An Improved Detection and Classification Technique of Harmonic Signals in Power Distribution by Utilizing Spectrogram

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 7, No. 1, February 2017, pp. 12~20
ISSN: 2088-8708, DOI: 10.11591/ijece.v7i1.pp12-20  12
Journal homepage: http://iaesjournal.com/online/index.php/IJECE
An Improved Detection and Classification Technique of
Harmonic Signals in Power Distribution by Utilizing
Spectrogram
M. H Jopri1
, A. R Abdullah2
, M. Manap3
, M. R Yusoff4
, T. Sutikno5
, M. F Habban6
1,2,6
Faculty of Electrical Engineering, Universiti Teknikal Malaysia Melaka, Malacca, Malaysia
3,4
Faculty of Engineering Technology, Universiti Teknikal Malaysia Melaka, Malacca, Malaysia
5
Department of Electrical Engineering, Universitas Ahmad Dahlan, Yogyakarta, Indonesia
Article Info ABSTRACT
Article history:
Received Oct 21, 2016
Revised Dec 21, 2016
Accepted Jan 13, 2017
This paper introduces an improved detection and classification technique of
harmonic signals in power distribution using time-frequency distribution
(TFD) analysis which is spectrogram. The spectrogram is an appropriate
approach to signify signals in jointly time-frequency domain and known as
time frequency representation (TFR). The spectral information of signals can
be observed and estimated plainly from TFR due to identify the
characteristics of the signals. Based on rule-based classifier and the threshold
settings that referred to IEEE Standard 1159 2009, the detection and
classification of harmonic signals for 100 unique signals consist of various
characteristic of harmonics are carried out successfully. The accuracy of
proposed method is examined by using MAPE and the result show that the
technique provides high accuracy. In addition, spectrogram also gives 100
percent correct classification of harmonic signals. It is proven that the
proposed method is accurate, fast and cost efficient for detecting and
classifying harmonic signals in distribution system.
Keyword:
Classification technique
Detection
Harmonic signal
Time-frequency distributions
Power distribution
Spectrogram
Copyright © 2017 Institute of Advanced Engineering and Science.
All rights reserved.
Corresponding Author:
M. H Jopri,
Advanced Digital Signal Processing Group, Faculty of Electrical Engineering,
Universiti Teknikal Malaysia Melaka,
Hang Tuah Jaya, 76100 Durian Tunggal, Melaka, Malaysia.
Email: hatta@utem.edu.my
1. INTRODUCTION
Power quality (PQ) disturbances analysis specifically for harmonics disturbance turns into an
exponentially expanding field of interest. The PQ analysis associates power engineering and power
electronics with digital signal-processing, artificial intelligence and optimization techniques [1]. PQ is
generally defined as any change in power (voltage, current, or frequency) that interferes with the normal
operation of electrical equipment. Furthermore, harmonics and inter-harmonics issues have been identified as
one of crucial issue in PQ [2]. To investigate these issues, the information is frequently accessible as a form
of sampled time function that is indicated by a time series of amplitudes. At the point when dealing such
information, the Fourier transform (FT) based approach is commonly utilized. FT assumes periodicity of a
given signal and loses the time axis account; nonetheless, FT not able in providing time information about
signal disturbances. Short time Fourier transform (STFT) offers both time and frequency information,
nevertheless it suffers severely from the Heisenberg uncertainty principle [2]. The Discrete Fourier transform
(DFT) is utilized for analyze frequency content in steady state and it is appropriate for harmonic analysis.
Conversely, it is not capable to fast changes in waveform. The DFT has foremost disadvantages such as
resolution, spectrum leakage as well as picket-fence effects [1]. The basics of wavelets and wavelet transform
can be denoted in [2]. The short time Fourier transforms (STFT) has the restriction of the fixed window

