基于智能在线虚拟参考反馈整定的控制方法研究

项目名称： 基于智能在线虚拟参考反馈整定的控制方法研究

项目编号： No.61304031

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 王灵

作者单位： 上海大学

项目金额： 22万元

中文摘要： 虚拟参考反馈整定(Virtual Reference Feedback Tuning，VRFT)是一种新颖的离线数据驱动控制方法。本项目针对VRFT现有的理论瓶颈与应用难题，提出智能VRFT方法，通过将VRFT从传统的参数辨识问题扩展为一个全局优化问题，采用改进的智能优化算法对其虚拟参考模型、控制器结构与参数协同全局优化，解决虚拟参考模型与控制器结构设计缺乏理论指导的不足，真正意义上实现VRFT最优控制器设计。在此基础上给出VRFT闭环控制稳定约束条件、优化算法流程并引入启发式策略，提出闭环稳定与在线VRFT方法，解决现有VRFT方法无法保证闭环稳定和难以有效应用于非线性、时变对象的缺陷。最终在电站控制中验证改进提出的智能VRFT方法。本项目从全局优化的角度思考、研究VRFT方法，采用智能优化算法进行求解，为VRFT的研究提供了一个全新的思路、方法与实现手段，具有重要的理论意义与应用前景。

中文关键词： 数据驱动；虚拟参考反馈整定；智能虚拟参考反馈整定；人类学习优化算法；

英文摘要： Virtual Reference Feedback Tuning (VRFT) is a novel off-line data-driven control method. To solve the theoretical problems of VRFT and extend its applications, a new intelligent VRFT is proposed in this project where VRFT is considered as a global optimization problem instead of an identification problem in the previous research work. Compared with the traditional VRFT methods, the virtual reference model (VRM) and the structure of controllers (SoC) of VRFT are also introduced and optimized by intelligent optimization algorithms together with the parameters of controllers, which can solve the problems on the choice of VRM and SoC and achieve the globally best control performance. By further developing the constrains of the closed-loop stability of VRFT, simplifying the process of VRFT and introducing heuristic strategies, we will present the closed-loop stable VRFT method and the online VRFT method which can guarantee the closed-loop stability of the designed controllers and control the time-variant nonlinear objects effectively and efficiently, respectively. Finally, the presented intelligent VRFT methods are improved and validated in power plant control. In this project, we study the VRFT methods as optimization problems and solve them by the improved intelligent optimization algorithms, which creates a brand

英文关键词： Data-driven；Virtual Reference Feedback Tuning；Intelligent Virtual Reference Feedback Tuning；Human Learning Optimization；