In order to effectively prevent and control goaf fire in coal mines, a goaf fire early warning index system based on multi-parameter fusion was established. So that the temperature characteristic index for early fire warning in Goaf could be found, thermogravimetric tests and programmed heating tests must be used to see how often coal spontaneously burns. Additionally, combined the temperature characteristic index and the Graham coefficient yields the primary gas index. The auxiliary gas index, which exhibits a significant correlation with goaf fire, was calculated through the utilisation of the grey correlation methodology. Furthermore, the integration of the early warning temperature index and other gas indicators was accomplished through the use of the D-S evidence theory. Additionally, the outcome of the integration process involved the establishment of a multi-parameter integration-based system that served as an early fire warning signal for coal mining sites. The validation of the early warning index system's effectiveness was then conducted by a programmed heating test. The results suggested that the process of coal spontaneous combustion could be divided into six discrete phases, specifically the latent phase, oxidation phase, critical phase, pyrolysis phase, fission phase, and combustion phase. The identification of these stages was based on the distinctive temperatures associated with spontaneous combustion of coal, which have been observed at 31.7°C, 54.8°C, 153.7°C, 204.5°C, and 241.6°C. 100Δ(CO)/ΔO2(%) can be used as the principal indicator of gas within the Goaf region, hence facilitating the prompt identification of fires. The compounds C3H8/CH4, C3H8/C2H6, C2H4/C2H6, and C2H2 possess the capability to function as additional gas indicators for the timely identification of goaf fires. The utilisation of the D-S evidence theory in weight distribution provides support for the creation of an early warning index system for goaf fires, which relied on the fusion of many parameters. The reliability of the early warning index system was shown by means of a controlled heating experiment.