You are on page 1of 42

Received: 12 September 2018 Revised: 31 December 2018 Accepted: 6 February 2019

DOI: 10.1002/2050-7038.12008

REVIEW ARTICLE

Power quality disturbance detection and classification using


signal processing and soft computing techniques: A
comprehensive review

Manohar Mishra

Department of Electrical and Electronics


Engineering, Institute of Technical
Summary
Education and Research, Siksha ‘O’ Power quality (PQ) studies have gained huge attention from the academics and
Anusandhan (Deemed to be University), the industry over the past three decades. The main objective of this article is to
Bhubaneswar, India
provide a comprehensive review on the state‐of‐the‐art techniques based on dig-
Correspondence ital signal processing (DSP) and machine learning for automatic recognition of
Manohar Mishra, Department of
PQ events. It is aimed to present extensive information on the status
Electrical and Electronics Engineering,
Institute of Technical Education and of detection and classification of PQ events to the academics following a line of
Research, Siksha ‘O’ Anusandhan investigation on the similar domain. On the other hand, microgrid is one of the
(Deemed to be University), Bhubaneswar
emerging architecture under the umbrella of smart grid infrastructure. In
751030, India.
Email: manohar2006mishra@gmail.com microgrid environment, the integration of renewable energy sources and distrib-
uted generators boosts the application of power electronic technology, which is
the main cause of various PQ issues. Therefore, various PQ detection and classi-
fication (PQD&C) schemes for microgrid application using DSP and machine
learning techniques have been reviewed in this article. Moreover, a comparative
assessment on various PQD&C techniques has been carried out in details consid-
ering several criteria, such as type of data used (synthetic or real world), type of
PQ disturbance studied (single or multiple), and performance in terms of accu-
racy (both noiseless and noisy environment). Hence, several types of research
work in PQD&C area, such as the detection principles, benefits, and weaknesses
of former works related to each PQD&C technique, are tinted in this manuscript.
Therefore, from this review, the opportunities in PQD&C research domain in the
power system can be explored further.

KEYWORDS
power quality, digital signal processing, machine learning techniques, detection and classification

1 | INTRODUCTION

Nowadays, the world has become more industrialized, and therefore, the uses of sensitive electronic equipment, solid‐
state switching devices, nonlinear load, power electronic converters, and relaying/protective devices have been
increased. The increasing application of such equipments/devices is the main source of unwanted deviation in voltage,
current, and frequency signals. Any deviation displayed in those power signals from the normal values is considered as a
power quality (PQ) disturbances. The integration of nonconventional energy sources (NCES) to the conventional power
grid is another source of power quality disturbances (PQDs). In short, power quality issues can cause failure or
Int Trans Electr Energ Syst. 2019;29:e12008. wileyonlinelibrary.com/journal/etep © 2019 John Wiley & Sons, Ltd. 1 of 42
https://doi.org/10.1002/2050-7038.12008
2 of 42 MISHRA

malfunctioning of protective electrical and electronic devices, damage to computer files and memory failure of sympa-
thetic loads (for example, computer, PLC controls, protection, and relaying apparatus), and unpredictable action of
automatic controllers.1 Hence, it is very essential to monitor these disruptions continuously. Uninterrupted monitoring
is mandatory because of the growing necessity of renewable power as mentioned in Khan2 and Ouyang and Wang,3 and
monitoring standards are also specified in Smith et al.4
As the PQ issues directly affect the overall continuation of electric transmission and distribution network, the unin-
terrupted monitoring of electric power systems (EPSs) has become highly necessary for the utility industry. Figure 1
shows the status of PQ detection field representing its popularity. The figure presents the numbers of articles reported
in the Scopus database annually,5 which are obtained by searching the exact phrase “power quality” in the title of the
article. It indicates that the PQ, as an ongoing issue, adds significant attention to power system researchers. In this
growing world, the integration necessities of NCES and distributed generation (DG) systems into the main grid boost
the use of power electronic technology (PET) which may be responsible for various power quality issues in EPS. Hence,
advance research development in the power quality analysis domain will be multiplied in the future owing to the
further applications of the PET used in DG and NCES.6
PQ indices are the justification of power quality standards for representing the adverse effect of electrical
disturbances, such as nominal frequency deviation, flicker, supply voltage variation and voltage transients, and
harmonics and interharmonics. Various most likely indices can be derived from the signal frequency spectrum like peak
values, crest factor, total harmonic distortion (THD), power factor, instantaneous distortion energy ratio, burst indices,
instantaneous frequency, etc. The standards allude to approaches in the frequency domain for endowing a tolerance in
the applied algorithms for example fast Fourier transform (FFT), chirp Z transform, Goertzel algorithm, zoom FFT, and
Welch algorithm, amongst others, which have been broadly utilised for monitoring power system parameters.7,8
However, at times, the info presented by means of the traditional standards is unsatisfactory.9 Furthermore, in Broshi,10
several additional issues which are not solved using the standards are illustrated. For example, one major drawback of
FFT‐based approaches is that they operate stationary signals, and therefore, their application is not appropriate for
sensing small spikes or transient in the signals. Thus, numerous investigations offer novel PQ indices for nonstationary
signals using time‐frequency transforms, like short‐time Fourier transform (STFT), wavelet transform (WT) and its
variants, S‐transform (ST) and its variants, Hilbert‐Huang transform (HHT), Kalman filters (KFs), Gabor transform
(GT) and their variants, etc.
There have been a number of review articles on different aspects of PQD&C.11-17 Flores11 reviewed the state‐of‐the‐art
methods based on SPTs for automatic recognition of PQ events. Ibrahim et al12 studied the uses of several developed
mathematical tools and wavelet transform in PQD&C approach. A widespread literature survey on applications of fuzzy
logic, expert systems, neural networks, and genetic algorithms in power quality was incorporated. Various techniques
and methodologies used for power quality assessment in EPS and classification of power disturbances were reviewed
by Granados‐Lieberman et al13 in general. A brief study on PQD&C approaches described earlier in the previous
studies14-16 but little insight into the widespread assessment of different schemes. An analytical literature survey on
PQD&C was discussed in Khokhar et al17 within a time span of 1990 to 2014. However, novel and more robust tech-
niques for the detection and classification of power quality events are presently accessible
Therefore, the tenacity of this manuscript is to provide a well‐organized, up‐to‐date, and an ample review on PQ
definitions, standards, PQ event types, underlying causes and negative effects, feature extraction, feature selection,

FIGURE 1 Numbers of article published in yearwise related to power quality analysis


MISHRA 3 of 42

and classification of PQ events. Moreover, this article provides an efficient overview on the existing methodology to
detect and classify the PQ events, in two aspects—application of signal processing techniques (SPTs) and application
of pattern recognition techniques (PRTs). For the application of signal processing techniques, different SPT‐based
approaches for PQD&C proposed in the present literature are conferred succinctly. Consequently, the article discusses
several concerns to apply SPTs and PRTs for PQ disturbance recognition with specific future research scopes.
Additionally, it discusses some real‐time application, and it is issued in the smart grid which can be addressed using
SPTs. Numerous future research scopes in this direction are highlighted. It is well known that the integration of
renewable energy sources and distributed generators boosts the application of PET, which is the main cause of
various PQ issues. Thus, various PQD&C schemes for microgrid application using SPTs and PRTs have been reviewed
and reported in this article.
In short, our main objective of this article is to deliver the following:

1. A comprehensive overview of PQ disturbances in terms of taxonomy, standards, and security issues.


2. A clear concept of feature extraction methodology based on SPT applications in PQ detection.
3. A state‐of‐the‐art review of PQ event classification methodology based on PRT applications.
4. An advanced assessment of feature selection (FS) techniques based on the optimization algorithm in automatic PQ
detection and classification.
5. A brief discussion on application of various PQD&C schemes for microgrid environment using SPTs and PRTs.
6. A comparative analysis between various existing PQD&C techniques.
7. Some useful remarks which can be used for the further improvement of the detection technique.

The subsequent sections of the manuscript are organized as follows: Section 2 describes the detailed overview on PQ
disturbances. Section 3 presents the taxonomy of PQ detection and classification techniques. In Section 4, a comparative
description analysis of various PQD&C techniques has been discussed. Lastly, Section 5 concludes the review on PQ
detection and classification technique with some suggested future scope.

2 | D E F I N IT I O N S

In general, this section includes the basic of PQ with respect to its definition(s), standards, taxonomy, causes, and effects
on consumer.

2.1 | Power quality

Several definitions of PQ were suggested by numerous sources. Some of the definitions are as follows:

• PQ is a concept of powering and grounding sensitive equipment in a matter that is suitable to the operation of that
equipment (IEEE dictionary).18,19
• Characteristics of the electricity at a given point on an electrical system, evaluated against a set of reference technical
parameters (IEC 61000‐4‐30).20
• PQ is the combination of voltage quality and current quality.21
• The ability of the power systems to deliver undistorted voltage, current, and frequency signals is termed as quality of
power supply.15
• Power quality is a set of electrical boundaries that allows a piece of equipment to function in its intended manner
without significant loss of performance or life expectancy.

2.2 | Standard on PQ
The power quality standards have recognized the reliable description of the PQ events. The insignificant operating
circumstances of the voltage, current, and their limits of variations within the EPS supply and the loading apparatus
are defined. Furthermore, the choice of the suitable monitoring apparatuses, their limits, application methods, and
the elucidation of outcomes has also been demonstrated. The classification of PQ events based on the threshold limit sets
for voltage and current signals variation in relation to nominal operative conditions during the disturbance
4 of 42 MISHRA

occurrence is stated by IEEE 1159 standard4 and the European EN 50160 standard.22 Moreover, the IEC 61000‐4‐30
standard23 has recognized the consistent techniques for measurement and interpretation of electronic parameters in
50/60 Hz EPS.

2.3 | Taxonomy of PQ events

IEEE standard 1159‐1995 suggested practice on monitoring electric PQ, terminology related to PQ and impact of bad PQ
on utility and consumer. This standard also states the variation of magnitude and time span for various PQ events.
Table 1 highlights the various types of PQ events based on the variation of voltage magnitude, time duration, and
spectral contents.4

2.4 | The effect of poor power quality on consumers

The primary types of PQ events and their negative effects to the consumers are described in Table 2 with the details of
underlying causes.

TABLE 1 Taxonomy of PQ events

Taxonomy of
Electromagnetic Standard Spectral Variation of Voltage
Phenomenon Contents Time Span Magnitude

Short duration variation


Interruption 0.5 Cycle‐3 s (momentary), 3 s‐1 min (temporary) <0.1 pu
Sag 0.5 Cycle‐30 cycles (instantaneous), 0.1‐0.9 pu
30 cycles‐3 s (momentary), 3 s‐1 min (temporary)
Swell 0.5 Cycle‐30 cycles (instantaneous) 1.1‐1.8 pu
30 Cycles‐3 s (momentary) 1.1‐1.1 pu
3 s‐1 min (temporary) 1.1‐1.2 pu
Long duration variation
Interruption (sustained) >1 min 0.0 pu
Undervoltage (UV) >1 min 0.8‐0.9 pu
Overvoltage (OV) >1 min 1.1‐1.2 pu
Waveform distortion
DC offset Steady state 0%‐0.1%
Harmonics 0‐100th H Steady state 0%‐20%
Interharmonics 0‐6 kHz Steady state 0%‐2%
Notching Steady state
Noise Broadband Steady state 0%‐1%
Transients
Impulsive <50 μs to >1 ms
Oscillatory
Low frequency <5 kHz 0.3‐50 ms 0‐4 pu
Medium frequency 5‐500 kHz 20 μs 0‐8 pu
High frequency 0.5‐5 MHz 5 μs 0‐4 pu
Voltage unbalance (VU) Steady state 0.5%‐2%
MISHRA

TABLE 2 PQ event type, underlying causes, and negative effect

Electromagnetic
Phenomenon Primary Sources Adverse Effect

Interruption Interruptions can be the consequence of EPS faults, apparatus 1. Instantaneous interruptions may impinge on electronic and lighting apparatus
breakdowns, and control errors. affecting malfunction or shutdown.
2. Momentary and temporary interruptions will nearly every time cause electric
apparatus to discontinue functioning. In certain cases, it may harm electronic
soft‐start tools.
Sag 1. Voltage sags are typically allied with EPS faults. A small speed variation of induction machines and a trivial decrease in output
2. It can also be caused by switching of large loads or from a capacitor bank can happen during sag.
starting of large motors.
Swell 1. These are commonly associated with EPS fault situations, but An additional rise in supply voltage to the EPS machinery above its nominal
they are much less common than voltage sags. rating may cause failure of the equipment subject to the frequency of incidence.
2. It can also be caused by load switching (OFF condition) Electronic gadgets, containing adjustable speed drives, central processing unit,
or/and capacitor switching (ON condition). and automatic controllers, may indicate instant breakdown modes in these
circumstances.
Undervoltage A load switching (ON), or a capacitor bank switching (OFF), 1. UVs with time span greater than 1 min can cause EPS apparatus to malfunction.
can produce an UVs till the voltage regulator acts to bring Motor regulators can failure during UV situations.
the voltage back to within limits. 2. Large heating loss in induction motors can occur because of the increased
motor current in case of long duration UV.
3. The effect of UV on capacitor banks result in a decrease of output power.
Overvoltage OVs can be the result of load switching like, load switching OV condition may cause equipment failure.
(OFF), or a capacitor bank switching (ON).
DC offset 1. This type of waveform distortion can happen as the
consequence of a geomagnetic disturbance.
2. It can also be caused by the result of half‐wave rectification.
Harmonics Its existence owing to the nonlinear characteristics of electric The harmonic distortion causes several adverse effect like the following:
equipment and loads on the EPS. 1. Overheating of rotational machines, static machines (for example, transformer),
and current‐carrying conductors.
2. Premature breakdown or action of protecting appliances (for example, fuses)
and harmonic resonance situations on the customers' EPS. This will further
help to weaken normal operation of EPS and metering imprecisions.
(Continues)
5 of 42
6 of 42

TABLE 2 (Continued)

Electromagnetic
Phenomenon Primary Sources Adverse Effect
Interharmonics The key resources of this type of waveform distortions are static frequency
converters, cycloconverters, induction motors, and arcing devices.
Notching It is an intermittent voltage distortion initiated by the usual operation of
power electronic devices when current is commutated between two
phases.
Noise Noise is an unnecessary electric quantity having broadband frequency
components within 200 kHz overlaid upon the EPS voltage/current in
phase conductors. Noise in EPS can be caused by PEDs, control circuits,
arcing apparatus, loads with solid‐state rectifiers, and switching power
supplies.
Impulsive transients The major reason of this type of transients is lighting. These transients are Transient voltages produced by means of switching operations or lightning can
usually caused by switching events. effect in degradation or instant dielectric breakdown in every type of equipment.
Oscillatory transients The major reason of this type of transients is lighting. These transients are Sharp magnitude and quick rise time causes the insulation failure in EPS
usually caused by switching events. apparatus such as electrical machines, power and measurement transformers,
capacitors, cables, and relay. Moreover, it may prompt nuisance tripping of
adaptable speed drives owing to the dc link OV protection electrical system.
Voltage unbalance Unbalanced loads are the main sources of VU. VU may also be the effect VU more usually arises in discrete customer loads caused by phase load
of capacitor bank abnormalities, like a blown fuse on one‐phase of a imbalances, particularly where heavy, one‐phase power loads are handled.
three‐phase bank.
MISHRA
MISHRA 7 of 42

3 | AUTOMATIC DETECTION AND C LASSIFICATION O F PQ


DISTURBANCES

The fundamental principle of PQD&C can be described using a flowchart as presented in Figure 2. Initially, the input dis-
turbance signal has been processed thru the preprocessing unit named as feature extraction unit. In the next step, the
extracted features are passed through the feature selection unit to provide the most distinctive, unique, and optimum fea-
ture vector with the least number of features. Then, the selected feature vector is used as input to the intelligent classifier in
the subsequent stage. The output of the intelligent classifier is used to make the final decision in the decision‐making stage.

3.1 | Feature extraction techniques

The performance of PQD&C approach generally depends on the accuracy of the intelligent classifier as illustrated in
Figure 2. The input to the intelligent classifier is the feature vector; therefore, the extraction of most distinctive features
from the original signal is furthermost important. The feature extraction process is generally carried out by using signal
processing techniques (SPTs) as suggested by various researchers. Several SPTs have been used for the feature extrac-
tion, eg, FT, WT, ST, HT, KF, and GT. An advanced taxonomy of the SPTs employed for the purpose of feature extrac-
tion from power quality events is presented in Figure 3. Table 3 is provided to establish a fair analysis of the stated SPTs
reported in this subsection, comprising key benefits and drawbacks of each technique.

3.1.1 | Fourier transform and its variants' based methods

The FT is the most widely used technique for frequency domain analysis. The input signal which is selected for analysis
can be described as a sum of essential sinusoids of different frequencies. Three variants of FT such as DFT, FFT, and STFT
were generally used for PQ disturbance recognition by various researchers in the last two decades. The PQ signals are gen-
erally nonstationary in nature. Thus, DFT is unsuitable to sense the immediate variations in PQ disturbances, for example,
their initial and final points. It is mostly apt for the stationary power quality events. The FFT technique provides a similar
result but in less time as compared with DFT.24 The FFT is extensively applied for harmonic estimation of PQD.25 Heydt

FIGURE 2 Block diagram representation of automatic detection and classification techniques of PQ events
8 of 42 MISHRA

FIGURE 3 Different feature extraction


methods based on SPTs

et al26 presented the techniques for PQ study based on windowed FFT. The STFT technique is another variant of FT, which
splits the waveform into small stationary segments. In this respect, STFT used to get the frequency/phase information of
the signals as they vary with respect to the time. Using a moving window, the relation between the time and frequency
variation can be recognized.27,28 Analysis of nonstationary signal by moving window STFT is very difficult but can be
achieved by fixed window.29 Gu and Bollen30 applied STFT for the study of nonstationary PQ disturbances. Beichang
et al31 proposed modified two‐stage FFT algorithm–based approach to detect harmonic and interharmonics which are
generated because of the integration of renewable energy sources.

