Partial Discharge Classification Research Articles

Towards Generalizable Classification of Partial Discharges in Gas-Insulated HVDC Systems Using Neural Networks: Protrusions and Particles

Towards Generalizable Classification of Partial Discharges in Gas-Insulated HVDC Systems Using Neural Networks: Protrusions and Particles

Novel Self-Organizing Probability Maps Applied to Classification of Concurrent Partial Discharges from Online Hydro-Generators

In this paper, we present an unprecedented method based on Kohonen networks that is able to automatically recognize partial discharge (PD) classes from phase-resolved partial discharge (PRPD) diagrams with features of various simultaneous PD patterns. The PRPD diagrams were obtained from the stator windings of a real-world hydro-generator rotating machine. The proposed approach integrates classification probabilities into the Kohonen method, producing self-organizing probability maps (SOPMs). For building SOPMs, a group of PRPD diagrams, each containing a single PD pattern for training the Kohonen networks and single- and multiple-class-featured samples for obtaining final SOPMs, is used to calculate the probabilities of each Kohonen neuron to be associated with the various PD classes considered. At the end of this process, a self-organizing probability map is produced. Probabilities are calculated using distances, obtained in the space of features, between neurons and samples. The so-produced SOPM enables the effective classification of PRPD samples and provides the probability that a given PD sample is associated with a PD class. In this work, amplitude histograms are the features extracted from PRPDs maps. Our results demonstrate an average classification accuracy rate of approximately 90% for test samples.

A Review on the Classification of Partial Discharges in Medium-Voltage Cables: Detection, Feature Extraction, Artificial Intelligence-Based Classification, and Optimization Techniques

Medium-voltage (MV) cables often experience a shortened lifespan attributed to insulation breakdown resulting from accelerated aging and anomalous operational and environmental stresses. While partial discharge (PD) measurements serve as valuable tools for assessing the insulation state, complexity arises from the presence of diverse discharge sources, making the evaluation of PD data challenging. The reliability of diagnostics for MV cables hinges on the precise interpretation of PD activity. To streamline the repair and maintenance of cables, it becomes crucial to discern and categorize PD types accurately. This paper presents a comprehensive review encompassing the realms of detection, feature extraction, artificial intelligence, and optimization techniques employed in the classification of PD signals/sources. Its exploration encompasses a variety of sensors utilized for PD detection, data processing methodologies for efficient feature extraction, optimization techniques dedicated to selecting optimal features, and artificial intelligence-based approaches for the classification of PD sources. This synthesized review not only serves as a valuable reference for researchers engaged in the application of methods for PD signal classification but also sheds light on potential avenues for future developments of techniques within the context of MV cables.

Classification of Partial Discharge in Vehicle-Mounted Cable Termination of High-Speed Electric Multiple Unit: A Machine Learning-Based Approach

This paper presents a machine learning-based approach to identify and separate partial discharge (PD) and two typical pulse interference (PI) signals in the vehicle-mounted cable terminations of high-speed electric multiple units (EMUs). First, a test platform was established to capture PD and two typical PI signals in these terminations. The acquired signals were then processed using the square envelope method to extract feature parameters, such as the rise time proportion, the left–right symmetry, and the upper–lower symmetry. PD signal classification was carried out on these signals, utilizing waveform parameters derived from a hierarchical clustering algorithm. The results validate that the extracted feature components effectively classify and separate PD and two typical PI signals in the vehicle-mounted cable terminations of high-speed EMUs.

Partial Discharge Pattern-Recognition Method Based on Embedded Artificial Intelligence

This paper proposes a method for detecting and recognizing partial discharges in high-voltage (HV) equipment. The aim is to address issues commonly found in traditional systems, including complex operations, high computational demands, significant power consumption, and elevated costs. Various types of discharges were investigated in an HV laboratory environment. Discharge data were collected using a high-frequency current sensor and a microcontroller. Subsequently, this data underwent processing and transformation into feature sets using the phase-resolved partial discharge analysis technique. These features were then converted into grayscale map samples in PNG format. To achieve partial discharge classification, a convolutional neural network (CNN) was trained on these samples. After successful training, the network model was adapted for deployment on a microcontroller, facilitated by the STM32Cube.AI ecosystem, enabling real-time partial discharge recognition. The study also examined storage requirements across different CNN layers and their impact on recognition efficacy. To assess the algorithm’s robustness, recognition accuracy was tested under varying discharge voltages, insulation media thicknesses, and noise levels. The test results demonstrated that the algorithm could be effectively implemented on a microcontroller, achieving a recognition accuracy exceeding 98%.

