Partager

Document IIF

Analyse assistée par IA des performances d’unités de conditionnement d’air mini-split : approches expérimentale et prédictive.

AI-assisted performance analysis of mini-split air conditioning units: Experimental and predictive approaches.

Auteurs : ASHOUR A. M., KAHDIM S. A., JAGHDAM I. H., HAMMOODI K. A., JAFFAR H. M., Al-GHEZI M. K. S., RAJHI W.

Type d'article : Article de la RIF

Résumé

Air-conditioning systems are crucial for modern society, supplying cooling for residential applications. Mini-split systems are widely embraced for their ease of installation, flexibility, and efficiency. Nevertheless, their operational COP is very sensitive to indoor temperature, humidity, and airflow variation. Traditional thermodynamic models are usually unable to accurately describe these nonlinear characteristics, as they often vary depending on the instantaneous operating points of this system. To address this research gap, an experimental study was conducted with a 3.5 kW mini-split unit under 16 steady-state operating conditions and a database containing 3093 data points from measurements from sensors that capture pressure, temperature, humidity, and airflow. Three machine learning models, such as XGBoost (Extreme Gradient Boosting), SVM (Support Vector Machine), and ANN (Artificial Neural Network) were used to predict COP. The best accuracy was achieved using the XGBoost model with R² = 0.992, RMSE = 0.085, and MAE = 0.051. The accuracy for the SVM model was R² =0.990, RMSE = 0.099, and MAE = 0.057, while that of the ANN model was the worst at R² = 0.976, RMSE =0.152, and MAE = 0.108. The research identifies the potential of machine learning for real-time systems’optimization and control in mini-split air-conditioning systems, increasing their efficiency for sustainability. Future research could aim at increasing the dataset to include other environmental parameters to further enhance the predictive capabilities.

Documents disponibles

Format PDF

Pages : 14 p.

Disponible

  • Prix public

    20 €

  • Prix membre*

    Gratuit

* meilleur tarif applicable selon le type d'adhésion (voir le détail des avantages des adhésions individuelles et collectives)

Détails

  • Titre original : AI-assisted performance analysis of mini-split air conditioning units: Experimental and predictive approaches.
  • Identifiant de la fiche : 30034475
  • Langues : Anglais
  • Sujet : Technologie
  • Source : International Journal of Refrigeration - Revue Internationale du Froid - vol. 181
  • Date d'édition : 01/2026
  • DOI : http://dx.doi.org/https://doi.org/10.1016/j.ijrefrig.2025.10.012

Liens


Voir d'autres articles du même numéro (33)
Voir la source

Indexation