Abstract: In order to solve the problem of big fluctuation of molten steel temperature in ultra-low sulphur steel refining process，a temperature prediction model of molten steel from tapping to casting was developed using the MATLAB language based on an investigation of molten steel temperature variation of the total procedure converter → LF refining → RH refining→ wire feeding → casting. The molten steel temperature was predicted by the model，and the ratio of the average error within ±8℃ was 87.5%. The results showed that the molten steel temperature of LF-in could be about 1 580℃ if the following conditions were satisfied: maintenance time of ladle less than 1 hour，fast baking 10 minutes，temperature of tapping not lower than 1 650℃ and transport time less than 25 minutes. Molten steel heating rate was 2.0℃/min in LF early power supply and 3.0—3.5℃/min in LF late power supply. Moreover，RH vacuum chamber inner liner was more than 1 250—1 300℃ to reduce molten steel temperature drop. It could meet the demand of refining time of LF no longer than casting time under the premise of heavy desulfurization task-performing to make LF refining and casting match.

Key words: ultra-low sulphur steel, molten steel temperature, optimization measures