Substation equipment abnormal detection based on acoustic fingerprint learning

Predicting and evaluating the operation status of the equipment online can not only reflect the personalization of the equipment but also meet the actual working needs of intelligent substations. Through acoustic fingerprint learning, potential defects and hidden dangers can be identified, and fault handling and emergency repair time can be shortened. This article proposed a convolutional neural network to learn the acoustic fingerprint of substation equipment and discussed the feature selection and feature preprocessing of the proposed machine learning model. We then conducted a simulation experiment and analyzed the right parameters selection for the proposed model. The experiment results show that the proposed model can achieve a recognition accuracy of above 90% on all the different abnormal voiceprint test sets. The recognition results showed the effectiveness of the voiceprint recognition model and can thus provide a solid guarantee for the safe and stable operation of the power system.