A review on power quality disturbance classification using deep learning approach

A REVIEW ON POWER QUALITY
DISTURBANCE CLASSIFICATION USING
DEEP LEARNING APPROACH
Muskan Rath
Computer Science and Engineering Dept
International Institute of Information Technology, Bhubaneswar
Abstract—Power Quality, a significant branch of power system
engineering, plays a crucial role in maintaining the power quality
supplied to consumers in the industry. The introduction of smart
grids further differentiates the significance of power output. A
single incident in power quality such as voltage drop triggered
by a transmission or distribution level failure will cost up to
millions of monetary losses for the affected industries. Power
Quality disturbances can be classified into Voltage sag, Voltage
Swell, Transient, Harmonic, Voltage Notch and Flicker. With the
help of digital techniques, at present, Power Quality disturbances
are tracked on-site and online. The primary objective of the
paper is to provide a thorough overview of the approaches in
deep learning for the automatic detection, identification and
classification of Power Quality Events, related to academics
following a line of investigation in the related area. The paper
further more gives insight on which of the techniques yields the
highest accuracy.
Keywords: Power Quality, Power Quality Disturbance, Deep
Learning, Convolutional Neural Network (CNN), Continuous
Wavelet transform, Sparse auto encoder, Stacked auto encoder
(SAE), Data classification, Multi-layer Convolutional neural Net-
work (MLCNN)
Index Terms—component, formatting, style, styling, insert
INTRODUCTION
Power Quality, in general, refers to the ability of electrical
equipment to consume the energy being supplied to it, ac-
cording to Captech. It has become a key concern as serious
problems begin to emerge that impact renewable energy,
energy efficiency and the atmosphere. Distributed generation
(DG) focused on alternate energy sources and traditional grids
is a challenge since it utilizes advanced power electronics
for monitoring, heavy non-linear loads, and microprocessor
and machine solutions [1-3]. There are available real-time
commercial PQ analyzer solutions. Fluke (Everett, WA, USA),
Yokogawa (Tokyo, Japan), and FLIR (Wilsonville, OR, USA)
are some of the principal manufacturers of PQ analyzers. The
simple functionality of PQ analyzer solutions is costly, but
the complex and comprehensive data can not be analyzed[4].
Due to fluctuations and loads, which alter the capacity of the
signals, non-stationary PQ disturbances happen. Disturbance
Identify applicable funding agency here. If none, delete this.
in power quality refers to the deviation exhibited in voltage
waveform, current waveform or natural frequency of the sys-
tem. In the power generation stage, disturbances can be caused
by renewable power suppliers like solar power and wind
power because of their inherent characteristic of oscillation,
arbitrariness, and irregularity. PQ disruptions may be due
to a rapid shift in frequency, amplitude, current and phase
angle. The output of renewable sources and the converters are
the main cause of disturbances leading to power unevenness
and voltage instabilities. In the transmission stage, the use
of battery-operated storages, the extensive usage of power
electronics devices like solid-state switches and converters
have severely increased the harmonic pollution. In the con-
sumption stage, the increased use of asynchronous motors,
electronic measurement devices, control devices, nonlinear
equipment and action of these tools decrease the quality of
the power supplied. Additionally, the synchronization and
optimum operation of local micro-grid systems overlap with
each other creating complex disturbances. We can solve this
problem by automated classification and identification of PQ
disturbances with appropriate methods[5-8]. Generally, the
process of identification of PQ disturbances consists of feature
extraction, feature selection, and classification[9]. The primary
types of power quality disturbances are Voltage sag, Swell,
Notch, Spike, Harmonics, Flickers and Interruption [10]. In
order to detect the power system disturbances can be detected
by employing several signal processing techniques like Fourier
transform (FT), Short-Time Fourier Transform (STFT), Dis-
crete Fourier Transform (DFT), Fast Fourier Transform (FFT),
Wavelet Transform (WT) for extraction, identification of fea-
tures. FT and STFT have been applied only to stationary
signals, STFT is an extension of DFT which does not apply
to non-stationary signals due to the fixed window size used.
The S-Transform (ST) also acts as a valuable tool for the
analysis of PQ disturbance; it blends WT with STFT and
provides a key mechanism for locating time frequencies [17].
A redundant representation of the time-frequency area[18] is
a significant downside to the S-Transform. Similarly, several
techniques for the analyzer of PQ are being studied, such

