The multi-conformational structures sampling of a molecule, considered as conformational information, is found to be crucial for applying COSMO-SAC to systems containing compounds with flexible molecular structures, such as alkylene glycol derivatives. In this study, a hybrid modeling approach that combines the COSMO-SAC model with classical molecular dynamics (MD) is proposed to investigate the phase behavior of systems containing alkylene glycol derivatives. The accuracy in predicting liquid–liquid equilibrium (LLE) was found to be strongly influenced by conformational populations obtaining from MD simulations under different solvation environments. The overall root mean square deviation (RMSD) in predicting LLE mutual solubility from COSMO-SAC for seven selected mixtures containing alkylene glycol derivatives reduces from 19.1% to 10.1% when conformation distribution is included. The new approach significantly improves the accuracy in predicting phase behaviors for alkylene glycol derivatives solutions and can be applied to systems containing compounds with flexible molecular structures.