You are on page 1of 6

1

Most Suitable Mother Wavelet for Measurement


of Power System Harmonics Using DWT in
View of IEEE Standard 1459–2000
Subhash Kashyap and Asheesh K Singh

eliminates these shortcomings of FT. Yoon and Devaney [3-4]


Abstract-- Fourier transform (FT) based techniques have been have used the Discrete Wavelet Transform (DWT) to define
traditionally used for power system harmonics analysis. A recent the power and root mean square (RMS) values of the voltages
alternative approach of Wavelet Transform (WT) is applied for and currents. In [5-6] reformulation of power components
non-sinusoidal signal analysis which has been found very useful
definitions contained in IEEE standard 1459-2000 have been
in power quality studies. The problem of Root mean square
(RMS) value calculation is an important issue in harmonic done. The selection of appropriate wavelet family along with
analysis. This paper uses a statistical approach for RMS suitable mother wavelet is very important to achieve the
calculation. The proposed method simplifies wavelet based accurate results. In [7] an algorithm for this, using Wavelet
harmonic measurement process and increases measurement Packet Transform (WPT) has been adopted. In both DWT and
accuracy. Case studies from practical power system’s voltage and WPT the signal is divided into several frequency bands and
current signals are demonstrated. The measurement results using
computation is performed on these bands, rather than on each
the Discrete Wavelet Transform (DWT) method for various
wavelet families are compared and discussed with the results of frequency point as in FT. WPT suffers from more
IEEE Std. 1459-2000 which forms the basis to propose the most computational burdens in comparison to DWT. This is due to
suitable mother wavelet for power system harmonics analysis. the fact that in WPT each frequency band is decomposed into
approximations and details, while in DWT signal is first
Index Terms-- Discrete wavelet transform (DWT), Power decomposed into approximation and detail, and successive
quality, Power system harmonics measurement, Root mean decomposition of approximations is considered and details are
square (RMS), Wavelets.
left out. This paper uses DWT for analysis of signals and a
statistical approach is adopted for RMS value calculation.
I. INTRODUCTION
DWT provides time-frequency information in comparison to

C ONTINUOUS widespread applications of power


electronic-based loads as industrial and commercial office
equipments and home appliances, due to their remarkable
loss of time information in FT and involves less computational
efforts as compared to WPT. Finally, the results obtained with
application of this method to different actual voltage and
advantages in efficiency and controllability, has tremendously current waveforms are compared with the results obtained
increased the harmonic levels (harmonic distortion) in the using the IEEE Std. 1459-2000, to recommend the most
power distribution system. Harmonic distortion is recognized suitable wavelet for power system harmonic analysis.
as an important factor in the degradation of power quality,
which may shorten equipment life and cause interference with II. WAVELET TRANSFORM (WT)
communication and control devices [1]. For effective
Wavelet transform (WT) is a powerful signal processing
harmonic mitigation, it is important to quantify the harmonic
tool used in power system analysis. The applications of
voltages and currents by real-time measurements.
wavelet analysis in power systems include analysis and
The change in sinusoidal waveshape of supply, due to
detection of electromagnetic transients, power quality, data
harmonic distortion makes the traditional definitions of power
compression, and fault detection [8]. It transforms a time-
components inappropriate. This has led to the several attempts
domain waveform into time-frequency domain and estimates
towards redefinition of power components by many
the signal in the time and frequency domains simultaneously.
researchers. The IEEE standard 1459-2000 [2] using Fourier
It uses wavelets, “small waves,” which are functions with
transform (FT) gives the quite convincing definitions under
limited energy and zero average [9], as shown in (1)
these new conditions. The frequency domain approach (FT)
+∞
suffers from time information loss and more computational
efforts. The wavelet transform (WT) based approach ∫ ψ (t )dt = 0 (1)
−∞
In WT a specific wavelet, in general referred as the mother
Subhash Kashyap is with the Indian Oil Corporation Ltd., INDIA.
wavelet, is selected initially. Dilated (stretched) and translated
Asheesh K Singh is with the Electrical Engineering Department, Motilal
Nehru National Institute of Technology, Allahabad, INDIA (e-mail: (shifted in time) versions of the mother wavelet are then
asheesh_k_singh@yahoo.com; asheesh@mnnit.ac.in).

