采用最优小波树和改进BP神经网络的感应电动机定子故障诊断

发布时间：2018-01-03 01:00

本文关键词：采用最优小波树和改进BP神经网络的感应电动机定子故障诊断　出处：《电工技术学报》2015年24期 　论文类型：期刊论文

【摘要】：为了准确及时地识别并排除感应电动机定子匝间短路故障,保障电动机设备的安全运行,提出了一种基于最优小波树和捕食搜索遗传算法优化神经网络的新型故障诊断方法。结合故障电流的特征,采用最优小波树,将滤除基波分量后的定子残余电流信号进行分解,提取表征信号内在规律最强的分解节点能量成分,作为BP神经网络的输入特征向量。采用BP神经网络进行分类,通过捕食搜索策略优化的遗传算法选择神经网络训练的初始权值和阈值,提升网络训练的速度和准确度。实验结果表明,该方法不但可以提取优于小波包方法的最优特征向量,同时可以准确识别三种故障下的电动机定子匝间短路故障。
[Abstract]:In order to accurately identify and exclude the induction motor stator interturn short circuit fault, ensure the safe operation of electric equipment, puts forward a new fault diagnosis method of neural network optimized by genetic algorithm to search the optimal wavelet tree. Combined with the characteristics and predation based on fault current, the optimal wavelet tree, the signal of stator residual current fundamental component after filtering the decomposition, decomposition of node energy extraction of signals the inherent law of the strongest, as input vectors of BP neural network. The classification based on BP neural network, neural network training through the initial weights and thresholds of predatory search genetic algorithm optimization, improve network training speed and accuracy. The experimental results show that the optimal this method not only can extract the feature vector is superior to the wavelet packet method, and can accurately identify three kinds of motor fault conditions Stator turn to turn short circuit fault.

【作者单位】：中国矿业大学信息与电气工程学院;
【基金】：教育部科学技术研究重大资助项目(311021)
【分类号】：TM346
【正文快照】： 0引言感应电动机转速高、容量大,是电力系统的重要支撑设备之一,其常见的故障主要有定子、转子、轴承及气隙偏心等故障。其中定子绕组匝间短路故障是感应电动机经常出现的故障类型,故障率约占电动机总故障的37%,因此能够在早期及时准确地预测出故障并排除具有重要的安全和经济

【参考文献】