as Kalman Filter (KF), Discrete Orthogonal ST (DOST),
Curvelet Transform (CT), Hilbert-Huang Transform (HHT),
Singular Value Decomposition (SVD), Wigner distribution
function (WDF), Gabor transforms (GT). Principal component
analysis (PCA) is a basic multivariate statistical approach
to finding a set of projection vectors that maximize data
variance [19-29]. The key objective of PCA is Data reduction
and Interpretation. The improved version of PCA is IPCA,
which is similar to PCA. Some of the applications of PCA
are face recognition, hyper spectral images, and electro car-
diogram [30-33]. PCA is the most suitable decomposition
technique for feature extraction and automatic detection of
PQ Disturbances [34-36].An essential part of power quality
classification is feature selection and classification[37]. Neural
network classifiers such as Probabilistic Neural network and
artificial neural network have already been investigated for PQ
Classification [38-41]. Although with respect to classification
accuracy and computational cost, Support vector machine
(SVM), fast extreme learning machine (FELM) and nearest
centroid neighbor (NCN) are more dominant techniques to
the traditional NN [42-44]; in deep learning, CNN is found
the most accepted one and successfully applied to the hyper
spectral image, large-scale audio face recognition and image
classification [45–48].
LITERATURE REVIEW
In this paper, we are discussing in brief, the various models
proposed for Power Quality Disturbances using Deep Learning
Approach and on the basis of certain criterion, conclude
which method is the best. There are various models proposed
for PQ Disturbance Classification using Deep Learning, of
which we will be discussing three major methods [49-51].
By now, we are aware of the fact that perceiving and catego-
rizing Power Quality Disturbances can help in resolving the
related risks. Unlike in traditional approach, which divides
the process into independent stages, which include, signal
analysis, feature extraction, optimal feature selection, and
finally classification on the basis of those selected features.
In the model proposed by Ebrahim Balouji and Ozgul Salor
the approach of using deep neural network for Power Quality
event the researchers described classifications and recognition,
collected reference work from LeNet-5, googlenet, ImageNet
classification with deep convolutional neural networks by
Krizhevsky, A., Sutskever, I., Hinton and CNN all these
almost the same format or type of weight convolutional layers
and they are tracked by one or more fully connected layer. The
architecture of the model used had a standard neural network
that gets a single vector as the input and also have the full
connection to all the neurons of the previous layer, neurons
with single independent layer function. This provides the class
classification score termed as Weight(W). When the amount
of weight is not marginable then it would be obtained in the
first layer. In order to avoid over fitting problems , Multi-label
convolutional neural network (MLCNN) is used as the deep
learning method . MaxPooling and Convolution are the two
functions that were used by the researchers in this Multi-label
convolutional neural network. The deep learning platform that
was used for this purpose was DIGITS deep learning platform
of NVIDIA [49]. In [50], a Convolutional Neural Network
(CNN) model based on deep learning methodology has been
proposed to replace the traditional models. This paper, by
following IEEE-1159 standard, has simulated a total of 16
classes of disturbances, including combined disturbances. The
salient features of paper [49] are as follows:
I. VARIOUS MATHEMATICAL MODELS USED ARE AS
FOLLOWS:
• Power Quality Disturbances: This type of disturbance
is most common is multi energy system and micro-grids.
There are a total of 16 disturbances that are taken into
consideration in this review. The following disturbances
considered here are regular pure sinusoidal, single type
disturbance like voltage swell, voltage sag, flicker, in-
terruption, oscillatory transient and complex disturbances
like interruption with harmonics, sag with harmonics,
swell with harmonics, flicker with harmonics, sag with
swell, sag with interruption, swell with interruption, sag
with interruption and harmonics, swell with interruption
and harmonics, flicker with interruption and harmonics.
• Continuous Wavelet Transform (CWT): In this paper,
CWT is used to convert the 1-D signal information to
2-D image representation to provide as the input to the
proposed CNN model.
• Data-set Preparation: The model here requires a large
data set hence this was observed using MATLAB.The
constraints are made according to the IEEE standards
1159 [1].
A. Description of the layers in proposed model
• 2-D convolution layers: The output of this layer depends
on the activation function used. This returns the max
value of the triggering function.
• Pooling Layer: This type of layer is used to minimize
its dimension and extract distinct features.
• Cross-channel normalization layer: It substitutes every
element with a standardized value obtained from the
parameters of several adjacent channels
• Dense layer: This layer consists of inter linkage of
multiple neurons.
• Soft-max Layer: This layer is fitted with the triggering
function called ‘softmax’.This ensures that the number of
neurons are equal to the number of output class.
In [51], the deep-learning framework, Stacked auto en-
coder was used in order to extract high level features for
Power quality disturbance classification. This paper addresses
the previously unsolved issue regarding selection of optimal
feature for PQD. The additional tools, which assisted this,
were variance of signals and Particle Swarm Optimization
Algorithm. A 10-fold cross-validation was done to validate
effectiveness of the proposed classification scheme; In other
words, the given approach is noise resistant and achieves
the required classification accuracy rate. The paper helps in

