Nonlinear System Identification using Opposition Based Learning Differential Evolution and Neural Network Techniques

Abstract

The slow convergence and local minima problems associated with neural networks (NN) used for non-linear system identification have been resolved by evolutionary techniques such as differential evolution (DE) combined with Levenberg Marquardt (LM) algorithm. In this work the authors attempted further to employ an opposition based learning in DE, known as opposition based differential evolution (OBDE) for training neural networks in order to achieve better convergence of DE. The proposed OBDE together with DE and neuro-fuzzy (NF) approaches to non-linear system identification has been applied for identification of two non-linear system benchmark problems. Results presented clearly demonstrate that the OBDE-NN method of non-linear system identification provides excellent identification performance in comparison to both the DE and NF approaches.