In this paper, a new strategy for robust control of temperature in a steel slab reheating furnace with large time delay uncertainty based on fractional-order controllers combined with a Smith predictor is proposed. A time delay model of the preheating zone of this process is used, obtained from an identification procedure applied in a real industrial furnace. It is shown that this process experiences very large time delay changes. A fractional-order integral controller embedded in a Smith predictor structure (FI-SP) is designed, which is robust to changes in such time delay. Simulated results of a standard Proportinal Integral Derivative (PID) controller, a PID controller embedded in a Smith predictor and the proposed FI-SP controller are compared. Six performance indexes have been used in this comparison. The analysis of these indexes shows that the designed FI-SP controller exhibits the most robust behavior (lowest indexes averaged in all the range of time delay variation) for ranges that include large time delays. Then the robustness features of the FI-SP controller outperform the other integer order controllers in the time responses both to set-point changes and to step disturbances. Therefore, this controller guarantees the best accuracy of temperature control. The designed FI-SP guarantees system stability and robust performance for a high range of plant time delay uncertainties.