Power grids serve as the backbone of critical infrastructures, enabling the efficient control and distribution of electricity. Subsequently, the number of cyber attacks against power grids continues to grow, especially during times of international conflict and uncertainty. Thus, we must continue to further the research that secures and ensures the stability of power grids. Related research has made progress to this end through the application of artificial intelligence-based systems, but such works suffer from the following limitations: (a) they strongly focus on attack detection, neglecting attack classification, (b) they produce complex systems requiring large amounts of computation without concern for reducing complexity, and (c) they lack concern for how the systems can be adapted when new attacks arise. These limitations motivate our work, where we develop efficient and high-performing classification systems that are adaptable to new attacks. Specifically, we develop diverse benign and attack datasets consisting of cyber and physical layer data using our power system testbed. Additionally, we adopt SHapley Additive exPlanations to reduce the total number of features required to accurately classify attacks by 83% while maintaining a superior accuracy of 97%. Lastly, we use transfer learning to enhance fine-tuning and adapt our classification system to classify new cyber attacks in power grid environments.

The growing popularity of electric vehicles (EVs) has rendered public EV charging stations (EVCSs) vital for alleviating range anxiety and supporting long-distance travel. However, recent studies reveal security vulnerabilities in EVs and EVCSs against attacks. This paper addresses these security concerns by introducing injection attacks on the front-end Vehicle-to-Grid (V2G) communication using the ISO 15118 protocol. Malicious EV owners or compromised EVCS supply equipment can inject harmful packets, potentially leading to runtime modifications and malware attacks. To counter this threat, we propose an innovative bidirectional recurrent attentive graph neural network (BiRAGNN)-based intrusion detection system (IDS) that dynamically captures spatiotemporal aspects and the bidirectional flow of information, while leveraging an attention mechanism to effectively detect injection attacks within EVs and EVCSs. The BiRAGNN model is founded on a probabilistic charging graph of real cities. Other deep learning and graph-based IDSs are also investigated as evaluation benchmarks. The IDSs are examined against a standalone system (from a single EVCS) and multi-node systems (from 8, 50, and 100-node EVCSs), all with packet-level, flow-level, and fused packet and flow-level data. The proposed BiRAGNN-based IDS offers a detection accuracy of 99% on the 100-node fused dataset, surpassing the benchmarks by$5 - 8\%$, offering EVs and public EVCSs resilience against cyber threats.

High voltage (HV) power cables are exposed to various stresses, such as chemical, electrical, mechanical, and thermal. These stresses contribute to the eventual failure of electrical power cables in an unpredictable manner. Cross-linked polyethylene (XLPE) is an essential component of a power cable's insulation layers, and unwanted stresses negatively affect the insulation layers' performance in various ways. In this paper, three 3D Maxwell finite element analysis (FEA) models of a single core, 30 kV DC power cable are built to study the effects of chemical stress, electrical stress, and mechanical stress on the characteristics of polymer insulation material. The main parameter utilized for determining insulation health in this work is the polarization index. Temporal evolution in the polarization index is investigated, as aging-induced stresses degrade the XLPE's structure. The severity of chemical stress, electrical stress, and mechanical stress is estimated and compared based on the polarization index, with a focus on chemical stress, highlighting the severity of the deformation caused within the inner polymeric structure of XLPE. The severity is obtained for the innermost and outermost insulation layers. It was discovered that the insulation layer's proximity to the cable's core is the main factor influencing degradation. Moreover, the obtained results successfully demonstrate the polarization index's ability to detect degradation after the 95-day mark, indicated by the substantial increase in polarization index values, common between all stress types applied.

One of the common causes of cross-linked polyethylene (XLPE) insulation aging and eventual cable failure is the presence of voids within the cable insulation. Gas-filled voids can be initiated in XLPE cables during their manufacturing, installation, and operating conditions. Once the size of the void reaches a certain critical limit, noticeable partial discharge (PD) activity develops. Therefore, the size of the void is one of the important factors considered when investigating PD activity in power cables. This paper presents a detailed examination of how different sizes and distributions of voids affect the cable's behavior. A 3D model of XLPE power cable insulation is carried out to improve the understanding of how the electric field distribution responds to different shapes, positions, and sizes of voids within the XLPE insulation. The effect of void size on PD behavior within the XLPE layer of an 18/30 kV cable specimen is studied. Also, the changes in artificial spherical void depths and diameters are investigated.

