In order to enriching the means of discriminating alfalfa and achieving the goal of nondestructive testing, terahertz time-domain spectroscopy (THz-TDS) was applied to explore and reveal the property characteristic of different alfalfa varieties in the same series. Six kinds of alfalfa were prepared for experiment and analysis, and these samples are classed as two series, namely caoyuan series and gannong series. In the result, the time-domain spectra were tested, and then the refractive indices and absorption coefficients was calculated, respectively. These spectrums showed an apparent difference between these two series, and to verify this classification, two statistical methods, partial least squares (PLS) and cluster analysis (CA), were performed to investigate. Finally, these methods yielded a classification result, and we found it classified gannong series and caoyuan series. All these result showed THz technology combined with statistical method can be an effective method for nondestructive identification of alfalfa breed with tiny different properties, and lay a foundation for establishing a forage database.