基于数据的受限最优迭代学习控制方法的设计和仿真

发布时间：2018-06-11 16:15

本文选题：数据驱动控制 + 迭代学习控制　；参考：《青岛科技大学》2017年硕士论文

【摘要】：许多的工业生产过程是受输入或输出范围限制的,针对此类受限系统,本论文提出了基于数据的受限最优迭代学习控制(constrained-DDOILC)方案,论文的主要工作及创新点如下:1)对传统的受限最优迭代学习控制(OILC)方法进行了分类整理,受限范围分为输入受限、输出受限和输入输出受限三方面,同时讨论了当控制系统中存在不确定参数时所采用的控制方法。2)针对一类带输入输出约束的非线性离散时间系统提出了constrained-DDOILC控制方案,该方案不需要任何的模型信息,仅仅利用I/O数据。通过引入一种新的迭代动态线性化方法,利用二次规划的最优控制工具,考虑I/O受限条件来进行设计。由于所提方案是数据驱动的,无需建模,因而设计过程中不存在未建模动态等问题,同时可以保证被控系统的鲁棒性,本文首先提出基于Lifted的技术进行控制设计,利用控制实例对其进行了仿真验证。3)针对同样的带输入输出约束的非线性离散时间系统,把基于Lifted技术的控制方案推广到Non-lifted技术上,并进行了相应的收敛性证明和仿真验证。4)针对工业生产过程中存在的多点跟踪问题,论文提出了基于数据的受限最优点对点迭代学习控制(constrained-DDOPTPILC)方案,该方案只需利用特定点的误差信息,跟踪给定系统期望的某一个或某几个点。此种方案大大减小了计算量,减轻了产业成本,提高了工作效率。通过与传统方法的仿真对比,进一步验证了所提方法的控制有效性。
[Abstract]:Many industrial processes are limited by input or output. In this paper, a constrained optimal iterative learning control scheme based on data is proposed for such constrained systems. The main work and innovation of this paper are as follows: (1) the traditional constrained optimal iterative learning control (OILC) method is classified. The restricted range is divided into three aspects: input constraint, output limitation and input / output limitation. At the same time, the control method. 2) for a class of nonlinear discrete-time systems with input and output constraints, a control scheme of constrained DDOILC is proposed, which does not require any model information. Using only I / O data. By introducing a new iterative dynamic linearization method and using the optimal control tool of quadratic programming, I / O constrained conditions are considered to design. Because the proposed scheme is data-driven and does not need to be modeled, there are no problems in the design process, such as unmodeled dynamics, and the robustness of the controlled system can be guaranteed at the same time. A simulation example is used to validate the proposed method. 3) for the same nonlinear discrete time system with input and output constraints, the proposed control scheme based on the lifted technique is extended to the Non-lifted technology. The corresponding convergence proof and simulation verification. 4) aiming at the multi-point tracking problem in industrial production process, a data-constrained optimal point-to-point iterative learning control scheme named constrained DDOPTPILC is proposed in this paper. The scheme only needs to use the error information of a given point to track one or more expected points of a given system. This scheme greatly reduces the calculation amount, reduces the industrial cost and improves the working efficiency. The control effectiveness of the proposed method is further verified by comparison with the traditional method.
【学位授予单位】：青岛科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP273

【参考文献】