Search

In order to maximise fuel cell reliability of operation and useful life span, an accurate online health assessment of the fuel cell system is essential. Existing algorithms for fault detection in fuel cell systems are based on sensing elements, control methods, and statistical/probabilistic models. In this paper, an artificial neural network (ANN) will be developed to detect and classify faults in proton-exchange membrane (PEM) fuel cell systems. As the ANN model developed within the PEM system relies on the input and output current and voltage, additional sensing devices are not required within the system. Based on an experimental setup using a 3-kW fuel cell system, it was found that the proposed model was able to detect faults associated with the reduction/increase of fuel pressure, H 2 consumption rate, and voltage regulation changes in the dc-dc converter with >90% accuracy. In the proposed model, historical data is required to train and validate the ANN algorithm, but after this is complete, no human intervention is required afterward. [Display omitted] • An ANN-based algorithm for identifying faults in PEM fuel cells. • Voltage regulation of dc-dc converters, fuel pressure, and hydrogen consumption rate can be detected. • A 3-kW fuel cell system was used to demonstrate the effectiveness of the ANN model. • The detection accuracy rate is always over 90% for all seven types of faults considered. • Performance of the ANN algorithm is compared with that of classical methods. [ABSTRACT FROM AUTHOR]

A wind turbine (WT) is an electromechanical system that often operates under a wide range of production conditions. These electrical systems are nowadays expanding rapidly, and they have considerable importance due to their efficiency as renewable energy sources. This led to proposing an innovative and efficient solution with intelligent systems to maintain and ensure the safe and stable operation of these dynamic systems. Maintenance tasks are based on the development of high-performance diagnostic tools, which consist in detecting and locating correctly and upstream the various failures affecting this wind machine. Where, the condition monitoring and supervision systems must rely on reliable fault diagnosis techniques in order to: avoid breakdowns and unscheduled shutdowns, improve their operation, and increase their energetic performances. In order to ensure adequate maintenance actions for the wind system, the purpose of this article is to propose and develop a robust and intelligent fault diagnosis structure. In this work, Kalman filters (KF) as state estimators are used to observe the output states of the sub-systems in order to generate the appropriate residuals evaluated by predetermined thresholds. Adaptive and hybrid network-based fuzzy inference systems (ANFIS) have been employed for the evaluation and classification stages of the detected faults to minimize the degradation of the wind turbine. All possible faults of wind turbine systems, sensors, and actuators are tested and investigated in all parts: pitch angle systems, drive, and generator with converter. The developed fault detection and identification structure are tested on a horizontal WT benchmark model using different scenarios and faults. The simulation results show the ability of the proposed and developed diagnostic methodology to detect the faults occurring efficiently and correctly in the machine. Thus, by using this robust diagnostic strategy, the condition monitoring system can maintain and ensure stable and safe operation to generate sufficient electrical power. [Display omitted] [ABSTRACT FROM AUTHOR]

Currently, microgrids are becoming more prevalent. Therefore, it is crucial to develop robust and reliable microgrid protection schemes. Researchers have recently explored various approaches to microgrid protection, including adaptive protection and AC microgrid protection. The study offers insights into fault detection and localization in microgrid protection, utilizing precise measurements from the point of common coupling (PCC). It distinguishes between fault occurrences and overload cases, addressing various challenges. Additionally, the analysis explores the role of Linear Discriminant Analysis (LDA) within the framework of multi-machine learning techniques, shedding light on its application in microgrid protection. Ultimately, the aim is to bolster microgrid resilience and reliability for end-users and stakeholders. The system model being tested has been created using Matlab/Simulink software and is based on a real system. The training and testing of the algorithms are developed and evaluated using MATLAB tools. The accuracy of the proposed method is demonstrated in the paper, indicating that it can be used to modernize the current protection apparatus in preparation for the eventual implementation of an advanced microgrid station protection system. • High Accuracy in Fault Identification: DTE and ANN models achieve 100% validation accuracy. • Efficient Overload Differentiation: Superior performance in managing faults and overloads. • Enhanced Fault Localization: Leveraging LDA for improved accuracy in fault localization. • Innovative Voltage Signature Signal: Novel composition enhances signal representation. • Promising Role of LDA: Proficient in fault localization, offering valuable insights. [ABSTRACT FROM AUTHOR]

Modeling and recognizing events in complex systems through machine learning techniques is a challenging task. Especially if the model is constrained to be explainable and interpretable, while ensuring high levels of accuracy. In this paper, we adopt a bilinear logistic regression model in which the parameters are trained in a data-driven fashion on a real-world dataset of power grid failure data. The bilinear white-box model – grounded on a specific neural architecture – has been proven effective in classifying faulty states with a performance comparable to several classifiers in technical literature. Additionally, the low computational complexity of the bilinear model, in terms of the number of free parameters, allows gaining insights into the fault phenomenon correlating the events that impact the power grid (exogenous causes) with its constitutive characteristics, thence eliciting the relational information hidden in the data. The proposed model is also able to estimate a vulnerability vector that can be associated, as a suitable characteristic "label", to power grid components, opening the way, as will be deeply demonstrated in the following, not only to predictive maintenance programs or condition monitoring tasks but also to risk assessment and scenario analyses in line with the explainable AI paradigm. • Introduced a bilinear logistic regression model trained on real-world power grid failure data for grid reliability assessment. • Emphasized the role of explainable AI in power grid analysis, highlighting the interpretability of the proposed model. • Developed a method to estimate a vulnerability vector for power grid components and a deep analysis related to the computed weight correlation matrix. • Stressed the importance of semantic specification in modeling complex systems. [ABSTRACT FROM AUTHOR]

[Display omitted] • A neural network is used to detect faults in photovoltaic modules. • According to the shape of the current vs voltage curve, the faults are classified. • The training uses synthetic curves and the test is performed also on experimental data. • The points are normalized and resampled radially to improve the results. • With two hidden layers of 100 neurons each, the hit rate reaches more than 98.5%. In this paper a model-based procedure for fault detection and diagnosis of photovoltaic modules is presented. A four-layered feedforward artificial neural network learns the correlation between the features of the current vs. voltage curve and the environmental variables, which are the irradiance and the temperature. This correlation describes the behavior of the module at normal conditions. Moreover, the effect of anomalous variation of some parameters is learnt and correlated to the shape of the same curve, thus associated to a specific failure mechanism and to some assigned ranges quantifying the fault severity. The neural network is trained by using synthetic curves simulated by employing the single diode model and some well assessed and validated translation formulae. The obtained results over the simulated set of curves with different failures allow to achieve a classification error lower than 1.5%. The proposed approach has been also validated for detecting anomalous increases of the series resistance in a large experimental set of curves; in this case, a classification error of 2.7% has been achieved. [ABSTRACT FROM AUTHOR]

Fault detection and classification are considered as one of the most mandatory techniques in nowadays industrial monitoring. The necessity of fault monitoring is due to the fact that early detection can restrain high-cost maintenance. Due to the complexity of the wind turbines and the considerable amount of data available via SCADA systems, machine learning methods and specifically deep learning approaches seem to be powerful means to solve the problem of fault detection in wind turbines. In this article, a novel deep learning fault detection and classification method is presented based on the time-series analysis technique and convolutional neural networks (CNN) in order to deal with some classes of faults in wind turbine machines. To validate this approach, challenging scenarios, which consists of less than 5% performance reduction (which is hard to identify) in the two actuators or four sensors of the wind turbine along with sensors noise are investigated, and the appropriate structures of CNN are suggested. Finally, these algorithms are evaluated in simulation based on the data of a 4.8 MW wind turbine benchmark and their accuracy approves the convincing performance of the proposed methods. The proposed algorithm are applicable to both on-shore and off-shore wind turbine machines. [ABSTRACT FROM AUTHOR]

