I
n
t
e
r
n
at
ion
al
Jou
r
n
al
of
E
lec
t
r
ical
an
d
Com
p
u
t
e
r
E
n
gin
e
e
r
in
g
(
I
JE
CE
)
Vol.
16
,
No.
5
,
Oc
tober
20
26
,
pp
.
2347
~
2356
I
S
S
N:
2088
-
8708
,
DO
I
:
10
.
11591/i
jec
e
.
v
16
i
5
.
pp
2
347
-
2356
2347
Jou
r
n
al
h
omepage
:
ht
tp:
//
ij
e
c
e
.
iaes
c
or
e
.
c
om
A
c
om
p
ar
at
iv
e
an
al
ysi
s of
h
y
b
r
id
F
FN
N
-
L
S
T
M
an
d
FF
N
N
-
RNN
ar
c
h
it
e
c
t
u
r
e
s f
o
r
sh
o
r
t
t
e
r
m
e
le
c
t
r
i
c
ity load
f
o
r
e
c
ast
in
g
T
e
m
it
op
e
Aki
n
ye
d
e
1
,
Jos
e
p
h
in
e
Adeni
k
e
Aki
n
y
e
d
e
2
,
P
au
l
Kehi
n
d
e
Ol
u
lop
e
3
,
E
m
m
an
u
e
l
T
aiwo
F
as
in
a
3
,
T
e
m
i
t
op
e
Adew
ale
Ol
om
in
u
4
1
D
e
pa
r
tm
e
nt
of
E
le
c
tr
ic
a
l
a
nd E
le
c
tr
oni
c
E
ngi
ne
e
r
in
g, E
ki
ti
S
ta
te
P
ol
yt
e
c
hni
c
,
I
s
a
n
-
E
ki
ti
, N
ig
e
r
ia
2
D
e
pa
r
tm
e
nt
of
E
le
c
tr
ic
a
l
a
nd E
le
c
tr
oni
c
E
ngi
ne
e
r
in
g, F
e
de
r
a
l
U
ni
ve
r
s
it
y of
T
e
c
hnol
ogy a
nd E
nvi
r
onme
nt
a
l
S
c
ie
nc
e
,
I
yi
n, N
ig
e
r
ia
3
D
e
pa
r
tm
e
nt
of
E
le
c
tr
ic
a
l
a
nd E
le
c
tr
oni
c
E
ngi
ne
e
r
in
g, E
ki
ti
S
ta
te
U
ni
ve
r
s
it
y A
do
-
E
ki
ti
,
A
do
-
E
ki
ti
, N
ig
e
r
ia
4
D
e
pa
r
tm
e
nt
of
S
ta
ti
s
ti
c
s
, F
e
de
r
a
l
U
ni
ve
r
s
it
y of
T
e
c
hnol
ogy Ak
ur
e
,
A
kur
e
, N
ig
e
r
ia
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
F
e
b
10,
2026
R
e
vis
e
d
Aug
11,
2026
Ac
c
e
pted
Aug
19,
2026
Sh
o
rt
-
t
erm
acc
u
rat
e
f
o
recas
t
i
n
g
o
f
el
ec
t
ri
c
i
t
y
d
eman
d
i
s
cru
c
i
al
f
o
r
p
o
w
er
-
s
y
s
t
em
o
p
erat
i
o
n
an
d
en
er
g
y
s
ch
e
d
u
l
i
n
g
a
n
d
t
h
e
e
q
u
i
l
i
b
r
i
u
m
b
e
t
w
e
en
el
ect
r
i
ci
t
y
g
en
era
t
i
o
n
a
n
d
c
o
n
s
u
m
p
t
i
o
n
.
T
h
e
p
red
i
ct
i
n
g
o
f
el
ec
t
ri
c
p
o
w
er
co
n
s
u
m
p
t
i
o
n
i
s
h
o
w
ev
er
d
i
ff
i
cu
l
t
b
eca
u
s
e
e
l
ect
r
i
ci
t
y
-
d
eman
d
p
ro
f
i
l
e
s
are
non
-
l
i
n
ear
an
d
t
i
me
v
ary
i
n
g
.
In
t
h
i
s
p
a
p
er,
w
e
co
n
s
i
d
er
t
w
o
h
y
b
r
i
d
d
ee
p
l
earn
i
n
g
arch
i
t
ect
u
res
feed
f
o
rw
ar
d
n
eu
r
al
n
e
t
w
o
rk
l
o
n
g
s
h
o
r
t
-
t
erm
memo
r
y
(FFN
N
-
L
ST
M)
an
d
feed
fo
r
w
ard
n
eu
ra
l
n
et
w
o
r
k
recu
r
ren
t
n
e
u
ral
n
e
t
w
o
rk
(FFN
N
-
RN
N
)
f
o
r
m
u
l
t
i
-
h
o
r
i
z
o
n
e
l
e
ct
ri
ci
t
y
-
l
o
a
d
f
o
r
eca
s
t
i
n
g
.
T
h
e
arc
h
i
t
ec
t
u
re
s
w
e
p
r
o
p
o
s
e
d
co
m
b
i
n
e
t
h
e
ab
i
l
i
t
y
o
f
FF
N
N
t
o
re
p
res
e
n
t
d
at
a
n
o
n
-
l
i
n
ear
l
y
w
i
t
h
t
h
e
s
e
q
u
e
n
t
i
al
m
o
d
e
l
l
i
n
g
cap
a
b
i
l
i
t
y
o
f
L
ST
M
an
d
RN
N
.
T
h
e
au
t
h
o
rs
as
s
e
s
s
t
h
e
mo
d
el
s
b
a
s
ed
o
n
h
i
s
t
o
ri
ca
l
el
ec
t
ri
c
i
t
y
-
l
o
a
d
o
b
s
erv
a
t
i
o
n
s
fro
m
A
d
o
-
E
k
i
t
i
,
a
t
t
h
ree
d
i
ffere
n
t
f
o
reca
s
t
i
n
g
h
o
ri
z
o
n
s
o
f
2
4
h
o
u
rs
,
7
2
h
o
u
rs
,
a
n
d
1
6
8
h
o
u
rs
.
Ro
o
t
mea
n
s
q
u
are
err
o
r
(RMSE
),
mean
a
b
s
o
l
u
t
e
erro
r
(MA
E
),
an
d
mea
n
ab
s
o
l
u
t
e
p
erce
n
t
a
g
e
erro
r
(MA
P
E
)
as
s
e
s
s
e
d
p
re
d
i
c
t
i
v
e
p
erfo
rma
n
ce.
T
h
e
fi
n
d
i
n
g
s
rev
eal
t
h
at
FFN
N
-
L
ST
M
ach
i
e
v
es
a
co
n
s
i
s
t
en
t
l
y
l
o
w
er
RMSE
acr
o
s
s
t
h
e
e
v
al
u
at
e
d
h
o
r
i
zo
n
s
,
re
s
u
l
t
i
n
g
i
n
g
reat
er
effec
t
i
v
en
e
s
s
t
o
c
o
n
f
i
n
e
rel
a
t
i
v
el
y
l
ar
g
e
f
o
recas
t
i
n
g
erro
r
s
.
O
n
t
h
e
o
t
h
er
h
an
d
,
t
h
e
s
ep
ara
t
e
L
ST
M
an
d
RN
N
mo
d
e
l
s
ac
h
i
e
v
e
l
o
w
er
MA
E
an
d
MA
PE
i
n
v
ari
o
u
s
s
cen
ar
i
o
s
,
s
u
g
g
es
t
i
n
g
s
t
ro
n
g
er
s
h
o
r
t
-
t
erm
react
i
o
n
t
o
el
e
ct
ri
c
i
t
y
d
eman
d
.
T
h
e
res
u
l
t
s
,
t
h
eref
o
re,
s
h
o
w
a
t
ra
d
e
-
o
f
f
b
et
w
een
f
o
recas
t
s
t
a
b
i
l
i
t
y
an
d
s
e
n
s
i
t
i
v
i
t
y
t
o
rap
i
d
l
o
a
d
ch
an
g
es
.
In
g
en
era
l
,
t
h
e
p
er
fo
rma
n
ce
o
f
h
y
b
r
i
d
arch
i
t
ec
t
u
re
i
s
mo
re
s
t
a
b
l
e
c
o
mp
are
d
t
o
t
h
e
s
t
an
d
al
o
n
e
rec
u
rren
t
mo
d
el
s
w
h
i
ch
are
mo
re
res
p
o
n
s
i
v
e
t
o
s
h
o
rt
-
t
erm
ch
a
n
g
es
w
h
en
w
e
l
o
o
k
at
m
o
s
t
o
f
t
h
e
f
o
recas
t
i
n
g
h
o
r
i
zo
n
s
.
A
cc
o
rd
i
n
g
t
o
t
h
e
res
earc
h
p
a
p
er,
t
h
e
fo
reca
s
t
i
n
g
arc
h
i
t
ect
u
res
ca
n
n
o
w
b
e
s
e
l
ect
e
d
fo
r
s
mar
t
g
ri
d
ap
p
l
i
cat
i
o
n
s
.
K
e
y
w
o
r
d
s
:
F
e
e
df
or
wa
r
d
ne
ur
a
l
ne
twor
k
L
ong
s
hor
t
-
ter
m
memor
y
M
e
a
n
a
bs
olut
e
e
r
r
or
M
e
a
n
a
bs
olut
e
pe
r
c
e
ntage
e
r
r
or
R
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
k
R
oot
mea
n
s
qua
r
e
e
r
r
or
S
hor
t
-
ter
m
load
f
or
e
c
a
s
ti
ng
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e.
C
or
r
e
s
pon
din
g
A
u
th
or
:
T
e
mi
tope
Akinye
de
De
pa
r
tm
e
nt
of
E
lec
tr
ica
l
a
nd
E
lec
tr
onic
E
nginee
r
i
ng,
E
k
it
i
S
tate
P
olyt
e
c
hnic
I
s
a
n
-
E
kit
i,
Nige
r
ia
E
mail:
takinye
de
@e
ks
poly.
e
du.
ng
1.
I
NT
RODU
C
T
I
ON
Ac
c
ur
a
te
f
or
e
c
a
s
ti
ng
of
e
lec
tr
icity
de
mand
p
lays
a
n
im
por
tant
r
ole
in
the
r
e
li
a
ble
a
nd
e
c
onomi
c
a
l
ope
r
a
ti
on
of
mode
r
n
powe
r
s
ys
tems
.
