TELKOM
NIKA
, Vol.13, No
.2, June 20
15
, pp. 556 ~ 5
6
2
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i2.1471
556
Re
cei
v
ed
Jan
uary 23, 201
5
;
Revi
sed Ma
rch 2
9
, 2015;
Acce
pted April 18, 2015
Virtual Instrument of Harmonics Detection Based on
Neural Network
Adaptive Filters
Xianfeng Zh
eng*
1
, Zheng
Fan
2
1
Department o
f
Electrical Eng
i
ne
erin
g,He
nan
Mec
han
ical a
n
d
Electrica
l
En
gin
eeri
ng C
o
ll
e
ge,
Hen
an,4
530
03
,Chin
a
2
Department o
f
Automatic Co
ntrol Eng
i
n
eeri
ng,He
na
n
Mec
han
ical a
nd El
ectrical En
gin
e
e
rin
g
Col
l
eg
e,
Hen
an, 45
300
3,Chi
n
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: hn
xxzxf@
12
6.com
1
,fanzhe
ng1
97
3@1
63.c
o
m
2
A
b
st
r
a
ct
T
h
is study inv
e
stigate
d
the a
dapt
iv
e detecti
on princ
i
p
l
e ba
sed on a sin
g
l
e
artificial n
eur
on, an
d
constructed
a
meth
od
for d
e
tecting
har
moni
cs usin
g th
e ar
tificial
ne
ural
n
e
tw
ork techni
q
ue. Bas
ed
on t
h
e
establ
ishe
d
me
thod, an
d by c
o
mp
re
hens
ivel
y processi
ng th
e obtai
ne
d har
mo
nics d
a
ta us
ing the
La
bVIEW
softw
are-devel
opi
ng e
n
vir
o
n
m
e
n
t of the
virtual i
n
stru
ment, the h
a
r
m
onic w
a
ves w
e
re d
e
tected
an
d
analy
z
e
d. Finally, t
he
analysis of current ball cr
usher
har
m
o
nics v
e
rified that the designed system
was
effective.
Ke
y
w
ords
:
Ha
rmo
n
ics Det
e
ction, Ne
ural
N
e
tw
ork, Virtual Instrume
nts
1. Introduc
tion
With the
wid
e
appli
c
atio
n
of all kin
d
s o
f
nonline
a
r
p
o
we
r ele
c
tron
ic devi
c
e
s
, h
a
rmo
n
ic
pollution ha
s becom
e in
cre
a
si
ngly seriou
s in po
wer n
e
two
r
ks, resulting
in the frequ
ent
occurre
n
ce o
f
various fa
u
l
ts and
acci
dents
ca
us
e
d
by harmo
nics o
c
cur.
To inten
s
ify the
treatment, m
anag
ement
and
ch
arge
for h
a
rm
oni
c pollutio
n
, t
he
re
sea
r
ch
on
the
po
wer
harm
oni
cs m
onitorin
g
sy
stems,
which a
r
e
capa
bl
e of
doing
real
-ti
m
e, accu
rate
and
contin
u
ous
measures of
power ha
rmo
n
ics, is
of gre
a
t theoretical and en
ginee
ri
ng pra
c
tical si
gnifica
nce [1].
2. Dete
ction
method for p
o
w
e
r harmo
n
i
cs base
d on artificial n
e
ural net
w
o
r
k
Harmoni
cs detection is a
key techni
qu
e of ac
tive powe
r
filter an
d harmo
nic
monitori
ng
system
s. Onl
y
when
harmonic cu
rr
en
ts are a
c
curately detecte
d in real tim
e
ca
n they
be
effectively an
alyzed,
comp
ensated, and
inhibit
ed. Existing po
we
r h
a
rmo
n
ics det
ection in
clu
d
e
s
those
metho
d
s b
a
sed
on
notch filters or
band
pa
ss an
alog
filters, F
r
yze
time dom
ain p
o
w
er
definition, fa
st Fou
r
ie
r transfo
rm
s, the insta
n
tane
ous
re
active
power th
eo
ry of three
-
p
hase
circuits,
simul
t
aneou
s dete
c
tion, wavel
e
t trans
fo
rm
s, and ada
ptive filters [2]-[3].
