Research on Fault Detection Method of Infrared Thermal Imaging for Power Equipment based on Deep Learning

Abstract

Abstract With the emergence of large image sets and the rapid development of computer hardware, especially GPU, it is a challenging problem to deploy convolution neural network (CNN) model on embedded devices with limited computing resources. The overheating fault of power equipment can be identified by collected infrared thermal imaging. Due to the propagation and fading of infrared radiation in the air, the infrared temperature measurement result is lower than the actual temperature value. In this paper, an efficient convolution neural network based on embedded equipment is proposed for thermal fault detection of power equipment. The backbone network in SSD algorithm is replaced by MobileNet, and Batch Normalization is merged with the previous convolution layer, so as to reduce model parameters, improve reasoning speed and make it run on lightweight computing platform. In order to solve the problem of propagation loss of infrared radiation in the air, an infrared temperature measurement correction unit based on BP neural network is proposed. Based on the above innovation, a thermal fault detection system for power equipment is designed. experiments and field applications show that the method has high accuracy and reasoning speed.