Abstract: In the biochemical process, such as ethanol fermentation, some problems such as high nonlinearity and poor stability cause more difficulties on the optimal control of the fed-batch process. In order to solve these problems and improve the optimization efficiency as well as maximizing the product concentration, rolling optimization strategy is proposed based on the core idea of predictive control. In the optimization process, the penalty function method was used to transform the original optimization problem with constraints into an unconstrained optimization problem. And hybrid optimization algorithm combining ant colony algorithm and iterative dynamic programming has been applied into the substrate flow rate control trajectory optimization. Compare this new optimization algorithm with ant colony algorithm, the simulation results show that some performances such as optimization speed, optimization performance have been improved greatly.

Key words: Receding horizon optimization, model predictive control, optimal control, ant colony algorithm, iterative dynamic programming

摘要： 针对非线性程度高、稳定性差的生化过程典型应用案例—酒精补料分批发酵过程的优化控制问题，从提高寻优效率和最大化产物浓度出发，提出了一种基于预测控制中的核心思想—滚动优化策略，利用罚函数法将原带有约束的优化问题转化为无约束的优化问题，从而通过蚁群算法与迭代动态规划相结合的混合优化算法来进行基质流加率控制轨迹动态优化。并将这种优化策略应用的结果与蚁群算法寻优效果比较，仿真结果显示，无论从寻优速度、优化性能等方面都有了较大的改善。

关键词: 滚动时域优化, 模型预测控制, 优化控制, 蚁群算法, 迭代动态规划