Partial Discharge Signals Research Articles

Combining multi-level feature extraction algorithm with residual graph convolutional neural network for partial discharge detection

Partial discharge (PD) identification is critical for the insulation diagnosis of cable conductors; however, there is still scope for enhancement in the existing adaptive extraction capabilities and feature utilization for PD signals. In this context, this paper introduces a method that integrates a one-dimensional convolutional neural network (1D-CNN) with a residual graph convolutional neural network (ResGCN) to recognize PD signals. Noise analysis is performed using various combinations of mother wavelets, and Bayesian optimization is employed to mitigate background noise. Key features of PD signals are progressively abstracted through a 1D-CNN-based multilevel automatic feature learning method, while signal timing attributes are maintained to minimize manual intervention. The graph data is constructed using the signal feature matrix and the signal timing feature similarity matrix. This is followed by the development of a ResGCN utilizing a graph attention mechanism to integrate node feature information and the topology of the PD graph data. This approach aims to fully exploit the correlation between local regions of the feature space and the temporal numerical properties of the signals. Additionally, it jointly optimizes feature extraction and model classification to facilitate adaptive diagnosis. The method is validated with extensive real experimental data obtained from medium voltage overhead power lines. It demonstrates exceptional performance and practicality, achieving an accuracy rate of 97.3% and a recognition rate of 96.1% for PD samples, thus offering reliable theoretical support for effective PD detection.

Partial Discharge Data Augmentation and Pattern Recognition Method Based on DAE-GAN

Accurate identification of partial discharge (PD) and its types is essential for assessing the operating conditions of electrical equipment. To enhance PD pattern recognition under imbalanced and limited sample conditions, a method based on a Deep Autoencoder-embedded Generative Adversarial Network (DAE-GAN) is proposed. First, the Deep Autoencoder (DAE) is embedded within the Generative Adversarial Network (GAN) to improve the realism of generated samples. Then, complementary PD data samples are introduced during GAN training to address the issue of limited sample size. Lastly, the model’s discriminator is fine-tuned with augmented and balanced training data to enable PD pattern recognition. The DAE-GAN method is used to augment data and recognize patterns in experimental PD signals. The results demonstrate that, under imbalanced and small sample conditions, DAE-GAN generates more authentic PD samples with improved probability distribution fitting compared to other algorithms, leading to varying levels of enhancement in pattern recognition accuracy.

A novel semi-supervised power transformer defect monitoring technique using unreliable pseudo-labels with highly imbalanced partial discharge signals

A novel semi-supervised power transformer defect monitoring technique using unreliable pseudo-labels with highly imbalanced partial discharge signals

Optimum feature selection for classification of PD signals produced by multiple insulation defects in electric motors

Partial discharges (PD) are initiated in electrical equipment during various points of the equipment’s lifecycle. The intensity of PD defects rises continuously with time, which can lead to insulation degradation and reduced operational life of the electrical equipment. The optimum feature selection of PD signals captured, from different insulation defects, can enhance the classification accuracy of PD defects and facilitate better visualization of PD parameters for electric motor (EM) insulation monitoring and diagnostics. This paper presents a hybrid approach, based on Maximize Relevancy and Minimize Redundancy (mRMR) and random forest (RF), for the optimum feature selection and classification of PD signals in EMs containing multiple defects. For this purpose, four PD defects are developed in the EMs insulation under laboratory conditions, and 800 PD signals are acquired using a conventional IEC-60,270 experimental platform. The severity of these defects is determined and investigated based on PD characteristic parameters. Several features of both PD sweep signals and conventional PD pulses are extracted. Consequently, the mRMR feature selection technique is implemented to select the significant features of the detected PD signals. To establish the plausibility of this technique, several other feature selection algorithms, including RefliefF, Gini Index (GI), and Information Gain (IG), are introduced for the same datasets. The performance of all these feature selection algorithms is validated using three commonly used classification techniques such as RF, support vector machines (SVM), and k-nearest neighbors (k-NN). In summary, the results show that the combination of mRMR and RF proves to be the most effective feature selection algorithm for the classification of insulation defects in EMs, achieving an accuracy of 99.875%. This accuracy is significantly better than other feature selection and classification techniques and indicates its potential for application to other power system components.

Comprehensive monitoring system for high voltage cables based on the ground current signal of the cable metal sheath

Abstract This paper has presented a comprehensive monitoring system for high voltage cables based on the ground current signal of the cable metal sheath. The system can collect and process the power frequency AC current and three high-frequency partial discharge signals in real time synchronously, which realize the synchronous detection of leakage current, relative dielectric loss, harmonic, partial discharge, and improve the integration of the hardware system.

