1
Novel Method for Tracking Time-Varying
Power Harmonic Distortions without Frequency
Spillover
C. Duque, Member, IEEE, P.M. Silveira, Member, IEEE, T. Baldwin, Senior Member, IEEE, P. F.
Ribeiro, Fellow, IEEE
Abstract-- Although it is well known that Fourier analysis is in
reality only accurately applicable to steady state waveforms, it is
a widely used tool to study and monitor time-varying signals,
which are commonplace in electrical power systems. The
disadvantages of Fourier analysis, such as frequency spillover or
problems due to sampling (data window) truncation can often be
minimized by various windowing techniques, but they
nevertheless exist. This paper demonstrates that it is possible to
track and visualize amplitude and time-varying power systems
harmonics, without frequency spillover caused by time-frequency
techniques. This new tool allows for a clear visualization of time-
varying harmonics which can lead to better ways to track
harmonic distortion and understand time-dependent power
quality parameters. It also has the potential to assist with control
and protection applications.
Index Terms—Harmonic distortion, harmonic decomposition,
filter banks, time-varying power harmonic distortion, multi-
resolution analysis, wavelet transform.
I. INTRODUCTION
W hile estimation technique is concerned with the process
used to extract useful information from the signal, such
as amplitude, phase and frequency; signal
decomposition is concerned with the way that the original
signal can be split in other components, such as harmonic,
inter-harmonics, sub-harmonics, etc..
Harmonic decomposition of a power system signal (voltage
or current) is an important issue in power quality analysis.
There are at least two reasons to focus on harmonic
decomposition instead of harmonic estimation: (a) if
separation of the individual harmonic component from the
input signal can be achieved, the estimation problem becomes
easier; (b) the decomposition is carried out in the
time-domain, such that the time-varying behavior of each
harmonic component can be observed.
This work was supported in part by CNPq and CAPES (Brazilian research
agencies). Carlos A. Duque, is with Electrical Engineering Dept., UFJF, Juiz de
Fora MG, Brazil, and CAPS / FSU (Center for Advanced Power Systems/
Florida State University) e-mail: carlos.duque@ufjf.edu.br, Paulo M Silveira is
with Itajubá Federal University, Itajubá, MG, Brazil, and CAPS/FSU e-mail:
pmsilveira@unifei.edu.br, Paulo F. Ribeiro is with CAPS/FSU and Calvin
College, e-mail: pfribeiro@ieee.org, Thomas L. Baldwin is with CAPS / FSU, e-
mail: tom.baldwin@ieee.org.
©2008 IEEE.
Some existing techniques can be used to separate
frequency components. For example, STFT (Short Time
Fourier Transform) and Wavelets Transforms [1], are two
well known decomposition techniques. Both can be seen as a
particular case of filter bank theory [2]. The STFT
coefficients filters are complex numbers which generates a
complex output signal whose magnitude corresponds to the
amplitude of the harmonic component being estimated [2].
The main disadvantage of this method is to set up an efficient
band-pass filter with lower frequency spillover. On the other
hand, although wavelet transform utilizes filters with real
coefficients, the common wavelet mothers do not have good
magnitude response in order to prevent frequency spillover.
Besides, the traditional binary tree structure is not able to
divide the spectrum conveniently for harmonic
decomposition.
Adaptive notch filter and PLL (Phase-Locked Loop) [3]-
[4] have been used for extracting time-varying harmonics
components. However, these methods work well only if a few
harmonics components are present at the input signal. In the
other case the energy of adjacent harmonics spills over each
other and the decomposed signal becomes contaminated. In
[5] the authors presented a technique based on multistage
implementation of narrow low-pass digital filters to extract
stationary harmonic components. Thus, no digital analytical
technique has been proposed or used to track time-varying
power systems harmonics without frequency spillover.
Consequently, the method proposed in this paper adequately
utilizes filters banks to avoid the frequency mixture,
particularly associated with time-varying signals.
Attempts to visualize time-varying harmonics using
Wavelet Transform have been proposed in [6] and [7].
However, the structures were not able to decouple the
frequencies completely.
The methodology proposed in this paper is constructed to
separate the odd and the even harmonic components, until the
15th. It uses selected digital filters and down-sampling to
obtain the equivalent band-pass filters centered at each
harmonic. After the signal is decomposed by the analysis
bank, each harmonic is reconstructed using a non-
conventional synthesis bank structure. This structure is
composed of filters and up-sampling that reconstructs each
 2

harmonic to its original sampling rate. multirate approach. The filters H0(z) and H1(z) are
The immediate use for this method is the monitoring of orthogonal filters [6], with the first one as a low-pass filter
time-varying individual power systems harmonics. Future use and the second one as a high-pass filter.
may include control and protection applications, as well as
inter-harmonic measurements.
The paper is divided into method description, odd/even
harmonic extraction and simulation results.