To effectively monitor and
eliminate po
wer
ha
rmoni
cs and therefore
redu
ce the d
a
mage
of power
harmonics, ha
rm
onics d
e
tecti
on metho
d
s
are
sup
p
o
s
ed
to pre
s
e
n
t the advanta
g
e
s
of
low
co
mputat
ion a
m
ount
s,
re
al-time
wel
l
, high
a
ccu
ra
cy an
d
relia
bi
lity, eas
e
of
realization, a
n
d
stron
g
a
dopti
v
e ability. With these
adv
antage
s,
the
fundam
ental
active curre
n
t, fundame
n
t
al
rea
c
tive cu
rrent, total harmonic
cu
rren
t, and t
he co
ntent of ea
ch harmoni
c can be d
e
tect
ed
simultan
eou
sl
y, which
be
nefits the
compen
satio
n
and a
nalysis of po
we
r harm
oni
cs.
By
con
s
id
erin
g the above
re
quire
ment
s a
nd co
mpa
r
in
g multiple d
e
tection m
e
thod
s, the po
wer
harm
oni
cs
d
e
tection
met
hod b
a
sed
o
n
an
ada
ptive
filter
wa
s
applie
d in th
e re
se
arch.
This
method n
o
t o
n
ly sho
w
s hi
gh dete
c
tion
accuracy, but
also
ca
n pe
rform tra
c
king
measureme
n
ts.
Additionally, it exhibits a
stron
g
ad
apti
v
e abilit
y. Therefo
r
e, thi
s
method
pre
s
ent
s the b
e
st
effects,
stron
gest a
daptiv
e ability and
bro
ade
st prosp
e
ct
s am
o
ng the afo
r
e
m
entione
d p
o
we
r
harmonic
s
detec
tion methods
[4]-[5].
A multi-cha
n
n
e
l adaptive filter is su
ppo
sed to
be ea
sily realize
d
an
d sho
u
ld also
be able
to perfo
rm re
al-time d
e
tect
ion for
po
wer harm
oni
cs.
B
e
cau
s
e
of
it
s si
mple
st
ru
ct
ure
an
d c
e
r
t
ain
adaptio
n and
self-lea
rnin
g
ability, only
a single a
r
ti
ficial neu
ro
n is used to buil
d
a multi-ch
a
nnel
adaptive filter. The stru
cture is
demo
n
st
rated in Figu
re
1.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Virtual Instrum
ent of Harm
onics Detecti
on
Based on
Neural Network .... (Xianfeng Zheng)
557
si
n(
)
nt
cos(
)
nt
si
n(
)
t
si
n
(
)
nt
Figure 1. Det
e
ction p
r
in
cipl
e for pow
er h
a
rmo
n
ics ba
sed on si
gnal
neuron
W
h
en
th
e pow
e
r
vo
lta
g
e
u
s
o
f
th
e s
i
ng
le
-
p
h
a
s
e
c
i
rc
u
i
t pa
ss
es th
r
o
ug
h th
e
loc
k
-
p
ha
se
c
i
rc
uit,
sin(
)
t
and
co
s
(
)
t
are obtai
ned.
Then by increasi
ng
the freque
ncy dou
bling, the sin
e
and
co
sine
si
gnal
s in the
numbe
r (2~N) are acqui
re
d, whi
c
h a
r
e
use
d
a
s
the
referen
c
e inp
u
t
sign
als of the
single
artifici
al neu
ron. Th
ese lin
ea
r co
mbination
s
of
the sine
and
co
sine
sign
a
l
s
are the net in
puts of the si
gnal artifici
al neuron [6]-[7]
.
