基于神经网络的受扰非线性系统最优控制及其应用

项目名称： 基于神经网络的受扰非线性系统最优控制及其应用

项目编号： No.60804005

项目类型： 青年科学基金项目

立项/批准年度： 2009

项目学科： 化学工业

项目作者： 高德欣

作者单位： 青岛科技大学

项目金额： 20万元

中文摘要： 研究受外界扰动非线性系统最优控制及其在化工工程领域的应用问题。主要研究成果如下：根据物理机理，给出了已知动态特性扰动外系统的数学模型，建立了化工过程反应器CSTR系统的非线性数学模型。研究了基于神经网络的受扰动非线性及时滞系统最优控制问题，设计扰动补偿器并给出控制器的扰动抑制优化方法；研究了基于反馈线性化的受扰动非线性及时滞系统的最优控制问题，利用微分同胚理论，将非线性系统精确线性化，并在此基础上设计扰动观测器，解决扰动不可测问题，利用LQR理论得到了系统最优控制、最优跟踪控制以及最优扰动抑制等控制算法，实现对被控对象的精确控制。为方便在化工工程等实际系统中实施，引入鲁棒性较强的PID控制器，设计从时域最优状态反馈到频域最优PID控制器参数的优化方法，获取系统最优的动态补偿网络，设计出最优PID整定参数，给出具有动态补偿反馈形式的最优扰动抑制控制算法。开发了仿真软件并对上述理论成果在CSTR模型上进行了数值仿真验证，检验了理论成果的有效性和实用性，发表了系列性学术论文，并获得山东省教育厅自然科学一等奖1项，申请专利5项，培养了6名硕士研究生。

中文关键词： 非线性系统；化工工程；最优控制；神经网络；PID控制

英文摘要： This project considers the optimal control problem for nonlinear systems with external disturbances in chemical processes. The main results are listed as follws. Based on the physical mechanism, external disturbances model is expressed by the exosystem, and mathematics models of the CSTR control systems for chemical processes are established. The theoretical issues of nonlinear systems with time-delay under disturbances based on neural network are researched, disturbances compensation controler is designed and the optimal disturbance rejection control approach is proposed. The optimal control problem for nonlinear system with time-delay based on feedback linearization are researched. Firstly, the nonlinear system model with external disturbances is converted to quasi-linear system model by differential homeomorphism,then design the optimal state-observe & disturbance-observer,and based on the theory of linear quadratic optimal control, we obtain design algorithms of optimal control, optimal tracking control, optimal disturbance rejection and other control algorithms for nonlinear systems with time-delay under disturbances. In order to facilitate the implementation of the project, the optimization of the state variables feedback in the time domain into the optimization of PID controller parametersin the frequency domain is designed, according to this method, the system optimal dynamic compensation network is obtained,the optimally tuned parameters of PID controller are designed and the realization algorithm is developed. The theoretical results are tested by numrical simulation of CSTR, and the validity and practicability are tested. We have published a series of papers, and a first natural science of Shandong Provincial Department of Education, and trained six graduate students.

英文关键词： nonlinear systems; chemical processes; optimal control; neural networ; PID control