Anda di halaman 1dari 8

Electric Power Systems Research 79 (2009) 14081415

Contents lists available at ScienceDirect


Electric Power Systems Research
j our nal homepage: www. el sevi er . com/ l ocat e/ epsr
A new reactive, distortion and non-active power measurement method for
nonstationary waveforms using wavelet packet transform
Walid G. Morsi, M.E. El-Hawary

Department of Electrical and Computer Engineering, Dalhousie University, PO Box 1000, Halifax, Nova Scotia, Canada
a r t i c l e i n f o
Article history:
Received 15 September 2007
Accepted 22 April 2009
Available online 7 June 2009
Keywords:
Power system harmonics
Wavelet packet transform
Nonstationary waveforms
a b s t r a c t
The denitions of power components that are contained in the IEEE Standard 1459-2000 [IEEE Std. 1459-
2000, Denitions for the measurement of electric quantities under sinusoidal, non-sinusoidal, balanced
or unbalancedconditions, 2000] are basedonthe Fourier transform(FT) whichis suitable only for the case
of stationary waveforms. However, for nonstationary waveforms, the FT produces large errors. Therefore,
the power components based on this transformbecome inaccurate. Anewapproach based on the wavelet
packet transform(WPT) for dening these power components is developed in this paper. The advantages
of using the wavelet transformare that it canaccurately represent andmeasure nonstationary waveforms,
providing uniform frequency bands while preserving both time and frequency information. In addition,
this paper addresses theproblemof choosingthemost appropriatemother wavelet for power components
measurements. The results of applying bothapproaches to stationary and nonstationary waveforms show
that the currently used denitions according to the IEEE Standard 1459-2000 result in large errors for the
case of nonstationary waveforms while the proposed approach (WPT based) gives more accurate results
in this situation.
Crown Copyright 2009 Published by Elsevier B.V. All rights reserved.
1. Introduction
Electric power system waveforms can be classied into two
types; stationary and nonstationary waveforms. Stationary wave-
formis dened as a waveformwhere all its statistical properties do
not change with time [2]. The main loads that generate stationary
waveforms are personal computers (PCs), televisions, energy sav-
ing lamps, HVDC links, exible ac transmission (FACT) and other
power-electronic converters.
On the other hand, nonstationary waveforms are those with sta-
tistical properties that change with time and are also called time
evolving waveforms. Electric arc furnaces (EAFs), electric welders,
capacitor switching, motor starting and transformers energization
are examples of loads that generate these nonstationary wave-
forms.
Power components such as active, reactive, distortion, non-
active and apparent power are dened by the IEEE Standard
1459-2000 [1] using fast Fourier transform (FFT). For the case of
stationary waveforms, the FFT can provide accurate results; how-
ever, for the case of nonstationary waveforms, the FFT introduces
large errors due to spectral leakage and picket fence phenomena
[3]. Fig. 1 shows a stationary current waveformof frequency 180Hz

Corresponding author. Tel.: +1 902 494 6198; fax: +1 902 422 7535.
E-mail address: elhawary@dal.ca (M.E. El-Hawary).
withamplitude 50A. The FFT spectrumshows exactly the third har-
monic at 180Hz and amplitude 50 with no spectral leakage. Fig. 2
shows an example of a nonstationary current waveform at 180Hz.
The current amplitude changes with time; it starts from an ini-
tial value of 50A for 0.083s and then the amplitude drops to 10A.
The FFT spectrum shows a 30A component at 180Hz (which does
not exist in the original waveform) plus other components that are
spread along the frequency axis. Spectral leakage is very clear here
which leads to inaccurate results. Moreover the time at which the
waveform drops from 50 to 10A is not accessible through the FFT
because it provides amplitudefrequency spectrum so therefore
time information is lost.
Basedonthis fact, the measurement of these power components
according to the IEEE Standard 1459-2000 [1] would be accept-
able only for the case of stationary waveforms but it would not
be acceptable for nonstationary waveforms which exist in large
amounts due to todays electric power system non-stop dynamic
operating conditions.
The wavelet transform, whichis considered as a timefrequency
transform, is capable of handling and accurately representing
nonstationary waveforms in too many applications in different dis-
ciplines [430]. The power of the wavelet stems from the fact that
it can provide variable frequency resolution while preserving time
information. This is an important requirement for the analysis and
measurement of the nonstationary waveforms whichpossess time-
variant characteristics and this is lost when using the FFT.
0378-7796/$ see front matter. Crown Copyright 2009 Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.epsr.2009.04.018
W.G. Morsi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 14081415 1409
Fig. 1. Stationary waveform. (a) Time-domain waveform and (b) FFT spectrum.
Our previous work for calculating and measuring the power
components was concerned with steady-state harmonic distortion
using discrete wavelet transform as in [31,32], while [33,34] were
concerned with harmonics and inter-harmonics distortion using
wavelet packet transform.
This paper, introduces a new approach for calculating and
measuring the active, reactive, non-active, distortion and appar-
ent power for the case of electric power system nonstationary
waveforms or time-varying harmonics using the wavelet packet
transform (WPT).
Fig. 2. Nonstationary waveform. (a) Time-domain waveform and (b) FFT spectrum.
Section 2 briey reviews the FFT based power components de-
nitions that have been adopted by the IEEE Standard 1459-2000 [1],
while Section 3 provides the basics for the wavelet transform and
wavelet packet transform. Section4contains the newapproachthat
has been developed for dening these power components using
WPT, Section 5 includes numerical examples and nally Section 6
contains our conclusions.
2. Power components according to the IEEE Standard
1459-2000
In nonsinusoidal situations the voltage and current can be
expressed according to [1] as follows:
v(t) = v
1
(t) +v
H
(t), i(t) = i
1
(t) +i
H
(t) (1)
v
1
(t) =

