In the continuous casting process, the temperature of liquid steel in tundish determines the casting speed and secondary cooling conditions, and then influences the billet quality. It's very important to measure the temperature of liquid steel in tundish quickly and accurately. However, the initial response lag of blackbody cavity sensor is inevitable since the time is required for the sensor inner wall and the liquid steel reaching thermal equilibrium by heat transfer. In this paper, in order to eliminate the initial response lag of sensor, a heat transfer model of sensor is established. The heat transfer characteristics and cavity integral emissivity of sensor with different depths immersed into liquid steel are analyzed. The analytical solution of sensor temperature is derived by separation of variables method and superposition principle, and is verified by the actual temperature measurement data. Then an innovative method of liquid steel temperature rapid identification is deduced and validated by the actual measurement data. The results show that the initial response lag of sensor is greatly shortened and the temperature measurement efficiency is improved. This study provides a theoretical method for improving the initial response speed of sensor.