AGA-BP神经网络用于变压器超高频局部放电模式识别

结合自适应遗传算法(AGA)和BP算法各自的优点，本文构造了AGA—BP混合算法作为神经网络的学习算法。分别采用BP、AGA和AGA—BP神经网络对实验室中变压器超高频局部放电自动识别系统检测列的五种放电类型进行了模式识别。实验结果表明，AGA—BP神经网络既解决了BP神经网络对初始权值敏感和容易局部收敛的问题，又提高了AGA神经网络的收敛速度、稳定性和求解质量，具有较高的识别率和较强的推广能力。     

PD source recognition by Weibull processing of pulse heightdistributions

The problem of identification of partial discharge (PD) phenomena occurring in an insulation system is addressed in this paper. PD distributions coming from different sources, such as internal voids, as well as surface and corona discharges, are compared. In particular, the investigation focuses on the Weibull probability function applied to pulse charge-height distribution. It is shown that the different discharge sources can be identified on the basis of the value of the shape parameter of the Weibull distribution and that identification holds even when two PD sources are combined, i.e. are active simultaneously. In this case, the application of the 5-parameter Weibull function permits separation of PD phenomena, and recognition by means of the shape parameter value associated to each phenomenon. Finally, the proposed procedure is applied to practical objects, i.e. insulation systems of rotating machines, with real insulation defects, showing promising on-field application prospective     

变压器超高频局部放电自动识别系统

本文研制了一套变压器超高频局部放电自动识别系统，详细介绍了系统的硬件和软件实现。硬件主要包括超高频天线和频谱分析仪，软件由信号采集，统计谱图生成，统计算子（指纹）计算，用基于遗传算法的神经网络进行模式识别四部分组成，在实验室中对变压器典型局放模型进行实验，结果表明，该系统可以很好地用于提取高频（中心频率在500－1000MHz之间可调）窄带（带宽5MHz)时域放电信号，并且具有强大的数据处理，显示、分析、识别等功能，从而为变压器超高频局部放电检测以及绝缘状态评估提供了丰富的信息。     

基于KELM-VPMCD方法的未知局部放电类型的模式识别

为了解决局部放电类型未知的样本无法被正确识别的问题,提出了一种基于核极限学习机变量预测模型(KELM-VPMCD)的未知局部放电类型的识别方法。通过KELM对已知局部放电类型的训练样本进行训练,然后对各局部放电类型已知的样本建立相应的变量预测模型。利用这些模型对测试样本进行回归预测。根据各样本的预测误差平方和,利用Otsu算法设置误差阈值,通过阈值识别各样本的局部放电类型。识别结果表明,所提方法对于未知的局部放电类型具有较高的正确识别率。     

PD recognition by means of statistical and fractal parameters and aneural network

A novel partial discharge (PD) defect identification method is described. Starting with PD data on different families of specimens, a suitable set of parameters are determined and then used as input variables to a neural network for the purpose of identifying the defects within the insulation. In this procedure the statistical Weibull analysis is performed on PD pulse amplitude histograms to obtain the scale parameter α and the shape parameter β. Thereafter, the two statistical operators (skewness and kurtosis) and two fractal parameters (fractal dimension and lacunarity) are evaluated from the PD phase on the discharge epoch histogram and from the 3 dimensional (pulse amplitude/phase/discharge rate) histogram, respectively. Following the exposition of the basic mathematical concepts regarding the above parameters, experimental results are reported on the recognition capability of the method in defining the defect category in a number of different specimens     

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     

Real-time frequency and harmonic evaluation using artificial neuralnetworks

With increasing harmonic pollution in the power system, real-time monitoring and analysis of harmonic variations have become important. Because of limitations associated with conventional algorithms, particularly under supply-frequency drift and transient situations, a new approach based on nonlinear least-squares parameter estimation has been proposed as an alternative solution for high-accuracy evaluation. However, the computational demand of the algorithm is very high and it is more appropriate to use Hopfield type feedback neural networks for real-time harmonic evaluation. The proposed neural network implementation determines simultaneously the supply-frequency variation, the fundamental-amplitude/phase variation as well as the harmonics-amplitude/phase variation. The distinctive feature is that the supply-frequency variation is handled separately from the amplitude/phase variations, thus ensuring high computational speed and high convergence rate. Examples by computer simulation are used to demonstrate the effectiveness of the implementation. A set of data taken on site was used as a real application of the system     