3.1.2 | Wavelet transform and its variants' based methods

WT is having several advantages over FT as presented in Addison.32 It is one of the most powerful feature extraction
methods for power quality signals by considering a multiresolution analysis (MRA) technique.33 The wavelet transform
coefficients possess the key attributes and characteristics of power quality disturbance signals in various frequency sub‐
bands. Several statistical parameters of PQ disturbance signals such as amplitude, mean, median, kurtosis, energy, stan-
dard deviation, and entropy can be calculated from the approximate and detailed coefficient of WT, which can be used
further to recognize the PQ events. Therefore, WT has been widely applied by the academic investigators to characterize
and classify of PQ disturbances. Three variants of WT such as continuous WT (CWT), discrete WT (DWT), and wavelet
packet transform (WPT) were generally used for PQ disturbance recognition by various researchers in last two decades.
Initially, the application of WT for detection of nonstationary PQ signals was carried out by the authors34-36 in 1996
and 1997. Pillay et al18 applied WT to reconstruct the stationary WT. Santoso et al34,36 proposed WT‐based approach to
detect and classify PQ events. Here, the multiresolution signal decomposition (MSD) technique was used to decompose
the signals into various frequency levels and from the WT coefficients unique features were extracted. In the previous
studies,26,37 Gaouda et al presented the extraction of standard deviation–based features and root mean square (RMS)–
based features, respectively, from WT coefficients using MSD techniques. Similarly, the MSD technique was used on
Xiangxun et al38 for the recognition of short duration power quality disturbances. Xiangxun39 proposed a WT‐based
method to detect, localize, quantify, and classify the short‐duration PQ disturbances. Angrisani et al proposed a mea-
surement method based on CWT using its modulus local maxima properties and DWT using MSD for PQ detection
and classification. Gencer et al40 presented PQ (only voltage sag) detection technique based on DWT. In Dekhandji,41
TABLE 3 Key benefits and drawbacks of SPTs
MISHRA

SPTs Ref. Benefits Drawbacks


24
DFT It is one of the most extensively used computation algorithm. The PQ events are generally nonstationary‐type signals. Thus, it is unsuitable to
DFT is generally considered for the steady‐state analysis sense the immediate variations in PQ disturbances, for example, their initial
of the stationary signals. It is carried out by extracting and finale points. It is mostly suitable for the stationary PQ events.
spectrum at particular frequencies.
25,26
FFT The outcome of FFT is similar to that of DFT, but faster The signal information such as amplitude frequencies and phases cannot be
execution time. It is extensively applied for harmonic study acquired correctly. This is because of the leakage, aliasing effects, and picket
of power quality disturbances. fence formed by FFT.
27-30
STFT STFT is used to determine the frequency and phase information It is hard to analyse nonstationary signal by STFT. The main drawback of this SPT
of the waveforms as they vary over time. Using a moving is the limited time‐frequency resolution. Moreover, it cannot give exact time
window, the relation between the time and frequency and frequency info simultaneously to categorize the power quality events as
variation can be recognized. Ease implementation is the stated by IEEE‐1159 standard.
advantage of STFT.
32-36
WT In contrast to Fourier transform, the wavelet transform can Experiencing spectral leakage and picket fence effects makes this technique
offer time and frequency info of a signal simultaneously. insignificant. Moreover, the increase levels of signal decomposition can lead to
high computational burden. Furthermore, this method is highly affected by the
noisy environment.
50,51
Wavelet packet WPT can provide a time frequency representation of any It provides high degree of freedom but with complex data structure algorithms.
transform (WPT) nonstationary waveform deprived of missing any time or Moreover, the process of selecting appropriate filters decreases the shift
frequency‐related information. Moreover, the WPT sensitivity. In general, WPT performs better with respect to fidelity of direction
decomposition provides uniform output frequency bands, other than improved directionality.
which can be used for the identification of harmonic
spectrum in EPS by choosing an appropriate sampling
frequency and the wavelet packet decomposition levels.
93-97
GT It is one of the improved SPT used for accurate phasor The computational burden directly depends on the sampling frequency. Moreover,
estimation. Furthermore, it can provide both time and the Wigner distribution will cause critical cross‐interference which promotes
frequency information of signals in contrast to FT. Good incorrect frequency components.
time‐frequency resolution is also an advantage of this
technique.
28,69-92
ST The ST is the advancement of WT and STFT by its phase This technique does not suit well in practical application as the widths of the
correction and a variable window, respectively. Similar to frequency windows in the ST are directly related to their central frequency,
WT, it can offer an improved time and frequency which causes an inappropriate measurement of harmonics. Moreover, one
representation of a signal. cannot guess the fixed Gaussian window to apt for all kinds of signals.
Moreover, the fixed modulating sinusoids respecting the Furthermore, the processing time of this technique is more contrasted with
time‐axis plus the scalable and movable Gaussian window other SPTs.
properties of the ST can be used for better recognition of
PQ events.

(Continues)
9 of 42
10 of 42

TABLE 3 (Continued)

SPTs Ref. Benefits Drawbacks


98-101
KF KF is useful to estimate the variation of frequency, amplitude, The state observation is nonlinear; it could instigate the KF to deviate in
and harmonic content of a signal. some badly selected initial conditions. This type of instability may
result in erroneous assessment of fundamental and harmonic
components of a sinusoidal signal in noisy environment. This is because
of linearization and inaccurate parameters, expensive calculations of
Jacobian matrices, and the biased nature of approximates. Here, the
decomposition of signal is not possible in both time and frequency
domains.
102-116
HHT The HHT provides a novel technique of analysing The EMD method is limited in distinguishing different components in
nonstationary and nonlinear time series data. Generally, narrow band signals. The other issue related to HHT includes end effects
this technique is the combination of two techniques such of the EMD, spline problems, best IMF selection, and uniqueness.
as empirical mode decomposition (EMD) and Hilbert
transform (HT).
122-131
Variational mode In compared with EMD, VMD can determine the related The crucial drawback of the VMD is with boundary effects and sudden
decomposition frequency bands more adaptively and estimates the signal onset in general. VMD does not work “out of the box” with
(VMD) corresponding modes simultaneously. nonstationary signals, like long‐term EEG records, for which the spectral
VMD method clearly outperforms EMD for the application in bands of modes can vary extremely upon time, and will universally
tone detection and separation and noise robustness. overlay.
133-147
Mathematical MM is a nonlinear signal processing tool. Computational
morphology (MM) burden involved with this method is very less, since the
method uses simple arithmetic calculation and set theory.
MISHRA
MISHRA 11 of 42

Dekhandji proposed DWT to detect various PQ disturbances such as sag, swell, interruption, transient, and harmonics.
In Deokar and Waghmare,42 the authors proposed a hybrid DWT‐FFT‐based PQ event detection and classification
approach. The average energy entropy of squared detailed coefficients (AEESDCs) as feature vector was calculated after
processed by DWT for the detection of four basic single PQ disturbances, such as sag, swell, interruption, and har-
monics. Afterward, the FFT is used to further classify the disturbances which are the combination of these disturbances
or multiple disturbances. Liao et al,43 another hybrid signal processing–based feature extraction method (DWT‐FFT),
was reported. Here, the fuzzy expert system (FES) was used to make a decision using the extracted features. In
Angrisani et al,44 the authors proposed a transient detector and classifier using WT and neural network, which was
capable to detect and classify the transients simultaneously and automatically.
Kaewarsa et al45 suggested multiwavelet transform based on MSD techniques using artificial neural network (ANN)
architecture to detect and classify the PQ disturbances. Garousi et al46 presented a combination of DWT and parallel neu-
ral network structures for multiple PQ disturbance classification. A new variant of WT named as frequency slice WT
(FSWT) was implemented in Subbarao and Samundiswary47 for nonstationary PQ disturbances. FSWT was found to be
superior to the earlier WT to enhance the time‐frequency resolution of different power signal disturbance waveforms.
Biswal and Mishra48 studied a nonstationary PQ disturbance recognition technique based on modified FSWT (MFSWT).
The MFSWT is an advanced version of FSWT which offers frequency‐dependant resolution with added window parame-
ters for improved localisation of the spectral features. Karimi et al49 suggested an online method of power signal distur-
bance detection using WT.
Chung et al50 proposed a hybrid method comprising of a WPT‐based Hidden‐Markov model (HMM) and a rule‐based
computational technique for the classification of PQ disturbances. Here, the time‐characterised disturbances (such as sag
and interruption) were classified by the rule‐based method, and the frequency‐characterised feature‐based PQDs (such as
transients and impulses) were categorized by using WPT‐based HMM. A PQD&C technique based on WPT and multiclass
support vector machine (SVM) was presented by Zhang et al51 to classify PQ disturbances. Initially, the disturbance signals
were processed by WPT and decomposed into several components of different frequency sub‐bands. Then, the energy
related to the wavelet packet coefficients was extracted as a feature vector. Subsequently, a multiclass SVM trained and
tested to classify the PQDs. Considering the advantages of Tsallies entropy compared with Shannon entropy, WPT com-
pared with WT, and singular value decomposition, Liu et al suggested a hybrid approach combing all these three to detect
the PQ disturbances.52 The automatic PQ monitoring system based on wavelet packet decomposition (WPD) was pre-
sented by Bhuiyan et al53 for the detection of five different types of disturbances in noisy environment. The suggested tech-
nique uses the interscale and intrascale relationship of the WPD coefficients for the recognition of the PQ disturbances
upon noise elimination. Ece and Gerek54 proposed a two‐dimensional discrete wavelet transform (2DDWT) method for
automatic recognition of PQ disturbances. Shareef et al55 proposed a new approach to visualize and detect the PQ distur-
bances using image processing technique (IPT) like greyscale images and binary images. Similarly, a two‐dimensional dis-
crete wavelet transform–based IPT was proposed by Krishna and Kaliaperumal56 for classification of PQ signals.
An effective smart PQ detection technique comparing DWT for feature extraction, k‐means apriori (KMA) algorithm
for optimal feature selection, and least square SVM (LS‐SVM) for classification process was given in Erişti.57 He and
Starzyk58 presented a new PQD&C using WT‐based MRA technique for feature extraction and self‐organizing learning
array for classification purpose. Chakraborty et al59 proposed a PQD&C technique where a variant of DWT named dual‐
tree complex WT was used for feature extraction and sparse representation for PQ disturbance classification. Zafar and
Morsi60 presented the undecimated WT to calculate feature indexes using complex wavelet coefficients. In Uyar et al,61
authors presented a wavelet norm entropy‐based efficient feature extraction technique for PQD&C problem. The distur-
bance classification scheme was achieved through wavelet neural network (WNN). The PQ detection scheme based on
WT in the field programmable gate array (FPGA) environment was proposed in Eristi.62 Classification of short duration
voltage variations utilizing WT‐based entropy criteria was presented in Priyadarshini and Sushama.63 Authors in Ray64
developed a technique based on DWT that can be able to detect and classify PQ disturbances in wind‐solar hybrid sys-
tem. Ucar et al65 proposed a PQ disturbance detection and classification technique using DWT‐based extreme machine
learning (ELM) technique. In Upadhyaya,66 maximum overlapping discrete wavelet transform (MODWT) was applied
along with the traditional DWT technique for the detection and localization of various classes of PQ disturbances. In
Gursoy et al,67 the PQD&C method was developed using empirical wavelet transform (EWT) and DWT methods.
Thirumala et al68 presented a novel tunable‐Q WT (TQWT)–based feature extraction method for automatic PQ event
detection and classification. The TQWT has the capability to regulate the Q factor (Q) and redundancy (r) of the wavelet
for extraction of a specific frequency component. Afterward, the multiclass SVM was used for classification of PQ
disturbances.
12 of 42 MISHRA

3.1.3 | Stockwell transform and its variants' based methods

The Stockwell transform (ST) is a SP tool that comprises the properties of both WT and STFT. It offers an
improved time and frequency representation of a signal. It exclusively combines a frequency‐dependent resolution
that at the same time localizes the real and imaginary spectra. The fixed modulating sinusoids respecting the
time‐axis plus the scalable and movable Gaussian window properties of the ST can be used for better recognition of PQ
events.
Considering the excellent time‐frequency resolution properties of ST, it was used initially for the analysis of
power quality disturbances in the year of the early twentieth century.28,69 Lee and Dash28 proposed a ST and neural
network–based PQD&C approach. Zhao and Yang70 and Mishra et al71 proposed ST as a simple and effective tool
for PQD&C and showed that ST can perform efficiently even under extremely noisy environment. The multiresolution
ST and fuzzy logic–based recognition system were used for optimal feature extraction and classification (FE&C) of
power quality events, respectively, in Chilukuri and Dash.72 Biswal et al73 proposed a PQD&C technique using
hyperbolic ST (HST) for feature extraction and genetic algorithm–based fuzzy C‐means (FCM) for automatic
recognition of events. The HST algorithm is one of the variants of ST where the pseudo‐Gaussian hyperbolic window
is used in place of a simple Gaussian window to provide improved time‐frequency resolution in low and high frequency
compared with ST. Nguyen and Liao74 proposed a hybrid approach comprising windowed DFT and ST for the feature
extraction and binary feature matrix technique for the automatic recognition of PQ events. The ST with module
time‐frequency matrix using maximum similarity principle was suggested by Xiao et al75 as a PQD&C approach. The
MRA‐based ST and rule‐based classifier was employed for optimal feature extraction and classification of PQ events,
respectively, in Rodríguez et al.76
Another variant of ST named as discrete S‐transform (FDST) was offered in Biswal and Dash77 to correctly extract the
time‐localized spectral features from the PQD signals. FDST with automatic scaling for correct detection of PQ distur-
bances and energy metering was digitally implemented in Jaiswal and Ballal.78 Here, a comparison of FDST‐based
energy metering process with existing algorithms like FFT and filter‐based design was carried out. The outcomes prove
that FDST‐based energy metering scheme performs outstanding compared with other recent algorithms in terms of
accuracy, adaptability, and complexity under PQ events. He et al79 suggested a hybrid PQD&C technique based on
ST and dynamics. To measure the effectiveness of the proposed technique, the classification accuracy and run time were
considered as the performance index. In this hybrid method, Dyn was initially used to find the position of the signal
components in the frequency spectrum output of FT and subsequently applied the inverse FT to a few signal compo-
nents. Afterward, several unique features from FT, ST, and dynamics were extracted, and a DT was operated for the
automated recognition of PQ disturbances.
Reddy et al80 offered a discrete orthogonal representation of ST (DOST) for PQD&C. The simulation result correspond-
ing to the stated PQD&C scheme was compared with other SP techniques, such as STFT, DWT, and ST, and proved that
the DOST outperforms these techniques with respect to the detection accuracy. Behera et al81 recommended ST and
hybrid particle swarm optimization‐fuzzy logic for power quality time series data mining. A novel PQD&C technique com-
prising ST and logistic model tree techniques was proposed by Moravej et al.82 A fast ST with modified Gaussian window
was utilized for the purpose of feature extraction from power quality disturbances signal in Biswal et al.83 Here, a combi-
nation of two optimization algorithms such as chemotactic differential evolution algorithm and bacterial foraging optimi-
zation algorithm was employed to increase the classification accuracy. Hasheminejad et al84 suggested a hybrid technique
using ST and the hidden Markov model for PQ event recognition. The HMM is strong and effective too on SP applications,
where it calculates the maximum likelihood probability among training and testing data signals for recognition. The ST
was combined with an ELM technique for automatic nonstationary power quality disturbance recognition in Erişti et al.85
Biswal and Dash86 proposed a new variant of ST (fast dyadic ST algorithm) and fuzzy DT–based PQD&C
technique. Here, the fast dyadic ST algorithm, DT, and fuzzy logic technique were used for feature extraction, for feature
selection, and for PQ disturbance classification, respectively. Similarly, Babu et al87 suggested an FDST and DT‐based
PQD&C technique. The classification of disturbances was carried out both noiseless and noisy condition. Kumar et al88
presented a ST with rule‐based DT and ANN for the classification of PQD signals. Li et al89 proposed a new PQ disturbance
detection technique using double‐resolution ST (DRST) and directed acyclic graph support vector machines (DAG‐SVMs).
DRST is a variant of ST. It has several advantages over ST such as improved time‐frequency resolution and greater robust-
ness and less computational burden deprived of mislaying the useful info of the original waveform. Eka et al90 proposed a
modified ST (MST) technique by modifying the dilation factor in Gaussian window function and applied to identify the
nonstationary power signals. Sahu and Choubey91 proposed a hybrid technique using MST and FCM‐based DE algorithm
MISHRA 13 of 42

for the automatic recognition of several nonstationary PQDs. A technique using ST was stated in Shaik and Mahela92 for
PQ assessment, identification of islanding and outage, and grid synchronization of NCES.

