
ent
s
a H
y
br
i
d
A
r
t
i
f
i
c
i
al
N
e
ur
al
N
et
w
or
k
(
H
A
N
N
)
f
or
c
hi
l
l
er
s
y
s
t
em
M
eas
ur
em
ent
and V
er
i
f
i
c
at
i
o
n (
M
&
V
)
m
odel
dev
e
l
opm
en
t
.
I
n t
h
i
s
w
or
k
,
hy
br
i
di
z
at
i
on of
E
v
ol
ut
i
o
nar
y
P
r
o
gr
am
m
i
ng (
E
P
)
and
A
r
t
i
f
i
c
i
a
l
N
e
ur
al
N
et
w
or
k
(
A
N
N
)
ar
e
c
on
s
i
d
er
ed
i
n
m
ode
l
i
ng t
he
ba
s
el
i
ne
el
ec
t
r
i
c
a
l
ener
gy
c
on
s
um
pt
i
on
f
or
a
c
hi
l
l
er
s
y
s
t
em
hen
c
e
qu
ant
i
f
y
i
n
g s
av
i
ng.
E
P
w
i
t
h
c
o
ef
f
i
c
i
ent
of
c
or
r
e
l
at
i
on (
R
)
obj
ec
t
i
v
e
f
un
c
t
i
on i
s
us
ed
i
n op
t
i
m
i
z
i
ng t
he ne
ur
al
net
w
or
k
t
r
ai
n
i
ng
pr
oc
es
s
an
d s
el
ec
t
i
ng t
he op
t
i
m
al
v
al
u
e
s
of
A
N
N
i
ni
t
i
a
l
w
ei
ght
s
a
nd
b
i
as
es
.
T
hr
e
e
i
n
put
s
t
hat
ar
e
af
f
e
c
t
i
ng
e
ner
g
y
u
s
e
o
f
t
h
e
c
hi
l
l
er
s
y
s
t
em
a
r
e
s
e
l
ec
t
ed;
1)
oper
at
i
ng
t
i
m
e,
2)
r
ef
r
i
ger
a
nt
t
onnag
e
an
d
3)
di
f
f
er
ent
i
al
t
e
m
per
at
ur
e.
T
he
out
put
i
s
h
our
l
y
e
ner
g
y
u
s
e
o
f
bui
l
di
ng
a
i
r
-
c
o
ndi
t
i
on
i
ng
s
y
s
t
e
m
.
T
he H
A
N
N
m
o
d
el
i
s
s
i
m
ul
a
t
ed w
i
t
h
16
di
f
f
e
r
ent
s
t
r
uc
t
ur
es
and
t
he
r
e
s
ul
t
s
r
ev
ea
l
t
h
at
a
l
l
H
A
N
N
s
t
r
uc
t
ur
es
pr
odu
c
e
hi
gh
er
pr
edi
c
t
i
o
n
per
f
or
m
an
c
e w
i
t
h R
i
s
a
bo
v
e
0.
977.
T
he
be
s
t
s
t
r
uc
t
ur
e
w
i
t
h
t
he
h
i
gh
es
t
v
a
l
ue
of
R
i
s
s
e
l
ec
t
ed
as
t
he
b
a
s
el
i
ne
m
od
el
h
enc
e
i
s
u
s
ed
t
o
det
e
r
m
i
ne
t
he
s
av
i
ng.
T
h
e a
v
oi
ded
en
er
gy
c
al
c
ul
at
ed
f
r
om
t
hi
s
m
ode
l
i
s
132
944.
59
k
W
h
t
h
at
c
ont
r
i
b
ut
es
t
o
1.
3
8%
of
s
av
i
ng per
c
ent
a
ge.
Ke
y
w
o
rd
s
:
N
eur
al
N
et
w
or
k
,
E
ner
gy
S
av
i
ng,
E
v
o
l
ut
i
onar
y
P
r
ogr
am
m
i
ng,
M
eas
ur
em
en
t
an
d
V
er
i
f
i
c
at
i
on.
C
o
p
y
r
i
g
h
t
©
2
01
7
In
s
t
i
t
u
t
e
o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
r
i
n
g
a
n
d
S
c
i
e
n
c
e
.
A
l
l
r
i
g
h
t
s r
es
er
ved
.
1
.
I
n
tr
o
d
u
c
ti
o
n
Mal
a
y
s
i
a i
s
a r
ap
i
d
l
y
d
ev
el
opi
ng c
oun
t
r
y
i
n A
s
i
a.
T
he G
r
os
s
D
o
m
es
t
i
c
P
r
oduc
t
(
G
D
P
)
i
n
Mal
a
y
s
i
a gr
e
w
b
y
20%
f
r
om
t
he
y
ear
201
0 t
o 201
3 a
nd ex
pec
t
e
d t
o i
nc
r
eas
e o
v
er
t
he
y
e
ar
s
[
1]
.
D
ue t
o t
hat
,
t
he n
um
ber
of
c
o
m
m
er
c
i
al
and
r
es
i
dent
i
a
l
ar
ea w
i
l
l
al
s
o i
nc
r
eas
e i
n
p
ar
al
l
el
w
i
t
h t
he
G
D
P
gr
o
w
t
h.
T
he
i
nc
r
eas
i
ng
n
um
ber
of
c
o
m
m
er
c
i
a
l
and
r
es
i
de
nt
i
al
ar
ea
i
n
Mal
a
y
s
i
a
has
i
nc
r
eas
ed
t
he
en
er
g
y
dem
and a
nd s
up
pl
y
.
I
n 20
14,
i
t
i
s
r
epor
t
e
d t
h
at
t
he n
um
ber
of
el
ec
t
r
i
c
i
t
y
c
o
ns
um
pt
i
on i
n
P
eni
ns
ul
ar
M
al
a
y
s
i
a
i
nc
r
eas
ed
b
y
7.
5%
f
r
om
20
12 t
o
20
14 a
nd
56%
w
as
f
r
om
c
o
m
m
er
c
i
al
and r
es
i
de
nt
i
a
l
s
ec
t
or
s
[2
]
.
T
he el
ec
t
r
i
c
i
t
y
c
os
t
w
as
t
he l
ar
ges
t
c
ont
r
i
bu
t
or
t
o t
he t
ot
al
oper
a
t
i
n
g c
os
t
of
a bu
i
l
d
i
n
g
[
3]
.
T
he needs
t
o r
e
duc
e e
ner
g
y
c
ons
um
pt
i
o
n
w
h
i
l
e m
ai
nt
ai
ni
ng pr
o
duc
t
i
v
i
t
y
ar
e i
m
por
t
ant
due t
o
t
he r
i
s
i
ng
of
el
ec
t
r
i
c
i
t
y
c
os
t
no
w
ada
y
s
.
T
hi
s
s
i
t
u
at
i
on
has
pr
om
pt
ed M
al
a
y
s
i
an
gov
er
nm
ent
t
o
t
ak
e s
ev
er
a
l
ac
t
i
ons
i
nc
l
ud
i
ng
en
er
g
y
ef
f
i
c
i
enc
y
t
o m
anag
e t
hes
e
pr
ob
l
em
s
.
B
y
i
m
pl
em
ent
i
ng
t
he
en
er
g
y
ef
f
i
c
i
enc
y
,
t
he
el
ec
t
r
i
c
a
l
po
w
er
dem
and
c
an
be
r
ed
uc
ed
a
nd
t
h
us
as
s
i
s
t
i
n
c
r
eat
i
n
g
an o
v
er
a
l
l
c
os
t
r
educ
t
i
on
.
E
ner
g
y
ef
f
i
c
i
enc
y
(
E
E
)
i
m
pl
e
m
ent
at
i
o
n r
em
ai
ns
f
ar
c
heaper
t
h
an i
n
v
es
t
i
n
g i
n an
add
i
t
i
ona
l
ge
ner
at
i
on
[
4]
.
R
egar
ds
t
o t
h
i
s
f
ac
t
or
,
t
he
ener
g
y
ef
f
i
c
i
enc
y
of
ener
g
y
c
ons
er
v
at
i
on
m
eas
ur
es
(
E
C
M)
pr
oj
ec
t
s
hav
e bee
n i
n
t
r
oduc
e
d b
y
t
he Ma
l
a
y
s
i
a
n go
v
er
nm
ent
.
E
C
M pr
oj
ec
t
s
hav
e
b
ee
n
i
m
pl
em
ent
e
d
w
i
t
h
t
h
e
ai
m
t
o
r
ed
uc
e
e
ner
g
y
c
ons
um
pt
i
o
n
i
n
t
he
bu
i
l
di
ng.
I
n
or
der
t
o
ev
a
l
u
at
e t
h
e i
m
pac
t
of
E
C
Ms
i
n E
E
,
t
he r
e
duc
t
i
on i
n
ener
g
y
c
o
ns
um
pt
i
on an
d ener
g
y
s
av
i
ng
m
u
s
t
be det
er
m
i
ned.
T
he ev
a
l
u
at
i
ons
ar
e
v
er
y
de
pe
nden
t
on
Meas
ur
em
ent
an
d V
er
i
f
i
c
at
i
o
n
(
M
&V)
a
c
t
i
v
it
ie
s
.
M&
V
m
et
hod i
s
a
t
oo
l
t
o
det
er
m
i
ne,
q
uant
i
f
y
a
nd
v
e
r
i
f
y
t
he
s
av
i
ngs
on e
ner
g
y
us
e.
M&
V
i
s
t
he
pr
oc
es
s
of
us
i
n
g m
eas
ur
e
m
ent
s
t
o r
el
i
ab
l
y
det
er
m
i
ne ac
t
u
al
s
av
i
ng
c
r
eat
e
d
w
i
t
hi
n a
n
i
nd
i
v
i
d
ual
ac
t
i
v
i
t
y
b
y
an
e
ner
g
y
m
ana
gem
ent
pr
o
gr
a
m
[
5]
.
