基于混合模型的间歇过程动态实时优化方法的研究

项目名称： 基于混合模型的间歇过程动态实时优化方法的研究

项目编号： No.61203103

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 自动化学科

项目作者： 贾润达

作者单位： 东北大学

项目金额： 23万元

中文摘要： 为了降低间歇过程的生产成本、改善产品质量以应对日益激烈的市场竞争，有必要对间歇过程的运行优化控制进行深入研究。目前针对间歇过程的动态优化大多基于严格机理模型，然而由于实际工业过程的复杂性，往往存在模型不匹配的现象，从而难以保证优化结果的可靠性。本项目基于混合建模思想，提出利用动态实时优化技术解决间歇过程的运行优化控制问题：通过研究动态混合模型的增量辨识策略，提出利用正则化方法实现不可测中间变量的精确估计；在上述研究结果的基础之上，结合离群点的识别方法，利用集成建模思想建立经验模型；通过对模型预测置信域的计算，指导混合模型的在线更新；通过联立算法求解优化模型，利用正交配置实现优化模型的完全离散化，并应用序贯二次规划求解所形成的大规模非线性规划问题；把所提出的相关方法和技术在已有的实验平台上进行验证与完善，并逐渐将其推广应用到1-2个典型的间歇工业生产过程。

中文关键词： 间歇过程；混合模型；模型更新；离群点；动态优化

英文摘要： To reduce the cost of production for batch processes, and improve the product qualities to cope with the increasingly fierce market competition, it is necessary to deeply research the operational optimal control of batch processes. Currently most dynamic optimization for batch processes was based on rigorous first principle models, but due to the complexity of the practical industrial processes, the situation is that the models are usually not well-matched, and it is difficult to guarantee the reliability of the optimization results. This project is based on the idea of hybrid model, and proposes to employ dynamic real-time optimization techinque to solve the operational optimal control problem of batch processes: By studying the incremental identification strategy for dynamic hybrid model, regularization method is used to accurately estimate the ummeasurable intermediante variables; On the basis of the aforementioned study resluts, integrated with outliers detection method, the idea of ensemble modeling is utilized to calibrate empirical models; Online update of the hybrid model is guided by calculating the prediction confidence region of the model; By introducing simultaneous strategy to solve the optimization model, orthogonal collocation is used to fully discrete the optimization model, and sequential quadra

英文关键词： batch process；hybrid model；model update；outlier；dynamic optimization