Power Transformer Voiceprint Operation State Monitoring Considering Sample Unbalance

The voiceprint characteristics of transformers are closely related to the operating conditions, but there is currently a lack of effective research on the voiceprint characteristics of transformers during operation. First of all, this article collects three operating conditions of load, light load, and no load on the basis of the transformer voiceprint signal acquisition platform. Secondly, in view of the characteristics of the transformer’s voiceprint, the 50Hz frequency multiplier component amplitude is extracted to form a feature vector, which solves the problem of low utilization rate of common feature extraction information. Finally, in view of the problem of transformer voiceprint failure and sample imbalance caused by fewer abnormal samples, a pattern recognition based on the RUSBoost algorithm is proposed. The algorithm has good recognition accuracy and applicability for transformer voiceprint samples with imbalance problems. The research results provide effective support for the monitoring and identification of the mechanical condition of transformers with sample unbalanced voiceprints, and the analysis of the operating conditions can effectively eliminate the errors that may be caused by their own different operating conditions.