一类参数不确定混杂系统的事件反馈闭环优化方法及其在炼钢生产运行中的应用

There exists a kind of industrial system such as Steelmaking-casting process system, in which there are so many uncertainty disturbances, e.g. raw material, operation conditions, and machines etc. The operation systems of such industrial systems, which are the interfaces between scheduling systems and process control systems, are the specific hybrid systems. The hybrid systems have the accurate DEDS models, which describe the scheduling sequences of jobs, and the inaccurate CVDS models, which describe the multi-physics fields. Therefore manual operation mode is chosen in real industrial process. However, manual operation cannot respond those disturbances in time. As a result, the operation performances of the process are worse with those problems as more energy consumption and worse product quality. The proposal will develop the event-feedback control structure, controller design and analysis methods for the specific hybrid systems with uncertainty parameters. First, we will propose the graph decomposition and decoupling technologies based on the hyper-structures of substance flow using dynamic color path graph, then propose the methods to optimize and compensate dynamically performance criterions with fuzzy set and DEA model. Second, we will present the uncertainty measure of hybrid system models and the analysis methods of MPA equations for reducibility, observability, controllability and stability. Third, we will propose the event-feedback control structure, design the event observer, and develop sequential duality of linear Programming algorithm and performance-sensitivity-based learning and optimization algorithm for the problem with MPA and DAE constraints. Finally, we will develop operation simulation experiment system and dynamic operation optimization prototype system for steelmaking plant.

以炼钢为代表的一类生产流程，受原料、设备等多种不确定因素干扰。其生产运行系统实现调度与过程控制的衔接，是一类特殊的混杂动态系统，即：调度过程对象可由离散事件系统模型准确描述；而过程工艺涉及多物理场耦合，难以建立准确连续时间模型。因此往往采取人工操作模式实现调度和过程控制的衔接，不能根据生产变化优化生产运行，出现能耗和质量等问题。 本项目针对一类参数不确定混杂系统，提出事件反馈闭环优化结构和设计分析方法: 以动态着色路径图描述物质流动态关系，提出图的解耦方法；建立基于生产数据的运行指标数据包络模型，实现指标的离线评价和在线补偿；建立过程模型不准确性的测度，提出系统对象MPA模型的可约性、可观性、可控性和稳定性等分析评价方法；建立事件驱动的闭环反馈滚动优化结构，提出事件观测器设计方法、MPA-DAE约束的序列对偶线性规划和性能指标灵敏度学习优化算法；以炼钢为背景开发原型系统。

以炼钢-连铸为代表的一类生产流程，受原料、设备等多种不确定因素干扰。其生产运行系统实现调度与过程控制的衔接，是一类特殊的混杂动态系统，即：调度过程对象可由离散事件系统模型准确描述；而过程工艺涉及多物理场耦合，难以建立准确连续时间模型。因此往往采取人工操作模式实现调度和过程控制的衔接，不能根据生产变化优化生产运行，出现能耗和质量等问题。. 本项目针对以炼钢-连铸生产过程为代表一类参数不确定混杂系统，提出了对其运行优化的事件反馈闭环优化结构、模型和算法: 1）以炼钢－连铸生产过程为研究对象，根据生产工艺流程的动态着色路径图和过程机理，对炼钢-连铸生产运行过程进行了分析和解耦；2）针对炼钢-连铸生产过程中不确定参数的类型，提出了相应的动态优化策略；3）提出基于MPC框架的炼钢-连铸调度与控制协调优化模型体系，建立了具有PDE约束的CPU滚动优化算法和GPU滚动优化算法，并对算法的性能进行了分析和工业数据验证。