The photovoltaic (PV) array is the most sensible element in PV plants, which is subject to different type of faults and defects. Thus, to keep these plants working efficiently they should be monitored and protected carefully. Some faults if they are not detected and isolated promptly they may lead to hazardous risks. The diagnosis of PV systems is widely addressed and recently machine learning (ML) and deep leaning (DL) methods drawn the attention of many researchers. Most applications of ML methods are based on the use of the I–V curves measurement, as enough information and features can be extracted from the curves, to detect and classify faults. These methods showed their capability to classify some faults, like line to line, degradation, disconnected PV modules, partial shading effect, and bypass diode faults. Another approach is based on the use of thermal or electroluminescence images of PV modules/arrays to detect and identify defects, such as hot spot, snails crack, and others. In this paper, different ML and ensemble learning (EL) methods are evaluated for fault diagnosis of PV arrays. The focus is mainly on the detection and classification of some complex faults that may affect the PV arrays, i.e., multiple faults, and faults with similar I–V curves, that are not evaluated before. The results showed the ability of the methods developed to detect faults with very good accuracy (classification rate = number of classified instances/total instances), within 99%, while the classification faults is done with an acceptable accuracy, within 81.73%. Through this study it is shown when really ML and EL methods should be used, and some recommendations, challenges and future directions in this topic are presented. [ABSTRACT FROM AUTHOR]

The low fault current contribution from inverter-interfaced distributed generators (IIDGs) and bidirectional flow of current in the power networks introduces protection challenges when islanded. The existing fault detection scheme based on the transient monitoring function (TMF) fails under such scenarios. A new decentralised machine learning (ML) based fault detection and classification scheme is proposed in this paper using terminal voltage information of IIDG (v o−abc). TMF of v o−abc , along with the zero sequence component of v o−abc , is utilised for training and testing of ensemble bagged trees ML algorithm. The proposed scheme is implemented inside each IIDG, and hence, it does not require communication. It is validated on a modified CIGRE test benchmark microgrid for all shunt faults at various locations with different loading, fault resistances, and inception angles. The accuracy of detection and classification of fault (DCF) during fault cases with unbalanced loads, noise, harmonic loads, and no-fault scenarios such as load and DG switching proves the effectiveness of the proposed method with varying scenarios. The performance of the proposed scheme is found to be superior when compared with the existing threshold based TMF and Hilbert–Huang transform (HHT) based techniques considering accuracy, dependability, security, and response time. • Queries by all the reviewer 3 are addressed in details. • Made suitable changes to the introduction to bring in more clarity. [ABSTRACT FROM AUTHOR]

• Existing PV fault detection techniques detect faults after occurrence. • Our proposed technique detect fault before occurrence. • Our proposed technique forecast fault based on solar cell parameters. • Deep learning-based transformer algorithm with U-Net classifier are used to predict faults. • Four faults with different severity levels are detected before occurrence. According to the US-based National Renewable Energy Lab (NREL), solar energy losses due to faults were 3.5 % in 2004, which increased to 17.5 % in 2018. Therefore, the fault prediction mechanism will enable PV practitioners to reduce losses effectively, enhancing the solar system's efficiency and power output. This paper proposes a deep learning-based Transformer model for robust fault prediction in photovoltaic. Transformer uses attention mechanism that considers data points as a language units "word" and learn dependencies among them to predict upcoming data points. Unlike other forecasting algorithms, our proposed approach does not rely on previous trends. In case of PV faults, trends do not exist. The proposed algorithm utilizes rate of change of solar cell parameters for establishing a trend to forecast faults, enabling proactive fault mitigation. It also classifies faults with different severity levels to identify the level of predictive maintenance required. The proposed approach is extensively evaluated using MATLAB on datasets of several faults with low, medium, and high severity levels. The proposed Transformer model achieves a forecasting mean average error (MAE) of 0.09377. Performance of the proposed forecasting and classification algorithm is compared with existing machine learning-based regression and classification techniques such as KNN, SVM, and NN, where proposed approach outperforms state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

• Detection and classification of faults including high impedance faults in microgrid. • Fault detection and classification from current signals at the substation. • Faulty feeder and sub-feeder identification within half cycle of post fault current signal. • Wavelet based power spectral density of current signals utilized to detect and classify fault and also identify faulty feeder. • Performance of the proposed algorithm verified under dynamic and challenging situations. The recent progress in the area of microgrid protection has led to the application of data-driven and signal processing techniques for fault detection in microgrids. The deployment of inverter interfaced distributed generators makes the traditional protection schemes ineffective. In this paper a wavelet based power spectral density (PSD) has been utilized to detect fault, identify faulty phases and locate fault in the microgrid. Wavelet based PSD decouples the frequency information from the time information and it is the optimal technique to reveal the change of distribution of harmonics in non-stationary faulty signals as for fault analysis it can consider the faulty frequency band instead of one particular harmonic. The proposed technique can detect fault and identify faulty phases by analyzing current signals from the substation only. The normalized value of PSD of three phase current signals has been considered for fault detection and probabilistic distribution of PSD has been considered for faulty phase identification. The normalized value of PSD of the current signals collected from both ends of the main feeders and one end of the non-DG sub-feeder have been used to locate the faulty feeder and sub-feeder. Appropriate threshold values have been considered to detect fault and to locate the faulty feeder. To locate faults, the technique does not require synchronized signals and thus provides a promising and economical solution for inverter interfaced microgrid. It does not require huge amount of data set for training and testing as it is essentially a threshold based technique and also does not require collection of data at high sampling rate. The proposed scheme has been extensively tested for high impedance fault (HIF), unbalanced loading, load switching, variation of DG parameters, measurement uncertainty and presence of noise in the signal with signal to noise ratio up to 10 dB. This technique provides accurate results for fault resistance from 0 Ω to 300 Ω and up to 10% measurement error. The performance of the proposed technique has been verified for feeder length variation up to 20%. Half cycle of current signal only after fault is required and three levels of wavelet decomposition has been considered. The simulation results and comparative assessment demonstrate the efficacy of the proposed scheme. [ABSTRACT FROM AUTHOR]

• Detecting the high resistance internal fault and identifying the correct faulted phase in shunt-compensated lines using phase currents only. • No symmetrical component and frequency-domain analysis are required which reduces the computational burden. • Validation of the proposed algorithm with filed data obtained from a 400 kV line. • Independent of compensation level, fault location, line length and its parameters. The proposed method performs correctly for time synchronization errors, variation in fault location, CT saturation, source impedance variations, different line lengths and power swing, etc. Static synchronous compensators (STATCOMs) are used in the power transmission line to improve the voltage profile and stability. For a fault on the line, additional uneven reactive current from the compensator causes the maloperation of local data-based relaying schemes. In this work, using the current of both ends of the line, internal faults and faulty phases are identified by the ratio of both end's current difference to the local one. Both the magnitude and angle of the ratio are utilized to enhance the dependability and security of the relay operation. The performance of the proposed technique is validated in Western System Coordinating Council (WSCC), a 9-bus system using data obtained from PSCAD. The authenticity of the proposed method is further validated using field data collected from a 400 kV shunt compensated line in the Indian power grid. The suggested strategy is resilient to changes in fault resistance, source impedance, fault location, current transformer saturation, various compensation levels, and high resistance faults (HRF). A comparison of the proposed Algorithm with existing methodologies demonstrates its superiority. [ABSTRACT FROM AUTHOR]

Photovoltaic (PV) power systems have a significant potential to reduce greenhouse gases and diversify the electricity generation mix. Faults and damages that cause energy losses are common during either the fabrication or lifetime of PV modules. The development of automatic and reliable techniques to identify and classify faults in PV modules can help to improve the reliability and performance of PV systems and reduce operation and maintenance costs. A combination of infrared thermography and machine learning methods has been proven effective in the automatic detection of faults in large-scale PV plants. However, so far, few studies have assessed the challenges and efficiency of these methods applied to the classification of different defect classes in PV modules. In this study, we investigate the effect of data augmentation techniques to increase the performance of our proposed convolutional neural network (CNNs) to classify anomalies, between up to eleven different classes, in PV modules through thermographic images in an unbalanced dataset. Confusion matrices are used to investigate the high within- and between-class variation in different classes, which can be a challenge when creating an automatic tool to classify a large range of defects in PV plants. Through a cross-validation method, the CNN's testing accuracy was estimated as 92.5% for the detection of anomalies in PV modules and 78.85% to classify defects for eight selected classes. • Defect classes as soiling, vegetation and cracking have a high within- and between-class variation. • CNNs combined with IR images are able to detect PV module defects with over 90% accuracy. • Convolutional layers can extract important IR patterns that can characterize specific classes of defects. • Diode defects, that cause up to 33% power loss, can be classified by our model with over 90% accuracy. • Oversampling techniques increase the number of IR images improving CNNs training and classification of PV modules defects. [ABSTRACT FROM AUTHOR]

