Predicting Greenhouse Microclimatic Parameters Using a Deep Learning Algorithm

Providing proper conditions for plant growth in the greenhouse requires precise management of resources concerning operating costs. Consequently, an automatic and efficient greenhouse weather control system is needed for accurate management and cost reduction. Traditionally, dynamic models have been valuable tools for controlling the greenhouse climate. In this research, the design of a system for predicting the environmental conditions of the greenhouse was studied using deep learning. The developed method was implemented to ensure precise conditions for the production of tomato crops in a glass greenhouse. The deep learning-based model successfully predicted the greenhouse temperature, relative humidity, and carbon dioxide concentration using inputs such as wind speed, the virtual sky temperature, cumulative outside global radiation, outside photosynthetically active radiation, outside temperature, outside relative humidity, and outside carbon dioxide concentration, with coefficients of determination of 0.81, 0.61, and 0.85, respectively. The performance of the deep neural network was significant due to the utilization of precise data controlled by expert operators. Compared to dynamic modelling, the advantages of the suggested framework include high stability, adaptability for use without the need for a previous model, the ability to make unlimited decisions, and low complexity in real-time training. Therefore, smart artificial intelligence methods can lead to finding the best solution for optimal greenhouse control, enhancing performance, and reducing costs while addressing other limitations.