IJECE ISSN: 2088-8708 
An Improved Detection and Classification Technique of Harmonic Signals in Power .... (M. H Jopri)
13
width, hence it is insufficient for the analysis of the non-stationary harmonic disturbances. The problem of all
above signal processing techniques is the principle of Heisenberg’s uncertainty in which one cannot
distinguish what spectral components be present at what instance of time. The distinctive features that
describe harmonic disturbances and methods to extract from logged disturbances are also obtainable [3]. The
STFT resolution problems resolved by utilizing wavelet transform (WT. This makes it significantly for
tracing changes in signal including fast changes in high-frequency [4]. Awkwardly, WT abilities are often
significantly degraded in real practice especially in noisy situations [5]. Based on the discussion, a good
method is required due to identify the most perfect, quick and economical way for analysis of harmonic
signals.
In this paper, the spectrogram technique which is time frequency analysis technique for
distinguishing the signals in time frequency domain is introduced. The spectrogram is a squared magnitude of
STFT gives the time waveform energy distribution in joint-time frequency domain where this technique is
used in many applications and widely used as an initial investigative tool with less computational complexity,
fast and accurate detection and classification technique [6]. The harmonic signals are analyzed and
represented in time frequency representation (TFR). Furthermore, the parameters such as RMS and
fundamental value, total harmonic distortion (THD), total non-harmonic distortion (TnHD) and total
waveform distortion (TWD) for voltage and current are evaluated from TFR and used for the classification of
harmonic and inter-harmonic signals. The performance of the proposed technique is verified by classifying
100 signals with numerous characteristics for every type of voltage variation signals and also based on IEEE
Std. 1159-2009 [7].
2. HARMONICS AND INTER-HARMONICS
According to (IEEE Std. 1159-2009), harmonics which can be caused by rectifiers and inverters, are
the inverse of inter-harmonics where they are frequency components of distorted voltages and currents that
are integer multiples of the fundamental frequency (50Hz for Malaysia). Harmonic combines with the
fundamental voltage or current produces waveform distortion. Harmonic distortion exists due to the nonlinear
characteristics of devices and loads on the power system. Meanwhile, Inter-harmonics are frequency
components of distorted voltages that are not integer multiples of the fundamental frequency, which resulted
from cyclo-converter, arc furnaces, and adjustable speed drives (ASDs) [7]. Inter-harmonics can lead to
excitation of low frequency mechanical oscillations, and failure in ripple control. The main sources of inter-
harmonic waveform distortion are static frequency converters, cycle-converters, induction motors, and arcing
devices. Furthermore, in order to verify the performance of the proposed method, 100 signals with various
characteristics of harmonics and inter-harmonics accordingly with refer to IEEE Std. 1159-2009 are used
respectively.
3. LINEAR TIME-FREQUENCY DISTRIBUTIONS
Time-frequency distributions are great techniques that represent a signal in, jointly, time and
frequency representation as known as time-frequency representation (TFR). From TFR, spectral information
of a signal can be observed with changes of time [8]. Therefore, the TFD are very applicable to analyse
power quality signals that consist of non-stationary and multi-frequency components signal. In the following
sections, one of TFDs which is discussed.
By windowing the signal at first and after that taking the Fourier transforms the time-localization
can be obtained correctly. Furthermore, this will lead to the rise of short time Fourier transform, (STFT) or
windowed Fourier transform [9]. The continuous-time STFT of a signal, x(t), is expressed as
 
detwxft fj
x
2
)()(),( 


 
(1)
where x(τ) is the signal which is used as an input and w(τ) is the window which is used as an observation.
However, the squared value of the STFT is, usually used in the signal analysis which called as a spectrogram.
The analysis of spectrogram is represented in TFR that represents three-dimensional imitation of the signal
energy when compared with time and frequency and is shown mathematically as:

 ISSN: 2088-8708
IJECE Vol. 7, No. 1, February 2017 : 12 – 20
14
2
2
)()(),(  
detwxftP fj
x