In this paper, a procedure aimed at the automatic extraction of the features from polymer electrolyte membrane fuel cell impedance spectra is proposed. An artificial neural network that is trained by exploiting the similarity learning concept has been used. The network learns the features of the impedance spectra and maps each of them into the embedding space by clustering them accurately and by emphasising differences among spectra corresponding to different faults. The siamese network structure is optimised and the quality of the learnt representation is evaluated by analysing the clusters obtained in the features space. The dataset of experimental spectra has been augmented in two different ways and the results are compared. The clustering quality of the proposed siamese network is compared with the one of other state of the art approaches. [Display omitted] • A Siamese Neural Network is used to analyse the impedance spectra of a PEM fuel cell. • The SNN is able to extract the features corresponding to each faulty condition. • Two different data augmentation approaches are used to train the SNN and their performance is analysed. • The approach is tested on a set of noisy impedance spectra. [ABSTRACT FROM AUTHOR]

• Novel method for the condition monitoring of gas turbines using acoustic emissions. • Wavelet analysis used to transform acoustic emission signals into images. • Deep convolutional neural networks used to classify turbine condition. • Network trained on experimental data with two healthy and two faulty conditions. Compared to vibration monitoring, acoustic emission (AE) monitoring in gas turbines is highly sensitive to changes that do not involve whole-body motion, such as wear, rubbing, and fluid-induced faults. AE signals captured by suitably mounted sensors can potentially provide early indications of abnormal turbine operation before such abnormalities manifest in structural vibration or emitted airborne noise. However, developing an online fault detection system requires extensive real-time data treatment to extract appropriate features and indicators from raw AE records. To build such a system for industrial turbines, researchers need to understand the AE-generating mechanisms associated with turbine operation and the sources of background noise. In this study, we aim to develop such an understanding using a small-scale turbine whose operational conditions can be modified safely to reflect both normal and faulty conditions. Our signal processing approach involves first extracting a time-series envelope using an averaging time selected to enhance major features and eliminate irrelevant noise. We then generate time–frequency features using a continuous wavelet transform, which are used to train a deep convolutional neural network to classify gas turbine conditions. The resulting model demonstrates high accuracy in classifying two normal running conditions and two faulty conditions at various turbine speeds. Overall, the proposed methodology offers a powerful tool for gas turbine condition monitoring, and we make all associated data available in open-source format to facilitate further research in this field.4 [ABSTRACT FROM AUTHOR]

Deep learning-based incipient fault diagnostic techniques have achieved surprisingly well in wind turbines. Due to component failures, wind turbines must undergo active maintenance, substantially influencing revenue and power generation. Unfortunately, there are consistently uneven data distributions between samples with faults and those without faults, resulting in incorrect fault classification. Wind turbine fault classification has a significant data imbalance problem, compromising learning attention for majority and minority classes. Machine learning methodologies based on Generative Adversarial Networks (GAN), over-sampling, and under-sampling techniques for generating synthetic data have been widely employed to address the imbalance data problem. However, the traditional synthetic minority oversampling technique (SMOTE) accomplishes oversampling using linear interpolation between close minority class samples, which could be confusing, subpar, and indistinguishable from the majority class. This study suggests combining over and under-sampling using adaptive SMOTE and edited nearest neighbors (ASMOTE-ENN) that incorporate over-sampling with adaptive SMOTE and under-sampling with ENN to improve the quality of the generated samples. With this resampling technique, noise in an imbalanced dataset is reduced on three levels by using an adaptive nearest neighbor selection algorithm to find the nearest neighbors that are visible. Then use SMOTE to create samples that precisely fall into the minority class, and later use the ENN technique to eliminate instances that contribute to noise afterwards. Resampling data created by combining over- and under-sampling approaches to match the data distribution over all classes is the foundation of the suggested method's efficacy. A hybrid ensemble method is used for effective classification, including boosting, bagging, and stacking techniques. The original unbalanced and balanced data using the ASMOTE-ENN algorithm were classified using the proposed hybrid ensemble method. The classification results show that the proposed strategy is more accurate than a few imbalanced fault diagnosis techniques. • To gain complete knowledge, characteristics of SCADA data are obtained using data analysis. • We present the adaptive SMOTE-ENN algorithm-based data resampling technique to deal with imbalanced data. • Create a hybrid ensemble classification technique to identify faulty and no-fault events effectively. • The results demonstrate that the suggested method outperforms the most popular ML techniques. [ABSTRACT FROM AUTHOR]

Due to the geographically widespread nature of the transmission lines, the chances of fault occurrence are more due to outdoor erection. This makes it inevitable to detect and isolate the faulty section of the system as quickly as possible. The existing fault detection and classification algorithms are limited to a specific configuration of transmission lines. This work has proposed a novel and simple methodology to recognize and classify faults on transmission lines by using single-end current data. A Harmonic Phasor Measurement Unit (HPMU) based on a non-recursive DFT algorithm is used to extract the fundamental and second-order harmonic components from current samples. The Equivalent Fundamental Component (EFC) and Δ EFC are estimated from the extracted data. Subsequently, an Index Matrix (P) is derived to recognize the fault. The non-zero value of the determinant of the index matrix detects the occurrence of a fault in the system. Second-order harmonic fault current coefficients are proposed for fault classification based on the second-order harmonic component extracted from HPMU. The proposed method accurately detects and classifies all types of conventional faults and cross-country faults (CCF) for any configuration of transmission network. The proposed method has also been validated under various abnormal conditions in PSCAD/EMTDC to ensure its efficacy. • | P | and SFCCs are proposed to detect and classify transmission line faults. • The proposed algorithm is simple and set free. • It can be implemented without changing the existing protection infrastructure. • The reliability of the proposed algorithm has been ensured in all abnormal conditions. [ABSTRACT FROM AUTHOR]

• Protection issues of TCSC compensated line integrated with DFIG based Wind Farm. • A novel protection algorithm is proposed to detect the faulty phase in a TCSC compensated line connected with DFIG based Wind Farm. • This proposed method is also validated in a real time platform using dSPACE DS-1104 based prototype processor. • Use of the ITD based technique for detecting faulty phase in TCSC compensated line connected to DFIG based wind farm. • This ITD based algorithm is sensitive for close in fault and high resistance fault in TCSC line with wind farm. Nowadays there is a trend for offshore wind farms to be connected with power grids through a TCSC series controlled compensating device employed in transmission lines for the improvement of large power flows. In such transmission lines, faulty phase identification and ground detection create a challenge for conventional distance relaying schemes because of non-linear output power fluctuation due to variations in wind speed and different firing angle operations of TCSC. Therefore, to overcome the problem of such transmission lines relaying, a novel time-frequency analysis-based intrinsic time scale decomposition (ITD) method is proposed in this paper. The proposed intrinsic time scale decomposition method uses only a three-phase grid side current from relaying end for faulty phase and ground detection and ITD supported random forest classifier for fault classification accuracy approach. The evaluation of the proposed scheme is tested on several fault cases with varying system parameters and non-fault transient events through the EMTDC/PSCAD platform during different wind speeds and varying firing angle operations of TCSC. The performance of the proposed scheme is also verified through a real-time interface dSPACE DS 1104 control prototype hardware. Finally, a comparison analysis with existing techniques demonstrated the effectiveness of the presented method to assess the mentioned problem of the conventional protective relaying schemes. The proposed method demonstrated a 100% accuracy for faulty phase selection and a 99.99% accuracy for fault classification. [ABSTRACT FROM AUTHOR]

