【作者】 杨国军；

【导师】 钱宇；

【作者基本信息】 华南理工大学 ， 化学工程， 2013， 博士

【摘要】 间歇生产过程已被广泛应用于生产高附加值产品,如药物、生物材料、聚合物、电化学品等。然而在实际生产过程中，由于间歇过程本身的动态特性以及在同一设备上运行不同的批次所产生的各批次之间的操作条件的变化会导致产品质量不高或重复性差等问题出现，甚至严重的会导致安全事故的发生，实时优化技术及先进控制技术的发展为解决这些问题带来了契机。针对实际操作过程中进料的变化、过程参数的不确定性以及各种扰动作用，过程操作需要进行实时的调整以保持过程的稳定性及性能，实时优化技术通过在线监测过程参数和产品质量参数的变化而进行实时的寻找过程最优操作点或轨迹，并控制过程在最优目标下运行，从而保证产品质量的最优与稳定，因此研究间歇过程的实时优化技术是产品质量控制研究领域的一个重要研究课题。然而实时优化技术是一个集成概念，其不仅涉及到多个技术的应用，还涉及到不同尺度的信息之间的交互融合，它主要包括动态模型建立、动态优化、在线监测、模型更新和在线控制等技术，其中最基本也是最主要的几个关键技术即是动态建模技术、动态优化算法及在线控制技术等。本研究以基础技术应用为主，研究开发新型实用的间歇过程建模和仿真方法、动态优化技术及非线性模型预测控制技术，为进一步实现间歇过程的实时优化控制技术的理论集成以及工业应用奠定一定的基础。本研究的主要内容有：（1）提出基于实验设计的方法进行间歇过程动态建模，以及提出基于xPC技术的动态过程实时仿真技术。以Mini-plant为实验对象，构建数据采集及监控平台，建立温度控制模型；（2）最优解和优化时间是实时优化技术需要考虑的问题，寻找更好的动态优化方法是实时优化中的一个关键问题。提出基于模型降阶的庞特里亚金最大值原理方法、基于改进的有限元正交配置法的实时动态优化技术及基于遗传算法的拟序贯动态优化技术，并以大量案例进行理论分析及验证；（3）在线控制技术在实时优化技术中主要负责具体输入操作的实施，其表现性能是直接决定实时优化技术能否成功应用的关键。针对传统的非线性模型预测控制技术进行改进，提出基于多模型的预测控制技术以提高在线计算效率，同时增强在各种扰动作用下的控制鲁棒性。本文提出了一系列的针对间歇过程建模、仿真、优化及控制的理论方法和技术，从理论的角度，这些方法更有效实用，可以作为实时优化集成理论的基本架构，从实际应用的角度，这些方法充分结合了实验测试方法，充分考虑实际过程中会出现的多种因素，相对来说更适合在实际中应用，同时也为将来的间歇过程实时优化技术的集成工业应用提供了一定工作基础。更多还原

【Abstract】 Batch processes have been widely applied to manufacture a large quantity ofvalue-added products such as pharmaceuticals, biological products, polymers and electronicchemicals. Increased demand for high product quality has motivated the interest in processmodeling, optimization, and advanced process control techniques that can improve theperformance of batch processes. Batch processes are inherently dynamic in nature and veryoften several batches are run in the same equipment. This may cause batch-to-batchvariation in operating conditions resulting in problems in safety, product quality andproductivity, which results in the development of real-time optimization technique andadvanced control technique.The operating strategy may not be the optimal strategy in presence of parametricuncertainty and disturbances. One approach to address these issues is to develop accuratemathematical models and use real-time optimization techniques to find the optimaloperating policies for each batch operation, and to ensure the optimal and stability ofproduct quality by controlling the process to track the optimal profiles. So integration ofreal-time optimization for batch process is very important for product quality control.However, the real-time optimization technique is a kind of integration strategy whichincludes techniques of dynamic modeling, dynamic optimization, on-line measurement,model update and control. The techniques of dynamic modeling, dynamic optimization andcontrol are the basic and main techniques of them. In this work, we mainly research on thenew practical techniques of batch process modeling and simulation, dynamic optimizationand nonlinear model predictive control (NMPC), which we hope will advance thedevelopment of real-time optimization technique and facilitate the implementation ofreal-time optimization technique in industrial batch processes. The main research contentsinclude:(1) Dynamic modeling for batch process with experiment design method based onMini-plant. By constructing the platform for data acquisition and process monitoring, atemperature control model has been built for the Mini-plant. Moreover, a dynamicsimulation technique based on xPC technique has been introduced. (2) Developing better dynamic optimization method for improving the computationalefficiency and the performance of the solution of real-time optimization problem. Threemethods, include dynamic optimization method based on Pontryagin’s maximum principlecombined model reduction, dynamic optimization method based on improved finite elementmethod and dynamic optimization method based on natural algorithm, have been developedin this work, and a large number of cases have been studied for validating the methods.(3) On-line control technique plays an important role in real-time optimization strategy,which in charge of the implementation of optimal trajectory tracking. Based on the standardNMPC technique, a real-time updated model predictive control technique has beenintroduced in this work, which can improve the computational efficiency of the solution ofNMPC problem and improve the robustness of the controller.In this work, we proposed a series of techniques of dynamic modeling and simulation,optimization and control for batch processes. From the view of theory, these techniques arepractical and efficient, which can become the basis of the real-time optimization strategy.Form the view of real application, since that a large number of experiment tests have beencombined into the methods and many practical problems which will be faced in real processhave been considered in the methods, these methods are more appropriate for real application.All of these works have provided some basic work for the industrial application of real-timeoptimization technique.更多还原