The prediction of the inside temperature and relative humidity of a greenhouse using ANN method with limited environmental and meteorological data

Abstract

In this paper, the prediction of the internal temperature (Tin) and relative humidity (Rhin) of a greenhouse located near Agadir, Morocco using artificial neural net-work (ANN) as machine learning method. First, an analyze of correlations be-tween inputs and outputs is studied in order to select the adequate input parameters. External temperature, relative humidity and solar radiations were the parameters that have the highest correlation coefficient with the outputs. They are thus selected as the only input parameters. The prediction of Tin and Rhin with the previously cited inputs gives a perfect coefficient of correlation (R=0.996). The aim of this study is to use only one measured input parameter (external temperature) and eliminate the two environmental parameters (relative humidity and solar radiation), by introducing the factor of time as input of the ANN model. Results were very satisfying and 20 neurons was sufficient to reach a correlation of about 0.98.