The research on near infrared spectroscopy of sodium in biological and medicine is significant. Sodion is the main component of electrolytes in human blood and electrolytes help maintain the body’s acid-base balance. In the present paper the concentration of sodium was determined with the use of NIR spectra. On the basis of NIR spectroscopic measurement mechanism of sodion, prediction models of the concentration of sodium were developed with linear regression using the absorbance at selected wavelengths. In order to reduce temperature perturbations to water bands with the measurement of sodium, Partial least squares regression (PLS) was adopted using select spectra area. The result shows that the determination coefficients (R2)=99.82%， the root mean square error of cross validation (RMSECV)=14.5, and the residual prediction deviation (RPD)=23.7, for the calibration model. It meets the daily requirements of biochemical detection accuracy. This technique can be applied to quantitative analysis of sodion in the hospital laboratory.