TSK动态网络及其在非线性动态系统中的应用  

TSK-DRFNN and Its Application in Nonlinear Dynamic Systems

作　　者：徐春梅[1] 尔联洁[2] 

机构地区：[1]北京交通大学电气工程学院,北京100044 [2]北京航空航天大学电气工程与自动化学院,北京100083

出　　处：《系统仿真学报》2006年第8期2358-2361,2365,共5页Journal of System Simulation

基　　金：国家自然科学基金(60304005)

摘　　要：针对非线性动态系统特点,提出了一种新型的基于TSK模糊模型的动态回归模糊神经网络DRFNN(Dynamic recurrent fuzzy neural networks),并给出了网络参数的迭代算法和基于李亚普诺夫稳定理论的收敛性证明。该动态回归网络由静态网络和内反馈动态回归网络组成,在结构上更好的拟合了非线性动态系统特点,应用于非线性动态系统的辨识和控制的试验结果也说明该动态回归模糊神经网络对解决非线性动态系统辨识和控制问题的有效性。A novel fuzzy neural networks -DRFNN based on TSK fuzzy model was proposed to nonlinear dynamic control systems. The premise and defuzification part is static networks while the consequent part is recurrent neural networks realized by llR filter. Beside the parameters, BP algorithm and convergence theory based on lyapunov methods were suggested. The structure of the proposed dynamic model is similar to that of nonlinear dynamic systems, so it can get better result when it applies to identification and control of nonlinear dynamic systems. Simulation results compared with TSK-FNN and PID were given and discussed indicting the effectiveness of the TSK-DRFNN.

关 键 词：非线性控制系统 辨识 控制 模糊神经网络 稳定性 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]