【作者】 罗国娟；

【导师】 吴刚；

【作者基本信息】 中国科学技术大学 ， 控制理论与控制工程， 2006， 博士

【摘要】 预测控制技术产生于二十世纪七十年代，在工业控制中获得了大量成功的应用，受到控制界和工业界的广泛关注。在过去近三十年里，预测控制无论是在理论研究还是工业应用中都取得了很大的进展。线性预测控制的理论已经比较成熟，非线性预测控制的理论研究也取得了一定的成果。预测控制的应用领域不断扩大，并且应用数量和被控对象的规模都持续增长。预测控制被誉为最有前途的先进控制算法，其应用前景是乐观的。 参数整定是控制器设计中的重要环节，控制效果与控制器参数有直接关系。但是预测控制的结构复杂，参数对闭环稳定性和闭环性能的影响不明确，参数整定缺乏透明性，并且待整定的参数比较多，参数之间有相互影响，这些都使得控制器的参数整定非常困难。参数整定的困难，限制了预测控制所能够获得的性能，甚至在模型匹配的情况下都很难调整参数使闭环稳定。为了使控制器能达到最佳控制效果，研究预测控制器参数之间的关系以及参数整定方法是非常必要的。 阶梯式预测控制，在预测控制中引入合理的控制量约束，简化了计算量，获得了许多成功的应用。本文针对阶梯式动态矩阵控制，深入分析了参数变化对闭环性能的影响，以及参数之间的相互关系，并提出基于多目标优化的预测控制器参数整定方法。本文的主要研究工作及创新之处包括： (1) 将工业控制中常用的一阶惯性加纯滞后形式的对象模型，以采样周期为单位进行了归一化处理。深入分析了动态矩阵控制器和阶梯式动态矩阵控制器参数变化对闭环性能的影响。由于阶梯因子的作用，其余参数对于闭环性能的影响也发生了变化。 (2) 研究了控制权重与柔化因子之间的关系，建立的量化表达式将控制权重取值与闭环时间常数间接联系起来，使控制权重值赋予了更直观更明确的物理意义，并指出控制权重的取值与模型增益的关系。 (3) 研究了控制权重的作用范围与控制器参数之间的关系，在某些情况下控制权重的取值下限值很大且随其余参数剧烈变化，可以仅用阶梯因子对未来控更多还原

【Abstract】 Model predictive control (MPC), which appeared in industry in the 1970s, has been applied successfully in many process industries, and gained much interesting by the control theoreticians as well as control practitioners. In the last nearly thirty years, much progress has been made in theory research of MPC. By now, linear MPC theory is quite mature, and many issues have been addressed in nonlinear MPC formulation with academic success. MPC technology can be found in a wide variety of application areas, and both the number of applications and the size of the largest application are increasing. As the most promising advanced control algorithm, MPC’s application in future will be optimistic.Tuning of parameters is an important step in the controller design process, since the closed-loop performance is related directly with values of the parameters. However, the MPC tuning is very difficult due to the complex structure of MPC, the implicit effect of parameters on closed-loop stability and performance, and the overlapped effect of the large number of parameters. The difficulty in tuning may degrade the control performance. Tuning MPC controllers for stable operation in the presence of constraints may be difficult, even when the process model is perfect. In order to achieve better performance, it is necessary to analyze the relationships among parameters and propose some efficient tuning methods.Stair-like MPC, which introduced reasonable constraints on control moves to reduce computational burden, has achieved many successful applications. Focused on stair-like DMC, the effects of parameters on closed-loop performance and relationships between parameters had been studied deeply, and then a multi objective optimization based tuning method for MPC was proposed. The main point and innovation of the thesis is listed as follows:1. The first-order plus dead time model, which is the most common model used to describe the main dynamic features of plants, was scaled by sample time. Both the effects of parameters on closed-loop performance of DMC and stair-like DMC were analyzed deeply. The stair-like factor changes the effect of parameters on closed-loop更多还原