Research on Voice Print Recognition of Electrical Faults Based on Attention-MFCC Algorithm

The sound of electrical equipment during operation contains a lot of information about the status of power equipment. In order to extract status information to identify the operating status of power equipment, a voiceprint recognition algorithm for electrical equipment was proposed in this paper. Gas Insulated Switchgear (GIS) were selected as research objects. The sound of GIS with different working conditions were captured. After preprocessing like framing and windowing, Mel frequency cepstral coefficient (MFCC) were used for audio feature extraction. For the sound characteristics of power equipment, the feature extraction method uses mid-time MFCC coefficients. Meanwhile, attention mechanism was introduced to improve the recognition performance of the algorithm. Then, Deep Neural Network (DNN) was used as the feature recognition algorithm. The research results show that this algorithm can effectively identify the operating status of power equipment. Based on the MFCC acoustic model, the F1 scores of GIS under different fault conditions generally exceed 90%. Compared with the original MFCC algorithm, the Attention-MFCC algorithm are more in line with the characteristics of smooth changes in the sound energy of the power equipment.