Temperature Difference Monitoring Method for Metal Devices of Electrical Equipment by Visible Images and Machine Learning

Abstract

In this paper, ordinary visible images were used to measure the temperature difference between two metal devices that were used in electrical equipment and transmission lines. The temperature-difference experiments of three kinds of metal devices used in 110kV transmission lines were conducted for collecting their visible images with varying temperature differences in a sunlight environment. The machine learning (ML) technique was applied for modeling the relationship functions between visible image features and temperature differences. The results showed that the corresponding temperature difference of metal devices in these images could be well predicted by trained ML models. Larger illuminance of sunlight helped to obtain higher prediction accuracy. This method could be used for monitoring the abnormal temperature rise of metal parts in electrical equipment by comparing the temperatures of two different phases. It would provide a new choice for intelligent temperature monitoring in power systems, helping save costs and improve inspection efficiency.