Identification of Wiener Model Using Least Squares Support Vector Machine Optimized by Adaptive Particle Swarm Optimization

摘要

In this paper, a novel approach is proposed for identifying Wiener model, in which the linear dynamic subsystem is represented by autoregressive exogenous model and the nonlinear static function by least squares support vector machine (LSSVM). The identifications of Wiener model are accomplished in a two-step procedure, in which the linear part is identified separating from that of nonlinear part via a small input signal. To obtain better approximation of static nonlinear function, a new adaptive particle swarm optimization is proposed to select the optimal hyper-parameters of LSSVM. Simulation results show that the proposed method is effective for identifying Wiener model...