By detecting the stable by-products of SF6 through ultraviolet spectroscopy, the present paper achieved the rapid detection of the GIS partial discharge fault. First derivative and the S-G filter were used for the spectral denoising and smoothing. The discharge experiment was used for validating feature selection. Principal component regression was used for the analysis of the concentration of SO2. The concentration of SO2 was used for fuzzy judge. By selecting the appropriate wavelength range (295～305 nm), ultraviolet spectrum can identify SO2 from the complex by-products of SF6. In this paper, firstly, the author reviewed the decomposition mechanism of SF6 under partial discharge, and then verified the rationality of detecting partial discharge by UV, and ultimately achieved the rapid detection of GIS partial discharge and fuzzy judgment of discharge time.