基于在线子空间辨识的自适应预测控制

针对实际工业工程中存在非线性、时变的特点,提出一种基于子空间辨识的自适应预测控制方法。利用滚动窗口在线更新R阵,得到新的预测模型参数矩阵,通过比较更新前和更新后的预测误差来决定是否更新预测模型。将此控制方法应用于2-CSTR过程控制的仿真试验,通过与自适应模糊控制、PID控制器的比较,说明了该方法的优越性。     

热风炉烟气温度广义预测控制器的设计与实现

某钢铁厂顶燃式热风炉升温过程具有滞后、时变和非线性的特点,现场热风炉烟气温度控制自动化程度低、控制品质差。为此,通过带遗忘因子的递推最小二乘法在线辨识时变模型参数,设计烟气温度升温过程的广义预测控制器。仿真和投运结果表明:控制效果良好,热风温度提高了3~5℃,同时节省了1.5%左右的煤气流量。     

非线性时变系统的小波预测控制

从小波分析原理出发,给出了一种正交小波网络辨识非线性动态系统的方法,并以此网络作为预测模型,同时通过递推最小二乘法在线修正模型;采用遗传算法优化控制器输出,通过对遗传算子和优化指标的修改,减少了在线寻优计算量.仿真实验表明该算法对非线性时变系统有良好的控制效果.     

一种基于VRFT的全格式无模型自适应控制

针对无模型自适应控制中的时变参数拟梯度向量估计值信息没得到充分利用的问题,提出一种基于VRFT（虚拟参考反馈调整方法）的无模型自适应控制方法.该方法利用VRFT中的参考模型思想,首先将非线性系统经全格式线性化后的线性模型与无模型自适应控制器构成的闭环传递函数作为参考模型,再以参考模型的输出与系统期望输出的误差作为控制器的输入,从而将拟梯度向量在过去时刻的估计值引入到无模型自适应控制设计中,提高了伪偏导数估计值信息利用率.同时消去控制律中的权重因子,避免因人为赋值而影响控制效果.     

化学反应器的神经元自适应观测和非线性控制

提出了一种基于神经网络的自适应观测和非线性控制策略,证明了自适应观测器的收敛件和非线性控制系统的稳定性,将其用于连续搅拌釜式放热反应器的浓度控制。根据可在线测量的反应温度,在线估计不可在线测量的反应物浓度和辨识Arrhenius指前因子,并利用重构的状态信息设计出带约束的非线性控制策略。仿真结果表明,观测器/控制器的组合提供了满意的闭环特性,证实了本文方法的有效性。     

基于神经网络的非线性时变对象自适应控制方法研究

针对工业控制中的非线性、时变和滞后被控对象,将神经网络辨识器和PID控制技术有机结合,构建自适应PID控制器,并引入知识先导学习算法,提高时间效率。仿真分析表明:该自适应PID控制器能够自动辨识被控对象模型、自整定控制参数,在辨识误差、响应时间及超调量等指标上均有良好表现。     

基于FGPC的单元机组协调控制系统研究

针对单元机组的大迟延、强耦合、参数时变且不确定性的特点,将T—s模糊模型引入预测控制中,作为预测模型。首先,用改进的模糊c一均值聚类算法和随机牛顿法辨识得到非线性系统的T-s模型;然后基于线性化后的系统模型设计模糊广义预测控制器,并对非线性对象进行在线控制。仿真结果表明：FGPC对于时变的非线性系统具有很好的控制效果。     

化工过程强非线性系统的变模型自适应预测控制

提出一种变模型自适应预测控制算法 ;基于非线性状态空间模型 ,通过每步在当前工作点 (非平衡点 )线性化获得线性化子模型 ,以此进行状态反馈预测控制 ,线性化子模型随工作点变化 ,且不限于平衡点。通过pH值控制的对比仿真实验 ,证明其对强非线性过程的控制效果优于传统的多模型预测控制。最后分析讨论了该控制算法存在的几个重要问题 ,并指出与之相关的未来研究方向     

