A SYSTEM FOR THE GENERATION AND DETECTION
OF ELECTRICAL DISTURBANCES
Iñigo Monede o1, Ca los León1, Jo ge Rope o1, José Luis de la Vega2,, Juan C. Mon año2, José Manuel Elena1
1School o Compu e Science and Enginee ing 2IRNAS
Elec onic Technology Depa men Campus Reina Me cedes
A da, Reina Me cedes s/n P.O. Box 1052
41012 Se ille (Spain) 41080 Se ille (Spain)
con ac email: imonede [email protected] phone: 34-954624711
Abs ac . Powe Quali y is de ined as he s udy o he
quali y o elec ic powe lines. The de ec ion and
classi ica ion o he di e en dis u bances which cause
powe quali y p oblems is a di icul ask which equi es a
high le el o enginee ing expe ise. Thus, neu al ne wo ks
a e usually a good choice o he de ec ion and
classi ica ion o hese dis u bances. This pape desc ibes a
powe ul sys em, de eloped by he Ins i u e o Na u al
Resou ces and Ag obiology a he Scien i ic Resea ch
Council (CSIC) and he Elec onic Technology
Depa men a he Uni e si y o Se ille, o he gene a ion
and de ec ion (by means o neu al ne wo ks) o elec ical
dis u bances.
Key Wo ds
Powe quali y, elec ical dis u bance, Wa ele ans o m,
neu al ne wo k.
1.- In oduc ion
Powe Quali y
Powe Quali y (PQ) is de ined as he s udy o he quali y
o elec ic powe lines. PQ has been a opic o
conside a ion o he las wo decades, and has ecen ly
acqui ed in ensi ied in e es due o he wide sp ead use o
elec onical de ices in complica ed indus ial p ocesses
and he gene alised powe quali y o comme cial elec ic
powe [1]. Thus nowadays, cus ome s demand highe
le els o PQ o ensu e he p ope and con inued ope a ion
o such sensi i e equipmen .
The poo quali y o elec ical powe is no mally a ibu ed
o powe line dis u bances such as wa eshape aul s,
o e ol ages, capaci o swi ching ansien s, ha monic
dis o ion and impulse ansien s. Thus, elec omagne ic
ansien s, which a e momen a y ol age su ges powe ul
enough o sha e a gene a o sha , can cause ca as ophic
damage suddenly. Ha monics, some imes e e ed o as
elec ical pollu ion, a e dis o ions o he no mal ol age
wa e o ms ound in ac ansmission, which can a ise a
i ually any poin in a powe sys em. While ha monics
can be as des uc i e as ansien s, o en he g ea es
damage om hese dis o ions lies in he loss o
c edibili y o he powe u ili ies is-a- is hei cus ome s.
The classi ica ion and iden i ica ion o each one o he
dis u bances is no mally ca ied ou om s anda ds and
ecommenda ions depending on whe e he u ili ies ope a e
(IEEE in he Uni ed S a es, UNE in Spain, e c).
Figu e 1.- S anda d IEEE 1159
The de ec ion and classi ica ion o he di e en
dis u bances which cause powe quali y p oblems is a
di icul ask which equi es a high le el o enginee ing
expe ise [2]. Due o he abo e men ioned di icul ies,
a i icial in elligence ools [3] eme ge as an in e es ing
al e na i e in he de ec ion o elec ical dis u bances. The
main in elligen ools o in e es include expe sys ems,
uzzy logic and a i icial neu al ne wo ks (ANNs) [4].
442-286
180
Neu al Ne wo ks on Powe Quali y
Fo he de ec ion and classi ica ion o dis u bances, ANNs
can be combined wi h ma hema ical analysis such as
Fou ie and Wa ele ans o ms o he gene a ion o
signal ea u es which se e as inpu s in he ne wo k [5].
Thus, ea u e ex ac ion by wa ele ans o ms p o ides
an unique cha ac e is ic which can ep esen e e y single
PQ dis u bance a di e en esolu ions using he
echnique called mul i- esolu ion signal decomposi ion o
mul i esolu ion analysis. In his way, while he de ec ion
o he powe quali y signals has ended o be easy, hei
classi ica ion is s ill a di icul ask in which ANNs play
an impo an ole [6][7][8].
