大型汽轮发电机在线监测与故障诊断技术的研究.pdf

摘 要
在大型汽轮发电机每根定子线棒出口水处和线棒层间都装设有测温元件，如何
合理地利用这些测温元件进行热故障在线监测与诊断故障是现阶段需要研究的课
题。鉴于此，本文建立发电机定子绕组动态温度水力模型，对该模型进行推导与简
化后，得到描述定子绕组动态温度水力模型的数学表达式，用径向基函数神经网络
（RBFNN）对模型参数进行识别，求取单调连续的两个负荷点间线棒温度的过渡时
间，以此作为监测发电机运行状况的判据。通过对某电厂一台 600MW 的汽轮发电机
进行计算，证明了该方法的正确性。用该种方法进行热故障在线监测最大的优点是
不受定子线棒温度延迟时间的影响。
另外，将小波多分辨率分析与小波能量谱分析方法运用到发电机转子发生匝间
短路时在定子并联支路产生的电压差与环流信号的分析中，并对信号进行了小波包
软阈值去噪。采用 Mallat 算法对信号进行快速分解与重构。采用了第 6 高频细节
的小波能量作为转子匝间短路故障的特征量。实验结果证明了该方法的正确性和有
效性。

关键词：在线监测,故障诊断,人工神经网络,小波变换,转子匝间短路
ABSTRACT
The elements of surveying temperature are fixed on the end of the stator bars and
between the stator bars in large generators. how reasonably uses these temperature
elements to monitor and diagnose the thermo fault is the topic which the present stage
needs to study. In view of this, a thermo hydraulic model of the water-cooled stator bars in
large generator is presented, through calculating and deducing, the formula reflecting the
dynamic model is given, then, radial-basis function neural network(RBFNN) is used to
identify the model parameters and calculate the transient time of the temperature which
occurred to water-cooled stator bars from load-1 to load- make the transient time as a
tool to judge the generator working in normal condition or in thermal fault condition. an
example which bases on a 600MW synchronous generator is used to prove this technique
correctly. And the distinct advantage of this technique is that the judgment result is not
affected by delay time which caused by temperature signal.
Multi-resolution analysis and Wavelet Energy spectrum are used to analysis the
signals of difference voltage and circular current when the rotor windings are faulty .and
using wavelet packet to reduce noise of signals. Mallat’s decomposition algorithm is
applied to decompose the signals, in the end, the wavelet energy o