Transformer voiceprint anomaly detection method based on multi-band self-supervision

Abstract

The invention relates to a transformer voiceprint anomaly detection method based on multi-band self-supervision. The method comprises the following steps: acquiring and preprocessing voiceprint data of a transformer; performing feature extraction on the collected voiceprint data; a self-supervised multilayer full-connection neural network model is designed, working voiceprint data features of the transformer are self-supervised and learned, and differences between the voiceprint data features are judged; in the self-supervised learning, when continuous voiceprint data features are input, the multi-layer full-connection neural network model outputs a continuous voiceprint data anomaly monitoring difference degree result; and identifying and detecting the abnormal frequency feature vector by using a template matching algorithm, and performing separate registration on the abnormal frequency feature vector to realize generalization identification and detection. According to the method, the key features in the transformer voiceprint are reserved, the parameters of the non-key frequency band are compressed, the feature vector dimension is reduced, the calculation burden is reduced, the system analysis speed is increased, the workload of professionals is greatly saved, and the precision deviation caused by manual intervention is reduced.

Description

technical field [0001] The invention relates to the technical field of power monitoring, in particular to a multi-band self-supervision-based method for abnormal detection of transformer voiceprints. Background technique [0002] The safety and stability of the power system is the basis of all activities in contemporary society, and plays a vital role in the normal operation of the city and the living security of residents. With the development of the country and the progress of society, the power load is further divided into industrial use Electricity, residential electricity, municipal electricity, etc., all kinds of social activities have further strengthened their dependence on the power system, and the demand for power consumption has further increased, which further increases the requirements for power supply quality, safety and reliability. In order to cope with this severe challenge, the construction of smart grid is imminent, which will not only help improve the lev...

AI Technical Summary

Problems solved by technology

The feature extraction process in the existing voiceprint anomaly detection technology usually uses Mel-frequency cepstral coefficients (MFCC) for voiceprint feature extraction to simulate the auditory characteristics of the human ear. However, this method is difficult to accurately represent the key frequencies in the transformer voiceprint detection task. characteristics, resulting in insufficient accuracy of the algorithm

In addition, due to the influence of load, voltage and other factors on the transformer, the working voiceprint of each transformer has its own characteristics. The existing technology requires separate parameter debugging for each device, and the detection accuracy is highly dependent on manual parameter settings, which consumes a lot of manpower. cost, and it is difficult to adjust in time when the working state of the transformer changes, resulting in low detection technology accuracy, poor robustness, and weak generalization ability

Images