Abstract: Measurement of calorific value using oxygen bomb calorimeter is time consuming.Visible-near infrared spectra technology was employed to collect spectral characteristic of agricultural crop residues.Differential spectral pretreatments such as smoothing，derivation，multiplicative scatter correction（MSC）and standard normal variate（SNV）were used to improve predicting models of calorific value based on partial least squares regression（PLR）and principal component regression（PCR），respectively.Results showed that the model pretreated with 10-point smoothing exhibited the highest correlation coefficient（R2）and root mean square error of prediction（RMSEP）of 0.853 7and 0.443 4，respectively.Predicting model based on PLR exhibited relative higher estimating accuracy than that of PCR.Also，PLR showed more stability in prediction than that of PCR.Excessive smoothing was observed when more than 15 points smoothened.Derivation and MSC pretreatments showed no enhancement for the model’s estimating performance.Results provided basic parameters and theoretical model for inventing biomass rapid GCV detection device.