Auto‐detection of partial discharges in power cables by discrete wavelet transform

One of the serious problems that can occur in power XLPE cables is destruction of insulator. The best and conventional way to prevent this is ascertaining partial corona discharges occurring at small voids in organic insulators. However, there are some difficulties in detecting those partial discharges due to the existence of external noise in detected data, whose patterns are hardly identified at a glance. For this reason, there have been a number of researches into detecting partial discharges by employing a neural network (NN) system, which is widely known as a system for pattern recognition. We have been developing an NN system for auto‐detection of partial discharges, and have input numerical data of the waveform itself and obtained appropriate performance. In this paper, we employed the discrete wavelet transform (DWT) to acquire more detailed transformed data in order to use them in the NN system. Employing the DWT, we were able to express the waveform data in time–frequency space, and achieved effective detection of partial discharges by the NN system. We present herein the results using DWT analysis for partial discharges and noise signals which we obtained. Moreover, we present results out of the NN system which dealt with those transformed data. © 2005 Wiley Periodicals, Inc. Electr Eng Jpn, 152(1): 24–30, 2005; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/eej.10315     

PD source identification with novel discharge parameters usingcounterpropagation neural networks

Computer aided partial discharge (PD) source identification using different multidimensional discharge patterns is widely regarded as an important tool for insulation diagnosis. In this paper, a neural network (NN) approach to PD pattern classification is presented. The approach is based on applying variants of the counterpropagation NN architecture to the classification of PD patterns. These patterns are derived from physically related discharge parameters, different from those commonly used. It is shown that considerable improvements of the classification quality can be obtained when an extended counterpropagation network with a dynamically changing network topology is applied to patterns that employ the voltage difference between consecutive pulses instead of the phase of occurrence as the main discharge parameter. Furthermore, using a particular parameter vector that takes the correlation between consecutive discharges into account also allows to solve the rejection problem with this type of NN     

静电放电电磁脉冲耦合的非线性优化建模

针对现有机理建模算法普遍存在计算电磁脉冲响应过程过于复杂的问题,为能够给电子设备静电放电电磁脉冲响应计算提供一种简便有效的能量耦合建模方法,设计了脉冲场强测试仪的静电放电辐射实验。用NARX神经网络代替传统NARX网络,依靠遗传算法对网络的初始权值、阈值进行优化,以3.5 kV静电放电实验数据作为建模数据对系统进行非线性辨识,并对4.5 kV静电放电电磁脉冲响应进行预测。建模结果表明,两种模型均能准确预测响应波形,但优化后的NARX神经网络模型精度更高。该建模方法计算过程简单。该方法同样适用于其他电磁脉冲响应建模。     

智能变电站电缆局部放电在线监测系统的研究

介绍了一种遵从于IEC 61850通信规约的智能变电站电缆局部放电在线监测系统。首先分析了电缆局部放电的检测原理及检测方法;然后设计了基于光纤同步触发和组网的监测硬件系统,以及基于用户数据报协议（user datagram protocol,UDP）和生产者/消费者模式的高速采集,并经过专家系统判别分析后能将监测结果通过IEC 61850规约实现共享的软件系统。最后用2种不同电极结构的局部放电类型,通过模拟实验对监测系统进行了检验。结果显示设备运行良好,监控平台能够准确地监视局部放电信号,能够识别出放电类型,并能通过IEC 61850通信规约将监测结果共享。     

电气设备长间歇性局部放电检测系统研究

为降低外界干扰在带电检测与在线监测中对电气设备内部发生的长间歇性、潜伏性局部放电的误判,避免因绝缘击穿而导致的设备故障停电.文中对长间歇性局部放电的放电特性进行研究,在现有带电检测与在线监测基础上,设计了一套便携式长间歇性局部放电监测系统.该系统不仅可以完成设备内部不同频带局部放电信号的采集与放电缺陷的初步定位,而且在...     

灭弧室内真空度与局部放电关系的研究

为研究真空断路器灭弧室内真空度与局部放电状态的关系,设计了一套模拟断路器运行的试验平台,采用特高频方法测量了灭弧室内不同真空度下的局部放电信号,并利用统计参数法统计出不同压强下的局放信号特征参数。试验结果表明随着灭弧室内真空度在0.5～300Pa之间变化,灭弧室局部放电量和次数有明显区别,压强在15～20Pa之间局部放电强度最大、次数最多,同时易发生间隙击穿。在此基础上通过对局部放电数据处理分析,得到不同压强下局部放电PRPD谱图,作为间接诊断断路器灭弧室真空度的依据。     

用于局部放电图象识别的统计特征研究

结合图象识别技术，提出一种采用局部放电灰度图象的统计特征区分局部放电类型的方法。局部放电灰度图象统计特征由图象的矩特征和相关统计特征构成；其中矩特征描述局部放电灰度图象基本灰度分布状态，相关统计特征描述局部放电正、负工频半波图象之间的相关程度。设计出模拟变压器内部放电与外部放电的五种放电模型，并通过试验获得大量放电样本数据，采用局部放电灰度图象统计特征和人工神经网络分类器，对于五种放电样本获得了较高的识别率，表明该方法具有良好的应用效果。     

