基于绕组短路冲击响应时频特性的多特征联合形变判定方法

Multi-Feature Joint Deformation Determination Method Based on Time-Frequency Characteristics of Winding Short-Circuit Impulse Response

为研究变压器短路冲击信号包含的物理特征与绕组形变之间的关系,提出一种基于绕组冲击响应时频特性的特征提取及形变判定方法。基于单自由度系统的相对位移模型构建绕组的短路冲击响应谱,得到其响应点频率偏移特征。对Hilbert能量谱进行非均匀频率带宽划分,修正累积效应的影响,提取不同频带范围内的能量带熵并计算加权能量带熵,表征特定频带内的能量混乱度;构建时频矩阵,量化能量密度分布、短时能量衰减变化等细节特征,对该矩阵进行奇异值分解,根据奇异值分布特性计算每个元素的权重值。据此提出基于加权能量带熵及基频奇异值畸变率的联合绕组状态判定方法,经三组共17次短路冲击试验数据验证,冲击信号的相对能量带熵增量变化与绕组产生的累积微应变呈高度正相关,基频奇异值畸变率与绕组发生形变的规律一致,发生形变时,该畸变率的值会大于2.0,证明了该判定方法的有效性。该方法可在变压器发生短路后对绕组的状态进行预判,进而考虑是否需要进一步处理,对变压器绕组状态的在线监测具有一定的指导意义。

The transformer will suffer many external short-circuit impact during long-term operation,and the winding will be affected by many accumulations,which will produce irreversible deformation.However,there are still some problems in the study of on-line monitoring of winding mechanical status.The operation of transformer is affected by the accumulative effect,and the relevant studies have not modified the relevant eigenvalues by de-accumulation.The fault diagnosis is mainly aimed at the fundamental frequency vibration signal of the transformer,but the frequency doubling component within 500 Hz is generally present in normal operation,and the transient state contains more higher harmonics.To solve these problems,a feature extraction and deformation determination method based on time-frequency characteristics of shock response of winding is proposed in this paper,which can accurately judge the deformation of winding.Firstly,based on the relative displacement model of the single degree-of-freedom system,the short-circuit shock response spectrum of the winding is constructed,and the distribution law of the vibration shock response of the winding considering the cumulative effect is obtained.The Fourier spectrum and shock response spectrum distribution of the measured signal are compared to determine the frequency analysis range.Secondly,the time-frequency energy distribution of the short-circuit impulse signal is calculated,and the frequency bands of the energy spectrum are divided according to the response distribution characteristics obtained by the impulse response spectrum to offset the influence of the cumulative effect,and the characteristic changes of the relative energy band entropy are obtained,and the weighted entropy increment is calculated.Thirdly,transform the energy spectrum into time-frequency matrix,decompose the signal by singular value decomposition(SVD),get the singular value vector,and calculate its weight according to the distribution characteristics of the singular value,can calculate the fundamental frequency singular value distortion rate of the signal.Finally,the calculated weighted entropy increment and the singular value distortion of fundamental frequency are compared with the measured cumulative microstrain.After converting the shock response spectrum into a planar strip,it can be seen that there is basically no accumulative effect when the short-circuit current is relatively small,so the first four shock response bands do not change significantly.Since the fifth time,due to the accumulative effect,the base-frequency short-circuit impulse response of A and C two-phase winding increases from about 100 Hz to 105.3 Hz and 104.7 Hz respectively,but the high-frequency part does not change significantly.The shock response of the B-phase fundamental short-circuit increases from 101.5 Hz to 107.6 Hz,and the peak distribution of the main response gradually increases to 200~550 Hz.In the range of 550~1000 Hz,the shock response of the winding is unevenly distributed.According to this,the frequency bands are divided,the accumulative effect is corrected,and the change of weighted entropy is calculated,which is highly correlated with the accumulative microstrain of the winding.However,because the threshold of entropy increment is not easy to judge,and the requirement of data acquisition is high,it can not be missed or miscollected,so the concept of fundamental frequency singular value distortion is proposed,and the distortion rate does not exceed 2.0 when there is no deformation,which can be used as an auxiliary judgment means.Through the analysis of the feature extraction and deformation determination methods proposed in this paper,the following conclusions are drawn:(1)Less high-frequency components in the spectrum can cause a higher level of shock response in the winding,and it is necessary to conduct a comprehensive analysis of signals within 1000 Hz.(2)Considering the offset degree of shock response points affected by cumulative effects,the relative energy band entropy in different frequency ranges is calculated,which is consistent with the law of winding burst deformation,indicating that it is correlated with winding deformation.(3)The weighted entropy increment of the short-circuit impact signal is calculated and compared with the cumulative microstrain of the winding.The two are linearly correlated.The calculated singular value distortion rate of the short-circuit impact signal does not exceed 2.0 when the winding is not deformed.The combination of the two can accurately predict whether the winding deformation occurs after short-circuit.