Design of Modified UWB Microstrip Antenna for UHF Partial Discharge Sensor

The development of printable ultrahigh-frequency (UHF) antennas as partial discharge (PD) sensors for high-voltage equipment has been extensively studied. However, achieving ultrawideband (UWB) UHF PD sensors frequently requires larger sizes, unsuitable for certain applications requiring compact sensors for dielectric windows in HV equipment. This research objective is to obtain PD sensors with a wider bandwidth (0.3–3 GHz) and a compact size fitting a less-than-100mm-length gas-insulated switchgear (GIS) dielectric window. A circular patch microstrip antenna (CPMA) was chosen for its small size and potential for UWB performance. This paper discusses the design modification of the CPMA to obtain a wider bandwidth for PD detection in GIS. Simulations and lab-scale experimental verifications were conducted to evaluate the optimized sensor. The modified sensor, with a size of 60 × 73 mm², achieved a bandwidth of 3.08–3.14 GHz, a reflection coefficient of -44 dB, and several resonant frequencies of 0.3–2.3 GHz. This is a seven-time wider bandwidth compared to earlier bowtie antennas while keeping a dimension of less than 100 mm². These properties allow for efficient PD detection in GIS and other insulating media. Experimental results indicate the sensor's capacity to reliably detect and analyze PD signals while responding appropriately to variations in voltage. Doi: 10.28991/ESJ-2024-08-05-03 Full Text: PDF

Impact of Harmonic Voltage on the Partial Discharge Properties of Eco-Friendly Nanofluid

The Insulation is crucial in transformers to maintain electrical isolation between components and windings, ensuring safe and efficient operation. Mineral oil acts as both insulation and coolant in transformers. It is a substance that does not naturally degrade and can persist in the environment for an extended period if accidentally released. Therefore, Sunflower oil is used as a substitute for mineral oil. However, harmonic distortion is a common issue in power systems, leading to increased dielectric loss and partial discharge activity, ultimately causing insulation failure and equipment damage. Hence, it is vital to analyze the partial discharge performance of pure Sunflower oil under varying harmonic AC voltages to ensure the reliable operation of electrical devices. The study examines the partial discharge characteristics of a sunflower oil-based nanofluid under harmonic AC voltages, which is a blend of pure sunflower oil and SiO2 nanoparticles. Two nanofluid mixtures were created, one with a 0.01% mass ratio and the other with 0.05% sunflower oil. Various PD characteristics were evaluated, including Partial Discharge significance, starting voltage, duration of increase, maximum amplitude, Phase-Resolved PD sample, and frequency and time graphs of the Partial Discharge signal. All tests were carried out under IEC regulations. A comparison was made between the outcomes of Nano blended sunflower oil and natural sunflower edible oil. Experiments were conducted on partial discharges in Sunflower oil using specific harmonic superimposed AC Voltages, utilizing the third and fifth harmonics. The findings reveal that: a) PDIV value decreases with increasing harmonic order when using AC voltage with superimposed harmonics, b) PD characteristics are closely related to the voltage waveform, and the addition of Silica to sunflower oil results in reduced PD amplitude, pulse duration, and time duration compared to pure sunflower oil.

Design and optimization for GIS UHF-Overvoltage composite sensor

Gas insulated switchgear (GIS) is widely used in power systems. Various built-in sensors have been designed to monitor internal partial discharges (PDs) and overvoltage, ensuring the safe operation of GIS. This paper presents a composite sensor for simultaneous measurement of PD and voltage, along with an effective optimization method that utilizes a Gaussian process (GP) surrogate model and self-adaptive evolution to optimize the complex composite sensors evaluated through simulations. The optimization method realizes accurate prediction of complex problems during the iteration with fewer training samples and is applied to the designed composite sensor, which requires determining additional parameters compared to the conventional sensors. Through iterations, the effective height of the composite sensor is significantly increased while maintaining a required measuring bandwidth. A test platform is constructed to assess the performance, and results demonstrate that the optimized sensor exhibits satisfying performance measuring both UHF PD signals and overvoltage in GIS. This study provides a design method for UHF-related sensors characterized by structural complexity.

Ultra-High Frequency Sensors for Detecting Partial Discharge in Power Cables

Partial discharge (PD) detection is vital for evaluating the health of power cable insulation, where failure can lead to significant disruptions. This study investigates the use of Ultra-High Frequency (UHF) sensors to detect PD signals, which typically occur in nanoseconds. The sensors operate in the GHz range and minimize interference by selecting frequencies above 400 MHz, where background noise is reduced. Through comparison with other detection methods, UHF sensors demonstrate superior sensitivity in measuring minor discharges, providing a reliable means for on-site condition assessment of high-voltage equipment. The study concludes that UHF sensors offer a practical solution for improving PD detection accuracy, particularly in challenging field conditions.