II. METHOD DESCRIPTION

Multirate systems employ a bank of filters with either a
common input or summed output. The first structure is known
as analysis filter bank [8] as it divides the input signal in
different sub-bands in order to facilitate the analysis or the
processing of the signal. The second structure is known as
synthesis filter bank and is used if the signal needs to be
reconstructed. Together with the filters the multirate systems Fig. 2 – Analysis bank filter to decomposed the input signal in its harmonics
must include the sampling rate alteration operator (up and components.
down-sampling). Figure 1 shows two basic structures used in
a multirate system, where the up and down-sampling factor, .
M and L, equal 2. Figure 1-a shows a decimator structure
composed by a filter following by the down sampler and Fig.
1-b the interpolator structure composed by a up-sampler
followed by a filter. The decimator structure is responsible to
reduce the sampling rate while the interpolator structure to
increase it. The filters H(z) and F(z) are typically band-pass
filters.

Fig.3- Multirate equivalent structure to the filter bank in Fig.2

Figure 4 shows the amplitude response for the analysis bank.

This figure was obtained using sampling rate equal to 256
samples by cycle and FIR filters of 69th order. These filters
Fig. 1- Basic structures used in multirate filter bank and its equivalent
representation for L=M=2. (a) Decimator; (b) Interpolator correspond to so-called orthogonal filter banks also known as
power-symmetric filter banks [8].
The direct way to build an analysis filter bank, in order to
divide the input signal in its odd harmonic component, is Analysis Filter Bank Frequency Response
0
represented in Fig. 2. In this structure the filter H k (z ) is a
H1 H3 H5 H7 H9 H11 H13 H15
band-pass filter centered in the kth harmonic and must be -5

projected to have 3dB bandwidth lower than 2f0 , where f0 is -10

the fundamental frequency. If only odd harmonics are
Magnitude (dB)

-15
supposed to be present in the input signal, the 3dB bandwidth
can be relaxed to be lower than 5f0. Note that Fig. 2 is not a -20

multirate system, because the structure does not include -25

sampling rate alternation, which means that there is only one
-30
sampling rate in the whole system.
The practical problem concerning the structure shown in -35
Fig. 2 is the difficulty to design each individual band-pass
-40
filter. This problem becomes more challenging when a high 0 2 4 6 8 10 12 14 16 18
frequency (harmonic number)
sampling rate must be used to handle the signal and the Fig. 4- Amplitude response in dB (frequency axis is labeled with the harmonic
consequent abrupt transition band. number)
In this situation the best way to construct an equivalent
filter bank is to utilize the multirate technique. Figure 3 shows The main difference between the structure shown in Fig. 2,
how an equivalent structure to Fig.2 can be obtained using the and the other one obtained using the multirate technique, is