A Contemporary Review of High Voltage Partial Discharge Detection and Recognition Techniques

This review article provides a summary of the most advanced approaches and advancements in the detection and recognition of high voltage partial discharge (PD). It discusses numerous detecting technologies, such as electrical, acoustic, and optical approaches, as well as their merits and disadvantages. It also discusses current developments in signal processing and pattern recognition algorithms used for PD detection and classification. Lastly, the study covers the challenges and limitations in high voltage PD detection and identification studies, as well as potential future solutions.

Partial Discharge Data Augmentation Based on Improved Wasserstein Generative Adversarial Network With Gradient Penalty

The partial discharge (PD) classification for electric power equipment based on machine learning algorithms often leads to insufficient generalization ability and low recognition accuracy. To solve the problem, this paper develops an improved Wasserstein generative adversarial network with gradient penalty (WGAN-GP) based data augmentation model. The improved WGAN-GP model can generate data samples to supplement the low-data input set in PD source classification. Firstly, an improved WGAN-GP model with conditional generation is trained and various new data samples are generated. Then the new data samples are utilized to expand the raw data set. Finally, the expanded data set is trained to get a new PD classifier. Experimental results demonstrate the proposed model can generate new high-quality data samples more stably. Moreover, the proposed method can suppress the over-fitting risk caused by low-data or imbalanced data distributions and the classification accuracy is effectively improved.

Partial Discharge Localization Techniques: A Review of Recent Progress

Monitoring the partial discharge (PD) activity of power equipment insulation is crucial to ensure uninterrupted power system operation. PD occurrence is highly correlated to weakened insulation strength. If PD occurrences are left unchecked, unexpected insulation breakdowns may occur. The comprehensive PD diagnostic process includes the detection, localization, and classification of PD. Accurate PD source localization is necessary to locate the weakened insulation segment. As a result, rapid and precise PD localization has become the primary focus of PD diagnosis for power equipment insulation. This paper presents a review of different approaches to PD localization, including conventional, machine learning (ML), and deep learning (DL) as a subset of ML approaches. The review focuses on the ML and DL approaches developed in the past five years, which have shown promising results over conventional approaches. Additionally, PD detection using conventional, unconventional, and a PCB antenna designed based on UHF techniques is presented and discussed. Important benchmarks, such as the sensors used, algorithms employed, algorithms compared, and performances, are summarized in detail. Finally, the suitability of different localization techniques for different power equipment applications is discussed based on their strengths and limitations.

Time-frequency Fusion Method via Convolutional Neural Network for Partial Discharge Classification

To improve the accuracy of partial discharge (PD) pattern recognition by jointing time-domain (TD) and frequency-domain (FD) information, a time-frequency (TF) fusion method via convolution neural network (CNN) is proposed in this paper. Firstly, PD signals are represented by PD waveform images and transformed into the envelope of variational mode decomposition-based Hilbert marginal spectrum (VHMS). Secondly, a fusion network, FuNet involving a 2-dimensional CNN (2D-CNN), a 1D-CNN, and a multilayer perceptron (MLP), is established to join TF information. In FuNet, the 2D-CNN inputted by PD waveform images and 1D-CNN adopted the envelope of VHMS of PD signal as its input are all improved by drawing on the complementary strengths of different convolution layers’ features. Then the MLP will fuse the extracted TD and FD features and classify the PD defects.