overcoming the limitation like difficulty in selection of feature
candidates for forming an optimal feature that arises due to
other methods. The existing methods were useful only in
the subsequent classification and could not be completely
reconstruct the original signal, which was a major drawback.
There were speculations of other features, possibly ignored,
which could have contributed to the classification. So, in this
paper, a profound learning-based strategy is brought into the
arrangement of intensity quality aggravations (PQDs). Stacked
auto encoder, as a learning system, is utilized to remove ele-
vated levels of PQDs. In this unique circumstance, a formerly
unsolved issue with respect to ideal highlights’ determination
for PQDs can be tended to. The input data can be collected
without a class label in the self-supervised learning procedure,
which enable reconstruction of input signal by extraction of
features by the sparse auto encoder. No information is omitted
from the input data, which makes feature extraction more
effective as compared to other proposed methods. Moreover,
if a deep-learning classification method is used, the difficulty
in selection of optimal feature can be overcome. This helps
in avoiding difficulty of feature selection and extraction, and
achieves better classification rates than the other methods.
Sparse Auto encoder has been used to extract the features
in each hidden layer of the network. The first section of the
model discusses in brief the concepts of sparse auto encoder
and stacked auto encoder and makes a comparative study of the
two. The second section describes in detail the various PQD
classification schemes and finally, evaluates and displays the
experimental results. A sparse auto encoder is an unsupervised
learning framework of the neural network; comprising of Input
Layer, a hidden layer and an output layer, in which the output
layer is the same as the input layer. The input data is the
activation unit for the first layer. By applying back propagation
algorithm, the output values are enabled to gradually approach
the inputs, with the aim to minimize the distance between the
inputs and the outputs. A sparsity constraint was introduced for
desired sparsity in the hidden layers, which forces the hidden
layers to remain as inactive as possible. In each iteration, the
contribution of every unit in the hidden layer is calculated
and the partial derivatives of the cost function with respect W
and b compared, where W is weight parameter and b refers
to bias term. It can be derived from the above procedural
computations that limited memory quasi-Newton method (L-
BFGS) has been adopted in the given paper. From the second
section of the model, we can find that the main goal is
to minimize the error between the input and reconstructed
signals. For designing the experiments, MATLAB m-files were
used to randomly generate disturbances in power systems
by varying the various parameters in the equations. The
researchers took the fundamental frequency of PQDs 50 Hz,
sampling frequency as 3200 Hz and length of each signal as
10 cycles and got the waveform for the seven signals. The
PQDs data generated were two thousand disturbance signals
of each type. All the data are labeled based on their types; the
types like sag, interruption, swell, harmonic, oscillation, sag
C harmonic , and swell C harmonic were specifically labeled
with label pairs (1,1), (1,2), (1,3), (2,0), (3,0), (4,0), and
(5,0), respectively. A 10-fold cross-validation has been done to
validate the effectiveness of the proposed method. The PQDs
data are divided into 10 equal-sized parts, in which each of the
10 parts was kept to be the test set and the remaining 9 parts
used as training data and finally, the average of 10 groups of
results were taken to produce a single estimation. Two hidden
layers were set-up for the architecture of SAE, each having
two hundred neurons. There were 640 neurons of the input
layer. Some parameter values like sparsity parameter p=0.08 ,
weight decay parameter(set to 0.003) and weight were sparsity
penalty(3) were redefined. The weight parameter and the bias
term were initialized before training the neural network (of
SAE), which acted like a mapping function. All bias terms
were initialized to 0 and each weight parameter was initialized
randomly to avoid all neurons of hidden layers learning the
same mapping function. The number of iterations in the PSO
algorithm was set to 300 and the inertia weight as 0.7982
and the acceleration factors as r1=r2=0.2 The conclusions that
were drawn from the proposed methods in [49-51] were as
follows:
• From [49], it was found that 85% of the data is used
for training in the procedure of MLCNN and the rest
15% was used for testing. Accuracy and loss functions
were used to accurately calculate number of epochs
in NVIDIA’s DIGITS platform. In terms of learning
rate, it was observed that epoch number between few
intervals were critical but after increasing beyond certain
level it became useless as accuracy and loss remained
approximately same. Classification of Interruption events
and Discriminating dip events with different amplitudes
and duration, obtained by Deep Learning corresponds
to almost 100% accuracy. These classifications were
compared with labels assigned by the Human Expert and
it was observed to be 100% accurate.
• Now,in [50], after the training set was input, we observed
that the output has the net Average accuracy of 99.6%
whilst the loss was close to 0.01%.For different type
of noise disturbance like noise-40dB, noise-30dB, and
noise-20dB, the model resulted 99.75%, 99.59% and
99.22% respectively for noise less environment a 100%
was observed.
• It was concluded from the experiment in [51] that deep
learning-based method could be exploited to classify
the types of PQDs; the classification accuracy were
high, even with an increase in the noise intensity. Deep
learning-based classification can resist the interference of
noise to some extent, by employing data compression in
which the process of mapping the input layer into the
hidden layer, at each level of the sparse auto encoder.
It suppressed the effects of noise in the input data to a
large extent. Thus, it can be concluded that the sparse
auto-encoder has a robustness against noise; by using
an SAE algorithm, the classification rates are as high as
99.75%, 99.607%, and 98.529% for noiseless, 30 and 20

