对角CARIMA模型输入输出约束自适应广义预测控制

Input and output constrained adaptive generalized predictive control for diagonal CARIMA model

为了简化输入输出约束存在时多变量广义预测控制算法的设计与实现问题,提出了对角结构形式的多变量CARIMA模型的约束广义预测控制算法。根据对角CARIMA模型的结构特点将多输入多输出对象的模型输出预报和参数辨识转化为多输入单输出子对象的模型输出预报和参数辨识,并详细分析了系统的输入输出约束形式,将模型输出预报、目标函数和约束条件中依赖于模型参数的项与依赖于过程历史数据的项完成分离,简化了模型输出预报的计算及优化问题的求解,从而简化了约束多变量广义预测控制算法的实现。在一个DCS控制的非线性液位装置上的对比实验结果表明了该方法的有效性。

An improved constrained mulitivariable generalized predictive control algorithm （ICMGPC） lor CARIMA （controlled autoregressive integrated moving average） model having diagonal matrics A and C was developed to simplify the design and implementation of ICMGPC when considering input and output constraints. The identification and model output predication problems of multivariable process with m inputs and n outputs are transformed into those of a series of subprocesses with m inputs and single output by utilizing the structure characteristic of the diagonal CARIMA model. The input and output constraints of the control system are also analyzed in detail. This method separates items depending upon model parameters from those depending upon input and output historical data in model output predic- tion, objective fuction and constraint condition, which simplifies the implementation of ICMGPC. Its validity is demonstrated by comparative experiment results obtained from a nonlinear liquid level equipment controlled by DCS.