Classification of Partial Discharges Recorded by the Method Using the Phenomenon of Scintillation

Classification is one of the most common methods of supervised learning, which is divided into a process of data acquisition, data mining, feature analysis, machine learning algorithm selection, model learning and validation, as well as prediction of the result, which was done in the current work. The data that were analyzed concerned ionizing radiation signals generated by partial discharges, recorded by a method using the phenomenon of scintillation. It was decided to check if the data could be classified and if it was possible to determine the defect of an electrical power device. It was possible to find out which classifier (algorithm) worked best for the task, and that the data obtained can be classified, as well as that it is possible to determine the defect. In addition, it was possible to check what effect changing the default values of the classifier’s parameters has on the effectiveness of classification.

Antenna contactless partial discharges detection in covered conductors using ensemble stacking neural networks

High impedance faults caused by vegetation are difficult to detect when covered conductors in medium voltage overhead power lines are used. Long-term contact of XLPE insulation with vegetation causes partial discharges (PDs) which damage the insulation. Although a cheap and easy to install, contactless detection method was developed using an antenna, there is a lack of classification algorithms for this method. Only two custom machine learning algorithms have been tested so far, and both rendered unsatisfactory results for the real application. This work investigates the use of neural network algorithms for this problem and the application of heterogeneous stacking ensembles using neural networks. We used real data collected from a number of detection stations in the Czech Republic. Also, we limited ourselves to supporting edge computing using devices such as Edge TPU. We propose the application of a heterogeneous stacking ensemble neural network to classify PDs obtained by the contactless method. The algorithm we propose is based on a stacking ensemble with a novel combination of base learners, and the Wide and Deep neural network is used as a meta-learner. We compared the results of our algorithm with other algorithms designated for time series classification. Also, an ablation study of the ensemble was conducted, and satisfactory results were obtained using the proposed algorithm. The ensemble outperformed all algorithms tested and is usable on the edge using AI HW accelerator as the ensemble is only feedforward and contains only well-used and known layers. This research improves our understanding of the classification of PDs using the contactless PD detection method and also introduces a stacking ensemble of convolutional neural network and autoencoders for a time series classification for the first time.

Novel data augmentation for improved insulation fault diagnosis under nonideal condition

Insulation fault diagnosis is essential because faulty insulation will lead to the interruption of power delivery. Partial discharge (PD) pattern classification is widely used in insulation diagnosis to identify incipient fault before catastrophic failure occurs. PD classification systems improved significantly due to the emergence of advanced machine learning techniques such as deep learning algorithms. However, classification systems trained in ideal noise-free condition suffer severe performance degradation when tested on-site where the condition is nonideal due to the presence of noise contamination. In this work, a novel data augmentation technique was implemented where simulated noise was added to create augmented data. The augmented data was used in tandem with ideal condition PD data during the training phase of four well known convolutional neural networks (CNN). For a realistic performance evaluation, the classification system was trained with ideal noise-free data and augmented data but tested with data overlapped with actual measured noise representing nonideal condition via K-fold cross-validation. The results showed that the proposed data augmentation technique improved the performance of CNN under nonideal condition by 15.83%–29.05% without compromising its performance under ideal condition.

The detection and classification of partial discharges in power transformer using the acoustic method

Partial discharges are the main causes of degradation of the insulation system in power transformers. Their permanent monitoring is therefore of fundamental importance, in the way that it ensures a better assessment of the level of electrical insulation. Several types of partial discharges are present in power transformers depending on the components of the electric field to which the insulation is subjected. Many on-line methods are used to the continuous measurement of partial discharges, including the acoustic method which doing both measurement of electrical parameters and localization of the sources of partial discharges. This article deals with both recognition and classification of partial discharges by analysis of acoustic signals. The goal is to reproduce various types of partial discharges present in the transformers (surface, internal, and corona discharges) by several structures of electrodes in a prototype transformer filled with insulating oil. The generation of partial discharges is done using an experimental setup consisting of a controlled 220V/50kV HV transformer connected to a prototype of transformer containing an electrode system. The experimental procedure are carried out according to the protocol of the IEC 61270 standard. The measurement of the acoustic signals on the tank is carried out using a conventional broadband sensor from the company ACOUSTIC EMISSION (GI150). Both signal analysis obtained in the frequency domain using the time-frequency response and high-resolution frequency analysis (MUSIC) makes it possible to clearly distinguish the type of discharge produced.