Pa e n ecogni ion in ANNs gene ally equi es
p ep ocessing o da a, ea u e ex ac ion and inal
classi ica ion. One o he mos impo an asks in he
design and de elopmen p ocess o an ANN is o gene a e
an adequa e numbe o aining pa e ns in o de o
app oxima e u u e inpu s. Some imes an op imal design
o he ANN is ound bu he limi ed numbe o ain
pa e ns does no gi e good esul s. In pa icula , in PQ a
g ea numbe s o elec ical pa e ns a e necessa y due o
he mul iple combina ions o di e en dis u bances which
can coincide in one o a ious samples. Ano he
addi ional p oblem wi h ANNs applied o PQ is he
impossibili y o ge ing eal pa e ns di ec ly om he
powe line due o he i egula i y in he appa i ion o
dis u bances.
2.- Elec ical Pa e n Gene a o
Fo he ask o aining neu al ne wo ks o he de ec ion
and classi ica ion o elec ical dis u bances we a e
de eloping an elec ical pa e n gene a o . The objec i e
o his gene a o is o c ea e an unlimi ed numbe o
pa e ns o be used by a classi ica ion sys em.
The Elec ical Pa e n Gene a o make i possible o
con igu e pa ame e s such as he du a ion o he sample,
he equency o he signal and he numbe o samples in
an ideal cycle (50Hz o 60Hz) and o add one o mo e
dis u bances. F om he selec ed pa ame e s, he gene a o
c ea es a ex ile wi h he ol age alues o he sample.
The s uc u e o he ile consis s o a heade wi h he ile
in o ma ion (name, numbe o sample cycles and
sampling pe iod) and a da a column co esponding o he
ol ages o each o he samples. One ile example om
he Elec ical Pa e n Gene a o can be seen in Figu e 2.
HEADER
Name: Impulse-01. x
Ideal cycles: 5
Sampling pe iod (ms): 0.156250
END HEADER
-28.171582
-12.934081
2.334579
17.597614
32.818256
47.959835
62.985875
77.860176
92.546906
107.010682
121.216659
...
Figu e 2.- File om he Elec ical Pa e n Gene a o
The ype o dis u bances includes: impulse, oscilla ion,
sag, swell, in e up ion, unde ol age, o e ol age,
ha monics and equency a ia ions. In ampli ude
dis u bances (impulses, sags, swells, in e up ions,
unde ol ages and o e ol ages), he ool allows us ha
pa ame e s such as ampli ude, s a ime, inal ime, ising
and alling slope, be con igu ed. The edi ion o ha monics
allows he con igu a ion o ampli ude and phases as a as
o y ha monic le els including he possibili y o adding
hem an o se .
Figu e 3.- Ideal signal (one-phase ep esen a ion)
Models o dis u bances
1.- Ha monic dis o ion
Ha monic dis o ion is de ined as he phenomenon in
which di e se sinusoidal signals wi h di e se equencies
which a e mul iples o he undamen al equency a e
supe posed on he ideal signal (Figu e 4).
181
Figu e 4.- Ha monic dis o ion
The ollowing ma hema ical model was implemen ed in
he gene a o :
1
() sin(2 )
N
iii
i
C A A
π
ϕ
=
=+ +
∑ (1)
• A: DC e m (V).
• Ai: Ampli ude o he i h ha monic o signal (V).
• i: F equency o he i h ha monic o signal (Hz)
• φi: Phase o he i h ha monic (Rad)
• i: Ha monic o de (i= 1,.., N).
In ou ha monic model C( ) is conside ed as consis ing o
a undamen al and 39 ha monic componen s.
2.- F equency de ia ion
F equency de ia ion is a signal dis u bance added o he
ha monic dis o ion. The model consis s o he equency
modula ion o he signal C( ) by means o he ca ie
signal M( ), which is named modula ing signal.
Figu e 5.- F equency de ia ion
The ma hema ical exp ession o his signal is:
) j sin(BB) (M jm
j
j
ϕπ
++= ∑
=
2
40
1
(2)
• B: DC e m (V).
• Bj: Ampli ude o he j h ha monic o signal (V)
• m: Fundamen al equency (Hz)
• φi: Phase o he i h ha monic (Rad)
Conside ing (1) and (2) he esul an signal would end up
as:
∑∑
==





+++++= 40
1
40
1
22
ij
jmjici ) j sin(BB i sinAA) (X
ϕπϕπ
(3)
which shows he ha monic con en and equency
modula ion.
The addi ion o ano he kind o dis u bance was ca ied
ou om he p e ious exp ession X( ). The e o e, he
p e ious exp ession would be he esul o an ideal
elec ical signal o a equency o /and ha monic dis u bed
signal.
3.- O e ol ages, swells, unde ol ages and sags.
In his kind o dis u bances he ampli ude o he signal
ises (o e ol ages o swells) o alls (unde ol ages and
sags) a ce ain alue along a ime in e al.