978-1-4244-1770-4/08/$25.00 ©2008 IEEE


2

generated. Dilation [9] is denoted by the scale parameter a level of decomposition and thereby computational attempts. It
while time translation is adjusted through b , as given by (2) is depicted in form of decomposition tree of DWT in Fig. 2.
⎛ t −b ⎞
1
ψ a ,b (t ) = ψ⎜⎟ (2)
a ⎝ a ⎠
AA2

where a is a positive real number and b is a real number. A1

The wavelet transform of a signal f(t) at a scale a and time AD2

translation b is the dot product of the signal f(t) and the Signal

specific mother wavelet, ψ a ,b (t ) . It is computed by circular DA2

convolution [9] of the signal with the wavelet function given D1


as
+∞
1 *⎛t −b ⎞ (3)
DD2

W{f ( a, b )} = f ,ψ a ,b = ∫ f (t ). ψ ⎜ ⎟ dt
−∞ a ⎝ a ⎠ Fig. 1. Decomposition Tree for Wavelet Packet Transform
With wavelet functions, only information of scale a <1
corresponding to high frequencies is obtained. In order to A5

obtain the low-frequency information necessary for full A4

representation of the original signal f(t), it is necessary to


A3 D5
determine the wavelet coefficients for scale a >1. This is
achieved by introducing a scaling function φ (t ) which is an A2 D4

aggregation of the mother wavelets ψ (t ) at scales greater than A1 D3

1. The scaling function can also be scaled and translated as the Signal D2

wavelet function, as shown in (4) D1


1 ⎛ t −b ⎞
φ a ,b (t ) = φ⎜ ⎟ (4) Fig. 2. Decomposition Tree for Discrete Wavelet Transform
a ⎝ a ⎠
With scaling function, the low-frequency approximation of III. DWT AND ITS IMPLEMENTATION
f(t) at scale a is the dot product of the signal and the DWT is a process of transforming a waveform from time-
particular scaling function [9], and can be computed by domain to time-frequency domain, with the help of suitable
circular convolution using (5) mother wavelet (wavelet family). Mother wavelet is chosen on
+∞ the basis of energy criteria [5-6]. This energy criterion states
φ* ⎛⎜
t −b ⎞
{ f (a,b)} =
1
L f ,φa,b = ∫ f (t). a ⎝ a
⎟ dt

(5) the percentage energy deviation at each level from the original
−∞ signal and is given by (7)
Implementation of these two transforms (3) and (5) can be E − Ean
done smoothly in continuous wavelet transform (CWT) or %δEn = a × 100 (7)
Ea
discretely in discrete wavelet transform (DWT).
where %δEn denotes percentage energy deviation at nth
In time-frequency domain the information regarding both
time and frequency are preserved making it better than FT, i.e. decomposition level from signal’s overall energy Ea , Ean is
a frequency domain approach. Also, the problem of fixed energy content of the signal at nth level of decomposition.
window size as in short time fourier transform (STFT) is Both high frequency and low frequency components are
solved in WT. The WT is applied on a signal and wavelet present in the signal.
coefficients are calculated. This is termed as CWT: Based on this energy criterion five mother wavelets have
∞ been selected for comparison namely-Discrete Meyer (dmey),
C(scale, position) = ∫ f (t )Ψ(scale, position, t )dt (6)
Daubechies 10 (db10) & Daubechies 7 (db7), Symlet 8 (sym8)
−∞
and Coiflet 5 (coif5). These wavelets are presented in Fig. 3.
where C corresponds to coefficients calculated at every scale
The selection of decomposition level is another important
and position. The coefficients also show how close the part of
aspect of DWT. The selection of suitable level becomes easier
the signal under consideration to the wavelet function ψ. But
by applying energy criterion. In addition to that, the signal is
CWT generates a huge amount of data in the form of wavelet
decomposed up to the level till fundamental component is
coefficients with respect to change in scale and position. This
separated from the harmonic components. The decomposition
leads to large computational burden. To overcome this
level also depends upon harmonic order. On continuous
limitation Wavelet Packet Transform (WPT) is used. In WPT
analysis of signals considered for this study, it is found that
the wavelet coefficients are calculated at uniform frequency
appropriate decomposition level is seven.
bands. The Decomposition Tree for WPT is shown in Fig.1.
Decomposition tree looks like Fig 2. At each level one
In view of the fact that WPT generates large number of
approximation & one detail is generated. The calculation of
nodes it increases the computational burden. In DWT only
RMS values at a particular level for either approximation or
approximations are further decomposed thus reducing the
detail is done for both voltage and current signals.
3

