State Identification Based on Dynamic T-S Recurrent Fuzzy Neural Network Observer

Abstract

Traditional fuzzy neural network is a static map, not suitable for induction motor state identification. To improve the accuracy of system identification, a dynamic TS recurrent fuzzy neural network observer was proposed. The dynamic back-propagation algorithm was derived from dynamic recurrent neural network observer model, which using Lyapunov Theorem to prove that the observer with global convergence. Simulation results show that: Because dynamic TS recurrent fuzzy neural network observer use the current data and historical data for state recognition at the same time, it has wonderful performance in the recognition accuracy and stability and better convergence than the traditional fuzzy neural network observer.