Figu e 6.- Sag
In he de elopmen o he dis u bance gene a o , a
apezoidal model o he ampli ude e olu ion (lineal
slope) was conside ed. The model makes i possible o
app oxima e he ampli ude dis u bances mos equen ly
encoun e ed in powe sys ems. Figu e 7 shows a g aphical
o he model used o o e ol ages o swells (in e se
apeze o unde ol ages and sags).
Figu e 7.- O e ol age and swell model
• p: ini ial sample o he apeze
• pip: slope o he ini ial amp
182
• p p: slope o he inal amp
• n1: numbe o samples o he ini ial amp
• n2: numbe o samples o he inal amp
• : numbe o samples when he climb is eached
• selec : o al numbe o samples
Figu e 8.- O e ol age
4.- T ansien s
The elec ical pa e n gene a o models ansien s as a
damped sine h ough a supe posed exponen ial unc ion,
which is added o X( ) a a ce ain poin .
Figu e 9.- T ansien
The implemen ed ma hema ical model obeys o:
) sin(Ae) (T
a
ϕπ
+= −2 (4)
• a: ansi o y exponen .
• A : ampli ude o he ipple (V)
• : equency o he ipple (Hz)
• φ : ini ial phase o he ipple (Rad)
5.- Noise
The gene a o makes i possible o add Addi i e Whi e
Gaussian Noise (AWGN) (Figu e 10) in o de o simula e
mo e ealis ic signals o he powe line.
Figu e 10.- A signal wi h AWGN
3.- Elec ical dis u bances classi ie based on
Neu al Ne wo ks.
In addi ion o he gene a ion sys em, we ha e de eloped a
i s e sion o a sys em o he de ec ion and
classi ica ion o elec ical dis u bances. The sys em is a
de ec o o powe line dis u bances based on a i icial
in elligence echniques (in pa icula , a i s e sion based
on ANNs). We ha e used he p e iously desc ibed
Elec ical Pa e n Gene a o in o de o ca y ou he ANN
aining. A he momen , we ha e gene a ed a ound one
housand pa e ns o elec ical signals which include all
he abo e-men ioned dis u bances.
The de ec ion sys em uses Wa ele ans o m o he
acqui ed signal o he gene a ion o signal ea u es
[5][6][7][8]. The aim o ea u e ex ac ion by Wa ele
ans o ms was o p o ide an unique cha ac e is ic which
can ep esen e e y single PQ dis u bance.
Fo p og amming asks we ha e used he MATLAB ool
which has some powe ul oolboxes (specialized
unc ions) o signal, Wa ele s and Neu al Ne wo ks.
The inpu ec o s o he ANN a e gene a ed ca ying ou a
numbe o ope a ions on he Wa ele ans o m. I is
known he Wa ele ans o m de ec s be e he slow
a ia ions in he las le els and as a ia ions in i s
le els. So, ou solu ion is based on he concep o ha he
ampli ude dis u bances would be be e de ec ed in he
i s le els o Wa ele ans o m while he equency ones
would be be e de ec ed in he las le els. The e o e, we
ha e used wo pa allel neu al ne wo ks. One o hem is
used o de ec he ampli ude dis u bances and he o he
one he equency dis u bances.
In pa icula , we ha e used he ollowing alues as inpu
ec o o he ampli ude neu al ne wo k: he VRMS o he
signal, he in eg al, he maximum and he VRMS o he
de ail coe icien s o 1,2 and 3 le el.
Capas
Capa
Capa
de en ada ocul as
de salida
Figu e 11.- Mul ilaye pe cep on
Inpu
laye
Hidden
laye s
Ou pu
laye
183
The kind o ANN is a mul ilaye pe cep on wi h 3 hidden
laye s wi h 20, 14 and 8 neu ons espec i ely. We ha e
eached his s uc u e a e es ing a g ea numbe o
di e en numbe s o neu ons and laye s. Al hough we
ha e es ed o he kinds o s uc u es like LVQ and RBF a
he momen he bes esul s ha e been eached wi h he
pe cep on.
In equency we ha e used he ollowing alues o he
inpu ec o o he ne wo k: he VRMS o he signal, he
in eg al, he maximum and he VRMS o he de ail
coe icien s o 6,7 and 8 le el. The s uc u e o he ANN
o equency is a mul ilaye pe cep on wi h 3 hidden
laye wi h 4,3 and 2 neu ons espec i ely.
In he aining o he ne wo ks we used 80% o he
gene a ed signals as aining pa e ns and 20% as es
pa e ns. In he i s ne wo k we eached he bes esul s
wi h a aining o 180 epochs ge ing he igh
classi ica ion o he 93.82% o he es signals. Besides,
mos o he inco ec ly de ec ed dis u bances we e in he
h eshold wi h o he kind o dis u bance. The kind o
di e en ampli ude dis u bances as well as he pe cen o
co ec ly de ec ed dis u bances in each one is shown in
able 1.
