Non-linear robust identification of a greenhouse model using multi-objective evolutionary algorithms

This paper presents a non-linear climatic model (temperature and humidity), based on first-principles equations, of a greenhouse where roses are to be grown using hydroponic methods. Fitting of model parameters (15 in all) is based on measured data collected during summer in the Mediterranean area. A multi-objective optimisation procedure for estimating a set of non-linear models ΘP (Pareto optimal), considering simultaneously several optimisation criteria, is presented. A new multi-objective evolutionary algorithm, phi-MOGA, has been designed to converge towards Θ^*p, a reduced but well distributed representation of ΘP since good convergence and distribution of the Pareto front J(ΘP) is achieved by the algorithm. The set Θ^*p can be used as the basis to choose an optimal model that offers a good trade-off among different optimality criteria that have been established. The procedure proposed is applied to the identification and validation of the greenhouse model presented in the paper.