基于EMD航空发电机定子绕组故障检测方法

航空发电机工作的可靠性关系到航空运输的安全性,针对航空发电机定子绕组内部故障诊断问题,提出了基于经验模式分解算法的检测方法。在介绍经验模式分解算法基本原理和同步发电机内部回路电路仿真方法的基础上,对航空发电机的定子电流信号进行了分析,提取了瞬时频率、幅值等瞬时故障特征,确定了故障发生时刻,并给出了诊断结果,仿真结果表明本文方法有效。     

A fault diagnosis method for three-phase rectifiers

In this article, we proposed a novel method based on principal component analysis and support vector machines was presented for fault diagnosis of three-phase rectifiers, in which the principal component analysis of fault signal is used to extract the features corresponding to various fault, then fault types are identified through the pattern recognition classifier based on support vector machines. The simulation result of fault diagnosis of a thyristor in a three-phase full-bridge controlled rectifier shows that the method can make an accurate identification of fault types as well as the location of the fault elements for power electronics circuits, and it has an excellent performance for noise robustness and calculation complexity. Therefore, it is quite practically valuable in the solution to the fault problems for power electronics rectifiers.     

A novel protection scheme for synchronous generator stator windings based on SVM

This paper proposes a novel scheme for detecting and classifying faults in stator windings of a synchronous generator (SG). The proposed scheme employs a new method for fault detection and classification based on Support Vector Machine (SVM). Two SVM classifiers are proposed. SVM1 is used to identify the fault occurrence in the system and SVM2 is used to determine whether the fault, if any, is internal or external. In this method, the detection and classification of faults are not affected by the fault type and location, pre-fault power, fault resistance or fault inception time. The proposed method increases the ability of detecting the ground faults near the neutral terminal of the stator windings for generators with high impedance grounding neutral point. The proposed scheme is compared with ANN-based method and gives faster response and better reliability for fault classification.     

三峡水力发电厂700MW水轮发电机故障诊断专家系统

三峡水力发电厂在我国电力系统中占有十分重要的地位。为提高三峡水轮发电机的安全运行水平、逐步实现优化检修,研制、开发了三峡水力发电厂发电机定子水冷状态监测及故障诊断专家系统。该系统是一个集在线实时、在线非实时和离线数据的采集与分析、状态监测、故障诊断、故障处理、故障统计分析于一体的多任务、多客户端、多信息处理、B-S结构的发电机定子水冷状态监测及故障诊断专家系统,已于2004年5月投入运行。实践表明:该系统的应用可显著地提高发电机的安全性和可靠性并为状态/优化检修提供科学的决策依据,降低发电成本,产生巨大的经济效益和社会效益。     

EMD小波变换水轮机故障奇异数据还原研究

水轮机组状态监测数据由于受到传感器、测量条件、环境等因素影响会造成数据失真,产生众多奇异点,对正确分析机组运行状态十分不利。本文提出通过经验模态分解与小波变换相结合的方法来分析水轮机故障信号的奇异性,该方法将原始信号经验模态分解后,利用小波变换检测出信号中的奇异点,并将剔除奇异点的信号重构,通过重构信号对机组进行分析。实例仿真表明,与直接对原信号进行小波分析相比,该方法提取的奇异性特征明显并能准确重构,在采用通用传感器信号准确描述机组状态和正确认识机组故障上有较好的应用价值。     

刘家峡水电厂1#机故障诊断分析

本文对水力发电机组由顶盖系统和机组推力系统耦合作用引起的故障进行了试验、理论分析和处理,这种故障是一种很特殊的故障,对机组的破坏是致命的。通过减少275t以上的水推力,使机组能够运行,挽回上千万元的经济损失。本文也提出了把动态故障分析和静态故障分析相结合的概念,对现场技术人员有一定的启发。实践证明,本文采取的理论分析和故障处理措施是非常有效的,它对于水力发电机组相似的故障处理和故障诊断分析具有指导和借鉴的作用。     

基于AM激励下谱分解的非线性电路故障特征提取

为解决非线性电路故障特征重叠、交叉造成的故障隔离度低的问题,提出一种基于调幅激励下电路响应谱分解的非线性电路故障特征提取方法．利用慢调幅激励下非线性电路Volterra响应谱的可分离性,通过带通滤波器分解出与Volterra级数对应的响应分量;根据线性电路的频响叠加性原理,分析了Volterra响应分量的故障全息性;再由Volterra响应分量提取非线性电路故障特征。最后,通过仿真实例说明该方法有效性。     

A parallel heuristic reduction based approach for distribution network fault diagnosis

For the large volume of data in power systems, existing approaches to rough sets reduction either run on a single machine, or are paralleledly achieved in an approximate manner. They also seldom consider rough sets based value reduction. These problems restrict them in applications of power systems. In order to accelerate attribute reduction and value reduction, and improve the efficiency of fault diagnosis analysis in power systems, in this paper, we present a parallel heuristic approach to exact attribute reduction and value reduction for fault diagnosis in distribution network. We obtain the diagnosis rules and diagnose and locate the faults in the distribution network. Our parallel algorithms have been implemented on the MapReduce platform. The experimental results show that our method can effectively improve reduction process and improve the accuracy of reduction results in dealing with a large volume of data sets.     