This paper proposes a self adaptive physics informed neural network (SAPINN) model to predict the degree of polymerization (DP) of oil-impregnated paper insulation to quantify the level of degradation and the remaining useful lifetime. The prediction is performed based on historical DP values and the corresponding prediction time step, which are used as input data points to the proposed model. The DP mathematical model is used to constrain the training phase of the AI-model through a weighted sum loss function. The weights of this loss function are adjusted for each epoch through a self-adaptive weighting method to determine the relative importance of the data component and the mathematical model throughout the training by defining these weights as trainable parameters. The trained model is then tested using different datasets which are not part of the training phase. The training and testing datasets are generated synthetically through an algorithm that considers the deviation from the ideal DP degradation curve and incorporates actual measurement noise. The performance of the proposed SAPINN is compared to the baseline PINN and NN (in the absence of physics) to highlight the importance of embedding the mathematical model and the self adaptation algorithm, and theses experiments demonstrate that SAPINN significantly enhances the DP prediction.

The adoption of electric bikes is growing exponentially worldwide. However, the existing electric bike (e-bike) chargers are costly and lack essential features such as proper authentication, real-time monitoring, parameter analysis, and effective maintenance. In this work, a bidirectional DC-DC converter is designed to improve the performance of a regenerative braking system and increase the battery lifetime in electric bicycles. Compared to the traditional design concepts, the proposed method includes higher-level control algorithms and enhanced control of the switching methods, which makes the system perform optimally. Also, a comparative study has been conducted with other topologies to clearly highlight the enhanced energy recovery, battery durability, and effectiveness of the proposed converter. The obtained results demonstrate the improved efficiency of the proposed topology, revealing its advantages over conventional topologies.

Induction motors are a necessity in many industries, which is why early fault detection is critical to account for damage and industrial downtime. Among the incipient damages, BF and stator winding faults are the most prevalent. Consequently, early detection and classification of these faults are gaining significant attention. This paper investigates the application of multiple gradient boosting machine learning (ML) algorithms, that are known for their robustness, and analyses the accuracy of the models on faulty induction motors (IM) using Motor Current Signal Analysis (MCSA). Five different Supervised machine learning algorithms such as Gradient Boosting Machines, XGBoosts, and LightGBM were used in this study and compared with strong models like RF and KNN. Overall, the experiments provided a classification accuracy of approximately 92% and were able to distinguish the normal, bearing, and stator winding faulty signals. The obtained results show that current signals are a viable option for observing IM electrical and mechanical faults with finetuned optimization of hyperparameters.

Discharges, such as partial discharges (PDs) and corona discharges (CDs) are the most common faults that occur in insulation materials used in high voltage (HV) equipment. A high repetition rate of discharge activity indicates the severity of the defects that shorten the lifetime of electrical equipment, leading to insulation failure. To solve this, this paper proposes an efficient classification technique for corona discharge defect intensity using features obtained from statistical parameters such as the ignition voltage of CDs. The Recurrent Neural Network (RNN) is proposed to identify the intensity of corona discharges. A comprehensive experimental evaluation is conducted, to demonstrate the capabilities of the proposed solution. The exceptional predictive abilities of the long short-term memory (LSTM) method are the primary benefit of the proposed approach presented, with a potential for enhancing the performance of CD detection systems. The obtained results demonstrate the accuracy of the proposed model, indicating its potential for deployment in practical applications. The innovative approaches utilized in this paper will help engineers and operators quickly determine the severity (sharpness and curvature) of the protrusions or surface defects that cause CDs, solely based on measurements of the ignition voltage.

Monitoring partial discharges (PDs) in high voltage (HV) electrical equipment is an effective tool to prevent unscheduled shutdown and eliminate possible failures. The inception of PDs can take various forms within the insulation system and can be detected using established approaches. The most common cause of HV rotating machine failure is the initiation of PDs in the stator winding insulation system. Various online and offline techniques have been proposed to monitor, detect, diagnose, and locate PDs in HV rotating machines. This paper gives an overview of existing PD detection and localization technologies in HV rotating machines. A comparison of online and offline PD detection approaches is discussed in this paper. Finally, a brief review of the methods of PD classification is provided.

