电力电子整流装置故障诊断方法的研究

发布时间：2019-01-12 14:13

【摘要】：鉴于电力电子功率元件的普遍使用和不断更新,电力电子设备在生产和生活方面广泛应用。电力电子整流装置是重要的组成部分,其可以完成能量的变化。但是一旦异常,系统可能损坏甚至威胁生命。因此,将诊断技术向智能化和精确定位方法发展是有意义的。本文以三相全控整流电路为例,就故障提取和识别两个重要技术进行优化,主要完成了以下工作:首先,研究了三相全控整流电路的故障模式,对可能出现的故障进行预测和整理。通过MATLAB平台搭建故障仿真模型,就常见的异常状态依次仿真,异常波形被得到,异常状态的输出电压从相应的测试点被得到。其次,故障点的特征提取分别用傅里叶变换和三层小波包分析针对整流输出电压进行。结果表明:经小波包分解后大量的异常信息分布在频率能量中,此方法能够合理提取异常特征。最后,被提取的故障特征作为BP神经网络和Elman神经网络的输入向量,即完成各种故障的识别和检测,结果表明:基于Elman神经网络诊断法可以很快完成训练、测试精度更高,达到了预期的效果。综上所述:用小波包分析和Elman神经网络组合的方式获得故障点的特征信息并进行分类是本文的一个亮点。实验结果表明:用此方法训练时间更短、测试误差更小。这是检测电力电子整流装置故障方面一种新的尝试,该方法的具有一定的理论意义和价值。
[Abstract]:Power electronic devices are widely used in production and daily life due to the widespread use and updating of power electronic components. Power electronic rectifier device is an important component, it can complete the change of energy. But in the event of an exception, the system can be damaged or even life-threatening. Therefore, it is meaningful to develop diagnostic technology to intelligent and accurate positioning method. Taking three-phase full-control rectifier circuit as an example, this paper optimizes two important techniques of fault extraction and identification. The main work is as follows: first, the fault mode of three-phase full-control rectifier circuit is studied. Forecast and sort out possible failures. The fault simulation model is built on the MATLAB platform. The usual abnormal state is simulated in turn. The abnormal waveform is obtained and the output voltage of the abnormal state is obtained from the corresponding test point. Secondly, the feature extraction of fault points is carried out by Fourier transform and three-layer wavelet packet analysis respectively for rectified output voltage. The results show that a large amount of abnormal information is distributed in the frequency energy after wavelet packet decomposition, and this method can reasonably extract abnormal features. Finally, the extracted fault features are used as input vectors of BP neural network and Elman neural network, that is to say, all kinds of faults are identified and detected. The results show that the diagnosis method based on Elman neural network can complete the training quickly and the test accuracy is higher. The expected effect has been achieved. To sum up, wavelet packet analysis and Elman neural network combination are used to obtain the feature information of fault points and classify them. The experimental results show that the training time is shorter and the test error is smaller. This is a new attempt to detect the fault of power electronic rectifier, and this method has certain theoretical significance and value.
【学位授予单位】：安徽理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM461

【相似文献】