3.1.4 | Gabor transform and its variants' based methods

The GT93 is an innovative SP tool employed for perfect phasor estimation. Compared with Fourier transform, the GT
offers better time and frequency info of a studied signal. It maps the time series data into both time and frequency
domains. A Gaussian window function incorporated with GT algorithm was employed to monitor power system tran-
sient harmonic variations in the time domain signal.94 Kawady et al95 proposed a hybrid method comprising GT and
ANN for the detection and classification of the arcing faults. Here, the GT was used as a SP tool for optimum feature
extraction. Gabor‐Wigner transform (GWT) is a hybrid SP tool combining GT and the Wigner distribution function
(WDF), and thus, it is more advantageous than both the individual techniques because it has less cross‐term difficulties
than the winger function and greater precision than the GT. Cho et al96 used this GWT as a time‐frequency–based fea-
ture extraction technique to detect the PQDs. Naderian et al97 developed a novel approach for recognition of PQDs using
discrete GT with a finite impulse response window (FIR‐DGT) and type 2 fuzzy kernel‐based SVM (T2FK‐SVM). The
FIR‐DGT and T2FK‐SVM techniques were used for feature extraction and classification, respectively.

3.1.5 | Kalman filter and its variants' based methods

The KF is a renowned SP tool, which is generally applied for the estimation of magnitude, phase angle, and frequency of
the noisy harmonic signal.98 In Abdelsalam et al,99 authors suggested a PQD&C method by combining of DWT and KF
with FES. Here, the WT was employed to detect the presence of noise in the extracted voltage signal, and KF was uti-
lized to accelerate its degree of convergence. Then, the output of the KF was feed as input to FES for the automatic PQD
recognition. A hybrid method comprising of S‐transform and extended KF is presented in Dash and Chilukuri100 for the
recognition of power quality events. The S‐transform was employed to identify and locate the waveforms while the
extended KF method was used for the estimation of the changes in amplitude, frequency, phase, and harmonic contents
of the distorted PQ waveform. Similarly, in Reddy et al,101 a hybrid approach using a new variant of KF named as
unscented KF and improved particle swarm optimization (PSO) algorithms was presented for sensing the magnitude,
phase, frequency, and harmonic contents of PQDs infected thru noise having of SNR value.

3.1.6 | Hilbert‐Huang transform and its variants' based methods

The HHT method has been developed in the recent year to study nonstationary PQ disturbances. Generally, this tech-
nique is the amalgamation of two techniques such as empirical mode decomposition and Hilbert transform.102 The
empirical mode decomposition (EMD) is a technique in which the input signal is decomposed into a series of small
components named as intrinsic mode functions (IMFs).103,104 Afterward, the IMFs are processed through Hilbert trans-
form (HT) for extracting various features related to frequency, amplitude, and phase of the signal.
Shukla et al105 proposed a PQD&C technique based on EMD and HT. The proposed HHT algorithm was compared
with ST‐based technique to prove the supremacy of the technique in sensing power quality disturbance events like
notches and flickers. Moreover, a probabilistic neural network algorithm was executed to categorise the PQDs. Norman
et al106 proposed frequency‐shifting wavelet decomposition via HT algorithm for PQ event detection. By this approach,
the spectra leakage problem can be completely diminished in the DWPT and employed for estimating power indices and
detection of flickers. Kumar et al107 proposed a PQD&C algorithm based on HHT and probabilistic neural network
(PNN) for single and multiple PQ disturbances. Yang et al108 and Manjula et al109 suggested a technique based on
EMD and HT to detect the causes of voltage sag. A PNN classifier was designed based on the features extracted through
HHT to classify the category of sag causes.109 Three categories of voltage sag causes were considered, such as short cir-
cuit fault, starting of the induction motor, and three‐phase transformer energization. Ozgonenel et al110 proposed a
PQD&C technique based on ensemble EMD (EEMD) and SVM classifier. The performance of EEMD technique was
compared by EMD technique for PQ disturbance detection. Authors111 presented a hybrid technique based on EMD,
HT, and fuzzy product aggregation reasoning rule (FPARR) to detect and classify the PQDs. Initially, the EMD tech-
nique was applied to remove the unwanted noise from the original extracted waveform, and then, the denoised signals
14 of 42 MISHRA

were processed through HT to extract features. The FPARR method was used to classify the PQ disturbances. Saxena
et al112 proposed an efficient technique to classify composite PQ disturbances using HHT. A comparative study between
PNN and SVM was carried out based on the classification accuracy.
HHT can analyse nonstationary signals with excellent time resolution. However, it may face trouble in determining
signals comprising components with closer frequency. To overwhelm the issue with components comprising closer fre-
quency, an innovative variant of HHT named as iterative HHT was presented by Afroni et al.113 Moreover, to over-
whelm the issue in detecting the spot of the transient occurrence in the signal, an innovative technique based on
symbolic aggregate approximation (SAX) scheme was presented. SAX converts the original waveform into symbols that
can be employed by a PRT to detect the boundaries of the stationary signals inside a nonstationary signal.114 Hafiz
et al115 proposed a PQD&C method using HHT and relevance vector machine. Zhan et al116 proposed HHT for detection
of multiple‐disturbance complex PQ signal. In Huang et al,117 authors proposed orthogonal HHT for the analysis of non-
linear and nonstationary PQ signals like voltage flickers. Huang et al suggested a SP technique using mathematical mor-
phology (MM) and HHT for PQ disturbance detection and analysis. In Senroy et al,118 the authors presented the HHT
technique with adjustment for time‐frequency study of distorted PQ events. Jayasree et al119 proposed a PQD&C tech-
nique based on HHT for feature extraction and radial basis function NNs (RBFNN) for calcification of PQ disturbances.
Aiming at the end effect and the mode mixing of original HHT, an improved HHT with adaptive waveform matching
extension was proposed in Peng et al.120 Both simulations and real‐time experimental verification were carried out for
the application of the proposed improved HHT for PQ disturbance detection in microgrid. Feilat et al121 presented a
PQD&C method based on envelope construction using HT and classification using ANN.

3.1.7 | Variational mode decomposition–based methods

Variational mode decomposition (VMD) is a SP tool that decomposes a signal into a collection of band‐limited IMFs.
The VMD has several advantages over EMD: recursive variational decomposition (RVD) and empirical wavelet trans-
form (EWT).122 In recent years, the VMD method was highly used for the study of power quality signals.123-131 In
Aneesh et al,123 the authors presented a comparison between the VMD and EWT‐based feature extraction technique
for PQD&C using SVM. The obtained results prove the superiority of VMD‐based feature extraction over EWT. Abdoos
et al124 proposed a hybrid SP method comprising VMD and ST for feature extraction and SVM to classify the power
quality events. Achlerkar et al125 presented VMD and DT‐based method for PQD&C in microgrid environment. To
enhance the detection rate and reduce the computational complexity of VMD‐based approach, the recursive FFT was
implemented with existing VMD by Soman et al126 for power signal analysis. Viswanath et al127 proposed VMD‐based
feature extraction technique for the detection of impulsive transients or spikes. Sahani and Dash128 presented VMD and
weighted online sequential ELM (WOSELM) for PQD&C in real‐time environment. Similarly, in Sahani and Dash,129
the authors used VMD and fast‐reduced kernal ELM for PQD&C in real‐time environment. Aneesh et al130 proposed
VMD as a feature extraction method for PQDs. The classification of PQ disturbances was carried out by SVM and ran-
dom kitchen sink (RKS) algorithm. Application suitability of VMD‐based method to discriminate both stationary and
nonstationary PQDs, such as harmonics, interharmonics, flicker, transient, and noise, was explored in Achlerkar
et al.125 Several distinctive features viz., mode central frequencies, relative mode energy ratios, instantaneous amplitude
(IA), and zero crossings were obtained for classification of both single and multiple PQ disturbances using a DT set of
rules. Chen and Cai131 presented VMD‐based harmonic detection for the application of active power filter.

3.1.8 | Mathematical morphology–based methods

MM is a nonlinear SP tool that modifies the shape of a signal.132 It was initially presented by Matheron133 and Serra.134
This technique is based on set theory and integral geometry. Unlike FT or WT, which obtains frequency info in signals,
MM operates mainly in time domain. The MM‐based feature extraction approaches are adopted by various authors to
identify and classify the stationary/nonstationary PQ disturbances in the previous studies.135-147
Lu et al135 presented a morphological transform technique to extract the features corresponding to the PQ disturbances
in time domain. Moreover, it also highlights the importance of the proposed technique to detect the location and duration
of disturbances. In Lu et al,135 the concepts of MM were applied to form an analyser, which was further applied to extract
significant features from the nonstationary PQ signals. Matz et al136 presented a digital filter and MM technique to detect
and classify the transient and waveform distortion. Using MM and complex WT, a new technique on PQD detection and
MISHRA 15 of 42

location was described in Xue et al.137 Sen and Zhen138 proposed an improved MM technique for PQD detection. Based on
MM and grille fractal, a new method on PQD detection and location was presented in Li et al.139 In Huang et al,140 the
authors presented a technique using MM and HHT, which was operated for the detection and analysis of PQDs. To suc-
cessfully cover up noises, the conventional Maragos MM filter was upgraded; therefore, the key features of input signal
were retained. Then, the denoised PQ signal was processed through HHT for optimal feature extraction. Similarly, the
MM and HHT techniques were adopted by Chen et al141 for PQD detection and classification application. In Zhang
et al,142 the authors proposed a new PQD detection approach using MM and radial coordinate visualization (Radviz).
Huang et al143 presented a voltage sag detection approach using a morphology filter and grille fractal. Chakravorti and
Dash144 presented MM and fuzzy DT–based technique for automatic PQD&C. Chen et al145 presented a morphology sin-
gular entropy–based PQ detection method. The MSE comprises three methods like MM, singular value decomposition
(SVD), and entropy theory. In Saputra et al,146 a novel approach in PQ disturbance identification was suggested employing
half multiresolution morphology gradient (HMMG). The abovementioned approach was based on multiresolution mor-
phology gradient, which is capable to find out the ascending and descending edges of the waveform. Igna et al147 proposed
a novel strategy of PQD&C using a vector matrix conversion with MM on a small‐window size.

3.1.9 | Miscellaneous SP‐based methods

In addition to the aforementioned techniques (Sections 3.1.1‐3.1.8), there are few other SP tools extensively employed
for the feature extraction process in power quality study.
Wang et al148,149 recommended a time‐frequency representation (TFR)–based feature extraction method via time‐
frequency ambiguity plane with kernel technique for PQD&C. Hu et al150 proposed Chirplet transforms (CTs) for PQ dis-
turbance detection. CT is the generalized representation of FT, STFT, and WT. In Janik and Lobos,151 the authors pro-
posed a novel technique of PQ disturbance detection using space phasors applied to three‐phase voltage signals.
Manikandan et al152 presented a novel PQD&C technique for single and multiple PQDs using sparse signal decomposition
(SSD). Hsieh et al153 proposed Slantlet transform for PQ disturbance detection. Moreover, some other SP techniques
applied for the PQ disturbance detection and classification are time‐time transform,154 Teager energy operator,155 princi-
ple curves (PCs),155 spectral kurtosis (SK),156 amplitude and frequency demodulation (AFD) with FPARR classifier,157
adaptive local iterative filter decomposition (ALIFD),158 delay small angle dq transform,159 generalized morphological fil-
ter, smart meter,160 PLL,161 ensemble EMD (EEMD),162 morphological lifting wavelet (AMLW),163 pencil matrix
method,164 and information theory.165

3.2 | Feature selection techniques for PQ disturbance monitoring

In pattern recognition task, the input vector contains a number of distinctive features. However, each feature has a unique
characteristic and also maintains a complex relationship among each other. Sometimes, an individual feature may show a
decent performance, but when operated simultaneously in a group may behave redundantly (or vice versa). Thus, the
selection of most essential features having diversified properties is a must. Mostly, the FS tasks are executed by the help
of optimization algorithms. In literature, FS procedures have been widely employed to extract the highly distinctive fea-
tures and get rid of the unwanted features. Some of the recent FS methods used for PQD&C application are genetic algo-
rithms (GAs),166-169 extension GA (EGA),170 genetic k‐means algorithm (GKA),171 PSO,172,173 Gram‐Schmidt orthogonal
transform for FS,174 mutual information (M) and relief (R),175 ant colony optimization (ACO),176 and combination of three
algorithms like GA, maximum relevance minimum redundancy (mRMR), and sequential forward selection (SFS).177

3.3 | Intelligent classifier for automatic classification of PQDs

Intelligent classifiers are the tools based on artificial intelligence (AI) technique, generally used for automatic classifica-
tion and/or decision‐making purpose. AI can be broadly defined as the automation of the actions related to human
thinking like learning, perceptions, reasoning, problem solving, and decisionmaking. In the recent years, several intel-
ligent classifiers have been used for automatic classification of PQ disturbances by various researchers such as ANN,
SVM, fuzzy logic, extreme learning machine, and k‐nearest neighbour.
16 of 42 MISHRA

3.3.1 | ANN‐based approach

In literature, the most used intelligent classifier for the application of PQD classification is ANN. The artificial neural
networks are very good for various applications like pattern recognition, classification, function approximation, optimi-
zation, and data clustering.61 Several variants of ANN are used for the classification of power system faults, for example,
multilayer perceptron (MLP), radial basis function (RBF), and probabilistic neural networks (PNN). The MLP is a class
of feedforward ANN that generates a set of output from a set of input. MLP is highly recognized for their learning and
classification skill where the signals cannot be expressed mathematically. Conversely, MLP has various issues like defin-
ing an appropriate architecture (number of hidden layers and nodes), time‐consuming, and extremely slow. A RBF is a
category of single hidden layer feedforward ANN. The RBF‐based NN shows similar properties as backpropagation net-
works such as generalisation ability and robustness. Moreover, it has some extra advantages like quick learning and the
capability to identify outliers during estimation.94 A PNN is a modified version of a radial basis function networks. It
comprises of a radial basis layer and a competitive layer. In Lee and Nam,178 the authors proposed a hybrid PQD&C
approach based on FFT, DWT, and MLP neural network (MLP NN). Similarly, Monedero et al179 presented a WT‐based
MLP NN with three hidden layers for the real‐time detection and classification of PQ waveforms. Uyar et al180 imple-
mented MLP NN on ST‐based input feature vector to classify PQ disturbances. Ghosh and Lubkeman181 applied
feedforward NN (FFNN) and time delay NN (TDNN), for the automatic PQ disturbance recognition.
He et al182 suggested a quantum NN (QNN)–based PRT using Dempster‐Shafer evidence (DSE) theory to classify the
PQ disturbance signals. Similarly, authors183 implemented learning vector quantization (LVQ) NN–based PRT using
DSE theory to classify the PQ disturbance signals. The spectral kurtosis and RBF NN were implemented in Liu
et al184 to detect and classify the transient PQDs.
Valtierra‐Rodriguez et al185 offered a twofold NN‐based procedure comprising of adaptive linear network
(ADALINE) and FFNN for the detection and classification of PQDs. The ADALINE was operated for the estimation
of the harmonic and interharmonic component, which helps to determine the RMS voltage and total harmonic distor-
tion. These parameters were further used to classify the PQDs such as sags, swells, outages, harmonics, and
interharmonics. Similarly, a FFNN was employed using the horizontal and vertical histograms of a specific voltage sig-
nal to categorize the rest of the PQ disturbances such as spikes, oscillatory transients, flicker, and notching,
Gaing186 suggested a DWT‐based PNN classifier to classify power disturbances such as momentary interruption, volt-
age sag/swell, harmonic distortion, capacitor switching, and flicker. The PNN classifier was integrated with ST to classify
eight types of PQ disturbances with two types of complex PQDs in Huang et al.187 An amalgamation of fully informed par-
ticle swarm (FIPS) and adaptive probabilistic neural network method called as PNN‐based FS was suggested in Lee and
Shen188 to classify the power quality waveforms. The input feature vector was extracted by ST and TT transform for train-
ing the PNN classifier. A hybrid method (dynamic wavelet network) based on WT and adaptive PNN was adopted by
Biswal et al189 for the classification of nonstationary PQ disturbances in a dynamic condition. The dynamic wavelet net-
work has the ability to reduce error by automatic tuning of learning cycles for unique kinds of power signals.
Bhende et al71 presented ST‐based modular NN to classify PQ disturbances. In Naik and Kundu,190 authors presented a
three‐module NN structure followed by a rule‐based classifier to detect PQ disturbances which occurred simultaneously.
Oleskovicz et al192 proposed a hybrid approach comprising of WT with MLP and RBF NNs to identify, locate, and classify
PQDs automatically. Dash et al191 applied three variants of NN such as PNN, FFNN, and RBF on two SP techniques (ST
and DWT) based input feature vector to classify the combination of single and multiple power quality events. Khadse
et al193 proposed conjugate gradient BPN‐based ANN for real‐time PQ assessment. A combination of ST algorithm and
NNs was used to detect and differentiate the PQ disturbances in Elango and Loganathan.194 Wang et al195 proposed an
analog‐digital mixing network (ADMN) for automatic data extraction and classification of PQ disturbances. In
Muthusamy and Ramanathan,196 the authors proposed a novel detection and classification method for PQ using least
mean square (LMS) and NN.

