An Approach for Estimation of Temperature Profile Inside a Reefer Container During Transportation

Food loss is a critical concern, with substantial losses occurring before reaching the retail stage. Maintaining optimal storage conditions for perishable food products during transportation is vital. This paper introduces a cost-effective and reliable solution for mapping the temperature distribution profile inside refrigerated containers during transportation, aimed at mitigating food losses. The research employs a Grey Box approach, integrating both White Box (comprehensive system knowledge) and Black Box (data-driven) methods. Three distinct models, including neural networks, XGBoost, and random forest, are trained with data from numerical simulations of the reefer container during transportation. Results reveal that the random forest model outperforms the others, displaying minimal errors in various evaluation metrics. The discrepancies between predicted and actual values are visualized, confirming the model’s reliability. This research provides a practical solution for monitoring and controlling environmental conditions within reefer containers. The model can be generalized by implementing and validating it with temperature data from other perishable supply chains. Furthermore, the model can undergo testing and validation using continuously streamed real-time temperature data, ultimately leading to the development of a digital twin model.