This paper presents an advanced deep learning framework for accurate localization of incipient partial discharge (PD) sources in a power equipment based on acoustic emission. In this regard, an experimental setup was configured to emulate single- and multiple-location partial-discharge activity. Five acoustic sensors (at five different wall sides) were employed to capture acoustic emission from PD sources during PD activity. The acoustic signals were analyzed in joint time-frequency space, and their representation was improved through the Tuned Multisynchrosqueezing Transform. A multi-sensor fusion deep neural network was configured to process time-frequency spectrograms corresponding to each acoustic sensor PD data simultaneously and predict PD source location. The deep neural network was trained on a large database and tested utilizing a cross-dataset learning strategy. Ablation study revealed superior performance of the proposed framework under a noisy background, suggesting it’s potential in real-life PD localization in power equipment.

This paper presents an efficient and straightforward machine learning (ML) algorithm designed for the rapid detection of the most frequently occurring disturbances in microgrid (MG) systems. A laboratory-based MG was constructed, including components such as nonlinear loads, compensating devices, inductive loads, grid synchronization features, a battery charging circuit, switching relays, instrument transformers, and a signal acquisition system. Initially, valuable informative data were filtered from the acquired signals using an unsupervised k-means clustering technique, followed by feature extraction through detrended fluctuation analysis (DFA). The feature matrix was then reduced by retaining only the most important principal components (PC) before applying the 1-Nearest Neighbour (1-NN) supervised ML algorithm. Finally, the proposed method was compared to eight other benchmark ML algorithms based on class-wise accuracy. Besides, the performance was also examined in noisy environments.

This article presents an Internet-of-Things (IoT)-based hardware-software model for detecting power system disturbances. A pre-trained model using a bidirectional long short-term memory (BiLSTM)-based algorithm has been used as a classifier. Initially, the proposed classifier has been trained using an input data matrix formed by the measured disturbance signals. With the optimal tuned hyperparameters, the classifier has been designed to render the maximum accuracy. This trained model has been installed in the processor at a remote convenient location. A couple of power system disturbances have been tested by feeding data in a compatible form to this classifier through a cloud-based server. These disturbances have been created in a laboratory-based distributed system. With the help of current transformer (CT), the line current has been sensed, and the respective digital data are stored in the memory of an ESP8266 microcontroller where the preliminary checking for disturbance has been carried out. Depending on internet speed and cloud congestion, the disturbance signal data has been sent to a remote monitoring system, where it has been processed and fed to the said (Power System Detection) PSD detection algorithm for identification.

Abstract The modernization of electrical power systems has accelerated the transition from conventional centralized grids to intelligent, cyber-enabled smart grids characterized by bidirectional energy flow, advanced sensing, and real-time data-driven control. However, the large-scale integration of inverter-interfaced renewable energy sources and nonlinear loads has introduced significant power quality (PQ) challenges, including voltage sag, swell, harmonics, flicker, transients, and composite disturbances under noisy conditions. Accurate and computationally efficient detection of such disturbances is critical for intelligent digital relaying and real-time monitoring applications. To address these challenges, this paper proposes a deep learning framework based on a Mamba-driven state-space model architecture for comprehensive PQ disturbance classification. The proposed approach converts voltage waveforms into structured two-dimensional patches using convolutional operations, which are then transformed into sequential representations for efficient long-sequence modeling via the Mamba module. Spatial feature extraction and temporal dependency learning are jointly achieved, while maintaining linear-time complexity and reduced memory requirements compared to transformer-based architectures. The extracted latent features are subsequently classified using a deep neural network (DNN)followed by s o f t m a x classifier. Eighteen PQ disturbance categories, defined in accordance with IEEE standards, are considered, including single and composite disturbances under noisy operating conditions. Experimental evaluation examines feature representations, convergence characteristics, confusion matrix performance, computational time, and memory usage. Comparative results demonstrate that the proposed Mamba-based framework achieves high classification accuracy with significantly lower computational complexity than conventional transformer models. The findings indicate that the proposed method is well-suited for real-time smart grid monitoring and intelligent protection systems.

