基于聚类多模型建模的多模态预测控制

多模型预测控制(MMPC)是解决非线性控制问题的重要手段，本文针对多模态控制器设计中模态匹配准则的选取问题，利用当前样本状态与各聚类建模子空间距离差异，提出了一种基于距离匹配的多模型控制器加权算法。然后，基于模态融合思想，提出了模态加权构建实时预测模型的控制策略。通过对pH中和过程进行仿真，结果表明：两种方法都提高了非线性系统的暂态响应，跟踪特性优良，体现了它们对非线性系统大范围控制的有效性。

Research on multi-mode MPC based on clustering multi-modeling

Multi-mode model predictive control (MMPC) is an important method for the control problem of a nonlinear systemIn this paper, aiming at the selection problem of model matching criterion in multi mode controller design, a weighted averaging method for multi-model manipulated variables was presented based on the ‘distance match’ criterion, in which the distance difference between current sampling state and each clustering center was used to obtain weighted averaging coefficientTo improve the precision of model, based on ‘model fusion’ technique, the predictive model was reconstructed online by modal weighted averagingThe simulation results in the pH neutralization process showed that both algorithms proposed in this paper could improve the transient response of a nonlinear system and perform good output tracking in a wide range.