Abstract: In view of the importance of heat state judgment for calcium carbide smelting, the concept of furnace heat index was presented by analyzing the smelting features. A calculation model of furnace heat index was established based on the two-stage thermal equilibrium, and a prediction model of hot calcium carbide temperature was constructed by using a BP neural network. Both the models can effectively judge the furnace heat state. Simulation results show that there is a significant linear correlation between hot calcium carbide temperature and heat surplus, and it is feasible to consider furnace heat index as a heat state's sign. The hit rate to hot calcium carbide temperature predicted by the prediction model reaches 86.7%.