I
nte
rna
t
io
na
l J
o
urna
l o
f
P
o
wer
E
lect
ro
nics
a
nd
Driv
e
S
y
s
t
em
(
I
J
P
E
DS)
Vo
l.
1
7
,
No
.
3
,
Sep
tem
b
er
20
2
6
,
p
p
.
2
0
8
6
~
2
1
0
0
I
SS
N:
2
0
8
8
-
8
6
9
4
,
DOI
: 1
0
.
1
1
5
9
1
/ijp
ed
s
.
v
1
7
.
i
3
.
p
p
2
0
8
6
-
2
1
0
0
2086
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
p
e
d
s
.
ia
esco
r
e.
co
m
Ada
ptive forg
et
-
g
a
ted
BiL
STM
enh
a
nced by DTW
b
a
sed
feature
select
io
n
i
n so
la
r P
V
foreca
sting
Are
Sa
m
ba
s
iv
a
Ra
o
1
,
K
un
a
da
Dha
na
Sree
Dev
i
2
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
G
I
TA
M
S
c
h
o
o
l
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
(
G
S
C
S
E)
,
G
I
TA
M
D
e
e
me
d
t
o
b
e
U
n
i
v
e
r
si
t
y
,
H
y
d
e
r
a
b
a
d
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
D
a
t
a
S
c
i
e
n
c
e
,
G
I
TA
M
S
c
h
o
o
l
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
(
G
S
C
S
E
)
,
G
I
TA
M
D
e
e
me
d
t
o
b
e
U
n
i
v
e
r
si
t
y
,
H
y
d
e
r
a
b
a
d
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ap
r
1
5
,
2
0
2
6
R
ev
is
ed
J
u
l 2
2
,
2
0
2
6
Acc
ep
ted
Au
g
1
4
,
2
0
2
6
G
ro
wi
n
g
e
x
h
a
u
st
i
o
n
o
f
f
u
e
l
re
s
e
r
v
e
s
a
n
d
t
h
e
ir
h
a
rm
fu
l
e
n
v
ir
o
n
m
e
n
t
a
l
i
m
p
a
c
t
s
h
a
v
e
d
ri
v
e
n
t
h
e
s
h
i
ft
t
o
wa
r
d
s
t
h
e
m
a
i
n
te
n
a
n
c
e
o
f
re
n
e
wa
b
le
e
n
e
r
g
y
s
o
u
rc
e
s
li
k
e
s
o
la
r
e
n
e
r
g
y
.
P
V
s
y
ste
m
s
,
h
o
we
v
e
r
,
s
ti
ll
fa
c
e
si
g
n
if
ica
n
t
c
h
a
l
le
n
g
e
s
w
h
i
le
in
te
g
ra
t
i
n
g
re
n
e
wa
b
le
s
o
u
rc
e
s
i
n
t
o
e
x
is
ti
n
g
u
ti
l
it
y
s
y
ste
m
s.
P
a
rt
ic
u
la
rl
y
,
e
n
v
ir
o
n
m
e
n
t
a
l
fe
a
t
u
r
e
s
l
i
k
e
tem
p
e
r
a
t
u
re
,
irr
a
d
ia
n
c
e
,
a
n
d
h
u
m
i
d
i
ty
a
re
n
o
t
p
e
r
fe
c
tl
y
we
l
l
c
o
o
r
d
i
n
a
te
d
w
it
h
t
h
e
p
o
we
r
o
u
t
p
u
t
,
a
r
e
a
lw
a
y
s
a
s
y
n
c
h
r
o
n
o
u
s
i
n
n
a
t
u
re
,
a
n
d
a
ffe
c
t
t
h
e
p
o
we
r
p
re
d
i
c
t
i
o
n
c
o
n
si
d
e
ra
b
l
y
.
Cl
o
se
o
b
se
rv
a
t
i
o
n
o
f
e
n
e
r
g
y
d
a
ta
se
ts
fr
o
m
m
a
n
y
P
V
p
l
a
n
ts
r
e
v
e
a
le
d
m
a
n
y
i
n
tr
i
n
si
c
e
n
v
i
ro
n
m
e
n
ta
l
v
a
r
ia
b
les
t
h
a
t
a
re
h
i
g
h
l
y
a
s
y
n
c
h
r
o
n
o
u
s
.
M
a
n
y
fo
re
c
a
st
i
n
g
m
o
d
e
ls
lea
r
n
re
d
u
n
d
a
n
t
fe
a
t
u
re
s
w
h
ic
h
m
i
g
h
t
se
e
m
u
se
f
u
l
,
a
n
d
t
h
e
re
b
y
t
h
e
tes
t
p
e
r
fo
rm
a
n
c
e
is
o
v
e
rf
it
ti
n
g
.
T
o
a
d
d
re
ss
t
h
e
n
o
n
l
i
n
e
a
r
a
n
d
a
s
y
n
c
h
r
o
n
o
u
s
b
e
h
a
v
io
r
o
f
e
n
v
ir
o
n
m
e
n
t
a
l
v
a
ria
b
les
,
th
e
re
is
a
s
e
r
io
u
s
re
q
u
irem
e
n
t
f
o
r
i
n
te
ll
i
g
e
n
t
fe
a
t
u
re
se
lec
ti
o
n
a
l
g
o
r
it
h
m
s
g
u
i
d
e
d
b
y
b
o
th
c
o
rr
e
la
ti
o
n
a
n
d
tem
p
o
ra
l
a
li
g
n
m
e
n
t
m
e
tr
ics
.
T
h
is
re
se
a
rc
h
p
r
o
p
o
se
s
a
n
o
v
e
l
a
d
a
p
ti
v
e
d
y
n
a
m
ic
t
im
e
wa
r
p
i
n
g
(A
-
D
W
T)
fe
a
t
u
re
se
lec
ti
o
n
w
i
t
h
a
n
a
d
a
p
ti
v
e
f
o
r
g
e
t
g
a
te
(
AFG
-
B
i
LS
T
M
)
t
o
a
d
d
r
e
ss
t
h
e
a
s
y
n
c
h
r
o
n
o
u
s
i
ss
u
e
s
.
E
x
p
e
r
ime
n
t
s
we
re
c
o
n
d
u
c
te
d
w
it
h
v
a
rie
d
e
n
v
ir
o
n
m
e
n
t
a
l
a
s
y
n
c
h
r
o
n
o
u
s
fe
a
t
u
re
s
,
a
n
d
t
h
e
re
s
u
lt
s
o
f
t
h
e
p
r
o
p
o
se
d
m
o
d
e
l
we
re
c
o
m
p
a
r
e
d
wi
t
h
t
ra
d
i
ti
o
n
a
l
B
iL
S
T
M
a
n
d
s
tac
k
e
d
B
iL
S
TM
m
o
d
e
ls
.
I
n
a
l
l
th
e
e
x
p
e
r
ime
n
ts
,
t
h
e
p
r
o
p
o
se
d
m
o
d
e
l
s
h
o
we
d
d
e
c
re
a
s
e
d
e
r
r
o
r
b
y
(
9
4
.
9
%
)
o
n
Da
tas
e
t
-
1
(
0
.
0
6
9
,
0
.
0
0
3
5
)
,
b
y
(
9
4
.
6
%
)
o
n
Da
t
a
se
t
-
2
(0
.
0
7
8
,
0
.
0
0
4
2
),
a
n
d
b
y
(9
0
.
9
%
)
o
n
Da
tas
e
t
-
3
(
0
.
0
6
9
,
0
.
0
0
6
1
)
w
h
e
n
c
o
m
p
a
re
d
t
o
sta
c
k
e
d
Bi
LS
T
M
.
T
h
e
M
S
E
l
o
ss
o
f
t
h
e
p
r
o
p
o
se
d
m
e
t
h
o
d
wa
s
o
b
se
r
v
e
d
t
o
b
e
b
e
tw
e
e
n
(
0
.
3
%
)
a
n
d
(
1
1
%
)
o
n
f
o
u
r
d
a
t
a
se
t
s.
K
ey
w
o
r
d
s
:
B
iLST
M
Dee
p
lear
n
in
g
Dy
n
am
ic
tim
e
war
p
in
g
Featu
r
e
s
elec
tio
n
PV so
lar
en
er
g
y
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Ar
e
Sam
b
asiv
a
R
ao
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
GI
T
AM
Sch
o
o
l o
f
C
o
m
p
u
ter
Scien
ce
an
d
E
n
g
in
ee
r
in
g
(
GSC
SE)
,
GI
T
AM
Dee
m
ed
to
b
e
Un
iv
er
s
ity
Hy
d
er
ab
ad
,
I
n
d
ia
E
m
ail: sar
e2
@
g
itam
.
in
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
g
lo
b
al
s
h
if
t
to
war
d
s
th
e
u
s
e
o
f
s
u
s
tain
ab
le
en
er
g
y
s
o
u
r
ce
s
h
as
r
em
ar
k
ab
l
y
ac
ce
l
er
ated
th
e
d
ev
elo
p
m
e
n
t
a
n
d
d
ep
lo
y
m
en
t
o
f
p
h
o
to
v
o
ltaic
(
PV)
p
lan
ts
.
T
h
e
co
n
tin
u
al
d
ep
letio
n
o
f
f
o
s
s
il
f
u
el
r
eser
v
es
a
n
d
in
ten
s
e
en
v
ir
o
n
m
en
tal
co
n
ce
r
n
s
,
s
o
lar
en
er
g
y
to
d
ay
h
as
em
er
g
ed
as
a
p
r
o
m
is
in
g
alter
n
ativ
e
f
o
r
clea
n
p
o
we
r
g
en
er
atio
n
.
Mo
d
er
n
PV
s
y
s
tem
s
to
d
ay
en
a
b
le
b
o
t
h
r
esid
en
tial
an
d
co
m
m
e
r
cial
co
n
s
u
m
e
r
s
to
g
en
er
ate
t
h
eir
o
wn
elec
tr
icity
,
tr
an
s
f
o
r
m
in
g
tr
ad
itio
n
al
co
n
s
u
m
e
r
s
in
to
p
r
o
s
u
m
er
s
.
T
h
is
d
y
n
am
ic
d
r
i
f
t
h
as
n
ec
ess
itated
r
ed
ir
ec
tio
n
o
f
c
o
n
v
e
n
tio
n
al
PV
s
o
lar
en
er
g
y
s
y
s
tem
s
to
war
d
s
th
e
u
s
e
o
f
m
o
r
e
in
te
llig
en
t
an
d
s
m
ar
t
ap
p
r
o
ac
h
es f
o
r
e
n
er
g
y
g
en
e
r
at
io
n
an
d
m
ain
ten
an
ce
o
f
PV sy
s
tem
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
Po
w
E
lec
&
Dr
i Sy
s
t
I
SS
N:
2088
-
8
6
9
4
A
d
a
p
tive
fo
r
g
et
-
g
a
ted
B
iLS
T
M e
n
h
a
n
ce
d
b
y
DTW b
a
s
ed
fe
a
tu
r
e
s
elec
tio
n
in
s
o
la
r
…
(
A
r
e
S
a
mb
a
s
iva
R
a
o
)
2087
A
cr
itical
p
r
o
b
lem
with
s
o
lar
p
o
wer
g
en
er
atio
n
s
y
s
tem
s
is
its
wo
r
k
in
g
u
n
d
er
v
a
r
ied
an
d
in
ten
s
e
atm
o
s
p
h
er
ic
co
n
d
itio
n
s
.
T
h
e
o
u
tp
u
t
o
f
PV
s
y
s
tem
s
ch
an
g
e
s
co
n
s
tan
tly
d
u
e
to
d
y
n
a
m
ic
atm
o
s
p
h
er
ic
en
titi
es,
m
ain
ly
b
ec
au
s
e
o
f
asy
n
c
h
r
o
n
o
u
s
en
v
ir
o
n
m
en
tal
f
ea
tu
r
es
lik
e
ir
r
ad
ian
ce
,
clo
u
d
co
v
e
r
,
tem
p
er
atu
r
e
,
an
d
atm
o
s
p
h
er
ic
h
u
m
id
ity
.
T
h
e
n
atu
r
al
d
y
n
am
ic
v
a
r
iab
ilit
y
th
r
o
u
g
h
o
u
t
th
e
en
v
i
r
o
n
m
e
n
tal
c
y
cles
cr
ea
tes
h
u
g
e
d
if
f
icu
lties
f
o
r
m
ain
tain
in
g
s
tab
le
an
d
r
eliab
le
PV
s
y
s
tem
s
f
o
r
ef
f
icien
t
en
er
g
y
g
en
e
r
atio
n
.
T
h
e
f
lu
ctu
atio
n
s
in
PV
p
lan
t
en
er
g
y
ca
n
tr
ig
g
er
cr
itical
is
s
u
es
lik
e
ex
tr
em
e
v
o
ltag
e
d
e
v
iatio
n
s
,
f
r
e
q
u
en
c
y
ir
r
eg
u
lar
ities
,
an
d
p
o
wer
co
n
v
er
s
io
n
c
h
allen
g
es,
all
o
f
wh
ich
co
m
p
r
o
m
is
e
th
e
in
teg
r
ity
o
f
PV n
etwo
r
k
s
.
Pre
cise
an
d
tim
ely
f
o
r
ec
asti
n
g
o
f
s
o
lar
p
o
wer
g
en
er
atio
n
h
as
th
er
ef
o
r
e
b
ec
o
m
e
ess
en
tial
f
o
r
ef
f
ec
tiv
e
p
o
wer
m
a
n
ag
em
e
n
t
f
r
o
m
PV
p
lan
ts
.
On
th
e
wh
o
le,
to
o
n
e
h
an
d
,
th
o
u
g
h
m
u
ltip
le
f
ac
to
r
s
in
f
lu
en
ce
PV
p
lan
t
s
o
lar
en
er
g
y
g
en
er
atio
n
,
th
e
p
r
im
ar
y
f
ac
to
r
c
o
u
ld
b
e
u
n
d
er
s
ta
n
d
in
g
th
e
d
ee
p
-
r
o
o
te
d
r
elatio
n
s
h
ip
s
an
d
d
y
n
am
ic
n
atu
r
e
o
f
atm
o
s
p
h
er
ic
f
ea
tu
r
e
s
s
u
ch
as
am
b
ien
t
tem
p
er
atu
r
e,
clo
u
d
co
v
e
r
,
atm
o
s
p
h
er
ic
d
ew
,
an
d
tech
n
ical
co
n
d
itio
n
s
o
f
PV
ce
lls
an
d
in
v
er
ter
s
,
wh
ich
m
ajo
r
ly
im
p
ac
t
th
e
o
v
er
all
s
y
s
tem
p
r
ed
ictio
n
p
er
f
o
r
m
an
ce
.
