In this research, the terahertz spectra data of the aging wheat processed under manual control environment by ATR accessory were collected. After the data diversity based on the composite score by PCA, the non-destructive identification models of aging wheat were developed by PLS-DA algorithm. The results showed that for the absorption coefficient spectrum, the accuracy of the experimental group, control group of the calibration set and cross validation set were 84.2%, 94.7%, 84.2% and 81.6% respectively, while the accuracy of the experimental group and control group of the external validation set were 73.7% and 100.0% respectively; for the refractive index spectrum, the accuracy of the experimental group, control group of the calibration set and cross validation set were 84.2%, 92.0%, 76.3% and 76.3% respectively, while the accuracy of the experimental group and control group of the external validation set were 84.2% and 89.5% respectively. The research indicates that ATR-THz technology should be of great potentials in the application in the non-destructive identification of aging wheat.