The quantitative analysis of the performance indexes of ethanol diesel was carried out by means of medium infrared spectroscopy and chemometrics. There were 96 samples in 32 different concentrations of ethanol diesel oil solutions. Using S-G, MSC, smooth differential processing (1st D and 2nd D), SNV of methods preprocessed spectral data, combined with the screening method of eight kinds (UVE, CARS,SPA,RPLS,UVE-SPA,UVE-CARS,SPA-CARS,UVE-SPA-CARS)processing ethanol diesel MIR spectral data, and PLSR model were established respectively with density of ethanol diesel oil, viscosity and ethanol content. The results showed that: comparing the eight variables screening methods, we found that UVE-SPA-CARS-PLS has the best modeling effect on ethanol content. The correlation coefficient was 0.978 1 and the root mean square error of prediction was 0.825 5, respectively. Compared with the model established by the original spectrum, variable selection not only reduced the number of model inputs, but also improved the prediction effect.