The
referen
c
e input ve
cto
r
()
Xt
, net input
()
s
t
, and
output
()
yt
o
f
the si
gnal
a
r
tificial
neuron a
r
e:
(
)
[s
i
n
,
c
os
;
,
si
n(
),
cos
(
)]
Xt
t
t
n
t
n
t
(1)
2
1
()
()
()
()
n
ii
i
st
w
t
x
t
t
(2)
)]
(
[
)
(
t
s
f
t
y
(3)
As
()
yt
can b
e
o
b
tained by the linear
com
b
ination of the referen
c
e input, the acti
on
function
of the artificial
ne
uron i
s
()
f
xx
. Because the alte
rnating
cu
rre
n
t (AC) tran
smissi
on
bus d
o
e
s
not contai
n dire
ct
current (DC)
comp
one
nts,
()
t
is ze
ro. Und
e
r this
conditi
on:
2
1
()
()
()
n
ii
i
yt
w
t
x
t
(4)
The lea
s
t mean sq
ua
re (LMS) algo
rith
m [8] was adopted in th
e learni
ng ru
le of the
neuron. As th
e algo
rithm a
pplie
s the e
s
timations
of
the releva
nt fun
c
tion
s of inpu
t vectors i
n
th
e
estimation
of
the in
stant
gradient
ve
cto
r
s, the
convergen
ce
rate
is
distinctly faster than found
in
the typical L
M
S algorithm
, with slight computation
complexity.
Next, the erro
r feedba
ck si
gnal
()
et
is used
to adjust the
weig
ht
()
i
wt
)]
1
(
)
(
[
)
(
)
(
)
(
)
1
(
t
w
t
w
t
x
t
e
t
w
t
w
i
i
i
i
i
(5)
Whe
r
e
is the
learning
rate
01
,
is the d
a
m
p
ing
coeffici
e
n
t, which a
ccelerate
s
learni
ng rate and re
du
ce
s vibration
01
.
The co
ncrete
pro
c
ed
ures a
r
e as follo
ws:
1) Initializatio
n. The
conve
r
gen
ce fa
cto
r
(0
1
)
and con
s
tan
t
are
set. The initial
value of wei
g
ht vector
n
W
is random, a
nd i
t
s dimen
s
io
n depe
nd
s on t
he num
ber
of harmo
nics
and interha
r
monics.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 556 – 56
2
558
2) The ite
r
ation numb
e
r i
s
set and the o
u
tput error is
comp
uted by:
1
()
;
0
,
1
,
2
,
1
qH
nn
n
N
ey
W
X
n
N
(6)
Whe
r
e
1
()
qH
n
W
is the
transpo
sed
matrix of the qth
learning result of the
weig
ht vector
matrix.
Acco
rdi
ng to the learni
ng rule of the artifici
al ne
uro
n
, the wei
ght ap
proa
ch
es the
optimal
value by several iteratio
ns. Under thi
s
condition,
2
[(
)
]
Ee
t
app
roache
s the minimum val
ue, which
indicates that
in the
le
ast
mean
squa
re
, the o
u
tput
()
yt
of the
neu
ron
optimally
ap
proximate
s
the noise inte
rfere
n
ce cu
rrent
()
it
, while the output
()
zt
of the syste
m
op
timally appro
x
imate
s
the sign
al
()
k
it
to be dete
c
ted
.
Furtherm
o
re, it proves
that the wei
ghts of ea
ch
sub
c
ircuit
adaptive filter optimally a
pprox
im
ate the pea
k val
ues of the
sine o
r
cosi
ne sig
nal
s that
corre
s
p
ond t
o
the sub
c
ircuit in the noise interfe
r
e
n
c
e current i(t) [9]-[10]. Therefore, the po
wer
harm
oni
cs
ca
n be dete
c
ted
usin
g the out
put
()
zt
of the sy
stem or the
weight of ea
ch
sub
circuit
of the artificia
l
neuro
n
ba
se
d
on the real
situation [11].