(a) (b) (c) (d) (e)


Fig. 3. Wavelets used for analysis. (a) db10, (b) dmey, (c) sym8, (d) coif5, and (e) db7

In this work, the statistical approach is adopted to calculate voltage and current are calculated for both examples. Table III
the root mean square (RMS) values of voltage and current shows DWT based values obtained for these five wavelets up
signals at each level for every approximation and detail using to level seven for Example 1. The percentage error between
(8): IEEE based values and DWT based values, calculated using
2
X rms = X 2 + σ X2 (8) step 6, is shown in Table IV, for example 1. Similarly, for
example 2 DWT based values with all five wavelets are given
where X rms is the rms value, X is the mean and σx denotes in Table V and Table VI presents the percentage error with
standard deviation. respect to IEEE values.
The steps involved in this study can be summarized as:
A. Example 1
Step 1: Consider the mother wavelet for analysis of the signals
under consideration. The signal taken is actual tested waveform obtained from a
Step 2: Apply energy criterion to the signal for the chosen combination of single-phase non-linear loads described
wavelet and obtain the energy deviation. earlier. The voltage and current waveforms captured using
Step 3: Similarly other wavelets are taken and energy power quality analyzer are shown in Fig. 4 and their
deviation for same signal is obtained. corresponding screens for numerical values of voltage and
Step 4: Now comparison of energy deviations is done for all current signals are depicted in Fig. 5
the wavelets, and wavelets with minimum deviation (less than
ten percent as found through continuous analysis) are selected
for the signal analysis.
Step 5: Using statistical approach, we get the mean values and
standard deviations for approximations and detail values of
current and voltage, from where RMS values have been
calculated using (8).
Step 6: Finally, percentage difference between IEEE standard
1459-2000 based values and DWT based values are
calculated.
The above steps are applied to both the case studies
considered in the given work.
Fig. 4. Actual tested waveforms for voltage and current signal for example 1

IV. CASE STUDIES


The approach adopted in this paper has been applied for
two different actual tested waveforms obtained by power
quality analyzer. The first example considers the case of
voltage and current signals obtained at a residential apartment
for a variety of new-generation non-linear loads including
Television, VCD player and Fridge one unit of each. The
second example signal is captured at a small computer (a) (b)
simulation laboratory for five personal computers (PCs)
Fig. 5. Numerical values of (a) voltage signal (b) current signal, as captured
processing specific programs simultaneously. by Power Quality analyzer, for Example 1
For both sets of voltage and current signals the RMS values
for voltages and currents are calculated using statistical For wavelet analysis, both the voltage and current signals
approach and from there THD values are obtained. Using are considered with harmonics up to 15th order. The signals
steps 1-4 energy deviations for both the signals are calculated are given as (9) and (10). The sampling frequency is taken as
for various wavelets and comparison has been done. Based on 25.6 kHz (512*50Hz). As, in this work harmonics up to 15th
minimum energy deviation, five specific wavelets as shown in order are considered, the suitable decomposition level is found
Fig. 3, have been selected. Energy at each level and energy to be seven. Energy at each level is obtained using energy
deviations are shown in Table I and Table II only for Example criterion and shown in Table I, whereas Table II contains the
1 under consideration. Using step 5 DWT based values for energy deviation at each level, calculated using (7).
4

v(t ) = 338.37sin ωot + 1.556sin(3ωot + 80o ) + 5.104sin(5ωot + 40o ) TABLE I


ENERGY CALCULATED AT EACH LEVEL FOR EXAMPLE 1
+ 4.232sin(7ωot + 42o ) + 4.108sin(9ωot + 178o ) + 0.996sin(11ωot + 91o ) Wavelets→ dmey db10 db7 sym8 coif5
o
+ 0.249sin(13ωot − 101 ) + 0.809sin(15ωot − 117 ) o
(9) Energy at each (%) (%) (%) (%) (%)
level↓

