I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
11
, N
o.
1
,
M
a
r
c
h
2022
, pp.
300
~
309
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
11
.i
1
.pp
300
-
309
300
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
M
ac
h
i
n
e
l
e
ar
n
i
n
g al
gor
i
t
h
m
s f
or
e
l
e
c
t
r
i
c
al
ap
p
l
i
an
c
e
s
m
on
i
t
or
i
n
g sys
t
e
m
u
si
n
g o
p
e
n
-
sou
r
c
e
sys
t
e
m
s
V
ie
t
H
oan
g D
u
on
g, N
am
H
oan
g N
gu
ye
n
D
e
pa
r
t
m
e
nt
of
I
ns
t
r
um
e
nt
a
t
i
on a
nd I
ndus
t
r
i
a
l
I
n
f
or
m
a
t
i
on, H
a
noi
U
ni
ve
r
s
i
t
y of
S
c
i
e
nc
e
a
nd
T
e
c
hnol
ogy, V
i
e
t
na
m
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
M
a
y 9
, 2021
R
e
vi
s
e
d
D
e
c
20
, 2021
A
c
c
e
pt
e
d
D
e
c
30
,
2021
Two
main
methods
to
minimize
the
impact
of
electricity
generation
on
the
environm
ent
are
to
exploit
clean
fuel
resources
and
use
electricity
more
effectively
.
In
this
paper,
we
aim
to
change
the
user'
s
electricit
y
us
age
by
providing
feedback
about
the
electrical
energy
consumed
by
each
device.
The
authors
introduced
two
devices,
load
monitoring
device
(LMD)
and
act
ivity
monitoring
device
(AMD).
The
function
of
the
LMD
is
to
p
rovide
feedback
on
the
operating
status
and
energy
consumpti
on
of
ele
ctrical
appliances
in
a
home,
which
will
help
people
consume
electrical
energy
more
efficiently.
The
parameters
of
LMD
are
us
ed
to
predict
the
on/o
ff
state
of
each
electrical
appliance
thanks
to
machine
learning
algorithms.
AMD
with
audio
sensors
can
assist
LMD
to
distinguish
electrical
devices
w
ith
the
same or vary
ing power over
time. The system wa
s tested
for three w
ee
ks and
a
chieved a s
tate predi
ction accu
racy of 93.
60%.
K
e
y
w
o
r
d
s
:
E
le
c
tr
ic
a
l
a
ppl
ia
nc
e
s
ta
te
s
M
a
c
hi
ne
l
e
a
r
ni
ng
O
pe
n
-
s
our
c
e
S
a
vi
ng e
ne
r
gy
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
N
a
m
H
oa
ng N
guye
n
S
c
hool
of
E
le
c
tr
ic
a
l
E
ngi
ne
e
r
in
g, H
a
noi
U
ni
ve
r
s
it
y of
S
c
ie
nc
e
a
nd T
e
c
hnol
ogy
N
o. 1 Da
i
C
o V
ie
t,
H
a
i
B
a
T
r
ung
D
is
tr
ic
t,
H
a
noi
, V
ie
tn
a
m
E
m
a
il
:
na
m
.nguye
nhoa
ng@
hus
t.
e
du.vn
1.
I
N
T
R
O
D
U
C
T
I
O
N
T
hi
s
pa
pe
r
i
s
a
n
e
xt
e
n
s
io
n
of
w
or
k
or
ig
in
a
ll
y
pr
e
s
e
nt
e
d
in
th
e
2019
I
nt
e
r
na
ti
ona
l
C
onf
e
r
e
nc
e
on
A
dva
nc
e
d
C
om
put
in
g
a
nd
A
ppl
ic
a
ti
on
s
(
A
C
O
M
P
)
[
1]
.
E
le
c
tr
ic
it
y
ge
ne
r
a
ti
on
pl
a
ys
a
s
ig
ni
f
ic
a
nt
r
ol
e
in
in
c
r
e
a
s
in
g gr
e
e
nhous
e
ga
s
e
m
is
s
io
ns
, a
nd t
hi
s
s
it
ua
ti
on i
s
ge
tt
in
g w
or
s
e
. I
n t
he
U
S
, e
le
c
tr
ic
it
y c
ons
um
pt
io
n i
n
2018
w
a
s
16
ti
m
e
s
hi
ghe
r
th
a
n
th
a
t
in
1950
[
2]
.
A
ls
o,
in
th
is
y
e
a
r
,
e
le
c
tr
ic
e
ne
r
gy
pr
im
a
r
il
y
c
a
m
e
f
r
om
th
r
e
e
s
our
c
e
s
:
r
e
s
id
e
nt
ia
l
(
39%
)
,
c
om
m
e
r
c
ia
l
(
36%
)
,
a
nd
in
dus
tr
y
(
25%
)
.
I
t
is
c
le
a
r
th
a
t
th
e
hi
ghe
s
t
pr
opor
ti
on
is
in
th
e
r
e
s
id
e
nt
ia
l
s
e
c
ti
on,
s
o i
t’
s
ne
e
de
d t
o t
a
k
e
m
e
a
s
ur
e
s
t
o r
e
duc
e
e
le
c
tr
ic
it
y
c
ons
um
pt
io
n f
r
om
t
hi
s
s
our
c
e
.
A
c
c
or
di
ng
to
[
3]
,
S
a
r
a
h
D
a
r
by
e
s
ti
m
a
te
d
th
a
t
c
oul
d
s
a
ve
up
to
15%
e
le
c
tr
ic
a
l
e
ne
r
gy
if
a
w
a
r
e
of
th
e
f
e
e
dba
c
k
on
th
e
pow
e
r
c
ons
um
pt
io
n
of
e
le
c
tr
ic
a
l
a
ppl
ia
nc
e
s
.
S
im
il
a
r
ly
,
S
é
ba
s
ti
e
n
H
oude
r
e
s
e
a
r
c
he
d
th
e
e
f
f
e
c
ts
of
pow
e
r
c
ons
um
pt
io
n
f
e
e
db
a
c
k
on
us
e
r
s
on
a
la
r
ge
s
c
a
le
(
1065
a
pa
r
tm
e
nt
s
in
e
ig
ht
m
ont
hs
)
[
4]
.
T
he
r
e
s
ul
ts
s
how
th
a
t
to
ta
l
pow
e
r
c
ons
um
pt
io
n
r
e
duc
e
d
by
a
bout
5
.7%
.
F
ur
th
e
r
m
or
e
,
C
a
r
r
ie
A
r
m
e
la
poi
nt
e
d
out
th
a
t
if
us
e
r
s
w
e
r
e
gi
ve
n
f
e
e
dba
c
k
on
th
e
e
ne
r
gy
c
ons
um
pt
i
on
of
e
a
c
h
e
le
c
tr
ic
a
l
a
ppl
ia
nc
e
,
th
e
y
c
oul
d
c
ons
um
e
12%
l
e
s
s
e
le
c
tr
ic
a
l
e
ne
r
gy
[
5]
.
I
ns
te
a
d
of
c
ha
ngi
ng
us
e
r
s
'
ha
bi
ts
,
T
s
a
i
e
t
al
.
bui
lt
a
s
y
s
te
m
to
a
ut
om
a
ti
c
a
ll
y
m
oni
to
r
a
nd
a
dj
us
t
in
door
e
le
c
tr
ic
it
y
us
a
ge
us
in
g
m
a
c
hi
ne
le
a
r
ni
ng
a
lg
or
it
hm
s
[
6]
.
R
e
s
e
a
r
c
h
s
how
s
th
a
t
e
le
c
tr
ic
a
l
a
ppl
ia
n
c
e
s
in
id
le
or
s
ta
ndby
m
ode
a
c
c
ount
f
or
3
-
11%
of
to
ta
l
in
door
e
ne
r
gy
c
ons
um
pt
io
n.
T
he
r
e
f
or
e
,
th
e
te
a
m
tr
a
in
e
d
th
e
s
ys
te
m
to
c
om
pl
e
te
ly
tu
r
n
of
f
e
le
c
tr
ic
a
l
d
e
vi
c
e
s
w
he
n
not
in
us
e
in
s
te
a
d
of
ope
r
a
ti
ng
in
id
le
or
s
ta
ndby
m
ode
.
T
o a
c
c
om
pl
is
h t
hi
s
, m
a
ny s
m
a
r
t
pl
ugs
w
it
h e
le
c
tr
ic
a
l
m
e
te
r
in
g a
nd on/of
f
f
unc
ti
ons
a
r
e
i
ns
ta
ll
e
d i
n t
he
hous
e
.
S
in
c
e
it
is
ti
m
e
-
c
ons
um
in
g
a
nd
c
os
tl
y
to
in
te
gr
a
te
e
a
c
h
s
m
a
r
t
pl
ug
in
to
e
ve
r
y
de
vi
c
e
,
in
th
is
pa
pe
r
,
th
e
a
ut
hor
s
m
ove
to
w
a
r
ds
c
ha
ngi
ng
u
s
e
r
s
'
ha
bi
ts
by
pr
ovi
di
n
g
f
e
e
dba
c
k
on
pow
e
r
c
ons
um
pt
io
n
le
v
e
ls
.
T
o
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
M
ac
hi
ne
l
e
ar
ni
ng algor
it
hm
s
f
or
…
(
V
ie
t
H
oang Duong
)
301
a
c
c
om
pl
is
h
th
is
pur
pos
e
,
th
e
a
ut
hor
s
in
tr
od
uc
e
d a
lo
a
d
m
oni
to
r
in
g
de
vi
c
e
(
L
M
D
)
to
m
oni
to
r
th
e
on/
of
f
s
ta
tu
s
a
nd
pow
e
r
c
on
s
um
pt
io
n
of
e
a
c
h e
le
c
tr
ic
a
l
de
vi
c
e
in
th
e
hou
s
e
. B
e
s
id
e
s
,
a
n
a
c
ti
vi
ty
m
oni
to
r
in
g
de
vi
c
e
(
A
M
D
)
w
it
h
a
n
a
udi
o
s
e
ns
or
is
a
ls
o
u
s
e
d
f
or
th
e
pur
pos
e
of
s
uppor
ti
ng
L
M
D
to
id
e
nt
if
y
e
le
c
tr
ic
a
l
de
vi
c
e
s
w
it
h
th
e
s
a
m
e
pow
e
r
or
c
ont
in
uous
ly
va
r
yi
ng powe
r
ove
r
t
im
e
.
2.
S
U
R
V
E
Y
O
N
M
O
N
I
T
O
R
I
N
G
E
L
E
C
T
R
I
C
A
L
A
P
P
L
I
A
N
C
E
S
H
a
r
t
in
tr
oduc
e
d
th
e
c
onc
e
pt
of
non
-
in
tr
us
iv
e
a
ppl
ia
nc
e
lo
a
d
m
oni
to
r
in
g
(
N
A
L
M
)
to
in
di
c
a
te
a
s
ys
te
m
th
a
t
id
e
nt
if
ie
s
e
le
c
tr
ic
a
l
a
ppl
ia
nc
e
s
us
in
g
onl
y a
s
in
gl
e
e
le
c
tr
oni
c
m
e
te
r
[
7]
.
E
a
c
h
e
le
c
tr
ic
a
l
de
vi
c
e
ha
s
di
f
f
e
r
e
nt
a
c
ti
ve
pow
e
r
(
P
)
a
nd
r
e
a
c
ti
ve
pow
e
r
(
Q
)
.
T
he
r
e
f
or
e
,
th
e
s
e
two
pa
r
a
m
e
te
r
s
a
r
e
us
e
d
a
s
"
S
ig
na
tu
r
e
"
f
or
e
a
c
h e
le
c
tr
ic
a
l
a
ppl
ia
nc
e
.
A
ppl
yi
ng
ha
r
t'
s
de
vi
c
e
id
e
nt
if
ic
a
ti
on
m
e
th
od,
W
e
is
s
e
t
al
.
u
s
e
d
a
c
om
m
e
r
c
ia
l
di
gi
ta
l
m
e
te
r
a
nd
s
of
twa
r
e
on
th
e
phone
to
pe
r
f
or
m
th
e
de
vi
c
e
r
e
c
ogni
ti
on
a
lg
o
r
it
hm
s
[
8]
.
