Objective To analyze the correlation between the structure of free fatty acid (FFA) and its retention time (RT) and establish and verify a RT prediction model. Methods The RT and structural information of 72 standard samples of FFAs were detected by UPLC-QTOF/MS platform. The correlation between RT and the length, number of double bonds, and locations of double bonds of the carbon chain of FFAs was investigated. The multiple generalized regression models were established and the optimal model was selected by cross-validation and Akaike information criterion. Results The structural information was nonlinear with RT. The optimal prediction model was established and the performance was evaluated. Conclusion The multi-element orthogonal polynomical regression model can well predict the RT of 72 standard samples of FFAs and can be used to predict the RT of other target substances. It is helpful for identifying FFAs without standard samples.