De
c
is
ions
r
e
l
a
ti
ng
to
ge
ne
r
a
ti
on
s
c
he
duli
ng,
r
e
s
e
r
ve
planning
,
e
lec
tr
icity
dis
tr
ibut
ion,
a
nd
de
mand
-
r
e
s
pons
e
pr
ogr
a
mm
e
d
de
pe
nd
c
ons
ider
a
bly
on
the
a
va
il
a
bil
it
y
o
f
de
pe
nda
ble
s
hor
t
ter
m
load
e
s
ti
mate
s
.
Ne
ve
r
thele
s
s
,
e
lec
tr
icity
c
ons
umpt
ion
is
dif
f
icult
to
p
r
e
dict
be
c
a
us
e
it
s
be
ha
vior
c
ha
nge
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2088
-
8708
I
nt
J
E
lec
&
C
omp
E
ng
,
Vol
.
16
,
No.
5
,
Oc
tober
20
26
:
2347
-
2356
2348
ove
r
ti
me
a
nd
is
in
f
luenc
e
d
by
int
e
r
a
c
ti
ng
f
a
c
tor
s
s
uc
h
a
s
we
a
ther
c
ondit
ions
,
ur
ba
n
de
ve
lopm
e
nt,
r
e
n
e
wa
ble
-
e
ne
r
gy
pe
ne
tr
a
ti
on,
e
lec
tr
ic
ve
hicle
us
a
ge
,
a
nd
the
incr
e
a
s
ing
de
ploym
e
nt
of
s
mar
t
gr
id
tec
hnologi
e
s
[
1]
–
[
8]
.
C
onve
nti
ona
l
s
tatis
ti
c
a
l
tec
hniques
,
including
r
e
gr
e
s
s
ion
a
nd
tr
a
dit
ional
ti
me
s
e
r
ies
models
,
r
e
main
us
e
f
ul
f
or
modell
ing
c
ompar
a
ti
ve
ly
s
im
ple
de
mand
s
tr
uc
tur
e
s
.
T
he
i
r
pe
r
f
or
manc
e
may,
howe
ve
r
,
de
ter
ior
a
te
whe
n
the
unde
r
lyi
ng
r
e
lations
hip
is
s
tr
ongly
nonli
ne
a
r
or
whe
n
c
ompl
e
x
tempor
a
l
int
e
r
a
c
ti
ons
a
r
e
pr
e
s
e
nt
[
9]
,
[
10]
.
T
his
li
mi
tation
ha
s
e
nc
our
a
ge
d
incr
e
a
s
ing
a
ppli
c
a
ti
on
of
mac
hine
-
lea
r
ning
a
nd
de
e
p
-
lea
r
ning
te
c
hniques
to
load
f
or
e
c
a
s
ti
ng.
I
n
pa
r
t
icula
r
,
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
ks
a
nd
long
s
hor
t
-
ter
m
memor
y
ne
twor
ks
a
r
e
we
ll
s
uit
e
d
to
s
e
que
nti
a
l
obs
e
r
va
ti
on
s
be
c
a
u
s
e
they
c
a
n
lea
r
n
r
e
lations
hips
that
e
volve
a
c
r
os
s
s
uc
c
e
s
s
ive
ti
me
s
teps
[
11]
–
[
15]
.
R
e
c
e
nt
r
e
s
e
a
r
c
h
ha
s
incr
e
a
s
ingl
y
c
ons
ider
e
d
hybr
id
a
r
c
hit
e
c
tur
e
s
in
whic
h
c
ompl
e
menta
r
y
pr
ope
r
ti
e
s
of
dif
f
e
r
e
nt
ne
ur
a
l
ne
two
r
k
s
tr
uc
tu
r
e
s
a
r
e
c
ombi
ne
d
withi
n
a
s
ingl
e
f
o
r
e
c
a
s
ti
ng
f
r
a
mew
or
k
.
I
n
s
uc
h
s
ys
tems
,
a
f
e
e
df
or
wa
r
d
ne
ur
a
l
ne
twor
k
(
F
F
NN
)
c
a
n
be
e
mpl
oye
d
to
t
r
a
ns
f
or
m
nonli
ne
a
r
input
r
e
lations
hips
,
while
r
e
c
ur
r
e
nt
laye
r
s
c
a
ptur
e
the
tempor
a
l
s
tr
uc
tu
r
e
e
mbedde
d
in
e
lec
tr
icity
-
de
mand
obs
e
r
va
ti
ons
.
P
r
e
vious
inves
ti
ga
ti
ons
ha
ve
r
e
por
ted
that
thi
s
type
of
int
e
gr
a
ti
on
c
a
n
i
mpr
ove
f
or
e
c
a
s
ti
ng
c
a
pa
bil
it
y
unde
r
dif
f
e
r
e
nt
ope
r
a
ti
ng
c
ondit
ions
a
nd
p
r
e
diction
hor
izons
[
1]
–
[
4]
,
[
16]
.
Ac
c
or
dingl
y,
thi
s
s
tudy
de
ve
lops
a
nd
c
ompa
r
e
s
F
F
NN
long
s
hor
t
-
ter
m
memor
y
(
F
F
NN
-
L
S
T
M
)
a
nd
F
F
NN
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
k
(
F
F
NN
-
R
NN
)
a
r
c
hit
e
c
tur
e
s
f
or
s
hor
t
-
ter
m
e
lec
tr
icity
-
load
f
or
e
c
a
s
t
ing.
T
he
models
a
r
e
e
va
luate
d
us
ing
the
s
a
me
his
tor
ica
l
da
t
a
s
e
t
a
nd
a
r
e
e
xa
mi
ne
d
a
t
f
o
r
e
c
a
s
ti
ng
hor
izons
of
2
4
hour
s
,
72
hour
s
,
a
nd
168
hou
r
s
.
T
he
i
r
pr
e
dictive
pe
r
f
or
m
a
nc
e
s
a
r
e
c
ompar
e
d
us
ing
R
oot
mea
n
s
qua
r
e
e
r
r
or
(
R
M
S
E
)
,
mea
n
a
bs
olut
e
e
r
r
or
(
M
AE
)
,
a
nd
mea
n
a
bs
olut
e
pe
r
c
e
ntage
e
r
r
or
(
M
APE
)
,
ther
e
by
pr
ovidi
ng
a
c
ons
is
tent
ba
s
is
f
or
e
xa
mi
ning
how
the
two
hyb
r
id
a
r
c
hit
e
c
tur
e
s
be
ha
ve
a
s
the
f
or
e
c
a
s
ti
ng
hor
izon
incr
e
a
s
e
s
.
E
v
e
n
th
ou
gh
h
yb
r
id
ne
u
r
a
l
a
r
c
h
i
tec
tu
r
e
s
a
n
d
tec
hni
q
ue
s
f
o
r
de
e
p
lea
r
ni
ng
ha
ve
in
c
r
e
a
s
in
gl
y
b
e
e
n
i
n
u
s
e
f
o
r
th
e
s
ho
r
t
-
te
r
m
f
or
e
c
a
s
t
in
g
o
f
e
le
c
t
r
ic
it
y
lo
a
d
,
the
r
e
a
r
e
s
ti
l
l
s
om
e
li
m
it
a
t
io
ns
in
t
he
wa
y
t
he
y
e
va
lu
a
te
t
he
i
r
pe
r
f
o
r
ma
nc
e
.
M
a
ny
o
the
r
s
t
ud
ies
f
oc
us
on
s
p
e
c
i
f
i
c
a
r
c
h
it
e
c
tu
r
e
o
r
e
v
a
l
ua
te
f
o
r
e
c
a
s
t
in
g
pe
r
f
o
r
manc
e
a
t
on
e
p
r
e
d
ic
ti
on
h
or
iz
on
o
nl
y
.
D
ue
to
d
i
f
f
e
r
e
n
c
e
s
in
da
tas
e
ts
,
p
r
e
p
r
oc
e
s
s
i
ng
,
m
ode
l
c
o
nf
i
gu
r
a
ti
on
s
a
nd
e
va
lu
a
t
io
n
,
r
e
po
r
ted
r
e
s
ul
ts
a
r
e
a
ls
o
n
ot
d
i
r
e
c
tl
y
c
o
mp
a
r
a
b
le.
As
a
r
e
s
u
lt
,
t
he
r
e
is
s
ti
l
l
a
r
e
qu
i
r
e
men
t
f
o
r
c
o
n
tr
o
ll
e
d
c
om
pa
r
is
o
ns
i
n
w
hi
c
h
a
l
te
r
n
a
t
iv
e
h
yb
r
id
a
r
c
h
it
e
c
tu
r
e
s
a
r
e
t
e
s
t
e
d
un
de
r
c
om
pa
r
a
bl
e
e
x
pe
r
im
e
n
ta
l
c
o
nd
i
ti
o
ns
.
T
he
c
ur
r
e
nt
wo
r
k
f
il
ls
th
is
ga
p
by
pe
r
f
or
mi
n
g
a
s
ys
tema
ti
c
c
ompar
is
on
of
the
two
hyb
r
id
a
r
c
hit
e
c
tur
e
s
F
F
NN
-
L
S
T
M
a
nd
F
F
NN
-
R
NN
f
or
m
ult
i
-
hor
izon
e
lec
tr
icity
load
f
or
e
c
a
s
ti
ng.
B
oth
a
r
c
hi
tec
tur
e
s
ha
ve
be
e
n
de
ve
loped
us
ing
the
s
a
me
da
tas
e
t,
in
put
va
r
iable
s
,
pr
e
pr
oc
e
s
s
ing
p
r
oc
e
dur
e
,
da
ta
pa
r
t
it
ioni
ng
s
tr
a
tegy,
a
nd
e
va
luation
f
r
a
mew
or
k.
