从非线性系统本身的物理背景出发 ,根据系统本身的内在特性、先验知识和经验建立系统辨识模型 ,提出了广义模糊神经网络 (GFNN) .文中证明了GFNN的函数逼近定理 ,并据此提出了GFNN的结构自组织和参数自学习算法 .GFNN在预设的辨识精度下能自动辨识系统的网络结构以及进行参数自学习 ,实现GFNN网络结构的真正在线自组织 .仿真结果表明 ,对于慢时变非线性对象 ,GFNN表现出了很强的非线性逼近能力 ,是模糊逻辑系统与人工神经网络两类方法的比较成功的融合 Based on the physical background of the nonlinear system, a generalized fuzzy neural network (GFNN) is proposed according to the inherent characteristics of the system itself, priori knowledge and experience, and a generalized fuzzy neural network (GFNN) is proposed. GFNN structure self-organization and parameter self-learning algorithm.GFNN can automatically recognize the network structure of the system and parameter self-learning under the preset recognition accuracy to realize the real online self-organization of GFNN network structure.The simulation results show that for the slow Variable non-linear objects, GFNN showed a strong nonlinear approximation ability, is a more successful fusion of fuzzy logic system and artificial neural network two kinds of methods