【作者】 陆妹；

【导师】 邵惠鹤；

【作者基本信息】 上海交通大学 ， 控制理论与控制工程， 2008， 博士

【摘要】 模型预测控制因为其具有处理受约束复杂系统的能力,已经被广泛应用于过程工业中。预测控制算法在处理时滞系统时,主要有两种方法:一是把优化时域比无滞后时延长d拍,其被控对象是纯滞后系统;二是将时滞系统转换成无时滞系统,其计算量将大大增加。工业被控对象不可避免的存在着不确定性、非线性、时滞和受扰动等现象。本文在设计预测控制器时直接考虑时滞现象和不确定性,系统地研究了时滞系统的预测控制算法。主要内容如下:1.分别针对范数有界和凸多面体两种不确定状态时滞系统,考虑输入饱和约束,给出了带有饱和特性的状态反馈预测控制研究。当被控对象存在凸多面体不确定性时,分别给出了常值状态反馈预测控制算法和增益调度预测控制算法,采用后者可获得较好的控制性能,但是该算法的计算量较大。为了减少计算量,给出了一种基于标称目标函数的预测控制算法。考虑到输入时滞的存在,针对凸多面体不确定状态和输入时滞系统,进行了带有饱和特性的状态反馈预测控制研究。通过仿真研究验证了算法的有效性。2.研究了适用于具有状态和输入时滞的非线性时滞系统的有限时域模糊预测控制算法。首先以公共Lyapunov -Krasovskii函数(CLF)为终端代价函数,分别以模糊状态反馈控制律和带有饱和特性的模糊状态反馈控制律为局部控制器,展开预测控制算法研究。为了进一步改善被控系统的性能,以模糊Lyapunov-Krasovskii函数(FLF)为终端代价函数,进行预测控制算法研究。通过对CSTR的仿真研究表明,有限时域模糊预测控制算法能够有效地控制具有状态和输入时滞的非线性时滞系统,而以FLF为终端代价函数、以带有饱和特性的状态反馈控制律为局部控制器的有限时域模糊预测控制算法具有较好的控制性能。3.研究了分段状态时滞系统的非脆弱状态反馈预测控制算法。控制器参数不确定性的普遍存在性,推动了非脆弱控制器的设计。分别以分段线性不确定状态时滞系统和分段模糊状态时滞系统为被控对象,采用分段Lyapunov-Krasovskii函数(PLF)方法,考虑加法和乘法两种形式的控制器增益摄动,展开非脆弱预测控制算法研究。通过对实例的仿真研究,验证了所提算法的有效性。4.研究了输入受限的受扰动线性状态时滞系统的滚动时域H∞控制算法。算法融合了模型预测控制的滚动优化原理,用系统当前状态刷新相应的LMI优化问题并在每个时刻在线求解,在满足约束条件的同时保证了闭环系统的H∞性能。考虑到系统状态可测和不完全可测两种情况,分别给出了状态反馈和动态输出反馈滚动时域H∞控制算法。仿真实例表明,该算法能实现对输入受限的受扰动线性状态时滞系统的有效控制。更多还原

【Abstract】 Model predictive control, which is a control algorithm able to handle constrained complex systems, has been widely used in process industry. When the predictive control algorithm was used to control systems with time-delay, two kinds of methods were applied. Firstly, the optimization horizon was d larger than that of algorithm which was used to control systems without time-delay. This method was adapted to time-delay systems with pure time delay. Secondly, systems with time-delay were transformed into systems without time-delay. As a result, the computation burden was increased. There always exist phenomena such as uncertainties, nonlinearity, time-delay and disturbances in industry. This dissertation investigates systematically predictive control algorithms taking time-delay and uncertainties into consideration explicitly in the course of controller design for systems with time-delay.The main contents of this dissertation are stated as follows:1. Saturating state feedback predictive control algorithms are investigated for norm-bounded and polytopic uncertain systems with state delay subject to input saturation. When polytopic uncertainty is considered, constant state feedback and gain-scheduling predictive control algorithms are presented. The latter can obtain better performance, but computation burden is high. To overcome this drawback, a predictive control algorithm based on nominal objective function is presented. For polytopic uncertain systems with both state and input delays, a saturating state feedback predictive control algorithm is also investigated. The simulation results demonstrate that the algorithms are effective.2. Finite horizon fuzzy predictive control algorithms for nonlinear systems with both state and input delays subject to input constraints are investigated. The predictive control algorithms applying common Lyapunov-Krasovskii function as the terminal cost function, applying fuzzy state feedback controller and fuzzy saturated controller as local controllers are presented. To improve the performance of the closed-loop system, a fuzzy Lyapunov-Krasovskii function is applied as the terminal cost function. From the application to continuous stirred tank reactor (CSTR), the effectiveness of the algorithms is guaranteed. And the system controlled by the algorithm whose terminal cost function is a fuzzy Lyapunov-Krasovskii function can obtain better performance.3. Non-fragile state feedback predictive control algorithm for piecewise system with state delay is investigated. Errors in the controller coefficients which always exist in industry speed the design of non-fragile controllers. In the algorithm, the controlled system is extended into uncertain piecewise linear system with state delay and nonlinear fuzzy system with state delay, piecewise Lyapunov-Krasovskii function is applied and both additive and multiplicative types of controller perturbations are considered. Simulation results show that the presented algorithms are effective.4. Receding horizon H∞control algorithms are investigated for systems with state delay subject to input constraints and disturbances. Combining receding optimization of model predictive control and H∞control, the algorithms are a LMI-based receding optimization problem with new measurement at each time instant. The disturbance attenuation levels of the controlled systems are guaranteed and the input constraints are satisfied. When not all states are available, a dynamic output feedback receding horizon H∞control algorithm is presented. The simulation verifies that both state feedback and dynamic output feedback receding horizon H∞control algorithms are effective to control the systems with state delay subject to input constraints and disturbances.更多还原