Research Article
Performance Prediction of Solar-Powered Domestic Fridge in Hot-Humid Climates Using Machine Learning
Joseph Kwarteng Acheampong*
,
Theophilus Frimpong Adu,
Daniel Marfo
Issue:
Volume 11, Issue 2, December 2026
Pages:
29-40
Received:
27 May 2026
Accepted:
29 June 2026
Published:
9 September 2026
DOI:
10.11648/j.ijecec.20261102.11
Downloads:
Views:
Abstract: The intermittent nature of solar energy and dynamic environmental conditions pose significant challenges to the performance and reliability of solar-powered refrigeration systems in hot-humid climates. While traditional thermodynamic models offer limited predictive capability under real-world variability, machine learning (ML) presents a promising alternative for addressing these complexities. This study develops and evaluates ML-based models to predict the performance of a 92-liter DC solar-powered domestic refrigerator, focusing on power input (Pin), coefficient of performance (COP), and refrigerating capacity (CR). Experimental data, including solar irradiance, ambient temperature, and system operational parameters, were collected at 10-minute intervals under hot-humid conditions. Three ML algorithms: Artificial Neural Network (ANN), Random Forest Regression (RFR), and Support Vector Regression (SVR) were trained and validated using a dataset of 500 samples. Key findings reveal that ambient temperature strongly influences system performance, exhibiting a negative correlation with COP (-0.65) and CR (-0.56), while solar irradiance shows a non-linear relationship with these metrics. Among the models, SVR demonstrated exceptional accuracy for power input prediction (R2 = 0.99, RMSE = 0.41), whereas ANN achieved the highest performance for COP prediction (R2 = 0.85, RMSE = 1.45). RFR emerged as the most robust for refrigerating capacity estimation (R2 = 0.71), though all models faced challenges due to the parameter's transient dependence. The results highlight the potential of hybrid ML approaches to optimize solar refrigeration systems in climate-specific scenarios. This study contributes a scalable, data-driven framework for performance prediction, offering practical insights for system design, adaptive control strategies, and policy recommendations to enhance the adoption of solar refrigeration in off-grid and energy-insecure regions. The limitations of dataset size and weather variability are acknowledged, suggesting avenues for future research, including long-term real-world validation and ensemble modeling.
Abstract: The intermittent nature of solar energy and dynamic environmental conditions pose significant challenges to the performance and reliability of solar-powered refrigeration systems in hot-humid climates. While traditional thermodynamic models offer limited predictive capability under real-world variability, machine learning (ML) presents a promising ...
Show More