T
he
e
xplo
r
a
t
ion
of
their
f
or
e
c
a
s
ti
ng
be
ha
vior
is
done
us
ing
a
R
M
S
E
,
M
AE
,
M
APE
,
a
nd
r
e
s
idual
be
ha
vior
a
t
thr
e
e
hor
i
z
ons
na
mely,
24
-
hour
s
,
72
-
hour
s
a
nd
168
-
hour
s
ha
ving
a
c
omm
on
e
xpe
r
im
e
ntation
f
r
a
mew
or
k
make
s
it
pos
s
ibl
e
to
a
s
s
e
s
s
dif
f
e
r
e
nc
e
s
in
pr
e
dictive
a
c
c
ur
a
c
y,
s
tabili
ty,
a
nd
r
e
s
pons
ivene
s
s
without
c
onf
ounding
va
r
iation
s
due
to
dif
f
e
r
e
nt
da
tas
e
ts
a
nd
modelli
ng
pr
oc
e
dur
e
s
.
2.
RE
L
AT
E
D
WORKS
S
hor
t
-
ter
m
load
f
or
e
c
a
s
ti
ng
(
S
T
L
F
)
is
a
n
e
s
s
e
nti
a
l
a
s
pe
c
t
of
powe
r
s
ys
tem
ope
r
a
ti
on
in
a
idi
ng
e
c
onomi
c
a
l
dis
pa
tch,
de
mand
r
e
s
pons
e
a
nd
g
r
id
r
e
li
a
bil
it
y
.
T
r
a
dit
ional
f
o
r
e
c
a
s
ti
ng
methods
li
ke
the
a
utor
e
gr
e
s
s
ive
int
e
gr
a
ted
movi
ng
a
ve
r
a
ge
(
AR
I
M
A
)
,
e
xpone
nti
a
l
s
moot
hing
a
nd
r
e
gr
e
s
s
ion
a
na
lys
is
a
r
e
a
ble
to
pe
r
f
or
m
we
ll
f
or
li
ne
a
r
load,
but
ha
ve
dif
f
icult
y
c
a
ptur
ing
the
non
-
li
ne
a
r
a
nd
non
-
s
tationar
it
y
f
e
a
tur
e
s
of
the
e
lec
tr
icity
de
mand
[
9]
,
[
10]
.
Ac
c
or
dingl
y,
the
de
e
p
lea
r
ning
method
ha
s
ga
ined
popular
it
y
a
s
it
a
utom
a
ti
c
a
ll
y
lea
r
ns
c
ompl
e
x
tempor
a
l
pa
tt
e
r
ns
f
r
o
m
the
his
tor
ica
l
load
da
ta.
F
F
NN
s
e
f
f
icie
ntl
y
model
nonli
ne
a
r
r
e
lations
hips
but
do
not
e
ntail
t
im
e
de
pe
nde
nc
ies
.
On
the
other
ha
nd,
R
NN
s
a
nd
L
S
T
M
s
a
r
e
f
or
s
e
que
nti
a
l
lea
r
ning
a
nd
the
us
e
of
L
S
T
M
s
ove
r
c
omes
the
va
nis
hing
gr
a
dient
pr
oblem
to
make
be
tt
e
r
f
or
e
c
a
s
ts
a
t
dif
f
e
r
e
nt
pr
e
diction
ho
r
izons
whic
h
wor
ks
we
ll
f
or
l
ong
-
ter
m
de
pe
nde
nc
ies
.
T
he
a
c
c
ur
a
c
y
of
f
or
e
c
a
s
ts
of
hyb
r
id
de
e
p
lea
r
ning
models
that
include
a
n
a
tt
e
nti
on
c
o
mponent
is
f
ur
ther
im
pr
ove
d
due
to
r
e
c
e
nt
s
tudi
e
s
.
I
n
r
e
c
e
nt
ye
a
r
s
,
hybr
id
a
r
c
hi
tec
tur
e
s
unit
ing
F
F
NN
s
’
nonli
ne
a
r
f
e
a
tur
e
e
xtr
a
c
ti
on
c
a
pa
bil
it
y
with
the
tempor
a
l
lea
r
ning
c
a
pa
bil
it
y
o
f
r
e
c
ur
r
e
nt
ne
twor
k
s
ha
ve
r
e
c
e
ived
incr
e
a
s
e
d
a
tt
e
nti
on.
T
he
incor
por
a
ti
on
of
his
tor
ica
l
loads
,
we
a
ther
va
r
iable
s
,
a
nd
c
a
lenda
r
inf
or
mation
e
nha
nc
e
s
f
or
e
c
a
s
t
r
obus
tnes
s
a
nd
a
c
c
ur
a
c
y
a
c
r
os
s
va
r
ious
hor
izons
in
the
model
[
2]
,
[
4]
,
[
17]
,
[
18
]
.
Ac
c
or
ding
to
[
1]
–
[
3]
,
[
16]
,
R
M
S
E
,
M
AE
a
n
d
M
APE
a
r
e
typi
c
a
ll
y
us
e
d
to
a
s
s
e
s
s
the
pe
r
f
or
manc
e
with
dif
f
e
r
e
nt
mea
s
ur
e
s
of
a
c
c
ur
a
c
y
a
nd
r
obus
tnes
s
of
f
or
e
c
a
s
t.
Additi
ona
l
c
ompar
a
ti
ve
s
tudi
e
s
de
mons
tr
a
te
that
hybr
id
models
ge
ne
r
a
ll
y
pe
r
f
o
r
m
be
tt
e
r
than
s
tatis
ti
c
a
l
models
a
nd
s
tanda
lone
de
e
p
lea
r
ning
models
in
ter
ms
of
pr
e
diction
a
c
c
ur
a
c
y
a
nd
ge
ne
r
a
li
z
a
ti
on
a
bil
it
y
[
1]
,
[
3]
,
[
19]
,
[
20
]
.
Ne
ve
r
thele
s
s
,
the
c
ompr
e
he
ns
ive
c
ompar
is
on
of
F
F
NN
-
L
S
T
M
a
nd
F
F
NN
-
R
NN
a
r
c
hit
e
c
tur
e
s
unde
r
the
s
a
me
e
xpe
r
im
e
ntal
c
ondit
io
ns
a
nd
in
va
r
ious
f
o
r
e
c
a
s
t
hor
izons
wa
s
s
c
a
r
c
e
.
T
h
i
s
ga
p
is
f
il
led
by
s
ys
tema
ti
c
a
ll
y
inves
ti
ga
ti
ng
both
hyb
r
id
models
us
ing
identica
l
p
r
e
pr
oc
e
s
s
ing,
tr
a
ini
ng,
a
nd
e
va
luation
f
r
a
mew
or
ks
in
thi
s
s
tudy.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
E
lec
&
C
omp
E
ng
I
S
S
N:
2088
-
8708
A
c
ompar
ati
v
e
analys
is
of
hy
br
id
F
F
N
N
-
L
ST
M
an
d
F
F
N
N
-
R
N
N
ar
c
hit
e
c
tur
e
s
…
(
T
e
mitope
A
k
inye
de
)
2349
2
.
1.
M
e
t
h
od
ology
T
he
c
ur
r
e
nt
s
tudy
de
ve
lops
a
nd
c
ompar
e
s
F
F
NN
-
L
S
T
M
a
nd
F
F
NN
-
R
NN
models
f
or
s
hor
t
ter
m
E
L
D
f
or
e
c
a
s
ti
ng.
Dur
ing
de
ploym
e
nt,
the
model
will
c
onduc
t
da
ta
c
oll
e
c
ti
on,
pr
e
pr
oc
e
s
s
ing
c
he
c
kpoint
s
,
model
de
ve
lopm
e
nt,
a
nd
r
e
por
t
tr
a
ini
ng
a
nd
pe
r
f
o
r
manc
e
output
.
T
he
M
AT
L
AB
de
e
p
lea
r
ning
tool
box
wa
s
us
e
d
to
r
e
a
li
z
e
the
models
.
2
.
2
.
Dat
a
c
oll
e
c
t
ion
an
d
p
r
e
p
r
oc
e
s
s
in
g
T
he
e
lec
tr
icity
dis
tr
ibu
ti
on
f
ir
m
ba
s
e
d
in
Ado
-
E
kit
i
,
Nige
r
ia
p
r
ovided
the
hour
ly
load
da
ta.
T
he
input
f
e
a
tur
e
s
c
ons
is
ted
of
pr
e
vious
hour
’
s
load,
p
r
e
vious
da
y’
s
load
,
p
r
e
vious
we
e
k’
s
load,
a
mbi
e
nt
tem
pe
r
a
tur
e
,
hour
of
da
y,
a
nd
da
y
of
we
e
k.
T
he
hou
r
ly
e
lec
tr
icit
y
load
is
the
tar
ge
t
va
r
iable
.
I
nter
polate
d
mi
s
s
ing
obs
e
r
va
ti
ons
a
nd
c
he
c
ke
d
a
bnor
mal
va
lues
be
f
or
e
model
tr
a
ini
ng
wa
s
done
.
Nor
maliza
ti
on
of
c
onti
nuous
va
r
iable
s
wa
s
pe
r
f
or
med
with
M
in
-
M
a
x
s
c
a
le:
=
−
−
T
he
obs
e
r
va
ti
ons
we
r
e
or
ga
nize
d
in
c
hr
onolog
ica
l
or
de
r
a
nd
we
r
e
a
r
r
a
nge
d
int
o
th
r
e
e
s
e
ts
:
tr
a
ini
ng,
va
li
da
ti
on,
a
nd
tes
ti
ng
in
the
pr
opor
ti
on
70%
:
15%
:
15%
.
2
.
3
.
S
t
r
u
c
t
u
r
e
of
t
h
e
m
od
e
l
T
he
F
F
NN
-
L
S
T
M
model
is
c
ompos
e
d
o
f
:
a
s
e
q
ue
nti
a
l
input
laye
r
,
a
f
ull
y
c
onne
c
ted
laye
r
with
50
ne
ur
ons
,
a
R
e
L
U
a
c
ti
va
ti
on
laye
r
,
a
n
L
S
T
M
laye
r
with
100
hidden
unit
s
,
a
f
u
ll
y
c
onne
c
ted
outp
ut
laye
r
,
a
nd
a
r
e
gr
e
s
s
ion
out
put.
T
he
F
F
NN
-
R
NN
model
ha
s
the
s
a
me
f
e
e
df
o
r
wa
r
d
f
e
a
tur
e
-
e
xtr
a
c
ti
on
s
tr
uc
tur
e
,
but
the
L
S
T
M
laye
r
is
s
ubs
ti
tut
e
d
with
a
s
im
ple
r
e
c
ur
r
e
nt
laye
r
ha
ving
100
hidden
unit
s
.