2V
1
sin(t
1
), i
1
(t) =

2I
1
sin(t
1
) (2)
v
H
=

h / = 1
V
h
sin(ht
h
) (3)
i
H
=

h / = 1
I
h
sin(ht
h
) (4)
where the subscript 1 indicates the fundamental component,
while the subscripts H and h refer to the non-fundamental
components andthe harmonic order. Here
1
,
1
represent the fun-
damental voltage and current phase angle, respectively, while
H
and
H
represent the harmonic voltage and current phase angle,
respectively.
The fundamental, harmonic and total active powers are dened
as
P
1
= V
1
I
1
cos
1
, P
H
=

h / = 1
V
h
I
h
cos
h
,
1
=
1

1
(5)
P = P
1
+P
H
(6)
where the RMS value for the voltage and currents are
V =

1
T
_
T
0
v
2
(t)dt, I =

1
T
_
T
0
i
2
(t)dt (7)
V
2
= V
2
1
+V
2
H
, V
2
H
=

h / = 1
V
2
h
(8)
I
2
= I
2
1
+I
2
H
, I
2
H
=

h / = 1
I
2
h
(9)
The fundamental reactive power, Budeanus harmonic reactive
power and Budeanus reactive power are dened as
Q
1
= V
1
I
1
sin
1
, Q
H
=

h / = 1
V
h
I
h
sin
h
, Q
B
= Q
1
+Q
BH
(10)
Theapparent power S, non-activepower NandBudeanus distortion
power D
B
are
S = VI, N =
_
S
2
P
2
, D
B
=
_
N
2
Q
2
B
(11)
S
2
= P
2
+Q
2
B
+D
2
B
(12)
3. Wavelet transform
3.1. Introduction
The wavelet transform is a timefrequency representation of
any waveform s(t) using a stretched (or squeezed) and translated
basis function called wavelet. This basis function is called mother
1410 W.G. Morsi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 14081415
Fig. 3. Analysis bank for extracting scaling and wavelet coefcients.
wavelet while its stretched (or squeezed) and translated versions
are called daughter wavelets. Choosing scale and time parameters
a and b, respectively, the wavelet coefcients in discrete form can
be obtained from:
C(2
j
, k2
j
) =
_
s(t),
j,k
_
= 2
j/2
_
s(t)

(2
j
t k)dt (13)
where
j,k
is the wavelet function for scale index j and time index
k, such that:
a = 2
j
, b = ka, j, k Z
In order to make the process of wavelet coefcients calculation
faster, Mallat [35] developed the multi-resolution analysis (MRA)
algorithm by introducing the two scale equations:
(2
j
t) =

k
h
j+1
(k)(2
j+1
t k) (14)
(2
j
t) =

k
g
j+1
(k)(2
j+1
t k) (15)
where is the scaling function, h and g are the low pass and high
pass lters. The scaling coefcients c
j,k
and the wavelet coefcients
d
j,k
at eachdecompositionlevel j andsampleindexkcanbeobtained
recursively by
c
j,k
=

