水轮发电机组振动与局放的监测与诊断

Hydroelectric generating set is the core device of the hydropower station, it consists of two parts, hydraulic turbine and hydro-generator. A lot of practices show that, 80% faults of the hydraulic turbine can be shown in the vibration signal, and the most faults of the hydro-generator can be find expression in the partial discharge (PD) of the insulating materials. This issue takes the two kinds of main faults: the vibration of the hydraulic turbine and the partial discharge of the hydro-generator as the focus topic. By monitoring the state of vibration and PD signal to clarifying the mechanism of the fault, studying the signal processing technology and intelligent diagnosis method. .Through collecting and analysis of the fault data of the vibration and PD signal, the fault samples of vibration and PD fault were obtained, then these data were analyzed and processed to set up the fault feature sets of hydroelectric generating set. By reference from latest research achievements of modern signal processing and artificial intelligence field, and by using the Hilbert-Huang transform, extracting the fault features based on the auto regressive model and taking the least squares support vector machines, and so on, the vibration and PD signal processing, de-noising, fault features extracting and fault diagnosis methods of the hydroelectric generating set were carried out to propose a intelligent diagnosis method for the vibration and PD fault of the hydroelectric generating set, and to develop a signal processing and fault intelligent diagnosis software which has the independent intellectual property rights, promoting the fault diagnosis study of the hydroelectric generating set.

水轮发电机组是水力发电厂的核心装置，包括水轮机和水轮发电机两大部分。大量的实践表明，水轮机的故障80%可以通过振动信号表现出来，而发电机的故障大部分以绝缘局部放电的形式表现出来。本课题以水轮机的振动和发电机的局部放电这两种最主要的故障类型为研究重点，通过对振动和局放状态的监测，弄清故障的机理，研究其信号处理技术与智能诊断方法。.通过收集水电站的振动和局部放电的故障信号数据，对故障信号进行分析处理，获取水轮发电机组的故障样本，建立水轮发电机组的故障特征集。借鉴现代信号处理与人工智能领域的最新研究成果，分别应用Hilbert-Huang变换、基于自回归（AR）的信号特征提取以及最小二乘支持向量机的方法，开展水轮发电机组振动以及局部放电故障的信号分析、抗干扰、故障特征提取和故障诊断方法的研究，提出适合水轮发电机组振动和局放在线诊断的智能诊断方法，开发具有自主知识产权的信号处理与故障智能诊断软件，

水轮发电机组是一个复杂的非线性动力系统，其运行过程中故障信号的出现和发展受水力、电气和机械等方面的耦合作用，给水电机组的故障诊断带来了一定的困难。项目的主要内容主要包括：研究了强噪声背景下水轮发电机组早期微弱故障信号检测方法和信号去噪技术：借助噪声辅助分析理论分析了噪声强度对随机共振现象的影响，并建立了二阶部分Duffing振子随机共振微弱信号检测模型；分析了局放信号与噪声相关性的差异并通过计算各阈值层故障概率值识别局放信号；构造了多尺度熵阈值代替传统的直接阈值思想，提出了多尺度熵EMD阈值信号去噪方法。研究了多振源耦合作用下的水轮发电机组非平稳信号故障特征提取技术：构建了二元树复小波特征熵提取水轮机尾水管动态特征信息；建立了基于复数据经验模态分解的特征提取方法；考虑到局放信号较强的非线性、时变性，利用相空间重构方法建立局放信号的高维相空间映射，计算每个分量的时频分布并构建局放信号时频流形，结合局部线性嵌入的维数约减结果实现局放信号在低维空间的特征提取；建立了基于多维度样本熵的局放信号特征提取方法。研究了水轮发电机组故障诊断方法和性能评估模型：建立了改进粒子群优化的支持向量机水轮发电机组故障诊断模型；针对故障诊断中的多分类问题构造了基于球结构的支持向量机算法，应用于局放信号特征的识别与诊断；以水轮发电机组监测信号和运行工况等参数为自变量，建立了基于H-K聚类逻辑回归模型的水电机组水导轴承磨损性能评估模型。项目中开展的信号处理方法和故障诊断模型研究已成功应用于黄河河口、炳灵等多个电站的机组状态监测与诊断系统中，取得了良好的应用效果，并为后续逐步实现以可靠性为中心的状态检修策略奠定了基础。