对角CARIMA模型多变量广义预测控制改进算法

为了增强多变量广义预测控制算法(MGPC)的实用性,对其实现形式进行了进一步的简化.利用对角CARIMA模型的结构特点,先对系统中单个输出变量期望值的自由响应部分进行分解推导,将其表达成自由响应项系数与系统输入输出变量已知值乘积的形式,得到此输出变量的预测表达式,然后将系统所有输出变量的预测表达式代入目标函数中,得到的控制增量等于控制器系数与参考轨迹、过程输入输出历史数据的乘积.控制器系数只与模型参数及设计参数有关,求解控制量时不再需要进行模型输出预报,控制器结构简单,实现容易.对比实验结果表明了该方法保持了常规MGPC方法的优秀控制性能.

Improved algorithm for multivariable generalized predictive control of diagonal CARIMA model

The implementation of multivariable generalized predictive control(MGPC)is simplified to improve its practicability.Each output prediction of a multivariable system can be separately obtained by utilizing the structure char-acteristic of the diagonal controlled autoregressive integrated moving average(CARIMA)model,and its free response part can be expressed as the product of coefficients and known input and output values by disassembling and deriving.Sub-stituting all these output prediction into the objective function produces the control law which is equal to the product of controller's coefficients and reference trajectory and system historical input and output data.These coefficients is deter-mined only by the model and tuning parameters,and the computation of output prediction requires no control increment.The controller structure thus obtained is simple and easy to be implemented.The comparative experiment results show this method has control performance as good as that the conventional MGPC.