In this study, the post-fault synchronized phasors acquired from the local phasor measurement units (PMUs) are used to provide comprehensive fault reporting in Three terminal transmission line system (TTTL). The reporting data incorporates the classification, section identification, and localization of the fault based on the post-fault measurements of positive, negative, and zero sequence quantities. For accurate reporting, a proposed algorithm computes transmission line parameters, on hourly basis, to account for fluctuations in characteristics caused by temperature influences arising from either the loading factor or the climatic vagaries. Upon fault occurrence, the algorithm firstly determines fault classification and faulty section. Thereafter, localization of symmetrical faults is achieved via the positive sequence quantities to solve the positive sequence fault loop equation while that of asymmetrical faults is obtained through the negative sequence quantities to solve the negative sequence fault loop equation. The fault localization speed of the reporting proposed algorithm is attributed to the fact that synchronized data are obtained after one and half cycles from the fault onset. PSCAD/EMTDC platform is dedicated for time-domain investigations, while the proposed algorithm is carried out in MATLAB. The proposed algorithm was tested against 1537 fault characteristics at different locations along the TTTL, including varying fault resistances, loading conditions, and network configurations, in addition to assessing its sensitivity to changes in the TTTL parameters. These tests confirmed the algorithm's reliability to be used with confidence in a variety of scenarios as the only 13 tests were observed with an error of estimation greater than 1%. • This method is reluctant to the TTTL parameters variations, network configurations and loading or environmental influences. • It encloses only two equations to localize eleven types of faults, which reduces the requested computing steps. In addition, no further mathematical calculations are entailed as in several studies to exclude incorrect fault distances and determine the targeted distance. • Its fault classification approach is reliable for all types of faults with different locations and fault resistances along the TTTL owing to the introduced tolerance angle. • As the fault loop equations disregard the fault path resistance, the proposed algorithm can effectively pinpoints faults with high resistances up to 1500 Ω. • Being independent on the zero sequence network, the uncertainty about estimating the zero sequence impedance along the TTTL is avoided. [ABSTRACT FROM AUTHOR]

With data in industrial processes being larger in scale and easier to access, data-driven technologies have become more prevalent in process monitoring. Fault classification is an indispensable part of process monitoring, while machine learning is an effective tool for fault classification. In most practical cases, however, the number of fault data is far smaller than normal data, and this imbalance of dataset would lead to the significant decline in performance of common classifier learning algorithms. To this issue, we propose a data augmentation method, which is based on Generative Adversarial Networks(GAN) and aided by Gaussian Discriminant Analysis(GDA), for enhancement of fault classification accuracy. To validate the effectiveness of this method for imbalanced fault classification, on toy data and the Tennessee Eastman (TE) benchmark process, common oversampling method and the basic GAN are compared to our method, with different classification algorithms. Besides, proposed method is deployed and parallelly trained on Tensorflow platform, which is suitable for applications like data augmentation and imbalanced fault classification in industrial big data environments. • A data augmentation method for imbalanced fault classification. • Generative Adversarial Networks aided by Gaussian Discriminant Analysis. • Superiorities of the model are analyzed and discussed in detail. • Parallel implementation of the model for big data in Tensorflow. • The performance of the method is verified on both toy data and TE process. [ABSTRACT FROM AUTHOR]

Stacked auto-encoder (SAE)-based deep learning has been introduced for fault classification in recent years, which has the potential to extract deep abstract features from the raw input data. However, SAE cannot ensure the relevance of deep features with the fault types due to its unsupervised self-reconstruction in the pretraining stage. To overcome this problem, a stacked supervised auto-encoder is proposed to pretrain the deep network and obtain deep fault-relevant features from raw input data. In each supervised auto-encoder, informative features are learned from the input data with the goal that they can largely distinguish different fault types. By stacking multiple supervised auto-encoders hierarchically, high-level fault-relevant features are gradually learned from raw input data, which can improve the classification accuracy of the classifiers. The proposed SSAE is tested on the Tennessee–Eastman (TE) benchmark process and a real industrial hydrocracking process. The results show the effectiveness and flexibility of SSAE. • A stacked supervised auto-encoder is proposed to pretrain deep network and obtain deep fault-relevant features. • In each supervised auto-encoder, informative features are learned from the input data with the goal that they can largely distinguish different fault types. • High-level fault-relevant features are gradually learned from raw input data by hierarchically stacking multiple supervised auto-encoders. • High classification performance of the proposed method is validated on TE process and an industrial hydrocracking process. [ABSTRACT FROM AUTHOR]

• A kurtogram based deep sequential model is examined for bearing fault classification. • The deep learning sequential model has a good ability to analyze the sequential data. • Analysing the kurtogram as sequential data helps in improving the performance. • Conversion of raw signal to kurtogram helps to generate good feature representations. The condition monitoring of rotating machinery has been widely accepted by the industrial system for intelligent fault diagnosis to achieve sustainability, high performance and provide safety to workers. Therefore, in recent years, artificial intelligence (AI) and signal processing (SP) methods are operated collectively for fault diagnosis. The complex and hybrid input feature set are constructed using SP methods for AI-based fault diagnosis. Thus, over the years the numbers of features in the feature space are increasing to represent the various faults as well as fault severities, and also, different feature selection techniques are operated on feature space to determine the ideal features. Consequently, it is a challenging task to design the dominant feature set for distinguishing the type of defects. Also, the requirement of a number of features is changing due to various working conditions of rotating machinery. Therefore, a new intelligent diagnosis method for fault classification build on the kurtogram and sequence models (SM) of deep learning is proposed in this paper. The kurtogram is a comprehensive tool for providing well-defined information about defects by organizing frequency domain kurtosis values in order at each level. Thus the SM analyzes the kurtogram as sequential data for fault diagnosis and hence helps to eliminate the feature selection exercise for identifying the dominant features from feature space. The proposed method has examined using two vibration datasets of bearing. The result demonstrates that the proposed method has a promising performance and achieves decent fault classification accuracy in comparison with other methods. [ABSTRACT FROM AUTHOR]

• K-means Bayes algorithm is newly proposed for imbalanced fault classification. • The proposed method neither adds samples nor reduces samples, ensuring the quality of the training data. • A MapReduce form of the proposed algorithm is further developed for big data application. • The superiority of the developed method is tested on a benchmark process. Fault classification is an important part of process monitoring in industrial processes. Most conventional fault classification methods are under the assumption that the amount of data in different classes are similar. However, in practice, most of the data collected from industrial process are normal data (majority) and only a few of them are fault data (minority). In other words, fault classification can be seen as an imbalanced data classification problem, which has not been considered in this area to date. In this paper, a K-means Bayes algorithm is proposed to deal with the imbalanced fault classification problem. After that, a MapReduce approach is further introduced to implement the method for fault classification in the big data case. Effectiveness of the proposed method is verified through experiments based on Tennessee Eastman (TE) benchmark process. [ABSTRACT FROM AUTHOR]

Highlights • A new protection scheme is proposed for digital relaying in MGs. • Extensive evaluations of the proposed method in noisy conditions are carried out. • A threshold is used for fault detection and classification using the UT to accommodate low SNRs. • The superior performance of the proposed method over existing methods is shown. Abstract This paper presents a new scheme for microgrid protection using the Time-Time (TT) transform. Initially, the bus currents are processed through the TT-transform to compute the TT-matrixes. To determine the fault patterns, an appropriate index based on the TT-matrix z-score vector is introduced. A threshold is used for fault detection and classification within a few cycles from the fault inception for both grid-connected and islanded modes. A threshold selection approach using the unscented transformation (UT) is used to accommodate low signal-to-noise ratios (SNRs). The protection scheme is evaluated through extensive simulations using PSCAD/EMTDC software for two different microgrids (MGs). Various test scenarios consisting of all shunt and high impedance faults are investigated in MGs with radial and loop topologies for grid-connected and autonomous modes of operation. The results verify that the proposed scheme can reliably detect and classify faults. Furthermore, the proposed method is robust to changes in fault parameters, and performs well under highly noisy conditions. [ABSTRACT FROM AUTHOR]

