Research on the Inversion of Hot Spot Temperature in GIS Equipment Considering Multiphysics Simulation Data

This study develops a multi-physics simulation model for electromagnetic, flow, and temperature fields based on a prototype three-phase Gas-Insulated Switchgear (GIS), and establishes a temperature rise test platform for the GIS busbar. Machine learning algorithms are used to train simulation data and invert the internal hotspot temperature. External temperature data from the test are used as inputs to the inversion network, and the inverted values are compared with the measured temperatures to validate accuracy. Additionally, the impact of different combinations of temperature measurement points is analyzed, leading to an optimized measurement layout for practical application. The study reveals that axial temperature measurement points on the GIS enclosure exert a more pronounced influence on the inversion accuracy compared to radial points, with the upper axial sensors demonstrating the most substantial impact.