【作者】 权玲；

【导师】 王学伟； 靳其兵；

【作者基本信息】 北京化工大学 ， 控制理论与控制工程， 2010， 博士

【摘要】 随着现代工业的飞速发展,生产过程变得越来越复杂,常常包含着非自衡对象,变结构对象,而过程对象中又常常存在不确定性、时滞性、多变量耦合、输入输出受约束等复杂特征。对于这些复杂的被控工业过程,采用常规的内模控制方法难以达到理想的控制效果。为了更好的发挥内模控制的优势,针对复杂系统中的各种特殊被控对象研究出简单、有效、实用的新型内模控制方法已成为当前研究领域及现代工业的迫切需求之一。本文正是基于这个立足点,以内模控制为基本控制方法,从内模控制结构的改进、内模结构的稳定性与鲁棒性分析、滤波器的改进等方向对复杂系统中一些难题提出了一些解决方案和改进措施,主要研究内容和已取得的研究成果如下：(1)阐述了系统辨识的原理和几种常见的辨识方法。重点研究了粒子群算法(PSO)和无约束多维寻优法(Powell),通过分析这两种算法的优缺点,提出了一种基于Powell的PSO辨识算法,通过对化工过程中某对象模型进行辨识,仿真实验证明了所提出算法的有效性和鲁棒性。研究了工业现场中常见的几种响应不规范情况,该方法能有效地解决在初值非稳态、终值非稳态、输出发散等响应不规范情形下的模型辨识问题。(2)介绍了内模控制的结构、原理和设计方法等,研究了内模控制和经典反馈控制的关系,以及滤波器的设计方法。研究了时滞环节的各种近似处理方法,对一阶Taylor近似、一阶Pade近似、二阶对称Pade近似、二阶非对称Pade近似以及全极点近似法进行了比较。基于以上五种近似方法针对一阶时滞系统和二阶时滞系统设计了内模控制器,并推导了IMC-PID控制器的参数整定公式。(3)研究了基于内模控制方法的多变量时滞系统的解耦控制问题,采用了两种内模控制方案：第一种是基于V规范型的多变量解耦内模控制,即让多变量内模控制器具有V规范型结构,使其同时具有解耦补偿器和内模控制器的作用,推导了控制器的设计方法,得到了控制器设计的一般通式；第二种方案是为了解决多变量时滞系统的传递函数矩阵难以进行因式分解的问题,通过求取过程传递函数矩阵的逆和最优对角分解矩阵,来实现闭环系统解耦以及控制器响应时滞最小的内模控制方法。最后,通过仿真和效果对比说明了这两种方法的有效性。(4)分析了传统内模控制方法在非自衡对象中应用的不足,提出了两种改进的内模控制结构。第一种方法,是在传统的内模控制结构中添加了一个比例控制器和一个比例微分控制器,分别用于镇定不稳定对象和控制系统的抗干扰特性及鲁棒特性,并采用全极点近似法对被控对象的纯滞后项进行近似处理；第二种方法,是为了克服外部干扰的影响,把反馈信号加在控制变量上,即在传统的一自由度内模控制结构的基础上添加了一个扰动估计器用于抑制干扰信号,实现了系统跟踪性能和抗扰性能的解耦控制,并有较好的动态特性。最后通过仿真实验结果证明了这两种方法的有效性。(5)针对化工过程中常见的非方系统,提出了两种非方系统的内模控制方法。第一种方法,是采用非方有效相对增益法对非方系统进行稳定性分析,选出输入输出关联性最强的最优子系统,再对最优子系统进行回路分析和配对,根据分析结果分别设计了两种内模控制器和一个强鲁棒性分散PID控制器；第二种方法,是针对含有时滞和非最小相位零点的非方耦合系统,提出了一种基于奇异值分解的解耦内模控制方法,该方法通过添加补偿项来实现对非方系统的解耦并消除不可实现因素。并根据奇异值分解法设计了一种非对角矩阵滤波器,使控制系统具有较好的高维解耦能力和强鲁棒性。最后通过仿真结果证明了这两种方法的有效性和可行性。通过对上述各种工业对象的研究,进一步发挥了内模控制方法的优势,在理论上使内模控制更为完善,在实际应用中也使其应用面更为广泛。更多还原