T
h
e
m
u
ltifa
ce
ted
n
atu
r
e
o
f
th
ese
a
tm
o
s
p
h
er
ic
f
ea
tu
r
es
n
ec
ess
itates
th
e
n
ee
d
a
n
d
u
s
ag
e
o
f
s
o
p
h
is
ticated
p
r
ed
ictio
n
m
o
d
els ca
p
ab
le
o
f
ca
p
tu
r
in
g
d
ee
p
r
elatio
n
s
h
ip
s
b
etwe
en
e
n
v
ir
o
n
m
en
tal
en
titi
es a
n
d
s
o
lar
p
o
wer
o
u
tp
u
t [
1
]
.
Dee
p
lear
n
in
g
-
b
ased
b
id
i
r
ec
tio
n
al
lo
n
g
-
s
h
o
r
t
ter
m
m
em
o
r
y
(
B
iLST
M)
m
o
d
els
r
en
d
er
ad
v
an
ce
d
m
eth
o
d
o
l
o
g
y
f
o
r
ca
p
tu
r
in
g
b
o
th
d
ee
p
p
ast
an
d
f
u
t
u
r
e
d
ep
e
n
d
en
cies
in
tim
e
s
er
ies
d
ata,
m
ak
in
g
t
h
em
h
ig
h
l
y
ef
f
ec
tiv
e
an
d
p
r
ef
er
a
b
le
f
o
r
s
o
lar
p
o
wer
f
o
r
ec
asti
n
g
.
T
h
ese
m
o
d
els
ar
e
g
o
o
d
at
ca
p
t
u
r
in
g
th
e
tem
p
o
r
al
d
ata
d
ep
en
d
e
n
cies a
n
d
r
etain
in
g
in
f
o
r
m
atio
n
o
v
er
ex
ten
d
e
d
lo
n
g
-
tim
e
s
eq
u
en
ce
s
,
th
u
s
m
ak
in
g
th
em
well
s
u
ited
f
o
r
m
o
d
ellin
g
th
e
tim
e
-
d
ep
e
n
d
en
t
n
atu
r
e
o
f
s
o
lar
p
o
wer
p
r
ed
ict
io
n
.
Mu
ch
o
f
th
e
r
esear
ch
o
n
B
iLST
Ms
s
h
o
wed
b
etter
tr
ain
in
g
ac
cu
r
ac
y
b
u
t
d
ep
leted
test
ac
cu
r
ac
y
.
So
m
e
o
f
th
e
ex
p
lo
r
ed
r
ea
s
o
n
s
f
o
r
d
ep
leted
test
ac
cu
r
ac
y
m
ig
h
t
b
e,
t
h
at
tem
p
o
r
al
m
is
alig
n
m
en
ts
m
ay
n
o
t
h
av
e
b
ee
n
s
t
u
d
ied
well
an
d
th
at
th
e
m
o
d
el
s
lear
n
ed
to
o
m
an
y
r
ed
u
n
d
an
t f
ea
tu
r
es u
n
k
n
o
win
g
ly
.
T
em
p
o
r
al
m
is
alig
n
m
e
n
t
is
esp
ec
ially
cr
itical
in
PV
f
o
r
ec
asti
n
g
b
ec
a
u
s
e
s
o
lar
p
o
wer
f
o
llo
w
s
a
tig
h
tly
d
eter
m
in
is
tic
d
iu
r
n
al
c
y
cle,
s
o
ev
en
s
m
all
lag
s
b
etwe
en
m
ete
o
r
o
lo
g
ical
in
p
u
ts
an
d
th
e
tar
g
e
t
s
ig
n
al
ca
n
d
is
to
r
t
th
e
lear
n
ed
m
a
p
p
in
g
.
Par
ticu
lar
ly
in
PV
d
atasets
,
th
er
e
ar
e
m
an
y
r
e
d
u
n
d
an
t
f
ea
tu
r
es
lik
e
h
u
m
id
ity
,
p
o
in
t
h
u
m
id
ity
,
clo
u
d
s
p
r
ea
d
,
clo
u
d
co
v
er
,
ir
r
ad
ian
ce
,
d
if
f
u
s
e
ir
r
ad
ian
ce
,
h
o
r
iz
o
n
tal
ir
r
a
d
ian
c
e
,
an
d
ev
e
n
ce
r
tain
tem
p
er
atu
r
e
m
ea
s
u
r
e
m
en
ts
,
wh
ich
ex
h
ib
it
s
tr
o
n
g
m
u
tu
al
co
r
r
elatio
n
s
an
d
ca
n
in
tr
o
d
u
ce
r
ed
u
n
d
an
c
y
in
to
p
r
ed
ictiv
e
m
o
d
els
if
n
o
t
p
r
o
p
e
r
ly
f
ilter
ed
[
2
]
.
T
h
ese
ex
p
lo
r
e
d
r
ea
s
o
n
s
ar
e
th
e
f
o
cu
s
ed
,
m
o
t
iv
ated
o
b
jectiv
es
in
o
u
r
p
r
o
p
o
s
ed
wo
r
k
.
T
h
is
r
esear
ch
p
r
o
p
o
s
es
a
m
o
d
el
f
o
r
PV
p
o
wer
p
r
ed
ictio
n
b
y
s
tack
in
g
m
u
ltip
le
B
iLST
Ms,
with
a
n
o
v
el
f
ea
tu
r
e
s
elec
tio
n
o
f
en
v
ir
o
n
m
en
tal
v
ar
iab
les.
T
h
e
B
iLST
Ms
in
th
is
s
tu
d
y
ar
e
p
r
o
p
o
s
ed
to
ca
p
tu
r
e
t
h
e
p
ast
an
d
f
u
t
u
r
e
tim
e
s
er
ies
d
ata
d
ep
en
d
e
n
cies
s
o
as
to
r
ai
s
e
th
e
p
r
ed
ictio
n
ac
cu
r
ac
y
.
A
f
ea
tu
r
e
s
elec
tio
n
ap
p
r
o
ac
h
,
a
d
ap
tiv
e
d
y
n
a
m
ic
ti
m
e
war
p
in
g
(
A
-
DT
W
)
with
v
a
r
ian
ce
en
tr
o
p
y
weig
h
t,
is
p
r
o
p
o
s
ed
to
a
d
d
r
ess
th
e
tem
p
o
r
al
m
is
alig
n
m
e
n
ts
an
d
t
o
ac
h
iev
e
b
etter
co
r
r
elatio
n
s
o
f
th
e
tim
e
s
er
ies
f
ea
tu
r
es
with
th
e
tar
g
et.
I
n
s
tead
o
f
ass
ig
n
in
g
id
en
tical
weig
h
ts
to
all
f
ea
tu
r
es,
A
-
DT
W
d
y
n
am
ically
ad
j
u
s
ts
ea
ch
f
ea
tu
r
e'
s
co
n
tr
ib
u
tio
n
ac
co
r
d
in
g
to
its
lo
ca
l
s
tatis
ti
ca
l
v
ar
iab
ilit
y
.
T
h
is
en
ab
les
t
h
e
alg
o
r
ith
m
t
o
em
p
h
asize
f
ea
tu
r
es
with
s
tr
o
n
g
tem
p
o
r
al
r
elev
a
n
ce
,
th
er
eb
y
p
r
o
d
u
cin
g
a
m
o
r
e
r
eliab
le
war
p
in
g
p
ath
an
d
im
p
r
o
v
in
g
t
h
e
ac
cu
r
ac
y
an
d
s
tab
ilit
y
o
f
f
ea
tu
r
e
s
elec
tio
n
f
o
r
s
o
lar
p
o
wer
f
o
r
ec
asti
n
g
.
R
ed
u
n
d
an
t
f
ea
tu
r
es
th
at
ar
e
ca
p
tu
r
ed
an
d
s
elec
ted
in
th
e
A
-
DT
W
m
o
d
u
le
ar
e
f
ilter
e
d
u
s
in
g
an
ad
ap
tiv
e
f
o
r
g
et
g
ate
(
AFG)
w
ith
lear
n
ab
le
weig
h
ts
,
wh
ich
i
s
in
teg
r
ated
in
to
th
e
p
r
o
p
o
s
ed
s
tack
ed
B
iLST
M
[
3
]
,
[
4
]
.
Un
lik
e
co
n
v
en
tio
n
al
f
o
r
g
et
g
ates,
th
e
p
r
o
p
o
s
ed
AF
G
ad
ap
tiv
ely
r
eg
u
lates
m
em
o
r
y
r
eten
tio
n
b
ased
o
n
tem
p
o
r
al
f
e
atu
r
e
r
elev
an
ce
,
th
er
eb
y
p
r
eser
v
in
g
in
f
o
r
m
ativ
e
f
ea
tu
r
es
wh
ile
s
u
p
p
r
ess
in
g
r
e
d
u
n
d
a
n
t
o
n
es.
T
h
e
n
o
v
elty
o
f
th
e
ap
p
r
o
ac
h
lie
s
in
co
m
b
in
in
g
A
-
DT
W
with
AFG
f
o
r
f
ea
tu
r
e
s
elec
tio
n
an
d
r
e
d
u
n
d
an
c
y
elim
in
atio
n
,
wh
ich
is
n
o
t c
o
m
m
o
n
ly
a
d
d
r
ess
ed
in
p
r
io
r
wo
r
k
s
.
On
e
o
f
th
e
m
ajo
r
g
ap
s
in
th
e
m
o
d
elin
g
o
f
d
ee
p
lear
n
in
g
al
g
o
r
ith
m
s
is
th
at
m
an
y
o
f
th
e
f
o
r
ec
asti
n
g
m
o
d
els
ass
u
m
e
in
s
tan
tan
eo
u
s
ef
f
ec
ts
o
f
en
v
ir
o
n
m
en
tal
v
a
r
iab
les
in
p
o
wer
p
r
e
d
ictio
n
an
d
f
ail
to
m
o
d
el
asy
n
ch
r
o
n
o
u
s
an
d
tem
p
o
r
ally
m
is
alig
n
ed
r
elatio
n
s
h
ip
s
b
etwe
en
en
v
ir
o
n
m
en
tal
f
ea
tu
r
es
an
d
s
o
lar
p
o
wer
o
u
tp
u
t
[
5
]
,
[
6
]
.
R
ea
l
-
tim
e
m
o
d
els
lack
tem
p
o
r
al
alig
n
m
en
t
f
r
am
ewo
r
k
s
th
at
d
y
n
am
ically
ad
ju
s
t
th
e
lag
s
f
o
r
d
elay
ed
im
p
ac
ts
[
7
]
.
B
o
th
f
ea
tu
r
e
co
r
r
elatio
n
s
an
d
tem
p
o
r
al
d
ep
e
n
d
en
cies
ar
e
d
if
f
i
cu
lt
f
o
r
tr
ad
itio
n
al
f
o
r
ec
asti
n
g
m
o
d
els
to
ad
eq
u
ately
ca
p
tu
r
e
[
8
]
,
[
9
]
.
Failin
g
to
f
o
cu
s
o
n
th
e
s
tu
d
y
o
f
asy
n
ch
r
o
n
o
u
s
an
d
im
p
er
f
ec
tly
co
o
r
d
in
ate
d
f
ea
tu
r
es c
r
itically
af
f
ec
ts
th
e
f
o
r
ec
as
tin
g
ac
cu
r
ac
y
an
d
m
o
d
el
i
n
ter
p
r
etab
ilit
y
.
T
h
is
s
tu
d
y
co
n
tr
ib
u
tes
to
th
e
f
ield
b
y
ad
d
r
ess
in
g
th
ese
g
a
p
s
th
r
o
u
g
h
t
h
e
in
tr
o
d
u
ctio
n
o
f
a
n
o
v
el
lear
n
ab
le
d
y
n
am
ic
tim
e
war
p
i
n
g
(
DT
W
)
m
o
d
u
le
th
at
d
is
co
v
er
s
an
d
a
d
ap
ts
o
p
tim
al
tem
p
o
r
al
alig
n
m
en
t
p
ath
s
b
etwe
en
en
v
ir
o
n
m
en
tal
f
ea
tu
r
es
an
d
p
o
wer
o
u
t
p
u
t,
ef
f
ec
ti
v
ely
ca
p
tu
r
in
g
th
e
v
ar
y
i
n
g
te
m
p
o
r
al
s
ca
les
an
d
in
h
er
en
t
d
elay
s
ac
r
o
s
s
d
if
f
er
e
n
t
m
eteo
r
o
lo
g
ical
v
ar
iab
les.
D
T
W
is
a
p
o
wer
f
u
l
alg
o
r
ith
m
ic
tech
n
iq
u
e
th
at
g
o
es
b
ey
o
n
d
tr
ad
itio
n
al
c
o
r
r
elatio
n
m
eth
o
d
s
.
I
n
lin
e
with
t
h
e
id
e
n
tifie
d
g
a
p
,
th
is
r
esear
ch
p
r
o
p
o
s
es
th
e
f
o
llo
win
g
o
b
jectiv
es:
d
e
v
elo
p
a
s
tack
ed
B
iLST
M
to
u
n
d
e
r
s
tan
d
a
n
d
ca
p
tu
r
e
t
h
e
p
ast
an
d
f
u
tu
r
e
tim
e
s
er
ies
d
ata
d
ep
e
n
d
en
cies
f
r
o
m
PV
d
ata;
d
ev
elo
p
a
f
ea
tu
r
e
co
r
r
elatio
n
m
eth
o
d
to
ca
p
tu
r
e
th
e
co
r
r
elate
d
e
n
v
ir
o
n
m
en
tal
f
ea
tu
r
es
with
wh
i
ch
th
e
B
iLST
M
is
tr
ain
ed
;
d
ev
elo
p
a
r
o
b
u
s
t
ap
p
r
o
ac
h
to
f
ilter
th
e
r
ed
u
n
d
a
n
t
en
v
ir
o
n
m
en
tal
f
ea
tu
r
es
o
f
th
e
PV
d
ata.
T
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
6
9
4
I
n
t J Po
w
E
lec
&
Dr
i Sy
s
t
,
Vo
l.
1
7
,
No
.
3
,
Sep
tem
b
er
20
2
6
:
2086
-
2
1
0
0
2088
s
ig
n
if
ican
ce
o
f
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
,
with
its
clea
r
ly
d
ef
in
ed
o
b
jectiv
es
,
lies
in
its
co
n
tr
ib
u
tio
n
to
th
e
r
esear
ch
co
m
m
u
n
ity
b
y
o
f
f
e
r
in
g
an
ef
f
icien
t
a
p
p
r
o
ac
h
f
o
r
t
em
p
o
r
al
f
ea
tu
r
e
c
o
r
r
elatio
n
an
d
r
ed
u
n
d
a
n
t
f
ea
tu
r
e
elim
in
atio
n
in
h
ig
h
-
d
im
en
s
io
n
al
en
v
ir
o
n
m
e
n
tal
d
atasets
.