T
her
e ar
e s
e
v
er
a
l
pr
ot
oc
ol
s
an
d
gui
del
i
n
es
f
or
M&
V
but
t
he
m
os
t
c
o
m
m
on and w
i
de
l
y
us
ed i
s
I
P
MV
P
[
2]
.
I
n or
d
er
t
o pr
oper
l
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
25
02
-
4
752
I
J
E
EC
S
V
o
l.
8
,
N
o.
1
,
O
c
t
o
ber
20
17
:
1
37
–
1
45
138
r
epor
t
s
av
i
ng
i
n M
&
V
,
t
he
b
as
el
i
ne e
ner
g
y
us
e pa
t
t
er
n
bef
or
e E
C
M
i
m
pl
em
ent
at
i
o
n m
us
t
be
f
i
r
s
t
s
t
udi
e
d
a
nd
de
v
e
l
o
ped
t
o
det
er
m
i
ne
t
h
e r
e
l
at
i
o
ns
hi
p
bet
w
e
en
en
er
g
y
us
e
an
d
i
nput
v
ar
i
a
bl
es
.
T
hen,
af
t
er
E
C
M
i
m
pl
em
e
nt
at
i
on,
t
h
i
s
b
as
el
i
ne
e
ner
g
y
m
odel
i
s
us
ed
t
o
es
t
i
m
at
e
ho
w
m
uc
h
ener
g
y
w
o
ul
d
h
av
e
us
ed
i
f
t
her
e
h
ad
bee
n
n
o
E
C
M
i
m
pl
em
ent
at
i
on.
T
hi
s
es
t
i
m
at
i
on
i
s
r
ef
er
r
i
ng
t
o
adj
us
t
ed
b
as
el
i
ne
ener
g
y
i
n
pos
t
-
r
et
r
of
i
t
ph
as
e.
E
ner
g
y
s
av
i
ng
c
an
be
det
er
m
i
ned
b
y
c
o
m
par
i
ng
t
h
e
adj
us
t
ed
b
as
el
i
ne
e
ner
g
y
w
i
t
h
t
he
p
os
t
-
r
et
r
of
i
t
m
eas
ur
ed
ener
g
y
.
R
ec
e
nt
l
y
,
r
egr
es
s
i
on
ana
l
y
s
i
s
i
s
t
he
m
o
s
t
c
o
m
m
on
m
et
hod i
n f
or
m
ul
at
i
ng t
he
bas
e
l
i
n
e e
ner
g
y
[6
],
[7
]
.
Ho
w
e
v
er
,
t
h
i
s
M&
V
r
egr
es
s
i
on m
odel
i
s
l
es
s
ac
c
ur
at
e es
pec
i
a
l
l
y
f
or
non
-
l
i
n
ear
c
har
ac
t
er
i
s
t
i
c
henc
e c
o
nt
r
i
but
es
a
l
ar
g
e s
t
and
ar
d er
r
or
[6
],
[7
]
.
T
he
M&
V
pr
oc
es
s
i
n
v
o
l
v
es
m
odel
i
ng,
m
et
er
i
ng
and
s
am
pl
i
ng
pr
oc
es
s
an
d
t
hes
e
ac
t
i
v
i
t
i
es
c
r
eat
e
unc
er
t
ai
n
t
y
i
n
r
ep
or
t
i
n
g
e
ner
g
y
s
a
v
i
ngs
.
It i
s
i
m
por
t
ant
t
o pr
ec
i
s
el
y
c
ons
i
d
er
i
n
g t
h
e ac
c
ur
ac
y
he
nc
e t
o
de
v
el
op
an
ac
c
ur
at
e
M&
V
b
as
el
i
ne
ener
g
y
m
odel
t
o o
v
er
c
om
e t
hes
e
i
s
s
ues
.
R
ec
ent
l
y
,
A
r
t
i
f
i
c
i
a
l
N
eur
a
l
N
et
w
or
k
s
(
A
N
N
)
has
been on
e of
t
he
m
os
t
p
opu
l
ar
f
or
ec
as
t
i
ng
t
ec
hni
ques
an
d
us
ed
t
o
s
o
l
v
e
v
ar
i
ous
eng
i
neer
i
ng
and
t
ec
h
no
l
og
y
pr
o
bl
em
s
[8
],
[9
]
.
T
he
m
ai
n ad
v
a
nt
a
ge of
A
N
N
i
s
t
h
e ab
i
l
i
t
y
t
o
per
f
or
m
c
om
pl
ex
pr
oc
es
s
i
ng t
as
k
i
n or
der
t
o f
i
n
d t
h
e
r
el
at
i
ons
h
i
p
be
t
w
ee
n
i
np
ut
s
an
d o
ut
p
ut
s
[
1
0]
.
I
n ot
h
er
w
or
ds
,
A
N
N
i
s
a
n ac
c
ur
at
e pr
edi
c
t
i
on
t
o
ol
t
hat
i
s
us
ed t
o
pr
ed
i
c
t
or
f
or
ec
as
t
f
ut
ur
e
out
p
ut
bas
e
d o
n pr
e
v
i
ous
dat
a.
G
e
ner
a
l
l
y
,
A
N
N
c
o
ns
i
s
t
s
of
t
he
i
n
t
er
c
onn
ec
t
ed
el
em
ent
s
pr
oc
es
s
i
n
g
de
v
i
c
es
k
n
o
w
n
as
n
eur
o
ns
.
A
N
N
i
s
t
r
ai
n
ed
t
hr
oug
h
t
he
adj
us
t
m
ent
of
w
e
i
ght
a
nd bi
as
es
par
am
et
er
s
bet
w
een n
eur
o
ns
.
F
i
gur
e
1
s
ho
w
s
Mu
l
t
i
l
a
y
er
F
eedf
or
w
ar
d N
e
ur
al
N
et
w
o
r
k
ar
c
hi
t
ec
t
ur
e t
ha
t
c
ons
i
s
t
s
of
t
hr
ee
t
y
pes
of
l
a
y
er
,
a
n i
np
ut
l
a
y
er
,
a
hi
d
den
l
a
y
er
an
d
an
o
ut
put
l
a
y
er
.
E
ac
h
l
a
y
er
c
ons
i
s
t
s
of
num
ber
o
f
neur
ons
w
h
i
c
h
i
s
c
onnec
t
e
d t
o
t
h
e ot
her
n
e
ur
ons
i
n t
h
e n
ex
t
l
a
y
e
r
.
E
ac
h ne
ur
on r
ec
ei
v
es
a
s
i
gna
l
f
r
om
t
he
neur
o
ns
i
n t
he pr
e
v
i
o
us
l
a
y
er
and bot
h ar
e c
on
nec
t
ed
t
o eac
h ot
her
b
y
a s
et
of
s
y
napt
i
c
w
e
i
g
ht
s
and b
i
as
es
.
A
s
c
an b
e s
e
en i
n
F
i
g
ur
e
1
,
i
s
a s
y
n
a
pt
i
c
w
e
i
gh
t
be
t
w
ee
n i
npu
t
and h
i
d
den
l
a
ye
r
,
i
s
a s
y
n
apt
i
c
w
e
i
g
h
t
bet
w
een
hi
dd
en a
nd
out
put
l
a
y
er
m
ean
w
hi
l
e
i
s
t
he b
i
as
.
E
ac
h neur
on i
n
t
h
e
pr
e
v
i
ous
l
a
y
er
i
s
m
ul
t
i
p
l
i
e
d
w
it
h
it
s
o
w
n
as
s
oc
i
a
t
ed
w
ei
g
ht
v
a
l
u
e.
T
hen,
t
he
w
e
i
g
ht
e
d
i
n
put
s
and
b
i
as
ar
e
s
um
m
ed
and
pas
s
ed
t
hr
o
ugh
t
r
ans
f
er
f
u
nc
t
i
on
,
w
hi
c
h
nor
m
al
l
y
m
odel
l
ed as
a
p
ur
e l
i
ne
ar
(
pur
el
i
n)
or
l
og
s
i
gm
oi
d (
l
ogs
i
g)
f
unc
t
i
on.
T
he pr
edi
c
t
ed
out
p
ut
m
a
y
obt
ai
n af
t
er
ap
p
l
y
i
ng
t
r
ans
f
er
f
unc
t
i
on t
o t
h
e
w
e
i
ght
ed
i
np
ut
a
nd
bi
as
.
Hidden layer
,
j
Output layer
,
k
Input layer
,
i
output
Input
,
x
i
W
ji
W
kj
2
n
1
2
3
n
1
1
b
.
.
.
.
.
.
b
F
i
gur
e
1
.
M
ul
t
i
l
a
y
er
F
ee
df
or
w
ar
d N
e
ur
al
N
et
w
or
k
ar
c
hi
t
ec
t
ur
e
Mos
t
of
t
he r
es
ear
c
her
s
i
m
pl
e
m
ent
ed t
he t
r
i
al
a
nd
er
r
or
t
ec
hni
q
ue t
o
det
er
m
i
ne an
o
pt
i
m
um
A
N
N
par
am
et
er
s
[
11]
,
[
12
]
.
T
her
ef
or
e,
t
o
g
et
a
bet
t
er
ac
c
ur
ac
y
of
A
N
N
pr
ed
i
c
t
i
on,
appr
o
pr
i
at
e A
N
N
p
ar
am
et
er
s
s
el
ec
t
i
on us
i
ng op
t
i
m
i
z
at
i
on t
ec
hni
que n
eed t
o be f
or
m
ul
at
ed.