After obtaini
n
g
the
data
of
each h
a
rm
oni
c, th
e p
o
w
e
r
n
e
t
w
o
r
k
pa
r
a
me
te
rs
ar
e calc
u
l
a
t
ed
usin
g the followin
g
formul
ae:
1) The effe
ctive value of the nth harm
oni
c:
22
()
/
2
nn
s
n
c
I
2) The p
h
a
s
e
angle of the nth harm
oni
c:
(/
)
nn
c
n
s
j
ar
c
t
g
W
W
3) Fun
dame
n
t
al active power:
11
1
cos
PU
I
4) Fun
dame
n
t
al reactive p
o
we
r:
11
1
si
n
QU
I
5) The
conte
n
t of the nth harmo
nic:
1
/1
0
0
%
nn
HRI
I
I
6) Ha
rmo
n
ic
conte
n
t:
2
2
H
n
n
II
7) Total ha
rm
onic di
stortio
n
:
1
/
100%
iH
TH
D
I
I
3. Sy
stem design and implementa
tio
n
3.1. O
v
erall
struc
t
ur
e of
the s
y
stem
The d
a
ta a
c
q
u
isition
ci
rcuit is
de
sign
ed
as
sh
own in
Figure 2. In
the figu
re,
DHPT an
d
DHCT are the mic
r
o hi
gh-acc
u
rac
y
voltage and
c
u
rrent s
e
ns
ors
,
respec
tively,
and have been
desi
gne
d to perfo
rm AC measurement
s for any
voltage belo
w
1
000V, cu
rre
nt in the range
of
1mA to 100
A, and freq
uen
cy in th
e ran
ge of
40 to10
00
Hz.AD620, a
s
a high
-a
ccuracy
instru
mentati
on amplifie
r, has
a wid
e
volt
age rang
e
and go
od o
u
tput linea
rity, and me
ets th
e
system requi
rements fo
r in
put sign
als by
adjusti
n
g
ba
ckwa
rd resi
st
ors a
nd bia
s
resi
stors.
Figure 2. Structure of
syste
m
hard
w
a
r
e
The m
odul
e
4010
6 i
s
a
Schmitt trig
ger,
whi
c
h
elimin
ates
and
sha
pes the
pea
ks of
sin
e
sign
als. The
module 4
046
is the phase
-
locke
d
l
oop
of monolithic integration,
whi
c
h mainta
ins
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Virtual Instrum
ent of Harm
onics Detecti
on
Based on
Neural Network .... (Xianfeng Zheng)
559
the co
nformit
y
of the outp
u
t sign
als
an
d input
sign
a
l
s by auto
m
a
t
ically trackin
g
the fre
que
ncy
variation of in
put sign
als in
a certai
n sco
pe.
After the loa
d
voltage
L
u
has b
een t
r
an
sform
ed by
DHPT
and fil
t
ered
by a lo
w-p
a
ss
filter, the obt
ained
sin
e
si
gnal i
s
a
m
plified u
s
in
g
AD620; afterwa
r
ds, it is shap
ed u
s
ing
40
2
06.
The si
gnal i
s
then sent into
the pha
se
-lo
c
ked
cont
rol
circuit 40
46 t
o
gen
erate
a
n
interruptio
n
o
f
the re
ctang
ul
ar si
gnal,
whi
c
h i
s
syn
c
h
r
o
nou
s with
L
u
. Therefo
r
e,
wh
en
L
u
frequen
cy chan
ge
s,
the cal
c
ul
atin
g step
si
ze i
s
adj
uste
d correspon
dingl
y. It is worth
mentionin
g
that the external
comp
one
nt o
f
4046 was
d
e
sig
ned b
a
se
d on the 5
0
Hz
of the ce
nter fre
quen
cy of the voltage-
controlled oscillator.
The loa
d
current
L
i
is tra
n
sfo
r
med to A
C
voltage u
s
ing
DHCT a
nd th
e sig
nal
conv
erter.