基于VHF的交联电缆附件局放检测的研究

为在线检测XLPE电缆附件局放,基于电磁耦合法设计了VHF宽频带钳型电流传感器,该传感器检测频3~60 MHz,灵活方便、操作安全,易于安装,还建立了XLPE电缆附件和几种典型的电缆局放模型(气泡放电,毛刺放电等)。试验表明,不同局放类型统计特征谱图不同;VHF检测技术与传统检测方法的检测结果一致性较好。     

自适应遗传算法在变压器超高频局放模式识别中的应用

采用自适应遗传算法(AGA)作为神经网络的学习算法,对实验室中变压器超高频局部放电自动识别系统检测到的5种放电类型进行了模式识别。实验结果表明,AGA神经网络解决了BP神经网络对初始权值敏感、收敛速度慢和容易局部收敛的问题,具有较高的识别率和较强的推广能力,可以很好的应用于变压器超高频局部放电的模式识别中。     

自适应遗传算法在变压器局部放电高频模式识别的应用

采用自适应遗传算法(AGA)作为神经网络的学习算法,对实验室中变压器局部放电超高频自动识别系统检测到的5种放电类型进行了模式识别.实验结果表明,AGA神经网络解决了BP神经网络对初始权值敏感、收敛速度慢和容易局部收敛的问题,具有较高的识别率和较强的推广能力,可以很好地应用于变压器局部放电的超高频模式识别中.     

自适应遗传算法在变压器局部放电超高频模式识别中的应用

采用自适应遗传算法(AGA)作为神经网络的学习算法，对实验室中变压器局部放电超高频自动识别系统检测到的5种放电类型进行了模式识别。实验结果表明，AGA神经网络解决了BP神经网络对初始权值敏感、收敛速度慢和容易局部收敛的问题，具有较高的识别率和较强的推广能力，可以很好地应用于变压器局部放电的超高频模式识别中。     

变压器内单一局放PRPD谱图统计参量的主成分分析

设计了4类由变压器油纸绝缘缺陷引起的"单一局放模式"模型:气隙放电模型、针板放电模型、介质表面金属杂质放电模型、油隙放电模型。通过标准化试验方法得到了不同模型局部放电相位分布模式(phase resolved partial discharge,PRPD模式)的二维谱图,并对谱图进行分析得到25个局部放电统计参量。采用主成分分析对统计参量的有效性进行了分析,得到了25个统计参量对不同类型放电信息表达的新特征参量组,以及对不同放电类型识别的贡献率。     

信息融合在变压器油纸绝缘局部放电识别中的应用

局部放电会引起变压器绝缘的老化和破坏,而变压器局部放电特性的研究能够很好反应变压器潜伏性缺陷,对其安全可靠运行具有重要意义,因而设计制作了模拟变压器沿面放电、气隙放电和电晕放电的3种缺陷模型,采用升压法进行相应的放电试验,通过分析油中溶解气体在局部放电发展过程中的变化规律,寻找出油中产生气体与不同局部放电的对应关系。引入局部放电的最大放电量相位特征谱图Hqmax(φ)和放电次数相位特征谱图Hn(φ),并提取基于谱图的统计特征参量,构建反向传播(back propagation,BP)神经网络和径向基函数(radical ba-sis function,RBF)神经网络对局部放电的放电类型进行初级识别,在此基础上,应用信息融合方法将初级识别结果融合油中产气特征以综合识别局部放电类型。实验结果表明,同一种溶解气体在不同放电模型中发展变化趋势是不一样的,根据统计特征参量和油中溶解气体变化规律,应用信息融合方法对变压器局部放电模式具有足够的识别能力。     

空气中电晕放电与绝缘液体中流注放电的测量和分析

Insulation is one of the most important parts in a high voltage equipment.There are gaseous,liquid and solid insulations which are commonly used.In a high voltage transformer for example the insulating materials are all used.During operation of a high voltage equipment high electric stress may occur.Under extreme condition failure of the insulation may take place.Excessive electric field in air may cause corona discharges while in liquid insulation discharges may take place in the form of streamer.This paper reports experimental results on the corona and streamer discharges in air and silicone oil.The discharges were artificially generated around a needle tip in a needle-plane electrode system with gap length of 4 mm under sinusoidal and triangular voltages.The needle was made of steel with tip radius of 3 μm and curvature angle of 30°.The needle was made by Ogura Jewelry.The discharge pulses were measured using personal-computer based partial discharge(PD)measurement system with sensitivity of better than 0.5 pC.The system is able to measure discharge in time sequential.Phase-resolved analysis of the discharges was done to interpret the physical processes behind the discharges.The experimental results showed that corona discharges took place at negative half cycles.The discharges were concentrated around 270° of phase angle of applied voltage.The discharge magnitude and discharge number of corona clearly dependent on the instantaneous of applied voltage.These were strongly supported by the application of triangular voltage.Streamer discharges occurred at both positive and negative half cycles.The discharges pulses concentrated around the peak of applied voltage at phase angle of 90° and 270°.Experimental results under sinusoidal and triangular voltages revealed that streamer discharge magnitude as well as probability of occurrence was strongly dependent on the instantaneous applied voltage.     

