Hybrid Modeling And Optimization Of GlcN Fermentation Process

Glucosamine(GlcN)is mainly used in biomedicine,food health care and cosmetics.In Europe and the United States and other developed regions,GlcN and its derivatives are widely used as dietary supplements.With the improvement of domestic living standards,the demand for GlcN products has been increasing.How to improve the product quality and concentration through the optimization control of fermentation process,so as to win the recognition of domestic and foreign markets,is the primary consideration of GlcN fermentation enterprises.The key of real time and accurate acquisition of fermentation process information is to realize the optimal control of fermentation process.Due to the lack of reliable biosensors for on-line measurement of key biological state parameters,it has become an important premise to find a suitable modeling method for the prediction of fermentation process biomass.This paper focuses on the problem of hybrid modeling and optimization of GlcN fermentation process,establishing a hybrid model of mechanism knowledge and improved Least Squares Support Vector Machine(LSSVM),and investigating the optimal control of feeding speed of fermentation process parameters.The main research contents are as follows:(1)The reaction mechanism of GlcN fermentation and the main factors affecting the fermentation process were studied.Based on the fermentation kinetics of GlcN,the kinetic equation of fermentation process was established,including the cell growth model,substrate consumption model and product generation model.Aiming at the parameter optimization of LSSVM model,an improved multiverse algorithm is proposed.By introducing Levy flight and adaptive adjustment strategy,the algorithm improves the shortcomings of multiverse algorithm,such as slow convergence speed,easy to fall into local optimum,poor overall stability and so on.The simulation results show that the improved multiverse algorithm is superior to the multiverse algorithm in convergence speed,optimization accuracy and stability.(2)The improved multiverse algorithm was applied to optimize the parameters of LSSVM model.Combined with the mechanism model,a hybrid model was established to predict the biological state parameters in the process of GlcN fermentation: cell concentration,substrate concentration and product concentration.In the selection of input data,principal component analysis algorithm is used to reduce dimension.The simulation results show that the hybrid model can obtain better prediction results than the pure mechanism model.Based on the hybrid model,Quantum Particle Swarm Optimization(QPSO)algorithm is used to solve the optimal feeding speed control curve.The simulation results show that the product concentration is increased by 5.8%.(3)A set of data acquisition and analysis system for GlcN fermentation process has been designed and developed by integrating C# programming technology,Oracle database and MATLAB software.The system can easily retrieve the real-time and historical data,and provide graphical analysis function.While collecting data,the system is connected to Matlab software for simulation and prediction to guide enterprise fermentation production.In this paper,the improved multiverse algorithm and QPSO algorithm are applied to the study of hybrid modeling and optimization of GlcN fermentation process.The simulation results demonstrate its advantages,and it has been effectively applied in a GlcN fermentation enterprise in Shandong Province,which provides the prediction of key biological state parameters for the production process and improves the quality of GlcN fermentation products.