Identification of dynamic systems using a differential evolution-based recurrent fuzzy system

This work presents the development of a simulation model based on a recurrent fuzzy system with structure and parameter identification by a differential evolution algorithm. The proposed model is formulated by state space equation, in which the state transition function is a recurrent fuzzy system with two feedback connections and adjustable delay operators and the output function is a linear function of the states. The identification process relies on two instances of the differential evolution algorithm in a hierarchical fashion. The outermost is considered for combinatorial structure optimization and the innermost for optimization of continuous parameters. The new model is evaluated in some benchmark problems and the results showed the model achieved good numerical performance. Moreover, the results demonstrated the ability of differential evolution algorithm to optimize both the parameters as well as the structure of the model.