作　　者： ; ; ;

机构地区： 华南理工大学电力学院

出　　处： 《控制理论与应用》 2011年第11期1645-1650,共6页

摘　　要： 变论域模糊控制器的控制函数被"复制"到后代中,往往存在着"失真"现象,这种现象的后果是造成算法本身的误差.针对这一问题,本文提出了一种基于Q学习算法的变论域模糊控制优化设计方法.本算法在变论域模糊控制算法基础上提出了一种利用伸缩因子、等比因子相互协调来调整论域的构想,且通过用Q学习算法来寻优参数使控制器性能指标最小,使其在控制过程中能够降低"失真率",从而进一步提高控制器性能.最后,把算法运用于一个二阶系统与非最小相位系统,实验表明,该算法不但具有很好的鲁棒性及动态性能,且与变论域模糊控制器比较起来,其控制性能也更加提高. When the control function of the variable-universe fuzzy controller is ＂copied＂ to the offspring,there usually exist some ＂distortion＂ phenomena which lead to the error of the algorithms.To deal with this problem,we present a novel optimal method of variable-universe fuzzy control based on Q-learning algorithms.This method adjusts the universe by the contraction-expansion factor and geometric proportional factors,and optimizes the parameters through Q-learning algorithms to minimize the performance index of the controller for reducing the ＂distortion rate＂ in the control process and improve the control performance.This method has been applied to a second-order linear system with non-minimum phase,resulting in desirable robustness and dynamic performance.The control performance is even better than that of the variable universe fuzzy controller.

关 键 词： 变论域模糊控制 学习算法 伸缩因子 等比因子

领　　域： [电气工程]