基于多物理量的GIS状态智能诊断研究

发布时间：2018-07-16 15:15

【摘要】：GIS设备(气体绝缘金属封闭开关设备)因其集成度高、运维方便、占地面积小等优点,应用程度越来越高,虽然运行可靠性较高,但是一旦发生故障,后果相当严重,检修工作繁杂,停电时间长,停电范围波及非故障元件。因此,必须加强GIS智能化和信息化建设,实现设备状态智能诊断,最终指导制定相应检修策略。GIS设备的智能化和信息化,主要是通过灵敏的状态监测手段、可靠的评价手段来判断GIS设备的运行状态,并且在设备运行状态出现异常时对设备进行故障分析,对故障部位、严重程度做出判断,并识别诊断故障初期征兆。目前国内外已应用的GIS设备状态带电检测和在线监测设备,多数采用单一种类的传感器对单一物理量进行检测和诊断,较少采用多参量综合检测的方法去研究GIS设备在运行过程中绝缘状态的变化过程与规律,如超声、特高频等。如何应用多种类型传感器,并有机地整合传感器和相应的数据采集、信号传输系统,对设备进行高效状态分析和故障诊断将是一个重要研究方向。本论文重点研究和探索基于声、电等信号传感器阵列实现GIS设备状态监测和智能评估诊断的方法和关键技术。首先,本文针对特高频和超声波信号在GIS设备中的传播特性以及设备中的绝缘盆、L型弯、T型分支等常见结构对于信号的影响,给信号检测和定位增加的复杂性进行了分析。第三章研究了基于高阶累积量和双谱估计的信号时延估计算法,通过以双指数振荡衰减函数模拟局部放电信号验证了该时延估计算法的准确性,最终利用该时延算法计算出实测特高频信号的时延,并将该时延序列应用于GIS设备局部放电定位,预测出局部放电源的空间位置。第四章研究了GIS状态诊断技术平台的设计,包括超声、特高频传感器及其阵列优化、监测IED装置架构设计、多通道同步采集装置研制,并研究了综合考虑时域脉冲和统计图谱的局放类型识别方法,通过分别利用主成分分析方法、信息增益方法及支持向量机回归消去方法等特征选择方法来降低特征空间维数,提高模式识别的效率,然后通过状态信息可视化监控平台实现GIS状态信息及故障类型预判展示。最后,通过建立GIS模拟实验平台,设计了 5种典型绝缘缺陷模型,利用本文设计的诊断平台进行模拟实验测试,对声电联合的局放精确定位算法进行了实物实验验证,并取得了主要故障类型指纹图谱和典型缺陷包络特征。
[Abstract]:Because of its advantages of high integration, convenient operation and maintenance, small area and so on, GIS equipment (gas insulated metal closed switchgear) is being applied more and more highly, although its operation reliability is high, but once it breaks down, the consequences are quite serious. Overhauling work, long blackout time, power failure range and non-fault components. Therefore, it is necessary to strengthen the construction of GIS intelligence and information, to realize the intelligent diagnosis of equipment status, and finally to guide the establishment of the corresponding maintenance strategy, the intelligence and information of GIS equipment, mainly through the sensitive state monitoring means. The reliable evaluation method is used to judge the running state of GIS equipment, and to analyze the fault of the equipment when it is abnormal, to judge the fault location and severity, and to identify the early symptoms of the fault diagnosis. At present, most of the existing GIS equipment used at home and abroad are used to detect and diagnose the single physical quantity by using a single kind of sensor. The method of multi-parameter comprehensive detection is seldom used to study the changing process and law of insulation state of GIS equipment during operation, such as ultrasonic, UHF and so on. How to apply various types of sensors and integrate sensors with corresponding data acquisition and signal transmission systems is an important research direction for efficient state analysis and fault diagnosis of equipment. This paper focuses on the research and exploration of the methods and key technologies of GIS equipment condition monitoring and intelligent evaluation and diagnosis based on acoustic and electrical signal sensor array. Firstly, this paper analyzes the influence of UHF and ultrasonic signal propagation characteristics in GIS equipment and the influence of the common structures such as the insulation basin L bending T branch on the signal, and the complexity of signal detection and localization is analyzed. In chapter 3, the time-delay estimation algorithm based on high-order cumulant and bispectral estimation is studied, and the accuracy of the time-delay estimation algorithm is verified by simulating partial discharge signals with double-exponential oscillation attenuation function. Finally, the delay of UHF signal is calculated by using the delay algorithm, and the delay sequence is applied to the location of partial discharge of GIS equipment, and the spatial position of local discharge power supply is predicted. The fourth chapter studies the design of GIS status diagnosis technology platform, including ultrasonic, UHF sensor and its array optimization, monitoring IED device architecture design, multi-channel synchronous acquisition device development. The method of partial discharge type recognition considering time domain pulse and statistical spectrum is studied. The feature space dimension is reduced by using principal component analysis method, information gain method and support vector machine regression elimination method, respectively. The efficiency of pattern recognition is improved, and then the GIS state information and fault type prediction are displayed through the state information visual monitoring platform. Finally, through the establishment of GIS simulation experiment platform, five kinds of typical insulation defect models are designed and tested by using the diagnostic platform designed in this paper. The fingerprint of main fault types and the envelope features of typical defects are obtained.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM595

【相似文献】