A new transformer condition diagnosis method integrating machine vision and auditory response is proposed to enhance the status monitoring and fault diagnosis capability of dry-type transformers.Firstly,potential fault locations of the dry-type transformer are identified and monitored by target detection algorithms to obtain their positional information.Secondly,the acoustic information of the potential fault location is calculated by integrating the delay-and-sum beamforming algorithm,and the acoustic information is used to monitor the occurrence of discharge defects in real time.Finally,for me-chanical faults that cannot be quickly diagnosed,the auditory peripheral model is used to obtain the auditory spectrum.Based on the characteristic frequency bands of the sound signal,multiple feature frequency auditory spectra are obtained,and the transformer fault type is identified by a convolutional neural network.The experimental results show that the pro-posed method has an accuracy rate up to 87.51%for the transformer state recognition under noisy conditions,which is bet-ter than other comparative time-frequency domain feature extraction methods.Therefore,the proposed method can diagnose the state of dry-type transformers more accurately,has better noise resistance feature,and effectively complement the current monitoring methods for transformers.
关键词
故障诊断/干式变压器/状态检测/机器视觉/声学成像/听觉外周模型
Key words
fault diagnosis/dry-type transformer/state detection/machine vision/acoustic imaging/auditory periph-ery model
维普
万方数据