and Voltage, V
o o
Ea1 100.00 100.00 100.00 100.00 100.00
i (t ) = 1.702sin ωo t + 0.357 sin(3ωo t − 69 ) + 0.326sin(5ωo t − 150 ) Ea2 100.00 100.00 100.00 100.00 100.00
+ 0.191sin(7ωo t + 119o ) + 0.157 sin(9ωot + 25o ) + 0.10sin(11ωot − 74o ) Ea3 100.00 100.00 100.00 100.00 100.00
Ea4 100.00 100.00 100.00 100.00 100.00
+ 0.047 sin(13ωot − 170o ) + 0.019sin(15ωo t − 85o ) (10) Ea5 99.98 99.97 99.98 99.97 99.97
Ea6 99.33 99.94 99.94 99.94 99.93
Ea7 99.27 99.89 99.86 99.83 99.90
ω o = 2π f o fo = 50 H z
,
Table III shows the DWT based results of harmonic Current, I
Ea1 100.00 100.00 100.00 100.00 100.00
analysis using five different wavelets together with the Ea2 100.00 100.00 100.00 100.00 100.00
corresponding IEEE Std. 1459-2000 based values of harmonic Ea3 100.00 100.00 100.00 100.00 100.00
components for the signals (9) and (10). Ea4 99.99 99.97 99.97 99.96 99.97
Ea5 98.90 98.36 98.25 98.16 98.77
There is a little difference in THD values shown in screen Ea6 95.33 93.71 93.08 92.81 93.95
and calculated DWT based values in Table III. This is because Ea7 92.18 90.08 89.97 90.04 90.57
the screen value is displayed for harmonic orders up to 51st
while in Table III the calculation of THD values is limited up TABLE II
ENERGY DEVIATION AT EACH LEVEL FOR EXAMPLE 1
to 15th harmonic order as considered in Example 1. The Wavelets→ dmey db10 db7 sym8 coif5
calculated percentage error between the IEEE Standard 1459- Energy deviation (%) (%) (%) (%) (%)
2000 based values and DWT based values for all the five at each level↓
wavelets considered here are presented in Table IV. Voltage, V
It is clearly visible from Table IV that among twenty δEa1=Ea-Ea1 0.00 0.00 0.00 0.00 0.00
parameters, for nine parameters db7 gives the least δEa2=Ea-Ea2 0.00 0.00 0.00 0.00 0.00
δEa3=Ea-Ea3 0.00 0.00 0.00 0.00 0.00
percentage error from IEEE based values, and for remaining δEa4=Ea-Ea4 0.00 0.00 0.00 0.00 0.00
parameters (for which db7 does not gives least percentage δEa5=Ea-Ea5 0.02 0.03 0.02 0.03 0.03
error) it has the difference in the range of 0.001% to 1.088% δEa6=Ea-Ea6 0.67 0.06 0.06 0.06 0.07
δEa7=Ea-Ea7 0.73 0.11 0.14 0.17 0.10
from the corresponding least percentage error values for the
wavelets other than db7. The minimum percentage errors are Current, I
presented in italics in Table IV. δEa1=Ea-Ea1 0.00 0.00 0.00 0.00 0.00
δEa2=Ea-Ea2 0.00 0.00 0.00 0.00 0.00
δEa3=Ea-Ea3 0.00 0.00 0.00 0.00 0.00
δEa4=Ea-Ea4 0.01 0.03 0.03 0..04 0.03
δEa5=Ea-Ea5 1.10 1.64 1.75 1.84 1.23
δEa6=Ea-Ea6 4.67 6.29 6.92 7.19 6.05
δEa7=Ea-Ea7 7.82 9.92 10.03 9.96 9.43

TABLE III
IEEE STANDARD 1459 – 2000 BASED VALUES AND DWT BASED VALUES USING FIVE DIFFERENT WAVELETS FOR EXAMPLE 1
IEEE Values DWT values dmey db10 db7 sym8 coif5

