Abstract: A stair-like generalized predictive control algorithm based on multiple models switching is proposed for systems with jumping parameters.Multiple fixed models,a conventional adaptive model and an assigned initial value adaptive model are established to identify dynamic characteristic in parallel. Multiple fixed models can improve transient performance,conventional adaptive model can ensure the stability of systems and assigned initial value adaptive model can further enhance transient performance.The best sub-model is selected as global model according to the switching index,and a stair-like generalized predictive controller is designed for this model.Consequently,the full-range operation is achieved.Finally,simulation shows that the control effect of proposed method is superior to a single controller.

Key words: multiple models, stair-like control, generalized predictive control, adaptive

摘要： 针对一类模型参数突变的系统，提出一种基于多模型切换的阶梯式广义预测控制算法。采用多个固定模型、一个常规自适应模型和一个可重新赋初值的自适应模型并行辨识系统的动态特性。多个固定模型可以提高系统的暂态性能，常规自适应模型可以保证系统的稳定性，可重新赋初值的自适应模型可以进一步提高系统的暂态性能。在每个采样时刻基于性能指标切换到最优的局部模型作为当前模型,设计阶梯式广义预测控制器，从而实现系统全局的控制。最后的仿真结果表明，其控制效果明显优于单一模型的控制器。

关键词: 多模型, 阶梯式控制, 广义预测控制, 自适应