3.3.2 | SVM‐based approach

SVM was initially presented by Vapnik.197 It is a supervised machine learning tool used for pattern recognition and
classification. This PRT is generally used the statistical learning concept. The successful application of SVM includes
various areas such as dependency estimation, forecasting, fault classification in power system, and constructing intelli-
gent machines. Handing a large dimensional input vector efficiently is the main advantage of SVM over other
MISHRA 17 of 42

conventional tools on large classification problems. Moreover, it has superior generalization properties in comparison
with the traditional classifiers. SVM was basically designed for binary classification, retaining the two classes of value
1 and −1 only. But in real‐time environment, it is very essential to classify the events having more than two classes.
Two different approaches are usually considered for the multiclass classification such as one against all (OAA) and
one against one (OAO).
In De Yong et al,198 the authors proposed WT‐based multiclass SVM to classify the PQ disturbances. Here, the OAO
approach was considered for multiclass SVM. In Janik and Lobos,151 authors proposed a novel technique of PQ distur-
bance detection using space phasor technique applied to three‐phase voltage signals for feature extraction and SVM as
classification technique. Lin et al199 implemented a combination of linear SVM and the disturbances versus normal
(DVN) approach to develop the multiclass SVM which was able to classify the complex PQ disturbances.
Eris and Demir200 presented a new automatic classification technique based on wavelet and SVM for the classifica-
tion of the three‐phase PQD signals. Hu et al201 proposed an automatic PQD&C technique using wavelet packet energy
entropy and weighted SVM. Moreover, several other techniques based on WT and SVM classifier for automated detec-
tion and classifications were presented in the literature.50,57,202-205 A PQD&C technique based on WPT and multiclass
SVM was presented by Zhang et al60 to classify PQ disturbances. Biswal et al206 applied a kernel‐based SVM to TT
transform–based feature vector to classify the PQ disturbances. An effective smart PQ detection technique comparing
DWT for feature extraction, KAM algorithm for optimal FS, and LS‐SVM for classification process was given in Erişti.57
Kapoor et al207 proposed a novel technique to differentiate PQ disturbances using the histogram of oriented gradients
(HOG) and SVM.

3.3.3 | Fuzzy expert system–based approach

FL scheme simplifies the standard binary logic for reasoning under ambiguity. The motivation behind the concept of FL
technique derives from perceiving human thinking to apply ideas and information. A fuzzy set is a function that maps
objects in the domain of concern to their membership values in the set. This function can be described as membership
function. An expert system using fuzzy set and fuzzy rule for the interpretation of data can be called as fuzzy expert
system.
The multiresolution ST and fuzzy logic–based classification system were employed for optimal feature extraction and
classification of PQDs, respectively, in the previous studies.72,81 Liao et al43 presented a hybrid SP‐based feature extrac-
tion method (DWT‐FFT) and FES as classification technique for PQD classification. A hybrid approach based on Fou-
rier linear combiner and a FES for the classification of transient PQ disturbances in a EPS was presented in Dash
et al.208 A WT‐based extended fuzzy reasoning scheme for PQD&C was proposed in Xiangxun.38 The FPARR classifier
was adopted for classification of PQ events in the previous studies,157,209 where AFD technique and DWT technique
were used as feature extraction techniques, respectively. The FPARR‐based PRT has the advantage of higher classifica-
tion accuracy because of its learning and generalization abilities. A fast ST with modified Gaussian window and fuzzy‐
based decision tree were used for feature extraction and classification of PQDs signal in the previous studies.83,85 A
hybrid PQD&C method based on the DWT, KF, and FES were presented in Rilling et al.103 The PQD&C based on a
fuzzy ARTMAP wavelet network were presented in Decanini et al.210 An approach using ST and FCM clustering was
proposed by Mahela and Shaik211 for PQD&C.

3.3.4 | Neuro‐fuzzy system–based approach

The authors212 proposed a neural‐fuzzy PRT using LVQ architecture and fuzzy associative memory (FAM). LVQ is very
well capable in pattern recognition, and FAM is highly flexible to handle uncertainties.
Reaz et al213 presented WT‐based feature extraction and univariate randomly optimized neural network with fuzzy
logic–based PRT to classify the PQ events. Liao and Yang214 presented a novel PQ disturbance recognition approach
using a noise suppression algorithm, WT‐based feature extraction algorithm, and neuro‐fuzzy classifier. The proposed
technique was tested on both noise ridding simulated PQ data and field data. Pires et al215 suggested a new PQD&C
approach using an automatic four‐step algorithm. In the initial step, the extracted voltage signals were represented
in a 3D space referential. The second and third steps correspond to the application of principal component
analysis (PCA) and extraction of features from obtained eigenvalues. Lastly, the extracted features were processed by
a neuro‐fuzzy–based PRT for automatic classification of PQ events.
18 of 42 MISHRA

3.3.5 | ELM‐based approach

Generally, the learning rate of FFNNs is much slower than necessity. The main causes behind it can be because of the
following: (1) the slow gradient‐based learning algorithms are widely employed to train NNs, and (2) all the required
parameters of the NNs are tuned iteratively by using such learning algorithms. The ELM was initially developed by
Huang et al.216 It is a single hidden layer feedforward neural network (SLFNs) which randomly selects hidden nodes
and systematically calculates the output weights of SLFNs. A noniterative solution of ELMs gives a speedup of five
and six orders of magnitude compared with MLP and SVM, respectively.
Ucar et al65 proposed a PQ disturbance detection and classification technique using DWT‐based ELM. The ST was
combined with an ELM technique for automatic nonstationary power quality disturbance recognition in Babu et al.87
Sahani and Dash128 presented VMD and an afresh developed WOSELM for PQD&C in real‐time environment. Simi-
larly, the authors129 used VMD and fast‐reduced kernal ELM for PQD&C in real‐time environment. In Sahani and
Dash,217 a combination of HHT technique and weighted bidirectional ELM (WBELM) classifier was used for detection
and classification of PQDs in real world. In Ahila et al,218 authors proposed hybrid method integrating MRA‐based WT
and ELM for PQ disturbance detection and classification.

3.3.6 | Deep learning–based approach

In recent years, researches on deep learning (DL) algorithms have gained a huge attention from the researchers. It has
been effectively used in various diverse research fields, such as speech recognition, human face recognition, computer
vision, signal, image, and information processing. The DL algorithms have the in‐built capability to learn optimal fea-
tures from original input signal automatically. Therefore, the time consumed by feature extraction process in other fea-
tures engineering has been avoided in this technique. The application of DL algorithms on automatic power system
fault detection and classification approaches was carried out by various authors in the previous studies.219-223
Ma et al219 proposed stacked autoencoder (SAE) as a DL architecture to extract advanced features from PQ distur-
bance signals for automatic PQD&C. Moreover, variances of signals and a PSO algorithm were employed to support
PQD&C process. Liu et al220 presented a new PQD&C method based on singular spectrum analysis (SSA), curvelet
transform (CT), and deep convolutional neural networks (DCNNs). The competency of DL algorithms to detect and
classify the PQ events was studied by Mohan et al.223 To know the usefulness of different DL tools, several architectures
viz. convolution neural network (CNN), recurrent neural network (RNN), identity RNN (I‐RNN), long short‐term mem-
ory (LSTM), gated recurrent units (GRU), and CNN‐LSTM were considered in this article. Balouji et al222 presented a
DL‐based approach for automatic voltage sag recognition. The technique includes a powerful LSTM architecture, which
is a variant of RNNs.

3.3.7 | Miscellaneous PRT–based approach

In addition to the aforementioned techniques (Sections 3.3.1‐3.3.6), there are few other PRTs extensively utilized for the
automatic recognition of the PQDs. Such PRTs for PQ disturbance classification are binary feature matrix (BFM),74 Wai-
kato Environment for Knowledge Analysis (WEKA),160 support vector data description (SVDD),224 k‐nearest neighbour
(KNN) algorithm,225 hidden Markov models (HHMs),50,66,226 random forest (RF) classifier,207 and group data handling.227

4 | DISC USS I ON

In Table 4, the existing PQD&C techniques are compared using several criteria, such as type of data used (synthetic or
real world), type of PQ disturbance studied (single or multiple), and performance in terms of accuracy (both noiseless
and noisy environment). Unluckily, some literature did not make available the info needed to extract all of the criteria.
Specific remark could be presented from this table as follows: The methods that use the synthetic data, in general, dem-
onstrate higher reliability compared with approaches based on real‐time PQ disturbances. This shows the necessity for
more practical analysis, so that the outcomes from real‐time and synthetic data‐based techniques become more compat-
ible. This analysis indicates that numerous researchers have employed PQD&C techniques only for single power quality
events; therefore, the study on multiple power quality disturbances needs a large amount of attention.
TABLE 4 Comparative review of various existing PQD&C methods
MISHRA

Type of Accuracy (%)


Feature Feature PQ Data Number of
With Noise
Extraction Classification Selection (Single/ Disturbances Without
Ref. Technique Technique Technique Data Multiple) Studied Noise 30 dB 20 dB Year Article
26
FFT × × Simulation Single × × × × 1999 Journal
30
STFT NA × Simulation Single × × × × 2000 Journal
43
FT+WT FES × Synthetic Single 9 99 × × 2004 Journal
43
FT+WT NN × Synthetic Single 9 97 × × 2004 Journal
42
DWT+FFT Threshold × Synthetic Single and 10 90.043 × × 2014 Journal
multiple
173
FT+WT Fuzzy system PSO Synthetic Single 8 98.5 93.625 2010 Journal
40
DWT Threshold × Simulation Single 1 100 × × 2010 Journal
59
DWT SOLAR × Synthetic Single and 7 94.93 92.2 × 2006 Journal
multiple
218
DWT ELM PSO Synthetic Single and 10 97.60 × × 2015 Journal
multiple
46
DWT NN × Synthetic Single 18 × 99.56 2016 Journal
65
DWT ELM × Synthetic Single and 06 100 96.22 × 2018 Journal
multiple
67
DWT ANFIS × Synthetic Single 6 100 × × 2018 Journal
67
EWT ANN × Synthetic Single 6 99.16 × × 2018 Journal
67
DWT NN × Synthetic Single and 7 95.71 × 89.92 2008 Journal
multiple
192
DWT RBF × Real time Single 5 99.51 × × 2009 Journal
186
DWT PNN × Synthetic Single 5 90 × × 2004 Journal
202
DWT SVM × Synthetic Single and 5 100 98 95.6 2009 Journal
multiple
204
DWT SVM × Synthetic Single and 09 98.89 97 96.33 2010 Journal
multiple
209
DWT Fuzzy rule × Synthetic Single and 8 98.75 × 90.61 2010 Journal
multiple
210
DWT Fuzzy‐ARTMAP FANN Synthetic Single and 08 99.66 × × 2011 Journal
neural network multiple
210
DWT HHM × Synthetic 14 99.46 × 93.83 2013 Journal
19 of 42

(Continues)
TABLE 4 (Continued)
20 of 42

Type of Accuracy (%)


Feature Feature PQ Data Number of
With Noise
Extraction Classification Selection (Single/ Disturbances Without
Ref. Technique Technique Technique Data Multiple) Studied Noise 30 dB 20 dB Year Article
Single and
multiple
172
WT RBF GKA Synthetic Single and 5 × × 93.89 2010 Journal
multiple
230
DWT Multirelevance × Synthetic Single and 09 99.03 × × 2011 Journal
vector machine multiple
170
DWT EGA Synthetic Single 06 96 2011 Journal
55
2‐D‐DWT × × Synthetic Single 8 97.33 × × 2012 Journal
45
DGHM NN × Synthetic Single 5 98.03 × × 2007 Journal
multiwavelet
48
MFSWT PNN × Synthetic Single 8 97.7 66.57 92.45 2013 Journal
60
DTCWT SRC Simulation Single 4 100 × × 2014 Journal
51
WPT SVM × Synthetic Single 8 97.7 93.4 × 2002 Journal
180
WPT QNN Synthetic Single and 10 96 × × 2012 Journal
multiple
169
WPT SVM GA‐FkNN Synthetic Single and 08 96.25 × × 2012 Journal
and real multiple
time
206
TT SVM × Synthetic Single and 10 99.9 × × 2013 Journal
multiple
176
TT DT PSO Synthetic Single and 09 91.22 × × 2011 Journal
multiple
176
TT DT ACO Synthetic Single and 09 95.97 × × 2011 Journal
multiple
28
ST PNN × Synthetic Single and 10 95.33 94 × 2003 Journal
multiple
180
ST NN × Synthetic Single and 09 99.67 99.11 94 2009 Journal
multiple
79
ST DT × Real time Single and 11 97.91 94.36 2013 Journal
multiple

(Continues)
MISHRA
TABLE 4 (Continued)
MISHRA

Type of Accuracy (%)


Feature Feature PQ Data Number of
With Noise
Extraction Classification Selection (Single/ Disturbances Without
Ref. Technique Technique Technique Data Multiple) Studied Noise 30 dB 20 dB Year Article
180
ST QNN × Synthetic Single and 10 96.6 × × 2012 Journal
multiple
171
ST Fuzzy C‐means APSO Synthetic Single and 09 96.33 × × 2009 Journal
multiple
72
ST Fuzzy rule × Synthetic Single 7 100 96.50 75 2004 Journal
76
ST Rule based × Synthetic Single and 11 99.5 99.3 72.7 2012 Journal
multiple
77
FDST DT DT Synthetic Single and 13 × 97.44 × 2013 Journal
and multiple
simulation
73
HST Fuzzy C‐means GA Synthetic Single 7 95.75 × × 2009 Journal
clustering
169
HST SVM‐DT GA Real time Single and 13 99.5 96.1 × 2014 Journal
multiple
174
HST+DWT SVM Gram‐Schmidt Synthetic Single and 10 99.77 99.77 99.22 2014 Journal
orthogonal multiple
transform
175
HST+DWT SVM Mutual forward Synthetic Single and 08 99.77 99.44 98.88 2014 Journal
selection multiple
74
ST Binary feature × Synthetic Single and 10 99.64 2009 Journal
matrix multiple
75
ST Maximum × Synthetic Single 7 100 99.5 2009 Journal
similarity
principle
81
ST FES PSO Synthetic Single and 12 99.2 2010 Journal
multiple
82
ST Logistic model tree × Synthetic Single and 09 99.11 97.79 2014 Journal
multiple
84
ST Hidden Markov × Synthetic Single and 7 × 95.04 91.86 2014 Journal
model multiple
83
MST Fuzzy clustering BFO+DE Synthetic 09 × 99.2 2012 Journal

(Continues)
21 of 42
TABLE 4 (Continued)
22 of 42

Type of Accuracy (%)


Feature Feature PQ Data Number of
With Noise
Extraction Classification Selection (Single/ Disturbances Without
Ref. Technique Technique Technique Data Multiple) Studied Noise 30 dB 20 dB Year Article
Single and
multiple
84
FDST Fuzzy decision × Synthetic Single and 13 × 97.94 × 2013 Journal
tree multiple
84
FDST Fuzzy decision × Real time Single and 13 × 92.69 × 2013 Journal
tree multiple
85
FDST DT × Real time Single and 11 × 95.3 × 2013 Journal
multiple
87
ST ELM × Real time Single 5 99.5 × × 2014 Journal
and synthetic
88
ST Rule‐based × Synthetic Single and 13 99.5 × × 2013 Journal
decision tree multiple
211
ST Fuzzy C‐means × Synthetic Single 10 99.2 × 98.5 2017 Journal
clustering
191
ST MLP × Synthetic Single and 09 96 95.5 × 2007 Journal
multiple
187
ST PNN × Synthetic Single and 08 100 93.25 × 2012 Journal
multiple
204
ST Modular‐NN × Synthetic Single and 11 99.96 × 2008 Journal
multiple
190
ST ANN × Synthetic Single and 09 99.5 97.51 91.04 2013 Journal
multiple
188
ST APNN PSO Synthetic Single and 11 96.3 96.1 90.8 2011 Journal
multiple
89
DRST DAG‐SVM × Synthetic Single and 09 99.5 × 97 2016 Journal
multiple
229
FDST DT × Synthetic Single and 13 99.54 × 98.35 2018 Journal
multiple
229
FDST Quad‐SVM × Synthetic Single and 13 99.42 × 94.14 2018 Journal
multiple

(Continues)
MISHRA
TABLE 4 (Continued)
MISHRA

Type of Accuracy (%)


Feature Feature PQ Data Number of
With Noise
Extraction Classification Selection (Single/ Disturbances Without
Ref. Technique Technique Technique Data Multiple) Studied Noise 30 dB 20 dB Year Article
229
FDST AdaBoost × Synthetic Single and 13 99.90 × 99.46 2018 Journal
multiple
229
FDST RF × Synthetic Single and 13 99.85 × 99.61 2018 Journal
multiple
102
HHT NN × Synthetic Single and 8 97.9 × × 2014 Journal
multiple
105
HHT PNN × Synthetic Single and 9 97.22 × × 2009 Journal
multiple
109
HHT PNN × Synthetic Single and 17 91.6 × × 2015 Journal
multiple
111
HHT Fuzzy‐based × Synthetic Single and 09 98 91.55 87.22 2014 Journal
intelligent multiple
classifier
112
HHT PNN × Synthetic Single and 06 100 × × 2016 Journal
multiple
119
HHT ELM × Real time Single and 16 96.75 × × 2017 Journal
multiple
152
SSD × Synthetic Single and 32 94.18 × × 2015 Journal
multiple
228
EMD NN × Synthetic Single and 11 99.64 95.66 93.25 2018 Journal
multiple
144
MM Fuzzy rule × Synthetic Single 5 98.566 × × 2016 Conference
123
VMD SVM × Synthetic Single and 6 100 × × 2015 Conference
multiple
124
ST+VMD SVM SFS Synthetic Single and 09 99.66 98.66 98.11 2016 Journal
multiple
125
VMD DT × Real time Single and 09 98.56 96.73 × 2018 Journal
multiple
128
VMD KELM × Synthetic Single and 15 98.82 × × 2017 Journal
and real time multiple
129
VMD WOSELM × 16 98.18 × × 2018 Journal

(Continues)
23 of 42
TABLE 4 (Continued)
24 of 42

Type of Accuracy (%)


Feature Feature PQ Data Number of
With Noise
Extraction Classification Selection (Single/ Disturbances Without
Ref. Technique Technique Technique Data Multiple) Studied Noise 30 dB 20 dB Year Article
Synthetic Single and
and real time multiple
130
VMD RKS × Synthetic Single and 6 88.89 × × 2015 Conference
multiple
126
VMD DT × Synthetic Single and 09 99.5 93.8 × 2018 Journal
and real time multiple
191
MST ANN × Synthetic Single and 14 97.2‐98.8 × 2016 Conference
multiple
78
FDST Threshold × Real time Single and 09 98.66 × 2017 Journal
multiple
225
KNN × Synthetic Single 04 96 × × 2017 Conference
97
FIR‐DGT SVM × Synthetic Single and 09 99.44 × 96.22 2017 Journal
multiple
185
ADALINE FFNN × Synthetic Single and 11 95.47 × 86.23 2014 Journal
multiple
196
LMS MLP‐NN × Synthetic Single 7 96.71 × × 2018 Journal
237
GT PNN × Synthetic Single and 9 99.51 × 98.40 2012 Journal
and multiple
simulation
238
FIR‐DGT T2FK‐SVM × Synthetic Single and 9 99.44 × 96.22 2016 Journal
and multiple
simulation
239
Higher‐order statistics Neuro‐tree × Simulation Single and 20 97.8 2018 Journal
multiple
240
PE, LSP and LP W‐ELM × Real data Single 5 96.43 2018 Conference
68
TQWT SVM × Synthetic and Single and 14 98.78 96.42 2018 Journal
simulation multiple
241
WT DT × Synthetic Single 5 96.0 2018 Conference

Abbreviations: DGHM, Donovan Geronimo Hardin and Massopust; MFSWT, modified frequency slice wavelet transform; SOLAR, self‐organizing learning array; SRC, sparse representation–based classification; DT,
decision tree; BFO, bacterial foraging optimization algorithm; RKS, random kitchen sink; DRST, double‐resolution S‐transform; APNN, adaptive probabilistic neural network; PE, permutation entropy; LSP, Lomb‐
MISHRA

Scargle periodogram; LP, local peaks.