T
hi
s
w
as
done
b
y
h
y
b
r
i
d
i
z
i
ng
v
ar
i
ous
opt
i
m
i
z
at
i
on
t
ec
hni
ques
w
i
t
h
A
N
N
m
odel
t
o
aut
om
at
i
c
al
l
y
f
i
nd t
he
opt
i
m
u
m
A
N
N
par
a
m
et
er
s
as
oppos
ed t
o
t
he
t
r
i
al
an
d er
r
or
t
ec
h
ni
que.
N
o
w
ad
a
y
s
,
t
he
m
aj
or
us
age
of
el
ec
t
r
i
c
i
t
y
i
n
c
om
m
er
c
i
al
an
d
r
es
i
d
ent
i
a
l
s
ec
t
or
s
c
om
e
s
f
r
o
m
t
he c
hi
l
l
er
p
l
a
nt
w
her
e i
t
pr
od
uc
es
c
hi
l
l
ed
w
at
e
r
f
or
t
he c
oo
l
i
ng s
y
s
t
em
t
o t
h
e b
ui
l
di
ng.
H
eat
i
ng,
V
ent
i
l
at
i
on
and
C
ool
i
ng
(
H
V
A
C
)
c
o
nt
r
i
b
ut
es
m
or
e t
han 2
4%
of
t
he
ene
r
g
y
us
e i
n
t
h
e
c
o
m
m
er
c
i
al
bui
l
di
ng
[
1
3]
.
I
t
i
s
es
s
ent
i
a
l
t
o i
m
pl
em
ent
ener
g
y
ef
f
i
c
i
enc
y
an
d ener
g
y
s
av
i
ng
i
n t
h
e
bui
l
d
i
ng t
o r
ed
uc
e t
he e
l
e
c
t
r
i
c
i
t
y
c
os
t
of
t
he c
hi
l
l
er
pl
a
nt
.
T
her
ef
or
e,
pr
edi
c
t
i
n
g t
he ener
g
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
EC
S
IS
S
N
:
2
502
-
4
752
D
ev
el
o
pm
ent
of
H
y
br
i
d A
r
t
i
f
i
c
i
a
l
N
e
u
r
al
N
et
w
or
k
f
or
.
.
.
(
W
an N
a
z
i
r
a
h W
an
M
d
A
d
n
an)
139
c
ons
um
pt
i
on bas
ed o
n
i
np
ut
v
ar
i
a
bl
es
af
f
ec
t
i
ng f
ac
t
o
r
i
s
ne
ede
d a
nd t
hi
s
i
s
on
e of
t
he
m
aj
or
ana
l
y
s
es
an
d f
oc
us
es
i
n t
hi
s
pap
er
.
A
n
y
c
h
ang
es
i
n
t
he i
np
ut
v
ar
i
ab
l
es
m
a
y
v
ar
y
t
he en
er
g
y
c
ons
um
pt
i
on.
A
l
t
hou
gh t
h
e A
N
N
has
be
en s
t
ud
i
ed
i
n m
an
y
app
l
i
c
at
i
o
ns
,
as
f
ar
as
t
he aut
h
or
s
ar
e
a
w
ar
e,
t
her
e
ar
e f
e
w
w
or
k
s
r
epor
t
ed
on
M&
V
m
odel
i
n
g of
c
hi
l
l
er
s
y
s
t
em
us
i
ng
A
N
N
.
T
he ai
m
of
t
hi
s
pa
per
i
s
t
o
de
v
e
l
o
p
an
ac
c
ur
at
e
M&
V
bas
e
l
i
ne
e
n
er
g
y
m
odel
us
i
ng
H
y
br
i
d
A
r
t
i
f
i
c
i
al
N
e
ur
al
N
et
w
or
k
(
H
A
N
N
)
f
or
c
hi
l
l
er
s
y
s
t
em
.
H
y
br
i
di
z
at
i
on of
A
N
N
w
i
t
h E
v
o
l
ut
i
o
nar
y
P
r
ogr
am
m
i
ng (
E
P
)
i
s
i
m
pl
em
ent
ed t
o opt
i
m
i
z
e t
h
e neur
a
l
ne
t
w
or
k
t
r
ai
ni
n
g pr
oc
es
s
and t
o s
el
ec
t
t
he o
pt
i
m
al
v
al
ues
of
A
N
N
p
ar
am
et
er
s
,
w
hi
c
h
ar
e
i
n
i
t
i
al
w
e
i
gh
t
s
and
b
i
as
es
.
T
hi
s
bas
el
i
ne
m
odel
us
i
n
g
a
t
es
t
dat
a
of
c
hi
l
l
er
s
y
s
t
em
i
n a
c
om
m
e
r
c
i
al
bui
l
d
i
n
g
i
n
K
u
al
a
L
um
pur
a
nd
t
hi
s
m
odel
us
e
d
t
o c
al
c
u
l
at
e
t
he
adj
us
t
ed
bas
el
i
n
e e
ner
g
y
h
enc
e t
o
qu
ant
i
f
y
en
er
g
y
s
av
i
ng.
T
he ov
er
al
l
s
t
r
uc
t
ur
e of
t
hi
s
paper
i
s
or
ga
ni
z
e
d as
f
ol
l
o
w
s
:
S
ec
t
i
on I
I
ex
pl
a
i
n
e
d t
he
m
et
hodol
o
g
y
i
nc
l
u
di
n
g bas
el
i
ne en
er
g
y
de
v
el
opm
ent
and s
av
i
n
g c
al
c
ul
at
i
on.
M
e
an
w
h
i
l
e,
s
ec
t
i
on
I
I
I
des
c
r
i
b
ed t
he r
es
u
l
t
s
,
a
n
al
y
s
es
and
di
s
c
us
s
i
o
n.
F
i
na
l
l
y
,
s
ec
t
i
on I
V
c
onc
l
ud
es
t
h
e pa
per
.
2.
M
&
V
M
o
d
e
l
D
e
v
e
l
o
p
m
e
n
t
T
he dev
e
l
o
pm
ent
of
M&
V
H
A
N
N
Mod
el
i
s
di
v
i
d
ed
i
n
t
o t
w
o p
has
es
,
1)
M&
V
B
a
s
el
i
n
e
E
ner
g
y
D
ev
el
opm
ent
phas
e an
d 2)
P
os
t
-
r
et
r
of
i
t
S
av
i
n
g C
a
l
c
ul
a
t
i
o
n p
has
e
as
i
n
F
i
gur
e
2
.
START
Load Baseline
Data
Load Post Retrofit
Data
AN
N
structure
determination
Adjusted Base
line
Modeling
Performance
Evaluation and
mode
l
selection
HAN
N
Optimization
Model Evaluation
Saving Calculation
EN
D
Input
:
operating time
,
refrigerant tonnage and
differential temperature
Output
:
ene
rgy use
(
Targetted output
)
Ene
rgy Avoided
(
Saving
)
Saving Percentage
EPHAN
N
Load HAN
N
N
e
twork
Saving Calculation
M
&
V B
aseline Energy Develo
pment
F
i
gur
e
2
.
M
&
V H
AN
N
f
l
o
w
c
h
a
r
t
In
t
hi
s
s
t
ud
y
,
M&
V
dat
a ar
e
c
ol
l
ec
t
e
d f
r
o
m
an aut
om
at
ed c
ent
r
al
i
z
ed c
ont
r
ol
of
bui
l
di
ng
’
s
a
ir
-
c
ond
i
t
i
oni
ng s
y
s
t
em
,
B
ui
l
di
ng
A
ut
om
at
i
on
S
y
s
t
em
(
B
A
S
)
t
h
at
i
s
i
n
on
e of
t
he c
om
m
er
c
i
al
bui
l
d
i
ngs
i
n
K
ua
l
a
L
um
pur
,
Mal
a
y
s
i
a.
T
her
e
ar
e
t
w
o
t
y
p
es
of
dat
a:
1)
2
,
92
8
ba
s
el
i
ne
d
at
a
f
r
om
S
ept
em
ber
2015
–
D
ec
e
m
ber
2015 an
d 2)
2,
1
84
pos
t
-
r
et
r
of
i
t
da
t
a f
r
om
S
ept
em
ber
2016
–
N
ov
em
ber
2016.
T
hr
ee
i
n
p
ut
v
ar
i
ab
l
es
ar
e
m
eas
ur
ed
i
n
d
ev
el
o
pi
ng
t
h
e
M
&
V
H
A
N
N
Mod
el
i
.
e.
1)
oper
a
t
i
n
g
t
i
m
e:
hour
of
t
he
d
a
y
,
f
r
o
m
1
t
o
24,
2)
r
e
f
r
i
ger
ant
t
o
nna
ge:
t
he
c
o
ol
i
ng
c
ap
ac
i
t
y
or
heat
r
em
ov
a
l
c
ap
ac
i
t
y
t
o
i
n
di
c
at
e
t
he
c
apac
i
t
y
or
s
i
z
e
of
t
he r
ef
r
i
ger
at
i
on
pl
ant
,
an
d 3)
d
i
f
f
er
ent
i
a
l
t
em
per
at
ur
e:
t
he d
i
f
f
er
enc
e i
n t
em
per
at
ur
e be
t
w
e
en
i
nl
et
t
em
per
at
ur
e (
t
em
per
at
ur
e of
c
ool
i
n
g
w
at
er
f
r
o
m
c
ool
i
n
g
t
o
w
er
i
n
t
o
c
ond
ens
er
)
and
out
l
et
t
e
m
per
at
ur
e
(
t
e
m
per
at
ur
e
of
c
ool
i
ng
w
at
er
f
r
o
m
c
ondens
er
t
o c
ool
i
n
g t
o
w
er
)
.
