Enhancing energy efficiency in supermarkets: A data-driven approach for fault detection and diagnosis in CO2 refrigeration systems

Masoud Kishani Farahani, Mohammad Hossein Yazdi, Mohammad Talaei and Abbas Rajabi Ghahnavieh

Applied Energy, 2025, vol. 377, issue PB, No S0306261924018622

Abstract: Supermarkets are significant consumers of energy, primarily due to refrigeration systems, which also contribute to global climate change through hydrofluorocarbon refrigerants. The transition to CO2 refrigeration systems (CO2-RS) offers a low-environmental-impact alternative; however, malfunctions can undermine these benefits. To prevent CO2-RS from malfunctioning, fault detection and diagnostics (FDD) are commonly employed. This study presents an innovative approach to developing an efficient data-driven FDD model for CO2-RS, emphasizing cost-effective sensor utilization and model interpretability. This new method is essential due to the limitations of existing FDD techniques, which often lack cost-effective sensor solutions and model interpretability, thereby hindering their practical application and effectiveness in identifying and diagnosing faults in CO2-RS. The approach focuses on diagnosing common faults in CO2-RS by developing virtual sensors, employing tree-based machine learning algorithms (Random Forest, XGBoost, CatBoost, LightGBM), selecting an optimal sensor set, and using SHapley Additive exPlanations (SHAP) for interpretability. The integration of three developed virtual sensors with pre-installed physical sensors, derived from physical relationships and existing sensors, enhances access to cost-effective sensors and improves the performance of data-driven FDD models. These virtual sensors, as well as the physical sensors needed to develop them, are selected as the optimal sensor set. Additionally, the data-driven FDD model, utilizing the random forest (RF) algorithm and the optimal sensor set, is introduced as an efficient model capable of classifying faults in CO2-RS, achieving an accuracy of 99.48 %, with precision and recall of 99.57 %, and an F1-score of 99.42 %. The SHAP technique is employed to enhance model interpretability, ensuring practical deployment in supermarket settings.

Keywords: Refrigeration systems; Machine learning; Fault detection; Fault diagnosis; Classification (search for similar items in EconPapers)
Date: 2025
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S0306261924018622
Full text for ScienceDirect subscribers only

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:appene:v:377:y:2025:i:pb:s0306261924018622

Ordering information: This journal article can be ordered from
http://www.elsevier.com/wps/find/journaldescription.cws_home/405891/bibliographic
http://www.elsevier. ... 405891/bibliographic

DOI: 10.1016/j.apenergy.2024.124479

Access Statistics for this article

Applied Energy is currently edited by J. Yan

More articles in Applied Energy from Elsevier
Bibliographic data for series maintained by Catherine Liu ().