【作者】 李玮；

【导师】 冯冬青；

【作者基本信息】 郑州大学 ， 检测技术与自动化装置， 2006， 硕士

【摘要】 间歇生产过程是以顺序的操作步骤进行批量产品生产的过程，广泛应用于精细化工、药品生产、生物制品、现代农业等领域，并随着工业生产柔性化的趋势和市场对产品多样化的要求，受到越来越多的重视。传统的间歇生产过程中人工操作占很大的比重，自动化水平普遍较低，因此，迫切需要运用先进的控制策略和优化方法来提高生产效率、节约生产成本。 由于许多间歇过程单元存在非线性、大滞后、时变和数学模型不确定等特征，采取传统的PID控制，难以达到理想的控制效果；同时，又由于间歇过程的运行一般是在没有稳态工作点的过渡状态下进行，其控制和优化问题十分复杂，采用简单的智能控制策略效果欠佳。因此非常有必要研究新的智能控制策略。考虑到间歇过程具有一个鲜明的特点，即过程运行是分批重复进行的，且每次的运行时间有限，这恰好与迭代学习控制(ILC)的适用特征相吻合。但是，传统的ILC是针对单输入单输出系统设计的，而且对过程控制中经常遇到的约束、耦合等问题的求解并不太适合。而模糊模型、预测控制在这些问题上具有各自的优势。因此，本课题在对上述三种方法进行综合应用并加以改进的基础上，给出了一种预测迭代学习新算法；研究了一种在无模型和无先验知识的情况下设计基于模糊预测迭代学习控制器的方案。编制了实现这种控制器算法的程序，通过仿真研究，验证了该控制器的性能。最后，考虑到农药生产属于典型的间歇过程，因此将本课题的研究结果应用于氧乐果合成反应温度过程的控制，较好的满足了控制要求。 本课题主要工作和研究内容如下： (1)分析了模糊控制、预测控制和迭代学习控制的发展现状。 (2)研究了模型预测控制的基本原理，深入分析了T-S模糊模型辨识方法，并将其与预测控制相结合，给出了模糊预测控制的两种结构。 (3)研究了迭代学习控制的基本原理，针对工业间歇过程常用的反馈-前馈迭代学习控制，分析比较了已有的两种反馈-前馈迭代学习控制存在的缺陷，在此基础上，研究了预测控制与迭代学习控制的结合技术，在迭代学习控制中引入预测的思想，给出了一种改进的迭代学习算法。 (4)将模糊模型辨识技术、预测控制和迭代学习控制三者相结合，设计了一种新的基于模糊预测的迭代学习控制器。 (5)以参数时变时滞后过程为被控对象，仿真研究了本文设计的控制器性能。更多还原

【Abstract】 Batch processes are batch-production processes following sequential operation steps. They are widely used in industrial domains such as fine chemical, pharmaceutical producing, biology engineering, modern agriculture etc. With the flexible trend of industrial manufacture and various requirements of market on products, they are paid more and more attention. Human operation dominates in traditional batch processes, so the corresponding automation is generally lack. As a result, advanced control strategy and optimization method are urgently required to develop productive efficiency and save productive cost.Because non-linearity, large time lag, time-varying, non-accurate mathematic model etc. exist in many batch process units, PID control method can not do as well as expected. Besides, as these batch processes always run under transitional condition without stability, their control and optimization are quite complex, as a result, simple intelligent control algorithm can not do well either. So, new suitable intelligent control strategy must be researched. The particular character of batch operations is that they are repetitive and errors in one batch are likely to repeat in the subsequent ones, which coincides with the application feature of iterative learning control (ILC). But the traditional ILC is only designed for SISO systems, furthermore, it’s not quite suitable for solving problems such as constraint, coupling etc, which are frequently encountered in process control. Meanwhile, fuzzy model and predictive control have their respective advantages in the problems above. Based on the combination and improvement of the three kinds of method above, a new algorithm of predictive iterative learning control is introduced in this paper. Moreover, a new method is given that the iterative learning controller based on fuzzy prediction can be designed without mathematic model and prior experience. In this paper, programs for the control algorithm are given, and the performance of the controller is tested by corresponding simulation. Finally, considering that pesticides production is a typical batch process, the controller is applied to omethoate synthesis temperature process control. The simulation result shows that it meets proposed standard better.The main contents are as follow: (1) The present condition of fuzzy control, predictive control and iterative learning更多还原