After being a
m
plified and
biased, the AC voltage si
g
nal cha
nge
s from bipol
ar to unipola
r
. When
it satisfies th
e req
u
ire
m
en
ts of the acq
u
isition
e
quip
m
ent, the sig
nal is ap
plied
as the on
e to b
e
detected to extract harmonics data, whi
c
h
will be processed usin
g a computer.
3.2. Soft
w
a
r
e
implementation of the
s
y
stem
Software i
s
th
e co
re of a vi
rtual in
strum
e
nt. To red
u
ce
developm
ent
efforts an
d i
m
prove
the desi
gne
d
system'
s
uni
versality and
expan
sion
ca
pabilitie
s, the most su
cce
s
sful and
wide
ly
use
d
software developm
e
n
t environme
n
t, LabVIE
W 2010,was u
s
ed to comp
print statem
ent,
rehe
nsively p
r
ocess
the co
llected
data.
This
a
llo
wed
for the
devel
opment
of a
n
integ
r
ated
a
nd
intelligent
det
ection
sy
ste
m
for po
we
r
harm
oni
cs, i
m
pleme
n
ting
multiple fun
c
t
i
ons,
su
ch
a
s
re
al-
time dete
c
tio
n
, display, an
alysis,
pre
d
ict
i
on, ala
r
m, a
nd p
r
ote
c
tion.
The
fun
c
tion
s a
r
e
de
scrib
e
d
in detail belo
w
:
1)
Data a
c
q
u
i
sition, in
cludi
ng the
actu
ation
an
d initial
i
zation
of dat
a acqui
sition
cards,
and the a
c
ce
ptance of acq
u
ired d
a
ta.
2) T
h
e
re
alization of
ada
ptive filtering
.
Th
is refers to us
ing a
c
o
mputer to
perform
adaptive filtering of the acq
u
ired d
a
ta in orde
r to
cal
c
u
l
ate each harmonic d
a
tum
in the sign
al.
3) The di
spl
a
y, storage an
d backu
p of the data.
4) An exce
edi
ng-limit ala
r
m
and prote
c
tio
n
for the data
.
5) Data q
u
e
r
i
e
s, statisti
cs,
and pri
n
t statement.
Becau
s
e L
a
b
V
IEW does n
o
t have the neural n
e
two
r
k function, am
ong its functi
ons, the
con
c
rete
com
putation of
a
daptive filteri
ng i
s
realize
d
by CI
N n
o
des,
or
by M
A
TLAB usi
n
g
the
MATLAB interface.
The structure of system software i
s
illust
rated in Figure 3.
Figure 3. Structure of
syste
m
softwa
r
e
Figure 4. Flow ch
art o
f
intelligent detection of
harm
oni
cs
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 556 – 56
2
560
Among
th
e a
f
oreme
n
tione
d
fun
c
tion
s, harm
oni
cs
de
tection and a
nalysi
s
sho
w
highest
requi
rem
ents and the larg
est co
mputati
on amo
unt
s
and real-time
levels, whi
c
h are the m
a
in
parts
of the software. Fi
rst
of all, harmon
i
cs d
e
tectio
n
wa
s pe
rform
e
d for the a
c
qu
ired d
a
ta usi
n
g
the above
ad
aptive filter al
gorithm, o
b
tai
n
ing the
dat
a
of ea
ch
harmonic,
as displayed in
Fig
u
re
4. Afterwards, based
on th
e re
ceived h
a
r
moni
c weig
h
t, the power
n
e
twork p
a
ra
m
e
ters, in
clu
d
i
n
g
the effective
value of the nth ha
rmonic, p
h
a
s
e angle, fu
ndame
n
tal a
c
tive po
wer and
fundame
n
tal
rea
c
tive po
wer, co
ntent o
f
the
nth ha
rmoni
c, the
harm
oni
c co
ntent, and to
ta
l
harm
oni
c di
stortion
we
re
analy
z
ed
a
nd di
splaye
d
usi
ng di
spl
a
y modul
e.