T
hes
e par
am
et
er
s
ar
e as
s
i
gned as
A
N
N
i
np
ut
and t
he
t
ar
get
e
d o
ut
p
ut
f
or
t
he
bas
el
i
ne
i
s
t
h
e h
our
l
y
e
l
ec
t
r
i
c
a
l
ener
g
y
c
o
ns
um
pt
i
on (
b
as
e
l
i
n
e m
eas
ur
ed
ener
g
y
)
,
k
W
h.
F
or
t
he pos
t
-
r
et
r
of
i
t
,
hour
l
y
e
l
ec
t
r
i
c
a
l
e
n
er
g
y
c
ons
um
pt
i
on (
pos
t
-
r
et
r
of
i
t
m
eas
ur
ed
ener
g
y
)
i
s
us
e
d t
o c
a
l
c
ul
at
e t
he s
a
v
i
ng.
F
i
g
ur
e
3
a
nd
F
i
gur
e
4
s
ho
w
t
he
hour
l
y
el
ec
t
r
i
c
al
e
ner
g
y
c
ons
um
pt
i
on f
or
bas
el
i
n
e a
nd p
os
t
-
r
et
r
of
i
t
r
es
pec
t
i
v
e
l
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
25
02
-
4
752
I
J
E
EC
S
V
o
l.
8
,
N
o.
1
,
O
c
t
o
ber
20
17
:
1
37
–
1
45
140
2.
1 M
&
V
B
as
el
i
n
e
E
n
e
r
g
y
D
ev
el
o
p
m
en
t
P
h
a
se
I
n t
h
i
s
ph
as
e,
A
N
N
s
t
r
uc
t
ur
e an
d p
ar
am
et
er
s
need t
o
be d
et
er
m
i
ned.
F
or
t
h
i
s
pap
er
,
t
he
num
ber
of
neur
ons
i
n t
h
e hi
dden
l
a
y
er
i
s
s
et
bet
w
ee
n 5
and 20 n
eur
ons
on
l
y
.
T
hi
s
m
eans
t
hat
a
t
ot
a
l
16 s
t
r
uc
t
ur
es
ar
e ev
al
u
at
ed
.
S
t
r
uc
t
ur
es
w
i
t
h o
ne hi
d
de
n l
a
y
er
ar
e c
hos
en as
s
ev
er
al
aut
h
or
s
f
ound t
ha
t
s
i
m
pl
er
net
w
or
k
s
ar
e bet
t
er
d
ue
t
o l
es
s
m
e
m
or
y
[1
4
], [1
5
]
T
hes
e A
N
N
s
t
r
uc
t
ur
es
ar
e
t
r
ai
n
ed
w
i
t
h
t
he
par
am
et
er
s
et
t
i
n
g
as
i
n
T
abl
e
1
.
T
he
t
r
ai
n
i
n
g
al
gor
i
t
hm
us
ed
and
r
ec
om
m
ended b
y
t
h
e
M
A
T
LA
B
a
nd m
os
t
l
y
us
ed
t
o
d
et
er
m
i
ne
t
h
e er
r
or
i
s
t
r
a
i
n
l
m
(
Lev
enber
g
-
Mar
qu
ar
dt
)
f
or
m
os
t
c
ondi
t
i
on a
nd d
ef
aul
t
al
gor
i
t
hm
[
16]
.
F
i
gur
e
3
.
B
a
s
e
lin
e
E
le
c
t
r
ic
a
l E
n
e
r
g
y
C
ons
um
pt
i
on
F
i
gur
e
4
.
Po
s
t
-
R
et
r
of
i
t
E
l
ec
t
r
i
c
al
E
ner
g
y
C
ons
um
pt
i
on
F
or
t
hi
s
bas
el
i
n
e m
odel
,
70
%
of
t
he dat
a i
s
a
l
l
oc
a
t
ed f
or
t
r
ai
n
i
ng,
1
5%
f
or
v
al
i
da
t
i
on an
d
15%
f
or
t
es
t
i
n
g.
T
he
s
e
l
ec
t
ed
t
r
a
ns
f
er
f
unc
t
i
ons
ar
e
l
o
gs
i
g
f
or
h
i
d
den
l
a
y
e
r
a
nd
pur
el
i
n
f
or
t
he
out
p
ut
l
a
y
er
.
T
he
i
np
ut
v
ar
i
abl
es
ar
e
nor
m
al
i
z
ed
i
n
t
he
r
ange
of
-
1 t
o 1
s
o t
h
at
al
l
t
he
i
np
ut
s
ar
e
at
a
c
om
par
abl
e
r
a
nge
an
d
t
o
ens
ur
e
t
ha
t
al
l
t
h
e
da
t
a
i
s
equ
al
l
y
di
s
t
r
i
but
ed
b
et
w
een
t
he
i
np
ut
v
ar
i
ab
l
es
an
d
t
h
e ou
t
put
s
[
17]
.
T
hen,
t
he
A
N
N
out
p
ut
s
ar
e
de
-
nor
m
al
i
z
ed
t
o g
et
t
he
pr
e
di
c
t
ed
el
ec
t
r
i
c
a
l
e
ner
g
y
c
ons
um
pt
i
on.
I
n t
he t
r
ai
ni
ng pr
oc
es
s
,
A
N
N
t
r
i
es
t
o f
i
nd t
he c
or
r
el
at
i
o
n bet
w
een
i
np
ut
an
d p
r
edi
c
t
ed
out
p
ut
ac
c
or
di
n
g t
o t
h
e gi
v
en s
et
of
i
np
ut
an
d t
ar
ge
t
ed out
put
.
A
N
N
c
r
eat
es
t
he i
n
put
-
o
ut
pu
t
m
appi
ng b
y
adj
us
t
i
ng t
he
w
ei
g
ht
s
and b
i
as
es
at
e
ac
h i
t
er
at
i
o
n t
o m
i
ni
m
i
z
e t
he er
r
o
r
bet
w
ee
n t
h
e
t
ar
get
e
d
a
nd pr
edi
c
t
ed out
p
ut
.
T
abl
e
1
.
A
N
N
P
ar
am
et
er
S
et
t
i
ng
T
r
ai
ni
ng
A
l
gor
i
t
h
m
Lev
enber
g
-
M
ar
qua
r
dt
(
LM
)
D
at
a
di
v
i
s
i
on
f
un
c
t
i
on
D
i
v
i
de bl
oc
k
(
70/
15/
15)
T
r
ans
f
er
f
unc
t
i
on
–
hi
dden l
ay
er
l
ogs
i
g
T
r
ans
f
er
f
unc
t
i
on
–
out
put
l
ay
er
pur
el
i
n
I
n
or
der
t
o
obt
a
i
n
t
he
o
pt
i
m
u
m
i
ni
t
i
a
l
w
e
i
g
ht
s
and
bi
a
s
es
par
am
et
er
s
,
t
he
A
N
N
need
t
o
be t
r
a
i
ne
d a
nd o
pt
i
m
i
z
e
d
us
i
ng
E
v
ol
ut
i
on
ar
y
P
r
ogr
a
m
m
i
ng (
E
P
)
w
i
t
h
t
he
obj
ec
t
i
v
e f
unc
t
i
o
n t
o
m
a
x
i
m
i
z
e C
o
ef
f
i
c
i
ent
of
C
or
r
el
at
i
on (
R
)
.
T
hi
s
h
y
br
i
d m
et
ho
d i
s
c
al
l
ed E
v
o
l
ut
i
o
nar
y
P
r
ogr
am
m
i
ng
H
y
br
i
d
w
i
t
h
A
r
t
i
f
i
c
i
al
N
eur
a
l
N
et
w
or
k
(
E
P
H
A
N
N
)
.
I
n
t
h
e
ot
her
w
or
ds
,
t
hi
s
E
P
H
A
N
N
i
s
t
r
ai
n
ed
t
o
m
i
ni
m
i
z
e t
h
e er
r
or
d
ur
i
n
g t
h
e t
r
ai
ni
n
g pr
oc
es
s
.
E
P
i
s
one
of
t
he
E
v
ol
u
t
i
o
n
ar
y
A
l
gor
i
t
hm
s
t
oc
has
t
i
c
op
t
i
m
i
z
at
i
ons
t
ec
hn
i
q
ues
,
or
i
g
i
nat
ed
f
r
o
m
t
he
r
es
ear
c
h
of
Law
r
e
nc
e
J
.
F
ogel
i
n
196
0.
I
t
i
s
i
ns
pi
r
ed
b
y
t
he
t
h
eor
y
of
na
t
ur
al
s
el
ec
t
i
o
n
and
e
v
o
l
ut
i
o
n
[
18]
.
W
ho
i
s
f
i
t
enough
t
o
c
op
y
t
he
m
s
el
v
es
w
i
l
l
s
ur
v
i
v
e
a
nd
w
ho
ar
e
u
nf
i
t
ev
e
nt
u
al
l
y
go ex
t
i
nc
t
.
EPH
AN
N
f
l
o
w
c
h
a
r
t
i
s
i
l
l
us
t
r
at
ed as
i
n
F
i
gur
e
5
.
E
P
H
A
N
N
s
t
ar
t
s
w
i
t
h t
he r
an
dom
num
ber
i
ni
t
i
a
l
i
z
a
t
i
o
n
of
i
n
i
t
i
al
w
ei
ght
s
and
bi
as
es
b
as
ed
on
t
he
num
ber
of
neur
o
ns
i
n
t
he
hi
d
den
l
a
y
er
s
.