Pra
ctica
lly
,
th
e
f
r
am
ewo
r
k
ca
n
e
n
h
an
ce
PV
s
y
s
tem
m
o
n
ito
r
in
g
an
d
p
er
f
o
r
m
a
n
ce
an
aly
s
is
b
y
im
p
r
o
v
in
g
d
ata
q
u
ality
an
d
s
u
p
p
o
r
tin
g
m
o
r
e
r
eliab
le
s
o
lar
en
er
g
y
m
an
ag
em
en
t sy
s
tem
s
.
T
in
a
et
a
l.
[
1
0
]
an
d
L
ai
et
a
l
.
[
1
1
]
,
in
th
eir
wo
r
k
,
p
r
esen
ted
a
s
y
s
tem
atic
ass
e
s
s
m
en
t
o
f
d
if
f
er
e
n
t
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
f
o
r
f
o
r
ec
asti
n
g
s
o
lar
p
o
wer
.
T
h
eir
wo
r
k
em
p
h
asized
th
e
p
er
f
o
r
m
a
n
ce
an
d
ef
f
ec
tiv
en
ess
o
f
en
s
em
b
le
h
y
b
r
id
m
o
d
els
in
id
en
tify
in
g
r
el
ev
an
t
p
atter
n
s
in
PV
d
ata.
T
ai
et
a
l
.
[
1
2
]
in
th
ei
r
s
tu
d
ies
u
s
ed
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
s
(
C
NNs)
alo
n
g
with
b
id
ir
ec
tio
n
al
L
STM
n
et
wo
r
k
s
f
o
r
ca
p
tu
r
in
g
tim
e
-
s
er
ies
tem
p
o
r
al
r
elatio
n
s
h
ip
s
o
f
s
o
lar
e
n
er
g
y
an
d
ir
r
ad
ian
ce
.
T
h
eir
m
eth
o
d
s
s
h
o
we
d
a
m
a
r
g
in
al
(
3
0
%)
in
cr
ea
s
e
in
f
o
r
ec
asti
n
g
ac
cu
r
a
cy
co
m
p
a
r
ed
to
t
r
ad
itio
n
al
alg
o
r
ith
m
s
wh
en
test
ed
o
n
d
if
f
e
r
e
n
t d
atasets
.
Pian
tad
o
s
i
et
a
l.
[
1
4
]
a
n
d
Kim
et
a
l.
[
1
4
]
,
in
th
eir
s
tu
d
ies
,
p
r
o
p
o
s
ed
a
tr
an
s
f
o
r
m
er
-
b
ased
s
o
lar
p
o
wer
p
r
ed
ictio
n
m
o
d
el.
T
h
is
m
o
d
el
u
s
ed
a
m
u
lti
-
h
ea
d
atten
tio
n
m
ec
h
an
is
m
to
f
in
d
te
m
p
o
r
al
p
at
ter
n
s
an
d
tr
en
d
s
in
s
o
lar
d
ata.
T
h
e
wo
r
k
o
f
Sto
ea
n
et
a
l.
[
1
5
]
p
r
o
p
o
s
ed
v
ar
io
u
s
h
y
p
er
p
ar
am
eter
tu
n
in
g
ap
p
r
o
a
ch
es
f
o
r
o
p
tim
izin
g
th
e
p
er
f
o
r
m
an
ce
o
f
d
ee
p
lear
n
in
g
f
o
r
ec
asti
n
g
m
o
d
els.
T
h
eir
ex
p
er
im
en
tal
s
tu
d
ies s
h
o
wed
a
co
m
p
ar
ativ
e
s
tu
d
y
wh
er
e
m
etah
e
u
r
is
tic
tech
n
iq
u
es
o
f
ten
o
u
tp
er
f
o
r
m
e
d
co
n
v
en
tio
n
al
ev
o
l
u
tio
n
ar
y
o
p
tim
izatio
n
alg
o
r
ith
m
s
,
wit
h
R
2
o
f
(
0
.
6
0
4
)
an
d
MSE
o
f
(
0
.
0
1
4
)
.
Su
laim
an
an
d
Mu
s
taf
f
a
[
1
6
]
in
tr
o
d
u
ce
d
a
n
ef
f
icien
t
d
i
f
f
er
en
tial
ap
p
r
o
ac
h
f
o
r
s
o
lar
p
o
wer
f
o
r
ec
asti
n
g
with
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
with
o
p
tim
ized
p
ar
a
m
eter
s
.
T
h
eir
wo
r
k
p
r
o
p
o
s
ed
a
m
u
tatio
n
alg
o
r
ith
m
th
at
ad
a
p
tiv
ely
ch
an
g
es th
e
s
ea
r
ch
p
ar
a
m
eter
s
ac
co
r
d
in
g
to
th
e
o
p
tim
izatio
n
p
r
o
g
r
ess
.
T
h
e
m
ax
im
al
er
r
o
r
g
i
v
en
b
y
th
eir
ap
p
r
o
ac
h
s
tan
d
s
at
(
7
.
3
0
3
%).
Fin
e
-
tu
n
in
g
d
ee
p
lea
r
n
in
g
m
o
d
els
f
o
r
tim
e
s
er
ies
f
o
r
ec
asti
n
g
was
p
r
o
p
o
s
ed
in
[
1
7
]
.
T
h
eir
wo
r
k
in
clu
d
e
d
an
it
er
ativ
e
,
f
in
e
-
tu
n
in
g
-
b
ased
ap
p
r
o
ac
h
f
o
r
p
a
r
am
eter
o
p
tim
izatio
n
to
av
o
id
lo
ca
l
m
in
im
a
wh
ile
tr
ig
g
er
in
g
a
g
lo
b
al
ex
p
lo
r
atio
n
.
T
h
e
ex
p
er
im
e
n
tal
s
tu
d
ies
s
h
o
wed
9
2
% p
r
ec
is
io
n
o
f
th
e
m
o
d
el.
L
ig
h
tweig
h
t
L
STM
m
o
d
els
wer
e
p
r
o
p
o
s
ed
f
o
r
f
o
r
ec
asti
n
g
s
o
lar
p
o
wer
,
wh
ich
h
as
r
aised
th
e
p
r
ed
ictio
n
ac
c
u
r
ac
y
[
1
8
]
,
[
1
9
]
.
T
o
q
u
an
tize
th
e
lig
h
tweig
h
t
m
o
d
el,
th
e
wo
r
k
s
u
s
ed
p
r
u
n
in
g
an
d
q
u
a
n
tizatio
n
m
eth
o
d
s
.
Dao
et
a
l
.
[
2
0
]
a
n
d
C
h
an
g
et
a
l.
[
2
1
]
in
tr
o
d
u
ce
d
lig
h
tweig
h
t
em
b
e
d
d
ed
.
A
n
o
p
tim
ized
ex
tr
em
e
lear
n
in
g
tech
n
iq
u
e
was
f
o
cu
s
ed
o
n
b
y
B
eh
er
a
et
a
l
.
[
2
2
]
,
wh
ich
d
y
n
am
ically
m
o
d
if
ie
d
m
o
d
el
c
o
m
p
lex
ity
b
ased
o
n
th
e
p
ar
am
eter
s
.
T
h
e
ir
m
o
d
els
u
n
d
er
v
ar
i
o
u
s
o
p
er
atio
n
al
s
ettin
g
s
in
cu
r
r
ed
lo
w
co
m
p
u
tatio
n
al
co
s
t,
with
a
MA
PE
o
f
2
.
9
4
%
at
1
5
m
in
o
f
tim
e
in
ter
v
al.
Niu
et
a
l.
[
2
3
]
,
Qu
et
a
l
.
[
2
4
]
,
a
n
d
Hu
an
g
et
a
l
.
[
2
5
]
u
s
ed
a
h
ier
ar
ch
ical
atten
tio
n
m
o
d
el
to
ca
p
tu
r
e
h
is
to
r
ical
tr
en
d
s
a
t
v
ar
io
u
s
tim
e
lev
els.
T
h
eir
ap
p
r
o
ac
h
es
s
h
o
wed
g
r
ea
ter
ac
cu
r
ac
y
with
ex
tr
em
e
m
eteo
r
o
lo
g
ical
v
o
latile
d
ata.
T
h
e
wo
r
k
o
f
Gan
et
a
l
.
[
2
6
]
p
r
esen
ted
a
u
n
i
q
u
e
atten
tio
n
ap
p
r
o
ac
h
with
a
m
u
lti
-
h
ea
d
ed
m
ec
h
an
is
m
f
o
r
s
o
lar
en
er
g
y
p
r
ed
ictio
n
with
h
is
to
r
ical
d
ata.
T
h
eir
ap
p
r
o
ac
h
es
d
em
o
n
s
tr
ated
h
ig
h
p
er
f
o
r
m
an
ce
in
co
m
p
ar
is
o
n
to
tr
ad
itio
n
al
L
STM
,
r
ed
u
cin
g
MA
E
b
y
3
8
.
5
3
%
,
an
d
in
m
u
ltis
tep
f
o
r
ec
asti
n
g
,
a
r
ed
u
ce
d
MA
E
o
f
5
4
.
2
6
%
was
o
b
s
er
v
ed
.
Gen
s
ler
et
a
l
.
[
2
7
]
in
th
eir
wo
r
k
u
s
e
a
h
y
b
r
i
d
L
STM
n
etwo
r
k
f
o
r
s
o
lar
en
e
r
g
y
p
r
e
d
ictio
n
in
co
m
p
ar
is
o
n
to
R
NN
m
o
d
els
an
d
s
h
o
wed
b
etter
p
er
f
o
r
m
an
ce
with
L
STM
s
in
u
n
d
er
s
tan
d
in
g
th
e
s
o
lar
ir
r
ad
iatio
n
f
ea
t
u
r
e.
T
h
eir
a
p
p
r
o
ac
h
r
e
d
u
ce
d
ab
s
o
lu
te
er
r
o
r
,
with
a
b
etter
p
e
r
f
o
r
m
an
ce
o
f
8
7
%.
Yu
et
a
l
.
[
2
8
]
an
d
Awa
is
et
a
l
.
[
2
9
]
i
n
th
eir
wo
r
k
s
u
s
ed
en
h
a
n
ce
d
L
STM
with
atten
tio
n
lay
er
s
,
wh
ich
h
av
e
th
e
ca
p
ab
ilit
y
o
f
u
n
d
er
s
tan
d
i
n
g
th
e
p
ast
an
d
p
r
esen
t
ir
r
ad
ian
ce
f
ea
tu
r
es
with
ad
ap
tiv
e
weig
h
ts
ass
ig
n
ed
.
T
h
ese
ad
ap
tiv
e
weig
h
ts
ar
e
th
en
u
s
ed
to
s
tu
d
y
th
e
f
ea
t
u
r
e
r
ele
v
an
ce
.
T
h
ey
co
n
d
u
cted
e
x
p
er
im
en
ts
u
n
d
e
r
ex
tr
em
e
wea
th
er
p
atter
n
s
an
d
h
ig
h
lig
h
ted
th
ei
r
m
o
d
el'
s
h
ig
h
p
er
f
o
r
m
an
ce
,
r
aised
b
y
m
o
r
e
th
an
1
5
%
wh
en
co
m
p
ar
e
d
to
tr
ad
itio
n
al
L
STM
m
eth
o
d
s
.
Z
h
o
n
g
et
a
l
.
[
3
0
]
p
r
esen
ted
a
B
iLST
M
m
o
d
el
with
te
m
p
o
r
al
atten
tio
n
f
o
r
m
u
ltis
tep
lag
s
tu
d
y
in
f
o
r
ec
as
tin
g
ir
r
ad
ian
ce
.
C
h
en
et
a
l
.
[
3
1
]
m
o
d
elled
a
s
liced
atten
tio
n
B
iLST
M
th
at
u
s
e
s
s
atellite
d
ata
f
o
r
r
eg
io
n
al
s
o
lar
en
er
g
y
f
o
r
ec
asti
n
g
.
T
h
eir
m
o
d
el
ex
h
i
b
ited
h
i
g
h
p
r
ed
ictio
n
a
cc
u
r
ac
y
with
a
lo
w
MA
E
o
f
1
.
1
5
.
T
h
e
wo
r
k
s
tu
d
i
ed
b
y
Yu
et
a
l
.
[
3
2
]
m
o
d
elled
a
B
iLST
M
with
d
o
u
b
le
-
d
ec
o
m
p
o
s
ed
la
y
er
s
f
o
r
s
o
lar
p
o
wer
p
r
ed
ictio
n
.
E
s
p
e
cially
f
o
r
lo
n
g
er
p
r
ed
ictio
n
h
o
r
izo
n
s
,
t
h
ese
s
tack
ed
lay
er
s
p
r
o
v
ed
b
etter
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
p
r
e
d
ictio
n
s
,
with
r
ed
u
ce
d
p
r
ed
ictio
n
e
r
r
o
r
b
y
5
6
.
0
9
% o
n
clo
u
d
y
d
ay
s
.
Stu
d
i
es
w
er
e
p
r
es
en
te
d
b
y
X
i
e
e
t
a
l
.
[
3
3
]
a
n
d
G
u
et
a
l
.
[
3
4
]
b
y
s
e
tti
n
g
u
p
a
m
u
lti
-
l
ay
er
ed
s
ta
c
k
e
d
B
iLST
M
wit
h
i
n
t
e
r
m
ed
iat
e
r
es
id
u
al
c
o
n
n
ec
t
io
n
s
.
T
h
e
ir
e
x
p
er
im
e
n
t
al
d
e
m
o
n
s
tr
ati
o
n
s
s
h
o
w
e
d
t
h
a
t
t
h
e
p
r
o
p
o
s
e
d
m
o
d
el
a
d
d
r
ess
e
d
t
h
e
v
a
n
is
h
i
n
g
g
r
a
d
i
en
t
an
d
a
ch
ie
v
e
d
c
o
n
s
i
d
e
r
a
b
l
e
p
e
r
f
o
r
m
a
n
c
e.
L
i
u
et
a
l
.
[
3
5
]
i
n
th
ei
r
w
o
r
k
in
t
r
o
d
u
ce
d
a
n
o
v
el
f
ea
t
u
r
e
e
n
g
in
e
e
r
i
n
g
m
e
th
o
d
i
n
a
s
t
ac
k
e
d
B
iLST
M
m
o
d
el
,
w
it
h
a
n
ac
cu
r
ac
y
i
m
p
r
o
v
e
m
e
n
t o
f
1
2
.
