A technology system for monitoring and predicting climate variables in an experimental greenhouse using IoT and Machine Learning

Nowadays, technology in agriculture allows the construction of intelligent greenhouses through control of the climatic variables to improve the quantity and quality of products. This work presents the implementation of a system for monitoring and predicting temperature and humidity to maintain an appropriate microclimate for cultures in an experimental greenhouse in the Tunshi station of the Escuela Politécnica de Chimborazo (ESPOCH), Ecuador. The electronic system comprises 4 sensors, an embedded board, IoT tech to store data in a local web server, a mobile app to enable remote access to evaluate in real-time, and a Machine Learning algorithm to predict these variables for decision-making to robust control of the microclimate. The control system switches on 2 single-phase squirrel cage motors to raise or lower the side curtains according to the climate variables of a specific crop. Results show 100% the integrity of information for wireless communication. The absolute error for temperature and humidity measured is 1.17%, and 1.56% between sensors and pattern equipment. The mean time spent raising and lowering the curtains is 20 seconds. The LSTM and CNN algorithms show R2 greater than 80%, and MSE is less than 2.0e-3 to train and test processes. This yields promising results to be accessible for students and local farmers due to the economic comparison and this system could be implemented for large-scale applications.