the decimator at the output of the bank, as shown in Fig. 5. In
fact, Fig.5 is equivalent to Fig. 3. It was obtained using the
multirate nobles identities [2] and moving the downsampler
factor inside Fig.2 to the right side. The decimated signal at
the output of each filter has a sampling rate 64 times lower
then the input signal. To reconstruct each harmonic into its
original sampling rate it is necessary to use the synthesis filter
bank structure.
H
xd (n)
1
1(z)
64
xd (n)
3
x(n)
H 3(z)
64
xd (n)
k Fig. 7 – Whole structure to harmonic extraction
H k (z) 64
Fig. 5- Equivalent multirate analysis filter bank
A. The synthesis filter banks
The synthesis filter bank used here is a different
implementation of the conventional one. As the interest is to
reconstruct each harmonic instead of the original signal, the
filter bank must be divided in order to obtain the
corresponding harmonic. Figure 6 shows the filter bank
utilized to reconstruct the harmonic components.
Fig. 6- Modified synthesis bank
It is important to remark that the frequency response of the
synthesis filter bank is similar to Fig. 4.
The analysis of Fig. 4 reveals that the filter bank in this
configuration is unable to filter even harmonics appropriately,
what means that these components will appear with the
adjacent odd harmonics. To overcome this problem an
additional filter stage is included. This filter is composed by a
second order Infinite Impulse Response (IIR) notch filter.
B. Extracting even harmonics
To extract the even harmonics the same bank can be used
together with a preprocessing of the input signal. Hilbert
transform is then used to implement a technique known as
Single Side Band (SSB) modulation [8]. The SSB modulation
moves to the right all frequencies in the input signal by f0 Hz.
In this way, even harmonics are changed to odd harmonics
3
and vice-versa.
Figure 7 shows the whole system for extracting odd and
even harmonics.
III. SIMULATION RESULTS
This section presents two examples: the first is a synthetic
signal, which has been generated in Matlab using a
mathematical model, and the second is a signal obtained from
the Electromagnetic Transient Program including DC systems
(EMTDC) with its graphical interface Power Systems
Computer Aided Design (PSCAD™). This program can
simulate any kind of power system with high fidelity and the
resulting signals of interest are very close to physical reality.
A. Synthetic Signal
The synthetic signal utilized can be represented by:
N
x(t ) = ∑ Ah sin(h ω0t). f (t ) + g (t ) (1)
h =1
Where h is the order (1 up to 15th) and A is the magnitude
of the component, ω0 is the fundamental frequency, and
finally, f(t) and g(t) are exponential functions (crescent, de-
crescent or alternated one) or simply a constant value ∈ ℜ.
Besides, x(t) is partitioned in four different segments in such
way that the generated signal is a distorted one with some
harmonics in steady-state and others varying in time,
including abrupt and modulated change of magnitude and
phase, as well as a DC component. Figure 8 illustrates the
synthetic signal.
The structure shown in Fig. 7 has been used to decompose
the signal into sixteen different harmonic orders, including the
fundamental (60 Hz) and the DC component.
Figure 9 shows the decomposed signal which its
corresponding components from DC up to 11th harmonic. The
left column represents the original components and the right
column the corresponding components obtained through the
filter bank.
For simplicity and space limitation the higher components
are not shown. However, it is important to remark that all
waveforms of the time-varying harmonics are extracted with
 4

efficiency along the time.

DC component DC component
Naturally, intrinsic delays will be present during the 2 2
transitions from previous to the new state. Figure 10 shows
1 1
some components of both the original and decomposed signal
0 0
in a short time scale interval. 0 1 2 3 4 0 1 2 3 4
1st 1st
5 1 1
4 0 0
-1 -1
3
0 1 2 3 4 0 1 2 3 4
2nd 2nd
2
1 1
Magnitude

1
0 0
0 -1 -1
0 1 2 3 4 0 1 2 3 4
-1 3rd 3rd
1 1
-2
0 0
-3
0 0.5 1 1.5 2 2.5 3 3.5 4 -1 -1
Time (s) 0 1 2 3 4 0 1 2 3 4
Fig. 8 – Synthetic signal used 4th 4th
1 1

0 0
B. Simulated Signal -1 -1
0 1 2 3 4 0 1 2 3 4
It is well known that during energization a transformer 5th 5th
can draw a large current from the supply system, normally 0.5 0.5
called inrush current, whose harmonic content is high. 0 0
Although today’s power transformers have lower harmonic -0.5 -0.5
0 1 2 3 4 0 1 2 3 4
content Table 1, shows the typical harmonic components 6th 6th
present in the inrush currents [9]. These values are normally 0.2 0.2
used as reference for protection reasons, but they do not take 0 0
-0.2 -0.2
into account the time-varying nature of this phenomenon.
0 1 2 3 4 0 1 2 3 4
7th 7th
TABLE 1- TYPICAL HARMONIC CONTENT OF THE INRUSH CURRENT 0.5 0.5

Order Content % 0 0

Dc 55 -0.5
0 1 2 3 4
-0.5
0 1 2 3 4
2 63 8th 8th
3 26.8 1 1

4 5.1 0 0
5 4.1 -1 -1
0 1 2 3 4 0 1 2 3 4
6 3.7 9st 9st
7 2.4 0.2 0.2
0 0
-0.2 -0.2
In recent years, improvements in materials and transformer
0 1 2 3 4 0 1 2 3 4
design have lead to inrush currents with lower distortion 10th 10th
content [10]. The magnitude of the second harmonic, for 1 1

example, has dropped to approximately 7% depending on the 0 0

design [11]. But, independent of these new improvements, it -1 -1

0 1 2 3 4 0 1 2 3 4
is always important to emphasize the time-varying nature of 11rd 11rd
0.5 0.5
the inrush currents.
In being so, a transformer energization case was simulated 0 0

using EMTDC/PSCAD and the result is shown in Fig. 11. -0.5

0 1 2 3 4
-0.5
0 1 2 3 4
Using the methodology proposed to visualize the harmonic Time s Time s
content of the inrush current, Fig. 12 reveals the rarely seen Fig. 9 – First column: original components, second column: decomposed signals.
time-varying behavior of the waveform of each harmonic
component, where the left column shows the DC and even
components and the right column the odd ones. This could be
used to understand other physical aspects not observed
previously.
 5