T
he
a
lg
or
it
hm
s
a
ll
ow
us
to
a
dd
a
ne
w
de
vi
c
e
th
r
ough
th
e
s
of
twa
r
e
on
th
e
phone
a
nd
th
e
n
s
a
ve
th
e
pa
r
a
m
e
te
r
s
of
th
is
de
vi
c
e
in
a
da
ta
ba
s
e
. T
he
s
ys
te
m
h
a
s
a
r
e
c
ogni
ti
on
a
c
c
ur
a
c
y
of
up
to
87%
.
L
a
ughma
n
e
t
al
.
us
e
d
ha
r
m
oni
c
s
a
s
a
n
id
e
nt
if
ie
r
f
or
e
a
c
h
de
vi
c
e
[
9]
.
T
he
a
ut
hor
s
s
how
th
a
t
a
c
om
put
e
r
a
nd
a
n
in
c
a
nde
s
c
e
nt
bul
b
a
nd
h
a
ve
s
im
il
a
r
P
a
nd
Q
.
H
ow
e
ve
r
,
onl
y
th
e
c
om
put
e
r
c
a
n
pr
oduc
e
a
th
ir
d
ha
r
m
oni
c
,
a
nd
th
e
in
c
a
nde
s
c
e
nt
la
m
p
doe
s
not
.
T
he
r
e
f
or
e
,
ha
r
m
oni
c
s
c
a
n be
us
e
d t
o di
f
f
e
r
e
nt
ia
te
t
he
s
e
t
w
o de
vi
c
e
s
.
A
not
he
r
m
e
th
od
de
ve
lo
pe
d
by
N
o
r
f
or
d
a
nd
L
e
e
b
is
a
na
ly
z
in
g
th
e
tr
a
ns
ie
nt
s
ta
te
[
10]
.
E
a
c
h
de
vi
c
e
w
it
h
a
di
f
f
e
r
e
nt
s
tr
uc
tu
r
e
w
il
l
ha
v
e
a
di
f
f
e
r
e
nt
s
w
it
c
hi
ng
pow
e
r
a
t
s
ta
r
t
-
up.
S
r
in
iv
a
s
a
n
e
t
al
.
u
s
e
d
h
a
r
m
oni
c
pa
r
a
m
e
te
r
s
a
s
in
put
s
to
va
r
io
us
m
a
c
hi
ne
le
a
r
ni
ng
a
lg
or
it
hm
s
f
or
e
le
c
tr
ic
a
l
de
vi
c
e
id
e
nt
if
ic
a
ti
on
[
11]
.
T
he
r
e
s
ul
ts
s
how
e
d
th
a
t
th
e
m
a
c
hi
n
e
le
a
r
ni
ng
m
e
th
od
m
ul
ti
la
ye
r
pe
r
c
e
pt
r
on
(
M
L
P
)
a
nd
r
a
di
a
l
ba
s
is
f
unc
ti
on
(
R
B
F
)
ga
ve
s
im
il
a
r
r
e
s
ul
ts
a
nd
w
e
r
e
be
tt
e
r
th
a
n
th
e
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
(
S
V
M
)
m
e
th
od.
P
a
te
l
e
t
al
.
obs
e
r
ve
d
noi
s
e
oc
c
ur
r
in
g
on
th
e
li
ne
e
ve
r
y
ti
m
e
a
de
vi
c
e
w
a
s
tu
r
ne
d
on/
of
f
or
ope
r
a
ti
ng
[
12]
.
T
he
a
ut
hor
s
poi
nt
out
th
a
t
th
e
r
e
s
is
ti
ve
lo
a
d
doe
s
not
c
a
us
e
noi
s
e
in
ope
r
a
ti
on
but
doe
s
c
a
u
s
e
tr
a
n
s
ie
nt
noi
s
e
w
he
n
tu
r
ne
d
on or
of
f
. I
nduc
ti
ve
a
nd
s
ol
id
-
s
ta
te
l
oa
ds
c
a
us
e
a
ddi
ti
ona
l
noi
s
e
dur
in
g ope
r
a
ti
on.
L
a
m
e
t
al
.
pr
opos
e
d
a
s
ol
ut
io
n
to
us
e
th
e
vol
ta
ge
-
c
ur
r
e
nt
tr
a
je
c
to
r
y
(
V
-
I
tr
a
je
c
to
r
y)
a
s
id
e
nt
if
ie
r
s
f
o
r
e
a
c
h
e
le
c
tr
ic
a
l
a
ppl
ia
n
c
e
[
13]
.
T
hi
s
m
e
th
od
pl
ot
s
th
e
gr
a
ph
s
o
f
vol
ta
ge
a
nd
c
ur
r
e
nt
ove
r
one
c
yc
le
a
nd
th
e
n
r
e
li
e
s
on
th
e
s
ha
pe
s
of
gr
a
phs
to
a
na
ly
z
e
a
nd
c
la
s
s
if
y
e
le
c
tr
ic
a
l
de
vi
c
e
s
.
A
ppl
yi
ng
th
e
V
-
I
t
r
a
je
c
to
r
y
m
e
th
od
i
n pr
a
c
ti
c
e
, B
a
e
ts
e
t
al
.
us
e
d c
onvolut
io
na
l
ne
ur
a
l
ne
twor
k
s
t
o a
na
ly
z
e
V
-
I
t
r
a
je
c
to
r
y i
m
a
ge
s
[
14]
.
B
e
c
a
us
e
th
e
a
ppl
ic
a
ti
on
of
m
a
c
hi
ne
le
a
r
ni
ng
a
lg
or
it
hm
s
in
N
A
L
M
is
ve
r
y
pot
e
nt
ia
l,
s
om
e
pa
pe
r
s
e
va
lu
a
te
d
th
e
pe
r
f
or
m
a
nc
e
of
s
om
e
m
a
c
hi
n
e
le
a
r
ni
ng
m
ode
ls
in
th
e
N
A
L
M
a
ppl
ic
a
ti
on
[
11]
,
[
15]
.
K
ol
te
r
a
nd
M
a
tt
he
w
ha
ve
bui
lt
a
m
a
s
s
iv
e
da
ta
b
a
s
e
f
or
th
e
de
ve
lo
pm
e
nt
of
id
e
nt
if
ic
a
ti
on
a
lg
or
i
th
m
s
[
16]
.
T
he
da
ta
ba
s
e
in
c
lu
de
s
in
f
or
m
a
ti
on
a
bout
th
e
e
ne
r
gy
c
ons
um
pt
io
n
of
m
a
ny
de
vi
c
e
s
in
1
0
hom
e
s
f
or
19
da
y
s
,
w
it
h
a
to
ta
l
da
ta
c
a
pa
c
it
y
of
up
to
1
te
r
a
byt
e
of
r
a
w
da
ta
.
I
n
c
ont
r
a
s
t,
to
r
e
duc
e
m
a
nua
l
la
be
li
ng
da
ta
,
K
ha
le
d
C
ha
hi
ne
de
ve
lo
pe
d a
s
y
s
te
m
t
ha
t
c
a
n
e
xt
r
a
c
t
th
e
s
ig
na
tu
r
e
s
a
nd l
a
be
l
th
e
m
a
ut
om
a
ti
c
a
ll
y
[
17]
.
D
ue
to
th
e
ha
r
dw
a
r
e
a
nd
s
of
twa
r
e
c
om
pl
e
xi
ty
in
th
e
N
A
L
M
a
ppl
ic
a
ti
on,
S
e
m
w
a
l
a
nd
P
r
a
s
a
f
oc
us
e
d
on
opt
im
iz
a
ti
on
a
lg
or
it
hm
s
us
in
g
m
in
im
um
f
e
a
tu
r
e
s
f
r
om
s
m
a
r
t
m
e
te
r
s
[
18]
.
I
ks
a
n
e
t
al
.
pr
opos
e
d
a
s
m
oot
hi
ng me
th
od f
or
f
il
te
r
in
g out pe
a
k s
ig
na
ls
[
19]
. T
he
s
ys
te
m
a
c
hi
e
ve
d be
tt
e
r
a
c
c
ur
a
c
y w
it
h t
hi
s
m
e
th
od.
I
ns
te
a
d
of
us
in
g
e
le
c
tr
ic
m
e
te
r
s
,
L
a
put
e
t
al
.
u
s
e
d
onl
y
a
pr
in
te
d
c
ir
c
ui
t
boa
r
d
(
P
C
B
)
w
it
h
m
ul
ti
pl
e
s
e
ns
or
s
[
20]
.
T
hi
s
bo
a
r
d
c
a
n
r
e
c
ogni
z
e
not
onl
y
e
le
c
tr
ic
a
l
de
vi
c
e
s
but
a
ls
o
in
door
a
c
ti
vi
ti
e
s
s
uc
h
a
s
ope
ni
ng/
c
lo
s
in
g
door
s
,
r
e
m
ovi
ng
ti
s
s
ue
pa
pe
r
,
a
nd
dr
a
in
in
g
t
he
f
a
uc
e
t.
I
n
pa
r
ti
c
ul
a
r
,
m
os
t
e
ve
nt
s
ha
ve
a
s
ig
ni
f
ic
a
nt
im
pa
c
t
on
th
e
m
ic
r
ophone
a
nd
th
e
a
c
c
e
le
r
om
e
te
r
.
T
he
te
s
t
w
a
s
c
onduc
te
d
in
m
a
ny
pl
a
c
e
s
a
nd
a
c
hi
e
ve
d a
n a
c
c
ur
a
c
y of
96%
.
A
l
ot
of
r
e
s
e
a
r
c
h a
bout
de
vi
c
e
s
ta
te
r
e
c
ogni
ti
on a
lr
e
a
dy publi
s
h
e
d. T
hos
e
r
e
s
e
a
r
c
he
s
m
a
in
ly
f
oc
us
on
ne
w
id
e
nt
if
ie
r
s
to
s
ol
ve
pr
obl
e
m
s
th
a
t
e
xi
s
t
w
he
n
u
s
in
g
ol
d
id
e
nt
if
ie
r
s
.
I
n
th
is
pa
pe
r
,
c
om
m
on
e
le
c
tr
ic
a
l
pa
r
a
m
e
te
r
s
a
nd
th
e
M
L
P
ne
twor
k
m
ode
l
a
r
e
u
s
e
d
to
id
e
nt
if
y
t
he
s
ta
te
s
of
e
le
c
tr
ic
a
l
a
p
pl
ia
nc
e
s
.
B
e
s
id
e
s
,
th
e
A
M
D
w
it
h
a
m
ic
r
ophone
is
a
ls
o
us
e
d
f
o
r
th
e
pu
r
pos
e
of
s
up
por
ti
ng
th
e
id
e
nt
i
f
ic
a
ti
on
of
e
le
c
tr
ic
a
l
de
vi
c
e
s
w
it
h t
he
s
a
m
e
pow
e
r
or
c
ont
in
uous
va
r
yi
ng powe
r
ove
r
t
im
e
.
3.
T
H
E
S
C
O
P
E
A
N
D
N
O
V
E
L
T
Y
O
F
T
H
E
P
A
P
E
R
T
he
a
ut
hor
s
c
onduc
te
d
th
e
e
xpe
r
im
e
nt
in
a
pr
iv
a
te
hou
s
e
in
s
te
a
d
of
a
la
r
ge
bui
ld
in
g.
W
e
do
not
pe
r
f
or
m
th
e
id
e
nt
i
f
ic
a
ti
on
of
e
le
c
tr
ic
a
l
a
ppl
ia
nc
e
s
w
hos
e
pow
e
r
s
c
ha
nge
c
ont
in
uous
ly
ove
r
ti
m
e
.
T
he
A
M
D
w
it
h a
m
ic
r
ophone
i
s
c
a
pa
bl
e
of
a
na
ly
z
in
g
a
c
ous
ti
c
noi
s
e
s
f
r
om
r
unni
ng a
ppl
ia
nc
e
s
t
o di
f
f
e
r
e
nt
ia
te
t
he
m
. T
hi
s
de
vi
c
e
ha
s
be
e
n
bui
lt
to
s
uppor
t
L
M
D
to
r
e
a
li
z
e
ti
m
e
-
va
r
yi
n
g
pow
e
r
a
s
w
e
ll
a
s
s
im
il
a
r
pow
e
r
a
ppl
ia
nc
e
s
.
