Temperature Prediction of a Greenhouse Equipped with Evaporative Cooling System Using Regression Models and Artificial Neural Network (Case Study in Kerman City)

Today's agriculture, greenhouse cultivation plays a key role in increasing the quantity and quality of products. Indoor conditions of the greenhouse depend on some external factors, which are usually not easily predictable. The purpose of this study was to estimate the air temperature inside the greenhouse in two modes of ventilation (non-ventilated conditions and evaporative cooling system) using artificial neural network and regression models. Some factors such as solar irradiance, ambient temperature, northern wall temperature and flow rate and temperature of the cooling air were employed as the inputs of the models. Verification of the models was conducted using statistical criteria of mean square error, correlation coefficient and mean absolute percentage error. In order to train the neural network from multilayer perceptron with the algorithm of post-error learning and using the Levenberg-marquart training algorithms, the Bayesian regression and the gradient conjugate scalar, and in the regression model of the progressive and forward method for determining regression equations were used. Comparison of the statistical criteria indicated that the artificial neural network method predicted the greenhouse temperature with a higher accuracy than the regression model.