In this study, terahertz time-domain spectroscopy (THz-TDS) and multivariate statistical methods were used to demonstrate the feasibility of identifying fourteen alfalfa forage varieties that look extremely similar. THz spectra parameters, such as refractive index and absorption coefficient, were calculated from 0.1 to -1.5 THz, and the test spectrum revealed that different kinds of alfalfa grass seeds are different in time delay, absorption intensity and average refractive index. Although these characteristics differences mentioned above mean that the THz-TDS are feasible to identify alfalfa forage varieties, the statistical methods, including cluster analysis (CA) and principal component analysis (PCA), were used to build models between THz parameters and different alfalfa forage varieties because there was no characteristics absorption peak as fingerprint identification basis. The Euclidean distances of CA between forage grasses, and the scores of the first principal component (PC1) in PCA method reflect the forage-dependent differences, indicating the consistency between CA and PCA. Consequently, the combination of THz technology and statistical methods can be an effective method for the rapid identification of alfalfa forage with different properties. Furthermore, this combination method also provides a favorable basis for establishing the THz spectrum database of forage species in the future.