油浸变压器端部垫块脱落故障准稳态模式研究

Research on Quasi-steady-state Pattern of End Block Shedding in Oil Immersed Transformer

变压器的声信号包含着丰富的信息,实现声纹信号的信息挖掘对变压器状态监测具有重要意义。以某变压器端部垫块脱落故障的异常声纹信号为切入点,通过对现场数据的特征提取,探索新的变压器声纹模式。研究发现,端部垫块脱落故障发生时变压器振动体现为两种交替的振动模式,模式1为平稳振动阶段,模式2为绕组碰撞发生阶段。首先,根据声纹特征引入准稳态的猜想,利用电磁–振动力学模型建模解释了准稳态模式切换过程及高频分量产生原因;其次,对准稳态声纹信号进行处理,采用Pearson自相关函数计算时频谱帧序列自相关度的方法找到区分两种模式的特征指标;再次,利用Mann-Kendal(M-K)突变检测法界定自相关曲线的突变时刻,完成信号分离,模式2证实了撞击引入高频分量,验证了理论模型;最后,根据准稳态声纹特点,通过划定现场和铁芯松动两数据集的Pearson自相关度均方根阈值范围实现准稳态状态检测。通过引入准稳态振动模式,为变压器状态判定提供了新的参考依据。

Acoustic signals of transformers contain abundant information,and the realization of information mining of voice print signals is of great significance to transformer condition monitoring.In this paper,the abnormal voice print signal of a transformer end block shedding fault was taken as the breakthrough point,and the new voice print pattern of transformer was explored through the feature extraction of data.Firstly,according to the voice print characteristics,the quasi-steady-state conjecture was introduced,and the switching process of two patterns as well as causes of high frequency components were explained by using the electromagnetic and vibration mechanical model.Secondly,the quasi-steady-state voice print signal was processed,and the Pearson autocorrelation function was used to calculate the autocorrelation degree of the spectrum frame sequence to find the characteristic index to distinguish the two patterns.Thirdly,the Mann-Kendal(M-K)mutation detection method was used to define the mutation time of the autocorrelation curve,so as to complete signal separation.Mode 2 confirms the introduction of high frequency components and validates the theoretical model.Finally,according to the characteristics of the quasi-steady state voice pattern,the quasi-steady state fault identification was realized by delimit the Pearson root mean square(RMS)threshold range of self-correlation between the field and core loosening data sets.Introducing quasi-steady-state vibration mode provides a new reference for transformer state determination.