l
c
j1,l
h(l 2k) =

s(t),
j1,k
(t)
_
(16)
d
j,k
=

l
c
j1,l
g(l 2k) =

s(t),
j1,k
(t)
_
(17)
Fig. 3 shows one decomposition level for extracting the scaling
and wavelet coefcients using low and high pass lter bank.
3.2. Wavelet packet transform
The wavelet packet transform is a generalization of the wavelet
transform where the decomposition is performed on both the
scaling and wavelet coefcients instead of only decomposing the
scaling coefcients as in the case of discrete wavelet transform
(DWT). This process offers the full decompositiontree whichmakes
the decompositionvery exible while providing uniformfrequency
sub-bands. The wavelet basis at any level j, tree node n and sample
time k can be generated using the following recursive formula [36]:

2n
j,k
(t) =

m
h(m)
n
j1,2km
(t) (18)

2n+1
j,k
(t) =

m
g(m)
n
j1,2km
(t) (19)
For n=0 (node zero),
0
j,k
(t) is the scaling function
j,k
(t) while at
any other node (n / = 0)
n
j,k
(t) is the wavelet function. Fig. 4 shows
the wavelet packet transform decomposition tree with all its tree
nodes (or tree leaves). Note that at the last decomposition level the
tree nodes are called terminal nodes (or terminal tree leaves).
The wavelet packet transform is more suitable than the fast
Fourier transform (FFT) or the short time Fourier transform (STFT)
whichhave xedwindowwidth, for analyzing nonstationary wave-
forms since it provides exible (or variable) window size therefore
preservingbothtime andfrequencyresolutions [3]. This means that
at any decomposition level in the WPT some nodes can be merged
or split in order to meet the required resolution, while in case of
STFT any attempt to change the frequency resolution, time resolu-
tionwill beaffectedaccordingtoHeisenbergs uncertaintyprinciple
[37].
Moreover, WPT is favored over DWT since it can provide
uniform frequency bands, offers exible decomposition through
mergingsplitting process of the nodes. Also since the WPT can be
considered as generalization of the wavelet transform, DWT there-
fore represent a special case of the WPT and can be obtained by
suitable merging of the details while keeping the decomposition
process being applied on the approximations only.
The advantages offered by the WPT come at the expense of the
computational burden; however, the following recursive algorithm
for coefcients calculation may greatly reduce the computational
effort to be the same as the FFT [36]:
d
2n
j
(k) =

m
h(m)d
n
j1
(2k m) (20)
d
2n+1
j
(k) =

m
g(m)d
n
j1
(2k m) (21)
3.3. Wavelet choice
It is clear from (16) and (17) that the values of coefcients vary
depending on the choice of the basis functions and . The prop-
erties of these basis functions along with the required application
mainly determine the choice of suitable basis functions. The follow-
ing properties are required when analyzing electric power system
waveform for computing the power components:
3.3.1. Orthogonality
If the basis functions are orthogonal then the analysis bank
becomes the time reverse of the synthesis bank.
3.3.2. Compactness support
The scaling and wavelet functions are time limited (not extend-
ing fromto +as the case in Fourier foundation). This property
Fig. 4. Wavelet packet transform decomposition tree.
W.G. Morsi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 14081415 1411
Fig. 5. Scaling function (phi) and wavelet function (psi) for (a) db1, (b) db10 and (c)
db43.
is very useful because it can offer good time localization without
affecting the frequency resolution therefore allowing the correct
detection and representation of waveform containing sudden and
rapid changes due to transients.
3.3.3. Number of vanishing moments
By letting the wavelet decrease very fast with decreasing the
scale a, therefore increasing the number of vanishing moments will
lead to more atness in the lters frequency response. This is very
important for identifying the sub-bands and also this reduces the
transitionbandhencesharper characteristics canbeachieved. Fig. 5
shows the scaling function (phi) and wavelet function (psi) for
Daubechies Wavelets; db1, db10 and db43 while Fig. 6 shows the
frequency response of the low and high pass lters.
Since Daubechies is able to handle all the aforementioned prop-
erties, it canbe concludedthat dbwithhighest number of vanishing
Fig. 6. Frequency response of low and high pass lters for db1, db10 and db43. (a)
and (b) Reconstruction bank while (c) and (d) decomposition bank.
1412 W.G. Morsi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 14081415
moments can be considered as the best candidate mother wavelet
for calculating the power components in case of nonstationary
power system waveforms.
Noting that going higher than db43 may lead to instability in
the algorithm used for the computation of dbN scaling lter which
affects the lters frequency response. This is due to that computing
the scaling lter requires the extraction of the roots of a polynomial
of order 4N[36]. So for N too large, the extraction process may lead
toinstability that affects the frequency response of the lowandhigh
pass lters. Moreover db higher vanishing moments means that
more lter coefcients will be processed which could inuence the
required memory size and the computational effort.
Accurate measurements of power components is a main objec-
tive because based onthese measured values the consumers will be
charged, the power quality will be evaluated and the amount and
type of compensators will be determined. For these reasons db43
will be favored in this study and its performance will be compared
with other db wavelet of lower number of vanishing moments.
4. Wavelet packet based power components
This section introduces a new approach for calculating and
measuring the active, reactive, non-active, distortion and appar-
ent power based on the wavelet packet transform (WPT). In the
timefrequency domain, voltage and current waveform with 2
N
sample point canbe expressedinterms of the wavelet packet trans-
form coefcients according to the following equations:
v(t) =
2
Nj
1