Vf 239.27 Va7 239.4025 239.10 239.00 239.300 239.3047


V3 1.1002 Vd7 2.8140 4.0500 4.3615 5.7280 3.0650
V5 2.6090 Vd6 4.9260 4.4880 4.3020 4.4130 4.8390
V7 2.9925 Vd5 3.0580 3.4430 3.6780 3.5710 2.9950
V9 2.9045 Vd4 0.3248 0.4524 0.5191 0.4403 0.4377
V11 0.7041 Vd3 0.0607 0.0499 0.0974 0.0938 0.0585
V13 0.1760 Vd2 0.0203 0.0155 0.0214 0.0231 0.0288
V15 0.5731 Vd1 0.0057 0.0047 0.0083 0.0068 0.0068
Vh 5.1248 Vd 6.4533 6.9718 7.1650 8.0771 6.4789
THDv 0.0214 0.0270 0.0292 0.0300 0.0338 0.0271

If 1.2038 Ia7 1.2010 1.2010 1.2040 1.2000 1.2110


I3 0.2527 Id7 0.2347 0.2413 0.2502 0.2369 0.2389
I5 0.2306 Id6 0.2414 0.2401 0.2260 0.2500 0.2456
I7 0.1352 Id5 0.1549 0.1532 0.1407 0.1501 0.1382
I9 0.1109 Id4 0.1089 0.1007 0.1101 0.1009 0.0932
I11 0.0709 Id3 0.0669 0.0801 0.0759 0.0799 0.0756
I13 0.0333 Id2 0.0292 0.0387 0.0400 0.0390 0.0341
I15 0.0133 Id1 0.0002 0.0000 0.0001 0.0001 0.0001
Ih 0.3923 Id 0.3934 0.3967 0.3908 0.3990 0.3899
THDi 0.3098 0.3163 0.3162 0.3103 0.3206 0.3139
5

TABLE IV
PERCENTAGE ERROR CALCULATION FOR EXAMPLE 1
v(t ) = 358.602sin(ωo t ) + 7.406sin(3ωo t − 90o ) + 0.747 sin(5ωo t + 179o )
Parameters %dmey %db10 %db7 %sym8 %coif 5
+ 6.473sin(7ωo t + 114o ) + 13.001sin(9ωo t − 173o )
RMS 0.0554 -0.0710 -0.1128 0.0125 0.0145 + 2.365sin(11ωo t − 124o ) + 1.99 sin(13ωo t − 28o )
Voltage 0.7163 1.2328 1.3630 1.9341 0.8212
0.9684 0.7853 0.7076 0.7540 0.9320 + 3.174sin(15ωo t + 177o ) (11)
0.0274 0.1883 0.2865 0.2418 0.0010 and
-1.0782 -1.0248 -0.9969 -1.0299 -1.0310
-0.2689 -0.2734 -0.2536 -0.2551 -0.2698 i (t ) = 5.354 sin(ωo t ) + 4.116 sin(3ωo t − 173o ) + 2.687 sin(5ωo t − 1o )
-0.0651 -0.0671 -0.0646 -0.0639 -0.0615
-0.2371 -0.2376 -0.2361 -0.2367 -0.2367 + 1.335sin(7ωo t − 178o ) + 0.097 sin(9ωo t − 92o )
0.5552 0.7719 0.8527 1.2339 0.5659
THDv 0.0055 0.0077 0.0086 0.0123 0.0057
+ 0.536 sin(11ωo t − 9o ) + 0.542 sin(13ωo t + 178o )

+ 0.326 sin(15ωo t + 5o ) (12)


RMS -0.2326 -0.2326 0.0166 -0.3157 0.5981
Current -1.4953 -0.9470 -0.2077 -1.3125 -1.1464
0.8972 0.7892 -0.3821 1.6116 1.2461 ωo = 2π fo , fo = 50 Hz
1.6365 1.4953 0.4569 1.2377 0.2492
-0.1661 -0.8473 -0.0665 -0.8307 -1.4703
-0.3323 0.7642 0.4154 0.7476 0.3904
-0.3406 0.4486 0.5566 0.4735 0.0665
-1.0882 -1.1048 -1.0965 -1.0965 -1.0965
0.0928 0.3657 -0.1306 0.5577 -0.1983
THDi 0.0065 0.0064 0.0005 0.0108 0.0041