新一代的自适应模型预测控制器

提出了新一代的自适应模型预测控制器,自适应MPC控制器由MPC控制模块、在线辨识模块、性能监控模块3个模块组成,相互协调配和来实现自适应MPC控制。除了控制器功能设计以外,其余过程均可自动进行。对于新建MPC应用,首先进行多变量测试与辨识,在模型符合控制要求时,自动进入控制器投运。通过控制器性能监视发现模型不满足控制要求精度时,触发一次多变量模型测试与辨识过程,替换原有模型进行控制,保证控制器性能始终处于最佳状态。自适应MPC控制器在PTA装置上的应用表明了算法的有效性。     

暖通空调系统基于SMITH预估自校正控制算法

目前暖通空调控制系统中普遍存在运行效率低下、能耗浪费严重的现象,主要原因在于暖通系统本身具有时变、时滞等非线性特性,使得传统的控制方法无法取得良好的控制性能。如何克服暖通空调系统时变和时滞特性所带来的不利影响,是提高暖通空调系统控制回路性能的重要途径之一。本文设计了一种基于SMITH预估的自校正控制算法,首先通过在线辨识包含时滞参数在内的暖通空调对象模型,进而采用SMITH预估器对时滞进行补偿,而SMITH预估器中的PI控制器则用于提高鲁棒性,并实现控制器参数的实时更新。仿真结果表明该算法性能要优于传统的控制方法。     

CONTROL OF A CONTINUOUS POLYMERIZATION REACTOR BY WIENER INPUT/OUTPUT DATA-BASED PREDICTIVE CONTROLLER WITH DIRECT INVERSE IDENTIFICATION

This article reports an experimental study for the identification and predictive control of a continuous methyl methacrylate (MMA) solution polymerization reactor. The Wiener model was introduced to identify the polymerization reactor in a more efficient manner than the conventional methods of Wiener model identification. In particular, the method of subspace identification was employed and the inverse of the nonlinear part was directly identified. The input variables in this work were the jacket inlet temperature and the feed flow rate, while the monomer conversion and the weight average molecular weight were selected as the output variables. On the basis of the identified model a Wiener-type input/output data-based predictive controller was designed and applied to the property control of polymer product in the continuous MMA polymerization reactor by conducting an on-line digital control experiment with online densitometer and viscometer. Despite the complex and nonlinear characteristics of the polymerization reactor, the proposed controller was found to perform satisfactorily for property control in the multiple-input multiple-output system with input constraints for both set-point tracking and disturbance rejection. This was also confirmed by simulation results.     

CONTROL OF A CONTINUOUS POLYMERIZATION REACTOR BY WIENER INPUT/OUTPUT DATA-BASED PREDICTIVE CONTROLLER WITH DIRECT INVERSE IDENTIFICATION

This article reports an experimental study for the identification and predictive control of a continuous methyl methacrylate (MMA) solution polymerization reactor. The Wiener model was introduced to identify the polymerization reactor in a more efficient manner than the conventional methods of Wiener model identification. In particular, the method of subspace identification was employed and the inverse of the nonlinear part was directly identified. The input variables in this work were the jacket inlet temperature and the feed flow rate, while the monomer conversion and the weight average molecular weight were selected as the output variables. On the basis of the identified model a Wiener-type input/output data-based predictive controller was designed and applied to the property control of polymer product in the continuous MMA polymerization reactor by conducting an on-line digital control experiment with online densitometer and viscometer. Despite the complex and nonlinear characteristics of the polymerization reactor, the proposed controller was found to perform satisfactorily for property control in the multiple-input multiple-output system with input constraints for both set-point tracking and disturbance rejection. This was also confirmed by simulation results.     

Adaptive linear quadratic gaussian (LQG) control of a bioreactor

The paper deals with the modelling and adaptive control of a continuous-flow fermentation process for the production of alcohol. The fermenter model has been developed from mass balance and leads to nonlinear differential equations. In practice, control strategies are difficult to derive using this non-linear model. The dilution rate and the substrate concentration have been considered as control and controlled variables, respectively. The adaptive control algorithm implemented is based on the linear quadratic control approach, where the associated Riccati equation is iterated until the system closed-loop poles belong to a restricted stability domain which is included in the unit circle. A single input/output model is used for control purposes. The model parameters are estimated on-line using a robust identification algorithm which includes: data normalization, time-varying forgetting factor, covariance matrix factorization, etc. Experimental results show the performance of this adaptive scheme and its ability to control biotechnological processes.     

