基于自适应进化多目标约束的青霉素发酵过程优化  

Optimization of penicillin fermentation process based on self-adaptive evolutionary constrained multi-objective approach

作　　者：王建林[1] 吴佳欢[1] 张超然[1] 于涛[1] 赵利强[1] 

机构地区：[1]北京化工大学信息科学与技术学院,北京100029

出　　处：《仪器仪表学报》2013年第12期2709-2714,共6页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61240047)资助项目

摘　　要：发酵过程多目标优化控制是提高发酵过程生产水平和经济效益的有效途径。提出了一种自适应进化多目标约束粒子群优化算法,并应用于青霉素分批补料发酵过程多目标优化。该算法根据不符合约束条件粒子的约束违反程度,修正了多目标粒子群优化算法的进化学习公式,提高了算法在约束边界区域的搜索能力;引入基于拥挤距离的Pareto最优解分布性动态维护策略,改进了Pareto前沿的分布性。实验结果表明,该算法能获得具有较好分布性的Pareto前沿,给出的底物补料策略能够使青霉素发酵过程在消耗更少底物的同时获得更多的产物产量,实现了发酵过程的多目标优化。The multi-objective optimization and control for fermentation process is an ettective way to improve me pro- duction level and economic benefit of fermentation process. A constrained multi-objective particle swarm optimization algorithm based on self-adaptive evolutionary is presented and applied in the multi-objective optimization of fed-batch penicillin fermentation process. The evolutionary learning formulas of the multi-objective particle swarm optimization algorithm are modified according to the constraint violation level of the particles that violate the constraints, and the algorithm search ability in the constraint boundary region is enhanced. Furthermore, a dynamic distribution mainte- nance strategy for the Pareto optimal solution based on the crowding distance is introduced to improve the distribution of the Pareto front. The experiment results show that the presented algorithm can obtain the Pareto front with better distribution, and give the substrate feed rate strategy that can make the penicillin fermentation process consume less substrate and produce more final penicillin product. The multi-objective optimization of the fermentation process is achieved.

关 键 词：多目标粒子群优化 自适应进化 拥挤距离 发酵过程 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]