【作者】 苏成利；

【导师】 王树青；

【作者基本信息】 浙江大学 ， 控制科学与工程， 2006， 博士

【摘要】 随着现代工业的发展和科学技术的进步,对产品的质量和产量要求的不断提高,对生产经济效益的不断追求以及工业生产过程日趋大型化、复杂化,工作点的变化范围大,使得以往采用工作点附近的线性化模型来设计控制器的线性预测控制方法已不能满足控制性能要求。因此,关于非线性模型预测控制的研究已成为控制工程界的重要研究命题。本文在前人研究工作基础上,从实际出发,对非线性预测控制的若干问题进行了较为深入的研究,其中包括: (1)针对一类Wiener模型描述的非线性系统,提出了一种改进的非线性预测控制方法。该方法将Wiener模型线性部分的控制增量信号描述成Laguerre级数展开式,将预测控制中在预测时域内优化求解未来控制输入序列转化为优化求解一组无记忆的Laguerre系数,以减少优化所需的计算量;利用静态T-S模糊模型来逼近Wiener模型的非线性部分,进而利用线性预测控制方法求解控制律。CSTR过程仿真结果表明该方法是可行的。 (2)针对MIMO非线性系统,提出了一种自适应模糊预测函数控制方法。该方法利用T-S模型进行建模,然后利用加权递推最小二乘法在线辨识模型后件参数,以克服模型失配的影响;在每一采样点进行单步线性化,将T-S模型描述的非线性系统转化为线性时变的状态空间模型,从而解决了非线性预测控制中如何获得预测模型以及非线性优化求解的问题。pH中和过程仿真结果表明该方法具有良好的跟踪性能和较强的鲁棒性。 (3)针对一类满足扇形界条件的不确定模糊模型描述的非线性系统,提出了一种输出反馈鲁棒预测控制方法。该方法不需要系统状态完全可测,仅仅利用系统测量输出和不可测状态的界限值来确定保证闭环系统稳定的预测控制器。仿真结果表明该方法能够获得较好的控制效果。 (4)针对一类具有状态时滞的离散不确定非线性系统,提出了一种扩展鲁棒预测控制方法。该方法将鲁棒预测控制的‘worst-case’目标函数优化问题转化为具有LMI约束的线性目标最小化问题。在LMI优化问题可行解存在条件下,该方法保证了闭环系统的鲁棒渐近稳定。CSTR过程的仿真结果显示了该方法是有效的、可行的。 (5)将微粒群优化(PSO)算法用于输入受限非线性系统,提出了一种基于PSO的非线性模型预测控制方法。该方法采用双模控制策略,在不变集外,利用PSO算法优化求解预测控制律,使系统状态进入不变集;在不变集内,利用线性状态反馈使系统状态渐近稳定。仿真结果表明该方法的优化效率优于遗传算法。 (6)针对采用常规控制出口温度波动大、燃烧状况差等特点,设计开发了以计算量更多还原

【Abstract】 With the development of modern industry and progress of science technology, the process industry is changed to more complex and with strong nonlinear characteristics so that linear model predictive control (LMPC) may not always obtain satisfactory control results. Predictive control based nonlinear model (NMPC) has become an important research issue in the control engineering fields. Some problems of nonlinear predictive control are researched in this dissertation, and the main research works are as follows:(1) An improved nonlinear predictive control approach based on Wiener model is proposed for a class of nonlinear systems. Laguerre functions are used to describe control input of linear section of Wiener model, optimization solutions of the future control input sequences are converted into optimization of a set of immemorial Laguerre coefficients in prediction horizon. Fuzzy static model is used to approximate nonlinear section of Wiener model, optimization of nonlinear predictive control is converted into linear optimization problem. Simulation results of CSTR show that the proposed approach is feasible.(2) An adaptive fuzzy predictive functional control approach based on T-S model for multivariable nonlinear systems is proposed. The premise parameters of T-S fuzzy model are kept constant;the model consequent parameters are identified online by using the weighting recursive least square method in order to overcome the influence of model mismatch on the performance of the control system. T-S fuzzy model is linearized to be time-varying state space model in each sample point. Two major difficulties in nonlinear predictive control to obtain accurate prediction model and to solve nonlinear optimization online are effectively solved. Simulation results of pH neutralization process show that the proposed approach is an effective control strategy with excellent tracing ability and robustness.(3) An output feedback robust predictive control approach based on uncertain fuzzy models satisfying the sector bounds is proposed for a class of nonlinear systems. The system states don’t need exactly measurable, only the measure outputs and the extreme values of the unmeasured states are used to determine the controller. Stability of the closed-loop system is demonstrated. Simulation results show that the proposed approach is valid.(4) An extended robust predictive control approach for discrete uncertain nonlinear systems with state time-delay is presented. The minimization problem of the ’worst-case’ objective function is converted into the linear objective minimization problem involving linear matrix inequalities constraints (LMIs). The state feedback control law is obtained by solving convex optimization of a set of LMIs. Sufficient condition for stability on robust performance index is given. Simulation results of CSTR process show that the proposedapproach is effective and feasible.(5) A particle swarm optimization (PSO)-based predictive control approach is proposed for input constrained nonlinear systems. The dual-mode control strategy and the invariant set theory are used. Outside the invariant set, predictive control law is solved by PSO algorithm. Inside the invariant set, the system states are stabilized gradually by linear state feedback control. Simulation results show that the optimal efficiency of the proposed approach is better than GA.(6) Aiming at the status of outlet temperature fluctuating greatness and thermal efficiency reduction based on conventional control, the advanced control system is designed and exploited for delayed coking furnace. Predictive functional control with little computing quantity and strong robustness is used as the core algorithm, in addition feedforward control and feedback control. CS3000 distributed control system (DCS) is used as the exploitation platform. All kinds of conventional control module, calculation module and logic module of DCS are used to make configuration of advanced control algorithm. The system is safe and credible, dispenses with appending new hardware. Since the system runs, outlet temperature and oxygen content of the furnace become much calmer, burning status is improved, and thermal efficiency is increased.更多还原