Parameter estimation of fuzzy neural network controller based on a modified differential evolution

  • Authors:

  • Hung-Ching Lu;Ming-Hung Chang;Cheng-Hung Tsai

  • Affiliations:

  • Department of Electrical Engineering, Tatung University, Taipei Taiwan;Department of Electrical Engineering, Tatung University, Taipei Taiwan;Department of Electrical Engineering, China University of Science and Technology, Taipei Taiwan

  • Venue:
  • Neurocomputing
  • Year:
  • 2012

Quantified Score

Hi-index 0.01

Visualization

Abstract

A tracking control of a nonlinear system is proposed in this paper via a fuzzy neural network (FNN) controller based on a modified differential evolution (MDE). The proposed modified differential evolution fuzzy neural network controller (MDEFNN) is composed of an FNN identifier, a hitting controller, a computation controller and a MDE estimator. First, the FNN identifier is used to estimate parameters of the nonlinear system. In order to compensate the uncertainties of the system parameters and achieve robust stability of the considered system, the hitting controller is adopted. The computation controller is used to sum up the outputs of the FNN identifier and hitting controller. Furthermore, there are two main learning phases in MDEFNN controller - the training phase and the online phase. In training phase, the mutation operation of the proposed MDE estimator according to fitness function effective produces a mutation vector. The MDE estimator is adopted to estimate the parameters of the MDEFNN controller. Therefore, there are several parameters such as the learning rates of the back-propagation (BP) algorithm, the parameters of error terms which are used in BP algorithm. The initial values of the FNN identifier and some preset parameters of MDEFNN controller can also be estimated by MDE estimator. After the best preset parameters are obtained, the nonlinear system is controlled by using MDEFNN controller. Further, the online parameter learning of the FNN identifier is based on the BP algorithm using error terms in the online phase. Finally, the simulation results are provided to demonstrate robustness, effectiveness and accurate tracking performance of the proposed MDEFNN controller under the conditions of external disturbance.