TinyML-powered ensemble modeling for greenhouse climate control using XGBoost and LightGBM

The cultivation of crops in smart greenhouses is experiencing a profound transformation, fueled by cutting-edge technological advancements in environmental control that significantly improve efficiency, sustainability, and productivity. Nonetheless, the intricate and ever-changing dynamics of microclimate conditions pose challenges in customizing environments to satisfy the specific requirements of various plants. Accurate prediction of these microclimate parameters emerges as a promising solution to this challenge. This study explores the integration of machine learning and TinyML platforms to create a groundbreaking ensemble approach for effectively forecasting microclimate conditions. We obtained exceptional prediction accuracy for temperature (R2 = 0.9972) and humidity (R2 = 0.9976) using a stacking ensemble of XGBoost and LightGBM models. We used Optuna for accurate hyperparameter optimization and thoroughly examined the best possible input variable combinations as part of our meticulous model construction approach. The results of this study demonstrate the revolutionary potential of machine learning in greenhouse climate management, opening the door for data-driven, intelligent agricultural systems that maximize crop yields while reducing energy consumption.