Random sampling and data processing for PD-pulse height and shapeanalysis

A measuring system for the digital acquisition of partial discharge (PD) pulse signals, based on the last generation of oscilloscopes, has been developed in order to perform both PD pulse shape and PD pulse height analysis. Wide-band, fast sampling rate and individually-triggerable memory blocks are available for the acquisition of PD pulse signals. However, a problem of sampling the population of PD pulses arises from the limited available on-line storage memory. Four sampling techniques are investigated and evaluated in order to record an amount of PD pulses which enables PD stochastic inference for a minimum of the on-line memory use. Some statistical indexes based on the pulse-height and pulse-phase distributions are used to compare performances of the different techniques. It is shown that a technique based on the Poisson law provides the most accurate sampling of PD pulses, while minimizing the memory, particularly in the presence of two simultaneously active PD phenomena. The developed procedure enables an accurate analysis of the shape of a large number of PD signals, but also stochastic processing of height distributions, which is becoming a reference for PD pattern analysis     

Advanced PD inference in on-field measurements. II. Identification of defects in solid insulation systems

Results of investigations performed to evaluate the effectiveness of a new inference method for the diagnosis of solid insulation systems, based on partial discharge (PD) measurements, are reported in this paper. Signal separation, noise recognition, and PD source identification are the main features of the proposed inference method. Techniques for signal separation and automatic noise rejection are reported in the 1st part of this paper, while the problem of the identification of PD phenomena, occurring in defects of insulation systems, is approached in this 2nd part. The identification is based on fuzzy logic and enables the recognition of PD generated from different basic sources, such as internal, surface and corona discharges. It is shown that the different source typologies can be identified by means of fuzzy rules applied to a selection of parameters derived from PD-pulse phase and amplitude distribution analysis, once PD phenomena have been clustered in homogeneous class through a fuzzy algorithm based on PD-pulse shape. The proposed identification procedure is finally applied to rotating machines and cables, affected by insulation defects, showing promising on-field applications.     

Partial discharge pattern recognition by neuro-fuzzy networks in heat-shrinkable joints and terminations of XLPE insulated distribution cables

An identification technique is described, based on a developed adaptive fuzzy logic network, that enables the recognition of partial discharges (PD) generated by different defects in heat-shrinkable joints and terminations of XLPE insulated power distribution cables. It is shown that different sources of PD can be identified on the basis of fuzzy rules applied to a selection of parameters derived from PD-pulse phase and amplitude distributions. A comparison with other PD pattern recognition techniques based on traditional neural networks is presented and discussed.     

气体绝缘组合电器多局部放电源的检测与识别

介绍了基于宽带检测的气体绝缘组合电器(gas insulated switchgear,GIS)多局放源检测与识别技术。该技术由脉冲群快速分类和基于最小二乘支持向量机(least square-support vector machine,LS-SVM)的基于相位分布的局部放电谱图(phase resolved partial discharge,PRPD)识别2个主要模块组成。其中脉冲群快速分类模块由基于脉冲波形的时频参数提取和竞争学习网络无监督聚类实现,它将脉冲群分为若干个由相同波形特征脉冲组成的子脉冲群。PRPD放电谱图识别模块对各子脉冲群对应的PRPD谱图进行放电指纹提取,并使用GIS绝缘缺陷特征数据库训练得到的LS-SVM判别函数对各子脉冲群进行识别。仿真和试验结果均验证了该技术的可行性和实用性。     

Effect of acquisition parameters on equivalent time and equivalent bandwidth algorithms for partial discharge clustering

