Machine learning-based modelling of heat and mass transfer in a dual-compartment moving-bed thermochemical energy storage
Journal of Energy Storage, cilt.174, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 174
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.est.2026.123128
- Dergi Adı: Journal of Energy Storage
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC
- Anahtar Kelimeler: Extreme gradient boosting (XGBoost), Heat and mass transfer, Machine learning (ML), Moving-bed reactor, Thermochemical energy storage (TCES)
- Yıldız Teknik Üniversitesi Adresli: Evet
Özet
Thermochemical energy storage (TCES) systems are vital for utilising renewable energy, offering high energy density and long-term storage capabilities compared to latent and sensible heat storage methods. To leverage these advantages, accurate modelling of the complex, transient heat and mass transfer during charging and discharging is essential for optimising system performance and design. This study addresses a gap in the literature by focusing on a novel dual-compartment moving-bed (DCMB) thermochemical reactor design, unlike the frequently studied fixed-bed reactors, and utilises advanced machine learning for performance prediction. Two distinct Extreme Gradient Boosting models were developed to predict system performance using a pumice-calcium chloride (PM-CaCl2) composite thermochemical material. The models were trained using a comprehensive experimental dataset comprising 1991 data points collected from tests performed under varying inlet conditions and flow rates. Model 1 predicted the heat output and mass transfer during the exothermic discharge phase, while Model 2 predicted the necessary heat input and mass transfer during the endothermic charge phase. The models exhibited outstanding predictive accuracy, validating their ability to generalise the complex dynamics. High coefficients of determination were obtained for all four output parameters, with values exceeding 0.980. Furthermore, the error magnitude was minimal, with the Mean Squared Error for the heat variables remaining below 1.8 × 10−3. These results validate the utility of the Extreme Gradient Boosting framework as a highly reliable and precise tool for characterising the transient charge and discharge behaviour of this innovative DCMB-TCES system, confirming its capability to handle complex coupled physical phenomena based solely on easily measurable inlet operating parameters.