This paper proposes a two degrees of freedom modified Smith predictor scheme for controlling the outlet temperature uniformity of a crude oil preheating furnace. A reliable mathematical model for describing the nominal process dynamics has been obtained from an identification procedure using realtime field data. This procedure yields a second order model with a dominant time-delay term (significantly higher than the time constant values). Moreover, disturbances have been identified as step responses of first order processes. A PI controller embedded in a modified Smith predictor scheme is therefore designed. A disturbance rejection term is designed for this scheme. This term is designed using a new methodology which reduces the effect of unknown disturbances on the process output. Simulations show that our proposal significantly outperforms the one degree of freedom Smith predictor and a well-known two degrees of freedom Smith predictor.