A data-driven modeling method using particle swarm optimization

Most of process systems such as chemical plants are considered as nonlinear systems. The global linear approximation for the systems with the strong nonlinearities might cause the large modeling error. It is then difficult to obtain the good control performance, even if the controllers are suitably designed based on the models. In this paper, a data-driven modeling method using particle swarm optimization has been proposed. In the proposed method, local linear models are designed with the multiple data-sets selected from the database when needed. Also, the time-variant system parameters are automatically adjusted by using the particle swarm optimization. Finally, the effectiveness of the proposed method is numerically evaluated through applications to the nonlinear systems and the time-variant systems.