The management of incipient faults in induction machines (IMs) is crucial for ensuring reliability and efficiency in diverse industrial applications, including power grids, electric vehicles, and manufacturing processes. This review explores advanced fault detection and diagnosis (FDD) strategies, empha-sizing deep learning (DL) methods such as convolutional neural networks (CNN), recurrent neural networks (RNN), and au-toencoders for fault detection and classification. Traditional machine learning (ML) approaches are also discussed, highlighting their integration with signal processing techniques like wavelet transforms and Fourier transforms to enhance FDD accuracy. Additionally, the potential of physics-informed neural networks (PINNs) is examined, demonstrating how incorporating physical knowledge into data-driven models can improve diagnostic precision. The paper presents an analysis of recent publications, identifies current research gaps, and proposes future directions, including the development of robust AI-based FDD systems and the consideration of stochastic industrial data for more accurate predictive maintenance. By offering a comprehensive overview of FDD techniques and highlighting key research areas, this review aims to advance the reliability and performance of IMs.

Condition monitoring of electrical Rotating Machines (RM) serves in structural changes detection during machine’s operation. However, the frequent fault occurrence reduces the RM remaining useful life and accelerates their deterioration. Therefore, this paper proposes an effective multi-fault classification system for the faults in electric rotating machines. The proposed method employs an Artificial Neural Network (ANN) and Synthetic Minority Over-sampling (SMOTE) technique for automatically detecting rotating machines failures. This model's efficacy stems from the use of the relief feature selection approach to identify the most affecting features and improve the model's performance. A case study analysis uses the Machinery Fault Dataset (MAFAULDA) to test the models' performance. Simulation results are obtained to demonstrate that the proposed paradigm provides outstanding performance based on a fair assessment using the MAFAULDA dataset and shows that the proposed model has a high potential to detect rotating machine state.

This paper proposes a method to analyze aging response variance in terms of statistical parameters and geometry specifications of medium voltage power cable. The electric stress distribution in the single-core medium voltage power cable model is logged quantitatively using finite element modeling. The dependable parameters such as electrical conductivity, thermal conductivity, ambient temperature, relative permittivity, defect position and size, act as variables in the model to analyze simulated electric stress in the power cable. The variation in electrical stress distribution under the influence of different factors impact distinctly on cable degradation over time. Moreover, the correlation between these factors is quantified using derived statistical parameters to form a correlation matrix. In this work, both the variation coefficients and mean stress intensity are utilized as response measurement variables. A multivariable linear regression is applied to derive the relationship between the model parameters and the response variables. This study shows that the stress variation is strongly correlated with electrical conductivity and thermal conductivity, while the maximum stress intensity is most impact by defect sizes and position. The implication and consequence of the rest of the model parameters on electric stress distribution are also analyzed and discussed.

Quadratic cost function such as Mean Squared Error (MSE) has been a widely used objective function for training deep neural networks to develop energy forecasting models in Smart Grids. In this work, Penalizing Underestimation Logarithmic Squared Error (PULSE), a novel objective function is proposed with the aim of reducing the tendency of deep learning models to underestimate the target variable. Stacked Long Short-Term Memory (LSTM) networks are adopted on the time series load demand data to investigate the performance of the proposed cost function against the widely used MSE cost function. The evaluation is performed using open-source real-world electricity load diagrams dataset covering a period of three years. The performance of the proposed scheme is examined with deep learning models through several experiments. The results demonstrate that the proposed scheme is able to eliminate the tendency to underestimate and provides competitively accurate load demand forecasting results. The results are additionally compared against the state-of-the-art machine learning models developed in the literature. The proposed cost function maintains the RMSE around 4*10-2kWh which is also the RMSE for deep learning models with MSE cost function and delivers 25% improvement in MAPE while also eliminating the underestimation of load demand.

One of the major factors that lead to energy losses for utility distribution systems is electricity or energy theft. Energy theft is tampering with smart meter reading to reduce customer energy usage and reduce electricity bills. A thief customer tends to consume more energy and hence, the theft negatively affects the power supply quality in the form of transformer overload, voltage unbalance, and voltage drop on system buses. Meanwhile, it also causes great economic losses for the business of electric utility. In order to enable efficient energy theft detection, data-driven approaches including utilizing trained deep neural networks are proposed in this paper. The machine learning approaches can detect energy theft involving stealthy connections or meter tampering at the level of smart meters or aggregated levels. In this work, the detection effectiveness of different approaches is evaluated on real case study data at the end consumer level. The challenges of class imbalance and the missing values (around 25% of the whole fields) are addressed in the LSTM-based methodology. In this paper, results are obtained on real energy consumption data to show the higher performance of the proposed solutions compared to previously presented work.