基于信息融合和神经网络的变压器内部故障快速诊断方法

变压器内部故障诊断通常需要利用油中溶解气体进行分析,但这些信息提取、检测分析过程繁琐,实时性较差。因此文章中提出了一种仅需要电气量的变压器内部故障快速诊断方法,采用小波包分析提取短路电流和差动电流的频域故障特征,采用最大值体现零序电流的故障特征,采用信息融合技术将所得到的所有故障特征进行融合,并利用BP神经网络算法对变压器内部电气故障类型进行诊断。在MATLAB/Simulink平台建立来仿真模型并进行了算例分析,结果表明文章中所提出的变压器内部电气故障诊断方法具有高准确性和高可靠性的优点。     

以槽电势为分析单元的大型发电机定子接地故障定位方法

已有的大型发电机定子绕组单相接地故障定位方法一般以线圈匝电势为分析单元,对于短距绕组的机组存在理论误差。为此,提出一种以槽电势为分析单元的大型发电机定子单相接地故障定位方法。首先,基于绕组连接顺序,以槽电势为单元建立定子绕组电势分布的解析表达式。然后,针对发电机是否配备注入式定子接地保护采取不同的故障定位方案。若配备注入式定子接地保护,则基于注入式设备的过渡电阻测量值构造故障评价指标。若未配备注入式定子接地保护,则引入3次谐波测量值构造不含过渡电阻的故障评价指标。最后,在故障相人为设置多个虚拟参考点,结合绕组电势分布计算各虚拟参考点的故障评价指标,将计算值最小的虚拟参考点视为故障位置,进而确定故障所在槽编号。利用在PSCAD中搭建的准分布参数仿真模型验证了所提方法的有效性。     

基于EEMD-RVM风力发电机故障诊断方法研究

针对风力发电机早期故障表征不明显、能采集有效数据量较少、诊断结果精度较低等问题,文中提出一种运用综合经验模态分解(Ensemble Empirical Mode Decomposition,EEMD)结合相关向量机的方法对风力发电机多类故障进行早期诊断。首先,利用EEMD结合灰色关联度的方法对风机各类故障的振动信号进行预处理,提取最优故障特征;再通过相关向量机(Relevance Vector Machine,RVM)对提取的故障特征训练,并建立相应的故障诊断模型进行诊断。在实例中将文中所提方法 EEMD-RVM与小波包分解(Wavelet Packet Decomposition,WPD)结合RVM以及EEMD结合最小二乘支持向量机LS-SVM(Least Square Support Vector Machine,LS-SVM)方法的诊断结果作对比,结果表明,EEMD-RVM方法具有可行性,且具有耗时短、精度高等优点。     

A novel fault location method for distribution networks with distributed generations based on the time matrix of traveling-waves

To improve location speed, accuracy and reliability, this paper proposes a fault location method for distribution networks based on the time matrix of fault traveling waves. First, an inherent time matrix is established according to the normalized topology of the target distribution network, and a post-fault time matrix is obtained by extracting the head data of initial waves from traveling wave detection devices. A time determination matrix is then obtained using the difference operation between the two matrices. The features of the time determination matrix are used for fault section identification and fault distance calculation, to accurately locate faults. The method is modified by considering economic benefits, through the optimal configuration of detection devices of traveling waves when calculating fault distances. Simulation results show that the proposed method has good adaptation with higher fault location accuracy than two other typical ones. It can deal with faults on invalid branches, and the error rate is under 0.5% even with connected DGs.     

一种基于 DSP的电力电子电路在线故障诊断方法

以三相桥式整流电路为例,首先对可控整流装置正常工作和故障运行时整流电压波形进行分析和归类,定义了一种"面积"并建立了故障模型,提出了一种改进的谱分析故障诊断方法,根据其特征值进行诊断定位;其次归纳出实验算法并用 DSP系统实现,实验表明该方法实现了可控整流装置的在线故障诊断和监测;最后总结了该方法的特点并推广至其它形式整流电路.     

