摘要
针对非线性动态系统辨识和控制的特点,对4层模糊神经网络进行了优化和改进.形成了动态模糊神经网络,提高了网络的稳定性和对动态系统的辨识能力,同时给出了基于Lyapunov函数稳定收敛定理的各权向量以及权矩阵学习速率的自适应调整算法.应用于非线性动态系统的辨识和控制仿真试验表明,改进后的动态模糊神经网络与模糊神经网络相比,可取得更好的辨识精度和跟踪控制效果.
A new dynamic recurrent fuzzy-neural network (DRFNN) is proposed based on the fuzzy neural networks (FNN). In inputting layer, a special mapping function is adopted to guarantee the stability when DRFNN is applied to control systems. Also, a dynamic recurrent layer is used to improve the stability and identification ability. An adaptive parameter learning algorithm is deduced based on Lyapunov function. Simulation results demonstrated that DRFNN is better than FNN used in identification and control of nonlinear dynamic systems.
出处
《控制与决策》
EI
CSCD
北大核心
2005年第2期226-229,共4页
Control and Decision
基金
国家自然科学基金项目(69874037).
关键词
动态模糊神经网络
控制
自适应学习算法
非线性动态系统
Adaptive algorithms
Computer simulation
Fuzzy sets
Learning algorithms
Lyapunov methods
Multilayer neural networks
Nonlinear control systems
Pattern recognition systems