MISHRA 25 of 42

4.1 | Implementation and comparative study of few PQD&C techniques

In this section, several most accepted techniques based on SP and PRTs for PQD detection and classification are studied
and compared closely considering diverse operating conditions. In general, 18 PQD signals, comprising both single and
multiple disturbances, are simulated based on IEEE‐1159 standard. Table 5 shows the PQD events used in this work and
its modelling equations with parameter ranges. A 2.5‐kHz sampling frequency is used on ten cycles of distorted wave-
forms for the feature extraction. The signal processing techniques considered for feature extraction process are STFT,
WT, ST, FDST, HHT, and VMD. Figure 4A shows a sample of simulated voltage signal associated to multiple PQDs such
as transient and sag. The detection using STFT is shown in Figure 4B. The detection results using db4 wavelet, ST,
FDST, HHT, and VMD are presented in Figure 4C to 4G, respectively. These results noticeably display an abrupt rise
in the magnitudes at the instant of disturbance initiation, thus clearly identifying the disturbance. A brief description
of feature vectors extracted using these SPTs is presented in Table 6. The classification techniques included in this work
are the ANN, KNN, naive Bayesian classifier (NBC), SVM, and ELM. Table 7 shows the parameters of these classifiers,
which are chosen to give the highest accuracy for the selected features. An aggregate of 250 PQD signals is simulated for
each disturbance, and feature extraction has been carried out. Then, each classifier is trained with a data set of 2700
(150 × 18) patterns consisting of 150 patterns for each class. Remaining data from (250 × 18) PQD signals are used
for testing purpose. The outcomes in terms of overall accuracy of the studied classification methods are presented in
Table 8 for both noiseless and noisy environments. It is easily seen that the combination of VMD and KNN provides
higher classification accuracy as compared with other approaches.

4.2 | PQ issues influenced by DGs and application of various PQD&C schemes for
microgrid environment

Microgrid is one of the emerging architecture under the umbrella of smart grid infrastructure. In microgrid environ-
ment, the integration of NCES and DG boosts the application of PET, which is the main cause for various PQ issues.
DG using NCES has amplified PQ issues as it requires, in numerous situations, solid‐state conversion and variations
in input power add new problems of voltage quality such as in solar PV generation and wind energy conversion systems.
DG also affects the current quality. With large penetration of DG, their tripping is a problem not only for the generator
owner but also for the system operator and other customers. The tripping of one discrete unit should not be a matter to
the EPS, but the simultaneous tripping of a large number of units is a thoughtful point.
The major PQ problems influenced by DGs are the following:

1. Sustained interruptions. This is the traditional reliability area. Lots of generators are designed to offer standby supply
to the consumer during power interruption. Nevertheless, a distributed generator has the potential to multiply the
number of interruptions in several cases.
2. Voltage regulation. This is often the most limiting factor for how much DG can be accommodated on a distribution
feeder without making changes.
3. Harmonics. Harmonics is a serious concern in case of both synchronous and inverter‐based DG. However, it is less
with inverter‐based DG utilizing advanced tools.
4. Voltage sags. Voltage sag is a unusual case since DG may or may not be responsible for it.

In recent years, research on PQ issues and its detection principle with the presence of DGs has been increased.
Several authors have proposed various PQD&C techniques in microgrid environment. Some of the most acceptable
researches on this direction are reviewed in this section as follows. Achlerkar et al125 presented VMD and DT‐based
method for PQD&C in grid‐connected distributed power generation system (DPGS). The DPGS considered for the
study consisting of PV and wind generation system. The proposed technique was tested for recognition of stationary
PQ disturbances (like harmonics, interharmonics, and flicker), nonstationary PQ events (transients), and noise. Dash
et al231 presented a fast modified recursive Gauss‐Newton (MRGN) technique to estimate PQ indices in microgrid sys-
tem in both islanded and nonislanded conditions. Ray et al169 presented PQ disturbance classification technique
using ST, SVM, and DT for DPGS. Various PQ events such as sag, swell, notch, and harmonics were considered
for the classification process. Martinez and Martin‐Arnedo232 analysed the impact of conventional synchronous
machine‐based DG on the current protection practices for DPGS. Moreover, the effect of small generation units on
26 of 42 MISHRA

TABLE 5 PQDs events, equations, and parameter

Pure (C1) Interruption (C2)

A sin (ωt) A(1 − ρ(u(t − t1) − u(t − t2))) sin (ωt)


where, A = 1(p. u.) and ω = 2π50 rad/sec where, 0.9 ≤ ρ ≤ 1, T ≤ (t2 ‐ t1) ≤ 9T

Sag (C3) Swell (C4)

A(1 − α(u(t − t1) − u(t − t2))) sin (ωt) A(1 + β(u(t − t1) − u(t − t2))) sin (ωt)
where, 0.1 ≤ β ≤ 0.8, T ≤ (t2 ‐ t1) ≤ 9T where, 0.1 ≤ β ≤ 0.8, T ≤ (t2 ‐ t1) ≤ 9T

Oscillatory transient (C5) Harmonics (C6)

−ðt−t Þ=τ
sinðωtÞ þ βe 1
* sinðωn ðt − t 1 ÞÞ*ðuðt − t 2 Þ − uðt − t 1 ÞÞ 7
where; 0:1 ≤ β ≤ 0:8; 0:5T ≤ ðt2 ‐t1 Þ ≤ 3T; sinðωtÞ þ ∑ αn sinðnωtÞ
n¼3
8ms ≤ τ30ms and 300Hz ≤ f n ≤ 900Hz

(Continues)
MISHRA 27 of 42

TABLE 5 (Continued)

Pure (C1) Interruption (C2)


where, 0.05 ≤ a3, a5, a7 ≤ 0.15 and ∑ an2 = 1
Flicker (C7) Notch (C8)

(1 + λ sin (κωt)) * sin (ωt) sinðwtÞ − signðsigðwtÞÞ sinðwtÞ


h i
where, 0.1 ≤ λ ≤ 0.2, 5 ≤ κ ≤ 50Hz 9
∑n¼0 K × fuðt − ðt 1 þ 0:2nÞÞ − uðt − ðt 2 þ 0:2nÞÞg
where; 0:1 ≤ K ≤ 0:4; 0:01T ≤ ðt2 ‐t1 Þ ≤ 0:05T and
0 ≤ t 2 ; t 1 ≤ 0:5T

Harmonics + sag (C9) Harmonics + swell (C10)

 7
  7

Að1 − αðuðt − t1Þ − uðt − t2ÞÞÞ sinðωtÞ sinðωt Þ þ ∑ αn sinðnωt Þ Að1 − βðuðt − t1Þ − uðt − t2ÞÞÞ* sinðωt Þ* sinðωtÞ þ ∑ αn sinðnωt Þ
n¼3 n¼3

Harmonics flicker (C11) Harmonics interrupt (C12)

(Continues)
28 of 42 MISHRA

TABLE 5 (Continued)

Pure (C1) Interruption (C2)


   
   7 7
1 þ λ sin ωf t * sinðωtÞ* sinðωtÞ þ ∑ αn sinðnωtÞ Að1 − ρðuðt − t1Þ − uðt − t2ÞÞÞ* sinðωt Þ* sinðωt Þ þ ∑ αn sinðnωt Þ
n¼3 n¼3

Harmonics + oscillatory transient (C13) Harmonics + notch (C14)

sinðωtÞ þ βe−ðt−t1 Þ=τ* sinðωn ðt − t1 ÞÞ*ðuðt − t 2 Þ − sinðwtÞ − signðsigðwt ÞÞ sinðwt Þ ×


   9 
7
uðt − t1 ÞÞ* sinðωt Þ þ ∑ αn sinðnωt Þ ∑ K × fuðt − ðt 1 þ 0:2nÞÞ − uðt − ðt2 þ 0:2nÞÞg ×
n¼0
n¼3
 7

sinðωtÞ sinðωt Þ þ ∑ αn sinðnωt Þ
n¼3

Flicker + sag (C15) Flicker + swell (C16)

Að1 − αðuðt − t1Þ − uðt − t 2 ÞÞÞ sinðωt Þ*ðð1 þ λ sinðκωt ÞÞ* sinðωt ÞÞ Að1 þ βðuðt − t1Þ − uðt − t2ÞÞÞ sinðωt Þ*ðð1 þ λ sinðκωt ÞÞ* sinðωtÞÞ
Oscillatory transient + sag (C17) Oscillatory transient + swell (C18)

Að1 − αðuðt − t1Þ − uðt − t2ÞÞÞ sinðωtÞ* Að1 þ βðuðt − t1Þ − uðt − t2ÞÞÞ sinðωt Þ sinðωt Þ*
h i h i
sinðωtÞ þ βe−ðt−t1 Þ=τ* sinðωn ðt − t1 ÞÞ*ðuðt − t 2 Þ − uðt − t 1 ÞÞ sinðωt Þ þ βe−ðt−t1 Þ=τ* sinðωn ðt − t 1 ÞÞ*ðuðt − t2 Þ − uðt − t 1 ÞÞ
MISHRA 29 of 42

FIGURE 4 PQD detection. A, Disturbed multiple PQ voltage signal (transient + sag). B, Detection results by using STFT. C, Detection
results by using WT. D, Detection results by using ST. E, Detection results by using FDST. F, Detection results by using EMD. G,
Detection results by using VMD

TABLE 6 Brief description of feature vectors extracted using SPTs

Feature Extraction Techniques (FETs) Extracted Feature Vector

STFT Energy, standard deviation, skewness, minimum, entropy, kurtosis and maximum of 50‐Hz
contour and entropy and mean of 150‐Hz contour.177
WT (db4 mother wavelet and Energy, Shannon entropy, kurtosis, minimum, skewness, standard deviation, and total
decomposition level equals to 4) harmonic distortion (THD) related to fourth level detail coefficient.
ST Maximum, minimum, mean, standard deviation, kurtosis of time‐maximum amplitude (TmA)
plot, THD, skewness and standard deviation of frequency‐maximum amplitude (FmA) plot
and amplitude factor.205
FDST Maximum, minimum, mean, standard deviation, kurtosis of time vs maximum amplitude
(TmA) plot, THD, skewness, and standard deviation of FmA plot242
HHT Energy, entropy, skewness, minimum, and maximum of amplitude curve from the first IMF,
standard deviation, skewness, and energy of phase curve.177
VMD (decomposition level equals to 4) Relative energy ratio (RER), mode instantaneous amplitude (IA), number of zero crossing, and
centre frequency (CF) from the each of the modes.125
30 of 42 MISHRA

TABLE 7 Summarized parameters of the PRTs.

PRTs Parameter

ANN Hidden layer sizes are 16, and Levenberg‐Marquardt back propagation is used as training function
KNN The number of nearest neighbor is 1 (K = 1), and Euclidean distance is used
NBC Gaussian distribution fit is used to model data
SVM Gaussian radial basis function is used as kernel function
ELM Sinusoid basis function is used as kernel function

TABLE 8 Comparative performance study of few PQD&C techniques

Noiseless Noisy Signals Noiseless Noiseless


FETs PRTs Signals SNR = 20 dB FETs Signals SNR = 20 dB FETs Signals SNR = 20 dB

FT ANN 87.55% 82.44% WT 96.33% 89.94% ST 95.88% 93.05%


KNN 87.77% 82.88% 97.05% 90.61% 96.77% 93.83%
NBC 87.33% 82.77% 96.72% 90.44% 96.72% 92.88%
SVM 87.11% 83.05% 96.83% 90.72% 97.05% 93.38%
ELM 88.11% 83.22% 97.00% 91.27% 97.38% 93.72%
HHT ANN 98.11% 95.83% FDST 95.88% 93.11% VMD 98.55% 95.66%
KNN 99.00% 97.38% 98.55% 93.5% 99.05% 96.77%
NBC 98.27% 95.83% 97.11% 93.61% 98.11% 95.88%
SVM 98.5% 96.11% 96.83% 93.72% 98.55% 96.11%
ELM 98.88% 96.55% 98.27% 93.88% 98.94% 96.05%

voltage sags was also discussed. Demirci et al233 presented a general real‐time monitoring system to monitor all elec-
trical parameter and PQ parameters of the electric power system including its interfaces with the DG system.
Chakravorti et al234,235 proposed a novel MM‐based disturbance detection and classification method for wind
farm–based microgrid system, where DT was adopted as a classifier for the disturbances like islanding, faults, load
switching capacitor switching, etc, in addition to some PQ events for example sag and swell. In grid‐connected DG
system, harmonic and interharmonic are very serious issues. A novel technique using MM filter and HHT was sug-
gested for the detection of harmonic and interharmonic caused by the electric elements in Chen et al.141 Mishra
et al236 presented a detection technique for islanding events based on HHT and classification of islanding events from
various nonislanding events (such as fault, load switching, capacitor switching, sag, and swell) using ELM classifier.
Considering the sensitivity issues such as stability and PQ caused by integration of renewable energy sources (RES) to
utility system, authors92 proposed a novel approach based on ST for PQ assessment and detection of islanding, out-
age, and grid synchronization of renewable energy sources. Ray et al6 presented the classification of PQ disturbances
produced not only by load switching but also by environmental characteristics such as variation in solar insolation
and wind speed. Several kinds of sag and swell events sourced by load switching, variation in wind speed, and solar
insolation were considered in the study. Here, ST and PNN were considered for feature extraction and classification
approach, respectively.

5 | CONCLUSION

A complete and thorough study on the applications of SP and PRTs to the detection of PQ disturbances is reported in
this manuscript. The main sources of PQ disturbances and its effect to the consumer are explained in detailed. The stan-
dard stated by various organizations related to PQ signal is also described in depth. The three key stages of automatic
recognition of PQ events are states as feature extraction, optimal feature selection, and classification. With this regards,
various feature extraction techniques based on SP techniques such as FT, WT, ST, HHT, KF, VMD, EMD, MM, and
some other miscellaneous techniques are reviewed for PQ event detection problem. Various intelligent classifiers for
automatic PQ disturbance classification are also reviewed like ANN, SVM, neuro‐fuzzy, FES, ELM, and some other mis-
cellaneous classifiers. Moreover, some feature selection technique used for the application of automatic PQD recogni-
tion is reviewed in details. The key benefits and drawbacks along with the effectiveness of the specific approaches
MISHRA 31 of 42

are also outlined in this paper. The comparative study between various existing techniques related to PQD&C has been
done and highlighted some useful remarks which can be utilized as future scope in the direction of PQ disturbance
characterization.
In short, the main contributions of this article can be summarized as follows:

1. More than 250 published works have been reviewed, and a critical overview of various PQ disturbances in terms of
taxonomy, standards, and security issues is provided.
2. Generally, the PQ detection technique comprises three major steps: feature extraction, feature selection, and classi-
fication. Therefore, various techniques related to feature extraction, selection, and classification involved in PQ
detection field are critically studied and presented in this article.
3. A brief discussion on application of various PQD&C schemes for microgrid environment using signal processing
techniques and pattern recognition has been presented.
4. Various useful remarks which can be used for the further improvement of the detection technique are highlighted.
5. A comparative assessment on various PQ detection and classification (PQD&C) techniques (more than 120 articles)
has been carried out in detail considering several criteria, such as type of data used (synthetic or real world), type of
PQ disturbance studied (single or multiple), and performance in terms of accuracy (both noiseless and noisy
environment).
6. Lastly, implementation and comparative study of several most accepted PQD&C techniques are carried out on 18
PQ events including single and multiple disturbances.

