As one of the main routes of vaccine production, cell suspension culture technology has become an inevitable trend in the development of biopharmaceutical industry. Its main advantage is that it can maximize the product quality while achieving high yield through accurate and effective process control means. As the basic material needed in the process of cell suspension culture, cell culture medium mainly provides energy through substances such as glucose. In order to obtain high-quality and high-performance cell products, it is necessary to continuously monitor the culture process. Traditional biochemical method can achieve the purpose of monitoring, but the detection cycle is long which making it hard to meet the needs of real-time monitoring. In contrast, Near-infrared (NIR) spectroscopy can detect glucose concentration in cell culture medium in real time without loss and it has the advantages of fast, accurate, and high sensitivity. In this study, a partial least squares regression (PLSR) model was established based on the near infrared spectrum of cell medium containing a certain amount of glucose to verify its validity and feasibility. Efforts have been made on the processing of sample outlier, spectrum preprocessing and wavelength optimization method. And good prediction results were obtained with the coefficient of determination (R2) of the model was 0.998, the cross-validation mean square error (RMSECV) was 0.0036 and the residual prediction deviation (RPD) value was 4.15. On this basis, in order to further improve the prediction accuracy of the model, this paper introduced glutamine variables to establish a multi-component fusion model and further optimized the combined wavelength selection method. The results showed that the model could achieve better prediction accuracy when using our proposed optimization method which indicated that the method of real-time detection of glucose concentration in cell medium based on near infrared spectroscopy was feasible.