| Yazarlar (4) |
Emrehan Gürsoy
Kardemir Karabük Iron Steel Industry Trade & Co. Inc., Turkey |
Doç. Dr. Mehmet GÜRDAL
Kastamonu Üniversitesi, Türkiye |
Dr. Öğr. Üyesi Muhammed TAN
Kastamonu Üniversitesi, Türkiye |
|
Engin Gedik
Karabük Üniversitesi, Turkey |
| Özet |
| This study investigates the effects of different metal foam (MF) configurations on the performance of latent heat thermal energy storage (LHTES) systems. Two distinct MF materials, aluminum (Al) and copper (Cu) were analyzed with varying porosity (ε=0.90–0.95) and pore density (ω=10–20 PPI) values to optimize the melting and energy storage characteristics of phase change materials (PCMs). A validated computational fluid dynamics (CFD) model using the enthalpy-porosity method (EPM) was implemented and a deep learning approach integrating long short-term memory (LSTM) networks and self-attention mechanisms was employed to optimize key parameters. The findings indicate that placing Cu MF near the heat source reduced the melting time by 28.2%, increased the total stored energy by 3.39%, and a solid phase state of 17% is observed in the Al MF case. Due to Cu's high thermal conductivity (401 W.m−1.K−1), overall heat transfer was 60% more efficient than with Al MF (202.4 W.m−1.K−1). Incorporating Cu MF with a porosity of ε=0.90 in the region adjacent to the heated wall resulted in a 51.7 % reduction in melting time compared to a porosity of ε=0.95. It was observed that the increase in porosity prolonged the melting time in LHTES, and the fastest melting occurred in RC22. The melting times of RC24, RC30, and RC32 were 10.3%, 21.5%, and 27.2% longer than RC22, respectively. The stored energy in the S2 regions varied according to the porosity status in S1, and the highest energy change occurred in RC24 with 5837.7 kJ.m−1. In contrast, variations in pore density (ω=10–20 PPI) had a negligible impact on system efficiency, with a difference of less than 1%. Based on the numerical results, the influence of the examined variables on the melting and energy storage characteristics of the LHTES system follows the order: porosity>material order>pore density. Additionally, a custom deep learning model composed mainly of LSTM and attention layers was developed, incorporating an Optuna-optimized hyperparameter tuning process over 9 hyperparameters with an objective considering both the validation loss and the generalization gap. The best hyperparameters were determined through the 100-trial Optuna search, ensuring optimal generalization and reducing overfitting. The model demonstrated high accuracy in predicting PCM temperature and enthalpy variations of the unseen test set, with coefficient of determination (R2) values exceeding 0.997, and mean absolute percentage errors (MAPE) lower than 3.0%. In conclusion, optimized MF configurations enhance melting uniformity, reduce thermal resistance, and improve stored energy amount, while the integration of advanced machine learning (ML) techniques provides an accurate and efficient predictive tool for LHTES performance in industrial and residential applications. |
| Anahtar Kelimeler |
| Enthalpy | Latent heat thermal energy storage | Machine learning | Metal foam | Phase change materials | Python |
| Makale Türü | Özgün Makale |
| Makale Alt Türü | SSCI, AHCI, SCI, SCI-Exp dergilerinde yayımlanan tam makale |
| Dergi Adı | Journal of Energy Storage |
| Dergi ISSN | 2352-152X Wos Dergi Scopus Dergi |
| Dergi Grubu | Q1 |
| Makale Dili | İngilizce |
| Basım Tarihi | 09-2025 |
| Cilt No | 130 |
| Sayı | 1 |
| Doi Numarası | 10.1016/j.est.2025.117361 |
| Atıf Sayıları |