In this paper machine learning (ML) models have been developed for the application of big data load forecasting using parallel computation. The load forecasting models' performance is directly linked to system execution capacity, memory, thread count, balancing the load, and available resources. This paper is focused on two main challenges. The first challenge is to reduce the execution time of the ML models and the second one is to choose the suitable tree-based model for effective load forecasting. The paper conducts a comprehensive evaluation of the load forecasting using real-world data on energy consumption. Comprehensive results are obtained to show that the performance of random search to tune the ML models exhibits competitive performances whilst not losing the accuracy of the models and gaining a competitive advantage on the run time.

The tight interaction between power grids and communication networks supports the advanced functionalities of smart grids and enables a more efficient utilization of the assets within the power system. Towards this objective, it is of utmost importance to develop a model that reflects such an interaction. Unfortunately, existing cyber-physical interdependent models are conceptual or are specific to certain regions and do not consider the spatial information and correlation of electrical elements. To address these limitations, in this paper, we propose an interdependent communication-power network model using tools from stochastic geometry, which takes into consideration the spatial locations of electrical buses and the required service quality within the communication network. A step-by-step process on building a communication network on top of the power grid while satisfying the average queuing delay and queuing stability constraints is described. A case study for a realization of this generative interdependent model is presented.

This paper proposes a novel sensorless vector control scheme for an open-end stator winding induction motor. The motor is fed by different three-voltage inverters with isolated DC sources. The inverters are controlled using space vector Pulse Width Modulation (PWM). A reduced-order sliding mode observer with smooth function is proposed for speed-sensorless control. The simulation results show the simultaneous correct estimation of both rotor flux and speed. The proposed control scheme is built within the Simulink environment combined with the Real-Time platform. The obtained results confirm the clear efficacy of the proposed control scheme under a variety of operating conditions.

In this paper, a second order sliding mode based Super Twisting (ST) controller is designed and implemented on sensorless five-level packed U cell inverter (PUC5). The aim is to enhance the inverter performance and achieve better robustness. A low complexity modulation strategy is employed. It contains a few logic blocks and two level-shifted carriers, which reduces the computation time and eases the implementation. Moreover, it ensures quick and self-balancing of the inverter capacitor voltage. Employing the proposed super-twisting controller leads to a significant enhancement of dynamic response and steady-state performance of the PUC5. It provides better accuracy in the tracking of the reference current, which results in high power quality, and helps in fast self-balancing of the capacitor voltage. The feasibility and the effectiveness of the proposed controller has been verified by simulation and experimentally. The obtained results show the higher performance of the super twisting controller in stand-alone mode of operation.

Accurate Short-Term Load Forecasting (STLF) has been considered a topic of extreme importance for efficient energy management, reliable energy transactions, and economic operation dispatch in smart grids. However, the continuous instability of the load demand essentially due to the high volatility of weather conditions and customers' demand behavior dramatically affects the STLF accuracy. In order to overcome this problem, five effective Deep Learning (DL) techniques are proposed for multivariate time series STLF based on Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), and stacked Auto-Encoder (AE). These DL based techniques are consolidated to build stacked Bidirectional GRU (BiGRU), Convolutional LSTM (ConvLSTM), stacked Bidirectional LSTM-AE (BiLSTM-AE), hybrid CNN-LSTM-AE (CNN-LSTM), and LSTM-AE (LSTM-AE) techniques. Simulation studies are conducted to demonstrate the performance superiority of BiLSTM-AE compared to the other DL models. The main contributions of this paper include 1) integrating a variety of deep neural networks for STLF; 2) employing time series as a benchmark to compare between heterogeneous DL architectures; 3) conducting the analyses on real data set.

This paper presents a novel state feedback controller design for a 5-level Packed U-cell (PUC5) inverter based on its average model. The state feedback is designed using a super-twisting sliding mode control (ST-SMC) and a PWM technique. The proposed super-twisting controller (STC) is used to regulate the capacitor voltage at the desired value in order to generate a 5-level output voltage waveform while feeding a smooth current to the load with low harmonics. A simulation and experimental comparative study between the proposed control technique and the conventional SMC and proportional-integral (PI) control techniques is presented. The presented results prove the higher performance of the proposed controller in terms of dynamic performances and power quality.