Reliable detection and localization of partial discharges (PDs) in transformers are critical for ensuring insulation integrity and extending equipment lifetime. Conventional acoustic and time–frequency (TF) analysis methods such as the Short-Time Fourier Transform, Continuous Wavelet Transform, Stockwell Transform often face limitations under low signal-to-noise ratio (SNR). This paper introduces the Powered Superlet Transform (PST), a novel TF representation that employs a powered growth schedule for the number of cycles in wavelet sets, resulting in increased inter location distance in TF plots and thus increased accuracy in localization. The PST is applied to acoustic signals measured from multiple PD sources in a controlled transformer tank setup. Subsequently, feature extraction through TF methods and classification through Support Vector Machine is performed. Comparative analysis with established TF techniques and acoustic localization methods demonstrates that PST is the most effective technique in terms of accuracy while both PST and Adaptive Superlet Transform (ASLT) provides superior noise robustness in terms of reduced localization error. Experimental results under varying SNR conditions show that PST can reliably localize PD sources with accuracy of 78.8% and approximate RMS localization errors 11.857 cm at SNRs as low as 2.5 dB, while established methods are less accurate. These findings highlight PST as a promising tool for advanced acoustic PD monitoring and condition assessment of high-voltage transformers.

Dissolved Gas Analysis is a well-established technique for diagnosing incipient faults in oil-immersed power equipment by analysing the concentrations and patterns of gases generated under thermal and electrical stresses within insulating media. Ratio-based methods offer simplicity and standardization however their reliance on fixed thresholds often leads to ambiguous or conflicting interpretations, particularly at low gas concentrations. Graphical techniques are better at improving interpretability but suffer from the fixed boundaries that cannot represent the nonlinear behaviour of gases and may overlap under complex fault conditions. AI-based methods may improve the diagnostic capability by modeling nonlinear relationships, but rely heavily on training data, are not physically interpretable, and can give inconsistent results. To address these limitations, a Hybrid Confidence Fusion Model is proposed to integrate multiple diagnostic outputs via a relational database and then using a confidence fusion voting scheme to fuse multiple diagnostic outputs. This model assigns adaptive confidence scores and incorporates severity-based prioritization to resolve conflicts. The proposed approach will make diagnosis more consistent, robust under sparse and mixed fault conditions and provide more reliable maintenance for modern power equipment.

In this work, a hardware module is developed to monitor real-time power consumption of home appliances at regular time intervals. The module comprises one microcontroller units, a voltage sensor, and five current sensors. This article introduces a cost-effective and easy to deploy hybrid power factor selection algorithm (HPFSA), which enables active power measurement with good accuracy and minimizes the required hardware components for this module. The developed module considers power factor (PF) computation error with supply frequency variation through HPFSA and gives accurate value of true power in real-time. Additionally, it can sense the total energy consumption of various appliances distributed throughout the home and send the consumed power data to the cloud for remote monitoring. This real-time load monitoring of home appliances is essential for applications such as energy conservation, smart home automation, and load management.

A study on power system disturbance (PSD) classification of a distributed generator (DG) based grid-connected network has been carried out in this article. A total of 24 different PSDs commonly occurring in the DG-based grid-connected network have been considered. The three signal processing tools, Discrete wavelet transform (DWT), Detrended Fluctuation Analysis (DFA) and Recurrence Quantification Analysis (RQA) have been applied to construct a feature matrix. A category-wise visual representation on a 2D plane has been provided using the Uniform Manifold Approximation and Projection (UMAP) technique, and the coefficients of the feature matrix have been processed using this UMAP technique. The non-linear model, obtained from UMAP, helps to reduce the dimension of the classification algorithm and provides a categorical projection in a 2D plane. The category-wise visual 2D representation clearly proves the effectiveness of the model in identifying and classifying each class of PSDs. This huge variety of disturbance classification by visual 2D representation using UMAP is a novel approach in PSD study.

This study proposes an automated methodology for the precise identification of both single and combined power quality (PQ) disturbances within an electrical power system network. To investigate this, a comprehensive dataset comprising 9 distinct single-event types and 27 mixed-event scenarios was synthetically generated. These simulated PQ signals were subsequently transformed into time-frequency representation (TFR) images using the multisynchrosqueezing transform (MSST), enabling effective encapsulation of their non-stationary characteristics. The resulting TFR images were then input into a pre-trained deep neural network for automatic feature extraction. To enhance classification efficiency, statistically significant features were identified using one-way analysis of variance (ANOVA). These selected features were further classified using various standard benchmark classifiers. Experimental results demonstrate that features extracted via the VGGNet16 architecture, when classified using a support vector machine (SVM), achieved a high detection accuracy of 99.58 %, which is on par with leading state-of-the-art techniques. Additionally, the proposed approach has been rigorously validated under noisy conditions as well as with real-world PQ disturbance data, confirming its robustness and practical applicability. The proposed framework (i.e. adaptive MSST -* transfer-learned deep features -* ANOVA selection -* lightweight classifier) provides a practical, computation-friendly alternative to training heavy end-to-end networks while improving discrimination for mixed PQ events.