S
ec
on
dl
y
,
t
he
f
i
t
nes
s
f
unc
t
i
on
i
s
e
v
al
uat
ed
w
her
e
A
N
N
i
s
t
r
a
i
n
ed
t
o
f
i
nd
t
h
e
m
ax
i
m
u
m
v
al
ue
of
R
.
T
he m
ax
i
m
u
m
and m
i
ni
m
u
m
v
al
ues
of
R
,
w
e
i
g
ht
s
and b
i
as
es
ar
e
det
er
m
i
ned i
n or
d
er
t
o
c
al
c
ul
a
t
e t
he n
ex
t
pr
oc
es
s
.
time (hours)
0
500
1000
1500
2000
2500
3000
Electrical Energy Consumption (kWh)
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
time (hours)
0
500
1000
1500
2000
2500
Electrical Energy Consumption (kWh)
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
EC
S
IS
S
N
:
2
502
-
4
752
D
ev
el
o
pm
ent
of
H
y
br
i
d A
r
t
i
f
i
c
i
a
l
N
e
u
r
al
N
et
w
or
k
f
or
.
.
.
(
W
an N
a
z
i
r
a
h W
an
M
d
A
d
n
an)
141
T
hen,
t
he
m
ut
at
i
o
n
pr
oc
es
s
s
t
ar
t
ed
w
h
er
e
eac
h
par
e
nt
r
epl
i
c
at
es
i
nt
o
a
ne
w
po
pu
l
at
i
on
(
of
f
s
pr
i
ng)
.
E
ac
h of
of
f
s
pr
i
ng i
s
m
ut
at
ed ac
c
or
di
n
g t
o
G
aus
s
i
an m
ut
at
i
o
n.
T
he A
N
N
i
s
t
r
ai
n
ed f
or
t
he s
ec
on
d t
i
m
e t
o det
er
m
i
ne t
h
e n
e
w
R
.
N
ex
t
,
par
e
nt
s
ar
e c
om
bi
ned
w
i
t
h
of
f
s
pr
i
ngs
bef
or
e t
he
s
el
ec
t
i
on
pr
oc
es
s
.
D
ur
i
ng
t
he
s
el
ec
t
i
on
pr
oc
es
s
,
par
e
nt
s
and
of
f
s
pr
i
ng
c
o
m
pet
e
t
o
s
ur
v
i
v
e
an
d
t
he
bes
t
s
o
l
ut
i
ons
w
i
t
h t
he
s
el
ec
t
ed p
ar
am
et
er
s
ar
e r
et
a
i
ne
d
t
o b
e
par
e
n
t
s
of
t
he
nex
t
gener
at
i
on.
B
ef
or
e s
t
ar
t
s
t
he
nex
t
e
v
o
l
ut
i
on
pr
oc
es
s
,
a
c
onv
er
g
enc
e t
es
t
i
s
ex
e
c
ut
ed
t
o
c
hec
k
w
het
h
er
t
o c
o
nt
i
nu
e
or
s
t
op
t
he
ev
ol
u
t
i
o
n pr
oc
es
s
.
START
Ge
nerate ANN
Initial
we
ights and biases
Calculate Fitness
2
(
Trai
n AN
N
to find R
)
Mutate parents to produce
offspring
(
Gaussian Distribution
)
Dete
rmine max and min R
Dete
rmine max and min AN
N
Initial
we
ights and biases
Calculate Fitness
1
(
Trai
n AN
N
to find R
)
Combine
parents
and offspring
Evaluate se
lection
Define
new
gene
ration
Fitness
(
R
)
converge
?
EN
D
N
O
YES
F
i
gur
e
5
.
EPH
AN
N
f
l
o
w
c
h
a
r
t
T
o
ev
a
l
uat
e
t
he
m
odel
p
er
f
or
m
anc
e
and
ac
c
ur
ac
y
,
A
N
N
pr
e
di
c
t
e
d
out
p
ut
w
i
l
l
b
e
c
o
m
par
ed
w
i
t
h t
h
e t
ar
g
et
e
d out
put
us
i
ng
s
ev
er
al
per
f
or
m
anc
e
f
unc
t
i
ons
.
F
or
t
h
i
s
s
t
ud
y
,
R
i
s
s
el
ec
t
ed
as
t
h
e obj
ec
t
i
v
e f
unc
t
i
on
t
o
opt
i
m
i
z
e t
h
e per
f
o
r
m
anc
e of
t
hes
e t
w
o
m
odel
s
.
R
m
eas
ur
es
t
he
s
t
r
en
gt
h
of
as
s
oc
i
at
i
o
n
and
t
he
d
i
r
ec
t
i
on
of
a
l
i
nea
r
r
el
at
i
ons
h
i
p
bet
w
een
t
w
o
v
ar
i
ab
l
es
.
T
he
hi
g
her
v
al
ue
of
R
(
t
he c
l
os
er
R
t
o
1)
i
nd
i
c
at
es
t
h
e s
t
r
o
ng l
i
n
ear
c
or
r
el
at
i
on
or
i
n ot
her
w
or
ds
,
t
he
hi
g
her
s
i
m
i
l
ar
i
t
i
es
bet
w
ee
n
t
he
t
ar
g
et
ed
a
nd t
he
pr
e
di
c
t
ed
ou
t
put
[
1
9]
.
O
ut
of
16
E
P
H
A
N
N
s
t
r
uc
t
ur
es
,
onl
y
one
i
s
s
el
e
c
t
ed as
t
h
e b
as
el
i
ne m
odel
bas
ed
on t
he h
i
g
hes
t
v
a
l
ue of
R
.
O
t
her
t
ha
n t
h
at
,
ot
her
p
er
f
or
m
anc
e c
r
i
t
er
i
a ar
e
al
s
o us
ed t
o
v
al
i
da
t
e t
he
m
odel
ac
c
ur
ac
y
,
w
hi
c
h
ar
e
Me
a
n
S
qu
ar
e
E
r
r
or
(
MS
E
)
,
S
t
and
ar
d
E
r
r
or
(
S
E
)
and
Mean
A
bs
ol
ut
e
P
er
c
ent
age
E
r
r
or
(
MA
P
E
)
bet
w
e
en t
h
e m
eas
ur
ed and pr
edi
c
t
ed v
al
u
es
.
T
he l
ow
er
v
a
l
ues
of
MS
E
,
S
E
and
MA
P
E
i
nd
i
c
a
t
e t
h
at
t
h
e m
or
e ac
c
ur
at
e t
h
e r
es
ul
t
s
.
T
he
m
at
he
m
at
i
c
al
r
epr
es
en
t
at
i
on
of
R
,
MS
E
,
S
E
and
MA
P
E
ar
e
s
ho
w
n
i
n
t
he
E
q
uat
i
on
(1
)
-
E
qu
at
i
on (
4)
.
=
∑
−
(
∑
)
∑
∑
−
(
)
∑
−
(
)
(
1)
=
∑
−
=
−
−
(
2)
=
∑
−
=
(
3)
=
∑
−
=
(
4)
w
he
re
i
s
t
h
e
num
ber
of
ob
s
er
v
at
i
on
i
n t
he
dat
a s
e
t
,
i
s
t
he
t
ar
ge
t
ed
o
ut
pu
t
d
at
a,
is
t
he pr
e
di
c
t
e
d o
ut
p
ut
d
at
a f
r
om
t
he A
N
N
,
and
i
s
t
h
e n
u
m
ber
of
i
nput
v
ar
i
a
bl
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
25
02
-
4
752
I
J
E
EC
S
V
o
l.
8
,
N
o.
1
,
O
c
t
o
ber
20
17
:
1
37
–
1
45
142
2.
2 P
o
st
-
R
e
tr
o
fi
t S
a
v
i
n
g
C
a
l
c
u
l
a
ti
o
n
P
h
a
s
e
I
n t
hi
s
p
has
e,
t
h
e pos
t
-
r
et
r
of
i
t
i
nput
dat
a i
s
us
ed t
o
det
er
m
i
ne t
he a
dj
us
t
ed b
a
s
el
i
n
e
ener
g
y
henc
e
t
o
qu
ant
i
f
y
s
av
i
ng.
T
he
pos
t
-
r
et
r
of
i
t
i
n
put
dat
a
i
s
l
o
ade
d as
an
i
np
ut
i
nt
o
t
h
e
s
el
ec
t
ed
E
P
H
A
N
N
bas
e
l
i
n
e m
odel
t
o pr
e
di
c
t
t
he
ou
t
put
.
T
hi
s
pr
ed
i
c
t
ed
out
put
i
s
k
now
n as
adj
us
t
ed
b
as
el
i
ne
ene
r
g
y
.
A
c
c
or
di
n
g
t
o
I
P
M
V
P
,
s
a
v
i
n
g
or
av
oi
d
ed
e
ner
g
y
us
e
i
s
obt
a
i
ne
d
f
r
o
m
t
he
di
f
f
er
enc
e bet
w
ee
n t
h
e adj
us
t
e
d
bas
el
i
n
e e
ner
g
y
a
nd
pos
t
-
r
et
r
of
i
t
m
eas
ur
ed e
ner
g
y
as
s
t
at
ed i
n E
qu
at
i
on
(
5)
.
=
–
−
(
5)
3.
R
e
su
l
t
s an
d
D
i
scu
s
si
o
n
T
hi
s
r
es
ul
t
par
t
i
s
di
v
i
de
d i
n
t
o t
w
o s
ec
t
i
o
ns
,
t
he bas
e
l
i
n
e ener
g
y
de
v
e
l
opm
ent
r
es
u
l
t
and
pos
t
-
r
et
r
of
i
t
s
av
i
n
g c
al
c
u
l
at
i
on r
es
ul
t
.
