基于模型重构的大规模复杂化工生产过程优化方法研究

Petrochemical processes become more and more large scale, centralized and continuous, which requires the processes running with higher efficiency and lower energy. Optimization of process is an important technology to improve efficiency. In order to obtain the optimal operation conditions, chemical production processes must be quickly optimized. So people can adjust the operation conditions and make them satisfy the constraints in time. But the scale of the chemical process is so large that the model is with more than one hundred thousand variables.The optimization of the model is too slow to meet the need of the real-time optimization. The project is proposed to solve this problem using the surrogate model based optimization method. Firstly the multi-source data should be processed and key decision variables can be obtained using dimensionality reduction method. Then the whole-plant model is developed using the mechanism and process data. Surrogate model is built based on Gaussian Process and updated dynamically in accordance with the whole-plant model. Swarm intelligent optimization based on probability densityis studied to decrease the time of optimization. In the end the large scale chemical process is optimized using surrogate model method under the evolution control strategy. The time of optimization is limited to less than one hour and will be applied in large scale industrial ethylene process. This project can generate the optimal operation condition for the large-scale chemical processes. It can also provide theoretical basis and technical support for energy saving of petrochemical industries.

石油化工过程日趋大型化、集中化和连续化，对生产过程的高效、低耗运行提出了更高的要求，而大规模生产过程的优化运行是提高生产效率的重要技术手段。针对大规模化工过程优化效率低的难题，本项目拟开展基于模型重构的优化策略研究。通过对大规模化工生产过程数据的多源信息处理及过程变量的维数约简得到建模所需关键决策变量，融合化工过程机理与生产运行信息建立大规模化工过程的全流程模型，探讨基于高斯过程的模型重构方法以保持与全流程模型的一致性并用于大规模过程优化从而减少计算时间，研究基于概率密度的群智能优化算法提高优化速度。在此基础上，提出“进化控制”策略下的基于模型重构的大规模化工过程优化方法，实现工业过程的及时优化，并在大型乙烯生产过程优化中进行工业验证。通过对大规模优化命题的研究，得到能够应用于大规模复杂化工生产过程的优化操作方案，为石油化学工业的节能降耗提供理论基础和技术支撑。

石油化工过程日趋大型化、集中化和连续化，对生产过程的高效、低耗运行提出了更高的要求，而大规模生产过程的优化运行是提高生产效率的重要技术手段。针对大规模化工过程优化效率低的难题，本项目以典型化工生产过程——工业聚乙烯生产过程为对象展开，深入研究了模型重构方法，建立了聚乙烯过程机理模型。从过程优化中的问题出发，开发了基于多级代理模型的改进优化算法、基于多级代理模型的迭代优化算法等多个高效优化算法。这些算法已应用于工业聚乙烯过程的优化，并在工业过程中进行了验证。基于特定的混合整数非线性问题，本项目开发了自适应竞争群优化算法，在一定程度上保证优化效果的基础上解决了化工过程换热网络同步综合问题计算量大的问题。通过对典型化工过程大规模优化问题的研究和工业验证，得到了能够应用于大规模化工生产过程的优化操作方案。