Advancing solar thermal utilization by optimization of phase change material thermal storage systems: A hybrid approach of artificial neural network (ANN)/Genetic algorithm (GA)

Solar energy is critical to the global shift towards sustainable and low-carbon energy solutions. Unlike fossil fuels, solar energy produces no emissions during operation, making it a key driver of environmentally friendly power generation. However, the intermittency of solar power remains a challenge, necessitating efficient energy storage systems to ensure a steady supply. Thermal energy storage systems utilizing phase change materials (PCMs) offer a solution by storing excess solar energy and releasing it when needed. This study focuses on enhancing the charging capacity of the PCM within a novel triplex tube heat exchanger (TTHE). The design incorporates bionic-shaped fins to compensate for thermal conductivity within the PCM. Three artificial neural network (ANN) models were employed to estimate the melting duration for the PCM to attain liquid fractions of 0.5, 0.8, and 1. The expansion angle (θ), initial length (D), and length of fin branches (Z) were systematically varied to investigate their influence on heat absorption. The ANN models exhibited high accuracy, with R² values of 0.998 for the liquid fraction of 0.5, 0.998 for the liquid fraction of 0.8, and 0.996 for the liquid fraction of 1, demonstrating their precise predictive capability. The results revealed that θ significantly reduced melting time, while Z consistently shortened the duration and eventually became the dominant factor. However, an excessive increase in these geometric parameters led to a counterproductive rise in the total melting time. Through optimization using a genetic algorithm, three optimal designs (design 1 (D1), design 2 (D2), and design 3 (D3)) were proposed. Compared to the fin-less TTHE, these designs reduced the full melting time by 71.58 %, 73.41 %, and 73.54 %, respectively. Faster melting improves the system's ability to store and release thermal energy more quickly, enabling more efficient utilization of solar power within the limited period of sunlight availability (usually 4–6 h a day). These findings underscore the potential of ANN-based approaches in optimizing the efficiency of solar energy storage systems.