Investigating the Capability of PD-Type Recognition Based on UHF Signals Recorded with Different Antennas Using Supervised Machine Learning

The article presents research on the influence of the type of UHF antenna and the type of machine learning algorithm on the effectiveness of classification of partial discharges (PD) occurring in the insulation system of a power transformer. For this purpose, four antennas specially adapted to be installed in the transformer tank (UHF disk sensor, UHF drain valve sensor, planar inverted F-type antenna, Hilbert curve fractal antenna) and a reference log-periodic antenna were used in laboratory tests. During the research, the main types of PD, typical for oil-paper insulation, were generated, i.e., PD in oil, PD in oil wedge, PD in gas bubbles, surface discharges, and creeping sparks. For the registered UHF PD pulses, nine features in the frequency domain and four features in the wavelet domain were extracted. Then, the PD classification process was carried out with the use of selected methods of supervised machine learning. The study investigated the influence of the number and type of feature on the obtained classification results gained with the following machine-learning methods: decision tree, support vector machine, Bayes method, k-nearest neighbor, linear discriminant, and ensemble machine. As a result of the works carried out, it was found that the highest accuracies are gathered for the feature representing peak frequency using a decision tree, reaching values, depending on the type of antenna, from 89.7% to 100%, with an average of 96.8%. In addition, it was found that the MRMR method reduces the number of features from 13 to 1 while maintaining very high effectiveness. The broadband log-periodic antenna ensured the highest average efficiency (100%) in the PD classification. In the case of the tested antennas adapted to work in an energy transformer tank, the highest defect-recognition efficiency is provided by the UHF disk sensor (99.3%), and the lowest (89.7%) is by the UHF drain valve sensor.

Recognition of Fused Partial Discharge Patterns in High Voltage Insulation Systems: A Hybrid DCNN and SVM Based Approach

The performance assessment of insulation materials within the components of a high voltage power system depends upon effective condition monitoring techniques of the insulation. One of the important phenomena to be considered in the diagnostics and performance assessment of high voltage insulation materials is the Partial Discharge (PD) measurement. The improved Deep learning techniques help in the effective identification and classification of various sources of PD mechanism. In this work, Deep Convolution Neural Network (DCNN) technique is proposed for the fusion of several features that are extracted from the PD signals. Three-dimensional Phase Resolved patterns of PD images can be used for training the DNN. For the classification of PD patterns, linear and non-linear type of multi-class support vector machine has been proposed. The results obtained are compared with linear SVM and non-linear SVMs with the polynomial kernel, Radial Basis Function (RBF) kernel, and Sigmoidal kernel. The SVM type with a higher pattern recognition rate is identified to be effective in PD pattern recognition and classification. The results of the proposed work show that the fusion approach of PD patterns supports applying huge PD data sets as input, generated by multiple faults, for effective PD pattern recognition.

Classification of Partial Discharges Originating From Multilevel PWM Using Machine Learning

A partial discharge (PD) is a small, localized breakdown which often appears in high-voltage insulation systems. PDs can be initiated within material defects such as voids or cracks when exposed to sufficiently high electric field levels. In this research work, we study and propose a machine learning (ML) approach to detect and classify PDs originating from multilevel pulsewidth modulation (PWM) waveforms as utilized in power electronic devices, such as inverters and variable frequency drives. Due to the increased use of power conversion units, new types of higher magnitude PDs have been observed to increase insulation degradation. By creating a cavity placed at different locations within the test object and exposing it to PWM voltage waveforms with two different rise times, a total of 345 660 PD events are recorded and organized into ten different classes. The maximum PD amplitude, duration, the time distance between consecutive PDs, and the area under the PD are used as features for PD classification. A unique way of concatenating a sequence of the extracted features to capture the temporal dependence of consecutive PDs is also presented. It is observed that when creating a sequence of information from consecutive observed PDs, a significant increase in classification accuracy can be obtained. Trained classifiers based on ensemble bagged decision trees (DTs) and long short-term memory (LSTM) architecture resulted in 95.3% and 98.5% average classification accuracy on test data respectively.