DC component bank structure demand high computational effort. By using

2
multirate techniques to implement the structure it is possible
1
to show that the number of multiplication by second to
0
2.99 3 3.01 3.02 3.03 3.04 3.05 implement one branch of the analysis filter bank is about one
2nd
1 million. Some low price DSPr (Digital Signal Processors)
0 available in the market are able to execute 300 Million Float
-1
2.99 3 3.01 3.02 3.03 3.04 3.05
Point operations per second. This shows that, despite the
3rd higher computational effort of the structure, it is feasible to be
1

0
implemented in hardware. However, new opportunities exist
-1
for overcoming the limitations and the development of
2.99 3 3.01 3.02
7th
3.03 3.04 3.05
improved and alternative algorithms. For example, the authors
0.5 are investigating the possibility of extracting inter-harmonic
0
components and developing a similar methodology based on
-0.5
2.99 3 3.01 3.02 3.03 3.04 3.05 Discrete Fourier Transform (DFT). The first results show
Time s similar visualization capabilities with the promise to provide a
Fig. 10 – Comparing original and decomposed component reduced transient time response and lower computational
burden. This process has the potential of addressing specific
14
protective relaying needs such as detecting a high impedance
12
fault during transformer energization and detection of ferro-
10 resonance.
DC component 1st
8
5
Magnitude kA

4
6 2 0
-5
4 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

2
2nd 3rd
1
2
0 0 0
-2
-1
-2 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (s) 4th 5th
Fig. 11 – Inrush current in phase A. 0.5
0.2
0 0
-0.2
TABLE 1- TYPICAL HARMONIC CONTENT OF THE INRUSH CURRENT -0.5
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
6th 7th
Order Content % 0.2 0.1
Dc 55 0 0
-0.1
2 63 -0.2
0 0.2 0.4 0.6 0.8 1 0 0.5 1
3 26.8
8th 9st
4 5.1 0.1
0.1
5 4.1 0 0
6 3.7 -0.1 -0.1
7 2.4 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

10th 11rd

IV. FINAL CONSIDERATION AND FUTURE WORK 0.05

0.05

0 0
The methodology proposed has some intrinsic limitations -0.05 -0.05
associated with the analysis of inter-harmonics as well as with 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

real time applications. The Inter-harmonics (if they exist) are 12th
0.05
13th
0.05
not filtered by the bank, so it would corrupt the nearby 0 0
harmonic components. The limitation regarding with real time -0.05 -0.05
applications is related to the transient time response produced 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4
15th
0.6 0.8 1

by the filter bank. For example, in the presence of abrupt 0.04

14th
0.05

change in the input signal, such as in inrush currents, the 0.02

0 0
-0.02
transient response can last more than 5 cycles, which is not -0.04 -0.05
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
appropriate for applications whose time delay must be as Time (s) Time (s)
short as possible. The computational effort for real time
Fig. 12 – Decomposition of the simulated transformer inrush current.
implementation is another challenge that the authors are
investigating. In fact the high order filter (69th) used in the
 6