I
n
t
he
b
as
el
i
ne
e
ner
g
y
de
v
el
o
pm
ent
,
E
P
H
A
N
N
i
s
de
v
el
o
ped
w
i
t
h
t
h
e
obj
ec
t
i
v
e
f
unc
t
i
o
n
t
o m
ax
i
m
i
z
e t
h
e v
al
u
e of
R
.
R
and
om
nu
m
ber
s
o
f
i
ni
t
i
a
l
w
ei
ght
s
an
d bi
as
es
ar
e i
ni
t
i
a
l
i
z
e
d
an
d
net
w
or
k
s
t
r
uc
t
ur
es
ar
e t
r
a
i
ne
d and o
pt
i
m
i
z
e
d w
i
t
h t
he di
f
f
er
ent
c
om
bi
nat
i
ons
of
neur
ons
i
n
hi
d
den
l
a
y
er
.
A
s
pr
e
v
i
o
us
l
y
m
ent
i
o
ned,
5
-
20
num
ber
s
ar
e n
eur
ons
ar
e c
o
ns
i
der
ed
f
or
t
hi
s
opt
i
m
i
z
at
i
o
n.
P
r
e
di
c
t
e
d
out
put
an
d
per
f
or
m
anc
e
f
unc
t
i
ons
ar
e
m
eas
ur
ed
and
r
ec
or
ded
f
or
eac
h
t
r
a
in
in
g
p
h
as
e.
I
n
s
el
ec
t
i
n
g t
he
bes
t
net
w
or
k
s
t
r
uc
t
u
r
e,
t
he
v
al
u
e of
R
i
s
e
v
al
uat
e
d an
d t
h
e
num
ber
of
neur
o
ns
,
as
w
el
l
as
t
h
e
i
n
i
t
i
al
w
e
i
gh
t
s
an
d
bi
as
es
,
ar
e
doc
um
ent
ed.
T
he
s
e
l
ec
t
i
on
of
t
he b
es
t
s
t
r
uc
t
ur
e
i
s
bas
e
d on
t
he
m
ax
i
m
u
m
v
al
ue
of
R
as
an
obj
ec
t
i
v
e
f
unc
t
i
on
,
as
w
el
l
as
MA
P
E
,
S
E
a
nd
MS
E
as
add
i
t
i
on
al
c
r
i
t
er
i
a.
La
t
er
,
i
n t
he
po
st
-
r
et
r
of
i
t
s
a
v
i
n
g c
al
c
u
l
at
i
on
,
t
he s
el
ec
t
ed s
t
r
uc
t
ur
e i
s
app
l
i
e
d t
o
obt
ai
n
t
he
adj
u
s
t
ed bas
e
l
i
ne e
ner
g
y
h
en
c
e t
o c
al
c
u
l
at
e
s
av
i
ng.
3.
1 B
asel
i
n
e
E
n
er
g
y D
ev
e
l
o
p
m
en
t
R
esu
l
t
s
F
i
gur
e
6
pr
es
e
nt
s
t
h
e a
v
e
r
age R
of
16 s
t
r
uc
t
ur
es
f
or
E
P
H
A
N
N
.
F
r
om
t
he gr
ap
h,
t
h
e
l
o
w
es
t
a
v
er
a
ge
R
of
0.
97
7
7857
8
i
s
o
bt
a
i
ne
d
f
r
om
t
he
c
om
bi
nat
i
o
n
of
5
n
eur
o
ns
i
n
h
i
dd
en
l
a
y
er
m
eanw
h
i
l
e
t
he
hi
ghes
t
v
a
l
ue of
av
er
ag
e R
i
s
0
.
98
14
08
9,
at
t
a
i
ne
d f
r
om
t
he c
om
bi
nat
i
on
of
19
neur
o
ns
.
T
her
ef
or
e,
hi
d
den
l
a
y
er
w
i
t
h
19
neur
o
ns
w
i
t
h
t
he
t
r
a
i
n
i
n
g
R
i
s
0.
9
793
8,
v
a
l
i
d
at
i
on
R
i
s
0.
98
63
an
d
t
es
t
i
ng
R
i
s
0.
9863
5
as
i
l
l
us
t
r
at
e
d
i
n
F
i
g
ur
e
7
i
s
s
el
ec
t
ed
as
t
h
e
be
s
t
per
f
or
m
anc
e
bas
ed
o
n
t
he
m
ax
i
m
u
m
v
al
ue
of
R
obj
ec
t
i
v
e
f
unc
t
i
o
n.
T
he
i
dea
l
R
i
s
one
and
as
c
an
be
s
een
,
t
he a
v
er
age
R
f
or
al
l
s
t
r
uc
t
ur
es
i
s
ab
ov
e 0
.
97
7.
T
he R
f
or
al
l
s
e
l
ec
t
e
d
v
al
ues
ar
e
hi
g
h an
d c
l
os
e
t
o
un
i
t
y
w
hi
c
h
c
an
be
c
ons
i
der
ed
g
ood
a
nd
ac
c
ep
t
ab
le
[
19]
.
A
par
t
f
r
o
m
t
hat
,
ot
her
per
f
or
m
anc
e
c
r
i
t
er
i
a
s
uc
h
as
MA
P
E
,
M
S
E
and
S
E
ar
e
al
s
o
e
v
a
l
uat
ed.
T
he
v
al
u
e
of
MA
P
E
i
s
8.
72
50%
,
M
S
E
i
s
1708
00.
3
5 an
d S
E
i
s
413
.
56 f
or
s
el
ec
t
ed s
t
r
uc
t
ur
e,
1
9 neur
o
ns
i
n t
he h
i
d
den l
a
y
er
as
i
n
T
abl
e
2
.
I
t
c
an
be c
onc
l
u
ded
t
ha
t
t
her
e
i
s
no
di
r
ec
t
c
or
r
e
l
at
i
o
n b
et
w
ee
n num
ber
of
neur
o
ns
and
R
.
F
i
gur
e
6
.
A
v
er
age
C
oef
f
i
c
i
e
nt
of
C
or
r
el
at
i
on,
R
f
or
16 E
P
H
A
N
N
s
t
r
uc
t
ur
es
.
0
.
9
7
5
0
.
9
7
6
0
.
9
7
7
0
.
9
7
8
0
.
9
7
9
0
.
9
8
0
.
9
8
1
0
.
9
8
2
5
6
7
8
9
1
0
1
1
1
2
1
3
1
4
1
5
1
6
1
7
1
8
1
9
2
0
C
o
ef
f
i
ci
en
t
o
f
C
o
rrel
a
t
i
o
n
, R
N
um
be
r
o
f
ne
ur
o
ns
i
n
hi
dde
n
l
a
y
e
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
EC
S
IS
S
N
:
2
502
-
4
752
D
ev
el
o
pm
ent
of
H
y
br
i
d A
r
t
i
f
i
c
i
a
l
N
e
u
r
al
N
et
w
or
k
f
or
.
.
.
(
W
an N
a
z
i
r
a
h W
an
M
d
A
d
n
an)
143
F
i
gur
e
7
.
T
r
ai
ni
ng,
v
a
l
i
dat
i
o
n an
d t
es
t
i
ng r
es
ul
t
s
f
or
E
P
H
A
N
N
T
he r
es
ul
t
s
of
opt
i
m
al
v
al
ues
f
or
E
P
H
A
N
N
i
s
i
l
l
us
t
r
at
ed i
n
T
abl
e 3
. In
T
abl
e
3
,
t
he
EPH
AN
N
o
p
t
i
m
a
l
v
a
l
ue
f
or
neur
o
ns
i
n
t
he
h
i
d
den
l
a
y
er
i
s
19
w
i
t
h
a
s
et
of
96
i
ni
t
i
a
l
w
e
i
ght
s
a
nd
bi
as
es
.
r
epr
es
ent
s
a s
et
of
w
ei
g
ht
s
b
et
w
e
en
i
n
put
and
hi
dde
n
l
a
y
er
and
i
s
a
s
et
of
w
ei
g
ht
s
be
t
w
ee
n
h
i
dd
en
l
a
y
e
r
a
nd
o
ut
p
ut
.
W
her
eas
,
1
and
2
ar
e
t
h
e
b
i
as
es
f
or
i
np
ut
-
h
i
dd
en
l
a
y
er
an
d h
i
d
den
l
a
y
er
–
ou
t
put
r
es
p
ec
t
i
v
e
l
y
T
hi
s
s
el
ec
t
ed
E
P
H
A
N
N
s
t
r
uc
t
ur
e
w
i
t
h 1
9 num
ber
of
neur
ons
g
i
v
es
s
m
al
l
er
r
or
ba
s
ed on
s
ev
er
a
l
per
f
or
m
anc
e c
r
i
t
er
i
a m
ent
i
one
d ab
ov
e an
d h
as
been
nom
i
nat
e
d as
ba
s
el
i
n
e en
er
g
y
m
odel
f
or
t
he pos
t
-
r
et
r
of
i
t
s
av
i
ng c
a
l
c
ul
at
i
on s
t
age.
T
abl
e
2
.
EP
H
AN
N
P
e
r
f
o
r
m
anc
e E
v
a
l
u
at
i
on F
or
19 N
e
ur
ons
I
n H
i
d
den
La
y
er
.
C
om
pu
t
at
i
on t
i
m
e
4355s
M
APE
8.
725%
MS
E
170800.
35
SE
413.
56
T
abl
e
3
.
O
p
t
im
a
l V
a
lu
e
s
O
f
E
P
H
A
N
N
.
N
um
ber
o
f
neur
ons
19
I
ni
t
i
al
w
ei
ght
s
I
ni
t
i
al
bi
a
s
es
=
⎣
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎡
0
.
4
7
1
9
9
7
0
.
6
7
3
6
9
4
0
.
6
2
5
8
3
3
0
.
