Research On Optimal Control For The Fermentation Process

Aiming at the complexity of the glutamic acid fermentation process, a neural network dynamic model of the fermentation process is established. The improved differential evolutionary algorithm (DEA) is used to the multi-variable optimal control of the fermentation process and the optimal control trajectories of operating variables are found out. Some improvements of the primitive DEA are made by the means of randomly selecting the mutation factor and the re-initialization of the individuals in the population on a suitable time, so that it can solve the constrained optimization effectively and avoid the problem caused by premature. Simulation results show the effectiveness of the method.Based on the same model, for the purpose of optimizing the conversion rate, the real-coded genetic algorithm (RCGA) is used to the multi-variable optimal control of the fermentation process, and the optimal control trajectories of operating variables are found out by the means of the whole process optimization. Considering the importance of fed-batch in the fermentation process, the start and end times of this operation are used as variables. Simulation results of comparison with the glutamic acid concentration optimization show that this method makes the conversion rate improved a great deal, and the glutamic acid concentration is closed to the latter.Then,multi-objective and multi-variable optimization is achieved aiming at the objectives of conversion rate and concentration in the glutamic acid fermentation process. According to the actual situation, the entire fermentation process is divided into two phases. The optimization in the first phase is a multi-variable optimization whose objective is just the concentration of the glutamic acid. The second stage's optimization is a multi-objective and multi-variable optimization, and the objectives are conversion rate and concentration of the glutamic acid. The improved Differential Evolutionary Algorithm is used into the first phase optimization. The non-dominated sorting and the Niche technique are introduced into the improved DEA, and then it is used in the multi-objective constrained optimization. Simulation results show that the modified DEA can effectively meet the needs of the two-stage fermentation process optimization control.Finally, based on the memory of this evolutionary algorithm,the research on dynamic optimal control methods is achieved. Two types of memory strategy to replace the individuals are used:(1)Identifying and replacing the worst individuals (2) Finding out the most similar individual with the optimal solution, it will be replaced if the fitness is worse than the optimal solution.