Contamination flashovers are the primary reason for insulator failure. Early and accurate prediction of insulator contamination severity can avert flashover incidents, thereby impacting the reliable operation of the electrical power system network. This paper introduces an innovative infrared image-aided deep transfer learning approach for remote monitoring of overhead line insulators. A large database is created based on the infrared image of artificially contaminated insulator samples captured at different environmental conditions. The infrared images were initially processed through a mask region- based convolutional neural network to mitigate the background effect. The infrared images corresponding to different contamination were fed to a benchmark convolutional neural network (CNN) model (Alex Net) for classification purposes. Transfer learning with a fine-tuning strategy was adopted to train the network. The result indicates that the proposed approach returns appreciably high classification performance with reduced computational time compared to existing techniques. This insight emphasizes the potential application of the proposed approach for non-contact monitoring overhead line insulators.

This work investigates the influence of TiO2 (titanium dioxide) fillers on ZnO (zinc oxide) based varistor using the leakage current analysis. For this purpose, a complete model of metal oxide surge arresters (MOSA) is developed using finite element analysis. In addition, different parameter variation of ZnO varistor grains (by varying electrical conductivity and relative permittivity) and grain boundaries under different TiO2 concentrations (0-15 mol%) have been incorporated in this model. The incorporated parameter values are considered depending on the formation of segregation layers and barrier voltages due to the effect of the corresponding compounds. Hence, the effectiveness of the developed model is validated through the comparison of the experimental results (without TiO2) and simulation results of the leakage current. Finally, the simulation results of leakage currents with increasing TiO2 content in ZnO based varistor shows that there is a reduction in the current through the ZnO column. These facts ensure the enhancement of grain boundary impedance. These findings suggest that tailored TiO2 doping in MOSA can effectively decrease the overall device dimensions by permitting reduced ZnO disk height for the same electrical stress, offering a pathway to improved surge protection reliability and compactness in high-voltage distribution networks.

The significant need to reduce the emission of green-house gases to improve the air quality in the atmosphere has led to an escalation in the development of electric vehicles (EVs). One of the major challenges in its growth of EV is battery charging technology. Nowadays, the most promising technology in this field is Inductive Resonance Wireless Power Transfer (IRWPT) where contact less charging takes place using strong magnetic coupling between the transmitter and receiver coils tuned at their resonant frequency. To optimize the performance of EV charging, the choice of an effective compensation technique is vital. This paper compares the quality factor and bandwidth of different compensation topologies, namely Series-Series (SS), Series-Parallel (SP), Parallel-Series (PS), and Parallel-Parallel (PP), using MATLAB SIMULINK. The paper also deals with the frequency response analysis of some advanced hybrid network configurations, which include inductor-capacitor-capacitor (LCC) and inductor-capacitor-inductor (LCL) topologies. The analysis further concluded that the power transmission capacity is highest when the wireless power transfer (WPT) circuit is simulated at the resonant frequency.

This article presents a novel approach to accurately measure moisture levels in power transformer oil-paper insulation by analyzing dielectric response currents under standard and nonstandard lightning impulse (NSLI) voltages. Laboratory-prepared samples emulating transformer paper insulation with varying moisture levels are impregnated in different oils (mineral, vegetable, or their nanofluids) and are subjected to lightning impulse testing. Response currents are recorded, and transfer functions are evaluated by examining applied voltage and response current. Five coefficients (three from the numerator, ${a}2$ , ${a}1$ , and ${a}0$ , and two from the denominator, ${b}1$ and ${b}0$ ) are found sensitive to moisture content variations. Therefore, empirical relationships between these coefficients and paper moisture content (%pmc) are developed and are validated using newly prepared samples with different moisture contents, yielding maximum errors below 2.75% for NSLI and 3.5% for SLI. These findings offer valuable insights into transformer insulation conditions, enabling timely preventive measures to mitigate transformer failure.

This work presents a methodology based on the Adaptive Superlet Transform (ASLT) for precise localization of Partial Discharge (PD) events within three-phase transformers. PD sources were artificially created at different locations of a three-phase potential transformer, which was installed inside a tank. To capture the acoustic PD signals, four acoustic sensors were positioned on the outside of the walls of the transformer encloser. The Time Frequency (TF) plots of these recorded acoustic PD signals were created using ASLT. Feature extraction was subsequently performed from the obtained ASLT plots. Using the acquired features, a Decision Tree (DT) based classifier was used to classify and, consequently, identify the locations of those PD sources. The trained model achieved a very high accuracy of 99.995%. The ASLT-based approach for identifying the location of partial discharge occurrence surpasses conventional TF methods regarding performance parameters.

