Performance improvement of a novel plate-finned latent thermal energy storage system by numerical simulation, artificial neural network, and genetic algorithm

Latent thermal energy storage based on phase change materials offers a promising solution to alleviate temporal and spatial mismatches between thermal energy supply and demand. However, conventional shell-and-tube devices suffer from inherent drawbacks such as large footprints and low energy storage densities, which limit system miniaturization and efficiency improvement. To address these challenges, this study proposes a tri-medium plate-finned latent thermal energy storage system, and develops a rapid optimization framework integrating numerical simulations with an optimization algorithm to improve its performance. Firstly, a validated numerical simulation model was employed to analyze the effects of fin thickness, fin height, and the percentage of phase change materials on power density, thereby generating a performance database. This database was then adopted to train artificial neural networks, yielding data-driven surrogate models that serve as efficient substitutes for complex and time-consuming computational fluid dynamics simulations. Subsequently, a genetic algorithm was applied to optimize key geometric parameters for maximum power density. The results show that the modified artificial neural networks surrogate model achieves high prediction accuracy, with a coefficient of determination of 0.9979 and a mean squared error of 0.295. The optimized system attains a power density of 95.62 kW m⁻³, representing a 29.71 % improvement over the baseline. In thermal release mode, it can supply up to 184 L of hot water at 50 °C for a duration of 7.68 min. Overall, this study integrates numerical simulations, artificial neural networks, and genetic algorithm to provide an effective methodology for performance prediction, enhancement, and optimization of plate-finned latent thermal energy storage systems.