对角CARIMA模型多变量广义预测控制

为了丰富多变量广义预测控制算法(MGPC)的建模能力、降低其求解难度、增强其控制参数选择的灵活性，采用对角形式的受控自回归积分滑动平均 (CARIMA)模型来改进MGPC.通过把CARIMA模型中的A与C矩阵构造成对角多项式矩阵的形式，把m个输入n个输出的多变量对象的参数辨识与模型预测问题分解成一系列m个输入单个输出子对象的参数辨识与模型预测问题.一方面简化了模型辨识问题，另一方面避免了模型预测中大量的矩阵运算，从而减轻了在线运算负担，简化了MGPC的实现，增强了MGPC的实用性.在一个由集散控制系统（DCS）控制的非线性液位装置上的对比实验结果表明，该方法保持了常规MGPC方法的优良性能.

Multivariable generalized predictive control for diagonal CARIMA model

A diagonal multivariable controlled autoregressive integrated moving average(CARIMA) model was used to improve the conventional multivariable model predictive control(MGPC),so as to enhance its capability for modelling,abating its difficulty for deriving,and boasting its agility in choosing tuning parameters.By constructing matrices C and A of CARIMA model to be diagonal,the identification and prediction problem of multivariable process with m inputs n outputs were transformed into that of n sub-processes with m inputs single output.This not only simplified the system identification,but also avoided many matrix operation,which can substantially ease the computational overhead associated with MGPC,simplify its implementation,and enhance its practicability.The comparison experiment results obtained from a nonlinear liquid system controlled by a distributed control system(DCS) show that this method retains the good performance of the conventional MGPC.