A distributed PD detection method for high voltage cables based on high precision clock synchronization

Partial discharge (PD) is a crucial cause of high voltage (HV) cable failures. Detecting and locating PD efficiently and effectively is paramount for HV cable maintenance. However, the attenuation of the PD signal after long-distance propagation makes it challenging to identify. This paper presents a distributed PD detection method that significantly reduces the propagation distance of PD signals before the sensor captures them. The technique uses the time-of-arrival to locate PD, whose localization accuracy is directly linked to the clock synchronization (CS) accuracy. CS between different sensors is achieved through three times interactions of the CS signal. The linear frequency modulation (LFM) signal is selected as the CS signal, which can generate a significant gain after pulse compression. It ensures that the CS signal can be effectively recognized after the attenuation of long-distance propagation. The paper also introduces a signal discrimination method based on the short-time Fourier transform that eliminates the influence of the LFM signal on PD identification. The cableless CS method makes it easier to set up before the onsite test. A prototype was implemented and tested. The CS accuracy was measured within 4 ns in the laboratory. An onsite test was carried out to evaluate the performance of the distributed PD detection system. A defect was found at a joint location, which was also confirmed by an additional ultrasonic PD test.

Denoising of partial discharges in switchgear insulation material using hybrid wavelet denoising-optimization-machine learning

Partial discharge (PD) diagnosis is essential for assessing the insulation status of power equipment, but onsite interferences often contaminate PD signals with noise, impacting diagnostic accuracy. This work proposes an adaptive wavelet threshold denoising technique, where the PD signal is first decomposed into wavelet coefficients using discrete wavelet transform (DWT). Traditional threshold selection methods rely on experience and statistical factors, challenging optimal threshold determination. To address this issue, Particle Swarm Optimization (PSO), Energy Valley Optimization (EVO) and Subtraction Average Based Optimization (SABO) are applied to achieve the best adaptive threshold. The proposed method is evaluated against traditional sqtwolog-based threshold methods using root mean square error (RMSE) and the recognition accuracy of classifiers, including Artificial Neural Networks (ANN), Support Vector Machines (SVM), Gradient Boosting Decision Trees (GBDT) and K-Nearest Neighbours (KNN). The results show that the proposed technique can find the best threshold and increase the recognition accuracy by 19% compared to the traditional method, demonstrating its superior performance.

Study on characteristics of health monitoring and critical warning based on partial discharge signals during the growth of electrical trees

Study on characteristics of health monitoring and critical warning based on partial discharge signals during the growth of electrical trees

Detection of failures in HV surge arrester using chaos pattern with deep learning neural network

AbstractAs a protective component of HV equipment, the primary function of a surge arrester is to mitigate the impact of surge voltages. When a surge arrester fails, the equipment it protects becomes vulnerable to damage. This study integrates chaotic systems with Convolutional Neural Networks (CNN) to diagnose faults in HV surge arresters. The Partial Discharge (PD) test was initially performed on six HV surge arrester fault models. The Discrete Wavelet Transform (DWT) was performed for filtering the PD signals. Subsequently, the Chen‐Lee chaotic system converted the filtered PD signals into a dynamic error scatter diagram, creating a feature map of various fault states. This feature map was then used as the input layer to train the CNN model. The results demonstrate that the proposed CNN achieved an accuracy of 97.0%, outperforming AlexNet and traditional methods using Histograms of Oriented Gradients (HOG) combined with Support Vector Machine (SVM), Decision Tree (DT), Backpropagation Neural Network (BPNN), and K‐Nearest Neighbor (KNN). This study also incorporates the LabVIEW graphic control software with a fault diagnosis system for HV surge arresters. The PD data can identify the fault type in real‐time, enhancing power equipment maintenance efficiency.

Typical Partial Discharge Signal Acquisition and Feature Extraction in High Voltage Switchgear

Typical Partial Discharge Signal Acquisition and Feature Extraction in High Voltage Switchgear

Acoustic Sensors for Monitoring and Localizing Partial Discharge Signals in Oil-Immersed Transformers under Array Configuration.

Partial discharge (PD) is one of the major causes of insulation accidents in oil-immersed transformers, generating a large number of signals that represent the health status of the transformer. In particular, acoustic signals can be detected by sensors to locate the source of the partial discharge. However, the array, type, and quantity of sensors play a crucial role in the research on the localization of partial discharge sources within transformers. Hence, this paper proposes a novel sensor array for the specific localization of PD sources using COMSOL Multiphysics software 6.1 to establish a three-dimensional model of the oil-immersed transformer and the different defect types of two-dimensional models. "Electric-force-acoustic" multiphysics field simulations were conducted to model ultrasonic signals of different types of PD by setting up detection points to collect acoustic signals at different types and temperatures instead of physical sensors. Subsequently, simulated waveforms and acoustic spatial distribution maps were acquired in the software. These simulation results were then combined with the time difference of arrival (TDOA) algorithm to solve a system of equations, ultimately yielding the position of the discharge source. Calculated positions were compared with the actual positions using an error iterative algorithm method, with an average spatial error about 1.3 cm, which falls within an acceptable range for fault diagnosis in transformers, validating the accuracy of the proposed method. Therefore, the presented sensor array and computational localization method offer a reliable theoretical basis for fault diagnosis techniques in transformers.

