In order to determine the vintages of the sesame flavor liquors, the three-dimensional fluorescence spectra of 11 aged liquor samples brewed from 1984 to 2012 are measured with an FLS920 multifunctional fluorescence spectrometer. Average fluorescence intensity, standard deviation of numerical matrix and maximal relative fluorescence intensity of the three-dimensional fluorescence spectra are extracted for polynomial fitting. Two prediction models for the vintages of aged liquors are established with partial least squares regression calibration and quadratic optimal regression calibration combined, respectively. The vintages of aged liquors of 1989, 1994, 1996, 2002, 2003, 2005, 2006, 2009, 2011 are predicted by the two models above, and the average absolute errors are 0.68 and 0.56, respectively. The results show that the established models can effectively determine the vintages of aged liquors. The nonlinear calibration model has better precision than the linear calibration model. The aged liquors are studied by three-dimensional fluorescence spectra combined with regression-calibrated polynomial functions, which paves a way for the determination of vintages of aged liquors.