 
(2)
In this approach, the Hanning window is chosen due to its lower peak side lope which is the narrow
effect on other frequencies around fundamental value which is 50 Hz and other frequency components. The
Hanning window function can be formulated as:
  






T
t
tw
2
cos1
2
1
(3)
4. SIGNAL PARAMETERS
Parameters of signals of voltage variation are estimated from the TFR. The signal constituents are
momentary RMS voltage and RMS fundamental voltage, instantaneous total waveform distortion (TWD),
instantaneous total harmonic distortion (THD) and instantaneous total inter-harmonic distortion (TnHD)
respectively. The following are parameters of signal that been obtained from TFR [10].
Instantaneous RMS Voltage, dfftPtV
sf
xrms 
0
),()(
(4)
Instantaneous RMS Fundamental Voltage, 
hi
lo
f
f
xrms dfftPtV ),(2)(1
(5)
Instantaneous Total Waveform Distortion,
)(
)()(
)(
1
2
1
2
tV
tVtV
tTWD
rms
rmsrms 

(6)
Instantaneous Total Harmonic Distortion,
(t)V
(t)V
tTHD
rms
H
h rmsh
1
2
2
,
)(
 

(7)
Instantaneous Total Nonharmonic Distortion, (8)
where Px(t, f) is the TFR of a signal, fs is sampling frequency, f0 is the fundamental frequency, V1rms(t) is
instantaneous RMS fundamental voltage, Vrms(t) is instantaneous RMS voltage and Vh,rms (t) is RMS harmonic
voltage. fhi=f0+25Hz, flo=f0-25Hz, 25 Hz is chosen for fhi and flo, it can represent the fundamental frequency
value and use for calculate the value of the frequency element.
5. SIGNAL CHARACTERISTIC
The determination of signal characteristics is from the calculated signal parameters. Furthermore, by
using the instantaneous RMS voltage, the signal properties for example the average of RMS voltage can be
calculated using equation 9 [11]. The information of the signal characteristics is used as input for classifier to
allocate the harmonic and inter-harmonic signals.
(9)
)(
)()(
1
0
2
,
2
tV
tVtV
TnHD(t)
rms
H
h rmshrms  



T
rmsaverms dttV
T
V
0
, )(
1

IJECE ISSN: 2088-8708 
15
Average of total harmonic distortion, THDave and total nonharmonic distortion, TnHDave can be
calculated from instantaneous total harmonic distortion, THD(t) and instantaneous total nonharmonic
distortion, TnHD(t), respectively. They can be defined as:
.
∫ ( ) (10)
∫ ( ) (11)
6. SIGNAL CLASSIFICATION
The rule-based classifier is a deterministic classification method that has been, widely, used in real
world application. The classification performance is much reliant on the best rules and threshold values [6].
In this analysis, since the signal characteristics give good previous information for the power quality signals,
the rule-based classifier which is rule is appropriate to be utilized for signal classification. The rule-based
classifier that been used in order to identify harmonics and inter-harmonics as in equation 12 and 13 [12].
> and < (12)
>= and < (13)
Meanwhile, the flow chart in Figure 1 shows a process of rule-based classifier of the harmonics and
inter-harmonics signals. Harmonic and inter-harmonic, their threshold value, THDthres and TnHDthres are set
based on several analyses made for these signals. THD or TnHD considered exist in the signal as their
magnitude is greater or equal than the threshold value. Thus, harmonic signal is classified as a signal that has
only THD while inter-harmonic signal has only TnHD.
Figure 1. The Implementation Flow Chart for Harmonics and Inter-Harmonics Detection
7. ACCURACY MEASUREMENT
The performance and viability of proposed method depend on the accuracy of this technique and the
evaluation of accuracy can be done as follows.
The accuracy of the analysis is identified by measuring the signal characteristics measurement
accuracy. The characteristics measurement is very important because it will be used as input for the classifier
of the signal. Low accuracy of the measurement will result in misinterpretation in exact signal characteristics.
To measure the exactness of the measurement, mean absolute percentage error (MAPE) is used as an
index [13].