k=0
d
0
j
(k)
j,k
(t) +
2
j
1

m=1

2
Nj
1

k=0
d
2m
j
(k)
m
j,k
(t)

= v
0
j
+
2
j
1

m=1
v
m
j
(22)
The root mean square (RMS) value of the voltage:
V =

_
1
2
N
2
Nj
1

k=0
(d
0
j
(k))
2
+
1
2
N
2
j
1

m=1
2
Nj
1

k=0
(d
m
j
(k))
2
=

_
(V
0
j
)
2
+
2
j
1

m=1
(V
m
j
)
2
(23)
Also the current is expressed in the timefrequency domain
using WPT as follows:
i(t) =
2
Nj
1

k=0
d
0
j
(k)
j,k
(t) +
2
j
1

n=1

2
Nj
1

k=0
d
2n
j
(k)
n
j,k
(t)

= i
0
j
+
2
j
1

n=1
i
n
j
(24)
The root mean square value of the current:
I =

_
1
2
N
2
Nj
1

k=0
(d

0
j
(k))
2
+
1
2
N
2
j
1

n=1
2
Nj
1

k=0
(d

n
j
(k))
2
=

_
(I
0
j
)
2
+
2
j
1

n=1
(I
n
j
)
2
(25)
where d
m
j
and d

n
j
represent the voltage and current WPT coef-
cients, respectively at level j, voltage node m and current node n.
The active power P can be calculated as
P = P
0
j
+
2
j
1

m = 1, n = 1
m = n
P
m
j
(26)
where P
0
j
is the active power at node zero and level j:
P
0
j
=
1
2
N
2
Nj
1

k=0
d
0
j
(k)d
0
j
(k) (27)
While the active power at any other node where m=n:
P
m
j
=
1
2
N
2
Nj
1

k=0
d
m
j
(k)d
n
j
(k) (28)
The total current can be decomposed into two components: the
active component I
n
aj
that is responsible for active power trans-
mission and residual current component I
n
resj
which represent the
non-active power, and this component should be compensated
I =

_
(I
0
aj
)
2
+(I
0
resj
)
2
+
2
j
1

n=1
(I
n
aj
)
2
+
2
j
1

n=1
(I
n
resj
)
2
(29)
where
I
0
aj
=

_
1
2
N

2
Nj
1
k=0
(d
0
j
(k))(d

0
j
(k))
2
2
Nj
1
k=0
(d
0
j
(k))
2

2
Nj
1
k=0
(d
0
j
(k))
4
(30)
I
n
aj
=

_
1
2
N

2
Nj
1
k=0
(d
n
j
(k))(d

n
j
(k))
2

2
Nj
1
k=0
(d
2n
j
(k))
2

2
Nj
1
k=0
(d
n
j
(k))
4
(31)
(I
0
resj
)
2
=
1
2
N
2
Nj
1

k=0
(d
0
j
(k))
2
+

2
Nj
1
k=0
(d
0
j
(k))
2
(d

0
j
(k))
2
2
Nj
1
k=0
(d
0
j
(k))
2

2
Nj
1
k=0
(d
0
j
(k))
4
2

2
Nj
1

k=0
(d
0
j
(k)

2
Nj
1
k=0
d
0
j
(k)d

0
j
(k)