H
ow
e
ve
r
, i
n t
hi
s
pa
pe
r
, w
e
ha
v
e
not
de
ve
lo
pe
d a
lg
or
it
hm
s
t
o c
o
m
bi
ne
da
ta
of
L
M
D
w
it
h A
M
D
.
W
it
h
th
e
pur
pos
e
of
de
ve
lo
pi
ng
a
n
a
c
c
e
s
s
ib
le
,
e
a
s
y
-
to
-
us
e
,
a
n
d
lo
w
-
c
os
t
m
e
a
s
ur
e
m
e
nt
de
vi
c
e
,
w
e
ha
ve
bui
lt
th
e
L
M
D
ba
s
e
d
on
ope
n
-
s
our
c
e
pl
a
tf
or
m
s
(
bot
h
ha
r
dw
a
r
e
a
nd
s
of
twa
r
e
)
.
T
h
e
ha
r
dw
a
r
e
s
c
he
m
a
ti
c
s
a
r
e
r
e
la
ti
ve
ly
s
im
pl
e
,
w
it
h
onl
y
vol
ta
ge
a
nd
c
ur
r
e
nt
m
e
a
s
ur
e
m
e
nt
c
ha
nne
ls
.
T
he
A
r
dui
no
s
of
twa
r
e
l
ib
r
a
r
y
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
11
, N
o.
1
,
M
a
r
c
h 20
22
:
300
-
309
302
e
xt
r
e
m
e
ly
a
c
c
e
s
s
ib
le
to
ne
w
c
om
e
r
s
.
T
he
r
e
f
or
e
,
it
is
m
uc
h
le
s
s
ti
m
e
-
c
ons
um
in
g
f
or
ne
w
c
om
e
r
s
to
be
f
a
m
il
ia
r
w
it
h t
he
de
vi
c
e
. A
ls
o, t
he
y c
a
n e
a
s
il
y c
us
to
m
iz
e
t
he
h
a
r
dw
a
r
e
a
nd de
ve
lo
p ne
w
f
e
a
tu
r
e
s
f
or
t
he
de
vi
c
e
.
A
s
oppos
e
d
to
ot
he
r
m
e
nt
io
ne
d
pa
pe
r
s
in
s
e
c
ti
on
2,
w
he
n
th
e
a
ut
hor
s
m
a
de
m
a
ny
e
f
f
or
ts
to
f
ig
u
r
e
out
ne
w
id
e
nt
if
yi
ng
c
ha
r
a
c
te
r
is
ti
c
s
f
r
om
a
ppl
ia
nc
e
s
,
w
e
us
e
onl
y
s
om
e
ba
s
ic
e
le
c
tr
ic
a
l
pa
r
a
m
e
te
r
s
(
e
.g.,
vol
ta
ge
,
c
ur
r
e
nt
,
a
nd
pow
e
r
)
th
a
t
c
a
n
be
obt
a
in
e
d
w
it
h
th
e
s
im
pl
e
ha
r
dw
a
r
e
to
r
e
c
ogni
z
e
hous
e
hol
d
a
ppl
ia
nc
e
s
.
A
s
de
m
on
s
tr
a
te
d
in
[
20]
,
m
os
t
e
ve
nt
s
ha
ve
a
n
im
pa
c
t
s
ig
ni
f
ic
a
nt
ly
on
s
ound
a
nd
vi
br
a
ti
o
n
s
e
ns
or
s
.
T
hus
,
in
s
te
a
d
of
us
in
g
a
dva
nc
e
d
e
l
e
c
tr
ic
a
l
pa
r
a
m
e
te
r
s
(
e
.g.,
ha
r
m
oni
c
s
,
noi
s
e
s
on
th
e
li
ne
)
th
a
t
r
e
qui
r
e
c
om
pl
ic
a
te
d
ha
r
dw
a
r
e
,
w
e
onl
y
a
dd
th
e
A
M
D
w
it
h
t
he
m
ic
r
ophone
to
s
uppor
t
L
M
D
to
r
e
c
ogni
z
e
a
ppl
ia
nc
e
s
.
4.
P
R
O
P
O
S
E
D
S
Y
S
T
E
M
D
E
S
I
G
N
4.1. Op
e
n
-
s
ou
r
c
e
d
e
s
ig
n
c
on
c
e
p
t
s
T
he
c
onc
e
pt
of
ope
n
-
s
our
c
e
ha
r
dw
a
r
e
w
a
s
f
ir
s
t
r
e
le
a
s
e
d
by
B
r
uc
e
P
e
r
e
ns
in
1997.
O
pe
n
-
s
our
c
e
ha
r
dw
a
r
e
m
a
ke
s
it
e
a
s
y
f
or
e
ve
r
yone
to
de
s
ig
n
a
nd
de
ve
lo
p
ne
w
f
e
a
tu
r
e
s
f
or
th
e
ir
ow
n
p
r
oj
e
c
ts
.
I
n
a
ddi
ti
on,
ope
n
-
s
our
c
e
s
of
twa
r
e
is
s
a
id
to
be
hi
ghl
y
r
e
li
a
bl
e
be
c
a
us
e
m
a
n
y
c
om
m
uni
ti
e
s
a
r
e
in
vol
ve
d
in
de
bugging
a
nd
te
s
ti
ng.
A
s
a
r
e
s
ul
t,
ope
n
-
s
our
c
e
w
il
l
m
a
ke
it
f
a
s
te
r
a
nd
e
a
s
ie
r
f
or
pe
opl
e
to
c
onduc
t
th
e
ir
pr
oj
e
c
ts
.
T
ha
n
ks
to
th
e
s
e
r
e
a
s
on
s
, ope
n
-
s
our
c
e
bo
a
r
ds
a
r
e
a
ls
o of
gr
e
a
t
u
s
e
f
or
e
duc
a
ti
on
[
21]
, [
22]
.
A
r
dui
no
is
a
c
om
pa
ny
th
a
t
s
pe
c
ia
li
z
e
s
in
pr
ovi
di
ng
boa
r
ds
w
it
h
ope
n
s
our
c
e
in
bot
h
ha
r
dw
a
r
e
a
n
d
s
of
twa
r
e
.
A
r
dui
no
pr
ovi
de
s
a
n
e
a
s
y
-
to
-
us
e
in
te
gr
a
te
d
de
v
e
lo
pm
e
nt
e
nvi
r
onm
e
nt
(
I
D
E
)
th
a
t
c
ont
a
in
s
m
a
ny
s
im
pl
e
f
unc
ti
ons
a
nd
li
br
a
r
ie
s
in
C
+
+
la
ngua
g
e
s
.
I
n
th
is
p
a
pe
r
,
in
or
de
r
to
ha
ve
a
n
e
a
s
y
-
to
-
c
us
to
m
iz
e
m
e
a
s
ur
in
g de
vi
c
e
, w
e
de
ve
lo
pe
d a
di
gi
ta
l
m
e
te
r
, L
M
D
, ba
s
e
d on the
A
r
dui
no D
ue
boa
r
d.
S
im
il
a
r
ly
, a
popula
r
ope
n
ha
r
dw
a
r
e
de
vi
c
e
is
S
he
nz
he
n
X
unl
ong
S
of
twa
r
e
'
s
O
r
a
nge
P
i
Z
e
r
o
boa
r
d.
T
hi
s
boa
r
d
us
e
s
a
hi
gh
-
pe
r
f
or
m
a
n
c
e
A
R
M
®
C
or
te
x™
-
A
7
m
ic
r
opr
oc
e
s
s
or
,
s
o
A
M
D
us
e
s
th
is
boa
r
d
to
a
na
ly
z
e
s
ound
a
nd
di
f
f
e
r
e
nt
ia
te
home
e
le
c
tr
ic
a
l
a
ppl
ia
nc
e
s
.
4.2. S
ys
t
e
m
ar
c
h
it
e
c
t
u
r
e
A
s
s
how
n
in
F
ig
ur
e
1,
th
e
r
e
a
r
e
two
ty
pe
s
of
m
o
ni
to
r
in
g
de
v
ic
e
s
:
L
M
D
a
nd
A
M
D
.
T
h
e
L
M
D
is
pr
ogr
a
m
m
e
d
t
o
m
e
a
s
ur
e
e
le
c
tr
ic
a
l
p
a
r
a
m
e
te
r
s
w
it
h
a
c
yc
l
e
of
on
e
s
e
c
ond
.
L
M
D
pa
s
s
e
s
t
he
s
e
d
a
ta
to
a
c
om
put
e
r
to
r
un
m
a
c
hi
ne
l
e
a
r
ni
n
g
a
lg
or
it
hm
s
th
a
t
pr
e
di
c
t
th
e
s
ta
t
e
of
e
le
c
tr
ic
a
l
a
ppl
ia
n
c
e
s
.
I
n
th
e
f
ol
lo
w
in
g
s
e
c
ti
on
s
of
th
e
p
a
pe
r
,
th
e
a
ut
hor
s
w
i
ll
a
na
l
yz
e
pr
obl
e
m
s
in
u
s
i
ng
L
M
D
to
id
e
nt
if
y
e
le
c
tr
ic
a
l
a
ppl
i
a
nc
e
s
.
T
h
e
f
ir
s
t
pr
obl
e
m
i
s
to
di
s
t
in
gui
s
h
de
vi
c
e
s
w
it
h
s
im
il
a
r
pow
e
r
c
ons
um
pt
io
n,
a
nd
th
e
s
e
c
ond
i
s
to
di
s
ti
ngui
s
h
de
vi
c
e
s
w
it
h
va
r
yi
ng
pow
e
r
ove
r
ti
m
e
.
T
h
e
r
e
f
or
e
,
A
M
D
e
qu
ip
m
e
nt
w
it
h
th
e
a
udi
o
s
e
ns
or
i
s
us
e
d
to
ov
e
r
c
om
e
th
e
s
e
pr
ob
le
m
s
i
n
th
e
f
u
tu
r
e
.
F
ig
ur
e
1
. T
he
pr
opos
e
d s
y
s
te
m
m
ode
l
5.
L
O
A
D
M
O
N
I
T
O
R
I
N
G
D
E
V
I
C
E
(
L
M
D
)
5.1. Har
d
w
ar
e
s
t
r
u
c
t
u
r
e
an
d
m
e
a
s
u
r
e
m
e
n
t
p
r
ogr
am
F
ig
ur
e
2
s
how
s
th
e
bl
oc
k
di
a
gr
a
m
of
th
e
L
M
D
.
B
e
c
a
us
e
th
e
L
M
D
is
in
te
gr
a
te
d
in
to
th
e
e
le
c
tr
ic
a
l
pa
ne
l,
L
M
D
a
r
e
pow
e
r
e
d
by
th
e
220
V
A
C
gr
id
.
T
he
L
M
D
m
e
a
s
ur
e
m
e
nt
c
ir
c
ui
t
in
c
lu
de
s
th
e
vol
ta
ge
a
nd
c
ur
r
e
nt
m
e
a
s
ur
in
g
c
ha
nn
e
l.
T
h
e
s
c
he
m
a
ti
c
s
of
th
e
s
e
m
e
a
s
ur
i
ng
c
ha
nne
ls
a
r
e
bui
lt
b
a
s
e
d
on
th
e
r
e
f
e
r
e
nc
e
de
s
ig
n
of
m
ic
r
oc
hi
p
te
c
hn
ol
ogy
[
23]
,
[
24]
.
S
im
il
a
r
ly
,
a
s
m
e
nt
io
ne
d
a
bove
,
th
e
A
r
dui
no
D
ue
boa
r
d
is
u
s
e
d
a
s
th
e
c
e
nt
r
a
l
pr
oc
e
s
s
in
g
uni
t
of
th
e
L
M
D
.