6
%.
W
a
v
el
et
d
ec
o
m
p
o
s
i
ti
o
n
s
w
er
e
u
s
ed
t
o
b
r
ea
k
d
o
w
n
ti
m
e
s
e
r
ies
i
n
t
o
i
n
t
er
m
e
d
i
at
e
f
r
e
q
u
e
n
c
y
co
m
p
o
n
en
ts
f
o
r
ad
ap
ti
v
e
f
e
at
u
r
e
e
n
g
i
n
e
er
i
n
g
.
S
ar
m
as
et
a
l
.
[
3
6
]
a
n
d
G
ao
et
a
l
.
[
3
7
]
e
x
p
lo
r
e
d
an
d
p
r
es
en
te
d
a
t
r
an
s
f
e
r
lea
r
n
i
n
g
a
p
p
r
o
a
c
h
t
o
s
tu
d
y
h
i
g
h
l
y
v
a
r
i
an
t
s
o
l
ar
d
ata
.
T
h
ei
r
e
x
p
e
r
im
e
n
ta
l
s
t
u
d
ies
c
o
n
d
u
cte
d
o
n
v
ar
i
o
u
s
d
atas
ets
s
h
o
wc
ase
d
th
e
m
o
d
el'
s
c
a
p
a
b
il
ity
i
n
u
n
d
e
r
s
ta
n
d
i
n
g
co
m
p
le
x
p
at
te
r
n
s
i
n
li
m
it
e
d
d
a
ta
.
C
o
n
v
en
tio
n
al
f
ea
tu
r
e
s
elec
tio
n
ap
p
r
o
ac
h
es
s
u
ch
as
p
r
in
cip
al
co
m
p
o
n
en
t
an
aly
s
is
(
PC
A)
an
d
co
r
r
elatio
n
-
b
ased
m
eth
o
d
s
h
a
v
e
b
ee
n
ex
ten
s
iv
ely
u
s
ed
f
o
r
d
im
en
s
io
n
ality
r
ed
u
ctio
n
a
n
d
r
e
d
u
n
d
a
n
cy
a
n
aly
s
is
.
Nev
er
th
eless
,
PC
A
p
r
o
jects
th
e
d
ata
o
n
to
o
r
th
o
g
o
n
al
co
m
p
o
n
en
ts
b
y
m
a
x
im
izin
g
v
a
r
ian
ce
with
o
u
t
p
r
eser
v
i
n
g
tem
p
o
r
al
d
ep
en
d
en
cies,
wh
er
ea
s
co
r
r
elatio
n
co
ef
f
icien
ts
ev
alu
ate
o
n
l
y
lin
ea
r
r
ela
tio
n
s
h
ip
s
b
etwe
en
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
Po
w
E
lec
&
Dr
i Sy
s
t
I
SS
N:
2088
-
8
6
9
4
A
d
a
p
tive
fo
r
g
et
-
g
a
ted
B
iLS
T
M e
n
h
a
n
ce
d
b
y
DTW b
a
s
ed
fe
a
tu
r
e
s
elec
tio
n
in
s
o
la
r
…
(
A
r
e
S
a
mb
a
s
iva
R
a
o
)
2089
s
y
n
ch
r
o
n
ize
d
o
b
s
er
v
atio
n
s
an
d
f
ail
to
ca
p
tu
r
e
tim
e
-
lag
g
ed
s
im
ilar
ities
[
3
8
]
,
[
3
9
]
.
C
o
n
s
eq
u
en
tly
,
t
h
ese
m
eth
o
d
s
ar
e
less
ef
f
ec
tiv
e
f
o
r
s
o
lar
p
o
wer
f
o
r
ec
asti
n
g
,
wh
er
e
m
eteo
r
o
l
o
g
ical
v
a
r
iab
les
o
f
t
en
ex
h
i
b
it
n
o
n
lin
ea
r
an
d
tem
p
o
r
ally
s
h
if
ted
in
ter
ac
tio
n
s
,
m
o
tiv
atin
g
th
e
ad
o
p
tio
n
o
f
DT
W
in
th
is
r
esear
ch
f
o
r
t
em
p
o
r
al
alig
n
m
en
t
an
d
s
im
ilar
ity
-
b
ased
f
ea
tu
r
e
s
elec
tio
n
.
DT
W
h
as
em
er
g
ed
as
a
p
o
wer
f
u
l
m
eth
o
d
f
o
r
ad
d
r
ess
in
g
th
ese
lim
itatio
n
s
o
f
d
is
tan
ce
m
etr
ic
s
.
Sam
ar
a
et
a
l
.
[
4
0
]
in
t
r
o
d
u
c
ed
DT
W
f
o
r
tim
e
s
er
ies
an
aly
s
is
.
Kim
an
d
Ho
n
g
[
4
1
]
,
in
th
eir
s
tu
d
ies
,
in
clu
d
e
d
DT
W
f
o
r
co
m
p
ar
in
g
p
atter
n
s
in
tim
e
s
er
ie
s
h
is
to
r
ical
d
a
ta.
T
h
ey
u
s
ed
DT
W
with
in
th
eir
p
r
o
p
o
s
ed
k
-
Sh
a
p
e
clu
s
ter
in
g
alg
o
r
ith
m
an
d
ex
p
er
im
en
tally
s
h
o
we
d
8
9
%
m
o
d
el
ac
cu
r
ac
y
in
f
o
r
ec
asti
n
g
tim
e
s
er
ies.
2.
DATA AN
D
M
E
T
H
O
DO
L
O
G
Y
Ma
n
y
o
f
th
e
cu
r
r
en
t
s
o
lar
e
n
er
g
y
p
r
ed
ictio
n
s
y
s
tem
s
p
er
f
o
r
m
p
o
o
r
ly
o
n
v
ar
ia
b
le
d
at
a
f
ea
tu
r
es
b
ec
au
s
e
th
e
m
o
d
els
ar
e
tr
ai
n
e
d
o
n
o
n
e
o
r
two
d
atasets
with
o
n
ly
a
f
ew
e
n
v
ir
o
n
m
en
tal
v
a
r
iab
les.
T
h
is
s
tu
d
y
in
clu
d
es
em
p
ir
ical
an
aly
s
is
o
n
4
d
if
f
er
e
n
t
s
o
lar
d
atasets
,
with
a
co
n
s
id
er
ab
le
n
u
m
b
e
r
o
f
en
v
ir
o
n
m
en
tal
v
ar
iab
les
to
u
n
d
er
s
tan
d
th
e
m
o
d
el
b
eh
a
v
io
r
an
d
v
ar
iab
ilit
y
with
th
e
f
ea
tu
r
es
o
f
th
e
d
a
ta.
Data
s
et
1
i
s
o
u
r
cu
s
to
m
d
ataset
co
llected
f
r
o
m
a
PV
p
la
n
t
in
s
talled
o
n
th
e
r
o
o
f
to
p
o
f
a
n
o
r
g
an
iz
atio
n
.
T
h
e
d
ataset
en
co
m
p
ass
es
ar
o
u
n
d
(
1
0
5
0
0
0
)
r
ec
o
r
d
s
co
llected
o
v
er
2
y
e
ar
s
with
1
3
f
ea
t
u
r
es.
Mo
s
t
o
f
th
e
f
ea
tu
r
es
in
th
is
d
ataset
ar
e
en
v
ir
o
n
m
en
t
-
r
elate
d
.
Data
s
et
2
is
r
ea
l
-
tim
e
d
ata,
tak
en
f
r
o
m
o
n
e
o
f
th
e
s
o
lar
p
o
wer
g
en
er
atio
n
p
lan
ts
.
T
h
e
d
ataset
h
as
ar
o
u
n
d
(
1
0
0
0
0
0
)
r
ec
o
r
d
s
an
d
2
2
f
ea
tu
r
es.
T
h
is
d
ataset
also
h
as
m
ajo
r
en
v
ir
o
n
m
en
tal
f
ea
tu
r
es.
Data
s
et
3
is
an
o
p
en
d
ataset
av
ailab
le
f
r
o
m
a
Git
r
ep
o
s
ito
r
y
.
T
h
e
d
ataset
h
as
ar
o
u
n
d
1
3
6
0
r
ec
o
r
d
s
with
1
7
f
ea
tu
r
es.
Data
s
et
4
is
r
ea
l
-
tim
e
d
ata,
tak
en
f
r
o
m
a
PV
p
lan
t
p
o
wer
g
e
n
er
atio
n
o
r
g
an
izatio
n
.
T
h
e
d
ataset
h
as
ar
o
u
n
d
6
9
4
8
0
r
ec
o
r
d
s
with
3
8
f
ea
t
u
r
es.
T
h
ese
d
a
tasets
ar
e
m
ad
e
av
ailab
le
in
t
h
e
lin
k
s
p
r
o
v
id
e
d
in
th
e
r
ef
er
en
ce
s
.
E
ac
h
d
ataset
u
n
d
er
wen
t
a
r
ig
o
r
o
u
s
p
r
o
ce
d
u
r
e
o
f
p
r
ep
r
o
ce
s
s
in
g
,
r
ef
in
i
n
g
,
tr
an
s
f
o
r
m
in
g
,
a
n
d
s
cr
u
tin
izin
g
th
e
r
aw
d
ata
to
g
en
er
ate
m
ea
n
in
g
f
u
l
r
elatio
n
s
h
ip
s
am
o
n
g
th
e
en
v
ir
o
n
m
e
n
tal
f
ea
tu
r
es.
Miss
in
g
v
alu
es
wer
e
id
en
tifie
d
f
o
r
ea
ch
f
ea
tu
r
e
an
d
im
p
u
ted
u
s
in
g
f
o
r
war
d
f
illi
n
g
f
o
r
c
o
n
tin
u
o
u
s
m
eteo
r
o
lo
g
ical
v
ar
iab
les.
Miss
in
g
r
ec
o
r
d
s
co
n
s
titu
ted
0
.
6
%,
1
.
4
%,
0
.
7
%
,
an
d
1
.
1
%
o
f
th
e
to
tal
o
b
s
er
v
ati
o
n
s
in
Data
s
et
s
1
–
4
,
r
esp
ec
tiv
ely
.
All
co
n
tin
u
o
u
s
v
ar
iab
les
wer
e
n
o
r
m
alize
d
u
s
in
g
m
in
-
m
a
x
n
o
r
m
aliza
tio
n
,
tr
an
s
f
o
r
m
in
g
ev
er
y
f
ea
tu
r
e
in
to
th
e
r
an
g
e
[
0
,
1
]
.
T
h
e
co
r
r
elatio
n
a
n
aly
s
is
was
co
n
d
u
cted
to
d
eter
m
in
e
t
h
e
s
tr
en
g
th
o
f
r
elatio
n
s
h
ip
s
b
etwe
en
en
v
ir
o
n
m
en
tal
v
ar
i
ab
les
an
d
en
er
g
y
o
u
tp
u
t
(
p
o
wer
g
en
er
ated
)
,
estab
lis
h
in
g
a
f
o
u
n
d
atio
n
f
o
r
s
u
b
s
eq
u
en
t f
ea
tu
r
e
s
elec
tio
n
a
n
d
m
o
d
el
d
ev
el
o
p
m
en
t.
2
.
1
.
P
r
o
po
s
ed
m
et
ho
do
lo
g
y
T
em
p
er
atu
r
e,
ir
r
ad
ia
n
ce
,
win
d
s
p
ee
d
,
an
d
o
th
e
r
en
v
i
r
o
n
m
e
n
t
al
v
ar
iab
les
m
ay
c
h
an
g
e
d
u
e
t
o
wea
th
er
,
g
eo
g
r
a
p
h
ic
l
o
ca
tio
n
,
tim
e
o
f
d
ay
,
a
n
d
s
ea
s
o
n
al
p
atter
n
s
,
m
ak
in
g
s
o
lar
en
e
r
g
y
g
en
e
r
atio
n
ex
tr
em
ely
v
a
r
iab
le.
T
h
e
p
o
wer
o
u
tp
u
t
f
r
o
m
PV
p
a
n
els
ca
n
f
lu
ctu
ate
r
a
p
id
ly
a
n
d
u
n
p
r
e
d
ictab
ly
o
v
er
s
h
o
r
t
o
r
l
o
n
g
p
er
i
o
d
s
,
d
r
iv
en
b
y
ch
a
n
g
es
in
ex
te
r
n
al
en
v
ir
o
n
m
en
tal
c
o
n
d
itio
n
s
,
th
u
s
c
au
s
in
g
th
e
en
v
ir
o
n
m
en
tal
v
a
r
iab
les
to
ch
an
g
e
d
y
n
am
ically
.
B
o
th
f
ea
tu
r
e
c
o
r
r
elatio
n
s
an
d
tem
p
o
r
al
d
ep
en
d
en
cies
ar
e
d
if
f
icu
lt
to
ca
p
tu
r
e
b
y
tr
ad
itio
n
al
f
o
r
ec
asti
n
g
m
o
d
els u
n
d
e
r
th
es
e
v
ar
iab
ilit
ies.
Un
d
er
s
tan
d
in
g
th
e
tem
p
o
r
al
d
ep
en
d
e
n
cies
is
h
ig
h
ly
n
ec
ess
ar
y
to
b
u
ild
m
o
d
els
th
at
ca
n
p
r
ed
ict
ch
an
g
es
in
p
o
wer
g
en
er
atio
n
.
Ou
r
r
esear
c
h
p
r
o
p
o
s
es
a
n
o
v
el
th
r
ee
-
s
tag
e
h
y
b
r
id
m
o
d
el
t
o
u
n
d
er
s
tan
d
t
h
ese
tem
p
o
r
al
d
ep
e
n
d
en
cies
,
wh
ich
co
m
b
in
es
th
e
co
n
tr
ib
u
tio
n
s
o
f
:
Stack
ed
B
iL
STM
f
o
r
ca
p
tu
r
in
g
p
ast
an
d
p
r
esen
t
d
ep
en
d
en
cies;
A
-
DT
W
f
o
r
f
ea
tu
r
e
co
r
r
elatio
n
a
n
d
s
elec
tio
n
;
n
o
v
el
B
iLST
M
wit
h
AFG
f
o
r
r
ed
u
n
d
a
n
t
f
ea
tu
r
e
elim
in
atio
n
.
W
e
u
s
ed
two
alg
o
r
ith
m
s
:
o
n
e
f
o
r
f
ea
tu
r
e
s
elec
tio
n
an
d
o
n
e
f
o
r
m
o
d
e
l
tr
ain
in
g
.
Fig
u
r
e1
s
h
o
ws
th
e
f
lo
w
d
esig
n
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
o
l
o
g
y
,
wh
i
ch
is
elab
o
r
ated
in
f
u
r
th
er
s
ec
tio
n
s
.
