Research on Transformer Fault Diagnosis Method Based on Small Sample Voiceprint Recognition Technology

As an important equipment in power system, it is very important to ensure the safe and stable operation of power transformer. The diagnosis effect of the fault is significantly affected by the quantity and quality of the fault voiceprint samples. However, the occurrence of transformer faults is a small probability event, the sample of abnormal state of equipment is scarce, and the density of data value is low, which makes it difficult for intelligent diagnosis algorithm to realize the effective diagnosis of transformer operation state under the condition of less data. Therefore, a transformer fault diagnosis method based on small sample voiceprint recognition technology is proposed, which has the advantage of not needing a lot of training data set. Firstly, according to the principle of voiceprint recognition technology, the Mel cepstrum features of transformer operation are extracted. Then, the fault voiceprint of the transformer is compared and classified by the small sample voiceprint recognition method. Finally, the analysis and verification are carried out under the measured data of a transformer factory. Compared with the traditional deep learning algorithm, the results show that the method has higher recognition accuracy, which is of great significance to engineering practice.