 ISSN: 2088-8708
16
It can be defined as:
(14)
where xi(n) is an actual value, xm(n) is measured value and N is the number of data. The smaller value of
MAPE offers more accurate results.
8. RESULTS AND ANALYSIS
This section will discuss the results for harmonics and inter-harmonics detection technique utilizing
spectrogram.
8.1. Detection of Harmonic Signal
Figure 2(a) and (b) present harmonic signal in time domain and its TFRs using spectrogram. The
TFRs indicate that the signal consists of two frequency component: fundamental frequency (50Hz) and 7th
harmonic component (350Hz). Signal parameters estimated from the TFR using spectrogram is shown in
Figure 2(b). Figure 2(c) shows that the harmonic voltage contributes to the increases of RMS voltage from
normal voltage which is 1.0 to 1.17 pu. However, it does not change the RMS fundamental voltage which
remains constant at 1.0 pu. Besides that, the signal also results the magnitude of TWD and THD remains
constant at 60% and zero percent for the TnHD as shown in Figure 2(d). Thus, the detection of harmonic
signal successfully carried out using the spectogram technique.
(a) (b)
(c) (d)
Figure 2. (a) Harmonic Signal from Simulation and its, (b) TFR using Spectrogram, (c) Instantaneous of
RMS and Fundamental RMS Voltage, (d) Instantaneous of Total Harmonic Distortion, Total Nonharmonic
Distortion And Total Waveform Distortion
%100
)(
)()(1
1
x
nx
nxnx
N
MAPE
N
n i
mi




IJECE ISSN: 2088-8708 
17
8.2. Detection of Inter-Harmonic Signals
Inter-harmonic signal and its TFR using spectrogram is shown in Figure 3(a) and (b), respectively.
As shown in Figure 3(b), the signal has two frequency components which are at fundamental frequency
(50 Hz) and inter-harmonic frequency of 375 Hz. Signal parameters estimated from the TFR using
spectrogram are shown in Figure 3(b). Figure 3(c) shows that the inter-harmonic voltage give rise to the
increases of RMS voltage from normal voltage from 1.0 to 1.17 pu. However, it does not change the RMS
fundamental voltage which remains constant at 1.0 pu. Besides that, the signal also results the magnitude of
TWD and TnHD remains constant at 60% and zero percent for the THD as shown in Figure 3(d). The
detection of inter-harmonic signal magnificently accomplished using the proposed technique.
(a) (b)
(c) (d)
Figure 3. (a) Interharmonic Signal from Simulation and its, (b) TFR using Gabor Transform,
(c) Instantaneous of RMS and Fundamental RMS Voltage, (d) Instantaneous of Total Harmonic Distortion,
Total Nonharmonic Distortion and Total Waveform Distortion
8.3. The Accuracy of the Analysis
Harmonic and inter-harmonics signals are tested and mean absolute percentage error (MAPE) of the
signal properties are calculated. Then, the results are averaged to identify the accuracy of the measurements
as shown in Table 1. The table shows that spectogram gives good accuracy for an average of RMS voltage,
THD and TnHD. In addition, spectogram has a good accuracy in harmonic and inter-harmonic signals
detection.