B
etter
tim
e
s
er
ies
lear
n
in
g
,
id
ea
l
f
ea
tu
r
e
s
elec
tio
n
,
r
ed
u
n
d
an
t
f
ea
tu
r
e
elim
in
atio
n
,
an
d
i
n
cr
ea
s
ed
p
r
e
d
ictiv
e
ac
cu
r
ac
y
ar
e
g
u
ar
an
teed
b
y
t
h
ese
tech
n
iq
u
e
s
.
2
.
1
.
1
.
DT
W
a
nd
A
-
DT
W
A
n
u
m
b
er
o
f
en
v
ir
o
n
m
en
tal
c
o
n
d
itio
n
s
lik
e
tem
p
e
r
atu
r
e,
cl
o
u
d
co
v
er
,
a
n
d
ir
r
a
d
iatio
n
h
a
v
e
a
d
ir
ec
t
im
p
ac
t
o
n
s
o
lar
p
o
wer
o
u
tp
u
t
.
T
h
e
im
p
ac
t
o
f
th
ese
v
a
r
iab
l
es
o
n
p
o
wer
p
r
o
d
u
ctio
n
is
n
o
t
alwa
y
s
p
r
ec
is
ely
co
o
r
d
in
ate
d
.
Fo
r
e
x
am
p
le,
a
s
ce
n
ar
io
w
h
er
e
i
r
r
ad
ian
ce
s
p
ik
es
d
u
e
t
o
th
e
s
u
n
ap
p
ea
r
in
g
f
r
o
m
b
e
h
in
d
a
clo
u
d
,
b
u
t
th
e
AC
p
o
wer
d
o
esn
'
t
in
s
tan
tly
s
p
ik
e
b
ec
au
s
e
o
f
d
elay
ed
p
an
el
h
ea
tin
g
.
T
h
is
ca
u
s
es
r
ed
u
ce
d
ef
f
icien
cy
an
d
ca
n
b
e
r
ea
lly
u
n
d
e
r
s
to
o
d
.
Su
ch
s
ce
n
ar
io
s
m
ay
b
e
in
d
ic
ato
r
s
th
at
th
er
e
ar
e
d
if
f
er
e
n
t
a
s
y
n
ch
r
o
n
o
u
s
d
elay
s
an
d
n
o
n
-
lin
ea
r
in
ter
ac
tio
n
s
in
h
o
w
ch
an
g
es
in
en
v
ir
o
n
m
en
t
al
cir
cu
m
s
tan
ce
s
im
p
ac
t
th
e
p
r
o
d
u
ctio
n
o
f
s
o
lar
en
er
g
y
.
T
h
u
s
,
a
th
o
r
o
u
g
h
u
n
d
er
s
tan
d
in
g
o
f
th
e
d
y
n
am
ic
i
n
ter
ac
tio
n
s
b
etwe
en
en
v
ir
o
n
m
en
tal
f
ea
tu
r
es
a
n
d
p
o
wer
g
e
n
er
atio
n
is
ess
en
tial
f
o
r
ac
cu
r
ate
an
d
p
r
ec
is
e
f
o
r
ec
asti
n
g
.
Fro
m
th
e
th
e
o
r
y
o
f
s
ta
tis
tics
,
co
r
r
elatio
n
an
aly
s
is
an
d
lag
g
ed
d
ep
e
n
d
en
cy
ex
p
lo
r
atio
n
ar
e
p
r
ef
e
r
r
ed
t
o
an
aly
ze
s
u
ch
b
eh
av
io
r
s
.
As
co
r
r
elatio
n
o
n
ly
co
n
s
id
er
s
a
o
n
e
-
tim
e
s
tep
,
it
m
ay
g
iv
e
a
lo
w
r
esu
lt
in
ca
p
tu
r
in
g
tem
p
o
r
al
m
is
alig
n
m
en
ts
.
T
h
is
m
ay
en
d
u
p
s
elec
tin
g
f
ea
tu
r
es
th
at
f
alsely
in
d
icate
m
ax
im
al
i
n
f
lu
en
ce
o
n
th
e
ta
r
g
et
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
6
9
4
I
n
t J Po
w
E
lec
&
Dr
i Sy
s
t
,
Vo
l.
1
7
,
No
.
3
,
Sep
tem
b
er
20
2
6
:
2086
-
2
1
0
0
2090
p
o
wer
g
e
n
er
atio
n
v
ar
iab
le.
T
h
is
is
th
e
ca
s
e
wh
er
e
we
n
ee
d
a
tech
n
i
q
u
e
th
at
c
o
u
ld
wr
a
p
th
e
tim
e
d
ela
y
s
o
f
v
ar
io
u
s
f
ea
tu
r
es
,
th
u
s
m
ak
in
g
th
em
id
ea
l
f
o
r
m
an
ag
in
g
th
e
asy
n
ch
r
o
n
o
u
s
im
p
ac
ts
o
f
en
v
i
r
o
n
m
en
tal
f
ea
tu
r
es
o
n
p
o
wer
o
u
tp
u
t.
As
en
v
ir
o
n
m
en
tal
f
ea
tu
r
es
lik
e
‘
tem
p
er
at
u
r
e’
,
‘
clo
u
d
co
v
er
’
,
a
n
d
‘
s
o
la
r
ir
r
ad
ian
ce
’
d
o
n
o
t
alwa
y
s
h
av
e
p
er
f
ec
tly
s
y
n
ch
r
o
n
ized
ef
f
ec
ts
o
n
p
o
wer
o
u
t
p
u
t,
th
e
war
p
in
g
tech
n
iq
u
e
b
e
co
m
es
cr
itical
an
d
h
elp
f
u
l in
t
h
e
co
n
te
x
t o
f
s
o
lar
en
er
g
y
g
en
er
atio
n
an
d
m
ay
lea
d
to
b
etter
m
o
d
el
p
r
ed
ictio
n
s
.
DT
W
is
a
p
o
wer
f
u
l
tech
n
iq
u
e
th
at
g
o
es
b
ey
o
n
d
tr
ad
itio
n
al
co
r
r
elatio
n
m
eth
o
d
s
.
B
y
en
a
b
lin
g
n
o
n
-
lin
ea
r
alig
n
m
en
t
o
f
tim
e
s
er
ies
d
ata,
DT
W
is
a
p
o
ten
t
ap
p
r
o
ac
h
th
at
s
u
r
p
ass
es
co
n
v
en
tio
n
al
co
r
r
elatio
n
m
eth
o
d
s
.
B
ec
au
s
e
o
f
its
ex
c
ep
tio
n
al
ab
ilit
y
to
alig
n
tem
p
o
r
ally
u
n
s
y
n
ch
r
o
n
ized
tim
e
s
er
ies
d
ata,
DT
W
is
id
ea
lly
s
u
ited
f
o
r
m
a
n
ag
in
g
t
h
e
asy
n
ch
r
o
n
o
u
s
im
p
ac
ts
o
f
e
n
v
ir
o
n
m
en
tal
v
a
r
iab
les
o
n
p
o
wer
o
u
tp
u
t.
T
o
f
in
d
th
e
b
est
alig
n
m
en
t b
etwe
en
tw
o
f
ea
tu
r
e
s
eq
u
e
n
ce
s
,
th
e
DT
W
ap
p
r
o
ac
h
will
co
m
p
r
ess
o
r
s
tr
etch
tim
e
s
er
ies,
in
co
n
tr
ast to
co
n
v
en
tio
n
al
c
o
r
r
el
atio
n
m
eth
o
d
s
,
an
d
is
g
iv
e
n
b
y
: D
T
W
(
X,
Y)
=
m
in
∑
(
,
)
(
,
)
∈
ℎ
.
DT
W
ass
u
m
es
all
f
ea
tu
r
es
eq
u
ally
co
n
tr
i
b
u
te
to
tim
e
s
er
ies
s
im
ilar
ity
,
th
er
eb
y
s
elec
tin
g
a
ll
f
ea
tu
r
es.
T
h
e
p
r
o
p
o
s
ed
a
d
ap
tiv
e
DT
W
(
A
-
DT
W
)
ex
p
an
d
s
o
n
th
e
id
ea
o
f
u
s
in
g
d
y
n
am
ic
weig
h
tin
g
p
ath
an
aly
s
is
to
p
ick
th
e
m
o
s
t
s
ig
n
if
ican
t
f
ea
tu
r
es.
I
t
o
f
f
er
s
a
d
ee
p
u
n
d
er
s
tan
d
i
n
g
o
f
t
h
e
co
m
p
le
x
r
elatio
n
s
h
ip
s
b
etwe
en
m
an
y
en
v
ir
o
n
m
en
tal
v
ar
iab
les
an
d
s
o
lar
p
o
wer
g
en
e
r
atio
n
b
y
ca
lcu
latin
g
n
o
t
o
n
l
y
th
e
d
is
tan
ce
b
u
t
also
ex
am
in
in
g
th
e
alig
n
m
en
t
o
f
f
ea
t
u
r
es
u
s
in
g
a
war
p
in
g
p
at
h
(
W
P).
T
h
e
v
alu
e
o
f
th
e
war
p
in
g
p
ath
f
r
o
m
DT
W
in
d
icate
s
h
o
w
clo
s
e
t
h
e
two
-
tim
e
s
er
ies
ar
e.
B
y
lo
o
k
in
g
at
th
e
DT
W
d
is
tan
ce
,
we
ca
n
an
aly
ze
h
o
w
d
if
f
er
en
tly
th
e
tim
e
s
er
ies
b
eh
av
e
,
an
d
t
h
e
p
ath
a
lig
n
m
en
t
s
co
r
e
ex
p
lo
r
es
h
o
w
co
n
s
is
ten
tly
th
e
f
ea
tu
r
es
tr
ac
k
ea
ch
o
th
e
r
.
As
a
wh
o
le
,
th
e
alg
o
r
ith
m
d
eter
m
i
n
es
co
r
r
elatio
n
s
tr
en
g
t
h
a
n
d
p
r
o
v
id
es
in
f
o
r
m
atio
n
ab
o
u
t
w
h
ich
en
v
ir
o
n
m
en
tal
f
ac
to
r
s
h
av
e
th
e
b
ig
g
est ef
f
ec
ts
o
n
s
o
lar
en
er
g
y
p
r
o
d
u
ctio
n
.
Fig
u
r
e
1
.
Pro
p
o
s
ed
m
et
h
o
d
o
lo
g
y
p
ip
elin
e
2
.
1
.
2
.
F
ea
t
ure
s
elec
t
io
n us
ing
a
da
ptiv
e
dy
na
m
ic
t
im
e
wa
r
pin
g
(
A
-
DT
W)
wit
h
v
a
ria
nc
e
ent
ro
py
weig
ht
Giv
en
a
tar
g
et
tim
e
s
er
ies
(
)
an
d
a
c
o
llectio
n
o
f
K
tim
e
s
er
ies,
(
1
≤
≤
)
,
A
-
DT
W
aim
s
to
f
in
d
a
tim
e
s
er
ies
∈
th
at
is
clo
s
est to
.
No
tatio
n
s
in
(
1
)
an
d
(
2
)
r
ep
r
esen
t th
e
A
-
DT
W
d
is
tan
ce
an
d
cu
m
u
lativ
e
m
in
im
al
ADT
W
d
is
tan
ce
s
.
(
,
)
(
1
)
(
,
)
(
2
)
Ma
th
em
atica
lly
,
A
-
DT
W
is
g
iv
en
b
y
(
3
)
.
(
,
)
=
.
(
−
)
2
(
3
)
W
h
er
e
∈
,
∈
:
=
1
+
(
|
−
|
+
+
∈
)
.
l
og
(
1
+
.
)
(
4
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
Po
w
E
lec
&
Dr
i Sy
s
t
I
SS
N:
2088
-
8
6
9
4
A
d
a
p
tive
fo
r
g
et
-
g
a
ted
B
iLS
T
M e
n
h
a
n
ce
d
b
y
DTW b
a
s
ed
fe
a
tu
r
e
s
elec
tio
n
in
s
o
la
r
…
(
A
r
e
S
a
mb
a
s
iva
R
a
o
)
2091
is
th
e
p
r
o
p
o
s
ed
ad
ap
tiv
e
wei
g
h
t
b
ased
o
n
v
ar
ia
n
ce
.
T
h
e
war
p
in
g
p
ath
W
is
th
en
co
n
s
tr
u
cted
,
wh
er
e
th
e
wr
ap
p
in
g
p
ath
W
is
a
s
eq
u
en
c
e
o
f
p
o
in
ts
th
at
alig
n
two
tim
e
s
er
ies
s
eq
u
en
ce
s
an
d
by
,
m
in
im
izin
g
th
e
cu
m
u
lativ
e
d
is
tan
ce
as sh
o
wn
in
(
5
)
.
=
{
(
1
,
1
)
,
(
2
,
2
)
,
…
…
,
(
,
)
}
(
5
)
W
h
er
e
(
,
)
r
ep
r
esen
ts
th
e
alig
n
m
e
n
t o
f
an
d
.
I
n
a
d
ap
tiv
e
DT
W
,
th
e
g
o
al
i
s
to
f
in
d
th
e
o
p
tim
al
wr
ap
p
i
n
g
p
ath
(
W
)
b
etwe
en
two
tim
e
s
er
ies
s
eq
u
en
ce
s
b
y
m
i
n
im
izin
g
th
e
cu
m
u
lativ
e
d
is
tan
ce
g
i
v
en
b
y
(
6
)
.
(
,
)
=
{
(
−
1
,
)
+
∗
(
,
)
(
,
−
1
)
+
∗
(
,
)
(
−
1
,
−
1
)
+
∗
(
,
)
(
6
)
T
h
ese
th
r
ee
m
i
n
im
al
cu
m
u
lativ
e
co
r
r
esp
o
n
d
to
wr
a
p
p
in
g
a
cr
o
s
s
v
er
tical
s
tep
,
h
o
r
izo
n
tal
s
tep
,
an
d
d
iag
o
n
al
s
tep
.
T
h
e
weig
h
t
as
s
h
o
wn
in
(
4
)
,
is
an
ad
ap
tiv
e
weig
h
t
th
at
v
ar
ies
with
th
e
v
ar
ia
n
ce
b
etwe
en
th
e
f
ea
tu
r
es
(
,
)
.
Hig
h
er
weig
h
ts
ar
e
co
n
s
id
e
r
ed
f
o
r
(
,
)
with
h
ig
h
v
ar
ian
ce
.