T
hi
s
bl
oc
k
w
il
l
pr
oc
e
s
s
th
e
s
ig
na
l
s
f
r
om
th
e
vol
ta
ge
a
nd
c
ur
r
e
nt
c
ha
nne
ls
,
di
s
pl
a
y
th
e
c
a
lc
ul
a
te
d
pa
r
a
m
e
te
r
s
on
th
e
li
ne
a
r
c
om
pl
e
m
e
nt
a
r
y
dua
l
(
L
C
D
)
,
s
e
nd
th
e
m
to
th
e
c
om
put
e
r
vi
a
W
i
-
F
i,
a
nd
s
to
r
e
th
e
e
ne
r
gy
c
ons
um
e
d
on
th
e
e
le
c
tr
ic
a
ll
y
e
r
a
s
a
bl
e
pr
ogr
a
m
m
a
bl
e
r
e
a
d
-
onl
y
m
e
m
or
y
(
E
E
P
R
O
M
)
.
L
M
D
c
a
n
m
e
a
s
ur
e
s
ix
e
le
c
tr
ic
a
l
pa
r
a
m
e
t
e
r
s
,
w
hi
c
h
a
r
e
vol
ta
ge
(
U
r
m
s
)
,
c
ur
r
e
nt
(
I
r
m
s
)
,
a
c
ti
ve
pow
e
r
(
P
)
,
r
e
a
c
ti
ve
pow
e
r
(
Q
)
,
pow
e
r
f
a
c
to
r
(
c
os
φ)
,
a
n
d
e
ne
r
g
y
c
on
s
um
pt
io
n
(
E
)
.
U
r
m
s
a
nd
I
r
m
s
a
r
e
c
a
lc
ul
a
te
d by (
1)
, a
s
s
how
n i
n F
ig
ur
e
2.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
M
ac
hi
ne
l
e
ar
ni
ng algor
it
hm
s
f
or
…
(
V
ie
t
H
oang Duong
)
303
F
ig
ur
e
2
. T
he
ha
r
dw
a
r
e
bl
oc
k di
a
gr
a
m
of
L
M
D
=
√
∑
2
(
)
−
1
=
0
;
=
√
∑
2
(
)
−
1
=
0
(
1)
A
c
ti
ve
pow
e
r
a
nd
r
e
a
c
ti
ve
pow
e
r
a
r
e
c
a
l
c
ul
a
te
d
f
r
om
u(
n)
a
n
d
i(
n)
,
a
s
s
how
n
in
(
2)
.
W
he
r
e
u(
n)
,
i(
n)
,
a
nd
i
90
-
de
gr
e
e
-
s
hi
f
t
(
n)
a
r
e
in
s
ta
nt
a
ne
ous
vol
ta
ge
,
in
s
ta
nt
a
ne
ou
s
c
ur
r
e
nt
,
a
nd
in
s
ta
nt
a
n
e
ous
c
ur
r
e
nt
s
hi
f
te
d
by
90
de
gr
e
e
s
.
N
is
th
e
num
b
e
r
of
s
a
m
pl
e
s
.
T
he
s
a
m
pl
i
ng
f
r
e
que
nc
y
is
2000Hz
,
th
e
m
e
a
s
ur
e
m
e
nt
da
ta
upda
t
e
r
a
te
i
s
1H
z
, s
o
N
=
2000.
=
∑
[
(
)
×
(
)
]
=
1
;
=
∑
[
(
)
×
90
−
−
ℎ
(
)
]
=
1
(
2)
5.2. Ap
p
li
an
c
e
s
t
at
e
s
d
e
t
e
c
t
in
g al
gor
it
h
m
I
n
r
e
c
e
nt
ye
a
r
s
,
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nt
(
A
I
)
ha
s
be
c
om
e
a
ph
e
nom
e
non
in
th
e
w
or
ld
.
M
a
c
hi
ne
le
a
r
ni
ng
(
M
L
)
is
a
s
ubs
e
t
of
A
I
.
M
a
c
hi
ne
le
a
r
ni
ng
is
tr
a
di
ti
ona
ll
y
di
vi
de
d
in
to
th
r
e
e
m
a
in
gr
oups
:
s
upe
r
vi
s
e
d
le
a
r
ni
ng,
uns
upe
r
vi
s
e
d
le
a
r
ni
ng,
a
nd
r
e
in
f
or
c
e
m
e
nt
le
a
r
ni
ng.
I
n
th
is
s
ys
te
m
,
w
e
c
om
bi
ne
th
e
s
up
e
r
vi
s
e
d
le
a
r
ni
ng
m
e
th
od
w
it
h
a
th
r
e
e
-
la
ye
r
M
L
P
to
tr
a
in
th
e
s
ys
te
m
in
r
e
a
li
z
in
g t
he
s
ta
te
of
e
le
c
tr
ic
a
l
a
ppl
ia
n
c
e
s
. A
s
s
ho
w
n
in
F
ig
ur
e
3
a
nd
(
3)
,
th
e
in
put
,
hi
dde
n,
a
nd
out
put
la
ye
r
s
a
r
e
r
e
p
r
e
s
e
nt
e
d
by
ve
c
to
r
s
x
,
a,
a
nd
y,
r
e
s
pe
c
ti
ve
ly
.
M
a
tr
ix
w
e
ig
ht
W
a
nd bia
s
b
r
e
p
r
e
s
e
nt
c
onn
e
c
ti
ons
be
tw
e
e
n t
he
two la
ye
r
s
.
F
ig
ur
e
3
.
T
hr
e
e
l
a
ye
r
s
m
ul
ti
la
ye
r
pe
r
c
e
pt
r
on (
M
L
P
)
=
[
1
2
.
.
.
]
;
(
)
=
[
1
(
)
2
(
)
.
.
.
]
;
=
[
1
2
.
.
.
]
;
=
[
11
(
)
12
(
)
.
.
.
21
(
)
22
(
)
.
.
.
.
.
.
.
.
.
(
)
]
;
=
[
1
(
)
2
(
)
.
.
.
]
(
3)
A
s
de
m
o
ns
tr
a
te
d
in
(
4)
d
e
s
c
r
ib
e
th
e
r
e
l
a
ti
on
s
hi
p
be
twe
e
n
t
he
in
put
a
nd
o
ut
put
la
y
e
r
.
W
he
r
e
l
is
la
ye
r
num
be
r
,
l
=
1,
2,
…
,
L
(
L
is
th
e
la
s
t
l
a
ye
r
,
a
(
0
)
i
s
in
put
v
e
c
t
or
,
a
(
L
)
is
out
put
ve
c
to
r
)
,
g(
z
)
i
s
a
c
ti
v
a
ti
on
f
unc
ti
on
w
hi
c
h
is
s
ig
m
o
id
f
unc
ti
on
s
ho
w
n i
n (
5)
.
(
)
=
(
)
(
−
1
)
+
(
)
;
(
)
=
(
(
)
)
(
4)
(
)
=
1
1
+
−
(
5)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
11
, N
o.
1
,
M
a
r
c
h 20
22
:
300
-
309
304
F
ir
s
t,
w
e
m
us
t
de
te
r
m
in
e
th
e
c
om
pone
nt
s
of
th
e
in
put
v
e
c
to
r
x
a
nd
th
e
out
put
ve
c
to
r
y
.
I
n
th
i
s
c
a
s
e
,
th
e
in
put
ve
c
to
r
is
th
e
e
le
c
tr
ic
a
l
pa
r
a
m
e
te
r
s
obt
a
in
e
d
f
r
o
m
th
e
L
M
D
.
T
he
out
put
ve
c
to
r
is
th
e
on
/o
f
f
s
ta
te
s
of
e
a
c
h
de
vi
c
e
.
T
h
e
ne
xt
s
te
p
is
to
de
te
r
m
in
e
th
e
opt
im
um
W
a
n
d
b
m
a
tr
ic
e
s
to
s
how
th
e
r
e
la
ti
on
s
hi
p
be
twe
e
n
th
e
i
nput
a
nd output
ve
c
to
r
s
. W
it
h t
he
s
upe
r
vi
s
e
d t
r
a
in
in
g m
e
th
od, a
t
r
a
in
in
g da
ta
s
e
t
is
us
e
d t
o f
in
d
th
e
s
e
t
w
o
m
a
tr
ic
e
s
. T
he
t
e
s
t
da
ta
s
e
t
is
t
he
n u
s
e
d t
o t
e
s
t
th
e
pr
e
di
c
ti
ve
a
c
c
ur
a
c
y of
t
he
s
ys
te
m
.
A
c
ti
ve
pow
e
r
P
a
nd
r
e
a
c
ti
ve
pow
e
r
Q
a
r
e
u
s
e
d
a
s
“
S
ig
na
tu
r
e
”
f
or
e
a
c
h
de
vi
c
e
.
T
a
bl
e
1
s
how
s
th
e
pow
e
r
of
s
om
e
i
ndoor
a
ppl
ia
nc
e
s
t
o be
pr
e
di
c
te
d i
n t
he
e
xp
e
r
im
e
nt
s
e
c
ti
on. W
e
c
a
n s
e
e
f
r
om
t
he
t
a
bl
e
t
ha
t
th
e
de
vi
c
e
s
ha
v
e
di
f
f
e
r
e
nt
P
a
nd Q
.
O
bs
e
r
vi
ng
th
e
lo
a
d
gr
a
ph
in
F
ig
u
r
e
4,
e
a
c
h
ti
m
e
a
de
vi
c
e
is
tu
r
ne
d
on
or
of
f
,
a
r
is
in
g
or
f
a
ll
in
g
e
dge
a
ppe
a
r
s
on
th
e
gr
a
ph.
T
he
a
m
pl
it
ude
of
th
e
e
dge
is
th
e
pow
e
r
c
ons
um
pt
io
n
of
th
e
de
vi
c
e
th
a
t
h
a
s
be
e
n
tu
r
ne
d
on/
of
f
.
W
e
de
ve
lo
pe
d
a
n
a
lg
or
it
hm
c
a
ll
e
d
“
e
dge
de
te
c
ti
on”
to
id
e
nt
if
y
th
e
pot
e
nt
ia
l
e
dge
s
.
E
ve
n
w
he
n
n
o
a
ppl
ia
nc
e
s
a
r
e
t
ur
ne
d on/of
f
, t
he
t
ot
a
l
pow
e
r
i
s
c
ons
ta
nt
ly
c
h
a
n
gi
ng. T
hus
, i
t
is
ne
c
e
s
s
a
r
y t
o de
f
in
e
a
t
hr
e
s
hol
d
va
lu
e
l
a
r
ge
e
nough f
or
t
he
a
lg
or
it
hm
t
o e
li
m
in
a
te
t
he
s
e
f
lu
c
tu
a
ti
ons
.
I
n t
hi
s
pa
pe
r
, t
he
os
c
il
la
ti
on
th
r
e
s
hol
d
f
or
P
a
nd Q
a
r
e
15W
a
nd 8VA
r
, r
e
s
pe
c
ti
ve
ly
.
T
a
bl
e
1
.
S
pe
c
if
ic
a
ti
ons
of
s
om
e
e
le
c
tr
ic
a
l
de
vi
c
e
s
u
s
e
d f
or
t
he
e
xpe
r
im
e
nt
No
A
ppl
i
a
nc
e
R
e
a
l
P
ow
e
r
(
W
)
R
e
a
c
t
i
ve
P
ow
e
r
(
V
A
r
)
P
ow
e
r
F
a
c
t
or
(
c
os
φ
)
1
H
a
i
r
dr
ye
r
(
m
ode
1)
455
13
0.99
2
H
a
i
r
dr
ye
r
(
m
ode
2)
893
31
0.99
3
K
e
t
t
l
e
1
1374
5
0.99
4
K
e
t
t
l
e
2
1958
50
0.99
5
L
E
D
l
a
m
p
22
11
0.89
6
C
om
pa
c
t
l
a
m
p
65
-
8
0.99
7
F
a
n (
w
i
t
h e
l
e
c
t
r
oni
c
c
i
r
c
ui
t
)
45
-
12
0.96
8
I
nc
a
nde
s
c
e
nt
l
a
m
p
60
0
1.0
9
C
ha
nde
l
i
e
r
202
0
1.0
10
H
e
a
t
i
ng l
a
m
p (
m
ode
1)
260
0
1.0
11
H
e
a
t
i
ng l
a
m
p (
m
ode
2)
526
0
1.0
12
F
l
uor
e
s
c
e
nt
l
a
m
p
30
69
0.4
13
H
e
a
t
i
ng ba
g
730
0
1.0
F
ig
ur
e
4
.
