This paper reports the usefulness of partial least squares (PLS) in the analysis of NIR spectra. Based on this work, the ethanol and methanol could be measured synchronously in the chemical industry. It is demonstrated that the proposed technique is quite convenient and efficient. PLS regression has been applied to establish a satisfactory calibration model via adopting optimum wavelength. In this model, the correlation coefficient in methanol determination is 0.99991 and the root mean square error of calibrate (RMSEC) is 0.431; in ethanol determination, the correlation coefficient is up to 0.99998, and the RMSEC is 0.193. Compared with the measured value of GC in the sample determination, the range of relative error is from 0.721% to 3.505%.