【Abstract】 With the rapid development of modern industry, the production process become much more complex than before, which always contains integrator and time delay process, variable structure object. And these controlled object often has the characteristics of uncertainty, time delay, multivariable coupling, input and output are constrained ect.. By using traditional internal model control method to solve these complex problems, the desired control effect can be hardly achieved. In order to fully play the advantages of internal model control, to develop some effective, simple and practical internal model control methods for a variety of special controlled objects has become one of the urgent needs for research area and modern industrial area. Based on this standpoint, taking internal model control as the basic control method, some solutions and improvements are proposed for the problems of complex systems. These solutions and improvements include the improvement of internal model control structure, the stability and robustness analysis of internal model structure, and the improvement of the filter. In this dissertation, the main study contents and the research results are as follows: (1) Represented the principles of system identification and several common identification methods. Mainly researched particle swarm optimization (PSO) and unconstrained multidimensional optimization algorithm(Powell), by analyzing the strengths and weaknesses of PSO algorithm and Powell algorithm, PSO-Powell identification algorithm was proposed. Through the simulation of chemical plant model illustrated the effectiveness and robustness of the proposed identification method. The unstandardized response situation of the industrial field had been researched, and this proposed method can carry out model identification in the condition of unsteady state initial value, unsteady final value, as well as Divergence output, ect.(2) The principle, structure and design method of internal model control are systematically represented. The relationship between IMC and classical feedback control, and the design method of filter are researched. A variety of time delay approximation methods had been researched. Based on these approximation methods internal model controllers were designed for the first order system with time delay and the second order system with time delay, derived the IMC-PID tuning formulas.(3) The decoupling and control problem of multivariable processes with time delays has been investigated. Two novel multivariable internal model control methods are adopted. The first method isⅤnorm multivariable decoupling internal model control. The multivariable internal model controller has V norm structure, which has the functions of decoupling compensator and internal model controller. The controller design procedures were deduced, and the general formula of controller design was obtained. The second method is Max-Max internal model control method for multivariable system with multiple time delays. By calculating the inverse of transfer function matrix and the optimal diagonal matrix, the closed-loop system can be decoupled and the control response with minimum time delay. Finally the simulation results illustrated the effectiveness of these two methods.(4) In view of the limitation of the traditional internal model control method in the integrator and time delay process. Two improvement IMC structure had been proposed. The first method is that a proportional controller and a proportional differential controller were added to the traditional internal model control structure, to stabilize unstable object and control the disturbance rejection performance and robustness of the system respectively. Meanwhile, All-pole approximation method was adopted to approximate the delay item of the system. The second method is that to overcome the impact of external interference, the feedback signal is added to the control variables, that is equal to added a perturbation estimator to traditional IMC structure, the decoupling control of the system was realized, and the system has good dynamic characteristics. Finally, the simulation results illustrated the effectiveness of these two methods.(5) The internal model control method for non-square system has been investigated in the dissertation. Two internal model control methods had been proposed. The first method is that to analysis the stability of non-square system by non-square effective relative gain, and chosen the optimal subsystem which has the most relevant inputs and outputs. After analyzing this optimal subsystem, two internal model controllers and one strong robustness decentralized PID controller were designed based on the analysis result. The second method is that for the non-square system with time-delay and non-minimum phase zero, based on analytical analysis and generalized inverse a decoupling internal model control was proposed. This method can realize decoupling of non-square processes and eradicating unrealizable factors by inserting compensated terms. Meanwhile, a non diagonal filter is designed on the basis of SVD matrix theory, which makes control system could bear the capacity of high-dimensional decoupling and fast response, and has strong robustness. Finally the simulation results illustrated the effectiveness of these two methods.Though the researches of the above industrial objects, the advantage of internal model control is further developed, complete the internal model control in theory, and made the application of internal model control much wider in practical application.更多还原