Controlling molten steel temperature in a tundish is crucial for stable and efficient production of steel products. In this research, a gray-box model is developed by combining a first-principle model and a statistical model. The first-principle model predicts the molten steel temperature in the tundish, and the statistical model predicts the prediction error of the first-principle model. The derived model was then used to derive the molten steel temperature in the Ruhrstahl-Heraeus process at the end of its operation from the target molten steel temperature in the tundish and other process variables. The result of applying the proposed model to real operation data has demonstrated its practicability.