离线优化在线查表的双层结构预测控制  

Off-line optimization and online table lookup for double-layer structure predictive control

作　　者：郑洪宇[1,2] 潘昊[1,3] 邹涛[1] 胡静涛[1] 于海斌[1] 

机构地区：[1]中国科学院沈阳自动化研究所信息服务与智能控制研究室,辽宁沈阳110016 [2]东北大学信息科学与工程学院,辽宁沈阳110819 [3]沈阳中科博微科技股份有限公司,辽宁沈阳110179

出　　处：《控制理论与应用》2016年第6期735-744,共10页Control Theory & Applications

基　　金：国家高技术研究发展计划("863"计划)(2014AA041802);国家自然科学基金(61374112);工信部智能制造综合标准化试验验证项目(智能制造系统能效评估方法标准研制及试验验证系统建设);中国科学院网络化控制系统重点实验室基金(WLHKZ2014010)资助~~

摘　　要：双层结构预测控制包括稳态优化和动态控制,计算复杂度较大,难以在实时性要求较高或者是主控制器计算能力较弱的场合应用.本文阐述了一种离线计算在线查表的稳态优化方法和详细的技术实施方案.首先采用枚举方式离线计算稳态优化值,然后通过在线查表得到实时优化值.当在线查表不可行时,通过查找距离该查表点最近的多个可查表点,以距离作为加权系数,近似计算最优稳态目标值.最后,通过性能分析和仿真实验证明了本文提出的该方法有效的降低了计算复杂度,减少了计算时间.Double-layer structure model prediction control, composed of steady-state target calculation and dynamic optimization, is restricted in the field of high real-time control and poor computational controller. A new method of steadystate target calculation, called off-line optimization and online table lookup, is proposed. Also the calculation method is given in detail. First, the steady-state optimization database is estimated through enumeration and off-line optimization.Then the steady-state optimization is addressed based on online lookup table. For the case that online table lookup is not feasible, the steady-state optimization is calculated approximately by seeking the multiple points which are nearest to the lookup table point in the lookup table, and selecting the distances as the weighting coefficient. Finally, The capabilities of the method in this paper are investigated through computational complexity analysis and computer simulation.

关 键 词：离线优化 模型预测控制 双层结构 稳态优化 计算复杂度 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]