 ISSN: 2088-8708
18
Table 1. MAPE Simulation Result for Spectrogram Analysis
Signal Characteristics MAPE (%)
Vrms,ave 0.1572
THDave 0.1551
TnHDave 0.1595
8.4. Detection and Classification of Harmonic and Inter-Harmonic Signals
As shown in the previous section, spectogram has high accuracy in identifying the signal
characteristics. Furthermore, the performance results of the signals classification using the spectogram are
shown in Table 2. 100 signals with various characteristics for each type of voltage signal is generated and
classified. The table shows that the classification results using spectogram give 100% correct classification
for all signals.
Table 2. Performance of Harmonic and Inter-Harmonic Signals Classification
Signal
Gabor Transform
Number of data sets % Correct Classification
Harmonics 100 100
Inter-harmonics 100 100
Normal 100 100
9. CONCLUSION
The performance evaluation of the signal analysis using spectogram are implemented in time
domain as well as frequency domain in terms of accuracy using MAPE. Furthermore, the verification of
detection and classification of harmonic signals are utilizing 100 signals with various characteristics for each
type of voltage signals. The results show that the spectogram gives good accuracy and gives 100 percent
correctness of signals classification. Hence, spectogram is an appropriate technique to be implemented for
harmonic signals detection and classification in power distribution system.
ACKNOWLEDGEMENTS
This research is supported by Advance Digital Signal Processing Laboratory (ADSP Lab). Special
thanks also to the Faculty of Electrical Engineering and Engineering Technology of Universiti Teknikal
Malaysia Melaka (UTeM), Ministry of Higher Education Malaysia (MOHE) and Ministry of Science,
Technology and Innovation (MOSTI) for giving the cooperation and funding for this research with grant
number 06-01-14-SF00119 L00025. Their support is gratefully acknowledged.
REFERENCES
[1] S. Khokhar, et al., “Automatic Classification of Power Quality Disturbances : A Review,” IEEE Student Coference
Res. Dev., no. December, pp. 16–17, 2013.
[2] O. Ozgonenel, et al., “A new classification for power quality events in distribution systems,” Electr. Power Syst.
Res., vol. 95, pp. 192–199, 2013.
[3] M. Gupta, et al., “Neural Network Based Indexing and Recognition of Power Quality Disturbances,”
TELKOMNIKA (Telecommunication Comput. Electron. Control., vol/issue: 9(2), pp. 227–236, 2011.
[4] S. A. Deokar and L. M. Waghmare, “Integrated DWT–FFT approach for detection and classification of power
quality disturbances,” Int. J. Electr. Power Energy Syst., vol. 61, pp. 594–605, 2014.
[5] M. V. Rodriguez, et al., “Detection and classification of single and combined power quality disturbances using
neural networks,” IEEE Trans. Ind. Electron., vol/issue: 61(5), pp. 2473–2482, 2014.
[6] A. A. Ahmad and A. Zuri, “Analysis and Classification of Airborne Radar Signal Types Using Time-Frequency
Analysis,” vol/issue: 2014(September), pp. 23–25, 2014.
[7] N. H. H. Abidullah, et al., “Real-Time Power Quality Disturbances Detection and Classification System,” World
Appl. Sci. J., vol/issue: 32(8), pp. 1637–1651, 2014.
[8] M. Manap, et al., “Comparison of Open and Short-Circuit Switches Faults Voltage Source Inverter (VSI) Analysis
Using Time-Frequency Distributions,” Appl. Mech. Mater., vol/issue: 752–753(APRIL), pp. 1164–1169, 2015.
[9] A. R. Abdullah, et al., “Short-circuit switches fault analysis of voltage source inverter using spectrogram,” Electr.
Mach. Syst. (ICEMS), 2013 Int. Conf., pp. 1808–1813, 2013.
[10] A. R. Abdullah, et al., “Power Quality Monitoring System Utilizing Periodogram And Spectrogram Analysis
Techniques,” Int. Conf. Control. Instrum. Mechatronics Eng., no. August, pp. 770–774, 2007.
[11] N. A. Abidullah, et al., “Real-time power quality signals monitoring system,” Proceeding - 2013 IEEE Student