R
is
in
g a
nd f
a
ll
in
g e
dge
w
he
n t
ur
ni
ng on/of
f
t
he
f
a
n
W
e
ha
ve
t
he
f
ol
lo
w
in
g i
nput
ve
c
to
r
of
t
he
M
L
P
ne
twor
k:
=
[
]
(
6)
A
to
ta
l
of
12
de
vi
c
e
s
a
r
e
us
e
d
dur
in
g
th
e
te
s
t,
in
w
hi
c
h
ha
ir
-
dr
ye
r
a
nd
he
a
ti
ng
la
m
p
de
vi
c
e
s
ha
ve
m
ul
ti
pl
e
ope
r
a
ti
ng
m
ode
s
.
T
hus
,
th
e
32
out
c
om
e
s
a
r
e
r
e
pr
e
s
e
nt
e
d
by
ve
c
to
r
s
,
a
s
s
how
n
in
T
a
bl
e
s
2
a
nd
3.
A
s
s
how
n
in
(
7)
s
pe
c
if
ic
a
ll
y
de
s
c
r
ib
e
s
a
pa
ir
of
in
put
ve
c
to
r
s
x
(
1)
a
nd
out
put
ve
c
to
r
s
y
(
1)
.
W
he
r
e
m
is
th
e
num
be
r
of
da
ta
poi
nt
s
c
ol
le
c
te
d,
th
e
m
a
tr
ic
e
s
X
a
nd
Y
a
r
e
th
e
pr
oduc
t
of
c
om
bi
ni
ng
a
ll
m
in
put
a
nd
out
put
ve
c
to
r
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
M
ac
hi
ne
l
e
ar
ni
ng algor
it
hm
s
f
or
…
(
V
ie
t
H
oang Duong
)
305
(
X
ϵ
ℝ
2×
m
,
Y
ϵ
ℝ
32×
m
)
.
T
he
s
e
da
ta
a
r
e
di
vi
de
d
in
to
two
gr
oups
a
s
th
e
tr
a
in
in
g
s
e
t
a
nd
t
e
s
t
s
e
t.
T
r
a
in
in
g
s
e
t
is
us
e
d t
o t
r
a
in
t
he
M
L
P
m
ode
l.
T
he
T
e
s
t
S
e
t
is
us
e
d t
o e
v
a
lu
a
te
t
he
pr
e
di
c
ti
on a
c
c
ur
a
c
y of
t
he
M
L
P
m
ode
l
a
f
te
r
it
ha
s
be
e
n
tr
a
in
e
d.
T
he
goa
l
of
th
e
t
r
a
in
in
g
pr
oc
e
s
s
is
to
f
in
d
two
opt
im
a
l
m
a
t
r
ic
e
s
W
a
nd
b,
s
o
th
a
t
th
e
pr
e
di
c
te
d
out
pu
ts
a
ppr
oxi
m
a
te
th
e
a
c
tu
a
l
out
put
s
.
T
he
di
f
f
e
r
e
nc
e
be
twe
e
n
th
e
s
e
two
out
put
s
i
s
e
va
lu
a
te
d
us
in
g (
8)
(
th
e
e
r
r
or
f
unc
ti
on)
.
T
a
bl
e
2
. O
ut
put
ve
c
to
r
y
O
ut
put
1
2
3
…
32
y
(
O
ut
put
ve
c
t
or
)
[
1
0
0
.
.
.
0
]
[
0
1
0
.
.
.
0
]
[
0
0
1
.
.
.
0
]
…
[
0
0
0
.
.
.
1
]
T
a
bl
e
3
. O
ut
put
ve
c
to
r
a
nd de
vi
c
e
s
ta
te
s
O
ut
put
S
t
a
t
e
1
H
a
i
r
dr
ye
r
i
s
on (
m
ode
1)
2
H
a
i
r
dr
ye
r
i
s
of
f
(
m
ode
1
)
…
…
32
H
e
a
t
i
ng ba
g i
s
of
f
(
1
)
=
[
455
.
36
12
.
19
]
;
(
1
)
=
[
1
0
.
.
.
0
]
;
=
[
|
|
…
(
1
)
(
2
)
.
.
.
|
|
…
]
;
=
[
|
|
…
(
1
)
(
2
)
.
.
.
|
|
…
]
(
7)
=
1
∑
∑
−
(
)
(
)
−
(
1
−
(
)
)
(
1
−
(
)
)
=
1
=
1
(
8)
W
he
r
e
a
k
L
is
pr
e
di
c
te
d
out
put
k
of
th
e
out
put
la
ye
r
,
y
k
is
th
e
c
or
r
e
s
ponding
la
be
l
k
in
T
r
a
in
in
g
S
e
t,
K
is
th
e
num
be
r
of
uni
ts
of
th
e
out
put
la
ye
r
(
K
=
32)
,
m
is
th
e
to
ta
l
num
b
e
r
of
c
ol
le
c
te
d
da
ta
in
T
r
a
in
in
g
S
e
t.
W
it
h
c
os
t
f
unc
ti
on
J
,
th
e
va
lu
e
J
is
s
m
a
ll
w
he
n
a
k
L
≈
y
k
. T
he
pr
obl
e
m
to
be
s
ol
ve
d
now
is
to
f
in
d
th
e
m
in
im
um
va
lu
e
of
th
e
f
unc
ti
on
J
;
f
r
om
th
e
r
e
,
w
e
ha
ve
th
e
m
a
tr
ic
e
s
W
a
nd
b,
r
e
s
pe
c
ti
ve
ly
.
F
in
di
ng
th
e
m
in
im
um
va
lu
e
of
J
by
s
ol
vi
ng
a
s
s
how
n
in
(
8)
is
a
c
om
pl
ic
a
te
d
ta
s
k,
a
nd
th
us
G
r
a
di
e
nt
D
e
s
c
e
nt
a
nd
B
a
c
kw
a
r
d
pr
opa
ga
ti
on
a
r
e
two
us
e
f
ul
a
lg
or
it
hm
s
t
o s
ol
ve
t
hi
s
pr
obl
e
m
[
25]
.
O
ve
r
a
ll
,
th
e
pr
oc
e
s
s
of
r
e
c
ogni
z
in
g
th
e
on/
of
f
s
ta
te
s
of
e
le
c
tr
ic
a
l
de
vi
c
e
s
i
s
de
s
c
r
ib
e
d
in
F
ig
ur
e
5.
T
he
c
om
put
e
r
w
il
l
a
lwa
ys
c
ol
le
c
t
th
e
e
le
c
tr
ic
a
l
p
a
r
a
m
e
te
r
s
f
r
om
th
e
m
e
te
r
ove
r
W
i
-
F
i.
O
nc
e
a
ll
th
e
pa
r
a
m
e
te
r
s
ha
ve
be
e
n
r
e
c
e
iv
e
d,
th
e
c
om
put
e
r
r
uns
th
e
"
e
dge
d
e
te
c
ti
on"
a
lg
or
it
hm
to
de
te
c
t
th
e
c
ha
nge
s
of
P
a
nd
Q
.
I
f
it
de
te
c
ts
a
n
e
dg
e
∆
P
a
nd
∆
Q
,
th
e
c
om
put
e
r
r
uns
th
e
m
a
c
hi
ne
le
a
r
ni
ng
a
lg
or
it
hm
to
id
e
nt
if
y
w
hi
c
h
a
ppl
ia
nc
e
ha
s
ju
s
t
be
e
n
tu
r
ne
d
on
or
of
f
.
A
f
te
r
th
a
t,
th
e
c
om
put
e
r
w
il
l
r
e
tu
r
n
to
c
ol
le
c
t
ne
w
da
ta
f
r
om
th
e
m
e
te
r
.
F
ig
ur
e
5
. P
r
oc
e
s
s
of
i
de
nt
if
yi
ng e
le
c
tr
ic
a
l
a
ppl
ia
nc
e
s
ta
te
6.
A
C
T
I
V
I
T
Y
M
O
N
I
T
O
R
I
N
G
D
E
V
I
C
E
(
A
M
D
)
T
he
di
s
a
dva
nt
a
g
e
w
he
n
id
e
nt
if
yi
ng
de
vi
c
e
s
f
r
om
pow
e
r
P
a
nd
Q
is
th
a
t
it
is
c
ha
ll
e
ngi
ng
to
id
e
nt
if
y
de
vi
c
e
s
w
it
h
th
e
s
a
m
e
or
c
ont
in
uous
ly
va
r
yi
ng
pow
e
r
s
ov
e
r
ti
m
e
.
A
M
D
e
qui
pm
e
nt
is
d
e
ve
lo
pe
d
f
or
s
m
a
ll
-
s
c
a
le
in
s
ta
ll
a
ti
ons
s
ol
e
ly
to
a
ddr
e
s
s
th
e
two
pr
obl
e
m
s
m
e
nt
i
on
e
d
a
bove
.
A
c
c
or
di
ng
to
[
20
]
,
th
e
two
s
e
ns
or
s
w
it
h t
he
m
os
t
da
ta
c
ha
nge
e
ve
r
y t
im
e
a
n e
ve
nt
oc
c
ur
s
:
th
e
s
oun
d s
e
ns
or
a
nd t
he
a
c
c
e
le
r
om
e
te
r
. T
he
r
e
f
or
e
, t
he
a
ut
hor
s
us
e
a
m
ic
r
ophone
f
or
A
M
D
to
im
pl
e
m
e
nt
s
y
s
te
m
s
up
por
t
f
or
de
vi
c
e
r
e
c
ogni
ti
on.
I
n
th
is
pa
pe
r
,
th
e
a
ut
hor
s
onl
y t
e
s
t
A
M
D
f
unc
ti
ona
li
ty
but
ha
ve
not
c
om
bi
ne
d L
M
D
t
o i
de
nt
if
y a
ppl
ia
nc
e
s
.
T
he
ha
r
dw
a
r
e
s
tr
uc
tu
r
e
of
A
M
D
i
s
il
lu
s
tr
a
te
d
in
F
ig
ur
e
6.
D
ir
e
c
t
c
ur
r
e
nt
(
D
C
)
pow
e
r
is
c
onv
e
r
te
d
f
r
om
a
220
V
a
lt
e
r
na
ti
ng
c
ur
r
e
nt
(
A
C
)
pow
e
r
li
ne
to
s
uppl
y
th
e
O
r
a
nge
P
i
Z
e
r
o
boa
r
d,
a
m
ic
r
ophone
,
a
nd
a
di
s
pl
a
y.
O
r
a
nge
P
i
Z
e
r
o
r
uns
U
bunt
u
A
m
bi
a
n
ope
r
a
ti
ng
s
ys
te
m
a
nd
s
uppor
ts
P
yt
hon
la
ngua
ge
.
W
e
de
pl
oy
P
yt
hon’
s
li
br
a
r
ie
s
r
e
la
te
d
to
th
e
f
a
s
t
f
ou
r
ie
r
tr
a
ns
f
or
m
(
F
F
T
)
a
lg
or
it
hm
f
o
r
a
na
ly
z
in
g
s
ounds
f
r
om
e
le
c
tr
ic
a
l
a
ppl
ia
nc
e
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
11
, N
o.
1
,
M
a
r
c
h 20
22
:
300
-
309
306
F
ig
ur
e
6
. T
he
ha
r
dw
a
r
e
bl
oc
k di
a
gr
a
m
of
A
M
D
W
e
pe
r
f
or
m
e
d
a
n
a
c
ous
ti
c
a
na
ly
s
is
f
r
om
th
r
e
e
ty
pi
c
a
l
e
le
c
tr
ic
a
l
a
ppl
ia
nc
e
s
,
th
e
ha
ir
dr
ye
r
,
th
e
f
a
n,
a
nd
th
e
m
ic
r
ow
a
ve
ove
n.
F
a
n
noi
s
e
m
a
in
ly
c
om
e
s
f
r
om
ba
ll
be
a
r
in
gs
a
nd
pr
ope
ll
e
r
s
w
h
e
n
r
ot
a
ti
ng.
