Abstract: In view of the high price and low detection accuracy of biosensors, which makes it difficult to obtain accurate and real-time biological parameters in the process of GlcN fermentation, the Least Square Support Vector Machine (LSSVM) model is established to predict the cell concentration, product concentration and substrate concentration. In order to improve the accuracy of the prediction model, the improved multiverse optimization algorithm based on Levy flight is utilized to optimize several parameters of the LSSVM model. On the basis of the model aiming at the maximum product concentration at the time of fermentation completion, the fermentation process parameters are optimized by the improved multiverse optimization algorithm. The simulation results show that the method achieves higher modeling accuracy and improves the final fermentation product concentration.

Key words: GlcN, least square support vector machine, multiverse optimization algorithm, fermentation process modeling, process parameter optimization