Feature extraction and reduction applied to anomalous state detection of power transformer
The existing vibration method for three-phase dry-type transformer state detection uses feature extraction from high-frequency vibration data to achieve detection.In order to mine the effective features of high-frequency vibration signals more accu-rately and improve the accuracy of diagnosis,for high-frequency vibration signals,the continuous signals are divided into inde-pendent periodic segments,and the time-domain,frequency-domain and time-frequency-domain features of the signals are extrac-ted from each cycle;Then,based on the analysis of transformer vibration principle and the correlation analysis,the feature selection strategy is given;Finally,the filtered features are used as fault eigenvalues to detect vibration anomalies in real-time.The results show that this method can effectively extract the characteristics of transformer vibration signals and improve diagnostic accuracy.
NETL
NSTL
万方数据
