Dynamic system modeling based on wavelet recurrent fuzzy neural network

In this paper, Combined recurrent neural network and wavelet-based fuzzy neural network, A new wavelet recurrent fuzzy network (WRFNN) is presented. In order to simplify parameters identification and improve model generalization ability, The premise and consequent coefficients are optimized separately. The premise parameters are optimized by LM algorithm, at the same time the consequent coefficients are updated by recursive least square estimation. Simulation results of a nonlinear dynamic system and a CSTR system modeling show that the WRFNN can catch system dynamic real-time.