【作者】 秦宇；

【导师】 陆益民；

【作者基本信息】 广西大学 ， 电力系统及其自动化， 2013， 硕士

【摘要】 由于电力电子功率元件的广泛应用和快速发展,电力电子诊断技术在当今的电力系统、工业自动化、计算机、电子信息等领域影响至深,研究电力电子故障诊断意义重大。本课题以新型的四相整流器为研究对象,进行基于人工神经网络的故障诊断算法研究。本文首先研究了四相全控整流桥的故障模式,对其故障形式和种类作了全面预测和统计整理。然后,选取BP和Elman神经网络两种不同的模式识别网络分别进行故障诊断。通过Matlab平台搭建四相全控整流桥电路得到了大量故障仿真实验数据,建立了故障诊断系统样本数据。接着,建立故障诊断系统,第一级诊断系统基于BP神经网络,BP网络诊断法以输出的直流电压作为故障测试点,采集的数据归一化后用频谱分析进行处理,然后再将所获取的变换数据进行BP网络识别分析；第二级诊断系统分别基于BP网络和Elman网络诊断法,以直流电压的小波分析为故障数据,分别通过BP网络和Elman网络算法对故障数据进行模式识别。最后,通过对两种故障诊断方法进行测试和对比,表明：BP网络在第二级诊断中无法达到指定的精度,而基于Elman网络的故障诊断系统对于对区分度不是很明显的数据组进行模式识别时,识别能力强,精度高,达到预期效果。更多还原

【Abstract】 The electric and electronic diagnosis technology influences to deep in today’s power system, industrial automation, computer, electronic information, etc duing to the power electronic and power device of the wide application and fast development. The research on the electrical power electronic fault diagnosis is of great significance. This topic is new type of four phase rectifier as the research object, with the focus on the fault diagnosis algorithms based on artificial neural network.This paper first researches the model circuit:four phase full controlled rectifier bridge of the specific work mode, the failure forms and species for a comprehensive forecast and statistical finishing.Secondly, we select different test points and different pattern recognition networks for fault diagnosis. Through the Matlab platform on four-phase full controlled rectifier bridge circuit, we get a lot of experimental data to establish the fault diagnosis system.Then, We establish the fault diagnosis system. Primary diagnosis system bases on BP neural network. BP network diagnostic uses DC output voltage as fault testing point, make spectrum analysis for processing after Normalizing Acquisition of data, then use the transformed data to make BP network discriminant analysis. Secondary diagnosis system respectively bases on BP network and the Elman network diagnosis method. We use DC voltage of the wavelet analysis as fault data to achieve the pattern recognition through the BP network and Elman network algorithm for fault data analysis. Finally, through test and comparison of two methods of fault diagnosis, Shows that the BP network can not meet the specified accuracy in the secondary diagnosis. And based on the Elman network fault diagnosis system for pattern recognition when data set Differentiation is not obvious, it has Recognition ability, high accuracy to achieve the desired effect.更多还原