7
3
2
2
8
0
0
.
4
2
7
7
5
6
0
.
2
1
1
7
7
2
0
.
9
8
0
4
6
9
0
.
8
3
2
6
6
2
0
.
9
1
5
3
7
4
0
.
6
6
6
6
7
5
0
.
1
7
1
1
7
8
0
.
1
3
4
0
9
4
0
.
6
8
2
9
5
2
0
.
4
4
0
5
1
0
0
.
9
6
7
8
8
5
0
.
8
7
5
1
7
8
0
.
7
5
2
2
2
5
0
.
6
1
9
7
0
5
0
.
5
0
7
5
8
4
0
.
6
1
5
3
2
9
0
.
6
4
0
6
2
7
0
.
4
6
7
7
6
4
0
.
2
5
4
0
2
0
0
.
0
1
3
3
4
3
0
.
2
4
4
6
3
7
0
.
3
5
0
2
7
5
0
.
7
8
9
3
0
5
0
.
5
6
5
8
2
7
0
.
6
1
2
9
0
5
0
.
8
9
9
4
1
9
0
.
7
6
0
5
9
1
0
.
3
3
7
5
9
0
0
.
0
0
2
0
0
7
0
.
3
4
5
0
3
0
0
.
6
3
5
8
1
5
0
.
5
1
6
1
9
9
0
.
4
4
7
4
5
2
0
.
5
1
6
4
8
9
0
.
6
2
4
2
4
2
0
.
5
1
5
6
5
6
0
.
5
1
9
3
8
8
0
.
2
6
5
0
7
7
0
.
1
5
7
1
5
5
0
.
6
2
6
2
8
7
0
.
3
9
8
8
9
8
0
.
9
8
0
6
8
5
0
.
9
4
9
0
2
7
0
.
9
7
7
4
8
8
0
.
2
9
8
5
9
6
0
.
2
2
3
3
0
2
0
.
3
0
3
5
4
8
0
.
1
5
4
4
1
0
0
.
9
9
4
3
8
2
0
.
5
6
6
6
0
8
0
.
9
3
5
5
5
2
0
.
8
7
7
2
4
2
0
.
2
1
5
9
2
3
⎦
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎤
=
[
0
.
1
3
5
2
1
1
0
.
9
7
8
4
1
8
0
.
2
1
4
1
6
7
0
.
1
3
5
5
7
4
0
.
0
5
2
4
6
7
0
.
2
6
4
5
1
8
0
.
7
5
7
1
2
8
0
.
4
3
5
5
4
5
0
.
5
0
6
5
3
1
0
.
4
9
4
5
6
6
0
.
0
9
4
0
6
9
0
.
9
7
0
9
5
3
0
.
4
2
4
7
7
9
0
.
2
7
0
0
4
6
0
.
9
3
5
1
2
2
0
.
2
2
0
5
7
1
0
.
1
8
9
3
9
8
0
.
2
8
0
4
3
2
0
.
6
7
3
9
0
9
]
1
=
⎣
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎡
0
.
5
9
8
0
6
9
0
.
3
8
8
7
1
8
0
.
9
3
5
6
3
2
0
.
8
2
6
8
3
3
0
.
0
5
1
1
4
3
0
.
8
2
7
5
3
8
0
.
3
8
7
6
5
5
0
.
4
4
5
6
2
4
0
.
0
0
9
0
2
8
0
.
2
8
0
0
1
7
0
.
2
9
1
8
7
6
0
.
4
2
9
7
8
7
0
.
7
0
1
1
6
7
0
.
0
1
0
1
5
0
0
.
2
3
6
8
9
2
0
.
4
0
1
3
0
7
0
.
0
8
7
5
6
6
0
.
8
8
3
4
0
7
0
.
4
3
6
4
0
2
⎦
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎤
2
=
[
0
.
3
0
8
6
5
1
]
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
25
02
-
4
752
I
J
E
EC
S
V
o
l.
8
,
N
o.
1
,
O
c
t
o
ber
20
17
:
1
37
–
1
45
144
3.
2 P
o
st
-
R
e
tr
o
fi
t S
a
v
i
n
g
C
a
l
c
u
l
a
ti
o
n
R
e
s
u
l
ts
T
he s
el
ec
t
ed
E
P
H
A
N
N
s
t
r
uc
t
ur
e i
s
app
l
i
ed t
o t
h
e p
os
t
-
r
et
r
of
i
t
dat
a
t
o c
a
l
c
ul
at
e t
he
adj
us
t
ed
b
as
el
i
ne
.
T
he
gr
aph
of
t
he
o
v
er
a
l
l
M
&
V
pr
oc
es
s
i
s
r
epr
es
ent
ed
i
n
F
i
gur
e
8
.
T
he
gr
aph
s
ho
w
s
t
he
m
eas
ur
ed
ener
g
y
c
ons
um
pt
i
on
f
or
bas
el
i
ne
and
pos
t
-
r
et
r
of
i
t
as
w
e
l
l
as
t
he
r
es
ul
t
s
of
bas
el
i
ne
e
ner
g
y
m
odel
an
d
adj
us
t
e
d b
as
el
i
ne
m
odel
f
r
om
E
P
H
A
N
N
.
T
he r
a
nge
of
dat
a f
r
om
0 t
o
200
i
s
i
l
l
us
t
r
at
e
d
f
or
bot
h
b
as
el
i
ne
a
nd
p
or
t
-
r
et
r
of
i
t
out
of
2,
928
d
at
a
f
or
bas
el
i
n
e
and
2,
18
4
f
or
pos
t
-
r
et
r
of
i
t
.
T
he
a
v
o
i
d
ed
ener
g
y
i
s
c
a
l
c
ul
at
e
d
f
r
om
t
he
di
f
f
er
enc
e
bet
w
ee
n
ac
t
ua
l
ener
g
y
c
ons
um
pt
i
on f
or
pos
t
-
r
et
r
of
i
t
and adj
us
t
e
d bas
e
l
i
ne
ener
g
y
.
A
s
c
an
be s
ee
n,
t
he adj
us
t
e
d
bas
el
i
ne
ener
g
y
i
s
s
l
i
g
ht
l
y
hi
g
her
t
ha
n t
h
e m
eas
ur
ed ener
g
y
c
ons
um
pt
i
on.
T
hi
s
i
nd
i
c
at
es
t
h
at
t
her
e
i
s
a
s
av
i
ng
i
n
t
h
e
pos
t
-
r
et
r
of
i
t
phas
e.
T
he
av
oi
de
d
ener
g
y
u
s
e
c
al
c
ul
a
t
ed
f
or
t
hi
s
m
odel
i
s
132,
944.
59 k
W
h t
hat
c
ont
r
i
but
es
t
o 1.
38%
of
s
av
i
n
g p
er
c
ent
ag
e
F
i
gur
e
8
.
M
eas
ur
em
ent
an
d V
er
i
f
i
c
at
i
o
n F
r
am
ew
or
k
4
.
C
o
n
c
l
u
s
i
o
n
T
he
m
ai
n obj
ec
t
i
v
e of
t
hi
s
w
or
k
w
as
t
o de
v
el
op a
n ac
c
ur
at
e bas
e
l
i
ne en
er
g
y
m
odel
f
or
a
c
hi
l
l
er
s
y
s
t
em
henc
e
t
o
det
er
m
i
ne
s
av
i
ng
.
I
n
t
h
i
s
w
or
k
,
M&
V
H
A
N
N
bas
e
l
i
n
e
en
e
r
g
y
m
odel
w
i
t
h
16 c
om
bi
nat
i
ons
of
neur
on
s
hav
e b
een
de
v
el
ope
d t
o
as
s
es
s
t
he opt
i
m
u
m
v
al
ue
of
t
he obj
ec
t
i
v
e
f
unc
t
i
on,
R
.
E
P
i
s
us
ed as
an opt
i
m
i
z
at
i
on t
ec
hni
qu
e
t
o opt
i
m
i
z
e t
he
A
N
N
i
n
i
t
i
a
l
w
ei
ght
s
a
nd
bi
as
es
.
F
or
t
h
e ef
f
ec
t
of
oper
at
i
n
g t
i
m
e,
r
ef
r
i
ger
ant
t
onna
ge a
nd d
i
f
f
er
ent
i
a
l
t
em
per
at
ur
e on
C
oef
f
i
c
i
ent
of
C
or
r
el
at
i
on,
t
he
opt
i
m
u
m
c
ondi
t
i
on
s
ug
g
es
t
ed
b
y
t
he
E
P
H
A
N
N
i
s
t
h
e
c
o
m
bi
nat
i
on
of
19 neur
ons
i
n a hi
dde
n l
a
y
er
.
F
or
f
ut
ur
e
w
or
k
s
,
ot
he
r
opt
i
m
i
z
at
i
on t
ec
hni
ques
s
uc
h as
P
ar
t
i
c
l
e
S
wa
r
m
O
p
t
i
m
i
z
at
i
on
and
A
r
t
i
f
i
c
i
al
B
ee C
ol
on
y
ar
e s
u
gges
t
e
d t
o b
e c
ons
i
d
er
ed
t
o de
v
e
l
op
an
ac
c
ur
at
e
bas
el
i
ne ener
g
y
m
odel
.
A
c
k
n
o
w
l
e
d
g
e
m
e
n
t
O
ur
ut
m
os
t
gr
at
i
t
ud
e goes
t
o Mal
a
y
s
i
a Mi
ni
s
t
r
y
of
E
duc
at
i
on an
d U
n
i
v
er
s
i
t
i
T
ek
nol
ogi
MA
R
A
(
U
i
T
M)
w
ho
ha
v
e s
p
ons
or
ed t
hi
s
p
aper
u
nder
L
E
S
T
A
R
I
G
r
ant
S
c
hem
e,
60
0
-
IR
M
I/D
A
N
A
5/
3/
L
E
S
T
A
R
I
(
016
7/
2
016)
.