T
he
s
ound
f
r
om
th
e
ha
ir
dr
ye
r
is
ge
n
e
r
a
te
d
by
th
e
he
a
te
r
,
f
r
ont
a
nd
r
e
a
r
gr
id
s
,
a
nd
th
e
a
ir
f
il
te
r
.
T
he
m
ic
r
ow
a
ve
noi
s
e
c
om
e
s
f
r
om
th
e
c
ool
in
g
f
a
n
to
m
a
ke
s
ur
e
th
e
m
a
gne
tr
on
doe
s
not
h
e
a
t
up.
F
ig
ur
e
7
c
om
p
a
r
in
g
m
e
a
s
ur
e
m
e
nt
r
e
s
ul
ts
in
th
e
s
p
e
c
tr
um
of
F
ig
ur
e
7
(
a
)
f
a
n
(
57
-
58
H
z
)
a
nd
F
ig
ur
e
7
(
b
)
ha
ir
dr
ye
r
(
900
-
940
H
z
)
a
f
te
r
te
n
ti
m
e
s
of
r
e
c
or
di
ng
a
nd
a
na
ly
z
in
g
th
e
s
ound
of
th
r
e
e
de
vi
c
e
s
a
t
a
di
s
ta
nc
e
of
1
m
e
t
e
r
.
F
ig
ur
e
8
c
om
pa
r
in
g
m
e
a
s
ur
e
m
e
nt
r
e
s
ul
ts
in
th
e
s
pe
c
tr
um
of
F
ig
ur
e
8
(
a
)
m
ic
r
ow
a
ve
ove
n
(
195
-
210
H
z
)
a
nd
F
ig
ur
e
8
(
b)
a
ll
th
r
e
e
a
ppl
ia
nc
e
s
a
t
th
e
s
a
m
e
ti
m
e
.
I
t
i
s
c
le
a
r
t
ha
t
e
a
c
h
d
e
vi
c
e
'
s
f
e
a
tu
r
e
s
ti
ll
a
ppe
a
r
s
c
le
a
r
ly
w
he
n
a
ll
of
th
e
th
r
e
e
a
ppl
ia
nc
e
s
op
e
r
a
te
s
im
ul
ta
ne
ou
s
ly
.
T
he
r
e
f
or
e
,
th
is
f
e
a
tu
r
e
c
a
n
be
u
s
e
d
a
s
th
e
“
S
ig
na
tu
r
e
”
f
or
e
a
c
h
de
vi
c
e
.
(
a
)
(
b)
F
ig
ur
e
7
. S
pe
c
tr
um
s
of
(
a
)
f
a
n V
I
N
A
W
I
N
D
Q
B
300Đ a
nd
(
b)
h
a
ir
dr
ye
r
P
H
I
L
I
P
S
H
P
4840
(
a
)
(
b)
F
ig
ur
e
8
. S
pe
c
tr
um
s
of
(
a
)
m
ic
r
ow
a
ve
ove
n D
A
E
W
O
O
K
O
G
-
1
A
4H
a
nd
(
b)
th
r
e
e
a
ppl
ia
nc
es
7.
E
X
P
E
R
I
M
E
N
T
S
F
ig
ur
e
9
c
om
pa
r
in
g
de
vi
c
e
P
C
B
s
of
F
ig
ur
e
9
(
a
)
th
e
L
M
D
boa
r
d
a
nd
F
ig
ur
e
9
(
b)
A
M
D
boa
r
d.
T
he
two
boa
r
ds
a
r
e
ba
s
e
d
on
op
e
n
-
s
our
c
e
d
e
s
ig
ns
f
r
om
M
ic
r
oc
hi
p
/Ar
dui
no
a
nd
th
e
O
r
a
nge
P
i
Z
e
r
o.
L
M
D
c
a
n
m
e
a
s
ur
e
U
r
m
s
,
I
r
m
s
,
a
c
ti
ve
pow
e
r
P
,
r
e
a
c
ti
ve
pow
e
r
Q
,
po
w
e
r
f
a
c
to
r
,
a
nd
e
ne
r
gy
c
ons
um
pt
io
n
E
.
T
he
c
om
put
e
r
w
il
l
r
e
c
e
iv
e
th
e
s
e
p
a
r
a
m
e
te
r
s
f
r
om
th
e
L
M
D
vi
a
W
i
-
F
i
to
pe
r
f
or
m
th
e
a
lg
or
it
h
m
in
F
ig
ur
e
5.
T
he
a
c
c
ur
a
c
y of
bot
h volt
a
ge
a
nd c
ur
r
e
nt
c
ha
nne
l
s
i
s
unde
r
1%
a
f
te
r
be
in
g c
a
li
br
a
te
d.
T
he
a
ut
h
or
s
te
s
te
d
th
e
m
oni
to
r
in
g
s
ys
te
m
a
t
th
e
pr
iv
a
te
hou
s
e
,
in
c
lu
di
ng
th
e
ba
th
r
oom
,
la
undr
y
r
oom
,
a
nd
be
dr
oom
.
T
he
c
ur
r
e
nt
t
r
a
ns
f
or
m
e
r
o
f
th
e
e
le
c
tr
oni
c
m
e
te
r
w
a
s
in
s
ta
ll
e
d
to
m
e
a
s
ur
e
pow
e
r
s
im
ul
ta
ne
ous
ly
in
th
r
e
e
r
oom
s
.
T
he
num
be
r
of
d
e
vi
c
e
s
is
li
s
te
d
in
T
a
bl
e
1;
s
om
e
of
th
e
m
h
a
ve
m
ul
ti
pl
e
ope
r
a
ti
ng mode
s
s
uc
h a
s
ha
ir
dr
ye
r
a
nd he
a
ti
ng l
a
m
p, s
o a
t
ot
a
l
of
32 on/of
f
c
a
s
e
s
ne
e
ds
t
o be
r
e
c
ogni
z
e
d. T
he
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
M
ac
hi
ne
l
e
ar
ni
ng algor
it
hm
s
f
or
…
(
V
ie
t
H
oang Duong
)
307
e
xpe
r
im
e
nt
w
a
s
pe
r
f
or
m
e
d
on
w
e
e
kda
y
e
ve
ni
ngs
(
m
onda
y
to
f
r
id
a
y)
.
A
t
th
e
e
nd
of
th
e
w
e
e
k,
th
e
a
ut
hor
s
r
a
n
th
e
te
s
t
a
ll
da
y.
T
he
to
ta
l
dur
a
ti
on
of
th
e
e
xp
e
r
im
e
nt
w
a
s
t
hr
e
e
w
e
e
k
s
.
T
he
M
L
P
m
ode
l
w
a
s
tr
a
in
e
d
to
r
e
c
ogni
z
e
32
on/
of
f
c
a
s
e
s
be
f
or
e
c
onduc
ti
ng
th
e
e
xp
e
r
im
e
nt
.
A
to
ta
l
of
215
da
ta
poi
nt
s
w
e
r
e
c
ol
le
c
te
d.
T
he
70%
poi
nt
s
f
or
th
e
T
r
a
in
in
g
s
e
t
a
nd
30%
poi
nt
s
f
or
th
e
T
e
s
t
s
e
t.
A
f
te
r
th
e
e
nd
of
30000
it
e
r
a
ti
ons
,
th
e
va
lu
e
of
t
he
e
r
r
or
f
unc
ti
on J
i
s
0.295. T
he
pr
e
di
c
ti
on r
e
s
ul
t
is
93.65%
a
c
c
ur
a
te
on t
he
t
e
s
t
s
e
t.
A
f
te
r
t
hr
e
e
w
e
e
ks
, t
he
s
ys
te
m
pr
e
di
c
te
d
a
to
ta
l
of
766
e
ve
nt
s
,
of
w
hi
c
h
49
w
e
r
e
w
r
on
gl
y
pr
e
di
c
te
d.
A
s
a
r
e
s
ul
t,
th
e
s
y
s
te
m
a
c
hi
e
ve
d
a
pr
e
di
c
ti
on a
c
c
ur
a
c
y of
93.60%
.
F
ig
u
r
e
10 s
how
s
t
he
a
c
ti
ve
a
n
d r
e
a
c
ti
ve
pow
e
r
gr
a
ph du
r
in
g t
he
4
th
te
s
t
da
y.
A
f
te
r
te
s
ti
ng,
th
e
a
ut
hor
s
f
ound
s
om
e
di
s
a
dv
a
nt
a
ge
s
of
th
e
c
ur
r
e
nt
a
lg
or
it
hm
.
F
ig
ur
e
11
c
om
pa
r
in
g
m
e
a
s
ur
e
m
e
nt
p
ow
e
r
of
(
a
)
w
a
r
m
e
r
ba
g
a
nd
(
b)
w
a
s
hi
ng
m
a
c
hi
ne
.
T
he
a
lg
or
it
hm
s
ba
s
e
d
on
P
a
nd
Q
pe
r
f
or
m
poor
ly
f
or
de
vi
c
e
s
w
it
h
c
ont
in
uous
ly
v
a
r
yi
ng
pow
e
r
.
F
or
e
xa
m
pl
e
,
obs
e
r
ve
F
ig
ur
e
11
(
a
)
,
th
e
pow
e
r
of
th
e
he
a
ti
ng
ba
g
gr
a
dua
ll
y
in
c
r
e
a
s
e
s
by
m
or
e
th
a
n
240
W
ove
r
ti
m
e
.
L
ik
e
w
is
e
,
th
e
pow
e
r
of
th
e
w
a
s
hi
ng
m
a
c
hi
n
e
c
ha
nge
s
c
ont
in
uous
ly
be
twe
e
n
1958
W
a
nd
2150
W
in
s
oa
ki
n
g
m
ode
a
s
s
how
n
in
F
ig
ur
e
11
(
b)
.
T
he
r
e
f
or
e
,
th
e
a
lg
or
it
hm
is
una
bl
e
to
obt
a
in
th
e
c
or
r
e
c
t
∆
P
va
lu
e
.
S
e
c
ond
,
th
e
s
ys
te
m
c
onf
us
e
s
de
vi
c
e
s
w
it
h
ne
a
r
ly
th
e
s
a
m
e
pow
e
r
,
s
uc
h
a
s
in
c
a
nde
s
c
e
nt
a
nd
c
om
pa
c
t
la
m
p
s
(
a
lm
os
t
id
e
nt
ic
a
l
a
c
ti
ve
pow
e
r
P
)
.
I
n
s
om
e
c
a
s
e
s
,
th
e
r
e
a
c
ti
ve
pow
e
r
Q
of
t
he
c
om
pa
c
t
la
m
p i
s
l
e
s
s
t
ha
n t
he
de
te
c
ti
on
t
hr
e
s
hol
d of
t
he
e
dge
de
te
c
ti
on a
lg
or
it
hm
of
8
V
A
r
.
T
he
a
lg
or
it
hm
ig
nor
e
s
th
is
pow
e
r
e
dge
,
th
u
s
in
c
or
r
e
c
tl
y
p
r
e
di
c
ti
ng
th
e
in
c
a
nde
s
c
e
nt
la
m
p.
W
e
bui
lt
A
M
D
w
it
h
th
e
m
ic
r
ophone
to
ta
c
kl
e
th
e
a
bove
two
pr
obl
e
m
s
i
n
th
e
f
ut
ur
e
.
T
he
f
ir
s
t
ve
r
s
io
n
of
A
M
D
u
s
in
g
th
e
O
r
a
nge
P
i
Z
e
r
o boa
r
d
.
T
he
t
hr
e
e
de
vi
c
e
s
us
e
d i
n t
he
r
e
c
ogni
ti
on e
xpe
r
im
e
nt
a
r
e
t
he
ha
ir
dr
ye
r
, t
he
f
a
n, a
nd
th
e
m
ic
r
ow
a
ve
ove
n.
T
h
e
te
s
t
w
a
s
c
ondu
c
te
d
a
t
a
di
s
ta
nc
e
f
r
o
m
0.5
to
2
m
.