dB signals, respectively.
II. CONCLUSION AND FUTURE SCOPE
It is observed that the proposed method in [50] is better
than many traditional methods having a net average accuracy
of 99.6% if much better CNN is used it can be improved by
replacing the traditional layers used. As of this model this
gives a better solution for handling the vast amount of real-
time monitored data for analysis of power quality disturbance
occurring in the power network.
REFERENCES
[1] Jiang, J.N.; Tang, C.Y.; Ramakumar, R.G. Control and Operation of
Grid-Connected Wind Farms: Major Issues, Contemporary Solutions,
and Open Challenges; Springer: Berlin, Germany, 2016.
[2] . Mundaca, L.; Neij, L.; Markandya, A.; Hennicke, P.; Yan, J. Towards
a Green Energy Economy? Assessing Policy Choices, Strategies and
Transitional Pathways; Elsevier: Amsterdam, The Netherlands, 2016.
[3] . Lin, B.; Liu, Y.; Wang, Z.; Pei, Z.; Davies, M. Measured energy use and
indoor environment quality in green office buildings in China. Energy
Build. 2016, 129, 9–18.
[4] Saini, K.M.; Beniwal, R.K. Detection and classification of power quality
disturbances in wind-grid integrated system using fast time-time trans-
form and small residual-extreme learning machine. Int. Trans. Electr.
Energy Syst. 2018, 28, e2519.
[5] Liu, H.; Hussain, F.; Shen, Y. Power quality disturbances classification
using compressive sensing and maximum likelihood. IETE Tech. Rev.
2017, 35, 359–368.
[6] . Shen, Y.; Hussain, F.; Shen, Y. Power quality disturbances classification
based on curvelet transform. Int. J. Comput. Appl. 2017, 40, 192–201.
[7] Niitsoo, J.; Jarkovoi, M.; Taklaja, P.; Klüss, J.; Palu, I. Power quality
issues concerning photovoltaic generation in distribution grids. Smart
Grid Renew. Energy 2015, 6, 148.
[8] Bollen, H.M.; Gu, I.Y. Signal Processing of Power Quality Disturbances;
John Wiley Sons: Hoboken, NJ, USA, 2006; Volume 30.
[9] Lee, C.-Y.; Shen, Y.-X. Optimal feature selection for power-quality dis-
turbances classification. IEEE Trans. Power Deliv. 2011, 26, 2342–2351.
[10] Reid, W.E. Power quality issues-standards and guidelines. IEEE Trans.
Ind. Appl. 1996, 32, 625–632.
[11] Heydt, G.; Fjeld, P.; Liu, C.; Pierce, D.; Tu, L.; Hensley, G. Applications
of the windowed FFT to electric power quality assessment. IEEE Trans.
Power Deliv. 1999, 14, 1411–1416.
[12] Liao, C.-C.; Yang, H.-T.; Chang, H.-H. Denoising techniques with
a spatial noise-suppression method for wavelet-based power quality
monitoring. IEEE Trans. Instrum. Meas. 2011, 60, 1986–1996.
[13] Poisson, O.; Rioual, P.; Meunier, M. Detection and measurement of
power quality disturbances using wavelet transform. IEEE Trans. Power
Deliv. 2000, 15, 1039–1044.
[14] Jurado, F.; Saenz, J.R. Comparison between discrete STFT and wavelets
for the analysis of power quality events. Electr. Power Syst. Res. 2002,
62, 183–190.
[15] Santoso, S.; Powers, E.J.; Grady, W.M.; Hofmann, P. Power quality
assessment via wavelet transform analysis. IEEE Trans. Power Deliv.
1996, 11, 924–930. b16 Morsi, W.G.; El-Hawary, M. Novel power
quality indices based on wavelet packet transform for non-stationary
sinusoidal and non-sinusoidal disturbances. Electr. Power Syst. Res.
2010, 80, 753–759.
[16] Dash, P.; Panigrahi, G.P.B. Power quality analysis using S-transform.
IEEE Trans. Power Deliv. 2003, 18, 406–411.
[17] Stockwell, R.G.; Mansinha, L.; Lowe, R. Localization of the complex
spectrum: The S transform. IEEE Trans. Signal Process. 1996, 44, 998–
1001.
[18] Dash, P.; Chilukuri, M. Hybrid S-transform and Kalman filtering ap-
proach for detection and measurement of short duration disturbances in
power networks. IEEE Trans. Instrum. Meas. 2004, 53, 588–596.
[19] Reddy, M.J.B.; Raghupathy, R.K.; Venkatesh, K.; Mohanta, D. Power
quality analysis using Discrete Orthogonal S-transform (DOST). Digit.
Signal Process. 2013, 23, 616–626.
[20] Liu, H.; Hussain, F.; Shen, Y.; Arif, S.; Nazir, A.; Abubakar, M. Complex
power quality disturbances classification via curvelet transform and deep
learning. Electr. Power Syst. Res. 2018, 163, 1–9.
[21] Shukla, S.; Mishra, S.; Singh, B. Empirical-mode decomposition with
Hilbert transform for power-quality assessment. IEEE Trans. Power
Deliv. 2009,
[22] 2159–2165.
[23] Li, T.-Y.; Zhao, Y.; Nan, L.; Fen, G.; Gao, H.-H. A new method for