Con
s
id
erin
g
the
necessa
ry legibility and intuition cha
r
acteri
st
ics of
the data,
harm
oni
cs d
a
ta are usu
a
lly
displ
a
yed in
the form of g
r
aph
s. The
r
ef
ore,
the LabVIEW s
o
ftware's
Waveform Graph control
wa
s use
d
to show the ef
fective value and c
onte
n
t of harmoni
cs. The protot
ype waveform,
fundame
n
tal
waveform, h
a
rmo
n
ic waveform, fun
d
a
m
ental a
c
tive
po
wer,
fund
amental
re
active
power, an
d total harm
oni
c distortio
n
we
re di
spla
ye
d usin
g the Wa
veform Ch
art
control, and
the
effective valu
e an
d
co
nten
t of ha
rmo
n
ics
we
re
prese
n
ted u
s
in
g
a
histo
g
ram
while oth
e
r dat
a
were displaye
d usin
g linetypes.
4. Opera
t
ion tes
t
After finishi
n
g
the
desi
gn,
operation te
st
s
we
re
co
ndu
cted fo
r th
e p
o
we
r
utilizatio
n of
a
micro-p
o
wde
r
plant. The m
a
jor lo
ad in th
e tests
wa
s a
ball cru
s
he
r that wa
s 2.2m
in diamete
r
; its
sampli
ng f
r
eq
uen
cy was 4.
8 kHz, an
d th
e samplin
g la
sted fo
r 1
0
0
m
S, whi
c
h l
a
sted fo
r a
bou
t 5
fundame
n
tal
wave
peri
o
d
s
. The
above
LMS al
gorit
hm was u
s
e
d
for lea
r
nin
g
; the fo
rgett
i
ng
fac
t
or was
0.99
u
, a
nd the co
nsta
nt was
0.01
.
By using the online mo
nito
ring sy
stem for po
we
r qua
lity,
the a-pha
se current wa
veform
of the ball cru
s
he
r in no
rma
l
operatio
n was obtai
ned,
as dem
on
stra
ted in Figure 5 and Figu
re
6.
Figure 5. Current wavefo
rm of the ball cru
s
h
e
r
Figure 6. Current fitting wa
veform of the ball crushe
r
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Virtual Instrum
ent of Harm
onics Detecti
on
Based on
Neural Network .... (Xianfeng Zheng)
561
By using the
above meth
o
d
, 19 si
ne
co
mpone
nts
we
re a
c
qui
re
d i
n
the current
sign
al of
the ball
cru
s
h
e
r; the e
s
tim
a
tions
of the
curre
n
t
para
m
eters of the
ball cru
s
he
r
are di
spl
a
yed
in
Table 1.
Table 1 Estim
a
ted re
sults o
f
current ha
rmonics of the
ball cru
s
h
e
r
Serial number
Freque
nc
y
/Hz
Amplitude
Phase
Harmonic rate
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
41.303
49.878
59.236
149.732
238.726
254.550
340.723
355.901
537.143
636.855
830.039
923.974
940.699
1121.311
1223.168
1784.413
1859.637
1930.749
2076.858
97.564
814.572
95.708
8.890
9.782
10.076
8.943
3.560
3.696
3.620
5.400
5.774
4.102
2.154
2.624
0.244
0.503
0.255
0.288
83.313
271.263
146.998
16.413
309.335
264.416
135.366
44.655
229.369
24.271
59.590
95.024
235.142
343.700
255.472
239.325
36.540
128.600
46.854
11.977%
100%
11.749%
1.091%
1.201%
1.237%
1.098%
0.437%
0.454%
0.444%
0.663%
0.709%
0.504%
0.264%
0.322%
0.030%
0.062%
0.031%
0.035%
THD
16.989%
Limited
by scre
en sp
ace,
Fi
gure 7
sh
ows spe
c
ific data of
1
0
harm
oni
cs wi
th
their
minimum freq
uen
cie
s
, whi
c
h we
re di
re
ctly display
ed
o
n
the u
s
er i
n
terface,
whil
e
the others
we
re
printed in a
statement.