L
a
ye
r
s
that
a
r
e
r
e
c
ur
r
e
nt
in
na
tur
e
maintain
pa
s
t
ti
me
s
tep
inf
or
mation
with
the
idea
o
f
lea
r
ning
s
e
que
nti
a
l
load
pa
tt
e
r
ns
.
F
igu
r
e
1
the
s
hows
wor
kf
low
of
the
hybr
id
F
F
NN
-
L
S
T
M
a
nd
F
F
NN
-
R
NN
s
hor
t
-
ter
m
load
f
or
e
c
a
s
ti
ng
f
r
a
mew
or
k
F
igur
e
1
.
W
or
kf
low
of
the
hyb
r
id
F
F
NN
-
L
S
T
M
a
nd
F
F
NN
-
R
NN
s
hor
t
-
ter
m
load
f
or
e
c
a
s
ti
ng
f
r
a
mew
or
k
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2088
-
8708
I
nt
J
E
lec
&
C
omp
E
ng
,
Vol
.
16
,
No.
5
,
Oc
tober
20
26
:
2347
-
2356
2350
2
.
4
.
M
od
e
l
t
r
ain
in
g
T
he
Ada
m
opti
m
ize
r
c
ombi
ne
d
wi
th
mea
n
s
qua
r
e
e
r
r
o
r
wa
s
the
los
s
f
unc
ti
on
us
e
d
to
tr
a
in
the
models
.
T
a
ble
1
pr
e
s
e
nts
the
model
a
r
c
hit
e
c
tur
e
a
n
d
tr
a
ini
ng
pa
r
a
mete
r
s
.
T
he
va
li
da
ti
on
s
e
t
is
us
e
d
to
ke
e
p
a
n
e
ye
on
the
model’
s
pe
r
f
o
r
manc
e
while
the
tes
ti
ng
s
e
t
is
lef
t
a
s
ide
f
or
f
inal
e
va
luation.
T
o
e
ns
ur
e
f
a
ir
ne
s
s
,
a
ll
models
we
r
e
ha
ndled
identica
ll
y
in
ter
ms
of
da
tas
e
t
pa
r
ti
ti
oning
a
nd
t
r
a
ini
ng
s
e
tt
ings
.
T
a
ble
1.
M
ode
l
a
r
c
hit
e
c
tur
e
a
nd
tr
a
ini
ng
pa
r
a
mete
r
s
P
a
r
a
me
te
r
V
a
lu
e
I
nput
f
e
a
tu
r
e
s
6
F
F
N
N
ne
ur
ons
50
L
S
T
M
/R
N
N
hi
dde
n unit
s
100
O
ut
put
uni
ts
1
O
pt
im
iz
e
r
A
da
m
M
a
xi
mum
e
poc
hs
250
M
in
i
-
ba
tc
h s
iz
e
16
I
ni
ti
a
l
le
a
r
ni
ng r
a
te
0.005
G
r
a
di
e
nt
t
hr
e
s
hol
d
1
L
e
a
r
ni
ng r
a
te
dr
op pe
r
io
d
125
L
e
a
r
ni
ng r
a
te
dr
op f
a
c
to
r
0.2
2
.
5
.
E
valu
at
io
n
of
t
h
e
p
e
r
f
or
m
an
c
e
T
he
r
e
wa
s
a
de
ter
mi
na
ti
on
of
f
or
e
c
a
s
t
pe
r
f
o
r
ma
nc
e
f
or
the
ne
xt
da
y
(
24
hou
r
s
)
,
3
da
ys
a
he
a
d
(
72
hour
s
)
a
nd
one
we
e
k
a
he
a
d
(
168
hour
s
)
us
in
g
r
oot
mea
n
s
qua
r
e
e
r
r
o
r
,
mea
n
a
bs
olut
e
e
r
r
or
a
nd
mea
n
a
bs
olut
e
pe
r
c
e
ntage
e
r
r
or
.
˗
T
he
M
AE
is
c
a
lcula
ted
a
s
=
1
∑
|
−
̂
|
=
1
˗
T
he
R
M
S
E
wa
s
c
a
lcula
ted
a
s
=
√
1
∑
(
−
̂
)
2
=
1
.
˗
T
he
M
APE
wa
s
c
a
lcula
ted
a
s
=
100
∑
|
−
̂
|
=
1
.
whe
r
e
a
nd
̂
de
note
the
mea
s
ur
e
d
a
nd
c
a
lcula
ted
e
lec
tr
ic
loads
r
e
s
pe
c
ti
ve
ly.
T
he
lowe
r
the
metr
ic
va
lue,
the
mor
e
a
c
c
ur
a
te
the
f
or
e
c
a
s
t
wil
l
be
.
T
he
inves
ti
ga
ti
on
a
ls
o
include
d
a
n
a
na
lys
is
o
f
f
or
e
c
a
s
t
p
r
of
il
e
s
a
nd
their
r
e
s
idual
be
ha
vior
f
or
pr
e
diction
bias
e
s
.
3.
RE
S
UL
T
S
T
he
L
S
T
M
,
R
NN
,
F
F
NN
-
L
S
T
M
a
nd
F
F
NN
-
R
N
N
models
’
f
or
e
c
a
s
ti
ng
pe
r
f
o
r
manc
e
wa
s
e
va
luate
d
us
ing
R
M
S
E
,
M
AE
,
a
nd
M
APE
ove
r
th
r
e
e
f
or
e
c
a
s
ti
ng
hor
izons
:
24
h,
72
h
,
a
nd
168
h.
R
e
s
idual
a
na
lys
is
a
ls
o
s
ugge
s
ted
that
mor
e
a
c
c
ur
a
te
f
o
r
e
c
a
s
ti
ng
models
s
howe
d
s
maller
a
nd
r
a
ndoml
y
s
c
a
tt
e
r
e
d
r
e
s
iduals
.
As
s
hown
in
F
igur
e
s
2
to
4
,
the
r
e
s
ult
s
of
the
f
o
r
e
c
a
s
ti
ng
a
nd
the
numer
ica
l
pe
r
f
or
manc
e
s
.
T
hr
ough
the
e
nti
r
e
f
or
e
c
a
s
ti
ng
ho
r
izon,
the
F
F
N
N
-
L
S
T
M
model
yielde
d
lowe
r
R
M
S
E
f
or
a
ll
the
pe
r
iods
c
ompar
e
d
to
the
L
S
T
M
model.
L
S
T
M
yiel
ds
lowe
r
M
APE
va
lues
f
or
24
-
hour
,
72
-
hour
a
nd
1
68
-
hour
f
or
e
c
a
s
ti
ng.
M
AE
pe
r
f
or
manc
e
va
r
ies
be
twe
e
n
the
two
models
de
pe
nding
on
the
f
or
e
c
a
s
ti
ng
hor
izon.
Ac
c
or
ding
to
the
ove
r
a
ll
r
e
s
ult
s
,
the
hybr
id
F
F
N
N
-
L
S
T
M
model
de
mons
tr
a
tes
mor
e
c
ons
is
tent
f
o
r
e
c
a
s
ti
ng
pe
r
f
or
manc
e
,
while
in
s
ome
c
a
s
e
s
,
the
s
tanda
lone
L
S
T
M
pe
r
f
or
ms
be
tt
e
r
in
ter
ms
of
pe
r
c
e
ntage
e
r
r
o
r
.
F
igur
e
5
c
ompar
e
s
the
24
-
hour
load
f
or
e
c
a
s
ti
ng
pe
r
f
or
manc
e
o
f
the
L
S
T
M
a
nd
hybr
id
F
F
NN
-
L
S
T
M
models
with
the
a
c
tual
load
.
B
oth
models
c
a
ptur
e
the
ove
r
a
ll
pa
tt
e
r
n
in
the
load;
howe
ve
r
,
L
S
T
M
mor
e
c
los
e
ly
f
oll
ows
the
quick
f
luctua
ti
ons
,
while
F
F
NN
-
L
S
T
M
pr
oduc
e
s
s
moot
he
r
pr
e
dictions
.
E
v
e
n
if
a
hybr
id
model
pos
s
e
s
s
e
s
s
upe
r
ior
pr
e
dictive
s
tab
il
it
y,
the
L
S
T
M
is
mor
e
r
e
s
pons
ive
dur
ing
s
udde
n
load
f
luctua
ti
ons
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
E
lec
&
C
omp
E
ng
I
S
S
N:
2088
-
8708
A
c
ompar
ati
v
e
analys
is
of
hy
br
id
F
F
N
N
-
L
ST
M
an
d
F
F
N
N
-
R
N
N
ar
c
hit
e
c
tur
e
s
…
(
T
e
mitope
A
k
inye
de
)
2351
F
igur
e
2
.
C
ompar
a
ti
ve
a
na
lys
is
of
R
M
S
E
f
o
r
L
S
T
M
a
nd
F
F
NN
-
L
S
T
M
F
igur
e
3
.
C
ompar
a
ti
ve
a
na
lys
is
of
M
AE
f
o
r
L
S
T
M
a
nd
F
F
NN
-
L
S
T
M
F
igur
e
4
.
C
ompar
a
ti
ve
a
na
lys
is
of
M
AP
E
f
o
r
L
S
T
M
a
nd
F
F
NN
-
L
S
T
M
F
igur
e
5
.
P
r
o
f
il
e
f
or
hybr
id
F
F
NN
-
L
S
T
M
a
nd
L
S
T
M
f
or
da
y
a
he
a
d
L
S
T
M
a
nd
F
F
NN
-
L
S
T
M
hybr
id
model
a
c
c
ur
a
c
y
ve
r
if
ica
ti
on
f
o
r
72
-
hour
f
or
e
c
a
s
ts
is
c
a
r
r
ied
ou
t
in
F
igur
e
6
a
ga
ins
t
a
c
tual
load.
B
oth
models
r
e
f
le
c
t
the
de
mand
tr
e
nd;
howe
ve
r
,
L
S
T
M
f
oll
ows
s
hor
t
-
r
un
f
luctua
ti
ons
mor
e
a
c
c
ur
a
tely
while
F
F
NN
-
L
S
T
M
ge
ne
r
a
tes
s
moot
he
r
pr
e
dictions
,
r
e
s
ult
ing
i
n
lowe
r
s
e
ns
it
ivi
ty
to
s
udde
n
load
f
luctua
ti
ons
.