LIST OF ABBREVIATIONS

PQ Power quality
PQD&C PQ detection and classification
DSP Digital signal processing
NCES Non‐conventional energy sources
EPS Electric power systems
PET Power electronics technology
DG Distributed generation
THD Total harmonic distortion
FFT Fast Fourier transform
STFT Short‐time Fourier transform
WT Wavelet transform
HHT Hilbert‐Huang transform
KF Kalman filters
GT Gabor transform
ST S‐transform
PRT Pattern recognition techniques
SPT Signal processing techniques
FS Feature selection
PQD Power quality disturbances
CWT Continuous‐WT
DWT Discrete‐WT
WPT Wavelet packet transform
MSD Multi‐resolution Signal Decomposition
FSWT Frequency slice WT
ANN Artificial neural network
HMM Hidden‐markov model
SVM Support vector machine
WPD Wavelet packet decomposition
IPT Image processing technique
32 of 42 MISHRA

EWT Empirical wavelet transform


MODWT Maximum overlapping discrete wavelet transform
Q Q‐factor
FDST Discrete s‐transform
HST Hyperbolic‐ST
FCM Fuzzy C‐means
FE&C Feature extraction and classification
DOST Discrete‐orthogonal s‐transform
DAG Directed acyclic graph
ELM Extreme machine learning
FIR Finite impulse response
WDF Wigner distribution function
GWT Gabor–Wigner transform
PSO Particle swarm optimization
EMD Empirical mode decomposition
IMF Intrinsic mode functions
PNN Probabilistic neural network
EEMD Ensemble empirical mode decomposition
FPARR Fuzzy‐product‐aggregation‐reasoning‐rule
SAX Symbolic aggregate approximation
MM Mathematical morphology
SP Signal processing
RBFNN Radial basis function neural network
VMD Variational mode decomposition
RVD Recursive variational decomposition
SVD Singular value decomposition
HMMG Half Multi‐resolution Morphology Gradient
TFR Time–frequency representation
SSD Sparse signal decomposition
PC Principle curves
AFD Amplitude and Frequency Demodulation
AMLW Morphological lifting wavelet
GKA Genetic k‐means algorithm
GA Genetic algorithms
EGA Extension GA
M Mutual information
R Relief
mRMR Maximum relevance minimum redundancy
MLP Multi‐layer perceptron
FFNN Feed‐forward‐NN
DSE Dempster–Shafer evidence
LVQ Learning vector quantization
ADALINE Adaptive linear network
FIPS Fully‐informed‐particle‐swarm
LMS Least mean square
OAA One‐against‐all
OAO One‐against‐one
FAM Fuzzy associative memory
WBELM Weighted‐bidirectional‐ELM
DL Deep learning
SAE Stacked auto encoder
GRU Gated recurrent units
MISHRA 33 of 42

SSA Singular spectrum analysis


WEKA Waikato Environment for Knowledge Analysis
SVDD Support vector data description
BFM Binary feature matrix
KNN K nearest neighbors
NBC Naive Bayesian classifier
RES Renewable energy sources
DGHM Donovan Geronimo Hardin and Massopust
SOLAR Self‐organizing learning array
DT Decision tree
RKS Random kitchen sink
DRST Double‐resolution S‐transform
APNN Adaptive probabilistic neural network
PE Permutation entropy
LSP Lomb–Scargle periodogram
LP Local peaks

ORCID
Manohar Mishra https://orcid.org/0000-0003-2160-4703

R EF E RE N C E S
1. Thapar A, Saha TK, Dong ZY. Investigation of power quality categorisation and simulating it's impact on sensitive electronic equipment.
In Power Engineering Society General Meeting, 2004. IEEE, 2004; 528‐533
2. Khan AK. Monitoring power for the future. Power Eng J. 2001;15(2):81‐85.
3. Ouyang S, Wang J. A new morphology method for enhancing power quality monitoring system. Int J Electr Power Energy Syst.
2007;29(2):121‐128.
4. Smith JC, Hensley G, Ray L. IEEE recommended practice for monitoring electric power quality. IEEE Std. 1995;1159‐1995.
5. 〈http://www.scopus.com/〉 [accessed 31.12.18].
6. Ray PK, Mohanty SR, Kishor N. Classification of power quality disturbances due to environmental characteristics in distributed gener-
ation system. IEEE Trans Sustainable Energy. 2013;4(2):302‐313.
7. Granados‐Lieberman D, Romero‐Troncoso RJ, Cabal‐Yepez E, Osornio‐Rios RA, Franco‐Gasca LA. A real‐time smart sensor for high‐
resolution frequency estimation in power systems. Sensors. 2009;9(9):7412‐7429.
8. Lara‐Cardoso J, de Jesus Romero‐Troncoso R. May. Low‐cost power harmonics analyzer of nonlinear loads based on FPGA. In Instru-
mentation and Measurement Technology Conference Proceedings, 2008. IMTC 2008. IEEE, 2008; 730‐735
9. Siahkali H. December. Power quality indexes for continue and discrete disturbances in a distribution area. In Power and Energy Con-
ference, 2008. PECon 2008. IEEE 2nd International, 2008; 678‐683
10. Broshi A. October. Monitoring power quality beyond EN 50160 and IEC 61000‐4‐30. In Electrical Power Quality and Utilisation, 2007.
EPQU 2007. 9th International Conference on, 2007; 1‐6
11. Flores RA. October. State of the art in the classification of power quality events, an overview. In Harmonics and Quality of Power, 2002.
10th International Conference on 2002; (Vol. 1, pp. 17‐20). IEEE.
12. Ibrahim WA, Morcos MM. Artificial intelligence and advanced mathematical tools for power quality applications: a survey. IEEE Trans
Power Delivery. 2002;17(2):668‐673.
13. Granados‐Lieberman D, Romero‐Troncoso RJ, Osornio‐Rios RA, Garcia‐Perez A, Cabal‐Yepez E. Techniques and methodologies for
power quality analysis and disturbances classification in power systems: a review. IET Gener Transm Distrib. 2011;5(4):519‐529.
14. Saini MK, Kapoor R. Classification of power quality events—a review. Int J Electr Power Energy Syst. 2012;43(1):11‐19.
15. Mahela OP, Shaik AG, Gupta N. A critical review of detection and classification of power quality events. Renew Sustain Energy Rev.
2015;41:495‐505.
16. Saxena D, Verma K, Singh S. Power quality event classification: an overview and key issues. Int J Eng Sci Technol. 2010;2(3):186‐199.
17. Khokhar S, Zin AABM, Mokhtar ASB, Pesaran M. A comprehensive overview on signal processing and artificial intelligence techniques
applications in classification of power quality disturbances. Renew Sustain Energy Rev. 2015;51:1650‐1663.
34 of 42 MISHRA

18. Jay F, Goetz JA. IEEE Standard Dictionary of Electrical and Electronics Terms. New York: The Institute of Electrical and Electronics
Engineers. Inc; 1984.
19. Dekhandji FZ. Signal processing deployment in power quality disturbance detection and classification. Acta Phys Pol, A.
2017;13(2pp):415‐419.
20. International Electrotechnical Commission. Electromagnetic compatibility (EMC)‐part 4‐30: testing and measurement techniques‐power
quality measurement methods. IEC 61000‐4‐30, 2003
21. Bollen MH. Understanding Power Quality Problems: Voltage Sags and Interruptions. New York: IEEE press; 2000. https://doi.org/10.1109/
9780470546840
22. Standard E. European Standard EN 50160. Voltage characteristics of electricity supplied by public electricity networks. CENELEC. 2010.
23. International Electrotechnical Commission. IEC 61000‐4‐30: testing and measurement techniques. Power quality measurement
methods, 2008
24. Szmajda M, Gorecki K, Mroczka J. October. DFT algorithm analysis in low‐cost power quality measurement systems based on a DSP
processor. In Electrical Power Quality and Utilisation, 2007. EPQU 2007. 9th International Conference on IEEE, 2007; 1‐6
25. Huang SJ, Hsieh CT, Huang CL. Application of Morlet wavelets to supervise power system disturbances. v. 1999;14(1):235‐243.
26. Heydt GT, Fjeld PS, Liu CC, Pierce D, Tu L, Hensley G. Applications of the windowed FFT to electric power quality assessment. v.
1999;14(4):1411‐1416.
27. Dash PK, Panigrahi BK, Sahoo DK, Panda G. Power quality disturbance data compression, detection, and classification using integrated
spline wavelet and S‐transform. v. 2003;18(2):595‐600.
28. Lee IW, Dash PK. S‐transform‐based intelligent system for classification of power quality disturbance signals. IEEE Trans Ind Electron.
2003;50(4):800‐805.
29. Wright PS. Short‐time Fourier transforms and Wigner‐Ville distributions applied to the calibration of power frequency harmonic ana-
lyzers. IEEE Trans Instrum Meas. 1999;48(2):475‐478.
30. Gu YH, Bollen MH. Time‐frequency and time‐scale domain analysis of voltage disturbances. v. 2000;15(4):1279‐1284.
31. Beichang G, Xiao W, Yingying L, Zhang Z, Fan Y. Influence research of renewable energy application on power quality detection. In:
IOP Conference Series: Earth and Environmental Science. Vol.168, No. 1 IOP Publishing; 2018, June:012033 https://doi.org/10.1088/
1755‐1315/168/1/012033.
32. Addison PS. The Illustrated Wavelet Transform Handbook: Introductory Theory and Applications in Science, Engineering, Medicine and
Finance. Boca Raton: CRC press; 2017. https://doi.org/10.1201/9781315372556
33. Nath S, Sinha P, Goswami SK. A wavelet based novel method for the detection of harmonic sources in power systems. Int J Electr Power
Energy Syst. 2012;40(1):54‐61.
34. Santoso S, Powers EJ, Grady WM, Hofmann P. Power quality assessment via wavelet transform analysis. IEEE Trans Power Delivery.
1996;11(2):924‐930.
35. Pillay P, Bhattacharjee A. Application of wavelets to model short‐term power system disturbances. IEEE Trans Power Syst.
1996;11(4):2031‐2037.
36. Santoso S, Powers EJ, Grady WM. Power quality disturbance data compression using wavelet transform methods. v.
1997;12(3):1250‐1257.
37. Gaouda AM, Salama MMA, Sultan MR, Chikhani AY. Application of multiresolution signal decomposition for monitoring short‐
duration variations in distribution systems. v. 2000;15(2):478‐485.
38. Xiangxun C. Wavelet‐based detection, localization, quantification and classification of short duration power quality disturbances. In
Power Engineering Society Winter Meeting, 2002. IEEE. 2002;2:931‐936.
39. Angrisani L, Daponte P, D'apuzzo M, Testa A. A measurement method based on the wavelet transform for power quality analysis. v.
1998;13(4):990‐998.
40. Gencer Ö, Öztürk S, Erfidan T. A new approach to voltage sag detection based on wavelet transform. Int J Electr Power Energy Syst.
2010;32(2):133‐140.
41. Dekhandji FZ. October. Detection of power quality disturbances using discrete wavelet transform. In Electrical Engineering‐Boumerdes
(ICEE‐B), 2017 5th International Conference on 2017, (pp. 1‐5). IEEE.
42. Deokar SA, Waghmare LM. Integrated DWT–FFT approach for detection and classification of power quality disturbances. Int J Electr
Power Energy Syst. 2014;61:594‐605.
43. Liao Y, Lee JB. A fuzzy‐expert system for classifying power quality disturbances. Int J Electr Power Energy Syst. 2004;26(3):199‐205.
44. Angrisani L, Daponte P, D'Apuzzo M. Wavelet network‐based detection and classification of transients. IEEE Trans Instrum Meas.
2001;50(5):1425‐1435.
45. Kaewarsa S, Attakitmongcol K, Kulworawanichpong T. Recognition of power quality events by using multiwavelet‐based neural net-
works. Int J Electr Power Energy Syst. 2008;30(4):254‐260.
46. Garousi MR, Shakarami MR, Namdari F. Detection and classification of power quality disturbances using parallel neural networks based
on discrete wavelet transform. J Electr Syst. 2016;12(1):158‐173.
MISHRA 35 of 42

47. Subbarao MV, Samundiswary P. January. Time‐frequency analysis of non‐stationary signals using frequency slice wavelet transform. In
Intelligent Systems and Control (ISCO), 2016 10th International Conference on IEEE, 2016;1‐6.
48. Biswal B, Mishra S. Power signal disturbance identification and classification using a modified frequency slice wavelet transform. IET
Gener Transm Distrib. 2014;8(2):353‐362.
49. Karimi M, Mokhtari H, Iravani MR, et al. v. 15(4):1212‐1220.
50. Chung J, Powers EJ, Grady WM, Bhatt SC. Power disturbance classifier using a rule‐based method and wavelet packet‐based hidden
Markov model. IEEE Trans Power Delivery. 2002;17(1):233‐241.
51. Zhang M, Li K, Hu Y. Classification of power quality disturbances using wavelet packet energy and multiclass support vector machine.
COMPEL Int J Comput Math Electr Electron Eng. 2012;31(2):424‐442.
52. Liu Z, Hu Q, Cui Y, Zhang Q. A new detection approach of transient disturbances combining wavelet packet and Tsallis entropy.
Neurocomputing. 2014;142:393‐407.
53. Bhuiyan SM, Khan J, Murphy G. WPD for detecting disturbances in presence of noise in smart grid for PQ monitoring. IEEE Trans Ind
Appl. 2018;54(1):702‐711.
54. Ece DG, Gerek ON. Power quality event detection using joint 2‐D‐wavelet subspaces. IEEE Trans Instrum Meas. 2004;53(4):
1040‐1046.
55. Shareef H, Mohamed A, Ibrahim AA. An image processing based method for power quality event identification. Int J Electr Power Energy
Syst. 2013;46:184‐197.
56. Krishna BV, Kaliaperumal B. Image pattern recognition technique for the classification of multiple power quality disturbances. Turk J
Electr Eng Comput Sci. 2013;21(3):656‐678.
57. Erişti H, Yıldırım Ö, Erişti B, Demir Y. Optimal feature selection for classification of the power quality events using wavelet transform
and least squares support vector machines. Int J Electr Power Energy Syst. 2013;49:95‐103.
58. He H, Starzyk JA. A self‐organizing learning array system for power quality classification based on wavelet transform. v.
2006;21(1):286‐295.
59. Chakraborty S, Chatterjee A, Goswami SK. A sparse representation based approach for recognition of power system transients. Eng Appl
Artif Intel. 2014;30:137‐144.
60. Zafar T, Morsi WG. Power quality and the un‐decimated wavelet transform: an analytic approach for time‐varying disturbances. Electr
Pow Syst Res. 2013;96:201‐210.
61. Uyar M, Yildirim S, Gencoglu MT. An effective wavelet‐based feature extraction method for classification of power quality disturbance
signals. Electr Pow Syst Res. 2008;78(10):1747‐1755.
62. Eristi B, Yildirim O, Eristi H, Demir Y. A real‐time power quality disturbance detection system based on the wavelet transform. In Power
Engineering Conference (UPEC), 2016 51st International Universities IEEE.2016;1‐5.
63. Priyadarshini MS, Sushama M. Classification of short‐duration voltage variations using wavelet decomposition based entropy
criteria. In Wireless Communications, Signal Processing and Networking (WiSPNET), International Conference on IEEE, 2016;
2192‐2196.
64. Ray P, Budumuru GK, Mohanty BK. A comprehensive review on soft computing and signal processing techniques in feature extraction
and classification of power quality problems. J Renewable Sustainable Energy. 2018;10(2):025102.
65. Ucar F, Alcin OF, Dandil B, Ata F. Power quality event detection using a fast extreme learning machine. Energies. 2018;11(1):145.
66. Upadhyaya S. Detection and classification of combined real time power quality disturbance signals with Hidden Markov Models incor-
porating wavelet features. Int J Scientific Res Sci Technol IJSRST. 2018;4(8):85‐101.
67. Gursoy Mİ, Ustun SV, Yilmaz AS. An efficient DWT and EWT feature extraction methods for classification of real data PQ disturbances.
Uluslararası Mühendislik Araştırma ve Geliştirme Dergisi. 2018;10(1):158‐171.
68. Thirumala K, Prasad MS, Jain T, Umarikar AC. Tunable‐Q wavelet transform and dual multiclass SVM for online automatic detection of
power quality disturbances. IEEE Trans Smart Grid. 2018;9(4):3018‐3028.
69. Dash PK, Panigrahi BK, Panda G. Power quality analysis using S‐transform. IEEE Trans Power Delivery. 2003;18(2):406‐411.
70. Zhao F, Yang R. Power‐quality disturbance recognition using S‐transform. v. 2007;22(2):944‐950.
71. Mishra S, Bhende CN, Panigrahi BK. Detection and classification of power quality disturbances using S‐transform and probabilistic
neural network. IEEE Trans Power Delivery. 2008;23(1):280‐287.
72. Chilukuri MV, Dash PK. Multiresolution S‐transform‐based fuzzy recognition system for power quality events. IEEE Trans Power
Delivery. 2004;19(1):323‐330.
73. Biswal B, Dash PK, Panigrahi BK. Non‐stationary power signal processing for pattern recognition using HS‐transform. Appl Soft Comput.
2009;9(1):107‐117.
74. Nguyen T, Liao Y. Power quality disturbance classification utilizing S‐transform and binary feature matrix method. Electr Pow Syst Res.
2009;79(4):569‐575.
75. Xiao X, Xu F, Yang H. Short duration disturbance classifying based on S‐transform maximum similarity. Int J Electr Power Energy Syst.
2009;31(7‐8):374‐378.
36 of 42 MISHRA