A disk-to-disk breakdown and arc modeling method for fault diagnosis of power transformers during impulse testing

Impulse tests are performed on power transformers to assess their insulation integrity. During these tests, impulse voltages are applied, and the resulting voltage and current waveforms are recorded. After postprocessing of these data, the presence of faults in the winding should be detected. Different methods and models have been developed for pattern recognition of transformer winding current and classification of impulse faults, . The drawback of these methods is the representation of a fault as a short circuit with constant conductance. In this paper, the breakdown and arc were modeled to represent a more realistic fault during impulse tests. Using this model, it is possible to generate more realistic data, which are necessary to develop more reliable fault-detecting algorithms and, as a result, better pattern recognition and classification of impulse test waveforms. The model has its roots in the classical Mayr equation, and it is simulated with TACS and MODELS modules in EMTP. The simulation results verify the ability of the suggested method to represent disk-to-disk breakdown and arc fault during impulse testing.     

基于 MADCNN 的故障诊断方法研究

针对旋转机械零部件进行故障诊断的方法包括传统方法和深度学习,传统方法往往需要大量的专家经验,且诊断精度欠佳,提出一种注意力机制改进多尺度深度卷积神经网络(multi-scale attention deep convolutional neural network, MADCNN)的故障诊断方法。MADCNN方法提供3个卷积通道,每个通道差异化的核尺寸原理有效拓宽网络,实现了对原始时域数据的多尺度特征提取。同时,CBAM对提取的特征进一步赋予权重,增强了模型对不同类型故障的区分度。采用凯斯西储大学(Case Western Reserve University, CWRU)轴承故障数据和行星齿轮箱实验台故障数据分别进行实验验证,与传统深度卷积模型相比,验证集准确率提高7.76%。实验结果表明,该方法的诊断精度高,泛化性能好。     

Research on fault diagnosis for MMC-HVDCSystems

Introduction: With the development of flexible HVDC technology, the fault diagnosis of MMC-HVDC becomes a new research direction. Based on the fault diagnosis theory, this paper proposes a robust fault diagnosis method to study the fault diagnosis problem of MMC-HVDC systems. Methods: By optimizing the gain matrix in the fault observer, fault detection with good sensitivity and robustness to disturbance is achieved. In the MMC-HVDC system, because of the inherently uncertain system and the presence of various random disturbances, the study of robust fault diagnosis method is particularly important. Results: Simulation studies during various AC faults have been carried out based on a 61-level MMC-HVDC mathematical model. The results validate the feasibility and effectiveness of the proposed fault diagnosis method. Conclusions: So this fault diagnosis method can be further applied to the actual project, to quickly achieve system fault diagnosis and accurately complete fault identification.     

基于计量异常事件组合的异常原因分析方法

现有的异常用电行为分析多是基于单个计量异常事件进行分析,存在工作量大、准确率低等问题。为了提高异常告警数据利用率和异常诊断准确率,提出了一种基于计量异常事件组合的异常原因分析方法。首先确定了一种基于异常发生日期间隔天数的关联度计算方法,然后通过设定关联度阈值甄别并发异常事件组合,最后将关联度与异常原因概率相结合,提出异常原因分析方法。结果表明该方法能够有效地提高故障诊断的准确率。     

基于改进的VMD和CNN神经网络的光伏逆变器软故障诊断方法研究

针对光伏发电系统中光伏逆变器电路复杂,出现故障时间短等问题,文中提出一种基于改进的变分模态分解和卷积神经网络相结合的故障诊断方法,可有效地解决故障特征提取困难,特征参数奇异性差,以及由于特征参数差而引起的故障诊断率低等问题.利用SIMULINK建立光伏逆变器软故障模型,并采集相关参数作为样本;使用VMD对参数进行变分模...     

一种基于频谱分析的可控整流电路故障诊断方法

以12脉波可控整流电路为例,简要介绍了大功率整流装置的工作原理,并对各种故障情况 下的整流波形进行详细分析和归类;定义了一种"面积",对故障波形进行逻辑预处理,建立了故障模 型,再利用DFT分析方法对模式向量进行谱分析,提出了故障诊断方法;最后归纳出故障诊断系统 的实验算法,总结了本方法的特点并将其推广至其他形式整流电路。     

基于RA-LSTM的轴承故障诊断方法

为解决轴承故障诊断时网络模型不能对一维振动信号中的特征按照重要程度分配不同的权重,导致无法提取具有代表 性意义的特征,进而影响诊断模型的精确度与鲁棒性的问题。 提出基于反向注意力机制( reverse attention mechanism,RA)的特 征突出处理方法,通过将特征进行注意力反向与剪枝,降低非重要特征占比,从而对重要特征进行凸显。 并通过长短期记忆网 络(LSTM)进一步学习特征之间的时间信息后通过全连接层进行故障类型分类。 通过实验选取了最优数据截取长度、剪枝超参 数并对信号添加噪声后模型的稳定性进行了验证。 实验结果表明所提出的 RA-LSTM 轴承故障诊断方法具有优异的故障诊断 性能,故障诊断精度能达到 100%,且在添加噪声后模型的诊断能力仍具有优异的鲁棒性。