power quality detection based on HHT. Zhongguo Dianji Gongcheng
Xuebao Proc. Chin. Soc. Electr. Eng. 2005, 25, 52–56. 24. Ozgonenel,
O.; Yalcin, T.; Guney, I.; Kurt, U. A new classification for power
quality events in distribution systems. Electr. Power Syst. Res. 2013,
95, 192–199.
[24] Cho, S.-H.; Jang, G.; Kwon, S.-H. Time-frequency analysis of power-
quality disturbances via the Gabor–Wigner transform. IEEE Trans.
Power Deliv. 2010, 25, 494–499.
[25] Abdullah, A.R.; Sha’ameri, A.Z.; Saad, N.M. Asia-Pacific Conference
on Power quality analysis using spectrogram and gabor transformation.
In Proceedings of the 2007 Asia-Pacific Conference on Applied Elec-
tromagnetics, Melaka, Malaysia, 4–6 December 2007.
[26] Manikandan, M.S.; Samantaray, S.; Kamwa, I. Detection and classifi-
cation of power quality disturbances using sparse signal decomposition
on hybrid dictionaries. IEEE Trans. Instrum. Meas. 2015, 64, 27–38.
[27] Lopez-Ramirez, M.; Ledesma-Carrillo, L.; Cabal-Yepez, E.; Rodriguez-
Donate, C.; Miranda-Vidales, H.; Garcia-Perez, A. EMD-based feature
extraction for power quality disturbance classification using moments.
Energies 2016, 9, 565
[28] Smith, L.I. A Tutorial on Principal Components Analysis; Technical
Report OUCS: Dunedin, Otago, New Zealand, 26 February 2002.
[29] Chawla, M.; Verma, H.; Kumar, V. ECG Modeling and QRS Detection
Using Principal Component Analysis. In Proceedings of the IET 3rd
International Conference MEDSIP 2006, Advances in Medical, Signal
and Information Processing, Glasgow, UK, 17–19 July 2006.
[30] Li, W.; Shi, T.; Liao, G.; Yang, S. Feature extraction and classification
of gear faults using principal component analysis. J. Qual. Maint. Eng.
2003, 9, 132–143.
[31] Moon, H.; Phillips, P.J. Computational and performance aspects of PCA-
based face-recognition algorithms. Perception 2001, 30, 303–321.
[32] Rodarmel, C.; Shan, J. Principal component analysis for hyperspectral
image classification. Surv. Land Inf. Sci. 2002, 62, 115–122.
[33] Ahila, R.; Sadasivam, V.; Manimala, K. Particle swarm optimization-
based feature selection and parameter optimization for power system
disturbances classification. Appl. Artif. Intell. 2012, 26, 832–861.
[34] Masoum, M.; Jamali, S.; Ghaffarzadeh, N. Detection and classification of
power quality disturbances using discrete wavelet transform and wavelet
networks. IET Sci. Meas. Technol. 2010, 4, 193–205.
[35] Biswal, B.; Mishra, S. Power signal disturbance identification and
classification using a modified frequency slice wavelet transform. IET
Gener. Transm. Distrib. 2014, 8, 353–362.
[36] Jamali, S.; Farsa, A.R.; Ghaffarzadeh, N. Identification of optimal
features for fast and accurate classification of power quality disturbances.
Measurement 2018, 116, 565–574.
[37] Kumar, R.; Singh, B.; Shahani, D.; Chandra, A.; Al-Haddad, K. Recog-
nition of power-quality disturbances using S-transform-based ANN
classifier and rule-based decision tree. IEEE Trans. Ind. Appl. 2015,
51, 1249–1258.
[38] Mishra, S.; Bhende, C.; Panigrahi, B. Detection and classification of
power quality disturbances using S-transform and probabilistic neural
network. IEEE Trans. Power Deliv. 2008, 23, 280–287.
[39] Gaing, Z.-L. Wavelet-based neural network for power disturbance recog-
nition and classification. IEEE Trans. Power Deliv. 2004, 19, 1560–1568.
[40] Wang, H.; Wang, P.; Liu, T. Power quality disturbance classification
using the S-transform and probabilistic neural network. Energies 2017,
10, 107.
[41] Weston, J.; Watkins, C. Multi-Class Support Vec-
tor Machines. Citeseer, 1998. Available online:
http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.50.9594
(accessed on 1 April 2019).
[42] Ucar, F.; Alcin, O.F.; Dandil, B.; Ata, F. Power quality event detection
using a fast extreme learning machine. Energies 2018, 11, 145.
[43] Mehta, S.; Shen, X.; Gou, J.; Niu, D. A New Nearest Centroid Neighbor
Classifier Based on K Local Means Using Harmonic Mean Distance.
Information 2018, 9, 234.
[44] Hu, W.; Huang, Y.; Wei, L.; Zhang, F.; Li, H. Deep convolutional neural
networks for hyperspectral image classification. J. Sens. 2015, 2015, 12.
[45] Hershey, S.; Chaudhuri, S.; Ellis, D.P.; Gemmeke, J.F.; Jansen, A.;
Moore, R.C.; Plakal, M.; Platt, D.; Saurous, R.A.; Seybold, B. CNN