(
,
)
o
f
(
6
)
in
cl
u
d
es
s
u
ch
lo
w
-
v
ar
ian
ce
p
o
in
ts
to
b
e
in
clu
d
e
d
in
t
h
e
o
p
tim
al
wr
ap
p
in
g
p
at
h
W
.
T
h
e
DT
W
d
is
tan
ce
b
et
wee
n
th
e
tar
g
et
an
d
s
elec
ted
tim
e
s
er
ies
f
ea
tu
r
e
(
,
)
is
th
e
m
in
im
u
m
cu
m
u
lativ
e
d
is
tan
ce
(
E
u
clid
ea
n
)
o
f
alig
n
in
g
all
p
o
in
ts
alo
n
g
t
h
e
war
p
in
g
p
ath
W
.
I
t is co
m
p
u
ted
as
(
7
)
.
(
,
)
=
∑
(
,
)
(
,
)
∈
(
7
)
T
h
is
m
in
im
al
d
is
tan
ce
f
r
o
m
(
7
)
s
h
o
ws
is
clo
s
est
to
a
n
d
ar
e
s
elec
ted
f
ea
t
u
r
es
with
im
p
o
r
tan
ce
.
Fig
u
r
e
2
s
h
o
ws
a
s
am
p
le
o
f
f
o
u
r
F1
,
F2
,
F3
,
a
n
d
F4
tim
e
s
er
ies
f
ea
tu
r
es
co
m
p
ar
ed
to
th
e
t
ar
g
et
f
ea
tu
r
e
f
o
r
s
im
ilar
ity
alig
n
m
e
n
t.
Fig
u
r
e
2
(
a
)
s
h
o
ws
in
itially
all
f
ea
tu
r
es
t
o
b
e
co
m
p
letely
d
is
s
im
ilar
to
th
e
tar
g
et
;
af
ter
DT
W
r
ea
lig
n
m
en
t
s
h
o
wn
in
Fig
u
r
e
2
(
b
)
,
th
e
f
e
atu
r
es
F2
an
d
F3
ar
e
s
elec
ted
to
b
e
s
im
ilar
to
th
e
tar
g
et,
wh
ich
ar
e
s
h
o
wn
in
Fi
g
u
r
e
2
(
c
)
.
T
h
ese
s
elec
ted
f
ea
tu
r
es
ar
e
s
h
o
wn
in
Fig
u
r
e
2
(
d
)
an
d
ar
e
tak
en
in
t
o
m
o
d
el
tr
ain
in
g
.
T
ab
le
1
s
u
m
m
ar
izes th
e
s
y
m
b
o
ls
an
d
a
b
b
r
e
v
iatio
n
s
u
s
ed
in
th
is
p
ap
er
.
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
2
.
Sam
p
le
tim
e
s
er
ies f
ea
tu
r
e
s
elec
tio
n
in
DT
W
alig
n
m
en
t: (
a)
o
r
ig
in
al
f
ea
tu
r
es,
(
b
)
DT
W
r
ea
lig
n
m
en
t,
(
c)
s
elec
ted
s
im
ilar
f
ea
tu
r
es,
a
n
d
(
d
)
f
ea
tu
r
es selecte
d
f
o
r
m
o
d
el
tr
ain
in
g
T
ab
le
1
.
Sy
m
b
o
ls
an
d
a
b
b
r
e
v
iatio
n
s
u
s
ed
in
th
e
p
ap
er
S
y
mb
o
l
D
e
scri
p
t
i
o
n
T
r
e
f
Ta
r
g
e
t
t
i
m
e
s
e
r
i
e
s
A
d
a
p
t
i
v
e
w
e
i
g
h
t
W
W
a
r
p
i
n
g
p
a
t
h
D
(
,
)
D
TW d
i
st
a
n
c
e
f
adp
(
t
)
P
r
o
p
o
se
d
a
p
p
r
o
a
c
h
:
f
o
r
g
e
t
g
a
t
e
α
Le
a
r
n
a
b
l
e
p
a
r
a
m
e
t
e
r
δ
S
mal
l
q
u
a
n
t
i
t
y
γ
S
h
i
f
t
sm
o
o
t
h
i
n
g
p
a
r
a
me
t
e
r
(
,
)
D
i
f
f
e
r
e
n
t
i
a
b
l
e
l
o
ss fu
n
c
t
i
o
n
Lo
ss fr
o
m
t
h
e
A
F
G
-
B
i
LST
M
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
6
9
4
I
n
t J Po
w
E
lec
&
Dr
i Sy
s
t
,
Vo
l.
1
7
,
No
.
3
,
Sep
tem
b
er
20
2
6
:
2086
-
2
1
0
0
2092
Alg
o
r
ith
m
1
.
Featu
r
e
s
elec
tio
n
with
A
-
DT
W
I
n
p
u
t:
T
im
e
s
er
ies
t
ar
g
et
an
d
(
1
≤
≤
)
1)
C
o
m
p
u
te
ad
ap
tiv
e
weig
h
ts
g
iv
en
=
1
+
(
|
−
|
+
+
∈
)
.
l
og
(
1
+
.
)
.
2)
C
o
m
p
u
te
war
p
in
g
p
ath
W
b
y
m
in
im
izin
g
(
,
)
u
s
in
g
:
(
,
)
=
min
{
(
−
1
,
)
+
∗
(
,
)
(
,
−
1
)
+
∗
(
,
)
(
−
1
,
−
1
)
+
2
∗
∗
(
,
)
3)
C
o
m
p
u
te
(
,
)
=
∑
(
,
)
(
,
)
∈
;
(
,
)
∈
.
4)
Select
f
ea
tu
r
es
,
with
m
in
im
al
(
,
)
f
r
o
m
s
tep
3
.
2
.
1
.
3
.
Nee
d
f
o
r
wra
pp
ing
in t
im
e
s
er
ies s
o
la
r
po
wer
(
ener
g
y
)
predict
io
n wit
h v
a
ria
nce
ent
ro
py
weig
ht
Me
asu
r
in
g
s
im
ilar
ity
b
etwe
en
two
tim
e
s
er
ies
f
ea
tu
r
es
i
s
ch
allen
g
in
g
b
ec
au
s
e
o
f
f
a
cto
r
s
lik
e
co
m
p
lex
ity
,
v
ar
iab
ilit
y
,
an
d
tem
p
o
r
al
n
atu
r
e.
T
im
e
s
er
ies
d
ata
o
f
ten
h
a
v
e
s
h
if
ts
in
tim
e,
m
ea
n
in
g
th
e
s
am
e
p
o
in
ts
m
ay
ap
p
ea
r
at
d
if
f
er
en
t
tim
estam
p
s
.
B
ec
au
s
e
o
f
th
e
tem
p
o
r
al
n
atu
r
e
o
f
tim
e
s
er
ie
s
,
th
e
f
ea
t
u
r
es
m
ay
in
itially
s
ee
m
to
b
e
d
is
s
im
ilar
,
b
u
t
clo
s
e
o
b
s
er
v
atio
n
b
y
s
tr
etch
in
g
o
r
co
m
p
r
ess
in
g
th
e
tim
e
ax
es
m
ay
r
ea
lly
h
elp
in
id
en
tif
y
in
g
t
h
e
r
elatio
n
s
h
ip
s
b
etwe
en
th
ese
tem
p
o
r
a
l
f
ea
tu
r
es.
I
n
s
ce
n
ar
i
o
s
o
f
s
o
la
r
en
er
g
y
p
r
e
d
ictio
n
,
s
im
ilar
ity
an
aly
s
is
will
f
etch
s
tr
o
n
g
p
r
e
d
icto
r
s
.
A
-
DT
W
wit
h
v
ar
ian
ce
en
tr
o
p
y
weig
h
t
ali
g
n
s
tim
e
s
eq
u
en
ce
s
b
y
co
m
p
r
ess
io
n
o
r
s
tr
etch
i
n
g
f
o
r
b
etter
a
n
aly
s
is
o
f
th
e
d
ata.
Ou
r
p
r
o
p
o
s
ed
A
-
DT
W
ap
p
r
o
ac
h
u
s
es
v
ar
ian
ce
en
tr
o
p
y
weig
h
t
,
as
s
o
lar
p
o
wer
g
e
n
er
atio
n
is
u
n
ce
r
tain
with
m
o
s
t
o
f
th
e
wea
th
er
v
ar
iab
les.
I
f
b
o
th
v
a
r
ian
ce
s
ar
e
h
ig
h
,
t
h
e
weig
h
t
ter
m
g
r
o
ws
en
o
r
m
o
u
s
ly
,
wh
ich
is
n
o
t
s
tab
le
f
o
r
th
e
b
est
f
ea
tu
r
e
s
elec
tio
n
.
Ou
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
p
en
alize
s
th
is
ex
p
lo
s
iv
e
b
eh
av
i
o
r
with
a
lo
g
(
)
ter
m
in
th
e
n
u
m
e
r
ato
r
,
as sh
o
wn
in
(
4
)
,
m
ak
in
g
t
h
e
m
o
d
el
less
s
en
s
itiv
e
to
ex
tr
em
e
v
ar
ian
ce
.
T
h
e
g
r
ap
h
in
Fig
u
r
e
3
(
a)
s
h
o
ws
th
e
r
elatio
n
s
h
ip
b
etwe
en
ir
r
ad
ian
ce
(
b
lu
e
cu
r
v
e)
a
n
d
p
o
wer
g
en
er
ated
(
g
r
ee
n
cu
r
v
e)
,
in
d
i
ca
tin
g
th
e
lag
p
h
en
o
m
en
o
n
.
T
h
e
p
o
wer
cu
r
v
e
s
h
o
ws
a
s
li
g
h
t
lag
,
wh
ich
co
u
ld
r
ed
u
ce
its
s
im
ilar
ity
to
ir
r
ad
ian
ce
.
Fig
u
r
e
3
(
b
)
s
h
o
ws
th
e
r
e
alig
n
m
en
t
o
f
th
e
s
am
e
lag
p
o
in
ts
af
ter
ap
p
ly
in
g
wr
ap
p
in
g
,
s
h
o
win
g
h
o
w
th
e
p
ea
k
s
co
u
ld
b
e
co
r
r
elate
d
f
o
r
d
ata
s
im
ilar
ities
.
Als
o
,
th
e
d
if
f
er
en
ce
s
b
etwe
en
th
ese
lag
s
ar
e
m
in
im
ized
an
d
m
ad
e
u
n
if
o
r
m
.
(
a)
(
b)
Fig
u
r
e
3
.
DT
W
alig
n
m
e
n
t o
f
i
r
r
ad
ian
ce
a
n
d
g
e
n
er
ated
p
o
we
r
: (
a)
b
e
f
o
r
e
alig
n
m
en
t a
n
d
(
b
)
af
ter
alig
n
m
e
n
t
2
.
2
.
P
r
o
po
s
ed
a
da
ptiv
e
f
o
rg
e
t
g
a
t
e
s
t
a
ck
ed
B
iL
ST
M
(
AF
G
-
B
iL
S
T
M
)
Fo
r
tim
e
s
er
ies
m
o
d
ellin
g
,
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
s
(
R
NNs)
an
d
th
eir
v
ar
ia
n
ts
h
av
e
b
ee
n
ex
ten
s
iv
ely
u
tili
ze
d
.
Ho
wev
e
r
,
b
ec
a
u
s
e
o
f
v
a
n
is
h
in
g
g
r
a
d
ien
ts
,
ty
p
ical
R
NNs
h
av
e
p
r
o
b
le
m
s
with
s
h
o
r
t
-
te
r
m
m
em
o
r
y
.
B
y
p
r
eser
v
in
g
lo
n
g
-
r
an
g
e
d
ep
e
n
d
en
cies,
lo
n
g
-
s
h
o
r
t
ter
m
m
em
o
r
y
(
L
STM
)
n
etwo
r
k
s
h
elp
to
allev
iate
th
ese
p
r
o
b
lem
s
.
T
h
e
L
STM
u
s
es th
r
ee
g
ates: f
o
r
g
et
(
f
t
)
,
in
p
u
t
(i
t
)
,
an
d
o
u
tp
u
t
(o
t
)
g
iv
en
b
y
(
8
)
-
(
1
0
)
.
=
(
+
ℎ
−
1
+
)
(
8
)
=
(
+
ℎ
−
1
+
)
(
9
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
Po
w
E
lec
&
Dr
i Sy
s
t
I
SS
N:
2088
-
8
6
9
4
A
d
a
p
tive
fo
r
g
et
-
g
a
ted
B
iLS
T
M e
n
h
a
n
ce
d
b
y
DTW b
a
s
ed
fe
a
tu
r
e
s
elec
tio
n
in
s
o
la
r
…
(
A
r
e
S
a
mb
a
s
iva
R
a
o
)
2093
=
(
+
ℎ
−
1
+
)
(
1
0
)
B
u
t
m
an
y
o
f
th
e
cu
r
r
e
n
t
r
esea
r
ch
wo
r
k
s
o
n
PV
d
ata
s
h
o
we
d
lo
w
p
e
r
f
o
r
m
an
ce
with
L
STM
,
as
th
ey
f
ailed
to
an
aly
ze
th
e
p
ast
an
d
p
r
esen
t
d
ata
d
ep
en
d
e
n
cies
o
f
tim
e
s
er
ies
s
eq
u
en
ce
s
,
p
ar
ticu
lar
ly
wh
e
n
p
r
e
d
ictin
g
p
o
wer
g
en
er
ated
with
m
a
n
y
o
f
th
e
e
n
v
ir
o
n
m
e
n
tal
p
r
ed
icto
r
v
ar
ia
b
les.
I
n
th
e
p
r
o
p
o
s
ed
s
tu
d
y
,
in
o
r
d
er
to
im
p
r
o
v
e
th
e
lear
n
in
g
o
f
b
o
th
p
ast
an
d
cu
r
r
en
t
d
ep
e
n
d
en
cies,
we
in
v
esti
g
ate
s
tack
ed
b
id
ir
ec
tio
n
al
L
STM
s
(
B
iLST
M
s
)
t
o
b
e
tr
ain
ed
with
th
e
f
ea
t
u
r
es
wh
er
e
m
an
y
en
v
ir
o
n
m
en
tal
v
ar
iab
les
ar
e
i
n
v
o
lv
e
d
.
C
o
n
v
en
tio
n
al
L
STM
n
etwo
r
k
s
h
an
d
le
in
p
u
t
in
a
f
o
r
war
d
o
r
ien
tatio
n
,
ef
f
icien
tly
ca
p
tu
r
in
g
p
r
ev
io
u
s
d
ep
en
d
en
cies
wh
ile
d
is
r
eg
ar
d
in
g
in
f
o
r
m
atio
n
th
at
m
ay
b
e
co
m
e
ac
ce
s
s
ib
le
in
th
e
s
eq
u
en
ce
in
th
e
f
u
tu
r
e.