A
t
e
a
c
h
di
s
ta
nc
e
,
A
M
D
pr
e
di
c
ts
th
e
de
vi
c
e
s
ta
te
m
ul
ti
pl
e
t
im
e
s
t
o e
va
lu
a
te
pr
e
di
c
ti
on
a
c
c
ur
a
c
y. T
a
bl
e
4 s
how
s
t
he
t
e
s
t
r
e
s
ul
ts
;
w
e
c
a
n s
e
e
t
ha
t
th
e
s
ys
te
m
i
s
m
o
s
t
a
c
c
ur
a
te
a
t
a
di
s
ta
nc
e
of
l
e
s
s
t
ha
n 1.5 m.
(
a
)
(
b)
F
ig
ur
e
9
.
D
e
s
ig
ni
ng of
t
he
(
a
)
L
M
D
boa
r
d
a
nd
(
b)
A
M
D
boa
r
d
F
ig
ur
e
10
. P
ow
e
r
c
ons
um
pt
io
n gr
a
ph i
n 4
th
te
s
t
da
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
11
, N
o.
1
,
M
a
r
c
h 20
22
:
300
-
309
308
(
a
)
(
b)
F
ig
ur
e
11
. P
ow
e
r
of
(
a
)
w
a
r
m
e
r
ba
g a
nd
(
b)
w
a
s
hi
ng ma
c
hi
ne
T
a
bl
e
4
. A
M
D
t
e
s
t
r
e
s
ul
t
s
0.5 m
1 m
1.5 m
2 m
F
a
n
40/
40 (
100%
)
40/
40 (
100%
)
35/
40 (
88%
)
19/
40 (
48%
)
H
a
i
r
D
r
ye
r
40/
40 (
100%
)
39/
40 (
98%
)
37/
40
(
93%
)
25/
40 (
62.5%
)
M
i
c
r
ow
a
ve
O
ve
n
20/
20 (
100%
)
20/
20 (
100%
)
19/
20 (
95%
)
13/
20 (
65%
)
8.
C
O
N
C
L
U
S
I
O
N
T
hi
s
pa
pe
r
pr
e
s
e
nt
s
t
w
o de
vi
c
e
s
c
a
ll
L
M
D
a
nd A
M
D
. L
M
D
i
s
t
he
e
le
c
tr
oni
c
m
e
te
r
t
ha
t
is
i
ns
ta
ll
e
d a
t
th
e
e
le
c
tr
ic
a
l
pa
ne
l
of
a
r
oom
or
a
hous
e
.
L
M
D
pr
ovi
de
s
in
f
o
r
m
a
ti
on
a
bout
pow
e
r
c
ons
um
pt
io
n
in
or
de
r
to
de
te
c
t
w
hi
c
h
e
le
c
tr
ic
a
l
a
ppl
ia
nc
e
s
a
r
e
r
unni
ng.
T
hi
s
d
e
vi
c
e
pr
ovi
de
s
e
le
c
tr
ic
it
y
c
ons
um
pt
io
n
in
f
or
m
a
ti
on
of
e
a
c
h
e
le
c
tr
ic
a
l
a
ppl
ia
nc
e
to
hom
e
ow
ne
r
s
s
o
th
a
t
th
e
y
c
a
n
a
dj
us
t
th
e
ir
us
a
ge
pl
a
ns
m
or
e
e
f
f
i
c
ie
nt
ly
.
M
e
a
nw
hi
le
,
A
M
D
is
e
qui
pp
e
d
w
it
h
th
e
m
ic
r
ophone
;
th
e
f
unc
ti
on
of
th
is
de
vi
c
e
is
to
pr
ovi
de
m
or
e
in
f
or
m
a
ti
on
on
a
ppl
ia
nc
e
s
w
it
h
ti
m
e
-
va
r
yi
ng
pow
e
r
or
s
im
il
a
r
pow
e
r
.
T
he
a
ut
hor
s
ha
ve
bui
lt
th
e
f
ir
s
t
ve
r
s
io
n
of
th
e
e
le
c
tr
ic
a
l
e
qui
pm
e
nt
s
t
a
te
r
e
c
ogni
ti
on
s
ys
te
m
us
in
g
a
n
ope
n
-
s
our
c
e
pl
a
tf
or
m
.
T
he
s
y
s
te
m
a
ppl
ie
s
th
e
s
upe
r
vi
s
e
d l
e
a
r
ni
ng me
th
od a
nd t
he
M
L
P
m
ode
l.
A
c
ti
ve
pow
e
r
P
a
nd r
e
a
c
ti
ve
pow
e
r
Q
a
r
e
us
e
d
a
s
S
ig
na
tu
r
e
s
f
or
e
a
c
h
de
vi
c
e
.
T
h
e
e
xpe
r
im
e
nt
w
a
s
c
onduc
te
d
ov
e
r
th
r
e
e
w
e
e
ks
in
th
r
e
e
di
f
f
e
r
e
nt
r
oom
s
.
T
he
a
ut
hor
s
tr
a
in
e
d
th
e
s
ys
te
m
to
id
e
nt
if
y
12
de
vi
c
e
s
in
w
hi
c
h
two
de
vi
c
e
s
ha
ve
m
ul
ti
pl
e
m
ode
s
.
T
he
a
c
c
ur
a
c
y
of
th
e
s
ys
te
m
is
93.60%
.
H
ow
e
ve
r
,
dur
in
g
te
s
ti
ng,
th
e
a
ut
hor
s
f
ound
t
ha
t
th
e
s
ys
te
m
w
or
ks
in
e
f
f
ic
ie
nt
ly
f
or
de
vi
c
e
s
w
it
h
c
ons
ta
nt
ly
c
ha
ngi
ng
pow
e
r
.
I
t
is
a
ls
o
di
f
f
ic
ul
t
f
or
th
e
s
ys
te
m
to
di
f
f
e
r
e
nt
ia
te
de
vi
c
e
s
w
it
h
s
im
il
a
r
pow
e
r
.
T
he
r
e
f
or
e
, t
he
a
ut
hor
s
i
nt
e
nd t
o c
om
bi
ne
da
ta
f
r
om
A
M
D
t
o s
ol
ve
t
he
t
w
o a
na
ly
z
e
d pr
obl
e
m
s
i
n t
he
f
ut
ur
e
.
R
E
F
E
R
E
N
C
E
S
[
1]
N
.
H
.
N
guye
n
a
nd
V
.
H
.
D
uong,
“
A
s
ys
t
e
m
f
o
r
m
oni
t
o
r
i
ng
t
he
e
l
e
c
t
r
i
c
u
s
a
ge
of
hom
e
a
ppl
i
a
nc
e
s
us
i
ng
m
a
c
hi
ne
l
e
a
r
ni
ng
a
l
gor
i
t
hm
s
,”
i
n
P
r
oc
e
e
di
ng
s
-
2019
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
A
dv
anc
e
d
C
om
put
i
ng
and
A
ppl
i
c
at
i
ons
,
A
C
O
M
P
2019
,
N
ov.
2019, pp. 158
–
164, doi
:
10.1109/
A
C
O
M
P
.2019.00032.
[
2]
“
E
l
e
c
t
r
i
c
i
t
y
e
xpl
a
i
ne
d
-
us
e
of
e
l
e
c
t
r
i
c
i
t
y,”
U
.S.
E
ne
r
gy
I
nf
or
m
at
i
on
A
dm
i
ni
s
t
r
at
i
on
,
2020.
ht
t
ps
:
/
/
w
w
w
.e
i
a
.gov/
e
ne
r
gye
xpl
a
i
ne
d/
e
l
e
c
t
r
i
c
i
t
y/
us
e
-
of
-
e
l
e
c
t
r
i
c
i
t
y.php (
a
c
c
e
s
s
e
d F
e
b. 2020)
.
[
3]
S
.
D
a
r
by,
“
T
he
E
f
f
e
c
t
i
ve
ne
s
s
of
f
e
e
dba
c
k
on
e
n
e
r
gy
c
ons
um
pt
i
on,”
E
nv
i
r
on
m
e
nt
al
C
hange
I
ns
t
i
t
ut
e
U
ni
v
e
r
s
i
t
y
of
O
x
f
or
d
,
pp.
1
–
21, 2006, [
O
nl
i
ne
]
. A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
w
w
w
.e
c
i
.ox.a
c
.uk/
r
e
s
e
a
r
c
h/
e
ne
r
gy/
dow
nl
oa
ds
/
s
m
a
r
t
-
m
e
t
e
r
i
ng
-
r
e
por
t
.pdf
.
[
4]
S
.
H
oude
,
A
.
T
odd,
A
.
S
uda
r
s
ha
n,
J
.
A
.
F
l
or
a
,
a
nd
K
.
C
.
A
r
m
e
l
,
“
R
e
a
l
-
t
i
m
e
f
e
e
dba
c
k
a
nd
e
l
e
c
t
r
i
c
i
t
y
c
ons
um
pt
i
on:
A
f
i
e
l
d
e
xpe
r
i
m
e
nt
a
s
s
e
s
s
i
ng
t
he
pot
e
nt
i
a
l
f
or
s
a
vi
ngs
a
nd
pe
r
s
i
s
t
e
nc
e
,”
E
ne
r
gy
J
our
nal
,
vol
.
34,
no.
1,
pp.
87
–
102,
J
a
n.
2013,
do
i
:
10.5547/
01956574.34.1.4.
[
5]
K
.
C
a
r
r
i
e
A
r
m
e
l
,
A
.
G
upt
a
,
G
.
S
hr
i
m
a
l
i
,
a
nd
A
.
A
l
be
r
t
,
“
I
s
d
i
s
a
ggr
e
ga
t
i
on
t
he
hol
y
gr
a
i
l
of
e
ne
r
gy
e
f
f
i
c
i
e
nc
y?
T
he
c
a
s
e
of
e
l
e
c
t
r
i
c
i
t
y,”
E
ne
r
gy
P
ol
i
c
y
, vol
. 52, pp. 213
–
234, J
a
n. 2013, doi
:
10.1016/
j
.e
npol
.2012.08.062.
[
6]
K
.
L
.
T
s
a
i
,
F
.
Y
.
L
e
u,
a
nd
I
.
Y
ou,
“
R
e
s
i
de
nc
e
e
ne
r
gy
c
ont
r
ol
s
y
s
t
e
m
ba
s
e
d
on
w
i
r
e
l
e
s
s
s
m
a
r
t
s
oc
ke
t
a
nd
I
oT
,”
I
E
E
E
A
c
c
e
s
s
,
vol
.
4, pp. 2885
–
2894, 2016, doi
:
10.1109/
A
C
C
E
S
S
.2016.2574199.
[
7]
G
.
W
.
H
a
r
t
,
“
N
oni
nt
r
us
i
ve
a
ppl
i
a
nc
e
l
oa
d
m
oni
t
or
i
ng,”
P
r
oc
e
e
di
ngs
of
t
he
I
E
E
E
,
vol
.
80,
no.
12,
pp.
1870
–
1891,
1992,
doi
:
10.1109/
5.192069.
[
8]
M
.
W
e
i
s
s
,
A
.
H
e
l
f
e
ns
t
e
i
n,
F
.
M
a
t
t
e
r
n,
a
nd
T
.
S
t
a
a
ke
,
“
L
e
ve
r
a
gi
ng
s
m
a
r
t
m
e
t
e
r
da
t
a
t
o
r
e
c
ogni
z
e
hom
e
a
ppl
i
a
nc
e
s
,”
i
n
2012
I
E
E
E
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
P
e
r
v
a
s
i
v
e
C
om
put
i
ng
and
C
om
m
uni
c
at
i
ons
,
P
e
r
C
om
2012
,
M
a
r
.
2012,
pp.
190
–
197,
doi
:
10.1109/
P
e
r
C
om
.2012.6199866.
[
9]
C
.
L
a
ughm
a
n
et
al
.
,
“
P
ow
e
r
s
i
gna
t
ur
e
a
na
l
y
s
i
s
,”
I
E
E
E
P
ow
e
r
and
E
ne
r
gy
M
agaz
i
ne
,
vol
.
1,
no.
2,
pp.
56
–
63,
M
a
r
.
2003,
doi
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
M
ac
hi
ne
l
e
ar
ni
ng algor
it
hm
s
f
or
…
(
V
ie
t
H
oang Duong
)
309
10.1109/
M
P
A
E
.2003.1192027.
