Robust identification of non-linear greenhouse model using evolutionary algorithms

This paper presents the non-linear modelling, based on first principle equations, for a climatic model of a greenhouse and the estimation of the feasible parameter set (FPS) when the identification error is bounded simultaneously by several norms. The robust identification problem is transformed into a multimodal optimization problem with an infinite number of global minima that constitute the FPS  . For the optimization task, a special evolutionary algorithm (εε-GA) is presented, which characterizes the FPS by means of a discrete set of models that are well distributed along the FPS  . A procedure for determining the norm bounds, such that FPS≠∅FPS≠∅, is presented.