76. Rodríguez A, Aguado JA, Martín F, López JJ, Muñoz F, Ruiz JE. Rule‐based classification of power quality disturbances using S‐
transform. Electr Pow Syst Res. 2012;86:113‐121.
77. Biswal M, Dash PK. Detection and characterization of multiple power quality disturbances with a fast S‐transform and decision tree
based classifier. Digital Signal Process. 2013;23(4):1071‐1083.
78. Jaiswal S, Ballal MS. FDST‐based PQ event detection and energy metering implementation on FPGA‐in‐the‐loop and NI‐LabVIEW.
IET Sci Meas Technol. 2017;11(4):453‐463.
79. He S, Li K, Zhang M. A real‐time power quality disturbances classification using hybrid method based on S‐transform and dynamics.
IEEE Trans Instrum Meas. 2013;62(9):2465‐2475.
80. Reddy MJB, Raghupathy RK, Venkatesh KP, Mohanta DK. Power quality analysis using discrete orthogonal S‐transform (DOST). Digital
Signal Process. 2013;23(2):616‐626.
81. Behera HS, Dash PK, Biswal B. Power quality time series data mining using S‐transform and fuzzy expert system. Appl Soft Comput.
2010;10(3):945‐955.
82. Moravej Z, Abdoos AA, Pazoki M. New combined S‐transform and logistic model tree technique for recognition and classification of
power quality disturbances. Electr Power Compon Syst. 2011;39(1):80‐98.
83. Biswal B, Behera HS, Bisoi R, Dash PK. Classification of power quality data using decision tree and chemotactic differential evolution
based fuzzy clustering. Swarm Evol Comput. 2012;4:12‐24.
84. Hasheminejad S, Esmaeili S, Jazebi S. Power quality disturbance classification using S‐transform and hidden Markov model.
Electr Power Compon Syst. 2012;40(10):1160‐1182.
85. Erişti H, Yıldırım Ö, Erişti B, Demir Y. Automatic recognition system of underlying causes of power quality disturbances based on S‐
transform and extreme learning machine. Int J Electr Power Energy Syst. 2014;61:553‐562.
86. Biswal M, Dash PK. Measurement and classification of simultaneous power signal patterns with an S‐transform variant and fuzzy deci-
sion tree. IEEE Trans Ind Inf. 2013;9(4):1819‐1827.
87. Babu PR, Dash PK, Swain SK, Sivanagaraju S. A new fast discrete S‐transform and decision tree for the classification and monitoring of
power quality disturbance waveforms. Int Trans Electr Energy Syst. 2014;24(9):1279‐1300.
88. Kumar R, Singh B, Shahani DT, Chandra A, Al‐Haddad K. Recognition of power‐quality disturbances using S‐transform‐based ANN
classifier and rule‐based decision tree. IEEE Trans Ind Appl. 2015;51(2):1249‐1258.
89. Li J, Teng Z, Tang Q, Song J. Detection and classification of power quality disturbances using double resolution S‐transform and DAG‐
SVMs. IEEE Trans Instrum Meas. 2016;65(10):2302‐2312.
90. Pujiantara AEM, Priyadi A, Pujiantara M, Penangsang O, Anggriawan DO, Tjahjono A. Improvement of power quality monitoring
based on modified S‐transform. In Intelligent Technology and Its Applications (ISITIA), 2016 International Seminar on IEEE, 2016;
539‐544.
91. Sahu G, Choubey A. Simultaneous power quality disturbances analysis using modified S‐transform and evolutionary approach. In
Microelectronics, Electromagnetics and Telecommunications Springer, Singapore, 2018, 305‐314.
92. Shaik AG, Mahela OP. Power quality assessment and event detection in hybrid power system. Electr Pow Syst Res. 2018;161:26‐44.
93. Qian S, Chen D. Discrete gabor transform. IEEE Trans Signal Process. 1993;41(7):2429‐2438.
94. Huang SJ, Huang CL, Hsieh CT. Application of Gabor transform technique to supervise power system transient harmonics. IEE Proc
Gen, Transm Distrib. 1996;143(5):461‐466.
95. Kawady TA, Elkalashy NI, Ibrahim AE, Taalab AMI. Arcing fault identification using combined Gabor Transform‐neural network for
transmission lines. Int J Electr Power Energy Syst. 2014;6:248‐258.
96. Cho SH, Jang G, Kwon SH. Time‐frequency analysis of power‐quality disturbances via the Gabor–Wigner transform. IEEE Trans Power
Delivery. 2010;25:494‐499.
97. Naderian S, Salemnia A. Method for classification of PQ events based on discrete Gabor transform with FIR window and T2FK‐based
SVM and its experimental verification. IET Gener Transm Distrib. 2017;11(1):133‐141.
98. Bertola A, Lazaroiu GC, Roscia M, Zaninelli D. September. A Matlab‐Simulink flickermeter model for power quality studies. In
Harmonics and Quality of Power, 2004. 11th International Conference on IEEE, 2014 4, 734‐738.
99. Abdelsalam AA, Eldesouky AA, Sallam AA. Classification of power system disturbances using linear Kalman filter and fuzzy‐expert
system. Int J Electr Power Energy Syst. 2012;43(1):688‐695.
100. Dash PK, Chilukuri MV. Hybrid S‐transform and Kalman filtering approach for detection and measurement of short duration distur-
bances in power networks. IEEE Trans Instrum Meas. 2002;53(2):588‐596.
101. Reddy JBV, Dash PK, Samantaray R, Moharana AK. Fast tracking of power quality disturbance signals using an optimized unscented
filter. IEEE Trans Instrum Meas. 2009;58(12):3943‐3952.
102. Biswal B, Biswal M, Mishra S, Jalaja R. Automatic classification of power quality events using balanced neural tree. IEEE Trans Ind
Electron. 2014;61(1):521‐530.
103. Rilling G, Flandrin P, Goncalves P. On empirical mode decomposition and its algorithms. In IEEE‐EURASIP workshop on nonlinear
signal and image processing, jin,2003; 3, 8‐11
MISHRA 37 of 42

104. Huang NE, Shen Z, Long SR, Wu MC, Shih HH, Zheng Q, Yen NC, Tung CC, Liu HH. The empirical mode decomposition and the
Hilbert spectrum for nonlinear and non‐stationary time series analysis. In Proceedings of the Royal Society of London A: mathematical,
physical and engineering science, The Royal Society.s, 1998; 454(1971): 903‐995.
105. Shukla S, Mishra S, Singh B. Empirical‐mode decomposition with Hilbert transform for power‐quality assessment. IEEE Trans Power
Delivery. 2009;24(4):2159‐2165.
106. Norman CF, Chan JY, Lau WH, Lai LL. Hybrid wavelet and Hilbert transform with frequency‐shifting decomposition for power quality
analysis. IEEE Trans Instrum Meas. 2012;61(12):3225‐3233.
107. Kumar R, Singh B, Shahani DT. Recognition of single‐stage and multiple power quality events using Hilbert–Huang transform and prob-
abilistic neural network. Electr Power Compon Syst. 2015;43(6):607‐619.
108. Yang L, Yu J, Lai Y. Disturbance source identification of voltage sags based on Hilbert‐Huang transform. In Power and Energy Engineer-
ing Conference (APPEEC), 2010 Asia‐Pacific (pp. 1‐4). IEEE, 2010, 1‐4.
109. Manjula M, Mishra S, Sarma AVRS. Empirical mode decomposition with Hilbert transform for classification of voltage sag causes using
probabilistic neural network. Int J Electr Power Energy Syst. 2013;44(1):597‐603.
110. Ozgonenel O, Yalcin T, Guney I, Kurt U. A new classification for power quality events in distribution systems. Electr Pow Syst Res.
2013;95:192‐199.
111. Shukla S, Mishra S, Singh B. Power quality event classification under noisy conditions using EMD‐based de‐noising techniques. IEEE
Trans Ind Inf. 2014;10(2):1044‐1054.
112. Saxena D, Singh SN, Verma KS. and K. Singh, S. HHT‐based classification of composite power quality events. Int J Energy Sect Manage.
2014;8(2):146‐159.
113. Afroni MJ, Sutanto D, Stirling D. Analysis of nonstationary power‐quality waveforms using iterative Hilbert Huang transform and SAX
algorithm. v. 2013;28(4):2134‐2144.
114. Hafiz F, Chowdhury AH, Shahnaz C. December. An approach for classification of power quality disturbances based on Hilbert Huang
transform and Relevance vector machine. In Electrical & Computer Engineering (ICECE), 2012 7th International Conference on (pp.
201‐204). IEEE.2012: 201‐204.
115. Önal Y, Turhal ÜÇ. The orthogonal Hilbert‐Huang transform application in voltage flicker analysis. In Power Engineering, Energy and
Electrical Drives (POWERENG), 2013 Fourth International Conference on (pp. 700‐704).IEEE, 2013; 700‐704.
116. Zhan W, Xiangjun Z, Xiaoxi H, Jingying H. The multi‐disturbance complex power quality signal HHT detection technique. In Innovative
Smart Grid Technologies‐Asia (ISGT Asia), 2012 IEEE (pp. 1‐5).IEEE, 2012; 1‐5.
117. Huang Y, Liu Y, Hong Z. Detection and location of power quality disturbances based on mathematical morphology and Hilbert‐Huang
transform. In Electronic Measurement & Instruments, 2009.ICEMI'09. 9th International Conference on (pp. 2‐319). IEEE, 2009; 2‐319.
118. Senroy N, Suryanarayanan S, Ribeiro PF. An improved Hilbert–Huang method for analysis of time‐varying waveforms in power quality.
IEEE Trans Power Syst. 2007;22(4):1843‐1850.
119. Jayasree T, Devaraj D, Sukanesh R. Power quality disturbance classification using Hilbert transform and RBF networks.
Neurocomputing. 2010;73(7‐9):1451‐1456.
120. Peng LI, Jing GAO, Duo XU, Chang WANG, Xavier YANG. Hilbert‐Huang transform with adaptive waveform matching extension and
its application in power quality disturbance detection for microgrid. J Modern Power Syst Clean Energ. 2016;4(1):19‐27.
121. Feilat EA, Aljarrah RR, Rifai MB. Detection and classification of voltage variations using combined envelope‐neural network based
approach. Jordan J Electr Eng. 2017;3(2):112‐124.
122. Dragomiretskiy K, Zosso D. Variational mode decomposition. IEEE Trans Signal Process. 2014;62(3):531‐544.
123. Aneesh C, Kumar S, Hisham PM, Soman KP. Performance comparison of variational mode decomposition over empirical wavelet trans-
form for the classification of power quality disturbances using support vector machine. Procedia Comput Sci. 2015;46:372‐380.
124. Abdoos AA, Mianaei PK, Ghadikolaei MR. Combined VMD‐SVM based feature selection method for classification of power quality
events. Appl Soft Comput. 2016;38:637‐646.
125. Achlerkar PD, Samantaray SR, Manikandan MS. Variational mode decomposition and decision tree based detection and classification of
power quality disturbances in grid‐connected distributed generation system. IEEE Trans Smart Grid. 2018;9(4):3122‐3132.
126. Soman KP, Poornachandran P, Athira S, Harikumar K. Recursive variational mode decomposition algorithm for real time power signal
decomposition. Procedia Technol. 2015;21:540‐546.
127. Viswanath A, Jose KJ, Krishnan N, Kumar SS, Soman KP. Spike detection of disturbed power signal using VMD. Procedia Comput Sci.
2015;2015(46):1087‐1094.
128. Chakravorti T, Dash PK. Multiclass power quality events classification using variational mode decomposition with fast reduced kernel
extreme learning machine‐based feature selection. IET Sci Meas Technol. 2017;12(1):106‐117.
129. Sahani M, Dash PK. Variational mode decomposition and weighted online sequential extreme learning machine for power quality event
patterns recognition. Neurocomputing. 2018;310(8):10‐27.
130. Aneesh C, Hisham PM, Kumar S, Maya P, Soman KP. Variance based offline power disturbance signal classification using support vector
machine and random kitchen sink. Procedia Technol. 2015;21:163‐170.
38 of 42 MISHRA

131. Chen Q, Cai W. A algorithm of VMD for the detection of APF harmonics. In Automation (YAC), 2017 32nd Youth Academic Annual
Conference of Chinese Association of IEEE, 2017, May. ; 1260‐1263.
132. [MM0]Gautam S, Brahma SM. Overview of mathematical morphology in power systems—a tutorial approach. In Power & Energy Soci-
ety General Meeting, 2009. PES'09. IEEE (pp. 1‐7). IEEE, 2009, 1‐7.
133. Matheron G. Random Sets and Integral Geometry. New York: Wiley; 1975.
134. Serra J. Image Analysis and Mathematical Morphology. New York: Academic; 1982.
135. Lu Z, Turner DR, Wu QH, Fitch J, Mann S. Morphological transform for detection of power quality disturbances. In Power System
Technology, 2004. PowerCon 2004.2004 International Conference on IEEE, 2004; 2: 1644‐1649.
136. Matz V, Radil T, Ramos P, Serra AC. Automated power quality monitoring system for on‐line detection and classification of
disturbances. In Instrumentation and Measurement Technology Conference Proceedings, 2007. IMTC 2007. IEEE, 2007; 1‐6.
137. Xue C, Hui‐jin L, Quan‐ming Z, Li‐ming Y, Qing‐fen L. Power quality disturbances detection and location using mathematical morphol-
ogy and complex wavelet transformation. In Industrial Electronics and Applications, 2008. ICIEA 2008. 3rd IEEE Conference on IEEE,
2008; 2263‐2268.
138. Sen O, Zhen R. Application of improved mathematical morphology method in the power quality monitoring. In Power System Technol-
ogy, 2006. PowerCon 2006. International Conference on IEEE, 2006; 1‐6.
139. Li G, Zhou M, Luo Y, Ni Y. Power quality disturbance detection based on mathematical morphology and fractal technique. In Transmis-
sion and Distribution Conference and Exhibition: Asia and Pacific, 2005 IEEE/PES IEEE., 2005: 1‐6.
140. Huang Y, Liu Y, Hong Z. Detection and location of power quality disturbances based on mathematical morphology and Hilbert‐Huang
transform. In Electronic Measurement & Instruments, 2009. ICEMI'09. 9th International Conference on, 2009, August (pp. 2‐319). IEEE.
141. Chen H, Sun Y, Cheng Y. Harmonic and inter‐harmonic detection of grid‐connected distributed generation based on modified mathe-
matical morphology filter and Hilbert‐Huang transformation. In Power Electronics and Motion Control Conference, 2009. IPEMC'09.
IEEE 6th International, 2009, May, (pp. 1155‐1160). IEEE.
142. Zhang LL, Ji TY, Li MS, Wu QH, Jiang L, Zhan JP. Disturbance identification based on mathematical morphology and radial coordinate
visualization. In Innovative Smart Grid Technologies Europe (ISGT EUROPE), 2013 4th IEEE/PES, 2013, October (pp. 1‐5). IEEE.
143. Huang RH, Wang FH, Zhang J, Duan RC, 2014, April. Detection and location of voltage sags based on mathematical morphology and
grille fractal. In T&D Conference and Exposition, 2014 IEEE PES (pp. 1‐5). IEEE.
144. Chakravorti T, Dash PK. 2016, December. Morphology based fuzzy approach for detection & classification of simultanious power quality
disturbances. In India Conference (INDICON), 2016 IEEE Annual (pp. 1‐6). IEEE.
145. Chen Y, Ji TY, Li MS, Wu QH. 2015, November. Power quality disturbance detection based on morphology singular entropy. In Inno-
vative Smart Grid Technologies‐Asia (ISGT ASIA), 2015 IEEE, 2015; 1‐5.
146. Saputra ID, Smith JS, Jiang L, Wu QH, October. Detection and classification of power disturbances using half multi‐resolution morphol-
ogy gradient. In PES Innovative Smart Grid Technologies Conference Europe (ISGT‐Europe), 2016 IEEE IEEE, 2016; 1‐5.
147. Igna DS, Smith JS, Wu QH. Detection of power disturbances using mathematical morphology on small data windows. In Telecommu-
nications and Signal Processing (TSP), 2016 39th International Conference on IEEE, 2016; 211‐214.
148. Wang M, Mamishev AV. Classification of power quality events using optimal time‐frequency representations‐Part 1: theory. v.
2004;19(3):1488‐1495.
149. Wang M, Rowe GI, Mamishev AV. Classification of power quality events using optimal time‐frequency representations‐Part 2: applica-
tion. v. 2004;19(3):1496‐1503.
150. Hu GS, Zhu FF, Tu YJ. Power quality disturbance detection and classification using Chirplet transforms. In: Asia‐Pacific Conference on
Simulated Evolution and Learning. Berlin, Heidelberg: Springer; 2006:34, 34‐41, 41.
151. Janik P, Lobos T. Automated classification of power‐quality disturbances using SVM and RBF networks. v. 2006;21(3):1663‐1669.
152. Manikandan MS, Samantaray SR, Kamwa I. Detection and classification of power quality disturbances using sparse signal decomposi-
tion on hybrid dictionaries. IEEE Trans Instrum Meas. 2015;64(1):27‐38.
153. Hsieh CT, Lin JM, Huang SJ. Slant transform applied to electric power quality detection with field programmable gate array design
enhanced. Int J Electr Power Energy Syst. 2010;32(5):428‐432.
154. Subasi A, Yilmaz AS, Tufan K. Detection of generated and measured transient power quality events using Teager energy operator. Energ
Conver Manage. 2011;52(4):1959‐1967.
155. Ferreira DD, de Seixas JM, Cerqueira AS, Duque CA. Exploiting principal curves for power quality monitoring. Electr Pow Syst Res.
2013;100:1‐6.
156. de la Rosa JJG, Sierra‐Fernández JM, Agüera‐Pérez A, Palomares‐Salas JC, Moreno‐Muñoz A. An application of the spectral kurtosis to
characterize power quality events. Int J Electr Power Energy Syst. 2013;49:386‐398.
157. Kapoor R, Saini MK. Hybrid demodulation concept and harmonic analysis for single/multiple power quality events detection and clas-
sification. Int J Electr Power Energy Syst. 2011;33(10):1608‐1622.
158. Yang D, Wang B, Cai G. Adaptive local iterative filter decomposition for analysis of non‐stationary waveforms in power quality. Int
Trans Electr Energy Syst. 2017;27(12):e2426.
MISHRA 39 of 42

