The detection of partial discharge by analyzing the components of SF6 gas in gas-insulated switchgears is important to the diagnosis and assessment of the operational state of power equipment. A gas sensor based on anatase TiO2 is used to detect decomposed gases in SF6. In this paper, first-principle density functional theory calculations are adopted to analyze the adsorption of SO2, SOF2, and SO2F2, the primary decomposition by-products of SF6 under partial discharge, on anatase (101) and (001) surfaces. Simulation results show that the perfect anatase (001) surface has a stronger interaction with the three gases than that of anatase (101), and both surfaces are more sensitive and selective to SO2 than to SOF2 and SO2F2. The selection of a defect surface to SO2, SOF2, and SO2F2 differs from that of a perfect surface. This theoretical result is corroborated by the sensing experiment using a TiO2 nanotube array (TNTA) gas sensor. The calculated values are analyzed to explain the results of the Pt-doped TNTA gas sensor sensing experiment. The results imply that the deposited Pt nanoparticles on the surface increase the active sites of the surface and the gas molecules may decompose upon adsorption on the active sites.