多尺度动态行为-网络化交互作用下多速率采样分布式预测控制策略研究

大型工业过程的分布式预测控制系统因子系统的维数和子系统间动态行为的耦合以及不同子系统动态行为间时尺度上的差异，对于由网络进行信息传输的控制结构，网络不确定性和不同尺度动态行为是影响控制性能的主要原因。本项目采用理论、仿真分析和实验等方法研究以反应器、存储罐和分离器组成的过程在网络连接形式下的分布式预测控制性能和多速率采样下的优化协调稳定性。从过程优化控制的理念出发，探讨时滞、数据丢包及子系统间时延动态耦合行为对分布式预测控制性能的影响。在此基础上，通过设计网络化分布式预测控制装置，在变采样速率下建立网络化的采样率-量化作用的子系统数学模型，掌握影响协调和稳定性的主要因素。深入理解网络化和子系统间的动态行为尺度共同作用的机理，本项目的研究有利于准确评定大型过程以网络化连接的分布式预测控制调节的综合控制性能。对提出更加合理的控制策略及扩大工业过程网络化分布式预测控制的应用具有重要的理论意义。

When distributed model predictive control(DMPC) is used for manipulating a large-scale industrial process, due to dimension difference, time-scale of dynamics, and dynamic coupling behavior among subsystems, uncertainty induced by network and time-scale of dynamics difference among subsystems are the main factor that affects the DMPC performance for the control structure in which information transmitted using the communication network, In this project, the theory analysis, simulation and experimental methods are used to investigate optimization coordination and stability performance of DMPC on process consists of reactor, storage tank and a separator process (RSS) in the network connection manner using multi rate sampling manner. Starting from the idea of process optimization control, the effects of time delay, data packet dropout and subsystems delay dynamic coupling behavior on the performance of DMPC are explored. Based on this, RSS network distributed predictive control device are designed, the networked mathematical model of RSS are established considering both sampling rate and quantization effects. The main factors that affect the stability of networked DMPC performance. To deeply understand the mechanism of interaction of networked uncertainty and multi-time scale dynamic behaviors among subsystems, the comprehensive control performance about networked DMPC for large-scale industrial process can be assessed accurately according to this project. Not only the more reasonable network distributed predictive control strategy can be proposed, but also the expansion of the industrial process networked distributed predictive control application has important theoretical significance and practical application value.