Abstract
In this paper a new identification algorithm based on a model expanded by a basis of automatic choosing functions (ACF) for continuous-time nonlinear systems is proposed. The Butterworth filter is introduced as a delayed state variable filter, in order to evaluate higher order derivatives of input and output signals. A data region or a whole domain of signals is divided into some subdomains and the unknown nonlinear function to be estimated is approximately represented by a linear local equation on each subdomain. Then these linear local equations are united into a single one by the ACF of sigmoid type smoothly. The resulting model is linear in unknown parameters, which are easily estimated by the least-squares method. The model structure and the state variable filter are properly determined by genetic algorithm. Simulation results are shown to demonstrate the effectiveness of the proposed method.
