基于EEMD的变压器振动与声音信号IMF峭度特征量提取方法

针对变压器振动问题,提出了一种基于集合经验模态分解(EEMD)的本征模函数(IMF)峭度特征量提取方法,并运用相关系数法、快速谱峭度图法提取敏感IMF分量。提取试验变压器正常、铁心松动故障状态下的振动、声音信号的特征量,研究变压器在正常、故障状态下这两种信号特征量分布情况;分析实际运行中出现铁心磁路故障、铁心多点接地故障状态的变压器的IMF峭度特征。结果表明,提出的特征量提取方法可同时反映频域、时域特性;在不同故障条件下,振动与声音信号的特征量变化不同,二者可相互补充,研究两种信号更有利于变压器状态的判定。

Feature Extraction Method for Transformer Vibration and Audio Signal Based on Ensemble Empirical Mode Decomposition

Aiming at the problem of transformer vibration, a kurtosis feature quantity extraction method was studied based on ensemble empirical mode decomposition (EEMD). The correlation coefficient method and fast spectral kurtosis method were used to extract sensitive IMF components. The method was used for feature extraction of vibration and audio signal of the test-transformer under normal and the core loosening fault condition. The distributions of the two signal features of the transformer under different conditions were studied. The IMF kurtosis characteristics analysis was performed on the transformers with core magnetic circuit faults and multi-point ground faults states in actual operation. The results show that the proposed feature quantity extraction method can reflect the frequency domain and time domain characteristics simultaneously. Under different fault conditions, the vibration and sound signal characteristics change differently, and the two can complement each other. Studying the two signals is more beneficial to the transformer state determination.