Figure 7. Use
r
interfa
c
e of the com
p
r
ehe
nsive dete
c
tio
n
system for
harm
oni
cs
Apart from th
e harm
oni
cs
detectio
n
and
analysi
s
, the manage
men
t
for harmoni
cs d
a
ta
is also very i
m
porta
nt. In the syste
m
, harmo
ni
cs dat
a mana
geme
n
t is reali
z
e
d
by the software
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 556 – 56
2
562
tool pa
ckage
SQL Tool
kit
,
which supp
orts
r
eal
-time
intera
ction
s
with the d
a
t
abase sy
ste
m
.
There a
r
e th
ree
step
s for
visiting the
d
a
taba
se
u
s
in
g O
D
BC
drivi
ng mo
de: a
s
signi
ng the
d
a
ta
sou
r
ce, writin
g the op
erating comma
nd
, and exe
c
uti
ng the
comm
and. Th
erefo
r
e, the datab
a
s
e
stru
cture ha
s to be det
ermin
ed first
,
bas
ed o
n
which data
base and d
a
ta sou
r
ce
were
establi
s
h
ed.
Afterwa
r
ds, t
he fun
c
tion
node
s of S
Q
L To
olkit
were a
dopte
d
to conn
ect
an
d
operate the d
a
ta sou
r
ce, thus effectively managi
ng ha
rmonics data.
Acco
rdi
ng to
the fun
c
tio
n
re
quireme
nts
of
the system,
the con
s
tru
c
ted
databa
se
contai
ns five data table
s
that store the
monito
ri
ng pa
ramete
rs, no
rmal ope
ration
powe
r
netwo
rk
para
m
eters,
power n
e
two
r
k p
a
ramete
rs befo
r
e a
n
d
after a fault
,
powe
r
net
work
pa
ramete
rs
before
and
after an
alarm,
and p
o
wer n
e
twork p
a
ram
e
ters befo
r
e
and after a tri
p
. The la
st fo
ur
instan
ce
s ap
ply the date (year, month, and day
) as the prim
a
r
y key and index, which is
conve
n
ient fo
r the que
ry, statistics, backup,
clea
ring,
and re
cove
ry of the databa
se.
5. Conclusio
n
s
Test re
sult
s verified that the system
perfo
rmed
well, realizi
n
g
all the functions in
deman
d an
al
ysis a
nd the
expecte
d de
sign obj
ecti
ve.
Comp
ared
with mo
st existing mo
nitori
ng
system
s
fo
r power harmo
nics,
t
he
system develo
p
ed in thi
s
p
aper exhibite
d the follo
wi
ng
advantag
es:
-
It detected sp
ecified h
a
rm
o
n
ic
by ch
angi
ng paramete
r
s onlin
e;
-
By adopting
the ada
ptive power
harm
onics d
e
tecti
on meth
od b
a
se
d on
a
si
ngle a
r
tificial
neuron, the system resp
onde
d quickl
y
with
high detectio
n
accura
cy and st
rong a
daptiv
e
ability;
-
The
com
p
re
h
ensive
monit
o
ring
sy
stem
demo
n
st
rate
s
stron
g
fun
c
tions, hi
gh fle
x
ibility, good
open
ne
ss a
n
d
expand
abili
ty through use of LabVIEW;
-
Its cost is lo
w and therefo
r
e can b
e
pop
ulari
z
ed o
n
a large
r
scal
e.
-
In conclusion, the integrated
and intelligent monitoring syst
em
for power harmonics
designed
in this pap
er
pre
s
ent
s broa
d appli
c
ation
pro
s
pe
cts.
Referen
ces
[1]
X
u
Z, Wakileh GJ.
Pow
e
r System H
a
rmon
i
cs: F
unda
me
n
t
als, Analysis
and F
ilter D
e
s
i
gn
. Ch
in
a
Machi
ne Press
.
2003.
[2]
W
a
kileh GJ. Harmonics
and r
o
tation mac
h
i
n
es.