159. Hong W, Xu Y, Tang Y. Research on voltage sags detection method based on delay small angle dq transform. In Power System Technol-
ogy (POWERCON), 2014 International Conference on IEEE, 2014; (pp. 2221‐2227).
160. Borges FA, Fernandes RA, Silva IN, Silva CB. Feature extraction and power quality disturbances classification using smart meters sig-
nals. IEEE Trans Ind Inf. 2016;12(2):824‐833.
161. Kumar R, Singh B, Shahani DT. Symmetrical components‐based modified technique for power‐quality disturbances detection and clas-
sification. IEEE Trans Ind Appl. 2016;52(4):3443‐3450.
162. Camarena‐Martinez D, Valtierra‐Rodriguez M, Perez‐Ramirez CA, Amezquita‐Sanchez JP, de Jesus Romero‐Troncoso R, Garcia‐Perez
A. Novel downsampling empirical mode decomposition approach for power quality analysis. IEEE Trans Ind Electron. 2016;
63(4):2369‐2378.
163. Wang XJ, Li MS, Ji TY, Wu QH. An adaptive morphological lifting wavelet and its application on power disturbances detection. In
Power and Energy Society General Meeting (PESGM), 2016: 1‐5.
164. Sheshyekani K, Fallahi G, Hamzeh M, Kheradmandi M. A general noise‐resilient technique based on the matrix pencil method for the
assessment of harmonics and interharmonics in power systems. v. 2017;32(5):2179‐2188.
165. Zhao J, He Z, Qian Q. Detection of power quality disturbances based on generalized morphological filter and information theory. In
Power and Energy Engineering Conference, 2009. APPEEC 2009. Asia‐Pacific IEEE, 2009; 1‐4.
166. Panigrahi BK, Pandi VR. Optimal feature selection for classification of power quality disturbances using wavelet packet‐based fuzzy k‐
nearest neighbour algorithm. IET Gener Transm Distrib. 2009;3(3):296‐306.209.
167. Manimala K, Selvi K, Ahila R. Hybrid soft computing techniques for feature selection and parameter optimization in power quality data
mining. Appl Soft Comput. 2011;11(8):5485‐5497.210.
168. Manimala K, Selvi K, Ahila R. Optimization techniques for improving power quality data mining using wavelet packet based support
vector machine. Neurocomputing. 2012;77(1):36‐47.211.
169. Ray PK, Mohanty SR, Kishor N, Catalão JP. Optimal feature and decision tree‐based classification of power quality disturbances in dis-
tributed generation systems. IEEE Trans Sustainable Energy. 2014;5(1):200‐208.212.
170. Wang MH, Tseng YF. A novel analytic method of power quality using extension genetic algorithm and wavelet transform. Expert Syst
Appl. 2011;38(10):12491‐12496.213.
171. Biswal B, Dash PK, Panigrahi BK. Power quality disturbance classification using fuzzy C‐means algorithm and adaptive particle swarm
optimization. IEEE Trans Ind Electron. 2009;56(1):212‐220.217.
172. Liao CC. Enhanced RBF network for recognizing noise‐riding power quality events. IEEE Trans Instrum Meas. 2010;59(6):1550‐1561.214.
173. Hooshmand R, Enshaee A. Detection and classification of single and combined power quality disturbances using fuzzy systems oriented
by particle swarm optimization algorithm. Electr Pow Syst Res. 2010;80(12):1552‐1561.216.
174. Hajian M, Foroud AA. A new hybrid pattern recognition scheme for automatic discrimination of power quality disturbances. Measure-
ment. 2014;51:265‐280.218.
175. Hajian M, Foroud AA, Abdoos AA. New automated power quality recognition system for online/offline monitoring. Neurocomputing.
2014;128:389‐406.224.
176. Biswal B, Dash PK, Mishra S. A hybrid ant colony optimization technique for power signal pattern classification. Expert Syst Appl.
2011;38(5):6368‐6375.219.
177. Jamali S, Farsa AR, Ghaffarzadeh N. Identification of optimal features for fast and accurate classification of power quality disturbances.
Measurement. 2018;116:565‐574.220.
178. Lee CH, Nam SW. Efficient feature vector extraction for automatic classification of power quality disturbances. Electron Lett.
1998;34(11):1059‐1061.
179. Monedero I, Leon C, Ropero J, Garcia A, Elena JM, Montano JC. Classification of electrical disturbances in real time using neural net-
works. v. 2007;22(3):1288‐1296.
180. Uyar M, Yildirim S, Gencoglu MT. An expert system based on S‐transform and neural network for automatic classification of power
quality disturbances. Expert Syst Appl. 2009;36(3):5962‐5975.
181. Ghosh AK, Lubkeman DL. The classification of power system disturbance waveforms using a neural network approach. v.
1995;10(1):109‐115.
182. He Z, Zhang H, Zhao J, Qian Q. Classification of power quality disturbances using quantum neural network and DS evidence fusion. Eur
T Electr Power. 2012;22(4):533‐547.
183. Santoso S, Powers EJ, Grady WM, Parsons AC. Power quality disturbance waveform recognition using wavelet‐based neural classifier. I.
Theoretical foundation. v. 2000;15(1):222‐228.
184. Liu Z, Zhang Q, Han Z, Chen G. A new classification method for transient power quality combining spectral kurtosis with neural net-
work. Neurocomputing. 2014;125:95‐101.
185. Valtierra‐Rodriguez M, de Jesus Romero‐Troncoso R, Osornio‐Rios RA, Garcia‐Perez A. Detection and classification of single and com-
bined power quality disturbances using neural networks. IEEE Trans Ind Electron. 2014;61(5):2473‐2482.
40 of 42 MISHRA

186. Gaing ZL. Wavelet‐based neural network for power disturbance recognition and classification. IEEE Trans Power Delivery.
2004;19(4):1560‐1568.
187. Huang N, Xu D, Liu X, Lin L. Power quality disturbances classification based on S‐transform and probabilistic neural network.
Neurocomputing. 2012;98:12‐23.
188. Lee CY, Shen YX. Optimal feature selection for power‐quality disturbances classification. v. 2011;26(4):2342‐2351.
189. Biswal B, Dash PK, Panigrahi BK, Reddy JBV. Power signal classification using dynamic wavelet network. Appl Soft Comput.
2009;9(1):118‐125.
190. Naik CA, Kundu P. Power quality disturbance classification employing S‐transform and three‐module artificial neural network. Int
Trans Electr Energy Syst. 2014;24(9):1301‐1322.
191. Dash PK, Nayak M, Senapati MR, Lee IW. Mining for similarities in time series data using wavelet‐based feature vectors and neural net-
works. Eng Appl Artif Intel. 2007;20(2):185‐201.
192. Oleskovicz M, Coury DV, Felho OD, Usida WF, Carneiro AA, Pires LR. Power quality analysis applying a hybrid methodology with
wavelet transforms and neural networks. Int J Electr Power Energy Syst. 2009;31(5):206‐212.
193. Khadse CB, Chaudhari MA, Borghate VB. Conjugate gradient back‐propagation based artificial neural network for real time power qual-
ity assessment. Int J Electr Power Energy Syst. 2016;82:197‐206.
194. Elango MK, Loganathan K. Classification of power quality disturbances using Stockwell transform and back propagation algorithm. In
Emerging Technological Trends (ICETT), International Conference on IEEE, 2016; 1‐7.
195. Wang M, Zhou H, Yang S, Jin L, Jiao L. Robust compressive features based power quality events classification with analog–digital
mixing network (ADMN). Neurocomputing. 2016;171:685‐692.
196. Muthusamy TA, Ramanathan N. An expert system based on least mean square and neural network for classification of power system
disturbances. 2018;1:308‐313.
197. Vapnik V. The Nature of Statistical Learning Theory. New York: Springer science & business media; 2013. https://doi.org/10.1007/978‐1‐
4757‐3264‐1
198. De Yong D, Bhowmik S, Magnago F. An effective power quality classifier using wavelet transform and support vector machines. Expert
Syst Appl. 2015;42(15‐16):6075‐6081.
199. Lin WM, Wu CH, Lin CH, Cheng FS. Detection and classification of multiple power‐quality disturbances with wavelet multiclass SVM.
IEEE Trans Power Delivery. 2008;23(4):2575‐2582.
200. Erişti H, Demir Y. A new algorithm for automatic classification of power quality events based on wavelet transform and SVM. Expert
Syst Appl. 2010;37(6):4094‐4102.
201. Hu GS, Zhu FF, Ren Z. Power quality disturbance identification using wavelet packet energy entropy and weighted support vector
machines. Expert Syst Appl. 2008;35(1‐2):143‐149.
202. Erişti H, Uçar A, Demir Y. Wavelet‐based feature extraction and selection for classification of power system disturbances using support
vector machines. Electr Pow Syst Res. 2010;80(7):743‐752.
203. Erişti H, Demir Y. Automatic classification of power quality events and disturbances using wavelet transform and support vector
machines. IET Gener Transm Distrib. 2012;6(10):968‐976.
204. Moravej Z, Abdoos AA, Pazoki M. Detection and classification of power quality disturbances using wavelet transform and support vector
machines. Electr Power Compon Syst. 2009;38(2):182‐196.
205. Mishra M, Rout PK, Routray SK, Nayak N. Power quality disturbance recognition using hybrid signal processing and machine intelli-
gence techniques. Int J Ind Electron Drives. 2014;1(2):91‐104.
206. Biswal B, Biswal MK, Dash PK, Mishra S. Power quality event characterization using support vector machine and optimization using
advanced immune algorithm. Neurocomputing. 2013;103:75‐86.
207. Kapoor R, Gupta R, Jha S, Kumar R. Detection of power quality event using histogram of oriented gradients and support vector machine.
Measurement. 2018;120:52‐75.
208. Dash PK, Mishra S, Salama MA, Liew AC. Classification of power system disturbances using a fuzzy expert system and a Fourier linear
combiner. IEEE Trans Power Delivery. 2000;15(2):472‐477.
209. Meher SK, Pradhan AK. Fuzzy classifiers for power quality events analysis. Electr Pow Syst Res. 2010;80(1):71‐76.
210. Decanini JG, Tonelli‐Neto MS, Malange FC, Minussi CR. Detection and classification of voltage disturbances using a fuzzy‐ARTMAP‐
wavelet network. Electr Pow Syst Res. 2011;81(12):2057‐2065.
211. Mahela OP, Shaik AG. Recognition of power quality disturbances using S‐transform based ruled decision tree and fuzzy C‐means clus-
tering classifiers. Appl Soft Comput. 2017;59:243‐257.
212. Huang J, Negnevitsky M, Nguyen DT. A neural‐fuzzy classifier for recognition of power quality disturbances. v. 2002;17(2):609‐616.
213. Reaz MBI, Choong F, Sulaiman MS, Mohd‐Yasin F, Kamada M. Expert system for power quality disturbance classifier. v.
2007;22(3):1979‐1988.
214. Liao CC, Yang HT. Recognizing noise‐influenced power quality events with integrated feature extraction and neuro‐fuzzy network. v.
2009;24(4):2132‐2141.
MISHRA 41 of 42

215. Pires VF, Amaral TG, Martins JF. Power quality disturbances classification using the 3‐D space representation and PCA based neuro‐
fuzzy approach. Expert Syst Appl. 2011;38(9):11911‐11917.
216. Huang GB, Zhu QY, Siew CK. Extreme learning machine: theory and applications. Neurocomputing. 2016;70(1‐3):489‐501.
217. Sahani M, Dash PK. Automatic power quality events recognition based on Hilbert Huang transform and weighted bidirectional extreme
learning machine. IEEE Trans Ind Inf. 2018;14(9):3849‐3858.
218. Ahila R, Sadasivam V, Manimala K. An integrated PSO for parameter determination and feature selection of ELM and its application in
classification of power system disturbances. Appl Soft Comput. 2015;32:23‐37.
219. Ma J, Zhang J, Xiao L, Chen K, Wu J. Classification of power quality disturbances via deep learning. IETE Tech Rev. 2017;
34(4):408‐415.
220. Liu H, Hussain F, Shen Y, Arif S, Nazir A, Abubakar M. Complex power quality disturbances classification via curvelet transform and
deep learning. Electr Pow Syst Res. 2018;163:1‐9.
221. James JQ, Hou Y, Lam AY, Li VO. Intelligent fault detection scheme for microgrids with wavelet‐based deep neural networks. IEEE
Trans Smart Grid. 2017.
222. Balouji E, Gu IY, Bollen MH, Bagheri A, Nazari M. 2018, May. A LSTM‐based deep learning method with application to voltage dip clas-
sification. In Harmonics and Quality of Power (ICHQP), 2018 18th International Conference on (pp. 1‐5). IEEE.
223. Mohan N, Soman KP, Vinayakumar R. 2017, December. Deep power: deep learning architectures for power quality disturbances classi-
fication. In 2017 International Conference on Technological Advancements in Power and Energy (TAP Energy) (pp. 1‐6). IEEE.
224. Lazzaretti AE, Tax DMJ, Neto HV, Ferreira VH. Novelty detection and multi‐class classification in power distribution voltage wave-
forms. Expert Syst Appl. 2016;45:322‐330.
225. Pan D, Zhao Z, Zhang L, Tang C. Recursive clustering K‐nearest neighbors algorithm and the application in the classification of power
quality disturbances. In Energy Internet and Energy System Integration (EI2), 2017 IEEE Conference on IEEE, 2017; 1‐5.
226. Abdel‐Galil TK, El‐Saadany EF, Youssef AM, Salama MMA. Disturbance classification using hidden Markov models and vector quan-
tization. IEEE Trans Power Delivery. 2005;20(3):2129‐2135.
227. Rajeshbabu S, Manikandan BV. Detection and classification of power quality events by expert system using analytic hierarchy method.
Cogn Syst Res. 2018;52:729‐740.
228. Saini MK, Beniwal RK. Recognition of multiple PQ issues using modified EMD and neural network classifier. Iranian J Electr Electron
Eng. 2018;14(2):188‐203.
229. Reddy MV, Sodhi R. A modified S‐transform and random forests‐based power quality assessment framework. IEEE Trans Instrum Meas.
2018;67(1):78‐89.
230. Moravej Z, Pazoki M, Abdoos AA. Wavelet transform and multi‐class relevance vector machines based recognition and classification of
power quality disturbances. Eur T Electr Power. 2011;21(1):212‐222.
231. Dash PK, Padhee M, Barik SK. Estimation of power quality indices in distributed generation systems during power islanding conditions.
Int J Electr Power Energy Syst. 2012;36(1):18‐30.
232. Martinez JA, Martin‐Arnedo J. July. Impact of distributed generation on distribution protection and power quality. In Power & Energy
Society General Meeting, 2009. PES'09. IEEE, 2009: 1‐6.
233. Demirci T, Kalaycıoglu A, Küçük D, et al. Nationwide real‐time monitoring system for electrical quantities and power quality of the elec-
tricity transmission system. IET Gener Transm Distrib. 2011;5(5):540‐550.
234. Chakravorti T, Patnaik RK, Dash PK. October. A morphological filter based disturbance detection and classification technique for
DFIG wind farm based microgrid. In Power, Communication and Information Technology Conference (PCITC), 2015 IEEE, 2015:
979‐985.
235. Chakravorti T, Patnaik RK, Dash PK. Detection and classification of islanding and power quality disturbances in microgrid using hybrid
signal processing and data mining techniques. IET Signal Proc. 2017;12(1):82‐94.
236. Mishra M, Sahani M, Rout PK. An islanding detection algorithm for distributed generation based on Hilbert–Huang transform and
extreme learning machine. Sustain Energy, Grids Networks. 2017;9:13‐26.
237. Moravej Z, Pazoki M, Niasati M, Abdoos AA. A hybrid intelligence approach for power quality disturbances detection and classification.
Int Trans Electr Energy Syst. 2013;23(7):914‐929.
238. Naderian S, Salemnia A. An implementation of type‐2 fuzzy kernel based support vector machine algorithm for power quality events
classification. Int Trans Electr Energy Syst. 2017;27(5):e2303.
239. Ribeiro EG, Mendes TM, Dias GL, et al. Real‐time system for automatic detection and classification of single and multiple power quality
disturbances. Measurement. 2018;128:276‐283.
240. Ucar F, Alcin OF, Dandil B, Ata F, Cordova J, Arghandeh R. Online power quality events detection using weighted extreme learning
machine. In 2018 6th International Istanbul Smart Grids and Cities Congress and Fair (ICSG) 2018, April, (pp. 39‐43). IEEE.
241. do Amaral Monteiro D, Zvietcovich WG, Braga MF. Detection and classification of power quality disturbances with wavelet transform,
decision tree algorithm and support vector machines. In 2018 Simposio Brasileiro de Sistemas Eletricos (SBSE), 2018, May, (pp. 1‐6).
IEEE.
42 of 42 MISHRA

242. Mishra M, Rout PK. Fast discrete s‐transform and extreme learning machine based approach to islanding detection in grid‐connected
distributed generation. Energy Syst. 2018. https://doi.org/10.1007/s12667‐018‐0285‐9

How to cite this article: Mishra M. Power quality disturbance detection and classification using signal
processing and soft computing techniques: A comprehensive review. Int Trans Electr Energ Syst. 2019;29:e12008.
https://doi.org/10.1002/2050‐7038.12008

You might also like