多变量加权多步预报控制*

对于线性多变量系统,本文给出了一种完全不同于以往的全状态反馈或观测器-控制器型的算法——多变量加权多步预报控制(MWLPC)算法。这种算法除引进了预测控制中的多步输出预报、滚动优化等机制外,最重要的是在二次型性能指标中引入了可调的多项式或有理分式矩阵权因子;适当选取这些权因子,便可按设计要求、仅用系统的输出信息反馈便能任意配置闭环系统的特征矩阵,从而保证闭环稳定性和其他优良性质。此外,该算法不改变原系统的零点,因而适用于非最小相位系统。

Multivariable Weighted Long-Range Predictive Control

This paper gives a new type of multivariabie controller-Multivariable Weighted Long-range Predictive Controller (MWLPC).Besides using some mechanics of predictive control in MWLPC,we introduce some weighting polynomial matrices or rational fraction matrices which can be adjusted by designer in the cost function.Only using the system output information feedback,we can set arbitrarily the closed-loop characteristic matrix by choosing the proper weighting factors,and so the closed-loop stability can be ensured and the steady-state tracking errors can be eliminated.In addition the MWLPC can be also used in a unminimum phase systems.