B. Example 2
In this case, the distorted voltage and current signals are
captured for five PC systems and are represented by (11)
and (12). For wavelet analysis, the sampling frequency is
taken as 25.6 kHz.
Fig. 6. Actual tested waveforms for voltage and current signal for
Again, the steps 1-6 are followed to obtain the DWT example 2
based values using five different wavelets which are
presented in Table V, along with IEEE Standard 1459-2000
based values. Table VI shows the percentage error of DWT
based values, calculated with respect to IEEE Standard
1459-2000 based values for example 2. The voltage and
current waveforms obtained using power quality analyzer
are shown in Fig. 6 and their corresponding screens for
numerical values of voltage and current signals are shown
in Fig. 7. (a) (b)
The signal is synthesized with harmonics up to 15th Fig. 7. Numerical values of (a) voltage signal (b) current signal, as
order, and the considered decomposition level is seven. captured by Power Quality analyzer, for Example 2
TABLE V
IEEE STANDARD 1459 – 2000 BASED VALUES AND DWT BASED VALUES USING FIVE DIFFERENT WAVELETS FOR EXAMPLE 2
IEEE Values DWT Values dmey db10 db7 sym8 coif5

Vf 253.5700 Va7 253.50 253.40 253.60 253.70 253.45


V3 5.2370 Vd7 6.2000 7.9130 6.0000 6.2300 7.2750
V5 0.5281 Vd6 4.7110 4.9220 3.0000 5.1980 5.0231
V7 4.5770 Vd5 7.1200 8.1325 7.0112 8.1043 8.8010
V9 9.1980 Vd4 6.5321 6.6210 6.6220 5.8421 6.0000
V11 1.6723 Vd3 0.1180 0.0671 0.1433 1.0690 0.0876
V13 1.4082 Vd2 0.0270 0.0194 0.0301 0.0266 0.0350
V15 2.2444 Vd1 0.0078 0.0057 0.0013 0.0089 0.0086
Vh 11.9614 Vd 12.4101 13.7471 11.7486 12.9145 13.8428
THDv 0.0472 0.0489 0.0542 0.0463 0.0508 0.0545

If 3.7860 Ia7 3.7740 3.7610 3.7890 3.7390 3.762


I3 2.9108 Id7 2.9160 2.9510 2.9100 2.9600 2.8843
I5 1.8999 Id6 1.8992 1.9101 1.9001 1.9200 1.9312
I7 0.9444 Id5 0.9461 0.9542 0.9441 0.9521 0.9311
I9 0.0687 Id4 0.0697 0.0601 0.0686 0.0701 0.0721
I11 0.3791 Id3 0.3788 0.3821 0.3786 0.3613 0.3823
I13 0.3835 Id2 0.3901 0.3657 0.4011 0.3732 0.3978
I15 0.2306 Id1 0.2401 0.2457 0.231 0.2271 0.2583
Ih 3.6501 Id 3.6550 3.6893 3.6513 3.6990 3.6458
THDi 0.6491 0.6957 0.7002 0.6939 0.7033 0.6959
6

