Abstract: Taking 120 t LF furnace as the research object, a new temperature prediction and control model combining mechanism model and linear regression method was designed, to address the insufficient control accuracy of molten steel temperature and hysteretic nature of manual adjustment during LF refining process. This model aimed to achieve high-precision prediction and dynamic control of molten steel temperature, in order to meet the strict requirements for temperature stability in modern steelmaking production. In the model design, the mechanism model was based on thermodynamic principles and constructed a theoretical framework for the change of molten steel temperature. Meanwhile, the linear regression model made full use of historical operation data to extract the linear relationship between key operation parameters and molten steel temperature. The research results showed that this composite model could predict the changes in molten steel temperature in real time and adjust the operation parameters dynamically according to the prediction results, thereby achieving precise temperature control. In actual production tests, when the operating conditions of the model were met and the production process was stable, the hit rate of the end-point temperature prediction error within ±5 ℃ exceeded 95%. Moreover, the model had been successfully integrated into the automatic control system of LF furnace, achieving the automatic regulation and closed-loop control of temperature. Combined with modern control technology, this system could automatically optimize operation parameters based on real-time data, reducing the uncertainty and hysteretic nature of manual intervention. Practical applications had shown that the model not only improved the stability of temperature control, but also reduced energy consumption and production costs, providing strong support for the intelligent production of LF furnaces. In the future, the model was expected to be further promoted to other types of refining furnaces, providing technical references for the intelligent transformation of steel industry.