Solvent extraction is an energy-efficient process to treat phenolic effluents in the industry, and a key step in designing an industrial extraction process is to screen a proper extraction solvent with high extraction efficiency and good physical properties. In this work, COSMO-SAC model was employed to screen the most promising extractant from 40 organic solvents, including alkanes, arenes, ethers, esters and ketones. The screening was performed based on a comparison of selectivity and solvent power, which were derived from the activity coefficient at infinite dilution. Moreover, the a-profiles of the solvents were used to analyze the interaction between solvents and phenol. Based on the results of screening, three ketones were selected for conducting LLE experiment, and all of them performed very well with high distribution coefficient and high selectivity. The NRTL and UNIQUAC models were successfully applied to correlate the experimental LLE data, with root mean square deviation less than 1.5%. The COSMO-SAC was also used to predict the tie-line data, showing quite good agreement with corresponding experimental data. Finally, the extraction process simulation was performed for the screened solvents. It showed that, the studied ketones are promising solvents for extracting phenol from wastewater. The extraction process treating an effluent with phenol concentration of 5000 ppm was simulated. High separation efficiency (the phenol concentration in the treated water < 10 ppm) can be reached with low stage number (e.g. 4) and solvent usage (e.g. extractant: wastewater = 1:25). (C) 2017 Elsevier B.V. All rights reserved.