摘要
变压器振动信号中包含了大量状态信息,但难以从中提取有效特征进行绕组松动状态识别。为此,提出了基于模糊自适应共振理论(fuzzy adaptive resonance theory,Fuzzy-ART)的变压器绕组松动状态识别方法。首先,设置9种绕组松动状态并进行短路实验,测取油箱表面振动信号;然后对振动信号进行4层小波包变换,提取有效测点状态特征频带的小波包能量构成特征向量;最后将特征向量作为Fuzzy-ART神经网络的输入,对不同绕组松动状态进行识别。实验结果表明,基于小波包的Fuzzy-ART神经网络能对绕组松动状态进行快速、稳定分类,可用于变压器绕组松动状态的在线监测与诊断。
Though there are great deal of state information in transformer vibration signals, it is hard to extract effective fea-tures to recognize winding looseness state of the transformer. Therefore, this paper presents a state recognition method for transformer winding looseness based on fuzzy adaptive resonance theory （Fuzzy-ART） neural network. Firstly, it sets nine kinds of winding looseness states for short-circuit test and measuring vibration signal on the surface of oil ^ plies four layers wavelet packet transform for vibration signal and extracts wavelet packet energy of state feature bands of ef-fective measuring points to form feature vectors. Finally, it takes feature vectors as input of Fuzzy-ART neural network for recognizing different winding looseness states. Test results indicate that the Fuzzy-ART neural network based on wavelet packet can rapidly and stably classify winding looseness state which is available for online monitoring and diagnosis on trans-former winding looseness state.
出处
《广东电力》
2017年第7期89-95,共7页
Guangdong Electric Power
基金
国网江苏省电力公司重点科技项目(J2014055)
关键词
变压器
绕组松动
振动信号
小波包能量
Fuzzy-ART神经网络
状态识别
transformer
winding looseness
vibration signal
wavelet packet energy
fuzzy adaptive resonance theory neural network
state recognition