IJECE ISSN: 2088-8708 
19
Conf. Res. Dev. SCOReD 2013, no. December, pp. 433–438, 2015.
[12] A. R. Abdullah, et al., “Performance Evaluation of Real Power Quality Disturbances Analysis Using S-Transform,”
Appl. Mech. Mater., vol. 752–753, pp. 1343–1348, 2015.
[13] K. R. Cheepati and T. N. Prasad, “Performance Comparison of Short Term Load Forecasting Techniques,” Int. J.
Grid Distrib. Comput., vol/issue: 9(4), pp. 287–302, 2016.
BIOGRAPHIES OF AUTHORS
Mohd Hatta Jopri was born in Johor, Malaysia on 1978. He received his B.Sc from Universiti
Teknologi Malaysia in 2000 and Msc. in Electrical Power Engineering from Rheinisch-
Westfälische Technische Hochschule Aachen (RWTH), Germany in 2011. Since 2005, he has
been an academia staff in the Universiti Teknikal Malaysia Melaka (UTeM) and currently he
pursuing his PhD in the area of power quality.
Dr. Abdul Rahim Abdullah was born in Kedah, Malaysia on 1979. He received his B. Eng.,
Master Eng., PhD Degree from Universiti Teknologi Malaysia in 2001, 2004 and 2011 in
Electrical Engineering and Digital Signal Processing respectively. He is currently an Associate
Professor with the Department of Electrical Engineering, Chief of Advanced Digital Signal
Processing (ADSP) Lab and Center of Excellent (COE) Coordinator for Universiti Teknikal
Malaysia Melaka (UTeM).
Mustafa Manap was born in Kuala Lumpur, Malaysia on 1978. He received his B.Sc from
Universiti Technologi Malaysia in 2000 and Msc. in Electrical Engineering from Universiti
Teknikal Malaysia Melaka (UTeM) 2016. Since 2006, he has been an academia staff in the
Universiti Teknikal Malaysia Melaka (UTeM). His research interests are power electronics and
drive, instrumentation, and DSP application.
Mohd Rahimi bin Yusoff was born in Kelantan, Malaysia on 1967. He received his M.Sc. in
Electrical Engineering from University of Nottingham, England in 1992 and B.Eng. in Electrical
& Electronics from University of Wales, Swansea, UK in 1990. Since 2010, he has been an
academia staff in Universiti Teknikal Malaysia Melaka (UTeM) and currently holding the post
as Dean of Faculty of Engineering Technology. His research interests are power quality, signal
processing, and DSP application.
Tole Sutikno is an Associated Professor in Department of Electrical Engineering at Universitas
Ahmad Dahlan (UAD), Indonesia. He is Editor-in-Chief, Principle Contact and Editor for some
Scopus indexed journals. He is also one of founder of the Indonesian Publication Index (IPI). He
has completed his studies for B.Eng., M.Eng., Ph.D. in Electrical Engineering from Diponegoro
University, Gadjah Mada University and Universiti Teknologi Malaysia, respectively. His
research interests include the field of industrial electronics and informatics, power electronics,
FPGA applications, embedded systems, data mining, information technology and digital library.

 ISSN: 2088-8708
20
Mohd Faiz bin Habban was born in Perak, Malaysia on 1987. He graduated with Bachelor of
Electrical Engineering Technology in Power Industry (Hons.) from Universiti Teknikal Malaysia
Melaka. He has doing the industrial training at Malakoff Berhad, Prai Power Sdn Bhd and
working at Masterplan Consulting Sdn Bhd, as Junior Engineer. Currently, he pursuing his
Master at Universiti Teknikal Malaysia Melaka (UTeM). His area of interest is Electrical Power
Engineering, Power Electronics & Drive, Digital Signal Processing and Energy Efficiency.

An Improved Detection and Classification Technique of Harmonic Signals in Power Distribution by Utilizing Spectrogram

More Related Content

What's hot (18)

Similar to An Improved Detection and Classification Technique of Harmonic Signals in Power Distribution by Utilizing Spectrogram (20)

More from IJECEIAES (20)

Recently uploaded (20)

An Improved Detection and Classification Technique of Harmonic Signals in Power Distribution by Utilizing Spectrogram