2
Nj
1
k=0
(d
0
j
(k))
2
2
Nj
1

k=0
d
0
j
(k) (32)
(I
n
resj
)
2
=
1
2
N

2
Nj
1

k=0
(d
2n
j
(k))
2
+

2
Nj
1
k=0
(d
n
j
(k))
2
(d

n
j
(k))
2

2
Nj
1
k=0
(d
2n
j
(k))
2

2
Nj
1
k=0
(d
n
j
(k))
4
2

2
Nj
1

k=0
(d
2n
j
(k)

2
Nj
1
k=0
d
n
j
(k)d

n
j
(k)

2
Nj
1
k=0
(d
n
j
(k))
2
2
Nj
1

k=0
d
2n
j
(k)

(33)
W.G. Morsi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 14081415 1413
The apparent power can be expressed in the timefrequency
domain using the wavelet packet based voltage, active and residual
current expressions as
S
2
=
2
j
1

m=0
(V
m
j
)
2
(I
m
aj
)
2
+
2
j
1

m=0
(V
m
j
)
2
(I
m
resj
)
2
+
2
j
1

m = 0, n = 0
m / = n
(V
m
j
)
2
(I
n
aj
)
2
+
2
j
1

m = 0, n = 0
m / = n
(V
m
j
)
2
(I
n
resj
)
2
(34)
S
2
=

2
j
1

m=0
V
m
j
I
m
aj

2
j
1

m=0
V
m
j
I
m
resj

2
j
1

m = 0, n = 0
m / = n
V
m
j
I
n
aj

2
+

2
j
1

m = 0, n = 0
m / = n
V
m
j
I
n
resj

(35)
S
2
= [P
2
] +[Q
2
] +[D
2
] (36)
where P is the total active power, Q is the reactive power and D is
the distortion power. Then the non-active power can dened as
N =
_
S
2
P
2
=
_
Q
2
+D
2
(37)
5. Numerical examples
This section includes two numerical examples in order to eval-
uate the proposed approach. The rst example considers the case
of stationary waveform while the second example considers non-
stationary waveform. In both examples the power components
were calculated using both approaches; the FFT based approach
according to the IEEE Standard 1459-2000 [1], and the WPT based
approach. The results were compared to the true values and the
errors associated with each method are calculated. Moreover, two
different mother wavelet are used in both examples (db10 and
db43) withthree decompositionlevels inorder tostudythe effect of
each on the measurement of the power components in both cases;
stationary and nonstationary waveforms.
5.1. Example 1 (stationary waveforms)
In this example the voltage and current waveforms are sta-
tionary; the voltage is sinusoidal (60Hz) with 50V amplitude and
zerophase angle. The current contains the fundamental component
with 25A amplitude, plus the fth harmonic with 10A amplitude.
The fundamental current component lags the fundamental volt-
age component by phase angle of 400. The sampling frequency is
1.92kHz (32 sample per 60Hz fundamental cycle) while the FFT
window is 10 cycles. Figs. 1 and 7 shows the time domain and the
FFT of the voltage and current waveforms, respectively.
TheFFTspectruminthis case, accuratelyidenties andmeasures
the different components included in the stationary waveform of
the voltage and the current. The true values and the calculated val-
ues of the power components based on both the IEEE Standard
1459-2000 (the FFT based approach) and the new approach (the
WPT based approach for db10 and db43) are listed in Table 1.
Fig. 7. Time domain waveform and its FFT spectrum for the current waveform (sta-
tionary case).
Theresults listedinTable1indicatethat theFFTis themost accu-
rate method for measuring power components when considering
stationary waveforms. Also it shows that the WPT based approach
gives better results when using db43 instead of using db10. In order
to compare the relative accuracy between the three methods, the
percentage error is plotted in Fig. 8 while considering the true val-
ues as the reference.
True values of the power components can be obtained using the
following time domain formula:
The active power:
P =
1
T
_
T
0
v(t) i(t)dt (38)
The apparent power is
S = VI, V =