Electric Po
w
e
r Systems Researc
h
. 200
3; 66: 5-14.
[3]
Y Kusuma L, YP Obulesu. Un
i
f
ied Po
w
e
r Qualit
y
Con
d
iti
o
n
e
r for voltage s
ag an
d harmo
n
i
c mitigatio
n
of no
nli
n
e
a
r l
o
ads.
T
E
LKOM
NIKA (T
elec
o
m
mu
nicati
on
Co
mp
uting
El
e
c
tronics
and
C
ontrol)
. 20
11;
1(1): 65-7
4
.
[4]
Soo HC, Gils
oo J.
T
i
me-freque
nc
y
ana
l
y
s
i
s of
po
w
e
r-q
u
a
lit
y
disturb
a
n
c
es via the Gabor-W
ig
ner
transform.
IEEE Transactions
on Power Delivery
. 2009; 2
5
(
1
): 494-4
99.
[5]
Han
Y,
Xu
L,
Muhamm
ad M
K
. Rob
u
st d
e
a
dbe
at co
nt
rol s
c
heme
for
a h
y
brid
APF
w
i
th
r
e
setting
filt
e
r
and Ad
ali
ne-
ba
sed harm
onic
estimatio
n
alg
o
r
ithm.
IEEE Transacti
ons on i
ndustri
a
l el
ectronics
. 20
11;
58(9): 38
93-
39
02.
[6]
Y Kusumal
a
th
a, Ch Sai
bab
u
.
Harmonics
Mitigati
on of Industri
a
l Moto
r Drives
w
i
th
Active Po
w
e
r
F
ilters in
Cem
ent Pl
ant-A C
a
se Stu
d
y
.
T
E
LKOMNIKA (T
elec
o
m
mun
i
cat
i
on
Co
mputin
g
Electro
n
ic
s
and C
ontrol)
. 201
2; 2(1): 1-8
.
[7]
Kuslj
e
vic MD.
Simulta
neo
us f
r
equ
enc
y
an
d
harmo
nic ma
g
n
itud
e estim
a
ti
on us
in
g d
e
co
u
p
le
d mo
dul
es
and mu
ltirate s
a
mpli
ng.
IEEE Transactions on Instru
mentation an
d Meas
urement
. 201
0; 59(4): 95
4-
962.
[8]
Barros J, Di
eg
o RI. App
licati
o
n of the
w
a
v
e
l
e
t-pa
cket tra
n
s
f
orm to the
esti
mation
of harm
onic
gro
ups
in curre
nt and
voltag
e
w
a
v
e
fo
rms.
IEEE
Transactions on P
o
wer Deliv
ery
. 200
6; 21(1): 53
3-53
5.
[9]
Hua
ng
L, W
u
S, Li
X.
Re
du
eed-r
ank M
D
L
meth
o
d
for
so
urce enum
erati
on in hi
gh-res
o
luti
on arr
a
y
Proeess
i
ng.
IEEE Transactio
n
s on Sig
n
a
l
Proeess
i
ng
. 20
0
7
; 55(22): 5
658
-566
7.
[10]
Suresh M, An
up KP. PI and
fuzz
y
lo
gic co
ntroll
er bas
ed
3-Phas
e 4-W
i
r
e
shunt
activ
e
filters for the
mitigati
on of c
u
rrent h
a
rmon
i
cs
w
i
t
h
the
id-i
q contro
l strateg
y
.
Jour
na
l o
f
Pow
e
r Electronics
. 20
11;
11(6): 91
4-9
2
1
.
[11]
W
ija
yaku
las
oor
i
y
a JV, Putrus GA.
F
a
st n
on-re
c
u
rsive e
x
tractio
n
of in
divid
u
a
l
harm
onics us
i
n
g
artificial
ne
ura
l
net
w
o
rks.
IEEE Proceedings-Gen
eration, Trans
m
i
ssi
on and
Distribution
. 2
005;
152(
4): 539-
54
3.
Evaluation Warning : The document was created with Spire.PDF for Python.