Table 1.- Resul s in ampli ude dis u bances
Dis u bance Tes
signals
Numbe
o e o s
Co ec ly
de ec ed %
Sag 51 2 96.08
Ideal 56 6 89.29
Swell 25 2 92.00
Unde ol age 21 0 100
O e ol age 38 4 89.47
F equency de ia ion 52 1 98.08
To al 243 15 93.83
On he o he hand, in he equency ne wo k we e
co ec ly classi ied he 97.53% wi h a aining o 261
epochs. Wi h he help o he gene a o o dis u bances, we
a e cu en ly wo king in he gene a ion o new pa e ns
o he aining o he ANN. E en when he cu en esul s
a e e y good, we expec o ge be e hem wi h he
inc ease o he numbe o pa e ns.
4.- Conclusions
Today i is known ha neu al ne wo ks a e a good choice
o de ec ing and classi ying elec ical powe
dis u bances. In he huge li e a u e [4][5][6][7] abou
de ec ion o elec ical dis u bances which we can ind,
o en he p oblem lies in gene a ing a su icien numbe o
aining pa e ns o ge ha neu al ne wo k ob ains good
esul s in u u e inpu s. Wi h he help o he gene a o is
possible o ca y ou a aining o he ANN su icien ly
en i e in o de o ge eliable esul s.
Thus, we ha e de eloped an elec ical pa e n gene a o
which is capable o gene a ing common dis u bances
which can be ound in a powe line wi h he aim o
making he aining o neu al ne wo ks easie . Besides,
we ha e de eloped a i s e sion o a classi ie sys em
based on ANNs.
Figu e 12.- E olu ion o aining o he equency ANN
Ou inal objec i e is o implan his de ec o as a eal-
ime sys em capable o classi ying dis u bances on line. In
o de o ge a as sys em, we a e cu en ly wo king in
educing and compiling he p ocessing o he signal
p e ious o ANN. We a e es ing o he possibili ies o
p ocessing o he signal and o he s uc u es o he ANN
in o de o ge be e he esul s.
Re e ences
[1] M. McG anaghan, B. Roe ge , Economic E alua ion
o Powe Quali y, IEEE Powe Enginee ing Re iew, Feb
2002.
[2] A.Hussain, M.H. Sukai i, A. Mohamed, R. Mohamed,
Au oma ic De ec ion O Powe Quali y Dis u bances and
Iden i ica ion o T ansien s Signals”, In e na ional
Symposium on Signal P ocessing and i s Applica ions,
Kuala Lumpu , Malaysia, 13-16 Augus 2001.
[3] R. Adapa, Powe Quali y Analysis So wa e, IEEE
Powe Enginee ing Re iew, Feb ua y 2002, 0272-
1724/02.
[4] W.R. Anis Ib ahim and M.M. Mo cos, A i icial
In elligence and Ad anced Ma hema ical Tools o Powe
Quali y Applica ions: A Su ey, 668 IEEE T ansac ions
on Powe Deli e y, Vol. 17, Ap il 2002
[5] G. Zheng, M.X. Shi, D. Liu, J. Yao, Z.M. Mao, Powe
Quali y Dis u bance Classi ica ion Based on Rule-Based
and Wa ele -Mul i-Resolu ion Decomposi ion,
P oceedings o he Fi s In e na ional Con e ence on
Machine Lea ning and Cybe ne ics, Beijing, 4-5
No embe 2002
184

[6] R. Daniels, Powe quali y moni o ing using neu al
ne wo ks, P oc.1s In . Fo um Applica ions Neu al
Ne wo ks Powe Sys ., 1991, pp. 195–197.
[7] C. Xiangxun, Wa ele -based Measu emen and
Classi ica ion o Powe Quali y Dis u bances, Powe
Enginee ing Socie y Win e Mee ing, 2002, ol. 2, 2002.
[8] D. Bo ás, M. Cas illa, N. Mo eno and J.C. Mon año,
Wa ele and neu al s uc u e: a new ool o diagnos ic o
powe sys em dis u bances, IEEE T ans. on Indus y
Applica ions, Vol. 37, No. 1, pp. 184-190, 2001.
Acknowledgemen s
The wo k desc ibed in his pape has been suppo ed by
he Spanish Minis y o Science and Technology (MCYT:
Minis e io de Ciencia y Tecnología) h ough p ojec
e e ence numbe DPI2002-04420-C03-03.
185