In this paper, based on the electric arc furnace material balance theory and the method of bp neural network, the realtime forecasting model of carbon content of molten steel in electric arc furnace steelmaking is set up with theoretical model and neural network. Through the model draw curves of carbon content in the smelting process, the carbon content of molten steel in the realtime monitoring is implemented. Because of the endpoint carbonoxygen equilibrium in decarburization reaction, the influencing factors of end point carbon content are analyzed. The forecast method based on BP neural network has reached accuracy requirements of EAF endpoint carbon content prediction.