This work proposes a Recurrence Plot (RP) based customized convolution neural network (CNN) method for the power system disturbance classification of (Distributed Generation) DG-based networks. All feasible disturbance-creating events including faults, different switching events and dual disturbances have been considered here. The proposed method has been applied to the disturbance signals generated from a laboratory-based experimental setup of DG. Those signals were converted to RP images to train the CNN algorithm. In addition, recurrence quantification analysis (RQA) was performed to train the Support Vector Machine (SVM) simultaneously for comparison. Performance indices have been estimated to validate the proposed method. This algorithm has also been tested under highly noisy conditions and performs satisfactorily.

This research investigated the influence of insulating (Al2O3), semiconducting (TiO2), and conducting (Fe3O4) Nanoparticles (NPs) on the breakdown strength of Nomex-based polymeric Nanocomposites (NCs). Seven samples were fabricated and thermally aged at 145 °C for 600 h to simulate accelerated degradation. Fourier Transform Infrared Spectroscopy (FTIR) was conducted to unaged, 300 h, and 600 h aging durations to examine the chemical alteration and understand the NP-induced resistance of meta-aramid polymeric chains against the thermal aging. The AC Breakdown Voltage (ACBDV) was measured every 100 h of aging, and the results were analyzed using the Weibull distribution to extract the scale parameters, shape parameters, and p-values. Cumulative breakdown probabilities at 1%, 10%, and 50% risk levels were evaluated to assess the impact of different NPs on the breakdown mechanism. After 600 h of aging, the incorporation of 1 wt% Al2O3 improved the BDV of pristine Nomex by 28.2%. At a 50% risk probability, BDV enhancements of 63.16% (Al2O3), 48.89% (TiO2), and 42.39% (Fe3O4) were achieved, demonstrating the superior performance of NP-reinforced Nomex insulation under thermal stress..

Battery Energy Storage Systems (BESS) play a crucial role in modern energy systems, driven by the increasing demand for grid stabilization, electric vehicles (EVs), and renewable energy integration. This review examines the materials and technologies supporting BESS, with a particular focus on Lithium-Ion (Li-Ion) batteries, their alternatives, and future prospects. Li-Ion batteries remain the dominant technology due to their high energy density and efficiency; however, they face challenges such as material scarcity, safety concerns, and energy limitations. Finally, key challenges and research opportunities in advancing BESS technology are discussed, emphasizing the need for higher energy densities, improved safety, and economic feasibility for large-scale energy storage solutions.

In this letter, a microcontroller-based smart fault handling system (FHS) is proposed which is capable of early sensing and managing the thermal runaway (TR) event in real-time battery management system (BMS) through online internal resistance (IR) computation method. In overcharging region of lithium-ion (Li-ion) batteries (LIBs), TR is one of the critical issues which occur when used in electric vehicles (EVs) and battery energy storage systems (BESSs). Therefore, a proper subsystem is utmost required in the BMS for detecting the TR event quite early, which will automatically prevent the battery modules from critical accidents like fire, explosion, etc. The developed smart FHS utilizes an efficient, cost-effective, and reliable online IR sensing-based early TR sensing (ETRS) system which detects the TR event similar to 3.9 min prior to the TR onset point (outperforming the other detection methods) and shuts down the charging mechanism. Additionally, this system sends an IoT-based short message service (SMS) alert notification to the users allowing them to take necessary preventive steps.

Fast and accurate detection of power system disturbance (PSD)-creating events is very much essential for the safe and reliable operation of today's power distribution network. This task becomes more challenging when two events occur simultaneously. In this work, a deep learning (DL) network-based classifier has been proposed to classify the events that are likely to occur in the power system. Here, a few normal events such as switching, load changing, all feasible faults, and five dual events have been considered. All acquired disturbance signals are allowed to go through three major stages. At the initial stage, signal-to-image conversion takes place using continuous wavelet transform (CWT) followed by a couple of convolution operations for feature extraction, and then, the extracted time-series data are fed to the input of bidirectional long short-term memory (BiLSTM) module. Finally, the softmax output classifier has been applied. Few other popular classifiers have also been studied for comparison. Using this proposed method, 98.57% accuracy has been achieved considering 21 feasible events.