Residential air-conditioning systems are subject to soft faults, which can reduce their efficiency, capacity, and life expectancy, without being detected by the occupants. Early detection and diagnosis of faults could potentially be done by using machine learning (ML) on measurable features of the air-conditioner, such as refrigerant temperatures and pressures. This paper describes a novel approach to applying ML, which is to train a fault classifier using data from one system, and apply it to different systems. If this is effective, it can significantly reduce the cost of widescale application of ML fault classification. The support vector machine (SVM) algorithm was used. An additional question for practical application is whether the number of sensors required for input feature measurement can be reduced. This work also tests the feature reduction effect, in combination with applying an ML classifier trained on one system, to classify faults in two other systems. The results are disappointing, showing that fault classification is poor when the SVM classifier is trained on a different system than it is applied to. However, with additional work, this method has potential for improvement, possibly to a point where it can be a viable approach for field deployment. [ABSTRACT FROM AUTHOR]

• Speed variation induces fault information imbalance in vibration signals. • Proposed a speed adaptive gate to address the information imbalance. • Empirically validated the effectiveness of the proposed speed adaptive gate. • Speed signals can facilitate deep learning-based fault classification. Rotating machines like wind turbines often work under varying speed conditions. Faults may occur in these machines as time goes on. It is important to detect faults early and classify the type and the severity of an occurred fault, so that timely maintenance could be scheduled. This paper attempts to improve the fault classification accuracy of existing deep learning models when they are used for the fault classification of rotating machinery that operates under varying speed conditions. To achieve this goal, we firstly investigate the effects of speed variation from the perspective of deep learning. We find that the fault information in vibration signals is imbalanced due to speed variation. The imbalance future deteriorates the fault classification accuracy of deep learning models. That is, the accuracy does not evenly distribute across speed but increases with speed. We then propose an auxiliary branch named speed adaptive gate (SAG) for existing deep learning models to address the speed induced fault information imbalance. The SAG takes the speed signal as the input and outputs speed adaptive gate values. The gate values control the information flow of deep learning models so that the effects of speed variation are mitigated. Case studies with two baseline models, i.e., a convolutional neural network (CNN) and a residual network (ResNet), over two experimental datasets, i.e., a planetary gearbox dataset and a fixed-shaft gearbox dataset, have validated the effectiveness of the proposed SAG and its superiority over existing approaches. [ABSTRACT FROM AUTHOR]

The high current carrying capacity of the power electronics based switches has gained much popularity to the VSC-based HVDC system. The challenge faced by the protection schemes is to identify and isolate the relevant fault within 5 to 6 ms. Moreover, the relay algorithm will not activate the circuit breaker to operate during other disturbances including external and non relevant internal faults. The Discrete Wavelet Transform (DWT) and Support Vector Machine (SVM) classifiers are used in this paper to formulate a protection method. Different types of faults are simulated on the using PSCAD/EMTDC software. Depending upon the relevant frequency components accompanied by the DC link voltage and current during fault, nine relevant features are judiciously selected. The acquired features are then used to train and test SVM for fault identification and classification. The proposed solution does not rely on communication as it only takes information from one end. Furthermore, the approach effectively distinguishes internal faults, external faults, and non-fault disturbances in a multi-terminal VSC-HVDC system. This method is capable of classifying faults in both fault-resistant and noisy situations. The extensive execution time study demonstrated that the proposed method is fast and computationally less expensive fault detection and classification technique. • The DC voltage and current are analysed to obtain the SVM-classifier features. • Method distinguishes internal faults, external faults, and non-fault disturbances. • Method is capable of classifying faults in both fault-resistant and noisy situations. • Execution time within 6ms proved the method is a computationally less expensive. [ABSTRACT FROM AUTHOR]

• The maximum fault detection time is found to be less than half-cycle, which helps in faster maintenance and classification accuracy is 100%. • The algorithm proposed for fault location is based on k-Nearest neighbor regression method. The achieved accuracy of fault location is found to be better than the methods proposed in the literature. • Numerous case studies have been done to check the robustness of the algorithm. This paper presents a novel protection algorithm based on K-medoids clustering and weighted k-Nearest neighbor regression. K-medoids clustering is used for fault detection and classification, while weighted k-Nearest neighbor regression is used to locate the fault. The three phase current signals are sampled at substation with frequency of 3.84 kHz and then decomposed with a db1 mother wavelet. Using the moving window of one cycle, k-medoids clustering is applied to wavelet approximate coefficients over a cycle to obtain two medoids. The difference of medoids is defined as fault index, which is compared with threshold to detect and classify the faults. Various statistical features are computed from post fault approximate wavelet coefficients obtained are fed to k-Nearest Neighbor classifier detect the two nearest buses. Followed by this detection the fault location between these two busses is estimated using weighted k-NN regression. The process of detection, classification and location of fault is accomplished within half a cycle. The various case studies used for established the robustness of the algorithm include type of fault, fault impedance and fault incidence angle and fault location. The proposed algorithm is proved to be not affected by DG trip, islanding and noise. [ABSTRACT FROM AUTHOR]

• A faulted phase selection method using locally measured instantaneous incremental currents. • Enhances the relay operating time and enable single-pole tripping schemes. • Does not need a communication channel or voltage transducers. • can be implemented in systems with low sampling rates and avoids the need to use transducers with high bandwidth. • Applicable to a wide range of line lengths, almost independent of fault resistance and not affected fault inception angle and require only three threshold settings. A novel, simple and fast method to identify the faulted phases using the maximum rate of change of incremental current (ROCOIC) is proposed. Six indices are calculated considering a pairwise comparison of ROCOIC. The decision-making process is based on selecting which two indices are the largest among all. The algorithm requires only three thresholds to detect fault inception and to identify the ground and three-phase faults. Simulation studies show that the method is robust and not affected by fault inception angle, fault location, and fault resistance. The proposed method only requires locally measured current signals. Since the method relies on the fundamental component of instantaneous incremental current, the proposed method can be implemented in systems with low sampling rates and avoids the need to use transducers with high bandwidth. [ABSTRACT FROM AUTHOR]

This paper presents the development of a novel data-driven fault detection and classification scheme for DC faults in multi-terminal HVDC transmission system which incorporates superconducting cables and modular multi-level converters. As the deployment of superconducting cables for bulk power transmission from remote renewable generation is progressively increasing in the future energy grids, many fault-related challenges have been raised (i.e., fault detection, protection sensitivity/stability). In this context, the applications of Artificial Intelligence techniques have started to be considered as a powerful tool for the development of robust fault management solutions. The proposed artificial intelligence-based method utilizes local current and voltage measurements to detect and classify all types of faults on the DC cables and DC buses, without the requirement of measurements exchange among different DC substations. The performance of the proposed scheme has been assessed through detailed transient simulation analysis and the results confirmed its effectiveness against a wide range of fault conditions (i.e., various fault types, fault locations and fault resistances). Furthermore, the feasibility of the developed scheme for real-time implementation has been validated using real-time software in the loop testing. The results revealed that the proposed algorithm can correctly, and within a very short period of time (i.e. less than 2 ms) detect and classify the faults within the protected zone and concurrently remain stable during external faults. Additionally, the generalization capability of the algorithm has been verified against influencing factors such as the addition of noise, highlighting the robustness of the presented scheme. • The deployment of DC Superconducting Cables in HVDC systems has started to be considered a promising solution for bulk power transmission from remote renewable generation units. • The inherent fault current limiting capability and the unique electro-thermal properties of the SCs have introduced many fault management issues in power systems. • For the protection of SCs in meshed HVDC systems, schemes with increased sensitivity, discrimination capability and high operational speed are required. • Artificial-Intelligence-based protection schemes overcome the limitations of existing protection methods in the data-rich power systems, providing fast, and accurate fault detection and classification. • XGBoost-based fault detection and classification algorithm performs effectively under various faults occurred on SCs and DC busbars, providing a reliable protection solution and accelerating the uptake of superconducting technology. [ABSTRACT FROM AUTHOR]