As
pe
r
the
r
e
s
ult
,
the
L
S
T
M
is
mo
r
e
a
da
ptable
to
dyna
mi
c
v
a
r
iations
while
the
hybr
id
F
F
NN
-
L
S
T
M
is
mor
e
s
table
but
l
e
s
s
dyna
mi
c
.
T
he
a
c
tual
load
va
lue
a
nd
the
168
-
hour
(
pe
r
we
e
k)
f
or
e
c
a
s
ted
va
lue
of
the
L
S
T
M
model
a
nd
F
F
NN
-
L
S
T
M
hybr
id
is
be
ing
c
ompar
e
d
in
F
igur
e
7
.
B
oth
models
f
oll
ow
the
ove
r
a
ll
t
r
e
nd
in
de
mand
ye
t
s
m
ooth
out
the
s
ha
r
p
f
luctua
ti
ons
obs
e
r
ve
d
in
a
c
tual
de
mand.
T
he
obs
e
r
ve
d
load
ha
s
a
c
los
e
r
a
gr
e
e
ment
with
a
n
L
S
T
M
model
than
a
F
F
NN
-
L
S
T
M
model.
T
his
is
e
s
pe
c
ially
tr
ue
whe
n
the
c
ha
nge
is
g
r
a
dua
l.
T
he
F
F
N
N
model
unde
r
e
s
ti
mate
s
the
pe
a
k
de
mand
due
to
the
lac
k
o
f
a
tt
e
nti
on
to
the
load
pr
o
f
il
e
.
L
S
T
M
model
s
hows
a
be
tt
e
r
tr
e
nd
r
e
pr
e
s
e
ntation
but
both
model
s
hows
les
s
c
a
p
a
bil
it
y
to
r
e
s
pons
e
s
udde
n
load
c
ha
nge
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2088
-
8708
I
nt
J
E
lec
&
C
omp
E
ng
,
Vol
.
16
,
No.
5
,
Oc
tober
20
26
:
2347
-
2356
2352
M
ode
ls
a
nd
r
e
s
ult
s
of
two
dif
f
e
r
e
nt
ne
ur
a
l
ne
twor
ks
.
T
he
R
NN
model
a
nd
the
hybr
id
F
F
NN
-
R
NN
model
yield
the
da
y
-
a
he
a
d
(
24
-
hour
)
,
thr
e
e
-
da
y
-
a
he
a
d
(
72
-
hour
)
,
a
s
we
ll
a
s
one
-
we
e
k
a
he
a
d
(
1
68
-
hour
)
f
or
e
c
a
s
ts
a
s
r
e
por
ted
in
F
igur
e
s
8
to
10
.
I
n
ge
ne
r
a
l
,
the
hybr
id
F
F
NN
-
R
NN
did
pos
s
e
s
s
lowe
r
R
M
S
E
f
or
both
24
-
hour
a
nd
168
-
hour
f
or
e
c
a
s
ts
.
I
n
c
ontr
a
s
t,
the
R
NN
pr
oduc
e
d
the
lowe
s
t
R
M
S
E
f
or
the
72
-
hour
f
or
e
c
a
s
ts
.
T
he
R
NN
c
ons
is
tently
p
r
oduc
e
s
lowe
r
M
AE
a
nd
M
APE
a
c
r
os
s
a
ll
thr
e
e
f
or
e
c
a
s
t
hor
izons
,
indi
c
a
ti
ng
be
tt
e
r
pe
r
c
e
ntage
e
r
r
or
pe
r
f
o
r
manc
e
.
T
he
hyb
r
id
F
F
NN
-
R
NN
(
i.
e
.
R
NN
with
the
F
F
NN
w
r
a
ppe
r
)
p
r
e
dicte
d
be
tt
e
r
ba
s
e
d
on
R
M
S
E
a
t
the
s
e
lec
ted
f
or
e
c
a
s
t
hor
izon.
How
e
ve
r
,
the
s
tanda
lone
R
NN
p
r
e
dicte
d
with
mor
e
s
tabili
ty
ba
s
e
d
on
M
AE
a
nd
M
APE
.
F
igur
e
6
.
P
r
o
f
il
e
f
or
hybr
id
F
F
NN
-
L
S
T
M
a
nd
L
S
T
M
f
or
th
r
e
e
-
da
ys
a
he
a
d
F
igur
e
7
.
P
r
o
f
il
e
f
or
hybr
id
F
F
NN
-
L
S
T
M
a
nd
L
S
T
M
f
or
one
we
e
k
a
he
a
d
F
igur
e
8
.
C
ompar
a
ti
ve
a
na
lys
is
of
R
M
S
E
f
o
r
R
NN
a
nd
F
F
NN
-
R
NN
F
igur
e
9
.
C
ompar
a
ti
ve
a
na
lys
is
of
M
AE
f
o
r
R
NN
a
nd
F
F
NN
-
R
NN
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
E
lec
&
C
omp
E
ng
I
S
S
N:
2088
-
8708
A
c
ompar
ati
v
e
analys
is
of
hy
br
id
F
F
N
N
-
L
ST
M
an
d
F
F
N
N
-
R
N
N
ar
c
hit
e
c
tur
e
s
…
(
T
e
mitope
A
k
inye
de
)
2353
F
igur
e
10
.
C
ompar
a
ti
ve
a
na
lys
is
of
M
AP
E
f
o
r
R
N
N
a
nd
F
F
NN
-
R
NN
An
R
NN
model
a
nd
a
n
F
F
NN
-
R
NN
hybr
id
we
r
e
c
ompar
e
d
a
ga
ins
t
the
a
c
tual
load
f
or
24
-
hour
f
or
e
c
a
s
ti
ng
a
s
s
hown
in
F
igu
r
e
1
1
.
B
oth
models
di
s
play
the
ge
ne
r
a
l
de
mand
t
r
e
nd
bu
t
da
mpen
os
c
il
l
a
ti
ons
in
the
load.
T
he
R
NN
f
oll
ows
the
load
pr
o
f
il
e
be
tt
e
r
,
whe
r
e
a
s
the
F
F
NN
-
R
NN
pr
oduc
e
s
s
moot
he
ne
d
va
lues
,
be
ing
les
s
s
e
n
s
it
ive
to
c
ha
nge
s
.
Ove
r
a
ll
,
the
R
NN
is
mor
e
a
li
gne
d
with
the
a
c
tual
load
a
nd
pe
r
f
or
ms
b
e
tt
e
r
f
or
ne
xt
-
da
y
f
or
e
c
a
s
ti
ng.
F
igur
e
1
1
.
P
r
o
f
il
e
f
or
hybr
id
F
F
NN
-
R
NN
a
nd
R
NN
f
or
thr
e
e
-
da
ys
a
he
a
d
3
.1
.
Dis
c
u
s
s
ion
T
he
s
tanda
lone
a
nd
hyb
r
id
de
e
p
lea
r
ning
models
we
r
e
us
e
f
ul
tool
s
f
or
s
hor
t
-
ter
m
e
lec
tr
icity
load
f
or
e
c
a
s
ti
ng
a
c
c
or
ding
to
the
c
ompar
a
ti
ve
a
na
lys
is
,
though
the
pe
r
f
or
manc
e
va
r
ied
a
long
the
f
or
e
c
a
s
t
hor
izon.
Out
of
thr
e
e
methods
a
ppli
e
d
on
wind
s
pe
e
d
da
t
a
s
e
t,
F
F
NN
-
L
S
T
M
hybr
id
a
r
ous
e
d
lea
s
t
R
M
S
E
r
e
f
lec
ti
ng
ove
r
a
ll
highes
t
pr
e
diction
a
c
c
ur
a
c
y.
B
ut
L
S
T
M
pr
oduc
e
d
les
s
M
AE
a
nd
M
APE
s
ugge
s
ti
ng
be
tt
e
r
point
-
wis
e
a
nd
pe
r
c
e
ntage
-
wi
s
e
e
r
r
or
pr
e
diction
a
c
c
ur
a
c
y.
I
n
a
s
im
il
a
r
wa
y,
f
o
r
24
hour
a
nd
168
hour
f
or
e
c
a
s
ts
,
hybr
id
F
F
NN
-
R
NN
a
c
hieve
s
lowe
r
R
M
S
E
than
the
s
tan
da
lone
R
NN
.
M
AE
a
nd
M
APE
s
how
R
NN
c
onti
nuous
ly
outper
f
or
mi
ng
the
hyb
r
id.
Upon
a
na
lyzing
the
r
e
s
iduals
,
it
wa
s
dis
c
ove
r
e
d
that
the
hybr
id
mo
de
l
wa
s
s
moot
he
r
with
les
s
tur
bulenc
e
,
while
the
r
e
pe
a
t
m
ode
l
c
a
ptur
e
d
load
c
ha
nge
s
be
tt
e
r
.
As
a
whole
,
th
e
f
indi
ng
he
r
e
s
ugge
s
ts
that
ther
e
a
r
e
s
tabili
ty
a
nd
r
e
s
pons
ivene
s
s
tr
a
de
-
of
f
s
a
s
s
oc
iate
d
with
pr
e
diction
a
c
c
ur
a
c
y.
I
n
f
a
c
t,
hybr
id
a
r
c
hit
e
c
tur
e
s
a
r
e
de
ter
mi
ne
d
be
s
t
s
uit
e
d
to
mi
nim
ize
ove
r
a
ll
f
or
e
c
a
s
t
e
r
r
or
while
s
e
lf
-
r
e
pe
a
ti
ng
type
models
be
tt
e
r
a
da
pt
to
load
va
r
iation
r
e
s
pons
e
.
3
.
2.
Nove
l
c
on
t
r
ib
u
t
io
n
s
of
t
h
e
s
t
u
d
y
T
he
major
c
ontr
ibut
ions
o
f
thi
s
p
r
ojec
t:
a.
A
hybr
id
f
or
e
c
a
s
ti
ng
model
that
is
de
ve
loped
by
c
ombi
ning
the
nonli
ne
a
r
f
e
a
tur
e
e
xt
r
a
c
ti
on
c
a
pa
bil
it
y
of
F
F
NN
a
nd
the
tempor
a
l
lea
r
ning
c
a
pa
bil
it
y
of
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
ks
is
us
e
d
f
or
s
hor
t
-
ter
m
load
f
or
e
c
a
s
ti
ng.
