纯滞后过程模糊预测控制研究

针对复杂工业过程控制,提出一种基于T—S模糊模型的预测控制方法,从参考轨迹和可测的过程变量提取特征信息,并利用最优控制理论,构成了具有模糊模型的纯滞后预测控制系统。经过跟踪调节和定值干扰调节实验仿真,仿真结果表明基于T—S模糊模型的预测控制方法的有效性和可行性,系统的跟踪效果良好,调节品质优于单纯的线性调节器。     

纯滞后过程的预估补偿控制

根据施密斯（Ｓｍｉｔｈ）预估器原理，提出一种被控对象数学模型未知的控制系统的实现方法，克服了被控对象数学模型不准确使控制性能变坏的弊病。同时，针对被控对象为一阶惯性、纯滞后环节时的控制，根据一系列被测参数值扒出它的预估值方法，预估过程与对象的时间常数、放大系统无关。实际运行取得了良好的控制效果。     

多重滞后系统的预测控制

研究了过程控制中常见的多重滞后系统的预测控制问题．同时考虑状态变量滞后和控制作用滞后,提出了一种改进的状态反馈预测控制算法,它假设在滞后时间内状态变量未来预测值保持其当前值不变,给出了其简化的递推算法．解决了因多重状态滞后问题的存在,使状态变量的未来值利用模型直接递推很难得到的问题．最后通过仿真实例表明了算法的有效性．该算法计算量小、易于实际工程应用．     

智能预测控制在工业滞后系统中的应用

为了克服工业滞后系统难以建立准确模型、控制误差大的问题,文章采用基于神经网络的预测控制方法,实现了滞后补偿,提高了抗干扰能力,从而降低了控制误差。仿真结果表明该方法在不引起震荡的同时大大提高了系统的稳定性能。     

一种大滞后过程的智能控制

本文针对一个带有大时滞的热工系统提出了一种递阶型智能控制器，其处于低层的ＰＩＤ控制在高层的实时专家控制系统管理下工作，该控制器的反馈输入是受控不同位置处事个温度的加权平均。     

基于灰色预测的大时滞过程的控制研究

针对传统PID控制对大时滞过程控制效果不佳的情况,采用基于GM(1,1)灰色模型对被控对象系统行为进行预测,并用等维新息、提高原始数据列的光滑度和改变初始条件三者结合的方法对该模型进行改进,得到一个改进的GM(1,1)模型.同时以GM(1,1)模型的发展系数a作为决定预测步长的依据,将其与传统PID结合,组成灰色预测控制系统对大时滞过程进行控制.仿真结果表明,与传统PID控制相比,该方法具有较强的适应性和鲁棒性,控制性能也得到了较明显的改善.     

基于预测函数控制的纯滞后系统的研究

针对大纯滞后系统难以控制的问题,提出了基于预测函数控制的大纯滞后系统。以上的分析研究均在JX-300X DCS中投入实际应用,取得了满意的应用效果。     

大纯滞后系统的自适应补偿控制

针对工业过程中普遍存在的纯滞后对象的控制问题,提出了一种带误差补偿环节的模型参考自适应控制方法。仿真结果表明,这种自适应控制器对于一类大纯滞后系统的控制具有比较好的控制效果,且结构简单,有一定的鲁棒性。     

纯滞后过程的预估补偿控制

根据施密斯（Smith）预估器原理，提出一种被控对象数学模型未知的控制系统的实现方法，克服了被控对象数学模型不准确使控制性能变坏的弊病。同时，针对被控对象为一阶惯性、纯滞后环节时的控制，根据一系列被测参数值推出它的预估值的方法，预估过程与对象的时间常数、放大系数无关。实际运行取得了良好的控制效果。     

时滞系统的多步预报控制

本文针对具有模型不确定性的时变大时滞系统，提出一种带监测及切换系统的多步预报控制算法。该方法的主要设计思想是对系统在包括时滞范围在内的未来时刻的输出作出预报，依此预报值进行滚动优化和闭环极点配置，同时，增加一层监测及切换系统，以保证整个系统的稳定性。理论分析及仿真实验表明，该方法能有效地用于时变大时滞对象的控制，是一种鲁棒性好的控制算法。     

A robust/cautious model predictive controller

In process control, a significant number of problems are encountered where there are hard and soft constraints on the measured process variables and/or on the value and rate of change of the manipulated variables. On-line implementation of traditional control strategies becomes unfeasible since they cannot explicitly deal with process constraints. Model predictive control offers a viable alternative; however, a prominent issue is the behaviour of these algorithms when the prediction model does not match the actual plant. A possible solution is to formulate the problem in a linear/non-linear programming framework, using the cutting plane technique to locate the ‘worst’ plant/model mismatch at every time interval. This results in a very practical cautious predictive controller that computes the next controller action based on expected model/plant mismatch. Control of a stirred tank reactor illustrates the method, using a one-step ahead predictive controller.     

