Open Access
ARTICLE
Modified PSO Algorithm on Recurrent Fuzzy Neural Network for System Identification
Chung Wen Hung, Wei Lung Mao, Han Yi Huang
Graduate School of Engineering Science and Technology and Department of Electrical Engineering, National Yunlin University of Science and Technology, 123 University Road, Section 3, Douliou, Yunlin 64002, Taiwan
* Corresponding Author: Chung Wen Hung,
Intelligent Automation & Soft Computing 2019, 25(2), 329-341. https://doi.org/10.31209/2019.100000093
Abstract
Nonlinear system modeling and identification is the one of the most important
areas in engineering problem. The paper presents the recurrent fuzzy neural
network (RFNN) trained by modified particle swarm optimization (MPSO)
methods for identifying the dynamic systems and chaotic observation
prediction. The proposed MPSO algorithms mainly modify the calculation
formulas of inertia weights. Two MPSOs, namely linear decreasing particle
swarm optimization (LDPSO) and adaptive particle swarm optimization (APSO)
are developed to enhance the convergence behavior in learning process. The
RFNN uses MPSO based method to tune the parameters of the membership
functions, and it uses gradient descent (GD) based scheme to optimize the
parameters of the conclusion part of the fuzzy system. The effectiveness of our
method is evaluated for three nonlinear system modelling and signal prediction,
including Henon system, nonlinear plant system and Mackey-Glass time series.
Simulation results show that the proposed RFNN with LDPSO algorithm can
provide more effective and accurate identification performances compared with
the APSO method in term of mean squared error (MSE).
Keywords
Cite This Article
C. W. Hung, W. L. Mao and H. Y. Huang, "Modified pso algorithm on recurrent fuzzy neural network for system identification,"
Intelligent Automation & Soft Computing, vol. 25, no.2, pp. 329–341, 2019. https://doi.org/10.31209/2019.100000093