1
T
_
T
0
v
2
(t)dt, I =

1
T
_
T
0
i
2
(t)dt (39)
The non-active power is
N =
_
S
2
P
2
(40)
It can be inferred from Fig. 8 that using db43 for measuring
power components may reduce the error by almost 50%. Also it
can be observed from Fig. 8 that the maximum error between the
Table 1
True values, FFT based values and wavelet packet transform values for stationary
waveform of example 1.
Power component True values FFT based
approach
WPT based approach
db10 db43
Active power 478.80 478.78 477.03 477.48
Reactive power N/A 401.74 408.48 405.44
Distortion power N/A 250.00 235.43 244.33
Non-active power 473.18 473.17 471.47 473.37
Apparent power 673.16 673.14 670.71 672.36
1414 W.G. Morsi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 14081415
Fig. 8. Percentage error plot of FFT, WPT (db10 and db43) based power components
(stationary case).
three methods is in the distortion power, while the minimumerror
is in the non-active and apparent power.
5.2. Example 2 (nonstationary waveforms)
In this example, the same sinusoidal source of Fig. 1 is used here
but the current consists of the fundamental component lagging the
voltage by 400 plus a fth harmonic component with time varying
amplitude. The fth harmonic amplitude is 10A starting fromtime
zero until time 0.08s and then the amplitude drops to 3A until the
end of the simulation time which is 0.167s. Fig. 9 shows the time
domain waveforms of the current and its FFT.
The time domain waveformshows a change in the waveshape at
0.08s which cannot be perceived in the FFT spectrum because the
FFT cannot provide any information concerning the time response.
Moreover, the FFT shows spectral leakage at the fth harmonic and
an error in estimating the magnitude of the fth harmonic.
Fig. 9. Time domain waveformand its FFT spectrumfor the current waveform(non-
stationary case).
Table 2
True values, FFT based values and wavelet packet transformvalues for nonstationary
waveform of example 2.
Power component True values FFT based
approach
WPT based approach
db10 db43
Active power 478.42 479.92 477.80 477.90
Reactive power N/A 401.04 405.84 403.94
Distortion power N/A 124.19 168.38 173.81
Non-active power 439.95 419.83 439.39 439.74
Apparent power 649.96 637.63 649.12 649.43
Table 2 lists the power components as calculated by the three
different methods. From Table 2 it can be inferred that the WPT
based approach gives better results than the FFT based approach
in this case. Also the value of the apparent power in the FFT based
approachis less thanthe true apparent power while that of the WPT
based approach (either db10 or db43) gives values that are close
to the true value. This indicates that the FFT produces enormous
errors in case of nonstationary waveform when measuring power
components.
The picture becomes clear through Fig. 10 which shows the
absolute percentage error between the three approaches. The g-
ure shows a large error in the distortion power component while
less error is introduced by the WPT approach when using db10. On
the other hand when using db43 the error is greatly reduced and
therefore best results can be obtained.
The error plot in this case may explain the large difference
between the true value of the apparent power and the FFT based
value. The difference comes from the large error introduced due
to the distortion power which stems fromthe spectral leakage that
occurred in the current fth harmonic due to the current waveform
nonstationarity.
This might conclude the unsuitability of the FFT based approach
that was adopted by the IEEE Standard 1459-2000 for measur-
ing the power components for nonstationary waveforms. The new
approach (WPT based approach) using db43 provide values that
are very close to the true values, therefore the new approach is
moresuitableandmoreaccuratethantheFFTbasedapproachwhen
dealingwithnonstationarywaveforms for power components mea-
surements.
Fig. 10. Percentage error plot of FFT, WPT (db10anddb43) basedpower components
(nonstationary case).
W.G. Morsi, M.E. El-Hawary / Electric Power Systems Research 79 (2009) 14081415 1415
6. Conclusions
Inthis paper, a newapproachbased onthe wavelet packet trans-
form has been introduced for measuring power components such
as active, reactive, distortion, non-active and apparent power. In
case of stationary electric power system waveform, the FFT based