基于神经网络误差修正的广义预测控制

本文基于ＢＰ结构神经网络,对系统的建模误差进行预测,并将其与模型预测相结合构成广义预测控制算法,目的在于抑制模型失配的影响,增强广义预测控制的鲁棒性;仿真结果表明了这一算法的有效性。     

PI型广义预测平均控制器及其仿真

研究一种新型的广义预测控制方法，并在计算机上进行了仿真。在控制器的设计中，进行了滚动优化控制序列的存储与平均，并且采用了比例积分型的目标函数，所以这种控制器对未建模动态很强的鲁棒性，对含有建模误差的过程进行了仿真研究，验证了该方法的有效性。     

化学反应器温度跟踪预测函数控制的研究及应用

对以预测函数控制算法为核心并结合PID控制的先进控制算法进行了研究,针对化学反应器的温度跟踪控制问题,给出了透明控制算法及预测函数控制器的具体形式,仿真结果表明,该算法具有较好的跟踪特性,抗干扰能力及鲁棒性,同时,该算法在SUPCON-JX300集散控制系统上实现,工业现场运行取得了很好的结果。     

一种新型鲁棒广义预测自校正控制器

本文提出了一种新的广义预测自校正控制器,这种控制器是基于一种新结构的预测模型——σ^-1模型的基础上设计的,其适应能力得到极大加强．     

Improved PID controller design for unstable time delay processes based on direct synthesis method and maximum sensitivity

In this paper, a proportional-integral-derivative controller in series with a lead-lag filter is designed for control of the open-loop unstable processes with time delay based on direct synthesis method. Study of the performance of the designed controllers has been carried out on various unstable processes. Set-point weighting is considered to reduce the undesirable overshoot. The proposed scheme consists of only one tuning parameter, and systematic guidelines are provided for selection of the tuning parameter based on the peak value of the sensitivity function (M_s). Robustness analysis has been carried out based on sensitivity and complementary sensitivity functions. Nominal and robust control performances are achieved with the proposed method and improved closed-loop performances are obtained when compared to the recently reported methods in the literature.     

基于小波网络动态补偿的广义预测控制器

由于广义预测控制器依赖于表征过程的线性模型,没有考虑未建模误差的影响,随着预测长度的增加,其预测误差也增大.因此,提出用小波网络建立误差的预测模型,并对模型预测进行动态补偿,以抑制模型失配的影响,增强鲁棒性.仿真结果表明了算法的有效性.     

Predictive Compensation for Stochastic Time Delay in Network Control Systems

This paper proposes a predictive compensation strategy to reduce the detrimental effect of stochastic time delays induced by communication networks on control performance. Values of a manipulated variable at the present sampling instant and future time instants can be determined by performing a receding horizon optimal procedure only once. When the present value of the manipulated variable does not arrive at a smart actuator, its predictive one is imposed to the corresponding process Switching of a manipulated variable between its true present value and the predictive one usually results in unsmooth operation of a control system. This paper shows: 1) for a steady process, as long as its input is sufficiently smooth, the smoothness of its output can be guaranteed; 2) a manipulated variable can be switched smoothly by filtering the manipulated variable just using a simple low-pass filter. Thus the control performance can be improved. Finally, the effectiveness of the proposed method is demonstrated by simulation study.     

网络控制系统中随机传输时延的预测补偿研究

This paper proposes a predictive compensation strategy to reduce the detrimental effect of stochastic time delays induced by communication networks on control performance. Values of a manipulated variable at the present sampling instant and future time instants can be determined by performing a receding horizon optimal procedure only once. When the present value of the manipulated variable does not arrive at a smart actuator, its predictive one is imposed to the corresponding process. Switching of a manipulated variable between its true present value and the predictive one usually results in unsmooth operation of a control system. This paper shows: 1) for a steady process, as long as its input is sufficiently smooth, the smoothness of its output can be guaranteed; 2) a manipulated variable can be switched smoothly by filtering the manipulated variable just using a simple low-pass filter. Thus the control performance can be improved. Finally, the effectiveness of the proposed method is demonstrated by simulation study.