The extractive distillation process is gradually being used for the desulfurization of fluid catalytic cracking (FCC) naphtha because of its excellent octane number protection performance during desulfurization. The key to this process is the separation of sulfides. However, the composition of FCC naphtha is complicated and easily changes, which seriously affect the separation performance. The operating conditions need to be adjusted frequently with the change in the different properties of feedstocks, which cost time, manpower, and investment. In this study, we introduce a simulation method using the universal quasi-chemical functional group activity coefficient (UNIFAC) model to predict the process of desulfurization. In the simulation, the group binary parameters for the UNIFAC model seriously affect the reliability of the method. Therefore, the lacking parameters of the solvent and sulfides were first regressively calculated by the experimental results of the binary vapor−liquid equilibrium (VLE). All the values of (D−J) by the Herington integral method were under 10, which means that the parameters obtained in this work were well correlated to those of the VLE experiments. Based on this, the prediction simulation method was established and the maximum error of the compositions of the products between simulation and real-oil experiment was only 0.5 wt %. According to the simulation, 98% of sulfides could be extracted under the optimal conditions (113 °C at the extraction column, 178 °C at the solvent recycle column, etc.). The rule of the extraction of sulfides was thiophene sulfur > mercaptan sulfur > thioether sulfur, which is consistent with the VLE study results. The sulfur content of the raffinate oil was decreased to 9.3 mg/kg. It well met the latest gasoline standard and is appropriate for the prediction of the desulfurization of FCC naphtha by the extractive distillation process.