一种基于模糊分类的模糊神经网络辨识方法

针对非线性辨识问题，基于改进的T-S模型，提出一种自适应模糊神经网络模型(AFNN)。首先，基于模糊竞争学习算法确定系统的模糊空间和模糊规则数，并得出每个样本对每条规则的适用程度。其次，利用卡尔曼滤波算法在线辨识AFNN的后件参数。AFNN具有结构简洁，逼近能力强，能够显著提高辨识精度，并且辨识的模糊模型简单有效。最后，将该AFNN用于非线性系统的模糊辨识，仿真结果验证了该方法的有效性。

Fuzzy Neural Network Identification Method Based-on Fuzzy Clusters

In accordance with modified T-S model,this paper proposes a adaptive Fuzzy Neural Network model.First,this network is utilized to determine the fuzzy space structure of system and the number of fuzzy rules based on fuzzy competitive learning algorithm and obtains the fitness degree of every rule contrast to every sample.Further, the parameters of AFNN are on-line identified by means of Kalman filtering algorithm.The proposed AFNN has the simple model structure?the ability of universal approach and improves greatly the precision of identification.The identified fuzzy model has the advantages of simplicity and effectiveness.The AFNN is applied to the fuzzy identification for a nonlinear system and the simulation results demonstrate the effectiveness of the proposed method.