[
10]
L
.
K
.
N
or
f
or
d
a
nd
S
.
B
.
L
e
e
b,
“
N
on
-
i
nt
r
us
i
ve
e
l
e
c
t
r
i
c
a
l
l
oa
d
m
oni
t
o
r
i
ng
i
n
c
om
m
e
r
c
i
a
l
bui
l
di
ngs
ba
s
e
d
on
s
t
e
a
dy
-
s
t
a
t
e
a
nd
t
r
a
ns
i
e
nt
l
oa
d
-
de
t
e
c
t
i
on
a
l
gor
i
t
hm
s
,”
E
ne
r
gy
and
B
ui
l
di
ngs
,
vol
.
24,
no.
1,
pp.
51
–
64,
J
a
n.
1996,
doi
:
10.1016/
0378
-
7788(
95)
00958
-
2.
[
11]
D
.
S
r
i
ni
va
s
a
n,
W
.
S
.
N
g,
a
nd
A
.
C
.
L
i
e
w
,
“
N
e
ur
a
l
-
ne
t
w
or
k
-
ba
s
e
d
s
i
gna
t
ur
e
r
e
c
ogni
t
i
on
f
or
ha
r
m
oni
c
s
our
c
e
i
de
nt
i
f
i
c
a
t
i
on,”
I
E
E
E
T
r
ans
ac
t
i
ons
on P
ow
e
r
D
e
l
i
v
e
r
y
, vol
. 21, no. 1, pp. 398
–
405, J
a
n. 2006, doi
:
10.1109/
T
P
W
R
D
.2005.852370.
[
12]
S
.
N
.
P
a
t
e
l
,
T
.
R
obe
r
t
s
on,
J
.
A
.
K
i
e
nt
z
,
M
.
S
.
R
e
ynol
ds
,
a
nd
G
.
D
.
A
bow
d,
“
A
t
t
he
f
l
i
c
k
of
a
s
w
i
t
c
h:
de
t
e
c
t
i
ng
a
nd
c
l
a
s
s
i
f
yi
n
g
uni
que
e
l
e
c
t
r
i
c
a
l
e
v
e
nt
s
on
t
he
r
e
s
i
de
nt
i
a
l
pow
e
r
l
i
ne
,”
i
n
U
bi
C
om
p
2007
:
U
bi
qui
t
ous
C
om
put
i
ng
,
S
pr
i
nge
r
B
e
r
l
i
n
H
e
i
de
l
be
r
g,
2007, pp. 271
–
288.
[
13]
H
.
Y
.
L
a
m
,
G
.
S
.
K
.
F
ung,
a
nd
W
.
K
.
L
e
e
,
“
A
nov
e
l
m
e
t
hod
t
o
c
ons
t
r
u
c
t
t
a
xonom
y
e
l
e
c
t
r
i
c
a
l
a
ppl
i
a
nc
e
s
ba
s
e
d
on
l
oa
d
s
i
gna
t
ur
e
s
,”
I
E
E
E
T
r
ans
ac
t
i
ons
on C
ons
um
e
r
E
l
e
c
t
r
oni
c
s
, vol
. 53, no. 2, pp. 65
3
–
660, 2007, doi
:
10.1109/
T
C
E
.2007.381742.
[
14]
L
.
D
e
B
a
e
t
s
,
J
.
R
uys
s
i
nc
k,
C
.
D
e
ve
l
de
r
,
T
.
D
ha
e
ne
,
a
nd
D
.
D
e
s
c
hr
i
j
ve
r
,
“
A
ppl
i
a
nc
e
c
l
a
s
s
i
f
i
c
a
t
i
on
us
i
ng
V
I
t
r
a
j
e
c
t
or
i
e
s
a
nd
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
ks
,
”
E
ne
r
gy
and B
ui
l
di
ngs
, vol
. 158, pp. 32
–
36, J
a
n. 2018, doi
:
10.1016/
j
.e
nbui
l
d.2017.09.087.
[
15]
K
.
K
ha
l
i
d,
A
.
M
oha
m
e
d,
R
.
M
oha
m
e
d,
a
nd
H
.
S
ha
r
e
e
f
,
“
P
e
r
f
or
m
a
nc
e
c
om
pa
r
i
s
on
of
a
r
t
i
f
i
c
i
a
l
i
nt
e
l
l
i
ge
nc
e
t
e
c
hni
qu
e
s
f
or
non
-
i
nt
r
us
i
ve
e
l
e
c
t
r
i
c
a
l
l
oa
d
m
oni
t
or
i
ng,”
B
ul
l
e
t
i
n
of
E
l
e
c
t
r
i
c
al
E
ngi
ne
e
r
i
ng
and
I
nf
or
m
at
i
c
s
,
vol
.
7,
no.
2,
pp.
143
–
152,
J
un.
2018,
doi
:
10.11591/
e
e
i
.v7i
2.1190.
[
16]
J
. Z
. K
ol
t
e
r
a
nd M
. J
. J
ohns
on, “
R
E
D
D
:
a
publ
i
c
da
t
a
s
e
t
f
or
e
ne
r
gy di
s
a
ggr
e
ga
t
i
on r
e
s
e
a
r
c
h,”
i
n
Sus
t
K
D
D
w
o
r
k
s
hop
,
2011, no. 1,
pp. 1
–
6.
[
17]
K.
C
ha
hi
ne
,
“
T
ow
a
r
ds
a
ut
om
a
t
i
c
s
e
t
up
of
non
i
nt
r
us
i
ve
a
ppl
i
a
nc
e
l
oa
d
m
oni
t
or
i
ng
–
F
e
a
t
ur
e
e
xt
r
a
c
t
i
on
a
nd
c
l
us
t
e
r
i
ng,”
I
nt
e
r
nat
i
onal
J
our
nal
of
E
l
e
c
t
r
i
c
al
and
C
om
put
e
r
E
ngi
ne
e
r
i
ng
,
vol
.
9,
no.
2,
pp.
1002
–
1011,
A
pr
.
2019,
doi
:
10.11591/
i
j
e
c
e
.v9i
2.pp1002
-
1011.
[
18]
S
.
S
e
m
w
a
l
,
R
.
S
.
P
r
a
s
a
d,
a
nd
P
.
J
une
j
a
,
“
I
de
nt
i
f
yi
ng
a
ppl
i
a
nc
e
s
us
i
ng
N
I
A
L
M
w
i
t
h
m
i
ni
m
um
f
e
a
t
u
r
e
s
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
E
l
e
c
t
r
i
c
al
and C
om
put
e
r
E
ngi
ne
e
r
i
ng
, vol
. 4, no. 6, pp. 909
–
922, D
e
c
. 2014, d
oi
:
10.11591/
i
j
e
c
e
.v4i
6.6715.
[
19]
N
.
I
ks
a
n,
J
.
S
e
m
bi
r
i
ng,
N
.
H
a
r
i
ya
nt
o,
a
nd
S
.
H
.
S
upa
ngka
t
,
“
R
e
s
i
de
nt
i
a
l
l
oa
d
e
ve
nt
de
t
e
c
t
i
on
i
n
N
I
L
M
u
s
i
ng
r
obus
t
c
e
ps
t
r
um
s
m
oot
hi
ng
ba
s
e
d
m
e
t
hod,”
I
nt
e
r
nat
i
onal
J
ou
r
nal
of
E
l
e
c
t
r
i
c
al
and
C
o
m
put
e
r
E
ngi
ne
e
r
i
ng
,
vol
.
9,
no.
2,
pp.
7
42
–
752,
A
pr
.
2019
,
doi
:
10.11591/
i
j
e
c
e
.v9i
2.pp742
-
752.
[
20]
G
.
L
a
put
,
Y
.
Z
ha
ng,
a
nd
C
.
H
a
r
r
i
s
on,
“
S
ynt
he
t
i
c
s
e
n
s
or
s
:
t
ow
a
r
ds
ge
n
e
r
a
l
-
pur
pos
e
s
e
n
s
i
ng,”
i
n
C
onf
e
r
e
n
c
e
on
H
um
an
F
ac
t
or
s
i
n
C
om
put
i
ng Sy
s
t
e
m
s
-
P
r
oc
e
e
di
ngs
,
M
a
y 2017, vol
. 2017
-
M
a
y, pp.
3986
–
3999,
doi
:
10.1145/
3025453.3025773.
[
21]
J
.
J
a
l
de
n,
X
.
C
.
M
or
e
no,
a
nd
I
.
S
kog,
“
U
s
i
ng
t
he
a
r
dui
no
due
f
or
t
e
a
c
hi
ng
di
gi
t
a
l
s
i
gna
l
pr
oc
e
s
s
i
ng,
”
i
n
I
C
A
SSP
,
I
E
E
E
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
A
c
ous
t
i
c
s
,
Spe
e
c
h
and
Si
gnal
P
r
oc
e
s
s
i
ng
-
P
r
oc
e
e
di
ngs
,
A
pr
.
2018,
vol
.
2018
-
A
p
r
i
l
,
pp.
6468
–
6472,
doi
:
10.1109/
I
C
A
S
S
P
.2018.8461781.
[
22]
C
.
H
oc
hgr
a
f
,
“
U
s
i
ng
a
r
dui
no
t
o
t
e
a
c
h
di
gi
t
a
l
s
i
gna
l
pr
oc
e
s
s
i
ng,”
2013,
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
pdf
s
.s
e
m
a
nt
i
c
s
c
hol
a
r
.or
g/
290a
/
8c
9a
a
b485d3b8f
4e
be
8e
08584bd300d766
6c
.pdf
.
[
23]
“
A
t
m
e
l
A
V
R
465
:
S
i
ngl
e
-
P
ha
s
e
P
ow
e
r
/
E
ne
r
gy
M
e
t
e
r
w
i
t
h
T
a
m
p
e
r
D
e
t
e
c
t
i
on,”
2013.
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
p:
/
/
w
w
1.m
i
c
r
oc
hi
p.c
om
/
dow
nl
oa
ds
/
e
n/
A
ppnot
e
s
/
A
t
m
e
l
-
2566
-
S
i
ngl
e
-
P
ha
s
e
-
P
ow
e
r
-
E
ne
r
gy
-
M
e
t
e
r
-
w
i
t
h
-
T
a
m
pe
r
-
D
e
t
e
c
t
i
on_A
p
-
N
ot
e
s
_A
V
R
465.pdf
.
[
24]
“
A
t
m
e
l
A
V
R
1631:
S
i
ngl
e
P
ha
s
e
E
ne
r
gy
M
e
t
e
r
us
i
ng
X
M
E
G
A
A
,”
2012.
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
p:
/
/
w
w
1.m
i
c
r
oc
hi
p.c
om
/
dow
nl
oa
ds
/
e
n/
A
ppnot
e
s
/
doc
42039.pdf
.
[
25]
T
. H
. V
u,
T
he
f
undam
e
nt
al
s
of
M
ac
hi
ne
L
e
ar
ni
ng
. 2018.
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Viet
Hoang
Duong
received
his
bachelor’s
degree
in
2016
and
M
aster’s
degree
in
2020
from
Hanoi
University
of
Science
and
Technology
(HUST)
,
Vietnam.
His
researc
h
interests
include
smart
grids,
machine
learning,
and
instrumentation
.
He
can
be
contacte
d
at
email:
viet.dhca180160@
sis.hust.edu.vn
.
Nam Hoang Ngu
yen
received his eng
ineer degree in 2
002 from Hano
i Universi
ty
of
Technology
(HUST),
his
Master'
s
degree
in
2004
from
Hendri
Poin
caré
University,
France,
and
his
Ph.D.
in
2009
from
Grenoble
Polytechnic
University,
Fr
ance.
He
is
currently
a
Lecturer
in
the
Depa
rtment
of
Industrial
Metrology
and
Informatics
(
3I),
School
of
Electrical
Engineering,
Hanoi
University
of
Technology
(HUST).
His
main
research
directions
are
intelligent mete
ring systems, I
oT and embe
dded systems,
IIoT, a
nd re
newable
energy sy
stems.
He ca
n be contacted at email:
nam.nguyenhoang@
hust.edu.vn
.
Evaluation Warning : The document was created with Spire.PDF for Python.