Abstract: Glucose concentration is an important parameter in biological fermentation processes. Precise control of glucose concentration can optimize fermentation conditions, improve production efficiency, and enhance product quality. However, in most cases, due to the presence of various substances in the fermentation broth, the Raman characteristic peaks of glucose in the fermentation broth are difficult to distinguish. For the quantitative analysis of glucose under such circumstances, the Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS) algorithm is first used to separate the Raman spectra of the target substance glucose from the matrix spectra of the fermentation broth. Then, the Competitive Adaptive Reweighted Sampling (CARS) algorithm is employed for dimensionality reduction, and Partial Least Squares (PLS) algorithm is used for concentration modeling. The results show that using this method for modeling significantly improves the predictive performance, with the R2 of the test set increasing from 0.86026 to 0.98031, and the root mean square error (RMSE) decreasing from 0.00183 to 0.00061.