R
ef
er
en
ces
[1
]
E
ner
gy
C
om
m
i
s
s
i
on.
M
al
ay
s
i
a
E
ner
gy
S
t
at
i
s
t
i
c
s
20
15.
2
014.
[2
]
M
i
ni
s
t
r
y
C
oor
di
na
t
or
o
f
S
t
r
at
eg
i
c
S
e
c
t
or
s
.
N
at
i
on
al
E
n
er
gy
B
a
l
anc
e 20
14.
2014
.
[3
]
N
Y
D
ahl
an,
M
S
S
haar
i
,
T
A
N
T
P
ut
r
a,
S
M
M
ohd S
hok
r
i
,
H
M
oham
m
a
d.
E
ner
gy
and en
v
i
r
onm
e
nt
a
l
audi
t
f
or
B
angu
nan
M
enar
a
S
er
i
W
i
l
ay
ah
and
B
ang
un
an
K
a
s
t
a
m
.
P
ut
r
aj
ay
a:
A
nal
y
s
i
s
a
n
d
r
ec
o
m
m
e
ndat
i
on
s
.
2
013.
[4
]
F
Bi
ro
l
.
En
e
rg
y
Ef
f
i
c
i
en
c
y
:
M
ar
k
et
R
e
por
t
2
015.
I
nt
.
E
ner
gy
A
genc
y
,
n
o.
d
ec
e
m
b
er
,
20
15.
[5
]
E
f
f
i
c
i
en
c
y
V
al
ua
t
i
on
O
r
gani
z
at
i
on.
I
nt
er
n
at
i
onal
P
er
f
or
m
an
c
e
M
eas
ur
em
en
t
an
d V
er
i
f
i
c
at
i
o
n P
r
ot
o
c
o
l
(I
PM
VP).
2
0
1
2
.
[6
]
O
A
k
i
ns
o
ot
o,
D
D
e C
anha
,
J
H
C
P
r
et
or
i
us
.
E
ner
g
y
s
a
v
i
n
gs
r
epor
t
i
ng an
d unc
er
t
ai
nt
y
i
n M
e
as
ur
em
en
t
& Ve
ri
f
i
c
a
t
i
o
n
.
i
n A
us
t
r
al
a
s
i
a
n U
ni
v
er
s
i
t
i
e
s
P
ow
er
E
ng
i
ne
er
i
ng C
onf
er
enc
e,
A
U
P
E
C
2
014.
C
ur
t
i
n
U
ni
v
er
s
i
t
y
,
P
er
t
h,
A
u
s
t
r
a
l
i
a,
20
14:
1
–
5.
time(hour)
0
50
100
150
200
250
300
350
400
Energy Consumption, kWh
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
Actual Baseline
Baseline Model
Actual Post Retrofit
Adjusted Baseline
Post-Retrofit Period
Baseline Period
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
EC
S
IS
S
N
:
2
502
-
4
752
D
ev
el
o
pm
ent
of
H
y
br
i
d A
r
t
i
f
i
c
i
a
l
N
e
u
r
al
N
et
w
or
k
f
or
.
.
.
(
W
an N
a
z
i
r
a
h W
an
M
d
A
d
n
an)
145
[7
]
S
M
A
r
i
s
,
N
Y
D
ahl
an,
M
N
M
N
aw
i
,
T
A
N
i
z
am
,
M
Z
T
ahi
r
.
Q
uant
i
f
y
i
ng
ener
gy
s
av
i
n
gs
f
or
r
et
r
o
fi
t
c
ent
r
al
i
z
ed hv
ac
s
y
s
t
e
m
s
at
S
el
ang
or
s
t
at
e s
e
c
r
et
ar
y
c
o
m
pl
ex
.
J
.
T
ek
no
l
.
,
2015
;
77(
5)
:
9
3
–
100.
[8
]
T
S
G
una
w
an,
I
Z
Y
aac
ob,
M
K
ar
t
i
w
i
.
A
r
t
i
f
i
c
i
al
N
e
ur
al
N
e
t
w
or
k
B
as
e
d F
a
s
t
E
dge D
e
t
ec
t
i
on
A
l
gor
i
t
h
m
f
or
M
R
I
M
edi
c
al
I
m
ag
es
.
I
ndo
n
es
.
J
.
E
l
ec
t
r
.
E
ng.
C
om
put
.
S
c
i
.
,
201
7;
7(
1)
:
12
3
–
1
30.
[9
]
S
R
A
R
ahi
m
,
I
M
us
i
r
i
n,
M
M
O
t
hm
an
,
M
H
H
us
s
ai
n.
M
ul
t
i
pl
e
D
G
pl
anni
ng
c
on
s
i
d
er
i
n
g di
s
t
r
i
but
i
o
n l
o
s
s
and p
enet
r
at
i
o
n
le
v
e
l
us
i
ng
E
M
E
F
A
-
A
N
N
m
et
hod.
I
nd
one
s
.
J
.
E
l
ec
t
r
.
E
ng.
C
om
put
.
S
c
i
.
,
2017;
7(
1)
:
1
–
8.
[
10]
A
R
i
s
ha
bh,
N
e
ur
al
N
et
w
or
k
s
.
2012.
[
11]
A
E
l
S
hah
at
,
R
J
H
adda
d,
Y
K
al
aan
i
.
A
n
A
r
t
i
f
i
c
i
al
N
eur
al
N
et
w
or
k
M
o
del
f
or
W
i
nd
E
ner
gy
E
s
t
i
m
at
i
on
.
i
n
I
EEE So
u
t
h
e
a
s
t
C
o
n
20
15,
2
015:
1
–
2.
[
12]
F
G
ebben,
S
B
ader
,
B
O
el
m
a
nn.
C
onf
i
g
ur
i
n
g
A
r
t
i
f
i
c
i
a
l
N
eur
a
l
N
et
w
or
k
s
f
or
t
he
P
r
edi
c
t
i
on
o
f
A
v
ai
l
abl
e
E
ner
gy
i
n S
ol
ar
-
P
o
we
r
ed S
en
s
or
N
od
es
.
i
n
I
EEE
,
20
15:
1
–
4.
[
13]
P
r
ogr
es
s
E
n
er
gy
.
C
hi
l
l
er
O
pt
i
m
i
z
at
i
o
n an
d E
ner
gy
E
f
f
i
c
i
en
t
C
hi
l
l
er
s
.
[
14]
M
Fa
s
t, T
Pa
l
m
.
A
ppl
i
c
at
i
on of
ar
t
i
f
i
c
i
al
ne
ur
al
ne
t
w
or
k
s
t
o t
h
e c
ondi
t
i
o
n m
oni
t
or
i
ng and di
a
gnos
i
s
of
a
c
om
bi
ne
d h
eat
a
nd p
ow
er
pl
a
nt
.
E
ner
gy
,
201
0
;
35(
2)
:
1
114
–
1120.
[
15]
A
K
um
ar
,
M
Z
am
an,
N
G
oel
,
V
S
r
i
v
as
t
av
a.
R
enew
abl
e
E
n
er
gy
S
y
s
t
em
D
es
i
gn
by
A
r
t
i
f
i
c
i
al
N
e
ur
a
l
N
et
w
or
k
S
i
m
ul
at
i
o
n A
ppr
oac
h
.
i
n 20
14 I
E
E
E
E
l
e
c
t
r
i
c
a
l
P
ow
er
and E
n
er
gy
C
onf
er
enc
e,
2
014:
1
42
–
147.
[
16]
N
T
ehl
ah,
P
K
aew
pr
ad
i
t
,
I
M
M
uj
t
aba.
A
r
t
i
f
i
c
i
al
n
eur
al
net
w
or
k
ba
s
e
d m
o
del
i
ng
and
opt
i
m
i
z
at
i
o
n o
f
r
ef
i
ne
d pa
l
m
oi
l
pr
o
c
es
s
.
N
e
ur
oc
om
put
i
ng
.
201
6;
2
16:
4
89
–
5
01.
[
17]
N
K
R
N
i
c
hol
as
.
F
or
e
c
a
s
t
i
n
g of
W
i
n
d S
peed
s
a
nd D
i
r
ec
t
i
on
s
w
i
t
h A
r
t
i
f
i
c
i
al
N
eur
a
l
N
et
w
or
k
s
.
2
012.
[
18]
D
B
F
ogel
,
E
C
W
a
s
s
on,
E
M
B
ought
on,
V
W
P
or
t
o.
A
s
t
e
p t
o
w
ar
d c
om
put
er
-
as
s
i
s
t
ed m
am
m
ogr
a
phy
us
i
n
g ev
ol
ut
i
o
nar
y
pr
ogr
am
m
i
ng an
d ne
ur
al
net
w
or
k
s
.
C
anc
er
Let
t
.
19
97;
1
19(
1)
:
93
–
97.
[
19]
M
H
S
hoj
aeef
ar
d,
M
M
E
t
ghani
,
M
T
ahani
,
M
A
k
bar
i
.
A
r
t
i
f
i
c
i
al
neur
al
net
w
or
k
ba
s
ed
m
ul
t
i
-
obj
ec
t
i
v
e
ev
ol
ut
i
o
nar
y
opt
i
m
i
z
at
i
on
of
a heav
y
-
dut
y
di
e
s
el
eng
i
ne.
In
t.
J
. A
u
t
o
m
o
t. E
n
g
.
,
20
12;
2(
4)
.
Evaluation Warning : The document was created with Spire.PDF for Python.