• A new method incorporating active learning to semi-supervised industrial fault classification is proposed to solve the lack of labeled training data. • The limited capacity of semi-supervised FDA can be effectively improved by introducing the active learning, and the semi-supervised FDA can utilize the information of unlabeled data to help active learning reduce the amount of labeled data. • At each iteration, the data having the greatest entropy will be labeled manually to update current model. • The iteration of active learning will stop when current model is confident of all unlabeled data. • Compared with semi-supervised FDA, ALSemiFDA can get better discriminant projection vectors and can handle more situations. Compared with active learning based FDA, ALSemiFDA needs less manually labeled data. The performance of Fisher discriminant analysis (FDA) method is highly depended on the labeled data. While obtaining the true labels of the industrial data is often time-consuming and expensive in practice. Although there are some researches on semi-supervised methods based on the FDA to solve this problem, they will fail when the labeled data are not satisfied with some special conditions. To improve the methods' applicability, this paper proposes an active learning based semi-supervised FDA model for industrial fault classification. This method bridges labeling important unlabeled data of active learning and learning from unlabeled data of semi-supervised learning. The performance has been greatly improved with the information from the new labeled data, and the amount of new labeled data has been reduced by additional unlabeled data. The experimental analysis on the two dimensional data sets indicates that active learning and semi-supervised learning are complementary for each other. Finally, the experiments carried out on the UCI benchmarks and Tennessee Eastman process (TEP) prove the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Highlights: • Fault transient voltage is analyzed under condition of distributed-parameter dc line. • Fault transient voltage is used to discriminate between internal and external faults. • Faulty pole selection is based on the change rate of positive- and negative-pole currents. Abstract Multi-terminal flexible HVDC grids are required due to the high penetration of renewable energy. However, they are vulnerable to dc faults. To ensure their high reliability and continuous operation, fast and selective protection schemes are indispensable. This paper analyses the transient voltages in the case of dc line faults in modular multilevel converter (MMC)-based HVDC grid. Then, making use of the boundary effect of current-limiting reactors, a fast non-unit dc line protection scheme based on fault transient voltage is proposed. In addition, the faulty pole identification criterion is constructed based on fault currents. A MMC-based four-terminal HVDC grid is modelled using PSCAD/EMTDC, and extensive simulations are conducted to validate the effectiveness of the proposed scheme. Simulation results show that the proposed protection method not only can detect faults rapidly and accurately, but is also effective in detecting large-resistance faults. [ABSTRACT FROM AUTHOR]

• A new scheme is developed to identify the fault using reactive power measurements. • The reactive-power superiority compared with active measurements is investigated. • The optimal mother wavelet is selected based fault identification scheme. • The proposed scheme is immune to different fault parameters. • Testing the fault identification scheme at different operating conditions. The process of fault-identification in teed transmission lines (TTLs) has considerable attention in recent years. In this line, this paper proposes a novel scheme for fault detection, classification and faulted-phase/s identification on the basis of reactive-power components of TTLs. The current and voltage measurements are collected from a single-end measuring system of teed circuits. Based on the filtering of measured currents and voltages signals using improved recursive wavelet transform, the reactive-power measurements are computed. Thereafter, the Wavelet Multi-Resolution Analysis Technique (WMRAT) is employed for decomposition of reactive-power measurements. An improved recursive wavelet (IRW) transform that has good time-frequency characteristics is proceed to remove dc offsets and harmonics from the measured signals. The ATP/EMTP package and MATLAB program are used to simulate the faulted cases and verify of operating the proposed fault-identification scheme. Simulation outcomes show that the proposed scheme has the capability to detect and classify different fault types in TTLs. The presented scheme is designated to be insensitive to different fault resistances, different fault locations, and inception angles. In addition, the proposed scheme is also tested under different technical aspects such as different sampling rates, different locations of the single-end measuring system, loading change and source parameters variation. The superiority of reactive-power compared with active power measurements during faults occurrence is also investigated to show the notability of the proposed scheme. The obtained results corroborate that the proposed scheme can be correctly identified different fault types and locations along the line. [ABSTRACT FROM AUTHOR]

Highlights • Shaded PV cells degrade the performance of PV systems. • I-V curves contain useful information for health monitoring. • Data driven method for fault diagnosis can deal with non-controlled solar irradiance. • Principal component analysis has proven its efficacy under real working conditions. Abstract Health monitoring and diagnosis of photovoltaic (PV) systems is becoming crucial to maximise the power production, increase the reliability and life service of PV power plants. Operating under faulty conditions, in particular under shading, PV plants have remarkable shape of current-voltage (I-V) characteristics in comparison to reference condition (healthy operation). Based on real electrical measurements (I-V), the present work aims to provide a very simple, robust and low cost Fault Detection and Classification (FDC) method for PV shading faults. At first, we extract the features for different experimental tests under healthy and shading conditions to build the database. The features are then analysed using Principal Component Analysis (PCA). The accuracy of the data classification into the PCA space is evaluated using the confusion matrix as a metric of class separability. The results using experimental data of a 250 Wp PV module are very promising with a successful classification rate higher than 97% with four different configurations. The method is also cost effective as it uses only electrical measurements that are already available. No additional sensors are required. [ABSTRACT FROM AUTHOR]

Highlights • A complementary fault classification stage for Thermal NDT-E of electrofusion PE joints. • Combined heat pulsing, IR imaging, and machine learning for classification accuracy. • The ovality diagnosis was more accurate than unalignment fault. • GLMNet (93.75%) was the best classifier, followed by Random Forests (84.21%) and k-means (70.37%). Abstract The capability of conveniently classifying the fault types in the electrofusion joints can certainly increase the security of polyethylene gas pipelines. Therefore in the current study, we use machine learning to push the horizons of our recent thermal pulsing and IR thermography method, to identify ovality versus unalignment faults. To do so, we extend our experimental IR-thermography data bank and then apply k -means, Random Forests and GLMNet algorithms in a two stage approach. The overall classification accuracy for k -means and Random Forests were 70.37% and 84.21% respectively; GLMNet could successfully outperform the others with a classification accuracy of 93.75%. [ABSTRACT FROM AUTHOR]

Abstract This paper presents a new approach for faults classification in analog integrated circuits using a multiclass adaptive neuro fuzzy inference system classifier. This is carried out to assist analog circuit's faults diagnosis suffering from inaccurate faults classification on one hand, and to lessen computational burden on the other hand. This has been achieved from features number reduction. These features serving as input feature vector are extracted from the selected circuits (CUT) frequency and transient responses under both fault free and faulty conditions. The considered faults are resistors and capacitors values variations of about 50% low and high from their nominal ones. The method accuracy has been validated with three experiment circuits, the Sallen Key band-pass, the four opamp biquad high-pass and the leapfrog filters. The obtained results reveal a high level of efficiency with an accuracy average reach to 99.76%. Hence, the proposed method has shown a good performance in term of fault classification accuracy when compared with those of both the Artificial Neural Networks (ANN) approach and the fractional Fourier transform (FRFT) method based on a statistical property. Highlights • A neuro-fuzzy classifier for faults classification in analog circuits is proposed. • Monte-Carlo simulation is used for features extraction. • Features selection technique is used to reduce the computational burden. • Classifiers fusion and winner takes all rule for multiclass fault discrimination. [ABSTRACT FROM AUTHOR]

This paper proposes a novel protection scheme for detection and classification of internal faults in bipolar HVDC transmission lines using the K -means data description (KMDD) method. In the proposed protection scheme, the inverter-side dc voltage and current signals are utilized. Relatively short-time windows are considered for these signals, and the sum values of data windows are calculated. In the preparation stage, for each dc fault type, the KMDD method is applied to some post-fault data generated in various conditions. Then, the obtained centroids and thresholds are used for detecting and classifying new internal dc faults even with unseen conditions. The performance of the proposed method is evaluated for 4320 internal and 2816 external fault cases in a 1000 km bipolar overhead HVDC transmission line under various conditions not seen in the preparation stage. Besides simplicity and low sampling frequency requirement of the proposed protection scheme, the obtained results show that it is fast and accurate enough for the internal dc faults, and also it is stable during the external ac faults and the pre-fault normal conditions. [ABSTRACT FROM AUTHOR]