UHF detection of leader discharges in SF6

The ultra-high frequency (UHF) technique of partial discharge detection has been used to study the partial discharges produced by electrode protrusions and surface contamination in a gas insulated substation (GIS). The paper describes the stages of discharge development from inception to breakdown. The work has shown that the UHF technique may be used to identify distinct phases of discharge activity corresponding to discharge inception and a prebreakdown condition. The prebreakdown activity is detected as an increase in the UHF signal magnitude which has been attributed to the formation of leader-type discharges. These discharges form an essential part of the breakdown process, and their detection can be used to identify when the system is close to breakdown. The paper describes a filtering technique that may be employed to optimize the sensitivity of UHF leader detection     

Application of artificial neural networks for optimization ofelectrode contour

In this paper artificial neural networks (NN) with supervised learning are proposed for HV electrode optimization. To demonstrate the effectiveness of artificial NN in electric field problems, a simple cylindrical electrode system is designed first where the stresses can be computed analytically. It is found that once trained, the NN can give results with mean absolute error of ~1% when compared with analytically obtained results. In the next section of the paper, a multilayer feedforward NN with a back-propagation algorithm is designed for electrode contour optimization. The NN is first trained with the results of electric field computations for some predetermined contours of an axisymmetric electrode arrangement. Then the trained NN is used to give an optimized electrode contour in such a way that a desired stress distribution is obtained on the electrode surface. The results from the present study show that the trained NN can give optimized electrode contours to get a desired stress distribution on the electrode surface very efficiently and accurately     

基于人工神经网络的变压器绝缘模型放电模式识别的研究

分析了变压器绝缘的主要放电形式,设计了模拟变压器放电的7种试验模型和3种模拟空气中放电干扰的模型,进行了不情况下模型的放电试验。使用数字化测量装置,取得了各种模型放电的放电量－相位信息。采用三维谱图提取放电指纹特征,并用人工神经网络ANN来识别不同的放电类型。研究结果表明,人工神经网络对油纸变压器绝缘放电有足够的识别能力。     

基于遗传算法的神经网络在发电机定子超高频局部放电模式识别中的应用

介绍了发电机定子局部放电高频在线监测系统的结构特点及对局部放电特征的提取,应用BP算法、自适应遗传算法AGA和AGA—BP神经网络对发电机定子超高频局部放电的三种放电类型进行了模式识别。结合AGA和BP算法各自的优点,构造了AGA—BP混合算法作为神经网络的学习算法。实验结果表明,AGA—BP神经网络解决了BP神经网络对初始权值敏感和容易局部收敛的问题,提高了AGA神经网络的收敛速度、稳定性和求解质量。     

Understanding surface partial discharges in HV coils and the role of semi-conductive protection

An electrical insulating system in high voltage rotating machines is subjected to several stresses that cause changes of material properties during its lifetime. This phenomenon reduces progressively ability of the insulation to withstand the service stresses. Internal partial discharges can be located inside the high voltage insulation and usually they are dangerous for life-cycle duration. Surface discharges are generated at the solid/air interface in the gaps of machine stator slots or at the slot exits. Surface discharges can cause more intensive deterioration of the insulation, accelerate aging processes and definitively take aim toward premature failure of the machine. The application of conductive or semi-conductive protection layer provides certain protection for solid insulation from these discharges. Semi-conductive corona protection tapes are used on the coils at the slot exit region to electrical field control and prevent surface discharges. Laboratory measurements of the surface partial discharge quantities in specimens of high voltage coils with mica insulation were performed. The aim of the paper is to describe theoretically the development of surface discharges in rotating machines and to present surface partial discharge development in HV motors when semi-conductive layer protection is missing. The results can be found useful to understand better the surface discharge development and importance of high-quality semi-conductive protection technology.     

Modeling PD inception voltage of epoxy resin post insulators usingan adaptive neural network

One of the parameters used to characterize the partial discharge (PD) behavior is its inception voltage. The partial discharge inception voltage (PDIV) should always be higher than the operating voltage to ensure that PD does not occur at or near operating voltage. Under these circumstances, other PD parameters such as apparent charge, energy dissipation need not be considered. This paper deals with modeling of PDIV of epoxy-resin post insulators using a neural network (NN). The PDIV is obtained experimentally for various shapes and sizes of post insulators, with a working voltage in the range of 3.3 to 33 kV. The electrode spacing d and creepage length l are the key parameters employed for the present modeling. An adaptively trained multilayer NN is employed for the modeling. Detailed studies are carried out to optimize the NN parameters for minimum error. The model results obtained closely follow the experimental data indicating the effectiveness of NN as an efficient tool in estimation of PDIV of epoxy-resin postinsulators