Due to high stress, high ground temperature, high moisture, and other factors in deep mines, the risk of coal spontaneous combustion (CSC) is enhanced, seriously affecting the safety of coal mining. To achieve early prediction of spontaneous combustion in the No. 3 coal seam at the Juye coalfield in the deep mine, this paper employs a temperature-programmed device to analyze the changing pattern of single-index gases and composite gas indices with temperature derived from the gas produced during csc. It also optimizes the index gas of coal sample spontaneous combustion. Simultaneously, the characteristics of coal temperature and a four-level warning indicator system for CSC are determined based on the analysis of indicator gas growth rate method, carbon-to-oxygen ratio, and the characteristics of the indicator gas. The composite index gases of the No. 3 coal seam in Juye coalfield are selected in the initial oxidation stage (Rco), accelerated oxidation stage (R1, G1), intense oxidation stage (R2, G1, G3), and oxidative decomposition stage (G3). This leads to the construction of a six-level warning system consisting of initial warning value, blue, yellow, orange, red, and black levels. Meanwhile, warning thresholds are also established.