【作者】 牛健；

【导师】 钱积新； 赵均；

【作者基本信息】 浙江大学 ， 控制科学与工程， 2010， 博士

【摘要】 预测控制是一种在实际应用中发展起来的控制算法,先以其易用性,鲁棒性,能处理输入输出约束以及多输入多输出系统的特点在实际工业过程中应用起来,而后引起理论界的重视,围绕预测控制算法展开了理论研究。预测控制是一种基于模型的控制算法,预测控制理论的发展都是围绕对模型的深入认识展开的。本文针对特殊的模型,提出了针对特殊模型的预测控制算法。本文的主要内容和创新点包括:(1)当模型具有状态空间形式时,注意到奇异摄动系统的特性不仅与系数矩阵有关,而且与输入矩阵有关。当输入矩阵变为特殊形式时,系统可以在输入方面解耦。(2)针对两类特殊的奇异摄动系统,输入双时标系统,输出双时标系统,在传递函数矩阵的形式下分别提出了针对这两种特殊模型的预测控制算法。并讨论了这几种预测控制算法的稳定性问题(3)对供应链预测控制问题进行了分析。提出了一种能够在线辨识需求模型的预测控制算法。(4)鉴于价格对需求的动态影响,把价格作为一个操作变量引入供应链预测控制算法中。由于订货和价格对库存的影响可能在不同的时间维度上,在价格通过需求对库存的影响远慢于订货量对库存的影响前提下,提出了针对这种模型的双时标预测控制算法。最后总结全文,指出今后有待进一步研究的内容。更多还原

【Abstract】 Model predictive control (MPC) originated from industry. Because MPC can be used to deal with multi-input-multi-output system and the application of MPC is not so easy, MPC has been introduced to process industry extensively. The theoretical research about MPC is a little later than the application. MPC is a model based control algorithm, and the development of MPC theory is closely connected with the understanding of the model. This thesis focuses on one special kind of model, and puts forward a series of special MPC algorithms.The main contents and renovation points in this thesis include the following:(1) When the model has the state space form, the characteristic of the singular perturbation system is not only determined by the coefficient matrix, but also affected by the input matrix. And when the input matrix possesses a special form, the model can be decoupled according to the input.(2) In the view of the transfer function matrix form, the singular perturbation system can be divided into two special types: input two-time scale system and output two-time scale system. And the special model predictive control algorithms for each special systems are proposed. And the stability of each special model predictive control algorithm is discussed in the thesis.(3) We analyze the application of model predictive control on the supply chain management, and a kind of model predictive control algorithm with on-line identified demand model is put forward.(4) Considering the relationship between demand and price, and the pricing is introduced as a manipulate variable, because the price and the ordering can affect the inventory on different time scales. We put forward a kind of two-time scale model predictive algorithm for a supply chain unit under the assumption that price has more slow effects on the inventory than the ordering on the inventory. At the end of the thesis, we give some suggestions to the further research in this field.更多还原