T
his
a
r
c
hit
e
c
tur
e
ha
s
two
models
-
F
F
NN
-
L
S
T
M
a
nd
F
F
NN
-
R
NN
.
b.
T
h
e
a
s
s
e
s
s
men
t
o
f
the
s
u
gg
e
s
t
e
d
m
ode
ls
oc
c
u
r
s
a
t
th
r
e
e
p
r
a
c
t
ica
l
f
o
r
e
c
a
s
t
in
g
h
o
r
i
z
o
ns
w
h
ich
a
r
e
a
t
2
4
h
o
ur
s
,
7
2
ho
ur
s
a
n
d
a
t
1
68
ho
ur
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2088
-
8708
I
nt
J
E
lec
&
C
omp
E
ng
,
Vol
.
16
,
No.
5
,
Oc
tober
20
26
:
2347
-
2356
2354
c.
I
n
or
de
r
to
e
va
luate
f
o
r
e
c
a
s
t
a
c
c
ur
a
c
y,
r
obus
tnes
s
a
nd
pr
e
diction
bias
,
the
model
pe
r
f
or
manc
e
ha
s
be
e
n
a
s
s
e
s
s
e
d
us
ing
R
M
S
E
,
M
AE
,
M
APE
,
a
nd
r
e
s
idual
a
na
lys
is
.
d.
Ac
c
or
ding
to
thi
s
r
e
s
e
a
r
c
h
s
tudy,
hyb
r
id
models
te
nd
to
a
c
hieve
r
e
latively
lowe
r
R
M
S
E
,
while
s
tand
a
lone
r
e
c
ur
r
e
nt
models
tend
to
a
c
hieve
lowe
r
M
AE
a
nd
lowe
r
M
APE
.
I
n
othe
r
wor
ds
,
the
latte
r
models
ha
ve
lowe
r
tr
a
de
-
of
f
.
e.
T
he
pr
opos
e
d
f
r
a
mew
or
k
is
pr
a
c
ti
c
a
ll
y
a
ppli
c
a
ble
f
or
e
lec
tr
icity
load
f
o
r
e
c
a
s
ti
ng
to
s
uppor
t
ge
ne
r
a
ti
on
s
c
he
duli
ng,
de
mand
r
e
s
pons
e
,
e
ne
r
gy
mana
ge
ment
a
nd
s
mar
t
-
gr
id
de
c
is
ion
-
making.
4.
CONC
L
USI
ON
T
he
r
e
s
e
a
r
c
h
c
ompar
e
s
the
hybr
id
model
a
r
c
hit
e
c
tur
e
s
of
F
F
NN
-
L
S
T
M
a
nd
F
F
NN
-
R
NN
a
r
c
hit
e
c
tur
e
s
f
or
s
hor
t
-
ter
m
e
lec
tr
icity
load
f
o
r
e
c
a
s
ti
ng
ove
r
24
hou
r
s
,
72
hour
s
,
a
nd
168
hour
s
.
Ac
c
o
r
ding
to
the
r
e
s
ult
s
,
the
F
F
NN
-
L
S
T
M
model
ha
s
the
lowe
s
t
R
M
S
E
va
lues
whic
h
s
ugge
s
t
that
thi
s
model
is
the
mos
t
a
c
c
ur
a
te
a
mong
the
other
s
.
How
e
ve
r
,
the
s
tand
-
a
lone
L
S
T
M
a
nd
the
R
NN
model
ha
ve
lowe
r
M
AE
a
nd
M
APE
va
lues
whic
h
s
ugge
s
t
that
thes
e
models
a
r
e
mor
e
s
e
ns
it
ive
to
s
hor
t
-
ter
m
load
f
luctua
ti
ons
.
T
he
f
indi
ngs
s
how
the
c
ompr
om
is
e
in
pr
e
diction
s
tabil
it
y
f
or
be
tt
e
r
a
da
pt
f
o
r
e
c
a
s
t
pe
r
f
or
manc
e
.
T
he
c
ompar
a
ti
ve
f
r
a
mew
or
k
p
r
opos
e
d
would
p
r
ovide
us
e
f
ul
inf
o
r
mation
to
s
e
lec
t
the
a
ppr
opr
iate
f
o
r
e
c
a
s
ti
ng
mo
de
ls
f
or
e
ne
r
gy
mana
ge
ment,
de
mand
r
e
s
pons
e
,
a
nd
s
mar
t
-
gr
id
a
ppli
c
a
ti
ons
.
I
n
the
f
utur
e
,
we
will
look
int
o
the
c
ombi
na
ti
on
of
mete
or
ology
va
r
iable
s
,
r
e
ne
wa
ble
e
ne
r
gies
da
ta,
tr
a
ns
f
or
mer
-
ba
s
e
d
models
,
a
tt
e
nti
on
mec
ha
nis
ms
,
a
nd
e
xplaina
ble
a
r
ti
f
icia
l
in
telli
ge
nc
e
to
incr
e
a
s
e
a
c
c
ur
a
c
y
a
nd
model
unde
r
s
tanding.
F
UN
DI
NG
I
NF
ORM
AT
I
ON
T
his
r
e
s
e
a
r
c
h
wa
s
no
t
f
unde
d
by
a
ny
gr
a
nt
f
r
om
a
ny
publi
c
,
c
omm
e
r
c
ial
,
o
r
not
-
f
or
-
pr
of
it
f
unding
a
ge
nc
y.
AU
T
HO
R
CONT
RI
B
U
T
I
ONS
S
T
AT
E
M
E
N
T
T
his
jour
na
l
us
e
s
the
C
ontr
ibut
o
r
R
oles
T
a
xo
nomy
(
C
R
e
diT
)
to
r
e
c
ognize
indi
vidual
a
uthor
c
ontr
ibut
ions
,
r
e
duc
e
a
utho
r
s
hip
dis
putes
,
a
nd
f
a
c
il
it
a
te
c
oll
a
bor
a
ti
on.
Nam
e
of
Au
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
T
e
mi
tope
Akinye
de
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
J
os
e
phine
Ade
nike
Akinye
de
✓
✓
✓
✓
✓
✓
✓
✓
✓
P
a
ul
Ke
hinde
Olulope
✓
✓
✓
✓
✓
✓
✓
✓
E
m
ma
nu
e
l
T
a
i
wo
F
a
s
in
a
✓
✓
✓
✓
✓
T
e
mi
tope
Ade
wa
le
Olomi
nu
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
li
z
a
ti
on
M
:
M
e
th
odol
ogy
So
:
So
f
twa
r
e
Va
:
Va
li
da
ti
on
Fo
:
Fo
r
ma
l
a
na
ly
s
is
I
:
I
nve
s
ti
ga
ti
on
R
:
R
e
s
our
c
e
s
D
:
D
a
ta
C
ur
a
ti
on
O
:
W
r
it
in
g
-
O
r
ig
in
a
l
D
r
a
f
t
E
:
W
r
it
in
g
-
R
e
vi
e
w
&
E
di
ti
ng
Vi
:
Vi
s
ua
li
z
a
ti
on
Su
:
Su
pe
r
vi
s
io
n
P
:
P
r
oj
e
c
t
a
dmi
ni
s
tr
a
ti
on
Fu
:
Fu
ndi
ng a
c
qui
s
it
io
n
CONF
L
I
CT
OF
I
NT
E
RE
S
T
S
T
AT
E
M
E
N
T
T
he
a
uthor
s
de
c
lar
e
that
no
c
onf
li
c
ts
of
int
e
r
e
s
t
e
xi
s
t.
DA
T
A
AV
AI
L
A
B
I
L
I
T
Y
T
he
da
ta
us
e
d
in
thi
s
s
tudy
a
r
e
a
va
il
a
ble
f
r
om
the
c
or
r
e
s
ponding
a
uthor
upon
r
e
a
s
ona
ble
r
e
que
s
t.
RE
F
E
RE
NC
E
S
[
1]
A
.
B
.
N
a
s
s
if
,
B
.
S
ouda
n,
M
.
A
z
z
e
h,
I
.
A
tt
il
li
,
a
nd
O
.
A
lm
ul
la
,
“
A
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
a
nd
s
ta
ti
s
ti
c
a
l
te
c
hni
que
s
in
s
hor
t
-
te
r
m
l
oa
d
f
or
e
c
a
s
ti
ng:
a
r
e
vi
e
w
,”
I
nt
e
r
nat
io
nal
R
e
v
ie
w
on
M
ode
ll
in
g
and
Si
m
ul
at
io
ns
,
vol
.
14,
no.
6,
pp.
408
–
430,
2021,
doi
:
10.15866/i
r
e
mos
.v14i6.21328.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
E
lec
&
C
omp
E
ng
I
S
S
N:
2088
-
8708
A
c
ompar
ati
v
e
analys
is
of
hy
br
id
F
F
N
N
-
L
ST
M
an
d
F
F
N
N
-
R
N
N
ar
c
hit
e
c
tur
e
s
…
(
T
e
mitope
A
k
inye
de
)
2355
[
2]
S
.
A
li
,
S
.
B
oga
r
r
a
,
M
.
N
.
R
ia
z
,
P
.
P
.
P
hyo,
D
.
F
ly
nn,
a
nd
A
.
T
a
ha
,
“
F
r
om
ti
me
-
s
e
r
ie
s
to
hybr
id
mode
ls
:
a
dva
nc
e
me
nt
s
in
s
hor
t
-
te
r
m
lo
a
d
f
or
e
c
a
s
ti
ng
e
mbr
a
c
in
g
s
m
a
r
t
gr
id
pa
r
a
di
gm,”
A
ppl
ie
d
Sc
ie
nc
e
s
(
Sw
it
z
e
r
la
nd)
,
vol
.
14,
no.
11,
p.
4442,
2024,
doi
:
10.3390/a
pp14114442.
[
3]
K
.
U
ll
a
h
e
t
al
.