基于快速滑窗QR分解的自适应子空间辨识

子空间辨识直接由输入输出数据辨识得到过程状态空间模型,在多变量系统的辨识中取得广泛应用。实现在线子空间辨识算法的关键在于快速、高效的QR分解及SVD分解更新算法。通过将Updating和Downdating操作有效结合,提出了一种快速的滑窗QR分解算法,减少了不必要的重复步骤,进一步提高了计算效率。复杂度分析结果表明,随数矩阵行数增加,快速滑窗QR分解算法比Updating、Downdating两步法可以减少8.3%的计算量。将快速滑窗QR分解算法用于PO-MOESP子空间辨识算法的自适应更新,并通过数值仿真实例验证了算法的有效性。     

基于广义预测控制策略的微生物燃料电池控制

针对MFC系统启动阶段输出响应不稳定以及调节时间较长的问题,结合微生物燃料电池自身特性,提出了基于广义预测控制(generalized predictive control,GPC)的微生物燃料电池(microbial fuel cell,MFC)控制策略。与加入PID控制方法对比得知,加入GPC的MFC系统输出能够避免响应出现大幅度的抖动,且响应速度快,动态调节鲁棒性好,保证了动态输出曲线快速准确地跟踪系统设定值。在给定外电阻为恒值和醋酸盐浓度随时间阶梯变化时,通过带遗忘因子的最小二乘法进行模型辨识,将所得线性模型作为预测模型,采用GPC算法进行控制。仿真表明,GPC能在控制响应速度方面取得好的控制效果以及系统调节过程中的鲁棒性也有了较大的改善。有效地实现了对微生物燃料电池系统的动态性能以及鲁棒性能的优化,验证了所提出的算法有效可行。     

基于动态补偿逆的非线性RBF内模控制及其应用

针对非线性不确定系统和传统的非线性内模控制在控制上存在的不足,提出一种基于动态补偿逆的非线性不确定系统RBF内模控制,在引入RBF建立逆模型的同时,将无模型自适应控制方法作为附加控制器,用于模型偏离被控对象时在线修正逆模型。仿真结果表明,本文提出的方法不仅对系统的常量摄动具有较好的鲁棒性,对时变不确定性仍然保持较好的跟踪效果,具有较好的实时性、鲁棒性和在线校正功能。     

基于自适应模糊神经网络的非线性系统模型预测控制

针对非线性动态系统的控制问题,提出了一种基于自适应模糊神经网络(adaptive fuzzy neural network, AFNN)的模型预测控制(model predictive control, MPC)方法。首先,在离线建模阶段,AFNN采用规则自分裂技术产生初始模糊规则,采用改进的自适应LM学习算法优化网络参数;然后,在实时控制过程,AFNN根据系统输出和预测输出之间的误差调整网络参数,从而为MPC提供一个精确的预测模型;进一步,AFNN-MPC利用带有自适应学习率的梯度下降寻优算法求解优化问题,在线获取非线性控制量,并将其作用到动态系统实施控制。此外,给出了AFNN-MPC的收敛性和稳定性证明,以保证其在实际工程中的成功应用。最后,利用数值仿真和双CSTR过程进行实验验证。结果表明,AFNN-MPC能够取得优越的控制性能。     

Recursive subspace identification with prior information using the constrained least squares approach

It is essential to develop high quality models for process control and other applications. The incorporation of prior information in subspace identification has been investigated to obtain improved model quality. One of the recent developments incorporates the prior information using the constrained least squares (CLS). In many online applications, the amount of process data for model identification grows with time, and it is therefore necessary to develop a recursive algorithm for online identification of process models and to address the time-varying characteristics of the systems. In this paper, a recursive subspace identification algorithm incorporating prior information is developed using the constrained recursive least squares (CRLS). It is shown via a simulation example that the state space model identified using the proposed algorithm possesses improved accuracy.