Frequency Resolved Partial Discharges Based on Spectral Pulse Counting

A partial discharge (PD) classification methodology based counting PD pulses in the spectral domain is proposed and presented in this paper. The spectral counting data are processed using the proposed PD Spectral Pulse Counting Mapping technique (PD-SPCM), which leads to a Frequency-Resolved Partial Discharges (FRPD) map. The proposed map is then used for PD detection and classification. In this work, corona and slot FRPDs are presented in frequency bands up to 500 MHz, obtained from laboratory measurements performed using two hydro-generator stator bars. The electromagnetic signals from the PDs were captured using a patch antenna designed for this purpose and a spectral analyzer. The corona and slot PDs were chosen because one can be mistakenly classified as the other because they may present similar Phase Resolved PD (PRPD) maps and may occupy shared spectral bands. Furthermore, corona and slot PDs can occur concurrently. The obtained results show that the corona and slot PDs can be properly identified using the developed methodology, even when they occur simultaneously. This is possible because, as it is experimentally demonstrated, corona and slot PDs have appreciable levels of spectral pulse counting in particular bands of the frequency spectrum.

A smart partial discharge classification SOM with optimized statistical transformation feature

Condition-based monitoring (CBM) has been a vital engineering method to assess high voltage (HV) equipment and power cables conditions or health levels. One of the effective CBM methods is partial discharge (PD) measurement or detection. PD event is the phenomenon that always associated with insulation healthiness. PD has been measured and evaluated in this paper to discriminate PD signals from a good signal. A mixed-signal being fed at an AI technique with statistical modified input data to do fast classification (less than five seconds) with nearly zero error. In this paper, an unsupervised neural network is applied for PD classification. The methods combine the self-organizing maps (SOMs) and feature statistical transformation. By the combination of these methods, the ‘range’ normalization method produced the best classification outcomes. This development decided that PD information was effectively correlated and grouped by means of MATLAB’s SOM Toolbox and transformation device to discriminate the normal signal from the PD signal.

Development of Hankel Singular-Hypergraph Feature Extraction Technique for Acoustic Partial Discharge Pattern Classification

Different types of classifiers for acoustic partial discharge (PD) pattern classification have been widely discussed in the literature. The classifier performance mainly depends on the measurement conditions (location and type of the PD, acoustic sensor position and frequency response) as well as extracted features. Recent research posits that features extracted by singular value decomposition (SVD) can exhibit the natural characteristics and energy contained in the signal. Though the technique by itself is not novel, in this paper, SVD is employed for PD classification in a revised way starting from data arrangement in Hankel form, to embedding the hypergraph-based features and finally to extracting the required set of optimal features. The algorithm is tested for various measurement conditions that include the influences of various PD locations and oil temperatures. The robustness of the algorithm is also tested using noisy PD signals. Experimental results show the proposed feature extraction method supremacy.

Noise invariant partial discharge classification based on convolutional neural network

Partial discharge (PD) pattern recognition is essential since it can help to identify the nature of the insulation defect. Numerous machine learning models have been utilized for PD classification applications in the past. However, traditional machine learning models rely on manual feature extraction to obtain training data. They are usually trained using clean PD data measured in the laboratory but are expected to work on-site where some degree of interference or noise is expected. When tested using clean PD data, most machine learning models can easily achieve above 90% accuracy. However, when tested using PD data overlapped with noise, classification accuracy reduces significantly. In this work, the development of a convolutional neural network (CNN)-based PD classification system using transfer learning was proposed. In order to achieve a more practical performance evaluation, a modified 10-fold cross-validation procedure was used where the CNN-based PD classifier was trained using clean PD data but tested using PD data that has been overlapped by noise. The results showed that CNN-based PD classifier was able to achieve up to 16.90% higher classification accuracy under noise contamination compared to traditional machine learning with manual feature extraction. This shows that the proposed method was able to retain higher classification accuracy in the presence of noise.