Identification of a reduced order model for greenhouse control based on CFD model results and experimental validation

The use of numerical models in the agricultural industry is becoming more common to solve different issues, for example design optimization or model based control design. Nevertheless, such numerical models are computationally expensive and impractical for control design. This work shows a methodology for obtaining a reduced order model that predicts the thermal behavior of a greenhouse, in order to design control systems to regulate the behavior of the internal microclimate of greenhouses. This model is obtained through system identification applied to the data obtained from a numerical greenhouse (NG), modelled using computational fluid dynamics (CFD) through Ansys Fluent 2019R3 software. The CFD model is experimentally validated by comparing the temperature distribution inside the greenhouse when turning on and off a set of fans, obtaining a 1.28% root mean square error. Subsequently, numerical simulations of turning on and off of fans are carried out to obtain temperature data and to obtain a model that relates the on time with the average temperature inside the greenhouse. Different structures were used in the IDENT tool of Matlab, where the best fit was found for a simple linear function with delay with 61% of relation to the simulated data from the NG and 62% when compared with experimental data.