Partial discharge monitoring by improved PGC-arctan algorithm

This paper proposes an improved PGC-Arctan demodulation algorithm for detecting partial discharge signals in transformers. Compared with the traditional PGC-Arctan algorithm, the improved algorithm extracts carrier frequency difference and carrier phase delay by fundamental mixing and using low-pass filtering on the interference signal, and compensates the phase modulation depth by high order frequency, reducing THD and enhancing SNR. The algorithm is theoretically analyzed, simulated and verified through experiments. In the 200 kHz acoustic emission signal detection experiment, the SNR and THD of the improved algorithm are 49.63 dB and 1.34%, respectively, which is 13.59 dB higher than the SNR of the traditional algorithm, and the THD is reduced by 76.61%, and the harmonic distortion is well suppressed. The system is used to detect weak ultrasonic signals generated during partial discharge of transformers and compare them with PZT. The results show that the system has the potential to be used to monitor the partial discharge signal in transformers.

Transformer partial discharge fault diagnosis based on improved adaptive local iterative filtering‐bidirectional long short‐term memory

AbstractInsulation deterioration, which is mainly caused by partial discharge (PD) occurring inside power transformers, is one of the prime reasons to cause transformer faults. Therefore, an effective diagnosis of PD is crucial to ensure the safe and stable operation of transformers. To extract more effective features that characterise transformers PD signals and enhance the recognition accuracy, a novel transformer PD fault diagnosis model based on improved adaptive local iterative filtering (ALIF) and bidirectional long short‐term memory (BILSTM) neural network is proposed. Addressing the issue of predetermined decomposition levels and accuracy in ALIF decomposition, the golden jackal optimisation (GJO) algorithm is introduced to optimise the parameters. The proposed fault diagnostic model extracts dominant PD features employing the improved ALIF and Refined Composite Multi‐Scale Dispersion Entropy and improves the diagnostic accuracy with the optimised BILSTM by introducing GJO. Experimental data evaluates the performance of support vector machine, long short‐term memory and BILSTM. The results verify the effectiveness and superiority of the proposed model.

Research on partial discharge signal denoising based on Kalman-WPT

Abstract Oil-immersed insulating paper will produce partial discharge under high-frequency pulse voltage conditions, so suppressing or reducing the generation of partial discharge signals is an important method for stabilizing the operation of power equipment. However, it has been found in experiments that various interference signals can affect the generation, capture, and measurement of partial discharge signals during various analyses. Therefore, how to capture partial discharge signals effectively and accurately through denoising algorithms has become the focus of this series of experiments. Kalman filtering and wavelet packet transform are the most basic and commonly used methods for denoising at this stage. By analyzing and comparing the basic principles and characteristics of the two methods, the author proposes a Kalman-WPT joint denoising method and conducts simulation analysis using MATLAB 2022b. Multiple controlled experiments have shown that the signal-to-noise ratio of selected discharge signals has increased from 1.3052 for Kalman filtering and 8.9553 for wavelet packet transform to 10.3702, which is sufficient to prove that this method enhances the denoising effect on the original partial discharge signal, reduces waveform distortion, and greatly improves signal detection accuracy.

Classification of Multi-Source Partial Discharge in GIS

Abstract Partial discharge (PD) analysis is an important method to identify electrical equipment faults. However, due to the influence of environmental electromagnetic interference and other equipment discharges in the power supply circuit during the on-site PD test, the collected PD signal is the combination of various PD and noise, which makes it challenging to identify the type or source of on-site partial discharge. A multi-source PD classification method based on the multiple threshold and K-means clustering algorithm is pro-posed to solve this problem. First, wavelet transform with an improved threshold is used to initially denoise the original signal. Then the multiple threshold method is used to filter the residual ripple between every two pulses. Finally, spectrum analysis is carried out for each pulse, and the maximum value of the spectrum is extracted as the feature quantity. Then, the K-means algorithm is used to classify the PD pulses, and the PRPD images of PD pulses in different frequency domains are obtained. The results show that the method adopted in this paper can denoise, extract and classify the PD signals collected on the site, and the PRPD diagram of PD pulse after classification is more obvious, which provides a basis for identifying the PD source on the site.

Wavelet-Based Convolutional Neural Network for Denoising Partial Discharge Signals Extracted via Acoustic Emission Sensors

Wavelet-Based Convolutional Neural Network for Denoising Partial Discharge Signals Extracted via Acoustic Emission Sensors