Thermal Enhancement of PCM-Based Latent Heat Storage With Perforated Fin Networks and Coupled Effects of Geometry, Natural Convection, and PCM Melting Using MLP/GA Optimization

Abstract

The growing demand for sustainable and efficient energy systems has significantly increased the importance of latent heat storage systems (LHSSs) in renewable energy applications, particularly in solar thermal technologies. Despite the high energy storage density of PCMs, their inherently low thermal conductivity remains a major obstacle that limits charging performance and practical utilization. In the present study, a novel perforated finassisted LHSS is proposed and numerically investigated to enhance the charging performance of PCM-based storage systems. The effects of three key geometrical parameters, including perforation length (L), tube diameter (D), and storage body inclination angle (α), are systematically examined through a full factorial design approach. Furthermore, a multilayer perceptron neural network is established to estimate the system's thermal behavior, while genetic algorithm-based optimization identifies optimal configurations for short- and long-term heat absorption. The optimized short-term energy absorption design (OSTE) absorbs 7763 kJ during the shortterm charging period. In contrast, the optimized long-term energy absorption design (OLTE) absorbs 7343 kJ, indicating that OSTE achieves approximately 5.72% higher short-term energy absorption than OLTE. Compared with the core design without perforated fins, which absorbs only 3077 kJ under the same conditions, the OSTE and OLTE configurations improve short-term energy absorption by approximately 152.3% and 138.6%, respectively. Under long-term operating conditions, OLTE absorbs 10,011 kJ, while OSTE absorbs 9957 kJ, corresponding to only 0.54% higher long-term energy absorption for OLTE. In contrast, the core design absorbs only 5395 kJ under the same conditions, demonstrating that OLTE and OSTE improve long-term energy absorption by approximately 85.6% and 84.6%, respectively. The results reveal that OSTE's superiority during most stages of the charging process is substantially greater than OLTE's slight long-term advantage. The findings demonstrate the strong potential of combining perforated fin networks, machine learning prediction models, and optimization algorithms to develop high-performance LHSSs for renewable energy applications.