The acquisition parameters of an unconventional Partial Discharge (PD) measuring system affect the way the PD pulses are recorded and in turn, the results of the data processing. The noise based on the oscilloscope’s vertical resolution is a feature of the sampled signal that is always present when a digital acquisition system is used. In PD unconventional systems, several parameters such as the sampling frequency Fs, the acquisition time T, the number of samples N and the vertical resolution VR of the digitizer result in a wide oscilloscope-based noise variation, that could be quantified by the signal to noise ratio (snr).The classification map is a tool that came available with the development of unconventional systems, that due to their wide bandwidth are able to resolve PD pulses in time and apply clustering techniques for PD source separation. The equivalent time T_eq and equivalent bandwidth W_eq, used to plot the classification map, attempts to extract features of the PD pulses to form clusters so that classification of sources can be achieved. The classification map is based on the ability of separating PD sources by resorting to the parameters T_eq and W_eq, that are believed to show significant differences for distinct PD sources, while they are clearly consistent for the same source.This paper conducts a set of theoretical analysis and laboratory measurements to evaluate the influence of the oscilloscope-based noise on the results of T_eq and W_eq. The results proved that the classification map is heavily influenced by the signal to noise ratio.     

A new methodology for the identification of PD in electrical apparatus: properties and applications

Applications of a new methodology, aimed at the identification of defects occurring in insulation systems of HV apparatus and based on partial discharge (PD) measurements, are presented in this paper. This methodology relies upon the digital acquisition of a large amount of PD pulses and separates the acquired pulses into homogeneous subclasses. Signal processing tools recognize the presence of noise among the different classes. Identification of basic PD source typologies (i.e., internal, corona and surface discharges) is then achieved, resorting to fuzzy algorithms. The proposed procedure is applied to measurements performed on different HV apparatus, such as cables, transformers and rotating machines. The purpose of this paper is to show that the identification process is robust, regarding the measuring circuit, and flexible, so that it can constitute an advanced tool for condition based maintenance, guiding maintenance experts in making decisions on the condition of the insulation system under test.     

Wavelet-based denoising of partial discharge signals buried in excessive noise and interference

Achieving acceptable levels of sensitivity during online and/or onsite partial discharge (PD) measurements still continues to remain a very challenging task, primarily due to strong coupling of external (random, discrete spectral and stochastic pulsive) interferences. Many analog and digital approaches have been proposed for suppressing these interferences, and amongst these, rejection of the pulsive type of interferences is known to be very difficult, if not impossible. The time and frequency characteristics of the pulsive interference being very similar to that of the PD pulses is the main reason posing difficulty in their separation. In this paper, a novel, semi-automatic, and empirical wavelet-based method (using multi-resolution signal analysis) is proposed to recover PD pulses, buried in excessive noise/interference comprising of random, discrete spectral, pulsive, and any combination of these interferences occurring simultaneously and overlapping-in-time with the PD pulses. A critical assessment of the proposed method is carried out, by processing both simulated and practically acquired PD signals. The results obtained are also compared with those from the best digital filter (infinite impulse response, IIR and finite impulse response, FIR) method proposed in literature. From the results it emerges that, the wavelet approach is superior and further, has the unique capability of successfully rejecting all the three kinds of interferences, even when PD signals and one or all interferences occur simultaneously and overlap-in-time.     

基于模糊聚类的XLPE电缆随机脉冲干扰抑制方法研究

为去除XLPE电缆局部放电现场检测中的随机脉冲干扰,采用等效时频法提取脉冲特征,对特征进行自适应模糊聚类,以相位集中度为指标区分局放与随机干扰.对现场数据分析处理的结果表明,该方法能够有效去除信号中的随机脉冲干扰,为评估XLPE电缆的绝缘状态提供了较好的理论和实践依据.     

基于高速数字电路的PD UHF信号纳秒级陡脉冲源研制

超高频法是目前电力设备局部放电在线监测广泛使用的监测方法,超高频电磁信号的定量、定位研究需要稳定可靠的UHF信号模拟发生装置,针对这一现实需求,文章设计了一种基于数字电路的纳秒级脉冲源,用于产生PD UHF模拟信号。文章详细分析了利用数字电路产生陡脉冲的原理,选用简单的高速逻辑器件构建了脉冲发生电路,并采用虚拟仪器产生频率可变的方波信号作为脉冲源的触发信号。用Pspice仿真分析了门电路的脉冲响应特性,并搭建了试验电路进行性能测试。实验结果表明,该脉冲源产生的陡脉冲信号幅值可达2 V,脉冲重复率为50 k Hz-20 MHz,上升陡度为1 ns,脉宽为3 ns,能有效模拟PD UHF信号,并进行相关的局部放电实验。     

