Improved Intelligent Identification of Hammerstein-Wiener Systems by Particle Swarm Optimization and K-Means Clustering

Abstract

This paper considers the robust identification of Hammerstein-Wiener systems in the presence of Gaussian or non-Gaussian noises. An improved intelligent identification scheme is exploited by combining particle swarm optimization (PSO) and K-means clustering. The proposed scheme has strong ability to keep the balance between exploration and exploitation. Its procedure is about “global particle swarm optimization search — K-means clustering — local particle swarm optimization search”. The proposed scheme can identify the parameters of the general Hammerstein-Wiener system with dead zone and saturation characteristics, and obtain a more accurate model for the actual production process. Relative to other improved particle swarm optimization methods, the accuracy of parameter estimation is improved by nearly 53 % at data length L=2000. In particular, the method can better model nonlinear dynamics and facilitate the precise implementation of control in chemical production.