,
“
S
ho
r
t
-
te
r
m
lo
a
d
f
o
r
e
c
a
s
ti
ng:
a
c
ompr
e
he
ns
iv
e
r
e
vi
e
w
a
nd
s
im
ul
a
ti
on
s
tu
dy
w
it
h
C
N
N
-
L
S
T
M
hybr
id
s
a
ppr
oa
c
h,”
I
E
E
E
A
c
c
e
s
s
, vol
. 12, pp. 111858
–
111881, 2024, doi:
10.1109/AC
C
E
S
S
.2024.3440631.
[
4]
M
.
M
.
A
s
ir
i,
G
.
A
ld
e
hi
m,
F
.
A
.
A
lo
ta
ib
i,
M
.
M
.
A
ln
f
ia
i,
M
.
A
s
s
ir
i,
a
nd
A
.
M
a
hmud,
“
S
hor
t
-
te
r
m
lo
a
d
f
or
e
c
a
s
ti
ng
in
s
ma
r
t
gr
id
s
us
in
g hybr
id
de
e
p l
e
a
r
ni
ng,”
I
E
E
E
A
c
c
e
s
s
, vol
. 12, pp. 23504
–
23513, 2024, doi:
10.1109/AC
C
E
S
S
.2024.3358182.
[
5]
X
.
W
e
n, J
.
L
ia
o,
Q
. N
iu
,
N
.
S
he
n, a
nd
Y
. B
a
o, “
D
e
e
p
le
a
r
ni
ng
-
dr
iv
e
n
hybr
id
mode
l
f
or
s
hor
t
-
te
r
m
lo
a
d
f
or
e
c
a
s
ti
ng
a
nd s
ma
r
t
gr
id
in
f
or
ma
ti
on ma
na
ge
me
nt
,”
Sc
ie
nt
if
ic
R
e
por
ts
, vol
. 14, no. 1, p.
13720, 2024, doi:
10.1038/s
41598
-
024
-
63262
-
x.
[
6]
M
.
K
ha
li
d,
“
A
r
e
vi
e
w
on
th
e
s
e
le
c
te
d
a
ppl
ic
a
ti
ons
of
b
a
tt
e
r
y
-
s
upe
r
c
a
pa
c
it
or
hybr
id
e
ne
r
gy
s
to
r
a
ge
s
y
s
te
ms
f
or
mi
c
r
ogr
id
s
,
”
E
ne
r
gi
e
s
, vol
. 12, no. 23, pp. 9
–
21, 2019, doi:
10.3390/en1223
4559.
[
7]
J
.
W
.
T
a
yl
or
a
nd
P
.
E
.
M
c
S
ha
r
r
y,
“
S
hor
t
-
te
r
m
lo
a
d
f
or
e
c
a
s
ti
ng
me
th
ods
:
A
n
e
va
lu
a
ti
on
b
a
s
e
d
on
E
ur
ope
a
n
da
t
a
,”
I
E
E
E
T
r
ans
ac
ti
ons
on P
ow
e
r
Sy
s
te
m
s
, vol
. 22, no. 4, pp. 2213
–
2219,
2007, doi:
10.1109/T
P
W
R
S
.2007.907583.
[
8]
H
.
L
.
W
il
li
s
a
nd
W
.
G
.
S
c
ot
t,
D
is
tr
ib
ut
e
d
pow
e
r
ge
ne
r
at
io
n
:
P
la
nni
ng
and
e
v
al
uat
io
n
.
B
oc
a
R
a
to
n,
F
L
,
U
S
A
:
C
R
C
P
r
e
s
s
,
2000.
[
9]
M
.
G
e
ur
ts
,
G
.
E
.
P
.
B
ox,
a
nd
G
.
M
.
J
e
nki
ns
,
T
ime
s
e
r
ie
s
anal
y
s
is
:
fo
r
e
c
a
s
ti
ng
and
c
ont
r
ol
,
5t
h
e
d.,
vol
.
14,
no.
2.
H
oboke
n,
N
J
,
U
S
A
:
W
il
e
y, 1977.
[
10]
R
.
J
.
H
yndma
n
a
nd
G
.
A
th
a
na
s
opoulos
,
F
or
e
c
a
s
ti
ng:
pr
in
c
ip
le
s
and
pr
ac
ti
c
e
,
3r
d
e
d.
M
e
lb
our
ne
,
A
u
s
tr
a
li
a
:
O
T
e
xt
s
,
2021.
[
11]
S
.
H
oc
hr
e
it
e
r
a
nd
J
.
S
c
hmi
dhube
r
,
“
L
ong
s
hor
t
-
te
r
m
me
mor
y,
”
N
e
ur
al
C
om
put
at
io
n
,
vol
.
9,
no.
8,
pp.
1735
–
1780,
N
ov.
1997,
doi
:
10.1162/nec
o.1997.9.8.1735.
[
12]
A
.
G
r
a
ve
s
,
A
.
R
.
M
oha
me
d,
a
nd
G
.
H
in
to
n,
“
S
pe
e
c
h
r
e
c
ogni
ti
on
w
it
h
de
e
p
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
ks
,”
in
I
C
A
SS
P
,
I
E
E
E
I
nt
e
r
nat
io
nal
C
onf
e
r
e
n
c
e
on
A
c
ous
ti
c
s
,
Spe
e
c
h
and
Si
g
nal
P
r
oc
e
s
s
in
g
-
P
r
oc
e
e
di
ngs
,
2013,
pp.
6645
–
6649,
doi
:
10.1109/I
C
A
S
S
P
.2013.6638947.
[
13]
H
.
Z
he
ng,
J
.
Y
ua
n,
a
nd
L
.
C
h
e
n,
“
S
hor
t
-
te
r
m
lo
a
d
f
or
e
c
a
s
ti
ng
us
in
g
E
M
D
-
L
S
T
M
ne
ur
a
l
ne
twor
ks
w
it
h
a
X
G
B
oos
t
a
lg
or
it
hm
f
or
f
e
a
tu
r
e
i
mpor
ta
nc
e
e
va
lu
a
ti
on,”
E
ne
r
gi
e
s
, vol
. 10, no. 8, p. 1168, 2017, doi
:
10.3390/e
n10081168.
[
14]
W
.
K
ong,
Z
.
Y
.
D
ong,
Y
.
J
ia
,
D
.
J
.
H
il
l,
Y
.
X
u,
a
nd
Y
.
Z
ha
ng,
“
S
hor
t
-
te
r
m
r
e
s
id
e
nt
ia
l
lo
a
d
f
or
e
c
a
s
ti
ng
ba
s
e
d
on
L
S
T
M
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
k,”
I
E
E
E
T
r
ans
ac
ti
ons
on Smar
t
G
r
id
, vol
. 10, no. 1, pp. 841
–
851, 2019, doi:
10.1109/T
S
G
.2017.2753802.
[
15]
D
.
L
.
M
a
r
in
o,
K
.
A
ma
r
a
s
in
ghe
,
a
nd
M
.
M
a
ni
c
,
“
B
ui
ld
in
g
e
ne
r
gy
lo
a
d
f
or
e
c
a
s
ti
ng
us
in
g
de
e
p
ne
ur
a
l
ne
twor
ks
,”
in
I
E
C
O
N
P
r
oc
e
e
di
ngs
(
I
ndus
tr
ia
l
E
le
c
tr
oni
c
s
C
onf
e
r
e
nc
e
)
, 2016, pp. 704
6
–
7051, doi:
10.1109/I
E
C
O
N
.2016.7793413.
[
16]
F
.
A
.
A
gga
,
S
.
A
.
A
bbou,
Y
.
E
l
H
oum,
a
nd
M
.
L
a
bba
di
,
“
S
hor
t
-
te
r
m
lo
a
d
f
or
e
c
a
s
ti
ng
ba
s
e
d
on
C
N
N
a
nd
L
S
T
M
de
e
p
n
e
ur
a
l
ne
twor
ks
,”
I
F
A
C
-
P
ape
r
s
O
nL
in
e
, vol
. 55, no. 12, pp. 777
–
781, 2022, doi:
10.1016/j
.i
f
a
c
ol
.2022.07.407.
[
17]
J
.
L
i,
Y
.
Z
ha
ng,
X
.
L
iu
,
a
nd
H
.
W
a
ng,
“
A
hybr
id
de
e
p
le
a
r
ni
ng
f
r
a
me
w
or
k
f
or
a
c
c
ur
a
te
s
hor
t
-
te
r
m
lo
a
d
f
or
e
c
a
s
ti
ng
in
s
ma
r
t
gr
id
s
,”
E
ne
r
gy
R
e
por
ts
, vol
. 10, pp. 1483
–
1497, 2024.
[
18]
A
.
K
.
S
in
gh,
S
.
M
is
hr
a
,
a
nd
R
.
K
uma
r
,
“
D
e
e
p
le
a
r
ni
ng
-
ba
s
e
d
mul
ti
-
hor
iz
on
s
hor
t
-
te
r
m
lo
a
d
f
or
e
c
a
s
ti
ng
us
in
g
hybr
id
ne
ur
a
l
ne
twor
ks
,”
E
le
c
tr
ic
P
ow
e
r
Sy
s
te
m
s
R
e
s
e
a
r
c
h
, vol
. 228, p. 1100
97, 2024.
[
19]
L
.
Z
ha
o,
Y
.
C
he
n,
a
nd
H
.
S
un,
“
W
e
a
th
e
r
-
a
w
a
r
e
de
e
p
le
a
r
ni
ng
f
r
a
me
w
or
k
f
or
mul
ti
-
s
te
p
e
le
c
tr
ic
it
y
lo
a
d
f
or
e
c
a
s
ti
ng,”
E
ngi
ne
e
r
in
g
A
ppl
ic
at
io
ns
of
A
r
ti
fi
c
ia
l
I
nt
e
ll
ig
e
nc
e
, vol
. 131, p. 107874, 202
5.
[
20]
Z
.
L
iu
,
X
.
W
a
ng,
J
.
Z
hou,
a
nd
Y
.
L
i,
“
M
ul
ti
-
ho
r
iz
on
e
le
c
tr
ic
it
y
lo
a
d
f
or
e
c
a
s
ti
ng
us
in
g
opt
im
iz
e
d
hybr
id
de
e
p
le
a
r
ni
ng
ne
twor
ks
,
”
A
ppl
ie
d E
ne
r
gy
, vol
. 378, p. 124932, 2025.
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Tem
i
t
o
pe
A
ki
n
y
ede
earn
ed
a
B.