B
y
ad
d
in
g
a
s
ec
o
n
d
L
STM
lay
er
th
at
p
r
o
ce
s
s
es
th
e
s
e
q
u
en
ce
in
r
ev
er
s
e,
b
id
ir
ec
tio
n
al
L
STM
s
(
B
iLST
Ms)
o
v
er
co
m
e
th
is
r
estrictio
n
an
d
p
r
o
v
id
e
a
m
o
r
e
th
o
r
o
u
g
h
co
m
p
r
e
h
en
s
io
n
o
f
s
eq
u
en
tial
d
ep
en
d
en
cies.
Fu
r
th
er
m
o
r
e
,
h
ier
a
r
ch
ical
f
ea
tu
r
e
ex
tr
ac
tio
n
is
m
ad
e
p
o
s
s
ib
le
b
y
s
tack
in
g
n
u
m
er
o
u
s
B
iLST
M
lay
er
s
,
wh
ich
en
h
an
c
es th
e
ex
p
r
ess
iv
en
ess
o
f
th
e
m
o
d
el.
Giv
en
an
in
p
u
t
s
eq
u
e
n
ce
=
[
1
,
2
,
3
,
…
…
]
,
a
B
iLST
M
co
m
p
u
tes
h
id
d
en
s
tates
(
h
t
)
in
b
o
th
d
ir
ec
tio
n
s
.
ℎ
⃗
=
(
,
ℎ
⃗
−
1
)
(
1
1
)
ℎ
⃖
⃗
=
(
,
ℎ
⃖
⃗
+
1
)
(
1
2
)
W
h
er
e
ℎ
⃗
an
d
ℎ
⃖
⃗
r
e
p
r
esen
t
f
o
r
war
d
an
d
b
ac
k
war
d
h
id
d
e
n
s
tates
at
tim
e
t.
T
h
e
f
in
al
r
ep
r
esen
tatio
n
at
ea
c
h
s
tep
is
th
e
co
n
ca
ten
atio
n
o
f
b
o
th
,
g
iv
en
b
y
(
1
3
)
.
ℎ
=
[
ℎ
⃗
,
ℎ
⃖
⃗
]
(
1
3
)
T
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
ad
d
s
n
o
v
elty
to
th
e
s
tack
ed
B
iLST
M
m
o
d
el
b
y
u
s
in
g
a
lear
n
ab
le
,
p
ar
am
etr
ized
f
o
r
g
et
g
ate
in
th
e
L
STM
lay
er
s
,
wh
ich
ad
ap
ts
to
th
e
d
ata
f
ea
tu
r
es
s
elec
te
d
f
r
o
m
th
e
A
-
DT
W
m
o
d
u
le.
T
h
e
co
n
v
en
tio
n
al
f
o
r
g
et
g
ate
in
L
STM
co
n
tr
o
ls
th
e
am
o
u
n
t
o
f
p
r
ev
io
u
s
m
em
o
r
y
r
etain
ed
o
r
d
is
ca
r
d
ed
u
s
in
g
f
ix
ed
lear
n
e
d
weig
h
ts
an
d
s
ig
m
o
id
ac
tiv
atio
n
f
u
n
ctio
n
s
.
I
ts
b
eh
a
v
io
r
r
em
ain
s
g
en
e
r
ally
u
n
if
o
r
m
ac
r
o
s
s
in
p
u
t
f
ea
tu
r
es
an
d
d
o
es
n
o
t
ex
p
licitly
co
n
s
id
er
tem
p
o
r
al
f
ea
tu
r
e
c
o
r
r
elat
io
n
o
r
r
ed
u
n
d
a
n
cy
.
Fo
r
ex
am
p
le,
b
ec
au
s
e
o
f
m
u
l
tip
le
lag
alig
n
m
en
ts
o
f
th
e
f
ea
tu
r
es,
in
th
e
A
-
DT
W
m
o
d
u
le
,
s
o
m
e
r
ed
u
n
d
a
n
t
f
ea
tu
r
es
lik
e
h
u
m
id
ity
an
d
p
o
in
t
h
u
m
id
ity
,
an
d
win
d
s
p
ee
d
an
d
win
d
f
lo
w
m
ay
en
d
u
p
s
h
o
win
g
a
m
in
im
al
DT
W
d
is
tan
ce
an
d
th
er
eb
y
b
e
s
elec
ted
b
y
th
e
A
-
DT
W
f
ea
t
u
r
e
s
elec
tio
n
m
o
d
u
le.
W
ith
a
co
n
v
en
tio
n
al
f
o
r
g
et
g
ate
,
th
ese
h
ig
h
ly
c
o
r
r
elate
d
o
r
r
ed
u
n
d
a
n
t
en
v
ir
o
n
m
en
tal
f
ea
tu
r
es
m
ay
s
till
b
e
p
r
o
p
a
g
ated
th
r
o
u
g
h
th
e
m
em
o
r
y
s
tates.
T
h
e
p
r
esen
ce
o
f
r
e
d
u
n
d
an
t
f
ea
tu
r
es
ca
n
lead
to
m
o
d
el
o
v
er
f
itti
n
g
,
ad
v
er
s
ely
im
p
ac
tin
g
m
o
d
el
p
er
f
o
r
m
an
ce
.
Ou
r
p
r
o
p
o
s
ed
a
d
ap
tiv
e
f
o
r
g
et
g
ate
lay
er
u
s
es
a
lear
n
ab
le
p
ar
am
eter
(
α
)
th
a
t
ad
ap
ts
to
th
e
d
ata
f
ea
tu
r
es
,
s
elec
ted
f
r
o
m
th
e
A
-
DT
W
m
o
d
u
le.
T
h
is
p
ar
am
ete
r
p
en
alize
s
th
e
r
ed
u
n
d
an
t
f
ea
t
u
r
es.
T
h
e
f
o
r
g
et
g
ate
(
)
in
o
u
r
ap
p
r
o
ac
h
is
g
iv
en
b
y
(
1
4
)
:
(
)
=
(
(
,
)
)
.
(
+
ℎ
−
1
+
)
(
1
4
)
with
α
as
a
lear
n
ab
le
p
ar
am
ete
r
th
at
ad
ap
ts
to
th
e
d
ata
f
ea
tu
r
es
,
s
elec
ted
f
r
o
m
th
e
DT
W
m
o
d
u
le.
Fro
m
(
7
)
,
(
,
)
is
th
e
DT
W
d
is
tan
ce
b
etwe
e
n
a
f
ea
tu
r
e
(
)
an
d
th
e
tar
g
et
tim
e
s
er
ies
(
)
.
T
h
e
lear
n
a
b
le
p
ar
am
eter
α
co
n
v
er
ts
th
e
DT
W
d
is
tan
ce
in
to
a
weig
h
t u
s
in
g
(
1
5
)
.
(
D
(
T
r
ef
,
X
j
)
)
=
1
1
+
e
D
(
T
r
ef
,
X
j
)
(
1
5
)
Fo
r
X
j1
,
X
j2
∈
X
j
two
s
im
ilar
r
ed
u
n
d
an
t
f
e
atu
r
es
s
elec
ted
in
th
e
A
-
DT
W
m
o
d
u
le
,
th
e
y
s
h
o
w
n
ea
r
ly
id
e
n
tical
weig
h
ts
,
i.e
(
(
,
1
)
)
≈
(
(
,
2
)
)
an
d
ar
e
class
if
ied
as
r
ed
u
n
d
a
n
t.
W
e
d
ef
in
e
th
e
r
ed
u
n
d
an
t
f
ea
tu
r
e
d
is
cr
im
in
atio
n
co
n
d
itio
n
(
R
FDC
)
as,
f
o
r
X
j1
,
X
j2
∈
X
j
two
r
ed
u
n
d
an
t f
ea
tu
r
es
:
|
(
D
(
T
r
ef
,
X
j1
)
)
−
(
D
(
T
r
ef
,
X
j2
)
)
|
≤
(
1
6
)
w
h
er
e
δ (
<0
.
0
5
)
is
a
s
m
all
q
u
a
n
tity
.
Un
lik
e
th
e
co
n
v
en
tio
n
al
L
STM
f
o
r
g
et
g
ate
(
8
)
,
wh
ich
c
o
m
p
u
tes
th
e
m
em
o
r
y
r
eten
tio
n
co
ef
f
icien
t
s
o
lely
f
r
o
m
th
e
c
u
r
r
en
t
in
p
u
t
an
d
p
r
e
v
io
u
s
h
id
d
en
s
tate,
th
e
p
r
o
p
o
s
ed
ad
ap
tiv
e
f
o
r
g
et
g
at
e
,
as
s
h
o
wn
in
(
1
4
)
,
ex
p
licitly
in
co
r
p
o
r
ates
th
e
te
m
p
o
r
al
s
im
ilar
ity
b
etwe
en
e
ac
h
s
elec
ted
f
ea
t
u
r
e
an
d
th
e
tar
g
et
tim
e
s
er
ies
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
6
9
4
I
n
t J Po
w
E
lec
&
Dr
i Sy
s
t
,
Vo
l.
1
7
,
No
.
3
,
Sep
tem
b
er
20
2
6
:
2086
-
2
1
0
0
2094
th
r
o
u
g
h
th
e
ad
ap
tiv
e
co
ef
f
icien
t
α
(
D(
T
ref
,
X
j
))
.
C
o
n
s
eq
u
en
tly
,
th
e
p
r
o
p
o
s
ed
g
ate
ca
n
b
e
ex
p
r
ess
ed
as
(
)
=
α
(
D(
T
r
ef
,
Xj)
)
,
in
d
icatin
g
t
h
at
it
is
a
DT
W
-
g
u
id
ed
s
ca
lin
g
o
f
th
e
co
n
v
en
tio
n
al
f
o
r
g
et
g
ate.
W
h
en
a
f
ea
tu
r
e
ex
h
ib
its
a
s
m
aller
A
-
DT
W
d
is
tan
ce
to
th
e
r
ef
er
en
ce
tar
g
et,
in
d
icatin
g
s
tr
o
n
g
er
te
m
p
o
r
al
co
r
r
elatio
n
,
α
ap
p
r
o
ac
h
es
u
n
ity
,
allo
win
g
th
e
f
ea
tu
r
e
to
r
etain
m
o
s
t
o
f
its
h
is
to
r
ical
m
em
o
r
y
.
C
o
n
v
er
s
ely
,
f
ea
tu
r
es
wit
h
lar
g
er
A
-
DT
W
d
is
tan
ce
s
r
ec
ei
v
e
s
m
aller
ad
a
p
tiv
e
c
o
ef
f
icien
ts
,
th
er
eb
y
r
ed
u
cin
g
th
eir
c
o
n
t
r
ib
u
tio
n
to
th
e
ce
ll
s
tate
an
d
p
r
ev
en
tin
g
r
ed
u
n
d
an
t
o
r
wea
k
ly
co
r
r
elate
d
in
f
o
r
m
atio
n
f
r
o
m
p
r
o
p
ag
atin
g
t
h
r
o
u
g
h
s
u
cc
ess
iv
e
m
em
o
r
y
u
p
d
ates.
T
h
e
p
r
o
p
o
s
ed
B
iLST
M
f
o
r
g
et
g
ate
is
t
r
ain
ed
to
d
is
ca
r
d
(
f
o
r
g
et)
o
n
e
am
o
n
g
th
ese
two
f
ea
tu
r
es
with
m
in
im
al
R
FDC
an
d
m
ax
im
u
m
co
r
r
elatio
n
b
et
wee
n
f
ea
tu
r
es
X
j
.
T
h
e
B
iLST
M
f
o
r
g
et
g
ate,
wh
ich
is
s
h
o
wn
in
(
1
4
)
,
iter
ativ
ely
tr
ain
s
its
elf
o
n
th
e
er
r
o
r
f
u
n
ctio
n
g
iv
en
in
(
1
8
)
.
Fig
u
r
e
4
s
h
o
ws
th
e
ar
ch
itectu
r
e
o
f
o
u
r
p
r
o
p
o
s
ed
m
o
d
el.
Fig
u
r
e
5
(
a)
s
h
o
ws
o
n
e
ce
ll
s
tate
o
f
th
e
AFG
B
iLST
M
f
o
r
g
et
g
ate
(
AFG)
.
Fig
u
r
e
5
(
b
)
s
h
o
ws
th
e
p
r
o
p
o
s
ed
A
-
DT
W
a
n
d
AFG
B
iLST
M
m
o
d
u
les.
T
h
e
s
tep
s
in
th
e
tr
ai
n
in
g
p
r
o
ce
s
s
o
f
AFG
-
B
iLST
M
ar
e
as
s
h
o
wn
in
Alg
o
r
ith
m
2
.
T
h
e
co
r
r
elatio
n
an
al
y
s
is
a
s
s
h
o
wn
in
T
ab
le
2
d
eter
m
in
es
th
e
s
tr
en
g
th
o
f
r
elatio
n
s
h
ip
s
b
etwe
en
en
v
ir
o
n
m
en
tal
v
ar
iab
les
an
d
en
e
r
g
y
o
u
tp
u
t
(
p
o
wer
g
e
n
er
ated
)
.
T
h
e
tab
le
in
clu
d
es
th
e
an
aly
s
is
o
n
two
d
atasets
.
T
h
e
t
ab
le
s
h
o
ws th
e
co
r
r
elatio
n
an
d
A
-
DT
W
d
is
tan
ce
s
.
Fig
u
r
e
4
.
Pro
p
o
s
ed
s
tack
ed
B
i
L
STM
a
r
ch
itectu
r
e
(
a)
(
b
)
Fig
u
r
e
5
.
Pro
p
o
s
ed
a
d
ap
tiv
e
f
o
r
g
et
g
ate
(
AFG)
B
iLST
M
ar
ch
itectu
r
e:
(
a)
AFG
B
iLST
M
f
o
r
g
et
-
g
ate
ce
ll st
ate
an
d
(
b
)
p
r
o
p
o
s
ed
A
-
DT
W
an
d
AFG
B
iLST
M
m
o
d
u
les
Alg
o
r
ith
m
2
.
Mo
d
el
t
r
ain
i
n
g
:
Stack
ed
AFG
-
B
iLST
M
I
n
p
u
t: Selec
ted
f
ea
tu
r
es Xj
f
r
o
m
Alg
o
r
ith
m
1
1.
T
r
ain
B
iLST
M
with
ad
ap
tiv
e
f
o
r
g
et
g
ate
g
iv
en
b
y
f
a
d
p
(
t)
2.
C
alcu
late
lo
s
s
D
(
T
r
ef
,
X
j
)
f
r
o
m
th
e
A
-
DT
W
m
o
d
u
le
in
(
1
7
)
.
