面向流程工业的分布式状态估计与输出反馈控制研究

With the rapid development in cyber-physical systems and the increasing requirement of complex processes on precise operations, the distributed process control architecture has emerged as the new trend in process control system design. While maintaining the structural flexibility of decentralized process control, distributed process control may achieve the centralized process control performance. Distributed process control systems have been widely recognized as the next generation advanced process control architecture for processes and been identified as one of the key fundamental technologies in smart manufacturing. When moving to distributed process control architecture, there are some important issues that need to be addressed including distributed state estimation, distributed output feedback process control and distributed fault tolerance control. Based on the research team’s previous results on robust distributed moving horizon state estimation and distributed model predictive control, in this project, we first develop a distributed state estimation algorithm that can handle processes exhibiting multi-time scales in dynamics which is a common phenomenon in industrial processes; subsequently, we further study distributed state estimation subject to unknown interactions between subsystem based on interaction prediction coordination; furthermore, we consider the integration of distributed state estimation and distributed output feedback process control and develop a systematic approach for the design of output feedback distributed process control systems. Rigorous analysis will be carried out to establish sufficient conditions for the convergence and closed-loop stability of the proposed state estimation and control algorithms. The developed algorithms will be applied to an experimental wastewater treatment plant. It is expected that this project will contribute significantly to the theoretical developments of distributed state estimation and control and will greatly enhance the application of distributed state estimation and control in processes.

随着信息物理系统快速发展和复杂流程工业对精细化操作需求日益严格，采用分布式控制系统是工业发展的必然趋势。分布式控制兼具集中控制的优越性能及分散控制的灵活架构，被广泛认为是新一代流程工业的标准操作模式，也被公认为是 “智能制造”的一个核心基础技术。分布式状态估计、分布式输出反馈控制、分布式系统容错控制等是分布式控制系统研究中具有挑战性的基础理论问题。本项目在前期鲁棒分布式状态估计及分布式预测控制工作的基础上，针对一类具有多时间尺度特征的分布式系统，提出分布式滚动时域状态估计算法并给出收敛性条件；进一步针对子系统关联未知的情况，建立基于关联预估协调方法的分布式状态估计架构、算法及收敛性条件；提出分布式输出反馈控制系统设计方法及收敛性、稳定性条件；开展上述理论成果在污水处理分布式控制系统中的应用验证。预期成果将极大拓展分布式控制系统的应用范围，强化分布式控制系统性能，具有重要的理论和应用价值。

分布式控制兼具集中控制的优越性能及分散控制的灵活架构，被广泛认为是新一代流程工业的标准操作模式，也被公认为是 “智能制造”的一个核心基础技术。分布式状态估计、分布式输出反馈控制、分布式系统容错控制等是分布式控制系统研究中具有挑战性的基础理论问题。本项目在前期鲁棒分布式状态估计及分布式预测控制工作的基础上，针对现有分布式滚动时域估计算法在多时间尺度过程中估计时域过大、计算量过高、性能不能保证等局限性，研究了多时间尺度情况及直方图方法，并对算法的收敛性进行了分析。在子系统关联未知的情况下，研究了非线性系统分布式状态估计中的子系统分解问题，提出了一种系统的子系统分解方案，包括整个系统的可观性测试，每个输出测量值的可观状态辨识，可测输出及状态变量间的相关度分析和敏感度分析等，并给出了收敛性条件。基于分布式切换观测器设计了输出反馈的分布式模型预测控制，并优化设计分布式切换状态估计器和分布式模型预测控制各个子系统之间的信息交换协议。最后将理论成果在污水处理分布式控制系统中进行了应用验证。研究成果拓展了分布式控制系统的应用范围，强化了分布式控制系统性能。