Conventional relays fail to detect the high impedance fault (HIF) in distribution lines, as the change in the current magnitude is very negligible compared to conventional relay settings. Moreover, the incorporation of distributed generators in the distribution lines changes the fault current level, which makes the HIF-detection more complex. In this paper, a fuzzy-based intelligent fault detection and classification scheme is developed for the distribution lines integrated with DGs. Two different fuzzy inference systems (FIS) are modelled in each phase to detect the fault. The first FIS identifies the high magnitude of fault current associated with normal shunt faults; and the second FIS identifies the small magnitude of current owing to occurrence of HIF. The proposed scheme uses the features extracted from the Teager energy operator. An extensive study is conducted and the response time is found to be around ¼-1 cycle. Results validate the efficacy of the proposed scheme. [ABSTRACT FROM AUTHOR]

Supervised learning for chemical process fault classifications need for a large amount of sampling data with fault labels which significantly consume expert manpower in general. It is seen that active learning is becoming an effective solution for this problem. To improve the quality of sampling data used for expert labeling, this paper proposes sampling evaluation metrics combing information entropy and similarity for active learning strategies in chemical process fault classifications. Specifically, in response to the insensitive direction of the cosine similarity, the Euclidean distance is merged with evaluate the sampling representativeness. Further, combing with information entropy, a comprehensive evaluation index (E-ECos) is established to select the most valuable samples for expert labeling, adding to the training set of classifiers, by labeling them by experts and adding them to the supervised learning training set of the classifier, the classification learning performance can be effectively improved with a relatively small number of actively labeled samples. On the TE process simulation platform, a deep learning network and an ensemble learning network based classifiers are employed to classify industrial process faults, respectively. Experimental results demonstrate the effectiveness of the proposed method. • This paper proposes a fault classification method with an improved active learning. To improve the quality of sampling data used for expert labeling, this method combing information entropy and similarity for active learning strategies in chemical process fault classifications. • Sample similarities: This paper comprehensively evaluates the similarity of samples by evaluating the cosine similarity and Euclidean distance between samples. On the one hand, Euclidean distance complements the problem that cosine similarity is insensitive to absolute direction. On another hand, the Euclidean distance is based on the absolute value of dimensional features. • E-ECOS: Taking advantage of the different sample selecting evaluation metrics of the two active learning algorithms, we establish a comprehensive evaluation index in terms of sample informativeness and sample representativeness. • Future potentiality: The following research is suggested to combine active learning with auto-machine labeling to further reduce the consumption of manpower for fault sample labeling. [ABSTRACT FROM AUTHOR]

In this work, a new fault classification and phase selection methodology for half-wavelength (HWL) transmission lines is developed. Initially, the basic principles of phase currents transformations are presented. Then, an innovative faulted phase selection methodology based on the variation of those currents is developed. The behavior of the electrical quantities related to short circuits in such long lines is very different when compared to conventional lines. The faulted phase selection method employed in this work is built on the fact that those decoupled currents are expected to not vary or vary less than the others, depending on how many phases are in short circuit. The robustness of the method was analyzed considering different load conditions and fault resistances. Furthermore, different sampling rates were considered for the phase currents. Extensive simulation studies showed that the developed method can accurately identify the fault types and the faulted phase with a relatively low sample rate. Moreover, the simulation results showed that the proposed method is not affected by the presence of inter-harmonics and responds precisely with high-resistance faults. The proposed algorithm presented errors smaller than 0.04%. • A new method for ground and phase fault classification based on decoupled currents by means of modal transformation was developed in this study. • The decoupled currents are obtained by means of a modal transformation. • The developed method is robust to faults with high resistance. • The developed method is not affected by the presence of interharmonics. [ABSTRACT FROM AUTHOR]

In this paper, a hierarchical clustering selection based weighted random forests scheme is proposed for fault classification in complex industrial processes. Model diversity and the strength of each model are deemed to be two key issues for the performance of ensemble learning method. To improve the diversity between classification trees and the performance of individual classification trees in random forests, the hierarchical clustering method is applied for offline model selection in random forests, which can simultaneously reduce the online fault classification complexity. Meanwhile, the weighted voting rule is used in random forests instead of majority voting, in order to boost the good performance models and weaken the bad ones. Detailed comparative studies between proposed method and conventional methods have been carried out through the Tennessee Eastman (TE) benchmark process. [ABSTRACT FROM AUTHOR]

This paper proposes a hierarchical sparse artificial neural network for classifying the faults in dynamic processes base on limited labeled data. The Stacked auto-encoders (SAE) is developed to extract features from different faults. Each neural network in the proposed SAE is given a sparse constraint to learn a Sparse Stacked auto-encoders (SSAE). Then, the Dynamic time window is combined into SSAE to build Dynamic Sparse Stacked auto-encoders (DSSAE). DSSAE model based semi-supervised fault classification scheme is then formulated to classify the dynamic faulty data. Simulation studies on the Tennessee–Eastman (TE) benchmark process evaluate the performance of the developed method, which indicate that the DSSAE method performs better than both SAE and SSAE. [ABSTRACT FROM AUTHOR]

A large amount of data is generated through monitoring, maintenance, repair and diagnostics of power transformer. However, all these data cannot preindicate the exact type and probability of failure. To overcome the problem this paper presents artificial intelligence based methodology for power transformers fault detection and classification. The possibility of presented monitoring methodology is to assist the operator’s engineers in decision making about urgency of intervention and type of maintenance of power transformer. The article analyzes the application of Mamdani-model and Sugeno-model in fuzzy expert system for fault diagnosis based on the current state of the power transformer. Paper presents two case studies with one unique and five separate controllers. In the first case inputs of controller are results of on-line and off-line transformer tests: age, the overheating temperature of the hot spot, frequency response analysis, temperature of insulation, dissolved gas-in-oil analysis, tgδ and polarization index. Second case study in addition to the existing inputs includes previous measurements. A fuzzy controller (FC) is designed to characterize the operating condition and to determine the urgency of intervention with possibility to indicate probability of specific type of failure. Cumulative probability of occurrence of the faults is also observed in second case study. FCs are tested based on real measurements from Serbian transmission system. The results show acceptable effectiveness in detecting different faults and might serve as a good orientation in the power transformer condition monitoring. [ABSTRACT FROM AUTHOR]

A bearing is an essential component in rotating machinery, one of its principal cause of failure, and its health condition is directly related to the safety and effective operation of such machinery. To the best of our knowledge, it is the first time that a probabilistic fuzzy system is applied to bearing fault classification. The type of probabilistic fuzzy classifier considered is a parsimonious fuzzy rule based model where each rule can diagnose a set of faults each one with its probability. For this kind of real world application, it is desirable to develop interpretable and accurate MIMO fuzzy systems, able to deal with the dimensionality and uncertainty present in data (vibration signals). For parameter estimation we adopt a two steps sequential state-of-the-art data-driven method. First, the antecedents of the rules are estimated using an iterative supervised clustering algorithm. Based on the antecedents the consequent parameters are then estimated. For this, a new method for consequent estimation is proposed. This is based on the observation that for defining a rule consequent not all training data within the region of applicability of that rule are equality relevant. Two criteria for selecting a relevant region in the feature space for consequent parameter estimation are proposed. Results show a statistically significant improvement on the performance of probabilistic fuzzy diagnosers trained with the proposed method, independently of the criterion used for defining the relevant region, when compared with the above mentioned state-of-the-art method. Moreover, the proposed consequent parameter estimation method practically has no overhead on the overall training of the diagnoser. Results show that an equilibrium can be found between the model level of detail and its accuracy. However, when accuracy is the sole comparison criterion, the proposed probabilistic fuzzy systems systematically matches the performance of other data-driven models like distance based methods (K-nearest neighbors), connectionists (probabilistic neural networks), or maximum margin classifiers (support vector machines). [ABSTRACT FROM AUTHOR]

Performance of the most existing fault detection and classification methods can only be guaranteed when each of their own assumptions are met. In other words, a method works well in one condition may not perform well in another. In this paper, a new analytic hierarchy process (AHP) based fuzzy decision fusion system is proposed to tackle the fault classification problem. The AHP approach is introduced to determine the priorities of different classifiers, which are further utilized as the weights in ensemble system. Comparing to conventional equal weighted fusion system, the proposed fuzzy fusion system is able to provide more rational and convincing fault classification result. Effectiveness of the proposed fuzzy fusion system with model evaluation is verified through the Tennessee Eastman (TE) benchmark process. [ABSTRACT FROM AUTHOR]