3.
C
alcu
late
lo
s
s
f
r
o
m
th
e
AFG
B
iLST
M
m
o
d
u
le
in
(
1
8
)
.
4.
C
alcu
late
to
tal
lo
s
s
,
ℒ
f
r
o
m
(
1
9
)
5.
R
ep
ea
t
tr
ain
in
g
to
m
in
im
ize
t
h
e
to
tal
lo
s
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
Po
w
E
lec
&
Dr
i Sy
s
t
I
SS
N:
2088
-
8
6
9
4
A
d
a
p
tive
fo
r
g
et
-
g
a
ted
B
iLS
T
M e
n
h
a
n
ce
d
b
y
DTW b
a
s
ed
fe
a
tu
r
e
s
elec
tio
n
in
s
o
la
r
…
(
A
r
e
S
a
mb
a
s
iva
R
a
o
)
2095
T
ab
le
2
.
C
o
r
r
elatio
n
Vs A
-
DT
W
d
is
tan
ce
o
n
Data
s
et
-
1
an
d
Data
s
et
-
2
D
a
t
a
s
e
t
F
e
a
t
u
r
e
C
o
r
r
e
l
a
t
i
o
n
F
e
a
t
u
r
e
A
-
D
TW d
i
s
t
a
n
c
e
D
a
t
a
s
e
t
-
1
:
Ta
r
g
e
t
f
e
a
t
u
r
e
-
p
o
w
e
r
_
g
e
n
e
r
a
t
e
d
i
r
r
a
d
i
a
n
c
e
0
.
9
8
i
r
r
a
d
i
a
n
c
e
0
.
4
9
a
i
r
_
t
e
m
p
e
r
a
t
u
r
e
0
.
7
4
h
u
m
i
d
i
t
y
0
.
8
3
a
mb
i
e
n
t
_
t
e
m
p
e
r
a
t
u
r
0
.
5
7
w
i
n
d
_
d
i
r
e
c
t
i
o
n
1
.
7
2
su
r
f
a
c
e
_
t
e
m
p
e
r
a
t
u
r
e
0
.
5
3
w
i
n
d
_
s
p
e
e
d
1
.
8
1
d
e
w
_
t
e
mp
e
r
a
t
u
r
e
0
.
5
0
a
i
r
_
t
e
m
p
e
r
a
t
u
r
e
2
.
0
1
w
i
n
d
_
s
p
e
e
d
0
.
1
4
su
r
f
a
c
e
_
t
e
m
p
e
r
a
t
u
r
e
2
.
1
4
a
t
m
o
sp
h
e
r
i
c
_
p
r
e
ssu
r
0
.
1
3
d
e
w
_
t
e
mp
e
r
a
t
u
r
e
2
.
2
0
w
i
n
d
_
d
i
r
e
c
t
i
o
n
0
.
0
9
a
t
m
o
sp
h
e
r
i
c
_
p
r
e
ssu
r
e
2
.
2
1
h
u
m
i
d
i
t
y
-
0
.
6
5
a
mb
i
e
n
t
_
t
e
m
p
e
r
a
t
u
r
e
2
.
2
6
v
a
p
o
u
r
_
p
r
e
s
u
r
e
-
0
.
9
1
v
a
p
o
u
r
_
p
r
e
s
u
r
e
2
.
4
1
C
l
o
u
d
_
c
o
v
e
r
-
0
.
9
6
C
l
o
u
d
_
c
o
v
e
r
2
.
6
9
r
e
l
a
t
i
v
e
_
h
u
m
i
d
i
t
y
-
0
.
9
7
r
e
l
a
t
i
v
e
_
h
u
m
i
d
i
t
y
2
.
8
8
c
l
o
u
d
_
p
r
e
c
i
p
i
t
a
t
i
o
n
-
0
.
9
8
c
l
o
u
d
_
p
r
e
c
i
p
i
t
a
t
i
o
n
3
.
0
4
D
a
t
a
s
e
t
-
2
:
t
a
r
g
e
t
f
e
a
t
u
r
e
-
p
o
w
e
r
g
e
n
e
r
a
t
e
d
S
o
l
a
r
i
r
r
a
d
i
a
n
c
e
0
.
9
7
S
o
l
a
r
i
r
r
a
d
i
a
n
c
e
0
.
6
2
D
i
f
f
u
se
s
o
l
a
r
r
a
d
i
a
t
i
o
n
0
.
6
8
D
i
r
e
c
t
n
o
r
ma
l
i
r
r
a
d
i
a
n
c
e
0
.
9
7
D
i
r
e
c
t
n
o
r
ma
l
i
r
r
a
d
i
a
n
c
e
0
.
6
5
D
i
f
f
u
se
s
o
l
a
r
r
a
d
i
a
t
i
o
n
0
.
9
7
U
V
R
a
d
i
a
t
i
o
n
0
.
6
2
U
V
r
a
d
i
a
t
i
o
n
0
.
9
7
W
e
t
b
u
l
b
t
e
m
p
e
r
a
t
u
r
e
0
.
1
3
R
e
l
a
t
i
v
e
h
u
mi
d
i
t
y
1
.
5
0
W
i
n
d
g
u
st
0
.
0
0
4
A
t
mo
s
p
h
e
r
i
c
p
r
e
ss
u
r
e
1
.
6
4
V
i
si
b
i
l
i
t
y
0
.
0
0
1
7
W
i
n
d
s
p
e
e
d
1
.
6
4
D
e
w
p
o
i
n
t
t
e
mp
e
r
a
t
u
r
e
0
.
0
0
1
5
C
l
o
u
d
c
o
v
e
r
(
%)
1
.
6
4
G
r
o
u
n
d
su
r
f
a
c
e
t
e
mp
e
r
a
t
u
r
e
0
.
0
0
1
4
A
i
r
t
e
m
p
e
r
a
t
u
r
e
1
.
6
4
S
u
r
f
a
c
e
a
l
b
e
d
o
0
.
0
0
1
1
S
u
r
f
a
c
e
A
l
b
e
d
o
1
.
6
5
P
r
e
c
i
p
i
t
a
t
i
o
n
r
a
t
e
0
.
0
0
0
9
To
t
a
l
sk
y
c
o
v
e
r
1
.
6
5
S
n
o
w
d
e
p
t
h
0
.
0
0
0
9
H
e
a
t
i
n
d
e
x
1
.
6
5
To
t
a
l
sk
y
c
o
v
e
r
0
.
0
0
0
6
W
i
n
d
g
u
st
1
.
6
5
A
i
r
t
e
m
p
e
r
a
t
u
r
e
0
.
0
0
0
2
D
e
w
p
o
i
n
t
t
e
mp
e
r
a
t
u
r
e
1
.
6
6
W
i
n
d
d
i
r
e
c
t
i
o
n
0
.
0
0
0
1
F
e
e
l
s
-
l
i
k
e
Te
m
p
e
r
a
t
u
r
e
1
.
6
6
F
e
e
l
s
-
l
i
k
e
Te
m
p
e
r
a
t
u
r
e
-
0
.
0
0
0
5
P
r
e
c
i
p
i
t
a
t
i
o
n
r
a
t
e
1
.
6
6
H
e
a
t
i
n
d
e
x
-
0
.
0
0
1
4
V
i
si
b
i
l
i
t
y
1
.
6
6
R
e
l
a
t
i
v
e
h
u
mi
d
i
t
y
-
0
.
0
0
2
3
W
i
n
d
d
i
r
e
c
t
i
o
n
1
.
6
6
W
i
n
d
s
p
e
e
d
-
0
.
0
0
2
8
W
e
t
b
u
l
b
t
e
m
p
e
r
a
t
u
r
e
1
.
6
6
C
l
o
u
d
c
o
v
e
r
(
%)
-
0
.
0
0
3
5
S
n
o
w
d
e
p
t
h
1
.
6
7
A
t
mo
s
p
h
e
r
i
c
p
r
e
ss
u
r
e
-
0
.
0
0
5
9
G
r
o
u
n
d
su
r
f
a
c
e
t
e
mp
e
r
a
t
u
r
e
1
.
6
7
S
o
l
a
r
Ze
n
i
t
h
a
n
g
l
e
-
0
.
6
0
8
1
S
o
l
a
r
Ze
n
i
t
h
a
n
g
l
e
1
.
9
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
All
ex
p
er
im
en
ts
ar
e
r
u
n
o
n
t
h
e
d
ee
p
lear
n
in
g
f
r
a
m
ewo
r
k
Py
T
o
r
c
h
,
wh
ich
allo
wed
f
lex
ib
ilit
y
in
d
esig
n
in
g
,
tr
ain
in
g
,
an
d
e
v
alu
atin
g
m
o
d
els.
T
h
e
im
p
lem
en
t
atio
n
was
ca
r
r
ied
o
u
t
in
Py
th
o
n
,
with
p
an
d
as
an
d
Scik
it
-
lear
n
lib
r
ar
ies
f
o
r
d
ata
p
r
ep
r
o
ce
s
s
in
g
an
d
ev
alu
atio
n
.
E
x
p
er
im
en
ts
wer
e
c
o
n
d
u
cted
o
n
a
co
m
p
u
te
r
eq
u
ip
p
e
d
with
an
NVI
DI
A
GP
U
an
d
C
UDA
s
u
p
p
o
r
t to
ac
ce
l
er
ate
tr
ain
in
g
.
T
h
e
m
o
d
el
co
n
f
ig
u
r
atio
n
s
an
d
th
eir
h
y
p
er
p
ar
am
eter
s
ar
e
s
et
as
f
o
llo
ws:
B
iLST
M
with
o
n
e
B
iLST
M
lay
er
with
1
2
8
h
id
d
en
u
n
its
,
r
eg
u
lar
f
o
r
g
et
g
ate,
d
r
o
p
o
u
t
0
.
3
,
R
MSp
r
o
p
o
p
tim
izer
with
f
i
x
ed
0
.
0
0
1
lear
n
in
g
r
ate,
an
d
MSE
lo
s
s
;
Stack
ed
B
iLST
M
u
s
es
1
2
8
u
n
its
p
e
r
l
ay
er
,
3
lay
er
s
o
f
B
iLST
M,
th
e
u
s
u
al
f
o
r
g
et
g
ate,
d
r
o
p
o
u
t
0
.
3
,
R
MSp
r
o
p
with
0
.
0
0
1
lear
n
i
n
g
r
ate.
T
h
e
p
r
o
p
o
s
ed
AFG
-
B
iLST
M
is
an
a
d
ap
tiv
ely
o
p
tim
ized
s
tack
ed
B
iLST
M
with
a
f
o
r
g
et
g
ate,
tu
n
ed
to
1
2
8
u
n
its
,
0
.
3
d
r
o
p
o
u
t,
f
in
e
-
tu
n
ed
with
th
e
A
d
am
o
p
tim
izer
,
an
d
tr
ain
ed
f
o
r
1
0
0
e
p
o
ch
s
.
A
7
0
-
3
0
tr
ai
n
-
test
s
p
lit
was
ad
o
p
ted
f
o
r
ea
ch
d
ataset,
with
th
e
f
ir
s
t
7
0
%
o
f
th
e
tim
e
-
s
er
ies o
b
s
er
v
atio
n
s
u
s
ed
f
o
r
tr
ain
in
g
an
d
th
e
r
em
ai
n
in
g
3
0
%
u
s
ed
f
o
r
test
in
g
.
T
h
e
m
ea
n
s
q
u
ar
e
d
e
r
r
o
r
(
MS
E
)
m
etr
ic
was
u
s
ed
to
co
m
p
a
r
e
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
m
o
d
els.
T
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
Ad
a
-
B
iLST
M
m
eth
o
d
with
AFG
was
s
t
u
d
ied
th
r
o
u
g
h
c
o
m
p
ar
is
o
n
a
g
a
in
s
t
f
o
u
r
d
atasets
a
s
d
is
cu
s
s
ed
in
s
ec
t
io
n
2
.
T
h
e
co
r
r
elatio
n
s
an
d
th
e
A
-
DT
W
d
is
t
an
ce
ar
e
ex
ten
s
iv
ely
s
tu
d
ied
ac
r
o
s
s
th
e
tar
g
et
an
d
th
e
f
ea
tu
r
e
v
a
r
iab
les
f
o
r
all
f
o
u
r
d
atasets
.
T
h
e
f
ea
tu
r
e
s
e
lectio
n
is
s
tu
d
ied
u
n
d
er
two
s
ce
n
ar
io
s
:
f
ea
tu
r
e
s
elec
tio
n
with
co
r
r
elatio
n
s
; f
e
atu
r
e
s
elec
tio
n
with
A
-
DT
W
d
is
tan
ce
.
T
h
e
ex
p
er
im
en
tal
r
esu
lts
o
f
co
r
r
elatio
n
s
an
d
A
-
DT
W
d
is
tan
ce
s
o
n
ea
c
h
d
ataset
ar
e
r
ec
o
r
d
ed
in
T
ab
le
2
h
ig
h
lig
h
ts
th
e
f
o
llo
wi
n
g
in
s
ig
h
ts
:
in
D
ataset
1
,
s
o
m
e
o
f
th
e
ess
en
tial
f
ea
tu
r
es
lik
e
‘
h
u
m
id
ity
’
,
‘
clo
u
d
co
v
er
’
,
‘
win
d
s
p
ee
d
’
,
a
n
d
‘
wi
n
d
d
ir
ec
tio
n
’
wer
e
o
b
s
er
v
e
d
t
o
b
e
v
e
r
y
wea
k
l
y
co
r
r
elate
d
w
ith
th
e
tar
g
et
p
o
wer
g
en
er
atio
n
a
n
d
a
r
e
n
o
t
a
m
o
n
g
th
e
to
p
-
5
s
tr
o
n
g
l
y
co
r
r
elate
d
f
ea
tu
r
es.
B
u
t
u
p
o
n
ap
p
l
y
in
g
A
-
DT
W
alig
n
m
en
t
,
th
ese
f
ea
tu
r
es
ar
e
in
th
e
T
o
p
-
5
,
h
ig
h
lig
h
tin
g
th
e
p
er
f
o
r
m
an
ce
o
f
o
u
r
n
o
v
el
f
ea
tu
r
e
s
elec
tio
n
ap
p
r
o
ac
h
.
W
h
er
ea
s
in
d
ataset
2
,
ess
en
tial
f
ea
tu
r
es
lik
e
‘
air
tem
p
er
a
tu
r
e’
,
‘
clo
u
d
co
v
er
’
,
‘
r
elativ
e
h
u
m
id
ity
’
,
‘
win
d
Evaluation Warning : The document was created with Spire.PDF for Python.