基于电力变压器的局部放电定位方法研究OA

Partial discharge is one of the main causes of insulation aging and performance degradation in power transformers,hence promptly determining the location of partial discharge is crucial for the safe operation of transformers.In this paper,based on the collected acoustic signals generated by transformer partial discharge using a microphone array,an audio classification and clustering system that integrates a Transformer encoder with a simple contrastive learning of visual representations self-supervised learning framework is constructed.In this method,multi-modal fusion is firstly performed by extracting the signal's Mel-spectrogram,Mel-frequency cepstral coefficients,and chroma features.Then,the Transformer encoder is utilized to capture global dependencies of time-frequency features,which are combined with contrastive learning to constrain the spatial distribution of features,making them compact within classes and separated between classes.Finally,a clustering algorithm is used to achieve precise spatial localization of partial discharge sources.Experimental results demonstrate that the proposed method can effectively locate partial discharge sources and performs well on multiple evaluation metrics.This method not only provides a new idea and approach for transformer partial discharge detection,but also offers valuable reference and technical support for the application of self-supervised learning in audio signal processing.