PREDICCIÓN SIMULTÁNEA DE TRES VARIABLES CLIMÁTICAS EN UN INVERNADERO

Abstract

The complexity of the greenhouse crop production, together with the environmental variables interaction, has encouraged the development of various models to predict and simulate variables in order to manage the greenhouse in an efficient way. One of the models that has been used successfully in prediction is the Artificial Neural Networks (ANN). The purpose of this paper was to predict the greenhouse inside conditions based on the outside environmental conditions. A data set of 19,960 values was used from the experimental greenhouse naturally ventilated, without CO2 enrichment, at Humboldt University of Berlin, from August 16 to October 24, 2007 and include the external variables: solar radiation, air temperature, wind velocity, relative humidity, and carbon dioxide concentration, heat transfer by heating system and ventilation opening; and the internal variables: air temperature, relative humidity and CO2 concentration. A three layer artificial neural network was trained and tested and validated using back conjugate gradient back propagation algorithm with a hyperbolic tangent function and momentum algorithm. The predicted values obtained from the ANN model were close to the measured values. These results showed that ANN model learnt the behavior and interactions between all variables. The aid of the ANN model developed was the simultaneous prediction of temperature, relative humidity and CO2 concentration inside of the greenhouse, which will be helpful in checking the accuracy of sensor readings.