Application of extension theory to PD pattern recognition in high-voltage current transformers

This paper presents a novel partial-discharge (PD) recognition method based on the extension theory for high-voltage cast-resin current transformers (CRCTs). First, a commercial PD detector is used to measure the three-dimensional (3-D) PD patterns of the high-voltage CRCTs, then three data preprocessing schemes that extract relevant features from the raw 3-D-PD patterns are presented for the proposed PD recognition method. Second, the matter-element models of the PD defect types are built according to PD patterns derived from practical experimental results. Then, the PD defect in a CRCT can be directly identified by degrees of correlation between the tested pattern and the matter-element models which have been built up. To demonstrate the effectiveness of the proposed method, comparative studies using a multilayer neural network and k-means algorithm are conducted on 150 sets of field-test PD patterns of 23-kV CRCTs with rather encouraging results.     

Building knowledge-based systems with an assembling technique

In the field of medical knowledge engineering, it is a common expectation that the number of diseases contained within a given system should constantly increase. The authors' effort to develop an enormous knowledge-based system (the Enormous Electronic-Brain Erudite, EBME) has extended for more than 10 years. The reason for such a long time-frame is that EBME has a huge knowledge base that consists of 1,001 diagnostic entities. It is not only time consuming and tedious but also very error-prone to build this type of database manually. To overcome this problem of time and accuracy, we put forward an assembly technique for knowledge-based systems, which we describe in this article. Our research direction is to develop a methodology to build an enormous knowledge-based system. The goals of this study are: (1) to enhance the efficiency of knowledge engineering by automating the knowledge engineering processes (2) to avoid repeated labor as much as possible (3) in an enormous knowledge-based system, to assemble different subsystems that not only meet the different needs of different users, but also are useful to avoid the occurrence of “combination explosion” (4) to advance the research of medical information processing standardization     

An effective preprocessing method for subspace face recognition using genetic‐based clustering in ideal and noisy conditions

Face recognition using principal component analysis (PCA) and linear discriminant analysis (LDA) suffer from the loss of accuracy when the number of classes becomes large. This paper presents an effective genetic‐based clustering algorithm (GCA) to preprocess a facial database into a two‐layer database. Then, face recognition is done to minimize the similarity criterion in a specific cluster as in the traditional PCA‐ and LDA‐based face recognition algorithms. Different from K‐means clustering, the proposed GCA introduces a novel distance and a balance factor. The distance is defined to measure the similarity effectively between a class and the centroid of each cluster, and the balance factor is designed to achieve balanced clustering results. Experimental results on the Yale‐B database in ideal and noisy conditions indicate that the proposed preprocessing method improves the recognition accuracy of the subspace recognition algorithms compared with K‐means clustering. The proposed preprocessing method is also applicable to other recognition algorithms. © 2012 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.     

电缆局部放电测试中本体放电和干扰的识别及放电点的准确定位

根据交联电缆放电模型,阐述了电磁波行波在电缆传输中的特性参数和识别电缆放电的图像特征,并介绍干扰信号的识别和排除方法。同时根据电缆的放电传播方式,介绍了电缆发生放电时放电点的定位方法,着重介绍了端头放电的波形识别和定位方法。实践证明该办法是有效的。     

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。     

Change in PD pattern with aging

The changes in the PD /spl Phi/-q-n pattern and the PD current shape due to the degradation of insulating materials have been investigated with a fast digital oscilloscope and a PD measurement system. In order to verify the effect of gas change, the volume of gas in a void was observed with the /spl Phi/-q-n pattern. It showed that the change in the maximum PD magnitude is related to the change in gas volume according to ageing time. The PD /spl Phi/-q-n pattern could be classified into the five stages. At stage 3 for example, which had a "rabbit-like pattern", the current shape of the first PD in each half-cycle had a large magnitude of the order of 100 mA and a narrow width. At stage 5, which had a "turtle-like pattern", the current shape had a slow rise-time and a slow decay-time. Such changes in PD characteristics will be applicable to the diagnosis of insulation degradation.     

