Raman spectroscopy has the advantages of high resolution, fast analysis, simple sample preparation, nondestructive, online measurement, etc. It is widely used to analyze the composition and molecular structure information of a wide range of organic and inorganic substances for qualitative and quantitative analytical measurement. At present, it is mainly used for quantitative analysis. When used for quantitative analysis, the poor reproducibility and the imperfect analysis theory of Raman spectroscopy are two main factors that limit its application. In the quantitative analysis based on intensity ratios, the emergence of Raman intensity normalization theory has provided a theoretical basis for its application. The choice of the reference peak/internal standard peak and the fitting method have a great influence on the measurement accuracy and stability. In this paper, the relative intensities of the characteristic peaks (C-C-O symmetrical stretching, 874 cm-1) of the Raman spectra of ethanol and other reference/internal standard peaks with different ethanol concentrations were investigated using a laser Raman system. A peak ratio method based on ethanol intrinsic peak and an internal standard method based on the characteristic peak position of CCl4 were developed, and both methods can effectively eliminate the effects of mutation noise and strong fluorescence background in the system through normalization. The reference peak/internal standard peak with the best accuracy and stability of the two methods was determined by statistical methods such as joint hypothesis testing for data differences within and between different groups. The F-test and t-test showed that the standard curve established with the characteristic peak at 1 446 cm-1 (CH3-asymmetric deformation) as the reference peak could more accurately invert the concentration of ethanol when the calibration was performed by the self-peak ratio method, while the standard curve with the Raman characteristic peak at 446 cm-1 as the internal standard peak had higher stability and accuracy when CCl4 was used as the internal standard. The retest within 30 days eliminates the need to measure and plot the standard curve again. The linear regression model established according to the two calibration methods can provide an experimental basis for the quantitative analysis of ethanol solution concentration. The ethanol concentration can be inverted more accurately in real-time, through the application of the model in the ethanol solution concentration detection system, to achieve accurate, rapid and real-time quantitative analysis of ethanol solutions in the high concentration range with strong fluorescence background interference.