具有输入增量分辨率的多变量预测控制策略  

Multi-variable predictive control strategy with input increment resolution

作　　者：邹涛[1] 刘博文 王美聪 孙威[1] ZOU Tao;LIU Bo-wen;WANG Mei-cong;SUN Wei(Key Laboratory of Networked Control System of CAS, Shenyang Institute of Automation of Chinese Academy ofSciences, Shenyang 110016,China;University of Chinese Academy of Sciences, Beijing 100049, China;College of Environment and Safety Engineering, Shenyang University of Chemical Technology, Shenyang 110142, China)

机构地区：[1]中国科学院沈阳自动化研究所中国科学院网络化控制系统重点实验室,沈阳110016 [2]中国科学院大学,北京100049 [3]沈阳化工大学环境与安全工程学院,沈阳110142

出　　处：《控制与决策》2019年第7期1529-1536,共8页Control and Decision

基　　金：国家重点研发计划项目(2017YFB0603703);国家自然科学基金项目(61773366,61503257);工信部工业互联网创新发展工程及智能制造综合标准化与新模式应用项目(时间敏感网络(TSN)与用于工业控制的对象链接与嵌入统一架构(OPC UA)融合关键技术标准研究与试验验证)

摘　　要：工业现场中不可避免地会出现噪声和不可测扰动等不利因素,造成控制器的控制变量波动,不利于加以控制和实现设备的长期使用。同时,对于现场中需要控制动作较大的阀门,在面对较小的控制动作时,无法准确跟踪控制。针对上述两种问题,结合当前工业控制目标为设定点和区间混合类型,引入动态增量阈值和静态增量阈值作为输入分辨率。当约束多变量系统通过QP规划算法求解出最优控制律后,对于动态阈值,若控制律超过阈值,则将此控制律添加到输入上,反之则忽视此控制律;对于静态阈值,若控制律超过阈值,则将此控制律添加到输入上,反之将控制律累加,直到累加值超出阈值后采用累加的控制律。仿真结果验证了所提策略能够有效保证控制变量的稳定,解决大控制动作阀门的跟踪控制问题。Unfavorable factors such as noise and unmeasurable disturbances are inevitable in the industrial field, which cause the fluctuations in the control variables of the controller and are not conducive to the actual control and long-term use of the equipment. At the same time, for valves that require large control action, it is impossible to accurately track and control when faced with small control action. For the above two problems, combined with the current industrial control target as the hybrid type with set point and interval, the dynamic incremental threshold and the static incremental threshold are introduced as the input resolution. When the constrained multi-variable system solves the optimal control law through the QP planning algorithm, for the dynamic threshold, if the control law exceeds the threshold, the control law is added to the input, otherwise the control law is ignored. For the static threshold, If the control law exceeds the threshold, the control law is added to the input, otherwise the control law is accumulated until the accumulated value exceeds the threshold, and then the accumulated control law is used. The simulation verifies that the strategy effectively guarantees the stability of the manipulated variable and solves the tracking control problem of the large control action valve.

关 键 词：多变量预测控制 区间预测控制 混合型目标 动态阈值 静态阈值 增量分辨率 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]