architectures for large-scale audio classification. In Proceedings of the
2017 IEEE International Conference on CNN Architectures for Large-
Scale Audio Classification, New Orleans, LA, USA, 5–9 March 2017.
[46] Hu, G.; Yang, Y.; Yi, D.; Kittler, J.; Christmas, W.; Li, S.Z.; Hospedales,
T. When face recognition meets with deep learning: An evaluation of
convolutional neural networks for face recognition. In Proceedings of
the IEEE International Conference on Computer Vision Workshops,
Santiago, Chile, 7–13 December 2015.
[47] Li, Q.; Cai, W.; Wang, X.; Zhou, Y.; Feng, D.D.; Chen, M. Medical im-
age classification with convolutional neural network. In Proceedings of
the 2014 13th International Conference on Medical Image Classification
with Convolutional Neural Network, Singapore, 10–12 December 2014.
[48] Balouji, Ebrahim, and Ozgul Salor. ”Classification of power quality
events using deep learning on event images.” 2017 3rd International
Conference on Pattern Recognition and Image Analysis (IPRIA). IEEE,
2017.
[49] Mishra, Pabitra Kumar, Umamani Subudhi, and Sarthak Jain. ”Power
Quality Disturbances Classification with Deep Learning Approach.”
2019 International Conference on Information Technology (ICIT). IEEE,
2019.
[50] Ma, Jian, et al. ”Classification of power quality disturbances via deep
learning.” IETE Technical Review 34.4 (2017): 408-415.

A review on power quality disturbance classification using deep learning approach

More Related Content

What's hot (20)

Similar to A review on power quality disturbance classification using deep learning approach (20)

More from MuskanRath1 (6)

Recently uploaded (20)

A review on power quality disturbance classification using deep learning approach