基于脉冲在线注入的变压器绕组变形特性无损监测与故障识别及预警方法的基础研究

Winding fault is one of the main fault types of the transformer, which has a great threat on the transformer and safety operation of the power grid. Based on the technology of pulse power generating and measuring, this project studies deeply on the winding fault characteristics, pattern recognition and early warning of the transformer and other key basic scientific issues. By using finite difference time domain method, three-dimensional transient propagation model of winding transient travelling wave in the is established, to reveal the propagation characteristics of pulse signal in the winding. Based on a controllable total-solid-state pulse generator with FPGA，pules signal would be injected online and actively and non-destructively， winding network response curve can be obtained. By exacting the characteristic parameters, the winding fault pattern recognition method is established. Coupled with working state and historical data，the relationship between characteristic parameters and winding faults would be studied deeply ，and winding faults analysis，identification and warning system would be developed.This project aims to reveal the principle and method about transformer winding faults non-destructive monitoring，identification and warning based on controllable all-solid-state pulse generator theory. It will lay the foundation for the new winding fault monitoring system with completely independent intellectual property rights and provide both theoretical and experimental basis for the winding fault monitoring and early warning system.It has important scientific significance and practical value,and will bring enormous social and economic benefits.

绕组故障是变压器最主要故障类型之一，是造成变压器损坏和威胁电网安全运行的重要原因。本项目基于脉冲功率发生和测量技术，对变压器绕组故障特性、模式识别和预警等关键基础科学问题进行深入研究。基于时域有限差分法仿真建立绕组暂态行波的三维瞬态传播模型，揭示绕组中脉冲信号的传播特性；基于FPGA灵活控制的全控型全固态脉冲发生器实现向变压器绕组无损、在线、主动注入脉冲信号，获取绕组网络响应曲线并提取特征参数，建立绕组故障的模式识别方法；结合运行工况和历史数据，深度挖掘特性参数与绕组故障的关系，研究特性参量与绕组故障之间的机理，不断完善绕组故障分析、识别及预警系统。本项目旨在揭示基于脉冲在线注入的绕组变形无损特性监测与故障识别及预警的原理和方法，为最终成功开发拥有完全自主知识产权的新型绕组变形故障监测及预警系统提供理论基础和实验依据，具有重要的科学意义和工程实用价值，将带来巨大的社会和经济效益。

绕组故障是变压器最主要故障类型之一，是造成变压器损坏和威胁电网安全运行的重要原因。本项目基于脉冲功率发生和测量技术，对变压器绕组故障特性、模式识别和预警等关键基础科学问题进行深入研究。取得的主要成果有：①对注入激励脉冲的类型及关键参数进行选择。利用套管结构设计了套管电容耦合传感器，仿真和耐压试验测试证实传感器的安装不影响套管正常运行；设计信号在线注入／测量电路；基于FPGA开发便携式绕组变形故障检测仪，具备可调脉冲发生、双通道信号采集、数据后处理等多项功能。②建立计及套管耦合回路的变压器绕组等值在线电路模型，利用阻抗关系解释在线频率响应曲线的形态特征；分析耦合电容变化和套管绝缘击穿故障对在线频率响应的影响，结果表明两者均引起频响曲线上下方向偏移，可用于区分套管耦合回路故障和绕组变形故障。③通过变压器离线测试证实本方法能够反映绕组的健康和故障状态，通过变压器加压测试实例证实检测方法和装置的安全性和较高的重复性，为该方法的实际应用奠定基础。将检测方法应用于750 kV高压并联电抗器绕组故障的检测中，首次将检测装置安装于运行的750 kV高压并联电抗器，初步实现检测方法的实际应用。④利用短时Fourier变换处理变压器暂态检测信号，得到脉冲频率响应曲线相比FFT方法清晰度更高，但也指出该方法本质更适合处理次稳定信号；进而研究利用复连续小波变换从暂态信号获取脉冲频率响应曲线，构造的暂态信号仿真和变压器等值电路模型仿真以及变压器实验表明，由小波变换得到的脉冲频率响应曲线较FFT方法更接近绕组扫频频率响应曲线，且曲线受噪声影响较小，利于实际绕组变形故障诊断。⑤开展变压器绕组变形故障模拟试验测试，提出以频率响应的有效极值点的频率变化和分频段均方误差指标构造故障诊断特征参量，利用支持向量机算法识别绕组变形故障类型，分析表明，使用两种特征参量，10组测试样本的总体识别准确率分别为78.1％和83.3％，均取得较满意的结果。