This paper presents a non-unit protection scheme for series capacitor compensated transmission lines (SCCTL) using discrete wavelet transform and k-nearest neighbor (k-NN) algorithm. All the protective relaying functions such as fault detection, fault classification, faulty phase identification and fault location estimation have been considered in this work. Such a comprehensive work providing all protective relaying functions for protection of double circuit SCCTL utilizing k-NN has not been reported so far. The signal processing and feature extraction are done using discrete wavelet transform due to its capability to differentiate between high and low frequency transient components. For fault detection and classification, only approximate wavelet coefficient of current signal up to level 1 has been used; while for k-NN location estimation, both voltage and current signals of the two circuits are decomposed up to level 3 have been used. Finally, the standard deviation of one cycle pre-fault and one cycle post-fault samples of the approximate wavelet coefficients are calculated to form the feature vector for the k-NN-based algorithm. The performance of the proposed technique is evaluated for large number of fault events with variation in fault type including inter-circuit faults, fault inception angle, fault location and fault resistance. The change in position of series capacitor and different degree of compensation has been discussed. The accuracy of the proposed k-NN-based fault detection and classification module is 100% for all the tested fault cases with a decision period of less than half cycle. The k-NN-based fault location scheme estimates the location of fault with ≤1% error for most of the tested fault cases, which is an exceptional attribute of the proposed scheme as compared with 10–15% error of conventional distance relaying scheme. [ABSTRACT FROM AUTHOR]

Several methods for the location and classification of faults in power transmission lines using computational intelligence and digital signal processing techniques have been described in literature. Artificial neural networks (ANNs) and wavelet transform (WT) have drawn significant attention lately, but they present some drawbacks when dealing with power systems faults where data are often contaminated by noise. This paper proposes an approach by combining independent component analysis (ICA) with travelling wave (TW) theory and support vector machine (SVM). The approach is adequate to locate and recognize faults in high-voltage (HV) transmission lines, while the acquired signals are noisy. Experiments performed for distinct types and locations of faults in a real transmission line model have shown that the proposed combined methods are able to provide excellent performance in fault location. The obtained errors are lower than 1% and accuracy is 100% for the classification of fault signals with noise. It can be stated that this method presents better performance than those regarding the main conventional techniques such as wavelets and neural networks in the presence of noise. [ABSTRACT FROM AUTHOR]

The fault classification of chemical processes based on supervised learning generally demands for enough labeled fault samples, which represents a big challenge for practical engineering issues. In the industrial field, when capturing an appearance of process abnormities, human operators usually take account of correlations of relevant variables so as to locate process variables associated with the abnormity before further classifying or diagnosing the corresponding faults. In this context, accurate abnormity localizations can be an effective help for classifying the fault type. Nowadays, deep learning based fault classifications have been extensively studied and widely circulated in the literature. However, the existing approaches rarely concern the supporting role of abnormity locations for fault classifications. Motivated by this observation, this paper proposes a multi-task learning (MTL) based convolutional neural network (CNN) model for concurrent abnormity locations and fault classifications for chemical processes. Specifically, fault samples are labeled with both abnormal variable locations and fault categories before constructing a multi-task learning CNN classifier performing an auxiliary task, abnormity locations, and a primary task, fault classifications. In addition, dynamic task weights are assigned to facilitate task contributions. This method can enjoy effectively reducing the necessary amounts of labeled fault samples in achieving satisfactory fault classification performances. Experiments are carried out on the TE process simulation platform and an industrial coal gasification plant along with comparisons with conventional fault classification methods, verifying the effectiveness of the proposed method. • A CNN model based on multi-task learning is proposed to better adapt to fault diagnosis of chemical industrial processes. • An adaptive multitask learning algorithm is proposed to balance the multi-task training process effectively. • Abnormity locations can improve the accuracy of fault classification and the convergence of the overall network training. • Abnormity locations task branches can share tags to the primary task thus reducing the impact of insufficient fault labels. • Future potentiality: Since MTL method doesn't require architectural constraints, it can be applied to other neural networks. [ABSTRACT FROM AUTHOR]

Rolling element bearings are essential components of a wide variety of industrial machinery and the leading cause of equipment failure. The prediction of Remaining Useful Life (RUL) and fault diagnosis is an essential component of equipment diagnosis and health management. Real-time Conditional Based Monitoring (CBM) faces a significant challenge in the Internet of Things and Industrial 4.0 eras due to automatically processing enormous quantities of data. A novel three-phase architecture based on deep learning has been proposed to perform real-time monitoring. The model is end-to-end adaptive, and the data is sequentially processed. The proposed hybrid approach incorporates change-point detection, RUL prediction, and fault classification. Adaptive data pre-processing is employed in conjunction with the unsupervised CPD module, followed by a model training on bearing data deterioration to increase efficiency. The autoencoder employed for CPD is also utilized to supply additional features from the bottleneck layer to acquire an enhanced Health Index (HI). The experiments were conducted on NASA's prognosis and diagnosis datasets, and the proposed method's effectiveness was compared to other benchmark approaches. The results show that the proposed methodology can execute real-time CBM efficiently and reliably, outperforming other methods. [Display omitted] • Real-time condition-based monitoring of bearings has been proposed. • The methodology can determine RUL as well as bearing fault categorization. • The methodology is based on a three-phase architecture that is end-to-end adaptive. • For real-time monitoring, data is sequentially processed between each phase. • Improved accuracy is achieved as the methodology integrates change-point detection. [ABSTRACT FROM AUTHOR]

• Fault classification and localization are carried out in a photovoltaic system. • A multi-output scalable machine- and deep-learning approach is developed. • A 50% reduction in the number of sensors needed per string is achieved. • Deep Learning models are more resilient to noise and are more accurate. • Accuracy up to 99.98% and cross-entropy loss as low as 0.002 are attained. Fault classification and localization are imperative to maintaining an efficient photovoltaic (PV) system. Due to the environmental factors that PV systems function in, they can be prone to multiple types of fault conditions such as string-to-string (SS), string-to-ground (SG), and open-circuit faults (OC). If left undetected, these faults can lead to power loss and eventually system failures. To help mitigate such outcomes, this paper presents an approach to detect, classify, and locate SS, SG, and OC faults using multi-output deep learning (DL) algorithms: convolutional neural networks (CNN), long short-term memory (LSTM), and bi-directional long short-term memory (Bi-LSTM) networks. The proposed approach reduces the number of sensors needed per string by 50% compared to the previous approaches in the literature. It also allows for increased scalability and the development of a multi-label dataset through an uncoupled modeling scheme. Fault classification and location are achieved with accuracies of 99.94% and 99.54%, respectively. Further, a comparison is conducted among the proposed DL and conventional machine learning algorithms using multiple evaluation metrics. The proposed method is validated by testing multiple datasets and adding noise. The results obtained reflect high reliability and demonstrate the effectiveness and scalability of the proposed multi-output DL approach. [ABSTRACT FROM AUTHOR]

The conventional distance relaying algorithms are unable to detect the inter-circuit faults, cross-country faults, high resistance faults which may occur in a double circuit line. This paper presents combined Discrete Wavelet Transform (DWT) and Support Vector Machine (SVM) based directional relaying and fault classification scheme including inter-circuit faults, cross-country faults and high resistance faults. SVM modules are designed for forward or reverse fault identification and fault classification using single terminal data. The 3rd level approximate discrete wavelet transform coefficients of three phase current signals only have been used. Proposed method is tested with variations in fault type, fault location, fault inception angle, fault resistance, inter-circuit faults, and cross-country faults. The proposed method based on SVM does not need any threshold to operate which is an exceptional attribute for a protective function. As SVMs are not based on comparing with some threshold, rather initially the SVMs are trained with the wide variety of fault patterns which is an offline process and then the trained SVMs are tested online to detect and classify the fault within short time. The test results show that all types of shunt faults can be identified within half cycle time. The proposed scheme offers both primary protection to 95% of the line section and also backup protection to 95% of the adjacent reverse and forward line section also. [ABSTRACT FROM AUTHOR]