The traditional electric power grid is moving toward smart grid, and with the advent of smart grids, a lot of data is generated in high volumes, velocity and variety. This brings several challenges with real time processing to get meaningful information for enhancing the benefits of smart grid. Furthermore, the application of short-term load forecasting poses additional challenges of being highly uncertain and volatile due to different load profiles. This paper conducts a study on the demand side management and load forecasting in electric power grids with help of the historical data obtained from smart grid. Machine learning models are developed using deep learning to ensure significant improvement in forecast accuracy when compared to benchmark Auto-Regressive Integrated Moving Averages ARIMA analysis. The short-term load forecast data has been merged with weather data from Application Program Interface (API). The discussed system uses an autonomous feedback loop to consider the historical load values as features for training and testing. This paper also applies deep learning methods like pooling based recurrent neural networks which could solve the curse of dimensionality that usually exists with increasing layers in traditional neural network methods. The paper proposes an averaging regression ensembles model for short-term load forecast.

In this paper, a data-driven approach has been used to identify and categorize fault in the electrical power system. The proposed methodology involves efficient analysis of the data with feature vectors including the area or zone of the bus. The training is done on machine learning models to classify and identify the location of the fault. Three-phase, line to ground, line-to-line to ground, line-to-line, loss of line with no fault and loss of load at bus faults are simulated to generate labeled data with type of fault and location of fault. Two algorithms have been proposed to choose the measurements selection strategy, and results have been stated. The proposed methodology proves its validity for identification of the fault without necessary measurement of the voltage of each node. The proposed approach works with a minimum number of buses required to be as few as 5-7% of the measured buses. The accuracy, capabilities, and limitations of the proposed algorithm are verified on IEEE 68 bus model. The highest classification accuracy attained on one of the test cases is 91%.

Smart grids systems generate a large amount of data. Big Data which is an essential element for improving the reliability, stability, and efficiency, and in decreasing the cost of energy use. Reliable operation of smart grid depends on the utilization of various real-time information related to monitoring, communications, control and management systems. This paper gives a common understanding of how big data can impact the reliability and stability of power grid, and investigates in detail a smart grid communication network architecture. In addition, the paper explores and collates the smart energy subsystem, the smart information subsystem, and the smart communication subsystem. This conceptual lens provides deep insights into the reliability challenges and effective solutions toward reliability issues in smart grid to reveal the big data role during the transition, and how it can fuel the organic growth of smart grid.

Big Data is an essential element for energy management and control decision toward improved energy security, efficiency, and decreasing costs of energy use. Power distribution network is required to deliver electric energy reliability with reduced complexity and to be part of future smart grid. Therefore, in this paper Big Data related to the distribution generation systems will be discussed and illustrated within the context of smart grid principle. The paper work is to study the impact of adopting big data on energy management systems and to show the importance of the big data in strategic decision-making. The paper will highlight the Big Data issues and challenges associated with it in the energy management and control decisions in power distribution networks.

Perfectly balanced supply voltages are not possible in practice. Therefore, detection, discrimination and diagnosis of stator winding turn fault in the presence of unbalanced supply voltages for three-phase induction motors is needed. In this paper a novel approach is presented for stator winding turn incipient faults detection in the presence of different levels of voltage unbalance and at different load conditions. The proposed method investigates and utilizes the ratio between third harmonic and fundamental voltage and current waveform. Fast Fourier Transform (FFT) magnitude components of the stator currents and voltages are utilized for detection and estimation of different insulation failure percentages in the presence of unbalanced supply voltages. The method uses artificial neural networks (ANN) and is tested through simulation and experimental investigations. The proposed approach presents a high degree of accuracy in detection and diagnosis of stator winding turn faults in the presence of unbalanced supply voltages condition. The method discriminates between the effect of incipient stator winding turn fault and those due to unbalanced supply voltage. In addition, the proposed approach gives a more significant and reliable indicator for detection and diagnosis of stator winding turn faults in the presence of unbalanced supply voltages conditions.

Motor drive system is considered the most important asset in industrial applications. Detection of broken rotor bars has long been important but difficult job in detection area of incipient motor faults. The need for highly efficient motor control drive systems becomes more and more important. Motors are controlled in closed-loop or open-loop modes of operation. This paper develops a novel approach for fault-detection scheme of broken rotor bar faults for three-phase induction motor using stator current signal. The empirical mode decomposition (EMD) combined with Wigner-Ville distribution (WVD) has been employed for the analysis of stator current signal. Artificial neural network is then used for pattern recognition of broken rotor bar signature. The proposed algorithm offers high performance in detecting broken rotor bar fault. Both simulation and experimental results show that stator current-based monitoring in conjunction with Winger-Ville distribution based on EMD yields a reliable indicator for detection and diagnosis of broken rotor bar faults using artificial neural network. All simulations in this paper are conducted using finite element analysis software. Experimental results validate the simulation and analytical results.