Advanced PD inference in on-field measurements. I. Noise rejection

Noise rejection, defect identification and degradation diagnosis in on-field partial discharge measurements are sought by industry, but hardly achieved in practice. This paper presents tools for automatic noise suppression in measurements performed by ultra wide band digitizers, able to record a large quantity of partial discharge (PD) pulse waveforms. Noise and PD signals are split in different classes on the basis of their shape by means of a fuzzy classifier. Tools used for establishing whether a given class of recorded signals is due to external noise or not are proposed. As an example, two kinds of noise are considered: random noise and rectifier-generated noise. A companion paper will explain how the same classification tools can be employed for the purpose of defect identification.     

Rejection of narrow-band noise and repetitive pulses in on-site PDmeasurements [corrected version]

This paper deals with the possibilities of noise suppression in partial discharge (PD) measurements under on-site conditions. After the presentation of previous work, a new concept of designing a digital FIR filter with a fixed order to suppress narrow-band disturbances is presented. The method is compared to a design using adaptive filters. A filter device, which was developed as an extension for a conventional PD detector, is presented. In the second part of the paper, a method for suppressing pulse shaped noise synchronous to the test voltage is discussed. The disturbing pulses are detected automatically in a PD signal by comparing several recorded cycles of the test voltage. There are two possibilities of suppressing the pulses. The first one is an off-line method that allows attenuation degrees up to 30 dB, the second one works on line, and can process continuous PD signals. For the latter case a suppression of 14 dB for worst case pulses is possible with the developed hardware. In order to demonstrate the abilities of the filters, several measurements are shown     

基于虚拟仪器的GIS局部放电声电联合检测系统

笔者设计了一套基于虚拟仪器的GIS局部放电声电联合检测系统,通过不同速率的数据采集模块对GIS局部放电过程中所产生的超高频、高频电流以及超声波信号进行多通道同步联合检测。系统基于LabVIEW虚拟仪器平台开发,可连续捕捉与采集存储局部放电脉冲波形信号、实时计算放电脉冲相位。利用采集的局放信号进行时频分析和模糊聚类实现了多源局放的有效分离和类型判断,同时基于多路局部放电超高频信号与超声波信号本身或两者之间的时差可实现局部放电源的定位。该系统已在多个220 kV及500 kV GIS变电站的局部放电现场检测中得以应用,检测结果验证了系统的性能。     

交流局部放电宽带检测与分析仪的研制

笔者系统地介绍了基于Measurement Studio软件平台研制的一种新型交流局部放电宽带检测与分析仪.其主要包括利用NI5112高速示波采集卡和前置信号处理单元实现的采样率为100 MS/s、模拟宽带为10 kHz～40 MHz的局部放电脉冲宽带检测技术,和基于脉冲波形时频特征的脉冲群实用快速分类技术以及基于PR...     

Development of standardized circuit solutions for control and data acquisition systems of a research complex on the basis of electron-beam technological installations

The structure, contents, and operations of electrical equipment, which are parts of a test bench for investigating the characteristics of technological electron guns, are reviewed. A list is given of critical problems that were solved with the equipment under which the basic requirements for system control and data acquisition were designed. Based on the classification of the control and recorded signals, the characteristics of input–output modules of digital and analog signals are defined. Taking into account the features of the equipment, standardized circuit solutions were developed that provide acceptable accuracy for control and signaling, as well as trouble-free operation of a computerized data acquisition system. The use of modem solutions providing galvanic insulation by optical transmission channel technology for both digital and analog signals is discussed. A method of selection of specialized microchips on the parameters of control and measurement signals is shown. Recommendations on the use of measuring complex for research and design of control systems of electron-beam welding installations are provided.     

