Based on a self-built Raman scanning system and the SERS substrate named SC silver sol which was prepared with sodium citrate reduction method, a quantitative prediction model of potassium sorbate in osmanthus wine was established. The 34 osmanthus wine samples were divided into calibration set and validation set, and the potassium concentration prediction model was established by multiple linear regression method. The determination coefficient and the root mean square error of the calibration set were 0.978 9 and 0.070 3 g·kg-1 respectively, the determination coefficient and the root mean square error of the validation set were 0.934 and 0.165 7 g·kg-1 respectively. The quantitative prediction model of potassium sorbate in osmanthus wine as the main spectrum model and by using the DS algorithm and the PDS algorithm, discussed the model transfer method of potassium sorbateintwo different wines. The K/S algorithm was used to sort the Raman spectra of bayberry wine. The main spectrum correction matrix was composed of osmanthus wine samples which had the same concentration ascalibration matrix. 15 bayberry wine samples were prepared to verify the effect of the transfer model. The results of the DS algorithm showed that RP and RMSEP were 0.906 1 and 0.215 0 g·kg-1 respectively. The results of the PDS algorithm showed that RP and RMSEP were 0.905 5 and 0.225 g·kg-1 respectively. DS algorithm and PDS algorithm can be achieved with a small number of samples for effective model transfer, and the best samples of the two methods were 4 and 3 respectively. In addition, window width of 5 was the best choice of PDS algorithm. Prediction model of potassium sorbate in osmanthus wine was suitable for the quantitative prediction of potassium sorbate in red bayberry wine by DS algorithm or PDS algorithm. The DS algorithm and PDS algorithm can achieve the transfer of prediction model for potassium sorbate in different wines. Potassium sorbatein red bayberry wine can be predicted by the prediction model of osmanthus wine.