In this article, a deep learning based automated framework has been proposed to sense minor stator winding inter-turn fault in an induction motor without having load level information. For this purpose, an experimental setup has been configured in laboratory through which three phase current signal of a 3-ph squirrel cage induction motor (IM) has been recorded at different minor stator winding inter-turn fault, different loading level and external voltage unbalance condition. In this proposed framework, three phase stator current profiles firstly are converted to 1-D Park’s vector modulus signal using Park’s transformation and Extended Park’s Vector Approach. Thereafter, Park’s vector modulus of three phase stator currents have been encoded into RGB image using recurrence plot technique and have been fed to a customized convolutional neural network module (RPCNNet) for detection of minor stator winding inter-turn fault severity. The recognition performance of the RPCNNet is also compared with four benchmark convolutional neural network (CNN) models namely “VGG16”, “AlexNet”, “ResNet50” and “DenseNet201” respectively. From the result, it is observed that RPCNNet has shown comparable performance and offers significantly reduced computational cost compared to the benchmark CNN models. Proposed framework has also outperformed the existing approaches by a significant margin and shown promising result under both noiseless and noisy environment. The proposed framework is also capable of differentiating minor stator winding inter-turn fault condition from unbalanced supply voltage condition. Hence, proposed framework is very much suitable for practical application.

Deposition of environmental pollutants like dust, salt, alkali etc. on the housing of metal oxide surge arrester can induces premature failure of it in long run. Therefore, reliability of an electrical system can be affected. Hence, accurate sensing of surface pollution severity of Metal oxide surge arrester is very much important. Considering the issue, this paper proposes an innovative approach to sense the pollution severity on metal oxide surge arrester housing using infrared thermal imaging technique. For this purpose, infrared thermal images of metal oxide surge arrester at different pollution severity have been captured. After suitable preprocessing, the captured infrared thermal images are fed to pretrained convolutional neural network architecture “ResNet50” for automatic feature extraction. The extracted deep features have been fed to 4 machine learning classifiers i.e., k-nearest neighbor, support vector machine, naïve Bayes and random forest for classification purpose. According to the result, the best performance has been achieved with the random forest classifier. It is also observed that proposed framework is very much capable to sense surface pollution severity of Metal oxide surge arrester with higher degree of accuracy. Therefore, proposed framework can be practically implemented to monitor surface condition of Metal oxide surge arrester.

In this article, an efficient technique has been proposed to estimate the contamination level of overhead polymeric insulators. Deposition of contamination on polymeric insulator surface, is a serious issue as it often results in the flashover and even insulator failure. For estimating the severity of contamination level, surface leakage current (SLC) signals of a 11kV polymeric insulator with contaminated surface has been analyzed in time-frequency domain through hyperbolic window stockwell transform (HST). HST is more flexible than classical stockwell transform. Also, HST can able to handle both the low and high frequencies adequately. Considering the advantage, HST has been used here to estimate contamination degree from SLC signature. HST analysis of SLC signal returned a 2d complex time-frequency HS matrix. The complex time-frequency HS matrix has been separated into magnitude and phase spectrum. Based on the phase and magnitude spectrum, 15 statistical features, namely HST features has been extracted. Thereafter, 5 relevant HST features have been selected through least absolute shrinkage and selection operator (LASSO) feature selection technique. Finally, these relevant HST features are fed to four machine learning classifiers for estimation of contamination degree. It has also been observed that, the proposed framework method offered better classification accuracy compared to other standard time-frequency analysis and existing methods available in literature.

With the fast development of power grid the usage of electrical equipments is increased which led to importance of power quality disturbance sensing for reliable and smooth operation. In this paper, a deep neural network has been designed using stacked autoencoder (SAE) for deep feature extraction from time-frequency spectrum of single and combined PQ disturbances in electrical power system network. For this purpose, synthetic PQ signals are analyzed in time-frequency domain through hyperbolic window stockwell transform (HWST). Thereafter, PQ signal converted HWST time-frequency matrix has been grouped into time-frequency blocks and subsequently fed as input to 3-layer stacked autoencoder model (SAE) for deep feature learning. Finally, the extracted deep features are classified through several machine learning classifier. The results indicate that proposed framework using XGboost classifier can classify 18 different single and combined PQ event with a 99.86% accuracy. The proposed framework also yields satisfactory outcome with real life PQ data. Therefore, proposed framework can be implemented for Power quality monitoring in electrical power system.

With the rapid development of electric power grid, the application of Metal Oxide Surge Arrester (MOSA) in electrical power system network has been enhanced. MOSA has been used as protective device in electrical power system (EPS) network. MOSA provides a basis of insulation coordination in EPS network. Premature failure of MOSA can affect the reliability of EPS network. Therefore, condition monitoring of MOSA on regular basis can enhance the reliability of EPS network. Currently various techniques are available of condition monitoring of MOSA. Out of them, leakage current signal-based monitoring is most popular and convenient way of MOSA monitoring as leakage current signal of MOSA can be acquired easily and it can give information of MOSA condition effectively. Therefore, most of the recent researches in the area of condition monitoring of MOSA are based on leakage current signal especially resistive leakage current analysis. This article presents a comprehensive review of leakage current signal employed methodologies used for condition monitoring of MOSA. The broad perspective of the review article is to provide various concepts used to extract resistive leakage current and find out suitable indicators, which are sensitive to MOSA condition. Advantages, limitations and challenges faced in the proposed methodologies are reported, which can be very useful for future research development in this field.

This paper presents an advanced method of sensing surface contamination of polymeric housed Metal Oxide Surge Arrester through resistive leakage current signal analysis. Surface condition of polymeric housed MOSAs often gets contaminated due to accumulation of dust and other pollutant. Accumulated pollutants can degrade the condition of the arrester due to overheating which may lead to explosion. Therefore, reliability of power system may get affected due to failure of MOSAs. Resistive leakage current analysis of MOSA is one of the conventional method for sensing surface contamination of surge arrester. In this article, Mathematical Morphology operator has been introduced to extract various features from the resistive part of leakage current signals measured at different surface contamination level. Further, the extracted features have been trained through Gaussian Naïve Bayes (GNB) and surface contamination level of MOSA has been identified through this classifier. Result shows that proposed technique provides satisfactory outcomes regarding condition monitoring of MOSA at different surface contamination level which in turn enhances the reliability of system. The proposed technique is generic in nature and well suited for any other similar kind of applications.