Base-integrated thermal storage system: Enhancing phase change material melting performance through geometric innovation and artificial neural network-genetic algorithm optimization

Abstract

Efficient and economically viable energy storage systems are crucial for enhancing the reliability of solar-based heating applications, particularly during periods of limited sunlight availability. In this study, a novel baseintegrated energy storage unit was introduced, in which heat transfer fluid tubes were embedded directly within the PCM box and supported by adjustable straight and tilted bases. This configuration enhances conductive pathways through continuous solid–solid contact among the tubes, bases, and device body, resulting in improved heat transfer performance while offering high structural flexibility and lower maintenance costs compared to conventional shell-and-tube systems. To systematically assess the influence of structural variables (the length of straight bases, the length of tilted bases, and the angle of tilted bases), a full factorial numerical analysis was carried out, generating distinct geometric designs. By the end of the 5-h charging period, the highest liquid fraction among the 27 parametric configurations was achieved by Design 5, reaching a value of 0.906, whereas the base-free design attained a liquid fraction of only 0.368. This indicates that Design 5 enhances the melting performance by approximately 146% compared to the base-free configuration. A similar trend is observed in terms of total stored energy. After 5 h, Design 5 stores 23,944 kJ of thermal energy, while the basefree system stores only 10,852 kJ. This corresponds to an improvement of approximately 120.6% in total absorbed energy for Design 5 relative to the base-free design. To generalize these findings, an artificial neural network prediction model was developed using MATLAB's feedforward network with the Levenberg–Marquardt training algorithm. The trained model was subsequently coupled with a genetic algorithm to carry out both single- and multi-objective optimization, with the goal of maximizing the PCM liquid fraction. The genetic algorithm yielded two optimal geometric configurations: Half-Melting Optimal for maximizing liquid fraction at 2.5 h and Full-Melting Optimal for maximizing liquid fraction at 2.5 h. The TOPSIS-selected design, named Balanced-Melting Optimal, falls between the values of Half-Melting Optimal and Full-Melting Optimal. Besides, three additional engineered Finned-designs were introduced to be compared with the optimal designs. After 5 h of charging, the liquid fraction reached 0.847, 0.913, and 0.881 for Half-Melting Optimal, Full-Melting Optimal, and Balanced-Melting Optimal, respectively. These values correspond to improvements of 130%, 148%, and 139%, respectively compared to the base-free design. Besides, Half-Melting Optimal, Full-Melting Optimal, and Balanced-Melting Optimal stored 23,236 kJ, 24359 kJ, and 23,810 kJ, respectively, compared to 10,852 kJ for the base-free system, yielding improvements of approximately 114%, 124%, and 120%. Among the optimized configurations, Full-Melting Optimal was proposed as the best overall optimal design, since it achieved the highest performance in terms of both liquid fraction and total stored energy. When compared directly with Design 5, the Full-Melting Optimal configuration still provided measurable improvements, with approximately 0.77% betterment in liquid fraction and about 1.73% in stored energy.