基于宽带检测的局放脉冲波形快速特征提取技术

根据电工设备绝缘存在多局放的工况,提出研制基于单个脉冲的局放宽带检测与模式识别在线监测系统.并指出研制该系统需要解决脉冲群快速分类这一关键技术,其由脉冲波形的快速特征提取以及基于波形特征的快速聚类分析组成.本文针对100Ms/S采样率获取的宽带脉冲波形、-时间序列,提出分别使用幅值参数法、等效时频法和时频熵法对局放脉冲波形的特征参数进行提取.在对人工设置的多局放源产生的脉冲群数据进行波形特征提取后,使用模糊聚类对脉冲群波形特征提取结果进行了对比分析.结果表明:三种波形特征提取方法均能在一定程度上很好地分离干扰源而提取出局放数据.GIS母线尖刺缺陷的试验数据处理同样证实了该技术的有效性和可行性.这为研制基于单个脉冲的宽带局放检测与模式识别在线监测系统提供了实用的脉冲波形快速特征提取技术.     

Continuous recording and stochastic analysis of PD

We describe the design and use of a digital partial discharge (PD) data recording system capable of continuous real-time recording of PD pulse trains. The recording system consists of a custom two-channel PD digitizer coupled to a personal computer via a 10-bit parallel interface. The digitizer is under software control with the resulting data being stored in binary files on the computer's hard disk. The stored data subsequently are subjected to stochastic analysis using appropriate computer software. Because all data are retained and the computer provides the desired stochastic analysis of data files, the new system is well suited to investigate non-stationary PD behavior such as encountered in aging studies. By way of illustration, the new system was used to determine the time-varying stochastic behavior of ac-generated PD from point-to-dielectric gaps in air where the insulation material was cast epoxy with aluminum oxide filler. The results confirm and extend previous measurements made with an analog stochastic analyser. With these sample results, we demonstrate how the system allows detailed stochastic analyses not possible with data obtained from existing conventional PD measurement systems     

Application of Maxwell solvers to PD Propagation. III. PD propagation in GIS

A transient electromagnetic field analysis by the method of moments (MoM) has been used to simulate the propagation of high-frequency PD pulses in electric power equipment. The analysis of transient electromagnetic fields described here enables the propagation of high-frequency signals in electric power equipment to be displayed. The motivation for this work is the expectation that such computational methods will provide an improved basis for the application of PD-based diagnostics to power apparatus.     

PD pattern recognition for stator bar models with six kinds ofcharacteristic vectors using BP network

The application of six different kinds of characteristic vectors to recognize PD sources is studied. Four kinds of model bars are used to simulate typical partial discharges in generator stator winding. The PD signals were measured by using a computer aided digital sampling system. The sampling results are processed by six kinds of feature extraction methods and different characteristic vectors are obtained. Then these vectors are used as input patterns for BP network. Recognition results using all six kinds of vectors are reasonable. Further analysis shows that vectors formed by moment features or fractal dimensions possess fairly good abilities of pattern identification and data compression     

Classification of multiple partial discharge sources in dielectric insulation material using Cepstrum analysis–artificial neural network

In high‐voltage equipment insulation, multiple partial discharge (PD) sources may exist at the same time. Therefore, it is important to identify PDs from different PD sources under noisy condition in insulations, with the highest accuracy. Although many studies on classifying different PD types in insulation have been performed, some signal processing methods have not been used in the past for this application. Thus, in this work, Cepstrum analysis on PD signals combined with artificial neural network (ANN) is proposed to classify the PD types from different PD sources simultaneously under noisy condition. Measurement data from different sources of artificial PD signals were recorded from insulation materials. Feature extractions were performed on the recorded signals, including Cepstrum analysis, discrete wavelet transform, discrete Fourier transform, and wavelet packet transform for comparison between the different methods. The features extracted were used to train the ANN. To investigate the classification accuracy under noisy signals, the remaining data were corrupted with artificial noise. The noisy data were classified using the ANN, which had been trained by noise‐free PD signals. It is found that Cepstrum–ANN yields the highest classification accuracy for noisy PD signals than the other methods tested. © 2016 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.