E
n
g
.
d
eg
ree
a
n
d
mas
t
er
i
n
el
ect
r
i
cal
a
n
d
el
ect
r
o
n
i
c
en
g
i
n
eer
i
n
g
fro
m
E
k
i
t
i
St
at
e
U
n
i
v
er
s
i
t
y
.
H
e
p
res
e
n
t
l
y
h
o
l
d
s
a
p
o
s
i
t
i
o
n
as
a
l
ec
t
u
rer
i
n
t
h
e
D
ep
art
me
n
t
o
f
E
l
ec
t
ri
ca
l
an
d
E
l
ec
t
ro
n
i
c
E
n
g
i
n
e
eri
n
g
,
E
k
i
t
i
St
a
t
e
Po
l
y
t
ec
h
n
i
c,
Is
a
n
E
k
i
t
i
,
N
i
g
eri
a.
H
e’s
a
reg
i
s
t
ered
e
n
g
i
n
eer
w
i
t
h
t
h
e
C
o
u
n
ci
l
fo
r
t
h
e
Reg
u
l
at
i
o
n
o
f
E
n
g
i
n
eer
i
n
g
i
n
N
i
g
eri
a
(CO
RE
N
)
an
d
a
memb
er
o
f
t
h
e
N
i
g
er
i
an
So
c
i
et
y
o
f
E
n
g
i
n
eers
(N
S
E
).
T
h
e
el
ec
t
ri
ca
l
l
o
a
d
fo
recas
t
i
n
g
,
p
o
w
er
s
y
s
t
em
s
,
ren
ew
a
b
l
e
e
n
erg
y
,
art
i
fi
ci
a
l
i
n
t
el
l
i
g
en
ce
a
p
p
l
i
ca
t
i
o
n
i
n
p
o
w
er
s
y
s
t
ems
a
n
d
s
m
art
g
r
i
d
h
as
b
ee
n
h
i
s
res
earc
h
area
o
f
i
n
t
e
res
t
.
t
a
k
i
n
y
e
d
e@
ek
s
p
o
l
y
.
ed
u
.
n
g
.
J
o
s
ephi
ne
A
den
i
ke
A
k
i
ny
ede
earn
ed
h
er
B.
E
n
g
.
an
d
mas
t
er
i
n
el
ec
t
ri
ca
l
an
d
el
ect
r
o
n
i
c
en
g
i
n
eeri
n
g
fro
m
t
h
e
U
n
i
v
ers
i
t
y
o
f
Il
o
ri
n
,
Il
o
r
i
n
,
N
i
g
eri
a.
Pres
en
t
l
y
,
t
h
e
i
n
d
i
v
i
d
u
al
i
s
a
Ph
.
D
.
s
t
u
d
en
t
i
n
el
ect
r
i
ca
l
an
d
e
l
ect
r
o
n
i
c
en
g
i
n
eer
i
n
g
at
t
h
e
Fe
d
eral
U
n
i
v
er
s
i
t
y
O
y
e
-
E
k
i
t
i
,
N
i
g
eri
a.
Sh
e
i
s
an
e
n
g
i
n
eer
reg
i
s
t
ered
w
i
t
h
t
h
e
Co
u
n
c
i
l
fo
r
t
h
e
Reg
u
l
a
t
i
o
n
o
f
E
n
g
i
n
eeri
n
g
i
n
N
i
g
eri
a
(CO
R
E
N
)
a
n
d
i
s
a
l
s
o
a
memb
er
o
f
t
h
e
N
i
g
er
i
an
So
c
i
et
y
o
f
E
n
g
i
n
eers
(
N
SE
).
H
e
r
res
earch
area
c
o
mp
r
i
s
e
s
p
o
w
er
s
y
s
t
ems
,
p
o
w
er
q
u
al
i
t
y
,
ren
e
w
ab
l
e
en
er
g
y
i
n
t
eg
ra
t
i
o
n
,
d
i
s
t
ri
b
u
t
e
d
g
en
era
t
i
o
n
a
n
d
s
mar
t
g
r
i
d
.
ak
i
n
y
ed
e
2
0
2
0
@
g
mai
l
.
co
m.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2088
-
8708
I
nt
J
E
lec
&
C
omp
E
ng
,
Vol
.
16
,
No.
5
,
Oc
tober
20
26
:
2347
-
2356
2356
P
a
ul
Kehi
nde
Ol
ul
o
pe
a
Pro
fes
s
o
r
o
f
E
l
ec
t
ri
ca
l
an
d
E
l
ec
t
ro
n
i
c
E
n
g
i
n
eer
i
n
g
at
E
k
i
t
i
St
at
e
U
n
i
v
ers
i
t
y
i
n
N
i
g
eri
a.
H
e
i
s
an
ex
p
er
t
at
p
o
w
er
s
y
s
t
em
en
g
i
n
eer
i
n
g
.
H
e
w
as
aw
ard
e
d
a
B.
E
n
g
.
d
eg
ree,
mas
t
er's
d
e
g
ree
fro
m
E
k
i
t
i
St
at
e
U
n
i
v
er
s
i
t
y
,
t
h
e
M.
E
n
g
.
H
e
h
as
a
B.
Sc.
d
eg
ree
fro
m
t
h
e
U
n
i
v
er
s
i
t
y
o
f
Ben
i
n
,
N
i
g
er
i
a,
an
d
t
h
e
Ph
.
D
.
d
eg
ree
i
n
p
o
w
er
s
y
s
t
em
en
g
i
n
eer
i
n
g
fro
m
t
h
e
U
n
i
v
er
s
i
t
y
o
f
Cap
e
T
o
w
n
,
So
u
t
h
A
fri
ca.
H
e
i
s
a
memb
er
o
f
IE
E
E
,
CO
RE
N
an
d
N
S
E
.
H
i
s
re
s
earc
h
i
n
t
ere
s
t
s
i
n
cl
u
d
e
t
h
e
a
p
p
l
i
ca
t
i
o
n
s
o
f
co
m
p
u
t
at
i
o
n
al
.
p
au
l
.
o
l
u
l
o
p
e@
ek
s
u
.
e
d
u
.
n
g
Em
m
a
nuel
Ta
i
w
o
F
a
s
i
na
o
b
t
a
i
n
e
d
t
h
e
B.
E
n
g
.
d
eg
re
e
i
n
el
ec
t
ri
ca
l
an
d
el
ec
t
ro
n
i
c
en
g
i
n
eer
i
n
g
fro
m
O
n
d
o
St
a
t
e
U
n
i
v
ers
i
t
y
A
d
o
E
k
i
t
i
,
N
i
g
eri
a
(n
o
w
E
k
i
t
i
St
at
e
U
n
i
v
er
s
i
t
y
,
A
d
o
-
E
k
i
t
i
),
i
n
1
9
9
8
a
n
d
M.
T
ech
n
o
l
o
g
y
a
q
u
a
l
i
f
i
cat
i
o
n
fro
m
L
A
U
T
E
C
H
,
O
g
b
o
mo
s
o
,
N
i
g
er
i
a.
In
2
0
1
9
,
h
e
o
b
t
ai
n
ed
a
P
h
.
D
.
fr
o
m
Card
i
ff
U
n
i
v
ers
i
t
y
i
n
t
h
e
U
n
i
t
e
d
K
i
n
g
d
o
m.
Cu
rren
t
l
y
,
h
e
i
s
a
n
A
s
s
o
c
i
at
e
Pro
fe
s
s
o
r
o
f
E
l
ec
t
r
i
cal
an
d
E
l
ec
t
ro
n
i
c
E
n
g
i
n
eeri
n
g
at
E
k
i
t
i
St
at
e
U
n
i
v
er
s
i
t
y
,
A
d
o
E
k
i
t
i
,
N
i
g
eri
a.
H
e
i
s
a
memb
er
o
f
t
h
e
N
i
g
eri
a
n
So
c
i
et
y
o
f
E
n
g
i
n
eer
s
,
an
d
a
re
g
i
s
t
ere
d
CO
R
E
N
en
g
i
n
eer.
Po
w
er
s
y
s
t
em
an
a
l
y
s
i
s
,
d
i
s
t
r
i
b
u
t
e
d
g
e
n
e
rat
i
o
n
,
el
ec
t
ri
c
v
eh
i
cl
e
s
,
el
ec
t
ri
c
i
t
y
t
ran
s
mi
s
s
i
o
n
p
ri
c
i
n
g
,
an
d
c
o
n
t
ro
l
are
h
i
s
mai
n
i
n
t
eres
t
s
.
emman
u
e
l
.
fas
i
n
a@
ek
s
u
.
ed
u
.
n
g
.
Tem
i
t
o
pe
A
dew
a
l
e
Ol
o
m
i
nu
o
b
t
a
i
n
ed
b
ac
h
el
o
r’s
d
eg
ree
fro
m
t
h
e
U
n
i
v
ers
i
t
y
o
f
Il
o
r
i
n
,
Il
o
r
i
n
,
N
i
g
er
i
a,
an
d
s
h
e
i
s
cu
rren
t
l
y
p
u
r
s
u
i
n
g
t
h
e
mas
t
er’
s
d
eg
ree
i
n
s
t
at
i
s
t
i
c
s
at
Fed
eral
U
n
i
v
er
s
i
t
y
o
f
T
ech
n
o
l
o
g
y
A
k
u
re,
N
i
g
er
i
a.
Sh
e
i
s
a
reg
i
s
t
ere
d
(CISO
N
)
s
t
a
t
i
s
t
i
ci
a
n
,
a
p
ro
fe
s
s
i
o
n
al
mem
b
er
o
f
(PSSN
)
a
n
d
a
res
earc
h
er
a
s
s
i
s
t
an
t
.
So
me
o
f
h
er
i
n
t
eres
t
i
n
cl
u
d
e
s
t
i
me
s
eri
e
s
an
al
y
s
i
s
,
eco
n
o
m
et
r
i
cs
,
s
t
at
i
s
t
i
ca
l
mo
d
e
l
l
i
n
g
,
mach
i
n
e
l
earn
i
n
g
,
el
ect
r
i
ci
t
y
l
o
a
d
fo
recas
t
i
n
g
,
an
d
d
at
a
a
n
al
y
t
i
cs
.
t
a
o
l
o
mi
n
u
@
ek
s
p
o
l
y
.
ed
u
.
n
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.