Sixty alfalfa samples, with different growth stage, cultivars, and preparing method (drying by oven, shade and sun), were selected to study the potential of determination of cellulose, hemicellulose and lignin content in the present research. The result showed that the correlation coefficient of cross-validation (R(cv)), determination coefficient of external validation(r2) and the ratio of standard deviation to root mean square error of prediction (RPD) of cellulose and lignin were 0.97, 0.97 and 4.44, and 0.94, 0.94 and 4.08, respectively. This indicated the feasibility of determining cellulose and lignin content of alfalfa using near infrared reflectance spectroscopy. Hemicellulose was not predicted accurately by NIRS in this study, due to the lowest accuracy (R(cv) = 0.39, r2 = 0.29, RPD = 1.09). The exact determination of cellulose and lignin using near infrared reflectance spectroscopy will be useful to quality control in alfalfa production and quickly analyzing the fiber composition of alfalfa samples breeding research.