V. CONCLUSIONS (EFEI), Brazil, in 1984 and 1991, respectively, and his DSc degree in electrical
engineering from University of Santa Catarina (UFSC), Brazil, in 2001. He is
This paper presents a new method for time-varying Associate Professor at Federal University of Itajubá, where he is also the Power
harmonic decomposition based on multirate filter banks Quality Study Group sub-coordinator. He was a Research Scholar at the Center
for Advanced Power Systems (CAPS) at Florida State University during 2007.
theory. The technique is able to extract each harmonic in the
His areas of interest are Power System Protection, Power Quality and Signal
time domain. The composed structure was developed to work Processing for Instrumentation.
with 256 samples per cycle and to track up to the 15th
harmonic. The methodology can be adapted through Thomas L. Baldwin (S’86-M’92-SM’01) received the B.S.E.E. and M.S.E.E
convenient pre-processing for different sampling rates and degrees from Clemson University, Clemson, SC, and the Ph.D. degree in
electrical engineering from Virginia Polytechnic Institute and State University,
higher harmonic orders. Blacksburg, VA, in 1987, 1989, and 1993, respectively. In 1992, he joined the
M. B. Haynes Corporation as a Design Engineer. He then joined ABB Electric
VI. ACKNOWLEDGMENTS Systems Technology Institute in 1994 as a Senior Engineer. He is currently an
Associate Professor at the FAMU-FSU College of Engineering at Florida State
University, Tallahassee, FL, and a Research Engineer at the Center for
The authors would like to thanks Dr. Roger Bergeron and Advanced Power Systems. He is a Registered Professional Engineer in the State
of North Carolina.
Dr. Mathieu van den Berh for their useful and constructive
suggestions. Paulo F. Ribeiro (M'78–SM'88–F'03) received a BS in Electrical Engineering
from the Universidade Federal de Pernambuco, Recife, Brazil, completed the
VII. REFERENCES Electric Power Systems Engineering Course with Power Technologies, Inc.
(PTI) 1, and received the Ph.D. from the University of Manchester, Machester-
[1] Yuhua Gu, M. H. J. Bollen, “Time-Frequency and Time-Scale Domain
UK. Presently, he is a Professor of Engineering at Calvin College, Grand
Analysis,” IEEE Transaction on Power Delivery, Vol. 15, No. 4, October
Rapids, Michigan, on sabbatical at the Center for Advanced Power
2000, pp. 1279-1284.
Systems(CAPS) at Florida State University, Tallahassee, Florida. Dr. Ribeiro is
[2] P.P. Vaidyanathan, Multirate Systems and Filter Banks, Prentice Hall,
active in IEEE, CIGRE and IEC working groups on power quality. He is
1993.
Registered Professional Engineer in the State of Iowa.
[3] M. Karimi-Ghartemani, M. Mojiri and A. R. Bakhsahai, “A Technique
for Extracting Time-Varying Harmonic based on an Adaptive Notch
Filter,” Proc. of IEEE Conference on Control Applications, Toronto,
Canada, August 2005.
[4] J. R. Carvalho, P. H. Gomes, C. A. Duque, M. V. Ribeiro, A. S.
Cerqueira, and J. Szczupak, “PLL based harmonic estimation,” IEEE
PES conference, Tampa, Florida-USA, 2007
[5] C.-L. Lu, “Application of DFT filter bank to power frequency harmonic
measurement,” IEE Proc. Gener. , Transm,. Distrib., Vol 152, No.1,
January 2005, pp. 132-136.
[6] P. M. Silveira, M. Steurer, .P F. Ribeiro, “Using Wavelet decomposition
for Visualization and Understanding of Time-Varying Waveform
Distortion in Power System,” VII CBQEE, August 2007, Brazil.
[7] V.L. Pham and K. P. Wong, “Antidistortion method for wavelet
transform filter banks and nonstationay power system waveform
harmonic analysis,” IEE Proc. Gener., Transm., Distrib., Vol 148, No.2,
March 2001, pp. 117-122.
[8] Sanjit K. Mitra, Digital Signal Processing – A computer-based
approach, Mc-Graw Hill 2006, 3ª Edition.
[9] C.R. Mason, “The Art and Science of Protective Relaying,” John
Wiley&Sons, Inc. New York, 1956.
[10] B. Gradstone, “Magnetic Solutions, Solving Inrush at the Source”, Power
Electronics Technology, April 2004, pp 14-26.
[11] F. Mekic, R. Girgis, Z. Gajic, E. teNyenhuis, “Power Transformer
Characteristics and Their Effect on Protective Relays”, 33rd Western
Protective Relay Conference, Oct 2006.

VIII. BIOGRAPHY

Carlos A. Duque (M’91) was born in Juiz de Fora, Brazil, in 1962. He

received the B.S. degree in Electrical Engineering from the Federal University of
Juiz de Fora, Brazil, in 1986, and the M.Sc. and Ph.D. degree from the Catholic
University of Rio de Janeiro, in 1990 and 1997, respectively, in Electrical
Engineering. Since 1989 he is a Professor in the Engineering Faculty at Federal
University of Juiz de Fora (UFJF), in Brazil. He is a Visiting Researcher at the
Center for Advanced Power Systems (CAPS) at Florida State University,
Tallahassee, Florida. His research interests are power quality analysis, digital
instrumentation and digital signal processing.

Paulo Márcio da Silveira was born in Itajubá/MG, Brazil in 1960. He received

his BSEE and MSc degrees from the Federal Engineering School of Itajubá