TABLE VI having harmonics up to 15th order are measured while only


PERCENTAGE ERROR CALCULATION FOR EXAMPLE 2
simulated signals with 3rd and 5th harmonics were considered
Parameters %dmey %db10 %db7 %sym8 %coif 5 in [5]. A statistical approach has been adopted for RMS value
calculation, which reduces the computational efforts
RMS -0.0276 -0.0670 0.0118 0.0513 -0.0473 significantly. It is found that the Daubechies 7 (db7) is the
Voltage 0.3797 1.0553 0.3009 0.3916 0.8037
most suitable mother wavelet for power system harmonics
1.6486 1.7328 0.9773 1.8417 1.7750
1.0029 1.4021 0.9600 1.3910 1.6658 measurement as it gives reasonably satisfactory results with
-1.0513 -1.0163 -1.0159 -1.3235 -1.2611 minimum percentage error with respect to IEEE Standard
-0.6128 -0.6330 -0.6030 -0.6263 -0.6250 1459-2000 as compared to other wavelets.
-0.5447 -0.5477 -0.5435 -0.5449 -0.5415
-0.8820 -0.8829 -0.8846 -0.8816 -0.8817
0.1769 0.7042 -0.0839 0.3759 0.7420 VI. REFERENCES
THDv 0.0017 0.0070 -0.0009 0.0036 0.0073
[1] V. E. Wagner, J. C. Balda, D. C. Griffith, A. McEachern, T. M. Barnes,
D. P. Hartmann, D. J. Phileggi, A. E. Emannuel, W. F. Horton, W. E.
RMS -0.3170 -0.6603 0.0792 -1.2414 -0.6339
Reid, R. J. Ferraro, and W. T. Jewell, “Effects of harmonics on
Current 0.1373 1.0618 -0.0211 1.2990 -0.6999
equipment,” IEEE Trans. Power Delivery, vol. 8, no. 2, pp. 672-680,
-0.0185 0.2694 0.0053 0.5309 0.8267
1993.
0.0449 0.2588 -0.0079 0.2034 -0.3513
[2] Definitions for the Measurement of Electric Quantities under Sinusoidal,
0.0264 -0.2271 -0.0026 0.0370 0.0898
Non-Sinusoidal, Balanced, or Unbalanced Conditions, IEEE Standard
-0.0079 0.0792 -0.0132 -0.4701 0.0845
1459–2000, 2000.
0.1743 -0.4701 0.4649 -0.2720 0.3777
[3] W.-K. Yoon and M. J. Devaney, “Power measurement based on the
0.2509 0.3988 0.0106 -0.0924 0.7316
wavelet transform,” IEEE Trans. Instrum. Meas., vol. 47, no. 5, pp.
0.1294 1.0354 0.0316 1.2916 -0.1136
1205–1210, Oct. 1998.
THDi 0.0016 0.0061 -0.0002 0.0092 0.0018
[4] W.-K. Yoon and M. J. Devaney, “Reactive power measurement using
the wavelet transform,” IEEE Trans. Instrum. Meas., vol. 49, no. 2, pp.
From Table VI it is found that out of eighteen parameters, 246–252, Apr. 2000.
db7 is giving least percentage error for fifteen parameters. For [5] Walid G. Morsi and M. E. El-Hawary, “Reformulating Power
Components Definitions Contained in the IEEE Standard 1459–2000
remaining parameters divergence from least percentage error Using Discrete Wavelet Transform,” IEEE Trans. Power Delivery, vol.
is in the range of 0.01% to 0.88%. The minimum percentage 22, no. 3, pp. 1910-1916, Jul. 2007.
errors are presented in italics in Table VI. [6] Walid G.Morsi and M. E. El-Hawary, “Reformulating Three-Phase
Power Components Definitions Contained in the IEEE Standard 1459–
Fig. 8 shows the comparison of number of parameters 2000 Using Discrete Wavelet Transform,” IEEE Trans. Power Delivery,
having least percentage error for different wavelets, which vol. 22, no. 3, pp. 1917-1924, Jul. 2007
precisely indicates that the use of db7 as mother wavelet for [7] Walid G. Morsi and M. E. El-Hawary, ”Suitable Mother Wavelet for
harmonic analysis gives least percentage error for most Harmonics and Inter-Harmonics Measurements Using Wavelet Packet
Transform” in Proc. Canadian Conference of Electrical and Computer
number of parameters. The comparison of results for both the Engineering, 2007, CCECE 2007, pp. 748-752, Apr. 2007
examples concludes that [8] H. Monsefa and S. Lotfifard, “Internal Fault Current Identification based
• Daubechies 7 (db7) is most appropriate wavelet for power on Wavelet Transform in Power Transformers,” Electric Power Systems
Research, Volume 77, Issue 12, October 2007, Pages 1637-1645
system harmonic analysis. [9] S. Chen and H. Y. Zhu, “Wavelet Transform for Processing Power
• Since values obtained from DWT based approach do not Quality Disturbances,” EURASIP Journal on Advances in Signal
differ significantly from IEEE Standard 1459-2000. It Processing, vol. 2007, Article ID 47695, 20 pages, Feb. 2007
proves the suitability of DWT based technique for
quantization of power system harmonic analysis.
No. of parameters with least
%age error

(a) (b)
Fig. 8. Comparison of number of parameters having least percentage error for
different mother wavelets for measurement of power system harmonic
components for (a) Example 1, and (b) Example 2

V. CONCLUSION
This paper presents the use of DWT based methodology for
power system harmonics analysis and proposes the most
suitable mother wavelet for this purpose. It has been applied
for two different actual tested waveforms as shown in example
1 and example 2. In the presented approach real time signals

You might also like