Studies On Genetic Algorithm Based Optimal Control Strategies For Batch Processes

With the explosive growth of market economy, there is ever an increasing demand for small-lot, varied and high value-added products that produced in batch processes of fine chemistry and pharmaceutical industries. It is acknowledged that advancements of automation technologies in most batch fermentation processes are still relatively limited. In order to accommodate the markets and promote the competitiveness of the products, as well as reduce the cost while taking into account the environmental protection, it is in desperate need of advanced control strategies and optimization methods for batch production processes.Industrial batch fermentation processes usually suffer severe nonlinearity, non-steady-state and enormous process variables, which are somewhat different from those of continuous processes. Therefore, many of advanced control algorithms and strategies adapted to continuous processes are no longer suitable for batch processes. Genetic algorithm is recognized as a random search and optimization approach able to efficiently deal with nonlinear processes, as well as the characteristics of global optimization and easy implementation. Taking account of the complexity and paying close attention to the intermediate states, of batch fermentation processes, this thesis explicitly develop a kind of novel optimal control strategy based on genetic algorithms, which could effectively achieve optimal operating strategies for batch processes by solving a series of sub-optimal control problems with unfixed endpoint time. Moreover, to overcome premature or local-best solution, a kind of improved adaptive genetic algorithms is introduced and applied to the optimal control strategies.A specified penicillin fermentation process is employed in the studies. Firstly, an objective function taking fully account of additive impacts of average specific growth rates and time intervals of feed flow on desired yields is established. An in-depth investigation on corresponding optimal control strategies is performed, along with algorithms realization, eventually leading to optimal solutions of the flow rate trajectory over the whole fermentation processes. Secondly, the proposed approaches are applied to the penicillin fermentation process and compared with level feed flow rate strategies through simulation studies, demonstrating the satisfied feasibility. Meantime, the improved adaptive genetic algorithms are introduced to the optimal control strategies, showing improvements in terms of local optimization and convergence rate through simulation experiments. Finally, the thesis analyzes the implementation of control strategies and discusses the impacts of pH on the fermentation process, resulting in pilot experimental conclusions. Additionally, an experimental simulation platform of penicillin fermentation process is developed, which can be utilized to tune parameters of control strategies, perform process simulations, as well as monitor real-time process variables in terms of data values and curves.