Temperature Prediction in Chinese Solar Greenhouse Based on Artificial Neural Networks Using Environmental Factors

Predicting environmental factors within the indoor agriculture environment, namely greenhouses, is important as they play essential roles in improving yields and reducing energy consumption. In particular, predicting temperatures because plants are sensitive to extreme cold and overheating could lead to potential losses of the crops. Specifically, the Chinese Solar Greenhouses (CSG) are completely passive, meaning that it is only heated by the sun and has no controllable heating system. Therefore, it is important to create an accurate prediction model for predicting the temperate in the CSG several hours before. By predicting the CSG temperature, suitable actions could be taken, such as rolling up/down the covering material and opening/closing the top and the bottom side for adjusting the ventilation to avoid losses of the crops. In this paper, a Multilayer Perceptron (MLP) neural network with the Levenberg-Marquardt (LM) algorithm to optimise the structural parameters of MLP further improves the prediction accuracy. The proposed LM-MLP model is tested and evaluated based on two real datasets collected from CSG during warm and cold seasons. The obtained results illustrated the ability of the proposed LM-MLP to predict the maximum and minimum temperature during two different seasons (warm and cold) with more accuracy compared to some of the existing prediction models. The performance of the proposed method is compared with the standard NNs and SVM techniques.