Intelligent Predictive Maintenance of Dry-Type Transformer Based on Vibration

The normal operation of the transformer is an important guarantee for safe whole power network. Usually, periodic maintenance work increases workload and personnel. In this paper, intelligent predictive system dry-type based on vibration proposed to monitor state transformer. fractional-order Kalman filter normalization method are used preprocess data reduce noise interference, then fault classified by one-dimensional convolution neural On basis, improved grey Verhulst model established predict time experimental results show that classification accuracy network can reach more than 95%, one week in advance.