approach that was adopted by the IEEE Standard 1459-2000 [1]
gives the best accurate results while the WPT based approach using
db43gives very small errors, whichis acceptable. Onthe other hand
the results of the numerical examples indicate that the use of db10
may introduce errors in the distortion power which leads to dif-
ference in the apparent power compared to the case when using
db43.
Innonstationary waveforms, the results showthat the FFT based
approach introduces very large error, which is due to the spectral
leakage associated with the FFT spectrum. The maximum error is
in the distortion power which greatly affects the apparent power
value, which leads to mismatch between the true value and the FFT
based apparent power values. On the other hand the WPT based
approach using db43 gives very accurate results that are very close
to the true values. The matching between the true value and the
WPT (with db43) value for the apparent power proves the accuracy
of the new approach and hence its suitability in this situation.
Fromthe computational burdenpoint of view, althoughthe WPT
has a computational complexity of n
2
which is greater than that
of the FFT, however, using the recursive algorithm mentioned in
Section 3.2, the computational complexity becomes only nlog
2
n as
the case in FFT.
Therefore, it can be concluded that the current denitions for
measuringpower components accordingtotheIEEEStandard1459-
2000 [1] are only suitable in case of stationary waveforms, however
in nonstationary waveforms they become unsuitable. The new
approach based on WPT and using db43 can be considered as a
comprehensive approach that may be suitable for all situations
including stationary and nonstationary waveforms.
Since the new denitions of these power components based on
WPT proves to be more accurate thanother methods, thenreducing
the error means saving money for the utility while metering the
consumers, help choosing the appropriate compensation devices
for improving the power factor, and applying suitable power qual-
ity mitigation techniques to enhance the power quality especially
under nonstationary operating conditions.
Acknowledgment
The authors are grateful to the Izaak Walton Killam Memorial
pre-doctoral scholarship for its nancial support of this work.
References
[1] IEEE Std. 1459-2000, Denitions for the measurement of Electric quantities
under sinusoidal, non-sinusoidal, balanced or unbalanced conditions, 2000.
[2] M.H.G. Bollen, I.Y.-H. Gu, Signal Processing of Power Quality Disturbances,
Wiley-IEEE, 2006.
[3] J. Driesen, T. Van Craenenbrock, R. Reekmans, D. Van Dommelen, Analyzing
time-varying power system harmonics using wavelet transform, in: Proc. 1996
Instrumentation and Measurement Technology Conference. IMTC-96, 1996, pp.
474479.
[4] S. Santoso, E.J. Powers, W.M. Grady, P. Hofmann, Power quality assessment via
wavelet transform analysis, IEEE Trans. Power Deliv. 11 (2) (1996) 824930.
[5] W.-K. Yoon, M.J. Devaney, Power measurement based on the wavelet transform,
IEEE Trans. Instrum. Meas. 47 (5) (1998) 12051210.
[6] L. Angrisani, P. Daponte, DApuzzo, A. Tesla, A measurement method based on
the wavelet transform for power quality analysis, IEEE Trans. Power Deliv. 13
(4) (1998) 990998.
[7] T. Zheng, E. Makram, A.A. Girgis, Power system transient and harmonic stud-
ies using wavelet transform, IEEE Trans. Power Deliv. 14 (4) (1999) 1461
1468.
[8] V.L. Pham, K.P. Wong, Wavelet-transform-based algorithm for harmonic anal-
ysis of power system waveforms, IEE Proc. Gen. Transm. Distrib. 146 (3) (1999)
249254.
[9] K.P. Wong, V.L. Pham, Analyzing power system waveforms using wavelet-
transform approach, in: Proceedings of the IEEE International Conference on
Advances in Power System Control, Operation and Management APSCOM, vol.
2, 2000, pp. 500504.
[10] K.H. Chul, R. Aggarwal, Wavelet transforms in power systems. I. General intro-
duction to the wavelet transforms, IEEE Power Eng. J. 14 (2) (2000) 8187.
[11] W.-K. Yoon, M.J. Devaney, Reactive power measurement using the wavelet
transform, IEEE Trans. Instrum. Meas. 49 (2) (2000) 246252.
[12] E.Y. Hamid, Z. Kawasaki, R. Mardiana, Wavelet packet transform for RMS and
power measurements, in: Proceedings of the IEEE Power Engineering Society
Summer Meeting, 2001, pp. 12431245.
[13] K.H. Chul, R. Aggarwal, Wavelet transforms in power systems. II. Examples of
