Dry-Type Transformer Fault Warning Based on Infrared Thermal Images

Dry-Type transformer is the core equipment of urban rail traction power supply system, its running state is directly related to the stability of the whole power supply system. Through infrared images, the overheating faults of a transformer can be directly reflected. In this paper, we first analyze and simulate four typical overheating faults: inter-turn short circuit, air duct blockage, long-term overload and terminal overheating, and then the infrared image dataset of faults is established. Next, a fault recognition model based on AlexNet is established. In contrast to the original AlexNet algorithm, batch normalization (BN) layer is added to reduce overfitting, and the number of parameters of the model is reduced by improving the convolution kernel parameters. The experimental results show that the number of AlexNet parameters is reduced by 83.99%, while the average classification accuracy reaches 95.41%, which is significantly higher than the original AlexNet algorithm.