| Cold storage, which involves preserving perishable products at low temperatures, plays a vital role in various economic activities. With advances in modern technologies, new solutions can now be developed to optimize both product preservation and the efficiency of refrigeration facilities. This paper presents the implementation of an IoT and AI-based system to optimize the management of positive temperature refrigeration systems. The IoT component ensures real-time monitoring of critical parameters, including temperature, humidity, and electrical variables. In parallel, AI models integrate XGBoost for anomaly detection and the identification of impending failures, along with LSTM for forecasting temperature and humidity variations, enabling proactive refrigeration system management. We trained and compared two classification models, XGBoost and Random Forest, to evaluate their performance and select the most suitable one for system implementation. The results show that XGBoost achieved an accuracy of 96.8%, outperforming Random Forest at 94.5%. For predictive modeling, LSTM demonstrated strong performance, with an MSE of 1.21 and an MAE of 0.65 for temperature prediction, as well as an MSE of 8.36 and an MAE of 2.39 for humidity prediction. Experimental tests validated the effectiveness of the system and highlighted its potential benefits in optimizing the management of positive temperature refrigeration systems. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.