Abstract: For the determination of the net calorific value of clean coal,the near-infrared diffuse reflection spectrums of 150 clean coal samples are collected. After the rejection of the abnormal samples,a Markov distance-based guantitative mathematical model is built through analysis of principal components and using different spectral preprocessing methods. Comparison with the industrial testing result indicates that an optimum modeling result can be obtained through multiple scattering correction. The correlation coefficient is found to be up to 0. 909 while the root-mean-square error of calibration and root-meansquare error of cross validation are 0. 001 31 and 0. 001 62 respectively. Then,through the reduction of the dimension of the spectral data using PCA method,the main components of the first three related samples are extracted. It is found that the contribution rate of the cumulative variance is as high as 93. 786%,well demonstrating the high reliability and prediction ability of the model. This provides an effective data processing method for the determination of net calorific value of clean coal through near-infrared diffuse reflection spectroscopic analysis.