预测控制优化变量的集结策略

Aggregation optimization strategy in model predictive control

为减少预测控制的在线计算量 ,提出了对优化变量进行集结的优化策略。通过一个集结矩阵将原优化变量与集结后的变量联系起来 ,从而为多种预测控制算法建立一个统一的框架。分析了现有的几种典型优化策略的集结表述 ,说明了所提出的集结优化框架具有一定的普适性 ,并在此框架下提出一种新的具有输入衰减形式的集结优化算法。

An aggregation optimization strategy is proposed to reduce the on line computational burden of model predictive control (MPC). A uniform framework is built for a class of simplified strategies, in which the original optimized variables are connected with the aggregated ones through the aggregation matrix, so the on line optimization of MPC is turned into a low dimensional programming of aggregated variables or the aggregated matrix. Several existing typical MPC algorithms are described and unified under this framework. Based on this framework, a new aggregation strategy with input decaying form is also presented.