超高压电缆局放在线监测局放信号提取技术研究及应用

局部放电检测已成为电缆绝缘在线检测技术的重要内容之一,由于局部放电信号微弱,且易受现场噪声干扰,因此如何准确识别并提取局放信号成为在线监测系统研究的重点和难点。本文通过运用联合抗干扰技术对局放信号进行处理,有效消除干扰并保留原始信号的必要形状特征;采用多相态谱图聚类分析法,从多相态谱图中提取放电特性,为局放在线监测提供判断依据和有效分析手段。经某水电厂330 k V电缆线路局部放电在线监测系统现场应用证明该技术具有较好的实用性和推广性。     

心音、心电采集系统设计与信号预处理

介绍了一种基于LabVIEW的心音、心电实时采集与存储系统,以及对存储后的信号用MATLAB进行信号预处理的信号处理解决方案.论文首先介绍了心音、心电采集电路的硬件设计;其次介绍了用LabVIEW和NI公司的USB-6009的数据采集卡完成信号采集、存储以及实时显示心音、心电信号的软件设计;最后介绍了对存储后的心音、心电信号用MATLAB进行预处理,包括ECG信号的小波去噪,心音信号的包络提取等.实现了信号的实时采集、存储以及预处理,为心音、心电数据库的构建和后续的信号处理奠定了基础.     

一起油浸式电流互感器局部放电缺陷的诊断和分析

本文中作者针对一起油浸式电流互感器油中溶解气体和介质损耗因数超注意值的缺陷,开展了局部放电、不同电压下高压介损和油品检测等诊断试验,结合例行试验结果,推测出了故障原因。     

变压器局部放电在线监测中局放电脉冲信号检测研究

介绍了变压器局放电脉冲信号在线监测中采用的抑制干扰算法.     

PD pulse sequence analysis and its relevance for on-site PD defect identification and evaluation

Digital partial discharge (PD) diagnosis has become a state of the art, but computer aided procedures based on pattern recognition principles are negatively effected by on-site disturbances. The benefits and limits of expert diagnosis compared to machine intelligent systems are discussed. It is pointed out that a sufficient noise resistivity and test voltage independency of the diagnosis concept is a must for a successful on-site PD defect identification and evaluation. Experimental results are presented, taken into account especially novel evaluation procedures, which are more independent from the applied PD sensors. It is shown that PD defect identification on site at operating voltages and respectively disturbed data is still a challenge. A method of resolution with an intelligent noise resistant diagnosis concept is discussed. A hierarchical and redundant approach is presented, which improves the on site PD source identification as well as on-site risk assessment.     

超声波局放监测在GIS耐压局放试验中的应用

介绍了基于超声波法局放监测系统的原理,同时,阐述局放脉冲在理论分析上可以用单指数震荡衰减模型（SEAOW）和双指数震荡衰减模型（DEAOW）来进行模拟。阐述了基于超声波法局放监测系统定位出500 kV 气体绝缘金属封闭开关设备（GIS）在交流耐压局放试验中的放电击穿位置,从而验证了基于超声波法局放监测系统在耐压局放过程中寻找放电点是行之有效的,可为及早地发现GIS设备绝缘缺陷提供了有效检测方法和依据。     

Fundamentals in computer aided PD processing, PD patternrecognition and automated diagnosis in GIS

The partial discharge (PD) measurement is a sensitive method for testing and monitoring the insulation condition of HV apparatus. Digital PD measurements have become state of the art. Computer-based PD diagnostic systems nowadays can be used to increase the reliability of modern gas-insulated system (GIS). Also on-line monitoring of HV equipment is becoming increasingly important to enable prevention of a sudden breakdown of PD affected components. On-line monitoring must be done on-site, but an important problem of on-site PD measurement is the superposition of PD data and interference. Furthermore, reliable recognition of the type or nature of the insulation defect is necessary for any risk assessment. Traditionally, this is achieved with reference to expert knowledge. An alternative approach is presented in this paper, which treats the fundamentals of automated computer-based PD diagnosis, taking into account that successful noise suppression is essential for a successful on-site diagnosis