application to actual power system transients, IEEE Power Eng. J. 15 (4) (2001)
193202.
[14] J. Driesen, R. Belmans, Active power lter control algorithm using wavelet-
based power denitions, in: Proceedings of the 10th International Conference
on Harmonics and Quality of Power, 2002, pp. 466471.
[15] E.Y. Hamid, R. Madriana, Z. Kawasaki, Method for RMS and power measure-
ments based on the wavelet packet transform, IEE Proc. Sci. Meas. Technol. 149
(2) (2002) 6066.
[16] L. Eren, M.J. Devaney, Calculationof power systemharmonics vis wavelet packet
decomposition in real time metering, in: Proceedings of the 19th International
Conference on Instrumentation and Measurement Technology. IMTC, vol. 2,
2002, pp. 16431647.
[17] J. Driesen, R. Belman, Wavelet-based power quantication approaches, IEEE
Trans. Instrum. Meas. 52 (4) (2003) 12321238.
[18] V. Shupletsov, I.V. Homchneko, Wavelet packet transformfor power quality fac-
tors measurement, in: Proceedings of the 5th International Serbian Workshop
on Electron Devices and Materials, 2004, pp. 145146.
[19] T. Croes, C. Gherasim, J. Van den Keybus, J. Ghijselen, J. Driesen, R. Belmans,
Power measurement using the wavelet transform of analytic signals, in: Pro-
ceedings of the 11th International Conference on Harmonics and Quality of
Power, 2004, pp. 338342.
[20] Z. Gaing, Wavelet-based neutral network for power disturbance recognition
and classication, IEEE Trans. Power Deliv. 19 (4) (2004) 15601568.
[21] T.X. Zhu, S.K. Tso, K.L. Lo, Wavelet-based fuzzy reasoning approach to power
quality disturbance recognition, IEEE Trans. Power Deliv. 19 (4) (2004)
19281935.
[22] S.A. Saleh, M.A. Rahman, Modelingandprotectionof a three-phase power trans-
former using wavelet packet transform, IEEE Trans. Power Deliv. 20 (2) (2005)
12731282.
[23] A.I. Megahed, A.M. Moussa, A.E. Bayoumy, Usage of wavelet transform in the
protection of series-compensated transmission lines, IEEE Trans. Power Deliv.
21 (3) (2006) 12131221.
[24] A.K. Sinha, A.K. Yadav, Designof HVtransmissionlineprotectionandcontrol sys-
tem using wavelet transform for microcontroller application, in: Proceedings
of the IEEE Power India Conference, 2006, p. 5.
[25] D. Das, N.K. Singh, A.K. Sinha, A comparison of Fourier transform and wavelet
transform methods for detection and classication of faults on transmission
lines, in: Proceedings of the IEEE Power India Conference, 2006, p. 7.
[26] T.X. Zhu, Effective value calculation in wavelet domain, IEEE Trans. Power Deliv.
19 (1) (2004) 400404.
[27] T.X. Zhu, Anewapproachto active power calculationusing wavelet coefcients,
IEEE Trans. Power Deliv. 21 (1) (2006) 435437.
[28] A. Sakar, S. Sengupta, A novel instantaneous power factor measurement based
on wavelet transform, Proc. IEEE Power India Conf. (2006), 6 pp.
[29] T. Tarasiuk, Hybrid wavelet-Fourier method for harmonics and harmonic sub-
groups measurementcase study, IEEE Trans. Power Deliv. 22 (1) (2007) 417.
[30] P. Lok-Fu, V. Dinavahi, G. Chang, M. Steurer, P. Ribeiro, Real-time digital time-
varying harmonic modeling and simulation techniques IEEE task force on
harmonic modeling and simulation, IEEE Trans. Power Deliv. 22 (2) (2007)
12181227.
[31] W.G. Morsi, M.E. El-Hawary, Reformulating power components denitions con-
tained in the IEEE Standard 1459-2000 using discrete wavelet transform, IEEE
Trans. Power Deliv. 22 (3) (2007) 19101916.
[32] W.G. Morsi, M.E. El-Hawary, Reformulating three-phase power components
denitions contained in the IEEE Standard 1459-2000 using discrete wavelet
transform, IEEE Trans. Power Deliv. 22 (3) (2007) 19171925.
[33] W.G. Morsi, M.E. El-Hawary, Timefrequency single-phase power components
measurements for harmonics and inter-harmonics distortion based on wavelet
packet transform-Part I: mathematical formulation, Can. J. Elect. Comp. Eng., in
press.
[34] W.G. Morsi, M.E. El-Hawary, Timefrequency single-phase power components
measurements for harmonics and inter-harmonics distortion based on wavelet
packet transform-Part II: mathematical formulation, Can. J. Elect. Comp. Eng.,
in press.
[35] S. Mallat, A Wavelet Tour of Signal Processing, Academic, New York, 1999.
[36] M. Misiti, Y. Misiti, G. Oppenheim, Wavelet Toolbox 4 for Use with Matlab, The
math works Inc., 2007.
[37] G. Strang, T. Nguyen, Wavelets and Filter Banks, Wellesley-Cambridge, Cam-
bridge, UK, 1996.